Nature CellBiology

Editorial officEs

london: [email protected] Macmillan Building, 4 Crinan Street, London N1 9XW Telephone: +44 207 843 4924; Fax: +44 207 843 4794 editor: Bernd Pulverersenior editors: Alison Schuldt & Sowmya Swaminathanassociate editors: Nathalie Le Bot, Silvia Grisendi & Christina Karlsson Rosenthalproduction editor: Karl Smart art editor: Denis Malletcopy editor: Tamra Pooruneditorial assistant: Alice Fuller ManagEMEnt officEs

nPg london: [email protected] The Macmillan Building, 4 Crinan Street, London N1 9XW Telephone: +44 207 833 4000; Fax: +44 207 843 4596/7 managing director: Steven Inchcoombe, publishing director: Alison Mitchellassociate directors: Jenny Henderson, Tony Rudlandeditor-in-chief, nature publications: Philip Campbelleditorial production director: James McQuatmanaging production editor: Donald McDonaldproduction director: Yvonne Strong production controller: Kelly Hopkinssenior marketing manager: Tim Redding

nPg nEw York: [email protected] 75 Varick St F1 9, New York, NY 10013-1917 Telephone: +212 726 9200; Fax: +212 696 9006executive editor: Linda Millerchief technology officer: Howard Ratner

nPg tokYo: [email protected] Japan K.K., Chiyoda Building 2–37 Ichigayatamachi, Shinjuku-Ku, Tokyo, 162-0843, Japan Telephone: +81 3 3267 8751; Fax: +81 3 3267 8746 asia-pacific publisher: Antoine E. Bocquet, manager: Koichi Nakamura asia-pacific sales director: Kate Yoneyamaasia-pacific sales: Ken Mikamimarketing/production manager: Takeshi Murakami

disPlaY advErtising

[email protected] (Us/canada) [email protected] (Europe) [email protected] (asia)

global head of display advertising: Andrew Douglas, Tel: 44 207 843 4960, Fax: 44 207 843 4996asia-pacific sales manager: Ken Mikami, Tel: 81 3 3267 8765, Fax: 81 3 3267 8746

disPlaY accoUnt ManagErs:western regional sales manager, west coast south/west coast canada:George Lui, Tel: (415) 781 3804, Fax: (415) 781 3805west coast north: Bruce Shaver, Tel: (415) 781 6422, Fax: (415) 781 3805new england/east coast canada:Sheila Reardon, Tel: (617) 399 4098, Fax: (617) 426 3717new york, mid-atlantic, southeast:Jim Breault, Tel: (212) 726 9334, Fax: (212) 696 9481midwest: Mike Rossi, Tel: (212) 726 9255, Fax: (212) 696 9481germany/switzerland/austria:Sabine Hugi-Fürst, Tel: 41 52761 3386, Fax: 41 52761 3419united kingdom/ireland/france/belgium/eastern europe:Jeremy Betts, Tel: 44 207 843 4968, Fax: 44 207 843 4749scandinavia/denmark/the netherlands/italy/spain/portugal/israel/iceland:Graham Combe, Tel: 44 207 843 4914, Fax: 44 207 843 4749greater china/singapore: Gloria To, Tel: (852) 2811 7191, Fax: (852) 2811 0743

world widE wEb

www.nature.com/naturecellbiologyweb production manager: Alex Thurrellweb production editor: Deborah Anthonyweb services: Anthony Barrera

natUrEjobs

Please refer to panel at the start of the naturejobs section at the back of the issue.

© 2009 Macmillan Publishers Limited. All rights reserved.

VOLUME 11 NUMBER 4 APRIL 2009

Nature Cell Biology® (ISSN 1465-7392) is published monthly by Nature Publishing Group (Porters South, 4 Crinan Street, London N1 9XW, UK). Editorial Office: Porters South, 4 Crinan Street, London N1 9XW, UK. Telephone: +44 (0)20 7843 4924. Fax: +44 (0)20 7843 4794. Email: [email protected]. North American Advertising: Nature Cell Biology, 75 Varick Street F19, New York NY 10013-1917, US. Telephone: +1(212) 726-9200. Fax: +1(212) 696-9006. European Advertising: Nature Cell Biology, Porters South, Crinan Street, London N1 9XW, UK. Telephone: +44 (0)20 7833 4000. Fax: +44 (0)20 7843 4596. New subscriptions/renewals/changes of address/back issues and all other customer service questions should be addressed to - North America: Nature Cell Biology, Subscription Dept, P.O. Box 5055, Brentwood, TN 37024-5055, USA. Outside North America: Subscriptions Department, Brunel Road, Basingstoke, Hants. RG21 6XS, UK. Annual subscription rates: Americas: US$225 (personal), US$3,060 (institutional); Europe: €287 (personal), €2,430 (institutional); UK: £185 (personal), £1,570 (institutional); Japan: ¥40,000 (personal), ¥345,000 (institutional). Back issues: US/Canada. $45 (Canada add 7% GST); Rest of World: surface US$43, air mail US$45. Application for periodical postage rate submitted and paid by New York, NY 10010 and additional mailing offices. Reprints: Nature Cell Biology Reprints Department, Porters South, Crinan Street, London N1 9XW, UK. Subscription information is available at the Nature Cell Biology homepage at http://www.nature.com/naturecellbiology POSTMASTER: Send address changes to Nature Cell Biology Subscriptions Department, Brunel Road, Basingstoke, Hants. RG21 6XS, UK or Nature Cell Biology Subscriptions Department PO Box 5054, Brentwood, TN 37024-5054, USA.

EDITORIAL

363 UK research funding

363 Turning points

TURNING POINTS

364 Coming in from the cold

Gottfried Schatz

REVIEW

365 Spindle orientation during asymmetric cell division

Karsten H. Siller and Chris Q. Doe

NEWS AND VIEWS

375 Double JMY: making actin fast

David W. Roadcap and James E. Bear

377 Skp2: caught in the Akt

Karin Ecker and Ludger Hengst

379 Targeting protein ubiquitylation: DDB1 takes its RING off

Sarah Jackson and Yue Xiong

381 SOC: now also store-operated cyclase

James W. Putney Jr

383 RESEARCH HIGHLIGHTS

p53-cofactor JMY nucleates actin filaments in cells. Expression of the actin nucleation region of JMY in U2OS cells induces the formation of actin filaments (phalloidin, red).

[letter p451]

nature cell biology i



ARTICLES

385 Two Beclin 1-binding proteins, Atg14L and Rubicon, reciprocally regulate autophagy at different stages

Kohichi Matsunaga, Tatsuya Saitoh, Keisuke Tabata, Hiroko Omori, Takashi Satoh, Naoki Kurotori, Ikuko Maejima, Kanae Shirahama-Noda, Tohru Ichimura, Toshiaki Isobe, Shizuo Akira, Takeshi Noda and Tamotsu Yoshimori

397 Phosphorylation by Akt1 promotes cytoplasmic localization of Skp2 and impairs APC–Cdh1-mediated Skp2 destruction

Daming Gao, Hiroyuki Inuzuka, Alan Tseng, Rebecca Y. Chin, Alex Toker and Wenyi Wei

409 Protein kinase DYRK2 is a scaffold that facilitates assembly of an E3 ligase

Subbareddy Maddika and Junjie Chen

N&V p379

420 Phosphorylation-dependent regulation of cytosolic localization and oncogenic function of Skp2 by Akt/PKB

Hui-Kuan Lin, Guocan Wang, Zhenbang Chen, Julie Teruya-Feldstein, Yan Liu, Chia-Hsin Chan, Wei-Lei Yang, Hediye Erdjument-Bromage, Keiichi I. Nakayama, Stephen Nimer, Paul Tempst and Pier Paolo Pandolfi

N&V p377

433 Store-operated cyclic AMP signalling mediated by STIM1

Konstantinos Lefkimmiatis, Meera Srikanthan, Isabella Maiellaro, Mary Pat Moyer, Silvana Curci and Aldebaran M. Hofer

N&V p381

LETTERS

443 Myosin IIIa boosts elongation of stereocilia by transporting espin 1 to the plus ends of actin filaments

Felipe T. Salles, Raymond C. Merritt, Jr, Uri Manor, Gerard W. Dougherty, Aurea D. Sousa, Judy E. Moore, Christopher,M.Yengo, Andréa C. Dosé and Bechara Kachar

451 p53-cofactor JMY is a multifunctional actin nucleation factor

J. Bradley Zuchero, Amanda S. Coutts, Margot E. Quinlan, Nicholas B. La Thangue and R. Dyche Mullins

N&V p375

nature cell biology ii

Apical extrusion of RasV12-transformed cells from a monolayer of normal epithelial cells occurs in a manner regulated by ROCK, myosin II and Cdc42.

[letter p460]

Interaction and cooperative function of myosin IIIa and espin 1 in promoting the formation of long actin protrusions. Myosin IIIa transports espin 1 to the plus ends of actin filaments, where the WH2 activity of espin 1 promotes the formation of very long filopodia.

[letter p443]



460 Characterization of the interface between normal and transformed epithelial cells

Catherine Hogan, Sophie Dupré-Crochet, Mark Norman, Mihoko Kajita, Carola Zimmermann, Andrew E. Pelling, Eugenia Piddini, Luis Alberto Baena-López, Jean-Paul Vincent, Yoshifumi Itoh, Hiroshi Hosoya, Franck Pichaud and Yasuyuki Fujita

468 Distinct regulation of autophagic activity by Atg14L and Rubicon associated with Beclin 1–phosphatidylinositol-3-kinase complex

Yun Zhong, Qing Jun Wang, Xianting Li, Ying Yan, Jonathan M. Backer, Brian T. Chait, Nathaniel Heintz and Zhenyu Yue

477 A mechanism for chromosome segregation sensing by the NoCut checkpoint

Manuel Mendoza, Caren Norden, Kathrin Durrer, Harald Rauter, Frank Uhlmann and Yves Barral

484 Modularity of MAP kinases allows deformation of their signalling pathways

Areez Mody, Joan Weiner and Sharad Ramanathan

492 STAT3 inhibition of gluconeogenesis is downregulated by SirT1

Yongzhan Nie, Derek M. Erion, Zhenglong Yuan, Marcelo Dietrich, Gerald I. Shulman, Tamas L. Horvath and Qian Gao

501 Absence of nucleolar disruption after impairment of 40S ribosome biogenesis reveals an rpL11-translation-dependent mechanism of p53 induction

Stefano Fumagalli, Alessandro Di Cara, Arti Neb-Gulati, Francois Natt, Sandy Schwemberger, Jonathan Hall, George F. Babcock, Rosa Bernardi, Pier Paolo Pandolfi and George Thomas

508 Erratum

NATURE CELL BIOLOGY CLASSIFIED

See back pages

nature cell biology iii

Acute loss of ribosomal protein S6 does not impair nucleolar integrity. Co-immunofluoresence of rpS6 (green) and rpL7a (red) in A549 cells transiently depleted of rpS6 by RNAi. Nascent rpS6 is absent from the intact nucleolus (marked by rpL7a).

[letter p501]


nature cell biology volume 11 | number 4 | APrIl 2009 363

E D I T O R I A L

UK research fundingIs the UK still committed to basic biology research?

In early March the Higher Education Funding Council for England (HEFCE), the agency responsible for the distribution of almost £8 billion (US$11.3 billion) of government funds for the academic year 2009–10, unveiled provisional funding allocations for research (£1.57 billion) and teaching (£4.78 billion). The budget, which accounts for roughly half of the funding at many UK universities, benefited from a 5.6% real-term increase. Awards for research are largely allocated according to the results of an evaluation of British universities carried out by the HEFCE every few years. The ‘Research Assessment Exercise’ (RAE) represents a thorough peer review evaluation of 52,400 researchers from 159 institutions. Although benchmarking against international research is not formally pursued, the international nature of the review process guarantees impartiality, and the RAE is well respected in the UK and abroad. Indeed, Enric Banda, President of Euroscience, suggests that central and southern Europe would benefit from adoption of similar achievement-oriented schemes. The latest RAE results, released last December, showed an increase in top-rated researchers in universities other than the two dozen that usually attract 80% funding (17% of researchers achieved the highest rating, and 37% the second highest; at least half the researchers from 118 universities fell into the top two categories). As a result, some worried that funding would be spread too thin to sustain world-class research, or that high priority subjects or application-oriented research would be favoured, while others worried that lower-ranked, but nevertheless valuable, research would fail to get support altogether. Universities are still evaluating the funding allocations, but it appears that while traditional top performers such as Cambridge and Oxford retained the bulk of research funding, several received real-term cuts, while the 24 ‘new universities’ created in the 1990s increased their funding share from 0.9 to 3.2%. At first glance, HEFCE seems to have remained true to its goal of “supporting and rewarding excellence in research of all kinds, in all subjects, wherever it may be found. This includes research that bridges traditional discipline boundaries, and applied and practice-based work, as well as purely curiosity-driven enquiry”.

Meanwhile, the UK government has decided to abandon the RAE scheme (apparently to save on the £12 million costs) in favour of a ‘lighter touch’ programme, the Research Excellence Framework, which will rely much more on bibliometrics, rather than peer review. We have discussed the inherent limitations of impact factors previously, in particular when comparing fields of dramatically divergent sizes and research activities (neither a good measure of the quality of an individual research programme). Comparisons against international benchmarks would help, but impact-factor related assessment already informs funding in many countries (the US may be an exception). Importantly, this move may encourage research in fashionable fields and application-oriented research over that in ‘blue skies’ and niche areas. While this move could save a few million, it may result in a less informed distribution of billions. It remains to be seen how far HEFCE will go in retaining independent expert reviews in the finalized assessment strategy due this autumn.

Much of the remaining research funding is awarded by seven UK Research Councils. Like HEFCE, these are independent of direct political

control. The Biotechnology and Biological Sciences Research Council (BBSRC) with an annual budget of £420 million, is the council responsible for most of the funding for the basic biological sciences and many cell biologists are among the 1,600 senior researchers it supports.

Last autumn, the BBSRC announced significant changes, which will come into force with the grant round now under evaluation: the number of research committees have been reduced to four with a wider remit and a more flexible set of experts drafted ad hoc to reflect the applications received. At the same time, ten new research and policy priorities came into effect. BBSRC stated that this list will “overarch all its activities”, and notably many of the topics have societal and economic relevance: ageing, bioenergy, environmental change, crop science and global security. The more basic topics are rather focused: technology development, bionanotechnology, systems biology and synthetic biology. These topics are juxtaposed with policy keywords including economic and social impact, reduction in animal research and international collaboration. The council will complement these by issuing occasional ‘highlight notices’. Some are concerned that the council that has traditionally supported the basic biosciences is en route to a more applied and narrow remit, and foresee starving for areas not readily supported by the medical charities. These concerns may yet be proven moot: BBSRC Chief Executive Douglas Kell assured that “the four new committees will take in the whole BBSRC remit, which is not changing, but will be bigger and more flexible. These changes are not about forcing researchers to work in industry. We want to encourage researchers to think about the strategic focus of their applications”. A more streamlined grant evaluation system is to be welcomed. However, basic research, including the blue sky variety without immediate applications, has served the UK well in maintaining its status as a global research leader and it is to be hoped that research priorities will also be interpreted with considerable flexibility by the BBSRC.

Further reading: Connotea.org/user/ncb/tag/ukresearchfunding

Turning pointsA series of essays describing pivotal events in the careers of cell biologists.

This month’s issue of Nature Cell Biology presents the first in a new series of short autobiographical essays by leading scientists entitled “Turning Points”. The articles offer a historical perspective of the career of the author and feature a first-hand recounting of a pivotal event that shaped their scientific future. Events may be as diverse as the unexpected generosity of a colleague, a move to a new destination or even arguments with peers that triggered a shift in research direction or led to the development of a new concept. We hope that the series will highlight some of the stories that are part of the folklore of cell biology — tales often recounted at the bar or beach during conferences, but which seldom find an audience in a more formal context. As such, we hope they will prove inspirational to scientists early in their career. The series launches on p364 with an account from Gottfried Schatz on how he was inspired to embark on a career devoted to studying mitochondria. The authors will be drawn from fields that are represented within the journal. If there is a particular cell, molecular or developmental biologist whom you would like to see featured in this series, please send your suggestions to [email protected]


Connotea.org/user/ncb/tag/ukresearchfunding

mailto:[email protected]

364 nature cell biology volume 11 | number 4 | APrIl 2009

TURNING PO INTS

Coming in from the cold: how answering a postcard can launch a scientific careerGottfried Schatz

“Here is that bio-something you wanted, Herr Doktor” mumbled the university librar-ian, pushing a pile of battered Physiologische Chemie textbooks in my general direction. His sarcasm was not lost on me, because he reserved ‘Herr Doktor’ for us students when we pestered him with extracurricular —hence frivolous — requests. Worse still, the bio-chemistry textbooks were pre-Second World War vintage. As we were then writing the year 1958, they were useless clunkers. After gradu-ating from high school in the Austrian city of Graz, I had wanted to become a biochem-ist, but in those days the University of Graz had no biochemists on its faculty and offered no biochemistry courses of any kind. As my parents could not afford to send me abroad and as international fellowships were virtu-ally nonexistent, I had decided to enrol as a chemistry student and to master biochemistry on my own. Thanks to my friendly librarian, I had just learned that textbooks from our uni-versity library were not an option. The book-stores in Graz carried only a single modern, but also prohibitively expensive, textbook by the Swiss biochemist Franz Leuthardt and were unwilling to find out what British or American publishers might have to offer. Such was the intellectual splendour of postwar Graz, which had once been home to such sci-entific giants as Otto Löwi, Karl Boltzmann, Ernst Mach and Erwin Schrödinger. After many false starts, I finally concocted the fol-lowing six-step biochemistry course: first, I worked my way through the Biochemistry Section of Chemical Abstracts, a now defunct

periodical that our library held. Second, I jotted down the names and addresses of the authors whose articles interested me. Third, I bought several dozen picture postcards of Graz and sent them to these authors with the lapidary handwritten request: “Please send me all your reprints”. Fourth, fifth and sixth, I waited, waited and waited, because I could not afford the luxury of airmail and had sent all postcards by land mail. Looking back, I am amazed that anybody answered them at all. Yet quite a few did, sending me one or two of their most recent reprints. Not so David E. Green, a leading researcher on the biochemistry of mitochondria, who ran a huge and highly successful laboratory with several dozen collaborators at the Enzyme Institute of the University of Wisconsin at Madison. Green liked to do things the big way and sent me a massive package with more than 200 reprints on the function and structure of mammalian mitochondria. Some of these papers are now classics, and all of them bore the mark of Green’s polished scientific prose. I devoured these articles, mostly on benches in our local park, and soon lost myself in an enchanted world of electron-conducting membranes and colourful cytochromes. What could be more exciting and important than the pathway that gave energy to life? My private biochemistry course had swung into high gear. Its balance of subjects may have been open to serious criti-cism, but it kindled my life-long fascination with cellular respiration and mitochondrial biogenesis.

I have never forgotten how anxiously I waited for replies to my postcards and how crucial David Green’s generous response was for my scientific career. To this day I promptly answer every letter or e-mail I receive, par-ticularly if it is from a young scientist whose

name is unfamiliar to me. Scientists of today are so overburdened with paperwork and mesmerized by competition that they often neglect this simple courtesy. A thoughtful letter to a young scientist in a remote corner of the globe may do more for science than a plenary talk at an international meeting. Many of these meetings have degenerated into a mélange of commercial trade show and sci-entific media event where everybody flocks around the scientific stars. Many of these stars are so eager to rub shoulders with their illustrious peers that they have little inclina-tion to waste their precious time talking to unknown young scientists. I have given my fair share of plenary talks and have usually enjoyed the limelight. But as I grew older, I became more conscious of the barrier between a plenary speaker and the younger congress participants. I often asked my hosts to excuse me from the after-speech presidential dinner table so that I could share my meal with groups of students I had never met before. Most of my hosts graciously accepted my excuse. The students usually rewarded me with refreshing insights and irreverent remarks about some of my colleagues, whereas I tried to reciprocate with advice on graduate schools, postdoctoral possibilities and faculty positions. And some-times our talks ventured into deeper waters, such as the meaning of science and its role in the modern world.

Scientific discoveries are usually children of solitude, yet are rarely born in isolation. Science is a supremely communal effort, which demands that scientists share their discoveries and help one another. Science is a covenant between generations. By hon-ouring this covenant so generously, David Green helped me get started in science and profoundly shaped my life.

Gottfried Schatz is Professor Emeritus at the Biozentrum of the University of Basel, Klingelbergstrasse 70, CH-5056, Switzerland and former president of the Swiss Science and Technology Council. e-mail: [email protected]



REV IEW

Spindle orientation during asymmetric cell divisionKarsten H. Siller1,2 and Chris Q. Doe1,3

Development of a multicellular organism from a fertilized egg depends on a precise balance between symmetric cell divisions to expand the pool of similar cells, and asymmetric cell divisions to create cell-type diversity. Spindle orientation can influence the generation of symmetric or asymmetric cell fates depending on how it is coupled to cell-intrinsic polarity cues, or how it is positioned relative to cell-extrinsic cues such as niche-derived signals. In this review, we describe the mechanism of spindle orientation in budding yeast, Drosophila melanogaster, Caenorhabditis elegans and mammalian neural progenitors, with the goal of highlighting conserved mechanisms and indicating open questions for the future.

The mitotic spindle consists of two spindle poles that nucleate microtu-bules from their minus-ends, and three classes of microtubules: kineto-chore microtubules that attach to chromosomes, interpolar microtubules that form an antiparallel array between the spindle poles and astral micro-tubules that radiate out from the spindle poles and probe the cytoplasm and cell cortex with their plus-ends. Interactions of astral microtubules with the cell cortex and cytoplasmic anchor sites are thought to be the main source of information for spindle alignment1, although cell shape can constrain the orientation of the linear mitotic spindle2. Spindle micro-tubules dynamically grow and shrink through the addition and removal of tubulin dimers, respectively, a property referred to as ‘dynamic insta-bility’. Microtubule dynamic instability allows probing for microtubule anchor sites, and can be coupled to spindle positioning force generation. Spindle positioning typically involves pulling-forces exerted on astral microtubules, which can be generated by; (1) plus-end depolymeriza-tion of astral microtubules that remain attached to the cell cortex, (2) cortically-attached microtubule minus-end directed motor activity or (3) translocation of microtubule plus-ends by attachment to actin-based motors. In these cases, the precise regulation of microtubule length is essential for productive spindle orientation, thus regulation of microtu-bule dynamic instability is critical for correct spindle positioning.

Budding yeastSpindle orientation is best understood in the budding yeast Saccharomyces cerevisiae, and many of the relevant yeast proteins are evolutionarily con-served. Thus, investigators curious about spindle orientation mechanisms in other cell types would be wise to pay careful attention to yeast.

The cell polarity axis of budding yeast is used to direct polarized growth of the daughter cell (bud), as well as to align the mitotic spindle along this axis to ensure proper DNA segregation to both mother and daughter cells. Establishment of cell polarity requires the localized cor-tical activation of Cdc42, a Rho GTPase family member, which marks the incipient bud site. Activated Cdc42 organizes polarized actin cables

extending from the bud site into the mother cell: subsequently a col-lar of septin proteins accumulates around the bud site and marks the bud neck3. Genetic analyses show that spindle positioning is controlled by two partially redundant pathways: an ‘early’ pathway that aligns the mitotic spindle along the bud axis of the mother cell before anaphase, and a ‘late’ pathway that translocates the aligned spindle through the bud neck during anaphase (Fig. 1).

The early pathway for spindle orientation. The first step in spindle orientation is the polarized transport of astral microtubule plus-ends along the actin cables into the bud, thereby positioning one spindle pole body (SPB) — the fungal centrosome — at the bud neck and leaving the other SPB at the base of the mother cell. Elegant genetic, biochemical and imaging studies have led to the following model for early spindle orientation. Adenomatous polyposis coli (APC)-related Kar9 is recruited to the daughter SPB by EB1-related Bim1. Kar9–Bim1 translocate to the microtubule plus-ends where Kar9 binds Myo2, a class V myosin, result-ing in polarized transport of the SPB and its microtubules along actin cables to the bud neck (Fig. 1a). In support of this model, mutants lack-ing Kar9, Bim1 or Myo2 have spindle alignment defects, as also caused by actin cable disruption4. Expression of a chimaeric Myo2–Bim1 fusion protein suppresses spindle alignment defects in kar9 mutants, indicat-ing that Kar9 functions as a linker between the microtubule-associated Bim1p and the actin-associated Myo2 (ref. 5). Furthermore, live imaging revealed that microtubule plus-ends move in sweeping motions along actin cables towards the bud neck and into the bud without microtu-bule shortening, suggesting that Myo2 provides the force for spindle orientation at this stage, rather than cortical microtubule capture and depolymerization5,6.

After actin-dependent spindle alignment, astral microtubules do shorten while attached to the cortex at the bud tip and bud neck, resulting in the final, precise positioning of the pre-anaphase spin-dle6–8. Cortical microtubule attachment requires Bud6, an actin- and

1Institute of Molecular Biology, Institute of Neuroscience, Howard Hughes Medical Institute 1254, University of Oregon, Eugene, OR 97403,USA. 2Current address: Department of Molecular and Cell Biology, University of California, Berkeley, CA 94720. USA.3Correspondence should be addressed to C. Q. D. (e-mail: [email protected])

nature cell biology VOLUME 11 | NUMBER 4 | APRIL 2009 365


[email protected]

R E V I E W

formin-binding protein localized to both the bud tip and neck9. How Bud6 captures microtubules is unknown, but premature Bud6 locali-zation to the bud neck increases microtubule capturing events at this site, indicating an instructive role7,9. Microtubule depolymerization at the bud neck is promoted by the septin-associated kinases Hsl1 and Gin4 (refs 7, 8), but it is unknown whether these proteins directly affect tubulin or exert their function on microtubule-associated proteins. Interestingly, Hsl1 and Gin4 are related to the MARK/Par-1 kinase, which can phosphorylate and inactivate the microtubule-stabilizing Tau protein10, raising the possibility that Hsl1 and Gin4 may desta-bilize microtubules by a similar mechanism. Despite these advances, several important questions remain. How are microtubules anchored to Bud6? How is astral microtubule-shortening regulated? Nevertheless, the existing data on the early pathway highlight the importance of actin-and microtubule -based molecular motors, microtubule–cortex interactions and regulated microtubule dynamics — mechanisms that are all used in higher eukaryotes as well.

The late pathway for spindle orientation. During anaphase, the SPB clos-est to the bud is translocated through the bud neck into the future daughter cell to establish its final position along the cell polarity axis. This process requires the microtubule minus-end directed dynein–dynactin motor

complex (Box 1, refs 11–13) and involves microtubule plus-end-directed transport of an inactive dynein–dynactin complex to the cortex, fol-lowed by activation of the cortically anchored complex (Fig. 1b). Proteins required to get the dynein–dynactin complex to the cortex include the associated proteins Bik1, Pac1 and Ndl1 (related to mammalian CLIP-170, Lis1 and Ndl, respectively). Bik1 recruits dynein–dynactin to the SPB and transports the complex to microtubule plus-ends through binding to the kinesin Kip2 (ref. 4). In the absence of Pac1, dynein fails to accumulate at microtubule plus-ends14,15, thus Pac1 may facilitate dynein–Kip2 binding, or may inhibit the minus-end directed motor activity of dynein.

Once dynein-loaded astral microtubules reach the cortex, dynein is activated by the membrane-bound pleckstrin-homology domain protein Num1 (refs 14–17). Activated dynein then pulls the SPB to the cor-tex using its microtubule minus-end directed motor activity6,17. Loss of dynein, dynactin or Num1 results in failure of the spindle to enter the neck, and the generation of a binucleate mother cell and an anucleate daughter cell6,11,12,18. In addition, Num1 mutants lack microtubule-sliding along the bud cortex and show increased dynein at microtubule plus-ends, consistent with a role for Num1 in cortical dynein activation14–17. Dynein may also be activated by the bud-cortex-localized Bud14 and associated Glc7 phosphatase: both Bud14 and Glc7 mutants show defects in spindle positioning, and Bud14 overexpression results in excessive

a

Mother cellDaughter

cell

Hsl1Gin4

Bud6

Pac1dyneindynactin

Myo2

Actincable

?

b

Num1

Cdc42complex

Kip2

Bim1

Kar9

Pre-anaphase

Anaphase/cytokinesis

Unknownlinker

Figure 1 Spindle orientation and positioning in budding yeast. (a) The early pathway (pre-anaphase). Bim1–Kar9 are recruited to the spindle pole body (SPB) translocate to microtubule plus-ends (pink arrowheads) and associate with the myosin motor Myo2. Myo2 motor activity (grey arrowheads) pulls attached microtubules into the bud resulting in positioning of the SPB at the bud neck. Bim1–Kar9 movement to microtubule plus-ends may require the Kip2 kinesin motor. The bud neck kinases Hsl1and Gin4 promote

microtubule shortening (blue arrow), which facilitates spindle alignment. (b) The late pathway (anaphase). The Kip2 kinesin transports the presumably inactive Pac1–dynein–dynactin complex from the SPB to microtubule plus-ends, and then the dynein complex is ‘off-loaded’ to the cortex, where it is activated by Num1. Cortical Num1–dynein–dynactin pulls the daughter centrosome to the centre of the bud cortex. Bold black arrows indicate the direction of the net spindle positioning force.

366 nature cell biology VOLUME 11 | NUMBER 4 | APRIL 2009


R E V I E W

microtubule-sliding and aberrant translocation of the entire spindle into the bud19. This phenotype is suppressed in dynein mutants, suggest-ing that the primary cause for aberrant spindle translocation is dynein hyperactivity19.

Spindle pole asymmetry. Proper spindle positioning requires SPB asymmetry, which ensures that only one SPB is pulled towards the bud1. This is accomplished by asymmetric localization of Kar9 and dynein specifically to the daughter SPB21–23. Kar9 localization to the daughter SPB requires the SPB-associated proteins Bim1, Bik1 and Stu2 (related to the microtubule-binding protein XMAP215), although none of these proteins are themselves restricted to the daughter SPB4,24,25. Kar9 localiza-tion to the daughter SPB also requires the early cyclins Clb4 and Clb5 and the associated Cdc28 kinase activity to prevent it from binding to the mother SPB23,26; this may be due to a daughter SPB-specific protein that binds phosphorylated Kar924,25.

In contrast, dynein localization to the daughter SPB occurs after asymmetric localization of Kar9, is Kar9-independent and requires the late cyclins Clb1/Clb2 and Cdc28 (ref. 21). Interestingly, the bud-neck-localized kinases Hsl1 and Gin4 are also required, suggesting that a signal is conveyed from astral microtubules contacting the bud neck back to the SPB to promote dynein asymmetry21. The identity of this putative signal and the substrates of the Cdc28, Hsl1 and Gin4 kinases are currently unknown.

C. elegans embryonic blastomeresThe C. elegans early embryo is unique in permitting powerful genetic analysis, as well as having a large cell size that facilitates mechanical experiments such as spindle severing and offers superb optical prop-erties for time-lapse imaging. The genetic attributes have led to the discovery of evolutionarily-conserved cell polarity proteins (Box 2);

whereas the large cell size and experimental accessibility have made this the premier system for understanding spindle force-generating mechanisms27,28.

The fertilized C. elegans zygote is elongated along the anterior-poste-rior axis, with the two juxtaposed pronuclei and associated centrosomes (nucleus centrosome complex, NCC) positioned in the posterior half of the zygote, the centrosomes aligned perpendicular to the anterior-posterior axis. Before the first mitosis, the NCC moves to the cell centre in a posterior to anterior direction (centration; Fig. 2a) and rotates 90˚ to align the centrosome pair along the anterior-posterior axis (spindle orientation; Fig. 2b). During early anaphase the mitotic spindle moves towards the posterior pole (spindle positioning; Fig. 2c), resulting in asymmetric cell division27.

Cortical polarity determines spindle orientation and position. The evolutionarily conserved Par complex (Par-3–Par-6–aPKC; Box 2) is localized to the anterior cortex of the zygote; whereas the posterior cortex is occupied by the PAR-1 kinase and the PAR-2 RING-finger protein27 (Fig. 2a). In par‑2 mutants, spindle orientation (NCC rota-tion) is absent leading to the assembly of a transversely oriented spindle, which ‘passively’ aligns along the anterior-posterior axis when it elongates during anaphase, owing to the elliptical shape of the zygote29,30. In contrast, par‑3 mutants show normal NCC rota-tion in elliptically shaped zygotes, but lack NCC rotation in spheri-cal blastomeres induced by egg-shell removal29,30. In addition, loss of cortical Par polarity disrupts anaphase spindle positioning, leading to the formation of equally sized anterior and posterior daughter cells27. Thus, both Par cortical polarity and cell shape regulate spindle ori-entation and position.

Dynein is a motor protein complex that uses ATP hydrolysis to trans-locate towards microtubule minus-ends. Dynein associates with the multiprotein dynactin complex which increases dynein processiv-ity and tethers dynein to its cargo proteins20. The dynactin complex includes the rod-shaped p150 dynactin protein, which directly binds dynein subunits and associates with microtubule plus-ends through its CAP-Gly domain; dynein complex structure and function is con-served from fungi to mammals. Other regulatory proteins include Lis1 (mutated in lissencephaly), a WD40 domain protein that directly binds dynein and dynactin subunits, and Nde (formerly mNudE) and Ndl (NUDEL), two coiled-coil proteins that bind Lis1. The dynein–dynactin complex regulates many processes, including organelle posi-tioning, centrosome separation and spindle orientation. In animal cells, dynein, dynactin and Lis1 are all required for positioning the mitotic spindle in response to cortical polarity cues. Coupling of cell polarity and dynein function can be mediated through the coiled-coil protein NuMA (homologous to C. elegans LIN-5 and Drosophila Mud), which mediates a physical link between dynein–dynactin–Lis1 and G-protein regulators LGN–AGS-3 (GPR-1/2 in C. elegans and Pins in Drosophila). These G-protein regulators show polarized locali-zation during asymmetric division in response to the activity of the Par protein complex in C. elegans blastomeres, Drosophila neuroblasts and mammalian cortical progenitor cells Drosophila (see Box 3).

BOX 1 The dynein–dynactin complex: an evolutionarily conserved spindle force generator

The establishment of cell polarity in many animal cells requires the partitioning defective (Par) complex, first discovered in pioneering genetic screens done by Ken Kemphues and collaborators in C. ele‑gans. The Par complex contains three proteins: Par-3, a PDZ domain scaffolding protein (Bazooka in Drosophila); Par-6, a CRIB and PDZ domain protein and atypical protein kinase C (aPKC; PKC-3 in C. elegans). Par-3 is required for the polarized cortical localization of Par-6–aPKC; Par-6 regulates the kinase activity of aPKC (Par-6 alone inhibits aPKC, but Par-6 bound to Cdc42 or Rac1 monomeric GTPases activates aPKC) and aPKC is the effector of the Par complex that regulates cortical polarity by phosphorylating and driving target proteins off the cortex. The Par complex localizes to the apical cortex in invertebrate and vertebrate epithelial cells (including epidermal and neural progenitors), to the anterior cortex in the C. elegans zygote and to the apical cortex in Drosophila neuroblasts. At present it is believed that the Par complex is required for establishing cortical polarity in all metazoan cell types that undergo regulated spindle orientation, and in each case defects in Par complex localization lead to aberrant spindle orientation. Par-dependent spindle positioning is mediated through polarized localization of the G-protein binding proteins LGN–AGS3 (mammals), GPR-1/2 (C. elegans) and Pins (Drosophila), and the associated NuMA (LIN-5, Mud) proteins. In mammalian and Drosophila cells, the ankyrin protein Insc provides a physical link between Par-3–Baz and LGN–AGS3–Pins, thereby coupling cortical cell polarity and spindle position (see Box 3).

BOX 2 The Par complex: a conserved regulator of cortical polarity



R E V I E W

Centrosome and spindle positioning forces. We will first briefly discuss NCC centration, and then focus on NCC rotation and spindle positioning mechanisms. NCC centration requires the dynein–dynactin complex but not cortical Par polarity31. This suggests that the force driving centration is provided by cytoplasm-anchored dynein–dynactin complex: the pos-teriorly located NCC generates longer microtubules in the anterior direc-tion, which consequently have more associated dynein–dynactin complex motors, thereby pulling the NCC anteriorly until all forces are balanced at centration32,33 (Fig. 2a). However, recent work shows that cortical polar-ity has a supporting role in centration. GPR-1/2 and LIN-5, components of the receptor-independent heterotrimeric G-protein pathway (Box 3), are transiently enriched at the anterior cortex at the time of centration, and are required for its timely occurence34. LIN-5 is known to interact with members of the dynein–dynactin complex35,36, which could provide an anterior-directed pulling force. In addition, disruption of microtubule interactions with the cortical acto-myosin network slows centration37. It seems likely that cytoplasmic dynein is sufficient for centration, with

cortical dynein and the acto-myosin network providing an additional anterior-directed force.

Data from spindle severing experiments, genetics and theoretical mod-elling show that both spindle orientation and positioning are driven by the attachment of astral microtubule plus-ends to cortically-anchored dynein–dynactin complex, resulting in pulling forces moving the centro-somes and spindle towards the cortex28,31,35,36,38–41 (Fig. 2b–c). In contrast to budding yeast, microtubules initiate end-on contact with the cortex in C. elegans42,43, although lateral microtubule-cortex interactions have also been proposed44. End-on engagement of cortical dynein must be coordinated with microtubule plus-end depolymerization to avoid coun-teracting force due to microtubule plus-ends pushing against the cortex, and indeed time-lapse imaging shows that microtubules maintain end-on contact with the cortex only transiently (≤ 1 sec) before they undergo catastrophe and depolymerize43. NCC rotation is due to differential pull-ing forces on each centrosome; however, both centrosomes seem to have similar initial positions relative to anterior-posterior cortical polarity cues

P0

Posteriorblastomere

Anteriorblastomere

aCytoplasmic anchors

dyneinLis1dynactin

Lin-5GPR-1/2Gα

LET -99Actinnetwork

PAR-3PAR-6PKC-3GPR-1/2

PAR-2PAR-1LET-99

NCC centration

b Spindle orientation

Spindle positioningc

Figure 2 Spindle orientation and positioning in the C. elegans zygote. (a) The nucleus/centrosomal complex (NCC) moves anteriorly to the cell centre (centration) primarily owing to activity of dynein–dynactin complex anchored to an unknown substrate in the cytoplasm (activation of cortical dynein by the GPR-1/2 complex may also contribute, b). (b) NCC rotation (spindle orientation) aligns the centrosomes along the anterior-posterior axis due to the combined activity of cortical Par polarity proteins, the cortical GPR-1/2 complex and associated cortical

dynein–dynactin complex. (c) At anaphase, GPR-1/2 is enriched at the posterior cortex, where it activates cortical dynein resulting in posterior spindle displacement and generation of a larger anterior and smaller posterior blastomere. LET-99 is enriched in a lateral cortical belt and restricts dynein activation in this domain by restricting cortical GPR-1/2 localization. Light red arrowheads indicate the direction of dynein motion; bold black arrows indicate the direction of the net spindle positioning force. Anterior is to the left and posterior to the right, in all panels.



R E V I E W

and, in contrast to yeast, no molecular or morphological centrosome asymmetry is apparent. Therefore it is likely that an initial stochastic difference in force on the two microtubule asters is amplified through a positive feedback loop that results in anterior-posterior alignment of the NCC, although molecular components of such a feed-back loop are unknown. In contrast, during anaphase spindle-positioning anterior and posterior microtubule asters are probing largely distinct cortical compart-ments, and the posterior spindle movement is due to larger net pulling forces acting on the posterior spindle pole as a result of this cortical polar-ity28,35,36,39,40. Dynein–dynactin complex and Lis1 are detected uniformly in the cytoplasm, at the cell cortex and on microtubules, throughout the cell cycle31,38, raising the question of how dynein pulling forces are temporally and spatially regulated to confer posterior-directed spindle positioning. A series of papers convincingly demonstrate that heterotrimeric G-proteins have a central role in this process.

Heterotrimeric G‑proteins control dynein‑dependent spindle posi‑tioning. The Gα(GAO-1 or GPA-16)–Gβ–Gγ heterotrimeric complex is inactive, but on dissociation both Gα and Gβ–Gγ are activated (Box 3). All three subunits show uniform cortical localization during the first zygotic division, and each is required for proper centrosome migration, spindle orientation and spindle positioning45–47. Reduction in Gα levels or activity results in spindle-positioning defects39,46–48, similar to those caused by depletion of dynein–dynactin35,36. In contrast, Gβ inhibition increases free Gα levels resulting in the opposite phenotype of exces-sive centrosome movements48–50. Thus Gα is the protein required for spindle positioning, whereas the Gβ–Gγ dimer attenuates Gα function by sequestering Gα into a non-functional complex. This raises the ques-tions of how cortical Par proteins regulate the receptor-independent heterotrimeric G-protein pathway, and how free Gα activates dynein–dynactin complex.

The answer to both questions involves the TPR/GoLoco domain protein GPR-1/2 (Pins in flies; LGN/AGS-3 in mammals). Par polarity cues result in the enrichment of GPR-1/2 at the anterior cortex during prophase when NCC centration and rotation occurs, and at the posterior cortex during anaphase spindle positioning34,48,50,51 (Fig. 2). Binding of GPR-1/2 to Gα activates both proteins: GPR-1/2 activates Gα by displac-ing Gβ–Gγ 48,51,52, and Gα activates GPR-1/2 by preventing TPR/GoLoco intramolecular interactions53,54, thus making the TPR domain available for intermolecular interactions. The ‘opened’ GPR-1/2 uses its TPR domain to bind LIN-5 (Mud in flies; NuMA in mammals)52, and LIN-5 and GPR-1/2 can associate with the dynein activator Lis1, and dynein itself, to exert a spindle pulling force35,36 (Box 3). This model is supported by in vivo and in vitro protein interactions, protein localization epista-sis, genetic interactions and the observation that reducing levels of Gα, GPR-1/2 or LIN-5 gives the same spindle positioning defects as loss of dynein–dynactin function31,35,36,38,46,47,52,55.

Recent work has helped define the mechanism that translates Par-polarity into cortical GPR-1/2 asymmetry, which is critical for polarized cortical dynein activation and directionality of cortex-spindle force pro-duction. LET-99, a DEP domain G-protein regulator, has a central role this process by inhibiting cortical association of GPR-1/2. It is localized in a lateral cortical belt in response to Par polarity, which promotes the exclusion of GPR-1/2 from this domain30,34,50 (Fig. 2a). The mechanism leading to cortical LET-99 enrichment and LET-99-dependent GPR-1/2 exclusion remains to be further investigated. Other questions also remain

unanwered. For example how do LIN-5–Lis1 interactions activate the dynein–dynactin complex? Is this function conserved in flies and mam-mals? Is this the only pathway required for spindle positioning?

Drosophila neuroblasts Neuroblasts are the stem cell-like progenitors of the Drosophila central nervous system. Embryonic neuroblasts delaminate as single cells from an apical/basal polarized neuroectoderm and divide asymmetrically perpendicular to the plane of the neuroectoderm to ‘bud off ’ a series of small ganglion mother cells (GMCs). Larval neuroblasts derive from embryonic neuroblasts but contact a glial cell rather than the neur-oectoderm59. Neuroblast asymmetric division can be subdivided into three steps: (1) apical/basal cortical polarity is established during late interphase/early prophase; (2) spindle orientation along the cell polarity axis is established by prometaphase and (3) spindle position is shifted towards the basal cortex during anaphase. This results in a molecularly and physically asymmetric cell division (Fig. 3).

Cortical polarity determines spindle orientation and position. Neuroblasts show no detectable cortical polarity during most of inter-phase. Apical cortical polarity is first seen at late interphase/early prophase for the Par-complex proteins Bazooka (Baz/Par-3)–Par-6–aPKC (Box 1) and the associated Inscuteable (Insc), Pins and Gαi proteins60 (Box 3). Basal proteins such as Miranda, Prospero and Numb are subsequently localized at prometaphase. In embryonic neuroblasts, loss of Baz or Insc leads to randomization of the spindle orientation relative to the overlying neuroectoderm, whereas loss of basal proteins has no effect on spindle orientation61–64. In larval neuroblasts, reduction of Pins or Gαi uncouples

Heterotrimeric G-protein complexes consist of α, β, and γ subunits tethered to the plasma membrane by lipid modifications on Gα and Gγ subunits. Canonical receptor-dependent heterotrimeric G-protein signalling is activated by ligand-binding to a seven-pass transmembrane receptor, which promotes dissociation of active Gα-GTP from Gβ–Gγ. In contrast, receptor-independent heter-otrimeric G-protein activity utilizes a GEF (Ric-8) to stimulate production of Gα-GTP, followed by a GAP (RGS-7 in C. elegans) generating Gα-GDP which is likely to be the active form in this pathway. Gα-GDP binds a tetratricopeptide (TPR)–GoLoco (Gα-binding) domain protein — for example, Pins (Drosophila), GPR-1/2 (C. elegans) or LGN–AGS3 (mammals) — and activates this protein by disrupting intramolecular TPR–GoLoco interac-tions. The ‘open’ TPR–GoLoco protein then binds a coiled-coil NuMA-related protein (LIN-5 in C. elegans, Mud in Drosophila and NuMA in mammals56–58). In this way, Gα-GDP triggers the formation of a tripartite protein complex (for example, Gα–Pins–Mud) that is required for spindle orientation. In Drosophila and mammals, the Insc protein can link the Pins–LGN–AGS3 to the Par complex. Whether the Pins–LGN–AGS3 TPRs can bind NuMA–Mud and Insc concurrently is unknown. The spindle posi-tioning function of this receptor-independent G-protein pathway is, at least in part, mediated through dynein–dynactin–Lis1, which has been shown to physically associate with mammalian NuMA and C. elegans LIN-5. This interaction is likely to be conserved in Drosophila, although it has not been tested.

BOX 3 Receptor-independent heterotrimeric G-protein pathway



R E V I E W

spindle alignment from the Baz cortical polarity axis54,62, whereas loss of Par-6, aPKC or basal cortical proteins have no effect on spindle orienta-tion65. Furthermore, loss of Baz and Pins together leads to failure to gener-ate spindle pole asymmetry, absence of basal displacement of the spindle and the production of two equally sized daughter cells66,67. From these observations, it seems that Baz, Insc, Pins and Gαi are the key polarity proteins regulating spindle orientation and spindle positioning, and that they may function through more than one pathway.

Spindle orientation pathways. Spindle orientation has been character-ized in both embryonic and larval neuroblasts. In both, centrosomes remain associated with the apical cortex during interphase68–70, despite the lack of any known cortical polarity cues. At prophase, one centri-ole pair moves basally to establish the bipolar mitotic spindle; although occasionally both centrosomes move 90 ̊ from the apical cortex during centrosome separation, one spindle pole always rapidly resumes contact with the apical cortex58,68,69,71. The mitotic spindle undergoes gentle rock-ing movements during metaphase, showing that microtubules are con-stantly probing the neuroblast cortex and exerting pulling forces, but the spindle never strays far from the apical/basal polarity axis58,68,69,71. These studies suggest that spindle orientation is fixed by the end of prophase, and remains stable despite the rocking movements.

Two pathways are known to regulate neuroblast spindle orientation: the Gα–Pins–Mud pathway and the Pins–Dlg–Khc73 pathway. Gαi, Pins and Mud (homologues of C. elegans Gα, GPR-1/2 and LIN-5, respectively) are members of an evolutionarily-conserved receptor-independent G-protein pathway (Box 3); Pins binds Gαi through its GoLoco domains72–75 and Mud through its TPR domain56–58 to form a tripartite protein complex. This pro-tein complex is linked to the apical Par complex by the adapter protein Insc, which binds Baz and the TPR domain of Pins64,72,74,75 . Recent data show that the PDZ-domain protein Canoe also associates with Pins in vivo76. Reducing the level of Gαi, Pins, Mud or Canoe prevents spindle alignment to the apical Par complex54,56–58,72–76. Mammalian and C. elegans orthologues of Mud (NuMA, LIN-5) associate with components of the dynein–dynactin complex35,36,77, and the dynein–dynactin complex plus Lis1 are required for dynamic spindle rocking and spindle orientation in Drosophila larval neuroblasts71. Thus, it is likely that the Gαi–Pins–Mud pathway works by recruiting the dynein–dynactin complex to the apical cortex, which exerts a pulling force to recruit and maintain one centrosome at the apical pole, thereby aligning the mitotic spindle along the apical/basal polarity axis. However, dynein–dynactin complex proteins have not been detected at the apical cortex of neuroblasts71,78, and an interaction between Mud and the dynein–dynactin complex remains to be documented.

A second spindle orientation pathway involves Pins, the tumor suppres-sor Discs large (Dlg; a PSD95 family member containing PDZ, SH3 and GK domains) and the microtubule plus-end-directed kinesin heavy chain 73 (Khc73; a kif13A-related protein). This pathway was discovered owing to the ability of astral microtubules and Khc73 to induce the formation of Dlg–Pins–Gαi crescents in embryonic neuroblasts lacking a functional Par complex62. The study showed that Khc73 localizes to microtubule plus-ends (after taxol stabilization), Khc73 binds Dlg in vitro and in vivo, similar to interactions between their mammalian orthologues79 and Dlg co-immunoprecipitates with Pins62. Thus, Khc73+ astral microtubules can induce Pins–Dlg cortical polarity. Interestingly, reducing Dlg or Khc73 levels leads to partial spindle-orientation defects without affecting apical Pins–Gαi cortical polarity, suggesting a ‘reverse’ signal flow with corti-cal Pins–Dlg directing spindle orientation through Khc73 (ref. 62). It is unknown whether the partial phenotype is due to residual Dlg or Khc73 protein remaining in the mutant or RNAi background, or whether it is because of redundancy with the Gαi–Pins–Mud pathway.

Spindle pole asymmetry. Drosophila larval neuroblasts show pro-nounced mitotic spindle pole asymmetry at anaphase: the apical spin-dle pole contains more pericentrosomal Centrosomin (Cnn), nucleates longer astral microtubules and the spindle is shifted towards the basal cortex68,69. Although spindle pole asymmetry is necessary for spindle orientation in yeast, it is not essential for spindle orientation in neurob-lasts. Mutants in cnn, sas‑4, asl or spd‑2 have morphologically identical spindle poles, yet most of these neuroblasts undergo normal asymmetric cell division80–83.

Spindle orientation and cell fate. Mutants in genes such as aurora‑A, mud and polo have both spindle orientation defects and increased neuroblast numbers56,84–86, suggesting that precise spindle alignment is required for normal neuroblast/GMC fate. However, aurora‑A and polo mutants have additional cell polarity defects, which complicates inter-pretation. In contrast, mud mutants show defects in spindle orientation and a moderate increase in neuroblast number yet have normal cortical

a

?

b

GαiPinsMud

dyneinLis1dynactin

Neuroblast(self-renewal)

Baz

Par-6aPKC

Cdc42

Insc

Dlg

Khc73

GMC (neuronal differentiation)

Adherensjunction

Actinnetwork

cCentrosomeseparation

Spindleorientation

Cytokinesis

Cnn

Figure 3 Spindle orientation and positioning in the Drosophila neuroblast. (a) Late interphase/prophase. Par-proteins (Baz, Par-6 and aPKC) and Cdc42 (associated through Par-6) are enriched at the apical cortex. One centrosome is anchored at the apical cortex by Gα–Pins–Mud (possibly through dynein–dynactin, although this remains to be tested) and by Pins–Dlg–Khc73. The second centrosome nucleates few microtubules and migrates basally. (b) Prometaphase/metaphase. Tight coupling of the spindle to the apical/basal polarity axis requires the motor proteins Khc73 and dynein; Khc73 binds Dlg and may facilitate cortical microtubule anchoring, whereas spindle positioning forces on microtubules are probably due to dynein complex activity. The Insc protein directly binds Baz and Pins, thereby coupling Par polarity with Mud- and Dlg–Khc73-dependent spindle positioning pathways. (c) Anaphase. The mitotic spindle becomes asymmetric leading to unequal sized daughter cells. Light red arrowheads indicate direction of dynein motion; bold black arrows indicate the direction of the net spindle positioning force.



R E V I E W

polarity56–58, leading to the model that divisions with a transverse spindle lead to the formation of two neuroblasts through symmetric division. This is an attractive model, but it has not been rigorously tested. Time-lapse imaging studies of neuroblasts undergoing transverse divisions are needed to determine the cell fate of their progeny, and whether cell fate is correlated with inheritance of the apical or basal cortical domain.

Mammalian neuroepitheliaIt has been proposed that spindle orientation regulates the determination of cell fate, the timing of neurogenesis and the evolution of brain size in mammals, but recent results suggest that spindle orientation might in fact have little or no effect on these processes. Here we review the evidence for regulated spindle orientation in the cerebral cortex and retina, highlight mechanistic similarities with other model systems and discuss the rela-tionship between spindle orientation and sibling cell-fate specification.

Cortical cell polarity. The mammalian cerebral cortex and retina both contain multipotent neuroepithelial progenitors with pronounced api-cal/basal polarity (Fig. 4a). Their very small apical cortical domain or

‘apical endfoot’ contains Cdc42–Par-3–aPKC–Par-6, and the transmem-brane protein Prominin (CD133)87. Their extensive basolateral domain or ‘basal process’ contains LGN protein (a mammalian GPR-1/2 and Pins orthologue; Box 3)88,89, and the two domains are separated by adherens junctions containing E-cadherin, α-catenin and β-catenin87.

Planar spindle orientation. When the mitotic spindle is aligned per-pendicular to the neuroepithelial progenitor apical/basal axis — that is, in the plane of the neuroepithelium — it is termed planar spindle orientation. True planar spindle orientation (‘planar symmetric’, Fig. 4a, left) results in a cleavage furrow that bisects the apical membrane domain to generate two molecularly identical neuroepithelial cells. However, some apparently planar divisions in fact partition the apical domain to just one cell, resulting in a ‘planar symmetric’ cell division. Only very recent studies using molecular markers have distinguished these two forms of planar cell divisions88–90. Planar spindle orientation requires the basolateral LGN protein, a member of the LGN/Pins family (Box 3). Inactivation of LGN function randomizes spindle orientation during the early proliferative phase of neuroepithelial progenitor divisions in

VZ

SVZ

Apical

Basal

IZ

CP

a

Gα (?)LGNNuMA (?)

dynactinLis1dynein

Par-3Par-6aPKC

mInsc

Adherensjunction

b

Planar symmetric Apical-basal

?

?

?

Planar

??

Apical-basal(mInsc misexpressed)

??

c

ASPM

Planar asymmetric

Cdc42

Figure 4 Spindle orientation and positioning in the mammalian neuroepithelium. (a) The mammalian cortex is a pseudostratified epithelium with morphologically distinct layers, including cortical plate (CP), intermediate zone (IZ), subventricular zone (SVZ) and ventricular zone (VZ). Neural epithelial cells or radial glia progenitors in the VZ (tan cells) can undergo planar and molecularly symmetric divisions (left); planar but molecularly asymmetric divisions (centre) or apical/basal molecularly asymmetric divisions (right). Molecularly asymmetric divisions generating a basal progenitor/neuron (green) and a neuroepithelial progenitor (tan) are shown, but the fate of sibling cells after each type of neuroepithelial progenitor division is now controversial (see text). (b) Planar divisions require basolateral LGN protein and Lis1 (dynein) function; it is likely but not proven that dynein activation involves Gα

signalling and is linked to LGN through NuMA. In addition, dynein controls planar apical spindle positioning through direct binding to adherens junction components. The centrosome-associated ASPM protein is required for planar spindle orientation, but the mechanism is unknown. (c) Misexpressed mInsc colocalizes with the apical Cdc42–Par complex and increases the frequency of apical-basal divisions. mInsc can associate with LGN and the related AGS3 protein, thus it is possible that apical/basal spindle reorientation involves relocalization of Gα–LGN(AGS-3)–NuMA to the apical cortex and subsequent dynein engagement of apically positioned astral microtubules. Support for this model awaits further experimental testing. Light red arrowheads indicate direction of dynein motion; bold black arrows indicate the direction of the net spindle positioning force.



R E V I E W

mouse88 and chick89. How does LGN induce planar spindle orientation? Apical/basal polarity markers and adherens junction markers are normal after LGN knockdown89, so LGN does not act by disrupting cell polarity. LGN binds NuMA, which associates with the dynein–dynactin com-plex77,91, suggesting that LGN recruits NuMA–dynein–dynactin to the basolateral domain (although this has not yet been shown). In support of this model, LGN recruits NuMA to the cortex overlying the spindle poles in epithelial Madin-Darby canine kidney (MDCK) cells53, and reducing the level of dynein regulators Lis1/Nde1 disrupts planar spindle orienta-tion in mouse cortical neuroepithelial progenitors92–94. How is the planar spindle located apically, and not randomly, within the extensive LGN+ basolateral domain? The subapical adherens junctions may provide an additional spindle orientation cue, as the adherens junction component β-catenin binds dynein95. Taken together, these data support a model in which subapical adherens junctions and basolateral LGN/NuMA utilize the evolutionarily-conserved spindle orientation dynein–dynactin com-plex to promote planar spindle orientation.

Proper spindle orientation requires both a cortical capture site and dynamic microtubules that probe the cortex, and indeed mutations in centrosomal and microtubule -binding proteins are known to affect pla-nar spindle orientation. Mice lacking the centrosomal protein ASPM (abnormal spindle-like microcephaly associated protein) or the micro-tubule -binding protein DCLK (Doublecortin-like kinase) have defects in planar spindle orientation and small brain size96–98. Consistent with these findings, mutations in the human ASPM gene, and five other genes, are linked to autosomal recessive primary microcephaly99. These genes may encode new components of the planar spindle positioning pathway, and two (CDK5Rap2 and CenpJ) have Drosophila orthologues (Cnn and Sas‑4) required for spindle orientation in neuroblasts80,83.

Apical/basal spindle orientation. When the neuroepithelial progeni-tor mitotic spindle is aligned parallel to the apical/basal axis it is termed ‘apical/basal spindle orientation’ (Fig. 4a, right). True apical/basal spin-dle orientation seems to be relatively rare during all phases of cortical neurogenesis88, although it may be more common in the retina100,101 (and in epidermal progenitors102). In the retina, mouse Insc (mInsc) is api-cally localized in progenitors undergoing apical/basal divisions, and a reduction in mInsc levels results in persistent planar spindle orienta-tion that expands the progenitor pool103. How does mInsc anchor one spindle pole at the apical cortex? The mechanism is likely to be highly conserved from flies to mammals. mInsc binds LGN–AGS3102,103, which has the potential to recruit the LGN–AGS3 planar spindle orientation pathway to the apical cortex, thereby reorienting the mitotic spindle. In the mouse cortex, depletion of AGS3 is reported to switch apical/basal to planar spindle orientation104. LGN–AGS3-associated NuMA may associate with microtubule -bound dynein77,91, to pull one spindle pole to the apical cortex. However, it remains to be seen whether NuMA or dynein–dynactin complex proteins are localized to the apical cortex in cells undergoing apical/basal spindle orientation.

Spindle orientation and cell fate. It has been proposed that spindle orientation regulates cell fate in the cortex and retina, with planar divisions generating two identical cell fates (neuroepithelial pro-genitor/neuroepithelial progenitor or neuron/neuron, depending on whether the division occurs early or late in neurogenesis), and apical/basal spindle orientation generating two different cell types

(for example, neuroepithelial progenitor/neuron)101,105,106. Testing this model has become possible with advances in live imaging methods, allowing neuroepithelial progenitor apical and basolateral domains to be followed from cell division to sibling cell fate specification. Initial experiments reported that inheritance of the apical cortical domain was a good predictor of neuroepithelial progenitor fate in the mouse cortex90,104, although more recent work shows that only cells that inherit both the apical and basal domain acquire neuroepithelial progenitor fate88. Other labs report that spindle orientation has no effect on progenitor fate, but merely regulates the position of the cells within the neuroepithelium; that is, cells lacking the apical domain and/or adherens junctions move away from the ventricular zone but retain neuroepithelial progenitor fate89. Consistent with this finding, reducing RhoA function increases apical/basal neuroepithelial pro-genitor divisions, but the basal daughter cell maintains neuroepithelial progenitor fate107.

In the retina, evidence for a causal relationship between spindle orien-tation and cell fate is more convincing. In the retinal neuroepithelium, Numb protein is localized apically (whether to the apical domain or the adherens junctions is unknown). Planar divisions give rise to two Numb+ photoreceptor neurons, whereas apical/basal divisions typically generate two molecularly distinct siblings (Numb+/Numb–) that assume differ-ent cell fates101,108. Reduction of mInsc reduces the frequency of apical/basal divisions and increases the frequency of symmetric proliferative divisions103. It is thought that Numb promotes neuronal differentiation, because overexpression of Numb in progenitor cells lead to formation of photoreceptor neurons at the expense of Muller glia101. Numb may act by inhibiting Notch, as in Drosophila, because overexpression of Notch results in the opposite phenotype of excess Muller glia at the expense of photoreceptors109.

COnCluSIOnSGreat progress has been made over the past few years in revealing con-served mechanisms of spindle orientation from yeast to mammals, yet much remains to be learned. The next few years should reveal more about the biochemical mechanisms used to assemble and activate the protein complexes regulating spindle orientation. More difficult will be defining the role of spindle orientation in specifying cell fate. Future studies will have to identify cell fate determinants, observe their distri-bution relative to spindle orientation and track the subsequent sibling cell fates. This is experimentally challenging, but advances in imaging technology have made these experiments possible in worms, flies and even the mammalian brain.

ACKnowleDgementSWe thank S. Siegrist for discussions; B. Bowerman, C. Cabernard, C. Johnston, K. Prehoda and S. Siegrist for comments on the manuscript and L. Chen in whose lab this work was completed. We apologize to all authors whose primary papers could not be cited because of space constraints.

Published online at http://www.nature.com/naturecellbiology/ Reprints and permissions information is available online at http://npg.nature.com/reprintsandpermissions/

1. Pearson, C. G. & Bloom, K. Dynamic microtubules lead the way for spindle positioning. Nature Rev. Mol. Cell Biol. 5, 481–492 (2004).

2. Thery, M. & Bornens, M. Cell shape and cell division. Curr. Opin. Cell Biol. 18, 648–657 (2006).

3. Pruyne, D. & Bretscher, A. Polarization of cell growth in yeast. I. Establishment and maintenance of polarity states. J. Cell Sci. 113 (Pt 3), 365–375 (2000).

4. Miller, R. K., D’Silva, S., Moore, J. K. & Goodson, H. V. The CLIP-170 orthologue Bik1p and positioning the mitotic spindle in yeast. Curr. Top. Dev. Biol. 76, 49–87 (2006).



http://www.nature.com/naturecellbiology/

http://npg.nature.com/reprintsandpermissions/


R E V I E W

5. Hwang, E., Kusch, J., Barral, Y. & Huffaker, T. C. Spindle orientation in Saccharomyces cerevisiae depends on the transport of microtubule ends along polarized actin cables. J. Cell Biol. 161, 483–488 (2003).

6. Adames, N. R. & Cooper, J. A. Microtubule interactions with the cell cortex caus-ing nuclear movements in Saccharomyces cerevisiae. J. Cell Biol. 149, 863–874 (2000).

7. Huisman, S. M. et al. Differential contribution of Bud6p and Kar9p to microtubule capture and spindle orientation in S. cerevisiae. J. Cell Biol. 167, 231–244 (2004).

8. Kusch, J., Meyer, A., Snyder, M. P. & Barral, Y. Microtubule capture by the cleav-age apparatus is required for proper spindle positioning in yeast. Genes Dev. 16, 1627–1639 (2002).

9. Segal, M., Bloom, K. & Reed, S. I. Bud6 directs sequential microtubule interactions with the bud tip and bud neck during spindle morphogenesis in Saccharomyces cer-evisiae. Mol. Biol. Cell 11, 3689–3702 (2000).

10. Drewes, G., Ebneth, A., Preuss, U., Mandelkow, E. M. & Mandelkow, E. MARK, a novel family of protein kinases that phosphorylate microtubule-associated proteins and trigger microtubule disruption. Cell 89, 297–308 (1997).

11. Eshel, D. et al. Cytoplasmic dynein is required for normal nuclear segregation in yeast. Proc. Natl Acad. Sci. USA 90, 11172–11176 (1993).

12. Li, Y. Y., Yeh, E., Hays, T. & Bloom, K. Disruption of mitotic spindle orientation in a yeast dynein mutant. Proc. Natl Acad. Sci. USA 90, 10096–10100 (1993).

13. Carminati, J. L. & Stearns, T. Microtubules orient the mitotic spindle in yeast through dynein-dependent interactions with the cell cortex. J. Cell Biol.138, 629–641 (1997).

14. Lee, W. L., Oberle, J. R. & Cooper, J. A. The role of the lissencephaly protein Pac1 during nuclear migration in budding yeast. J. Cell Biol. 160, 355–364 (2003).

15. Sheeman, B. et al. Determinants of S. cerevisiae dynein localization and activation: impli-cations for the mechanism of spindle positioning. Curr. Biol. 13, 364–372 (2003).

16. Farkasovsky, M. & Kuntzel, H. Cortical Num1p interacts with the dynein intermedi-ate chain Pac11p and cytoplasmic microtubules in budding yeast. J. Cell Biol. 152, 251–262 (2001).

17. Heil-Chapdelaine, R. A., Tran, N. K. & Cooper, J. A. Dynein-dependent movements of the mitotic spindle in Saccharomyces cerevisiae do not require filamentous actin. Mol. Biol. Cell 11, 863–872 (2000).

18. Farkasovsky, M. & Kuntzel, H. Yeast Num1p associates with the mother cell cortex during S/G2 phase and affects microtubular functions. J. Cell Biol. 131, 1003–1014 (1995).

19. Knaus, M. et al. The Bud14p-Glc7p complex functions as a cortical regulator of dynein in budding yeast. EMBO J. 24, 3000–3011 (2005).

20. Schroer, T. A. Dynactin. Annu. Rev. Cell Dev. Biol. 20, 759–779 (2004).21. Grava, S., Schaerer, F., Faty, M., Philippsen, P. & Barral, Y. Asymmetric recruitment of

dynein to spindle poles and microtubules promotes proper spindle orientation in yeast. Dev. Cell 10, 425–439 (2006).

22. Liakopoulos, D., Kusch, J., Grava, S., Vogel, J. & Barral, Y. Asymmetric loading of Kar9 onto spindle poles and microtubules ensures proper spindle alignment. Cell 112, 561–574 (2003).

23. Maekawa, H., Usui, T., Knop, M. & Schiebel, E. Yeast Cdk1 translocates to the plus end of cytoplasmic microtubules to regulate bud cortex interactions. EMBO J. 22, 438–449 (2003).

24. Moore, J. K., D’Silva, S. & Miller, R. K. The CLIP-170 homologue Bik1p promotes the phosphorylation and asymmetric localization of Kar9p. Mol. Biol. Cell 17, 178–191 (2006).

25. Moore, J. K. & Miller, R. K. The cyclin-dependent kinase Cdc28p regulates multiple aspects of Kar9p function in yeast. Mol. Biol. Cell 18, 1187–1202 (2007).

26. Maekawa, H. & Schiebel, E. Cdk1-Clb4 controls the interaction of astral microtubule plus ends with subdomains of the daughter cell cortex. Genes Dev. 18, 1709–1724 (2004).

27. Schneider, S. Q. & Bowerman, B. Cell polarity and the cytoskeleton in the Caenorhabditis elegans zygote. Annu. Rev. Genet. 37, 221–249 (2003).

28. Grill, S. W., Gonczy, P., Stelzer, E. H. & Hyman, A. A. Polarity controls forces governing asymmetric spindle positioning in the Caenorhabditis elegans embryo. Nature 409, 630–633 (2001).

29. Cheng, N. N., Kirby, C. M. & Kemphues, K. J. Control of cleavage spindle orientation in Caenorhabditis elegans: the role of the genes par-2 and par-3. Genetics 139, 549–559 (1995).

30. Tsou, M. F., Hayashi, A., DeBella, L. R., McGrath, G. & Rose, L. S. LET-99 determines spindle position and is asymmetrically enriched in response to PAR polarity cues in C. elegans embryos. Development 129, 4469–4481 (2002).

31. Gonczy, P., Pichler, S., Kirkham, M. & Hyman, A. A. Cytoplasmic dynein is required for distinct aspects of MTOC positioning, including centrosome separation, in the one cell stage Caenorhabditis elegans embryo. J. Cell Biol. 147, 135–150 (1999).

32. Kimura, A. & Onami, S. Computer simulations and image processing reveal length-dependent pulling force as the primary mechanism for C. elegans male pronuclear migration. Dev. Cell 8, 765–775 (2005).

33. Reinsch, S. & Gonczy, P. Mechanisms of nuclear positioning. J. Cell Sci. 111 (Pt 16), 2283–2295 (1998).

34. Park, D. H. & Rose, L. S. Dynamic localization of LIN-5 and GPR-1/2 to cortical force generation domains during spindle positioning. Dev. Biol. 315, 42–54 (2008).

35. Nguyen-Ngoc, T., Afshar, K. & Gonczy, P. Coupling of cortical dynein and G a pro-teins mediates spindle positioning in Caenorhabditis elegans. Nature Cell Biol. 9, 1294–1302 (2007).

36. Couwenbergs, C. et al. Heterotrimeric G protein signaling functions with dynein to promote spindle positioning in C. elegans. J. Cell Biol. 179, 15–22 (2007).

37. Goulding, M. B., Canman, J. C., Senning, E. N., Marcus, A. H. & Bowerman, B. Control of nuclear centration in the C. elegans zygote by receptor-independent Gα signaling and myosin II. J. Cell Biol. 178, 1177–1191 (2007).

38. Cockell, M. M., Baumer, K. & Gonczy, P. lis-1 is required for dynein-dependent cell division processes in C. elegans embryos. J. Cell Sci. 117, 4571–4582 (2004).

39. Grill, S. W., Howard, J., Schaffer, E., Stelzer, E. H. & Hyman, A. A. The distribution of active force generators controls mitotic spindle position. Science 301, 518–521 (2003).

40. Pecreaux, J. et al. Spindle oscillations during asymmetric cell division require a thresh-old number of active cortical force generators. Curr. Biol. 16, 2111–2122 (2006).

41. Schmidt, D. J., Rose, D. J., Saxton, W. M. & Strome, S. Functional analysis of cyto-plasmic dynein heavy chain in Caenorhabditis elegans with fast-acting temperature-sensitive mutations. Mol. Biol. Cell 16, 1200–1212 (2005).

42. Labbe, J. C., Maddox, P. S., Salmon, E. D. & Goldstein, B. PAR proteins regulate micro-tubule dynamics at the cell cortex in C. elegans. Curr. Biol. 13, 707–714 (2003).

43. Kozlowski, C., Srayko, M. & Nedelec, F. Cortical microtubule contacts position the spindle in C. elegans embryos. Cell 129, 499–510 (2007).

44. Tsou, M. F., Ku, W., Hayashi, A. & Rose, L. S. PAR-dependent and geometry-dependent mechanisms of spindle positioning. J. Cell Biol. 160, 845–855 (2003).

45. Zwaal, R. R. et al. G proteins are required for spatial orientation of early cell cleavages in C. elegans embryos. Cell 86, 619–629 (1996).

46. Miller, K. G. & Rand, J. B. A role for RIC-8 (Synembryn) and GOA-1 (G(o)α) in regulat-ing a subset of centrosome movements during early embryogenesis in Caenorhabditis elegans. Genetics 156, 1649–1660 (2000).

47. Gotta, M. & Ahringer, J. Distinct roles for Gα and Gβγ in regulating spindle position and orientation in Caenorhabditis elegans embryos. Nature Cell Biol. 3, 297–300 (2001).

48. Colombo, K. et al. Translation of polarity cues into asymmetric spindle positioning in Caenorhabditis elegans embryos. Science 300, 1957–1961 (2003).

49. Afshar, K. et al. RIC-8 is required for GPR-1/2-dependent Gα function during asym-metric division of C. elegans embryos. Cell 119, 219–230 (2004).

50. Tsou, M. F., Hayashi, A. & Rose, L. S. LET-99 opposes Gα/GPR signaling to generate asymmetry for spindle positioning in response to PAR and MES-1/SRC-1 signaling. Development 130, 5717–5730 (2003).

51. Gotta, M., Dong, Y., Peterson, Y. K., Lanier, S. M. & Ahringer, J. Asymmetrically distrib-uted C. elegans homologs of AGS3/PINS control spindle position in the early embryo. Curr. Biol. 13, 1029–1037 (2003).

52. Srinivasan, D. G., Fisk, R. M., Xu, H. & van den Heuvel, S. A complex of LIN-5 and GPR proteins regulates G protein signaling and spindle function in C elegans. Genes Dev. 17, 1225–1239 (2003).

53. Du, Q. & Macara, I. G. Mammalian Pins is a conformational switch that links NuMA to heterotrimeric G proteins. Cell 119, 503–516 (2004).

54. Nipper, R. W., Siller, K. H., Smith, N. R., Doe, C. Q. & Prehoda, K. E. Gαi gener-ates multiple Pins activation states to link cortical polarity and spindle orientation in Drosophila neuroblasts. Proc. Natl Acad. Sci. USA 104, 14306–14311 (2007).

55. Skop, A. R. & White, J. G. The dynactin complex is required for cleavage plane speci-fication in early Caenorhabditis elegans embryos. Curr. Biol. 8, 1110–1116 (1998).

56. Bowman, S. K., Neumuller, R. A., Novatchkova, M., Du, Q. & Knoblich, J. A. The Drosophila NuMA Homolog Mud regulates spindle orientation in asymmetric cell divi-sion. Dev. Cell 10, 731–742 (2006).

57. Izumi, Y., Ohta, N., Hisata, K., Raabe, T. & Matsuzaki, F. Drosophila Pins-binding protein Mud regulates spindle-polarity coupling and centrosome organization. Nature Cell Biol. (2006).

58. Siller, K. H., Cabernard, C. & Doe, C. Q. The NuMA-related Mud protein binds Pins and regulates spindle orientation in Drosophila neuroblasts. Nature Cell Biol. (2006).

59. Doe, C. Q. Neural stem cells: balancing self-renewal with differentiation. Development 135, 1575–1587 (2008).

60. Knoblich, J. A. Mechanisms of asymmetric stem cell division. Cell 132, 583–597 (2008).

61. Shen, C. P., Jan, L. Y. & Jan, Y. N. Miranda is required for the asymmetric localization of Prospero during mitosis in Drosophila. Cell 90, 449–458 (1997).

62. Siegrist, S. E. & Doe, C. Q. Microtubule-induced Pins/Gαi cortical polarity in Drosophila neuroblasts. Cell 123, 1323–1335 (2005).

63. Kraut, R., Chia, W., Jan, L. Y., Jan, Y. N. & Knoblich, J. A. Role of inscuteable in orient-ing asymmetric cell divisions in Drosophila. Nature 383, 50–55 (1996).

64. Wodarz, A., Ramrath, A., Kuchinke, U. & Knust, E. Bazooka provides an apical cue for Inscuteable localization in Drosophila neuroblasts. Nature 402, 544–547 (1999).

65. Rolls, M. M., Albertson, R., Shih, H. P., Lee, C. Y. & Doe, C. Q. Drosophila aPKC regu-lates cell polarity and cell proliferation in neuroblasts and epithelia. J. Cell Biol. 163, 1089–1098 (2003).

66. Fuse, N., Hisata, K., Katzen, A. L. & Matsuzaki, F. Heterotrimeric G proteins regu-late daughter cell size asymmetry in Drosophila neuroblast divisions. Curr. Biol. 13, 947–954 (2003).

67. Cai, Y., Yu, F., Lin, S., Chia, W. & Yang, X. Apical complex genes control mitotic spindle geometry and relative size of daughter cells in Drosophila neuroblast and pI asymmetric divisions. Cell 112, 51–62 (2003).

68. Rusan, N. M. & Peifer, M. A role for a novel centrosome cycle in asymmetric cell divi-sion. J. Cell Biol. 177, 13–20 (2007).

69. Rebollo, E. et al. Functionally unequal centrosomes drive spindle orientation in asym-metrically dividing Drosophila neural stem cells. Dev. Cell 12, 467–474 (2007).

70. Siegrist, S. E. & Doe, C. Q. Extrinsic cues orient the cell division axis in Drosophila embryonic neuroblasts. Development 133, 529–536 (2006).

71. Siller, K. H. & Doe, C. Q. Lis1/dynactin regulates metaphase spindle orientation in Drosophila neuroblasts. Dev. Biol. 319, 1–9 (2008).



R E V I E W

72. Parmentier, M. L. et al. Rapsynoid/partner of inscuteable controls asymmetric division of larval neuroblasts in Drosophila. J. Neurosci. 20, RC84 (2000).

73. Schaefer, M., Petronczki, M., Dorner, D., Forte, M. & Knoblich, J. A. Heterotrimeric G proteins direct two modes of asymmetric cell division in the Drosophila nervous system. Cell 107, 183–194 (2001).

74. Schaefer, M., Shevchenko, A. & Knoblich, J. A. A protein complex containing Inscuteable and the Gα-binding protein Pins orients asymmetric cell divisions in Drosophila. Curr. Biol. 10, 353–362 (2000).

75. Yu, F., Morin, X., Cai, Y., Yang, X. & Chia, W. Analysis of partner of inscuteable, a novel player of Drosophila asymmetric divisions, reveals two distinct steps in inscuteable apical localization. Cell 100, 399–409 (2000).

76. Speicher, S., Fischer, A., Knoblich, J. & Carmena, A. The PDZ protein Canoe regulates the asymmetric division of Drosophila neuroblasts and muscle progenitors. Curr. Biol. 18, 831–837 (2008).

77. Merdes, A., Ramyar, K., Vechio, J. D. & Cleveland, D. W. A complex of NuMA and cyto-plasmic dynein is essential for mitotic spindle assembly. Cell 87, 447–458 (1996).

78. Siller, K. H., Serr, M., Steward, R., Hays, T. S. & Doe, C. Q. Live imaging of Drosophila brain neuroblasts reveals a role for Lis1/dynactin in spindle assembly and mitotic checkpoint control. Mol. Biol. Cell 16, 5127–5140 (2005).

79. Hanada, T., Lin, L., Tibaldi, E. V., Reinherz, E. L. & Chishti, A. H. GAKIN, a novel kinesin-like protein associates with the human homologue of the Drosophila discs large tumor suppressor in T lymphocytes. J. Biol. Chem. 275, 28774–28784 (2000).

80. Megraw, T. L., Kao, L. R. & Kaufman, T. C. Zygotic development without functional mitotic centrosomes. Curr. Biol. 11, 116–120 (2001).

81. Giansanti, M. G., Bucciarelli, E., Bonaccorsi, S. & Gatti, M. Drosophila SPD-2 is an essential centriole component required for PCM recruitment and astral-microtubule nucleation. Curr. Biol. 18, 303–309 (2008).

82. Giansanti, M. G., Gatti, M. & Bonaccorsi, S. The role of centrosomes and astral micro-tubules during asymmetric division of Drosophila neuroblasts. Development 128, 1137–1145 (2001).

83. Basto, R. et al. Flies without centrioles. Cell 125, 1375–1386 (2006).84. Lee, C. Y. et al. Drosophila Aurora-A kinase inhibits neuroblast self-renewal by regulat-

ing aPKC/Numb cortical polarity and spindle orientation. Genes Dev. 20, 3464–3474 (2006).

85. Wang, H. et al. Aurora-A acts as a tumor suppressor and regulates self-renewal of Drosophila neuroblasts. Genes Dev. 20, 3453–3463 (2006).

86. Wang, H., Ouyang, Y., Somers, W. G., Chia, W. & Lu, B. Polo inhibits progenitor self-renewal and regulates Numb asymmetry by phosphorylating Pon. Nature 449, 96–100 (2007).

87. Gotz, M. & Huttner, W. B. The cell biology of neurogenesis. Nature Rev. Mol. Cell Biol. 6, 777–788 (2005).

88. Konno, D. et al. Neuroepithelial progenitors undergo LGN-dependent planar divisions to maintain self-renewability during mammalian neurogenesis. Nature Cell Biol 10, 93–101 (2008).

89. Morin, X., Jaouen, F. & Durbec, P. Control of planar divisions by the G-protein regula-tor LGN maintains progenitors in the chick neuroepithelium. Nature Neurosci. 10, 1440–1448 (2007).

90. Kosodo, Y. et al. Asymmetric distribution of the apical plasma membrane during neuro-genic divisions of mammalian neuroepithelial cells. EMBO J. 23, 2314–2324 (2004).

91. Du, Q., Stukenberg, P. T. & Macara, I. G. A mammalian Partner of inscuteable binds NuMA and regulates mitotic spindle organization. Nature Cell Biol. 3, 1069–1075 (2001).

92. Pawlisz, A. S. et al. Lis1-Nde1-dependent neuronal fate control determines cerebral cortical size and lamination. Hum. Mol. Genet. 17, 2441–2455 (2008).

93. Yingling, J. et al. Neuroepithelial stem cell proliferation requires LIS1 for precise spindle orientation and symmetric division. Cell 132, 474–486 (2008).

94. Feng, Y. & Walsh, C. A. Mitotic spindle regulation by Nde1 controls cerebral cortical size. Neuron 44, 279–293 (2004).

95. Ligon, L. A., Karki, S., Tokito, M. & Holzbaur, E. L. Dynein binds to β-catenin and may tether microtubules at adherens junctions. Nature Cell Biol 3, 913–917 (2001).

96. Fish, J. L., Kosodo, Y., Enard, W., Paabo, S. & Huttner, W. B. Aspm specifically main-tains symmetric proliferative divisions of neuroepithelial cells. Proc. Natl Acad. Sci. USA 103, 10438–10443 (2006).

97. Kouprina, N. et al. The microcephaly ASPM gene is expressed in proliferating tis-sues and encodes for a mitotic spindle protein. Hum. Mol. Genet. 14, 2155–2165 (2005).

98. Shu, T. et al. Doublecortin-like kinase controls neurogenesis by regulating mitotic spindles and M phase progression. Neuron 49, 25–39 (2006).

99. Woods, C. G., Bond, J. & Enard, W. Autosomal recessive primary microcephaly (MCPH): a review of clinical, molecular, and evolutionary findings. Am. J. Hum. Genet. 76, 717–728 (2005).

100.Cayouette, M., Barres, B. A. & Raff, M. Importance of intrinsic mechanisms in cell fate decisions in the developing rat retina. Neuron 40, 897–904 (2003).

101.Cayouette, M. & Raff, M. The orientation of cell division influences cell-fate choice in the developing mammalian retina. Development 130, 2329–2339 (2003).

102.Lechler, T. & Fuchs, E. Asymmetric cell divisions promote stratification and differentia-tion of mammalian skin. Nature 437, 275–280 (2005).

103.Zigman, M. et al. Mammalian inscuteable regulates spindle orientation and cell fate in the developing retina. Neuron 48, 539–545 (2005).

104.Sanada, K. & Tsai, L. H. G protein βγ subunits and AGS3 control spindle orienta-tion and asymmetric cell fate of cerebral cortical progenitors. Cell 122, 119–131 (2005).

105.Chenn, A. & McConnell, S. K. Cleavage orientation and the asymmetric inherit-ance of Notch1 immunoreactivity in mammalian neurogenesis. Cell 82, 631–641 (1995).

106.Noctor, S. C., Martinez-Cerdeno, V., Ivic, L. & Kriegstein, A. R. Cortical neurons arise in symmetric and asymmetric division zones and migrate through specific phases. Nature Neurosci.7, 136–144 (2004).

107.Roszko, I., Afonso, C., Henrique, D. & Mathis, L. Key role played by RhoA in the bal-ance between planar and apico-basal cell divisions in the chick neuroepithelium. Dev. Biol. 298, 212–224 (2006).

108.Cayouette, M., Whitmore, A. V., Jeffery, G. & Raff, M. Asymmetric segregation of Numb in retinal development and the influence of the pigmented epithelium. J. Neurosci. 21, 5643–5651 (2001).

109.Furukawa, T., Mukherjee, S., Bao, Z. Z., Morrow, E. M. & Cepko, C. L. rax, Hes1, and notch1 promote the formation of Muller glia by postnatal retinal progenitor cells. Neuron 26, 383–394 (2000).



news and v iews

Double JMY: making actin fastDavid W. Roadcap and James E. Bear

The assembly of actin networks is dependent on nucleation-promoting factors. a new study identifies JMY as a protein containing two separate nucleation-promoting activities that shuttles between the nucleus and the cytoplasm and promotes cell migration. These observations indicate that JMY is an important factor controlling actin dynamics in motile cells.

Actin filaments provide the structural basis for much of cell motility and are therefore critical to numerous physiological processes such as morphogenesis, wound healing and immune response. Abnormal cell migration also has a role in disease states such as autoimmune disorders and metastatic cancer. To under-stand these processes better, a comprehensive knowledge of the mechanisms for promoting, inhibiting and regulating actin dynamics is required. The first, rate-limiting, step in form-ing actin filaments is the de novo nucleation of actin filaments from actin monomers. This reaction is strongly kinetically disfavoured by the presence of proteins that sequester actin monomers within cells. Thus, protein cofac-tors that promote actin filament nucleation are required for the generation of actin networks at specific locations within the cell, such as at the distal lamellipodium.

One mechanism by which new filaments are nucleated is by creating branches off the sides of existing filaments by means of the Arp2/3 complex. This highly conserved seven-protein complex has intrinsically low activity and requires activating protein cofactors. The best-studied of these cofactors is the WASP family of nucleation-promoting factors (NPFs), which is regulated by Rho-family GTPases1. Originally there were only two known activators of the Arp2/3 complex (WASP and SCAR), but that number has grown recently to include several other proteins2,3. This increasing complexity of

Arp2/3-complex activators suggests that cells use subcellular and context-specific activation of Arp2/3 for a more robust and precise regula-tion of branched actin networks.

A more recently discovered mechanism for generating new actin filaments is through a protein called Spire4, which promotes filament nucleation by bringing monomers together with four tandem actin-monomer-binding WH2 domains. The four bound actin mono-mers are lined up end-to-end and mimic a short single strand of a nascent filament. Together, these monomers form the pointed end of a new filament to which free mono-mers then bind to grow the nascent filament. Spire-mediated nucleation does not result in branched actin filaments, and the mechanism may thus be used either to jump-start network formation or in circumstances in which a stiff branched network is not necessary.

Using protein homology searching, Zuchero et al.5 have identified the p53 cofactor JMY as possessing a potential Arp2/3 regulatory sequence. JMY is known to bind to p300/CBP and cooperates with it to activate p53-dependent transcription6, but no connection with the actin cytoskeleton had previously been suspected. Zuchero and colleagues purified JMY and demonstrated biochemi-cally that it activates Arp2/3-induced actin polymerization in a dose-dependent fashion. Somewhat surprisingly, they also found that JMY was able to catalyse actin polymerization in the absence of Arp2/3. Further examina-tion revealed that JMY, like Spire, was able to nucleate new filament formation through tan-dem WH2 domains. This is the first instance of these two biochemical activities being united in one protein. By both increasing the speed at

which new filaments are formed and harness-ing the amplification of polymerization that occurs after activation of Arp2/3, JMY seems to be capable of inducing very rapid assembly of new actin networks.

Zuchero and colleagues’ examination of JMY in a cellular context reveals a prima-rily nuclear localization for JMY in most cell types, as would be expected for a p53 regu-lator. In primary human neutrophils, how-ever, JMY co-localizes with actin filaments at the leading edge and is excluded from the nucleus. This localization pattern correlates with motility, because JMY moves from the nucleus to the cytoplasmic compartment when HL60 cells are differentiated from non-motile cells into highly motile neutrophil-like cells. Furthermore, overexpression and knockdown studies demonstrated that JMY expression promotes the rate of cell migration in wound-healing assays. These data are con-sistent with a role for JMY in controlling actin dynamics in highly motile cells (Fig. 1).

As with any newly discovered protein activ-ity, questions quickly outstrip available answers. One interesting question is whether JMY has a role in regulating nuclear actin in addition to its known function as a transcriptional regulator. Although not well understood, nuclear actin has been linked to transcription, chromatin structure and nuclear transport7. The charac-teristics of nuclear actin networks, however, are fundamentally different from those used dur-ing cell motility, so it is unclear how JMY may function in this context, or what consequences any activity would have.

JMY’s relationship with other nucleators also remains unclear. There may be cytoplas-mic competition for actin monomers, and

David W. Roadcap and James E. Bear are at the Lineberger Comprehensive Cancer Center and Department of Cell & Developmental Biology, University of North Carolina-Chapel Hill, Chapel Hill, North Carolina 27599, USA.e-mail: [email protected]




n e w s a n D v i e w s

migration of JMY into that compartment could diminish the supply of actin available for other nucleators such as WASP–Arp2/3 or the formins (another class of actin-nucleating factors). Such competition might complicate the analysis of JMY depletion phenotypes as a result of potential over-activation of other nucleators in the absence of JMY. An alternate possibility is that the nucleators act independ-ently, and introduction of JMY can be used simply as a context-dependent catalyst. In this fashion, JMY may be part of a network of nucleators and NPFs that act in concert to fine-tune cytoskeletal dynamics.

Another unanswered question concerns the mechanism by which JMY activity is regulated.

Given its potent biochemical activity, perhaps it is not surprising that the primary localization of JMY is in the nucleus. Rather than regulating JMY by phosphorylation or other post-transla-tional modification, sequestration of the protein away from most actin could be an ideal way to keep its activity moderated until required. Both the method of transport from the nucleus and its potential triggers remain to be discovered. Similarly, the supply of JMY protein could be regulated by proteasome-mediated degrada-tion. DNA damage causes an accumulation of JMY protein, whereas Mdm2-catalysed ubiquit-ylation targets JMY for proteasome-dependent degradation8. Cell motility-related cues could also tap into this mechanism and contribute to

its availability to alter actin dynamics. It will be interesting to determine whether localiza-tion and protein degradation are indeed used to control JMY activity, and to see whether other mechanisms also contribute.

Another complex issue is reconciling the involvement of JMY in two very different cel-lular processes: transcription and the regula-tion of actin dynamics. It will be important to examine both pathways in the future when examining the function of JMY. Specifically, it will be necessary to test whether any given phenotype is attributable to one activity or a combination of both. Such dichotomy is not unprecedented, because the β-catenin pro-tein is known to have important roles as a cytoskeletal linker mediating cell adhesion, as well as acting as a component of the Wnt signalling pathway that translocates to the nucleus after pathway activation9. Perhaps this system could offer insights into the best path to follow in understanding how JMY balances such discrete functions.

Although there are many unanswered ques-tions about how JMY functions and is regu-lated, it is clear that Zuchero et al. have added another entry to the list of important actin regulatory proteins. Their work has there-fore helped refine our understanding of actin dynamics and cell motility.

1. Chesarone, M. A. & Goode, B. L. Curr. Opin. Cell Biol. 21 (1), 28–37 (2009).

2. Campellone, K. G., Webb, N. J., Znameroski, E. A. & Welch, M. D. Cell 134, 148–161 (2008).

3. Linardopoulou, E. V. et al. PLoS Genet. 3, e237 (2007).

4. Quinlan, M. E., Heuser, J. E., Kerkhoff, E. & Mullins, R. D. Nature 433, 382–388 (2005).

5. Zuchero, J. B., Coutts, A. S., Quinlan, M. E., La Thangue, N. B. & Mullins, R. D. Nature Cell Biol. 11, 451–459 (2009).

6. Shikama, N. et al. Mol. Cell 4, 365–376 (1999).7. Vartiainen, M. K. FEBS Lett. 582, 2033–2040

(2008).8. Coutts, A. S., Boulahbel, H., Graham, A. & La Thangue,

N. B. EMBO Rep. 8, 84–90 (2007).9. Perez-Moreno, M. & Fuchs, E. Dev. Cell 11, 601–612

(2006).

JMY

JMY

WWWCA

JMY

JMY?

Spirelike

nucleation

Arp2/3activation

Newfilament

formation

Actinbranch

formation

Actin network formation

JMY

Nucleus

Cytoplasm

p300

p53

Actin

Figure 1 JMY functions both in and out of the nucleus. JMY functions in concert with p300 to activate p53-dependent transcription. In highly motile cells JMY is transported to the cytoplasm, where it promotes the formation of actin filament networks by means of two separate biochemical activities. It is able to nucleate new filament formation through a Spire-like mechanism that is dependent on its tandem actin-monomer-binding WH2 domains. In addition, JMY promotes actin branch formation by activating the Arp2/3 complex that is dependent on its three tandem actin-monomer-binding WH2 domains (WWW). In addition, JMY promotes actin branch formation by activating the Arp2/3 complex with at least one WH2 domain, a central domain (C) that binds actin and Arp2/3, and an Arp2/3-binding acidic domain (A). It remains undetermined whether JMY regulates nuclear actin dynamics.




skp2: caught in the aktKarin Ecker and Ludger Hengst

To control cell proliferation, signal transduction needs to regulate the cell-cycle machinery. Recent findings show that akt — a major kinase that coordinates diverse signalling pathways — phosphorylates skp2, a subunit of the sCF-skp2 ubiquitin ligase that targets key cell-cycle regulators. akt1-dependent phosphorylation activates sCF-skp2 through multiple mechanisms.

Progression through the cell cycle is orches-trated mainly by ubiquitin-mediated pro-teolysis of key regulatory proteins, such as cyclins and cyclin-dependent kinase (CDK) inhibitors. Ubiquitin ligases (E3) function in the last step of a three-enzyme cascade, which leads to covalent attachment of ubiquitin or polyubiquitin chains to Lys residues of sub-strates. Two types of ubiquitin ligases, SCF complexes and the anaphase-promoting com-plex/cyclosome (APC/C), ubiquitylate many cell-cycle regulators and are essential for cell-cycle progression. SCF-type ubiquitin ligases are composed of three invariable subunits, Skp1, Cul1 and Rbx1, and one of about 70 dif-ferent F-box proteins. These proteins bind to Skp1 through their F-box domain and deter-mine target selection by binding to substrates, often in a manner that depends on substrate phosphorylation1. On pages 420 and 397 of this issue, Lin et al.2 and Gao et al.3, respec-tively, show that the F-box protein Skp2 is itself phosphorylated by Akt. Phosphorylation regulates formation and ubiquitin ligase activ-ity of the SCF-Skp2 complex, Skp2 localization and stability, cell migration, cell proliferation and tumorigenesis. This adds an important direct link in the complex regulatory network between phosphatidylinositol-3-kinase (PI(3)K)/Akt signalling and cell-cycle control.

Among the different substrates of SCF-Skp2, the CDK inhibitor p27Kip1 was identified as a key target, and SCF-Skp2 is a major ubiquitin ligase for CDK/cyclin-bound p27 (refs 1, 4). Skp2 is considered a proto-oncogene as its overexpres-sion causes increased proliferation, at least in part through increased p27 proteolysis. Activation of the PI(3)K/Akt pathway by diverse extracel-lular signals triggers a cascade of responses, including cell growth, proliferation, survival

and motility. PI(3)K is antagonized by several lipid phosphatases, including PTEN. The PI(3)K/Akt pathway is known to regulate the Skp2/p27 axis at multiple levels. For example, Akt con-trols p27 transcription, translation, localization, complex formation and stability through direct and indirect mechanisms4. In addition, PTEN inhibits Skp2 expression5 and Akt induces Skp2 transcription6,7. PTEN also inhibits SCF-Skp2 complex formation indirectly by inhibiting Cul1 association with Skp1 or Skp2 (ref. 8).

Both, Lin et al. and Gao et al. observed that Akt1 binds directly to Skp2. Binding was lost on removal of the 90 amino-terminal amino acids of Skp2 (ref. 3), a region dispensable for the assembly of active SCF-Skp2 ligase complexes9. They found that Akt1, but not related kinases such as Akt2, SGK or S6k3, phosphorylates Skp2 on Ser 72 (refs 2, 3). Ser 72 had recently been identified as one of two major Skp2 phosphorylation sites in vivo and, interestingly, Ser 72 phosphorylation was maximal in M phase10. Not only does Ser 72 phosphorylation induce p27 degrada-tion, indicating that Akt stimulates SCF-Skp2 activity, but phosphorylated Skp2 translocates to the cytoplasm2,3.

Lin et al. found that Skp2 phosphoryla-tion at Ser 72 is essential for its ability to pro-mote cell proliferation and tumorigenesis. A phospho-deficient mutant, Skp2S72A mark-edly impaired Skp2-induced cell proliferation in vitro and tumorigenesis in a mouse model2. Therefore Skp2, phospho-Akt and PTEN lev-els were analysed in human prostate and colon tumour microarrays. Skp2 cytosolic localiza-tion correlates strongly with activated Akt, low PTEN levels and lymph node metastasis in colon cancers. This supports a potential role for Akt signalling and Skp2 cytoplasmic localization in tumour metastasis2.

Lin et al. and Gao et al. report different mechanisms that may activate Skp2. Lin et al. found that Ser 72 phosphorylation promotes SCF-Skp2 assembly and enhances ubiquit-ylation of p27 (ref. 2). In contrast, Gao et al. observed that Ser 72-phosphorylated Skp2 is stabilized through its inability to bind to Cdh1, an activator of the APC/C ubiquitin ligase3 that ubiquitylates Skp2 in G1 (refs 11, 12).

Ser 72 is flanked by two phosphorylation sites, Ser 64 and Ser 75 (ref. 3), all located within a region of Skp2 required for Cdh1-binding11. Within this cluster, Ser 64 is most

Karin Ecker and Ludger Hengst are in the Division of Medical Biochemistry, Biocenter, Innsbruck Medical University, Fritz-Pregl-Str. 3, A-6020 Innsbruck Austria.e-mail: [email protected] Published online 8 March 2009; DOI:10.1038/ncb1859

SPPRKRLKSKGSKGS

p27

S S

Cdk2(Cdk1)

Akt1 CK1

40 90

48 57 64 72 75

Putative NLS

D-box

1 100 150 208 390 424

14-3-3Akt1cyclin A

Cul1Skp1

Cdh1

LRRF-box

Figure 1 Schematic representation of the Skp2 protein. Functional domains (D-box, F-box and the Leu-rich repeats, LRR) are indicated. Positions of phosphorylation sites are shown with their assigned kinases. All of these sites cluster within a potential regulatory domain, which contains a putative nuclear localization sequence (NLS, underlined). This domain is involved in binding of Skp2 to various proteins and, in its unphosphorylated form, may serve as an inhibitory domain preventing SCF complex formation. Ser 48 and 57 have only been identified by mass spectrometry analysis16.





highly conserved phylogenetically and found in all Skp2 orthologues, from vertebrates to insects10. Ser 72 is conserved in Skp2 from most mammals, including humans, but altered to Gly in mice and not conserved in other ver-tebrates such as birds, frogs and fish. Ser 75 seems to be the least conserved residue and is found in some mammals, including primates and rodents, but not in dog, cow or boar or non-mammalian vertebrates.

Skp2 is nearly quantitatively phosphorylated in vivo on Ser 64 (ref. 10) by Cdk2 and possibly Cdk1 (refs 10, 13, 14). Others have previously reported that Skp2 phosphomimetic mutants Skp2S64D and, to a lesser extent Skp2S72D, were stable and proposed that Skp2 phosphoryla-tion, mainly on Ser 64, stabilized the protein by weakening its interaction with Cdh1 (ref. 10). Gao et al. confirmed that phosphorylation of Ser 64 stabilizes Skp2, but found that Ser 64 mutants (phosphomimetic S64D or S64A) still interacted well with Cdh1 (ref. 3). These data suggest that Ser 64 phosphorylation may

stabilize Skp2 by a Cdh1-independent mech-anism or that additional modifications may cooperate with Ser 64 phosphorylation to pre-vent Cdh1 binding. Moreover, Gao et al. found that combined phosphorylation of Ser 72 and Ser 75 inhibited Skp2 binding to Cdh1. They observed that Ser 72 phosphorylation primes Skp2 for Ser 75 phosphorylation and phosphorylation of both sites permits Skp2 to escape APCCdh1-mediated ubiquitylation and subsequent degradation3.

Lin et al. identified a different mechanism for Skp2 activation by Akt. They found that Ser 72 phosphorylation is required for effi-cient complex formation and ubiquitin ligase activity of SCF-Skp2. For example, a phos-pho-deficient Skp2S72A mutant bound poorly to Skp1 and Cul1, and inhibition of PI(3)K by LY294002 prevented SCF-Skp2 complex formation. It is of note, however, that in these experiments PI(3)K inhibition also prevented binding of Skp1 to Cul1 (ref. 2), suggesting a broader effect of PI(3)K inhibition on all

Cul1-containing SCF complexes. A possible explanation may be an earlier observation that inhibition of PI(3)K blocked SCF-Skp2 complex assembly by promoting Cul1 seques-tration in CAND1 complexes, which block Cul1 accessibility to Skp1 and Skp2 (ref. 8). Therefore PI(3)K/Akt seems to affect SCF-Skp assembly through more than one pathway.

At first glance the observation that Ser 72 phosphorylation is required for SCF-Skp2 complex formation is surprising. Previous studies have shown that the initial 100 amino acid region of Skp2 is dispensable for SCF complex assembly and ubiquitin ligase activity9, and that the Skp2 F-box is suf-ficient to form a quaternary complex with Cul1, Rbx1 and Skp1 (ref. 15). One attractive model that can reconcile the findings of Lin et al. with previous work is to propose that the unphosphorylated N-terminal region prevents the interaction of Skp2 with Skp1 and Cul1; Ser 72 phosphorylation induces a structural change that permits SCF assembly.

PKB/AktP

Skp2

Skp2

Skp2

CKI

14-3-3

Skp2

Importin α5; α7

Skp2

Skp2Skp2

InactiveSkp2

SC(F)

SC(F)

Active

UbUb

UbUb

APCCdh1

Skp2Ub

UbUb

Ub

Skp2Ub

UbUb

Ub

Skp2

Skp2

P

p27

PS72

PS72 PS72

PS72

PS72

PS72PS72

PS72

Figure 2 Model for Akt1-dependent regulation of Skp2 activity and localization. Akt1 phosphorylates Skp2 at Ser 72, leading to cytoplasmic retention of the protein by promoting binding to 14-3-3 and inhibiting binding to the nuclear import receptors, importins. Ser 72 phosphorylation primes Skp2 for subsequent phosphorylation at Ser 75 by casein kinase 1 (CK1). Phosphorylation on both sites interferes with the Skp2/APCCdh1 interaction and stabilizes Skp2. Phosphorylation at Ser 72 also permits an enhanced interaction with Skp1 and Cul1, and activates the SCF-Skp2 ubiquitin ligase.




Consistent with this model, a Skp2 mutant lacking the N-terminal 90 amino acids formed a SCF complex more efficiently than full-length Skp2 (ref. 2).

In contrast to the findings of Lin et al. that Ser 72 phosphorylation is required for SCF-Skp2 assembly and activity2, others found recently that simultaneous mutation of amino acids 64 and 72 (S64D/S72D and S64A/S72A) did not affect SCF assembly or its activity10. In these experiments, Skp2 and all other SCF sub-units, as well as the cofactor Cks1, were over-expressed10, whereas Lin et al. only expressed Skp2. Several experimental differences may explain the opposite outcome. For example, overexpression of all SCF subunits may favour SCF-Skp2 complex formation and ligase activity even in the absence of Ser 72 phosphorylation.

Adding to this complexity, Gao et al. and Lin et al. both found that Ser 72 phosphor-ylation also translocates the protein to the cytoplasm. Again, two different mechanisms seem to contribute to this. First, Ser 72 is located within a putative nuclear localiza-tion sequence (NLS) and its phosphorylation impairs Skp2 binding to nuclear import recep-tors3. Second, Lin et al. and Gao et al. found that Ser 72 phosphorylation facilitates Skp2 binding to 14-3-3 proteins2,3. Skp2 cytoplas-mic localization required 14-3-3β2.

Overexpressed Skp2S72D shows only partial cytoplasmic localization2,3, and a Skp2S64D/S72D double mutant was mainly nuclear10; however, endogenous Ser 72-phosphorylated Skp2 was predominantly cytosolic, but not nuclear2.

Akt1-phosphorylated Skp2 is bound in 14-3-3 complexes, which anchors the protein in the cytoplasm2. Overexpression of Skp2 may exceed the 14-3-3 pool, permitting partial nuclear localization. The cytoplasmic locali-zation of Skp2 raises a number of interesting questions. For example, can Skp2/14-3-3 inte-grate into active SCF complexes? If so, does this alter target substrate selection and what are central substrates of cytosolic SCF-Skp2? If nuclear, how can Ser 72-phosphorylated Skp2 escape 14-3-3 and how is the protein imported? Of note, a Skp2–NES (nuclear export signal) fusion protein was predominantly cytosolic but unable to form an active SCF or to ubiq-uitylate p27 ref. 2). As Cdh1 is usually nuclear, is inhibition of Cdh1 binding physiologically significant as long as most Ser 72-phosphor-ylated Skp2 resides in the cytoplasm?

Skp2 overexpression by gene amplification is frequently observed in metastatic tumours5. Lin et al. found that Skp2–/– MEFs showed a profound defect in cell migration, which could be compensated by Skp2S72D but not Skp2S72A (ref. 2), and a predominantly cytosolic Skp2–NES fusion protein rescued migration of null MEFs. These findings suggest that cytoplasmic Skp2 has a potential function in metastasis. Although p27 has a well-established role in cell migration4, regulation of cell motility by cytoplasmic Skp2 seems to be independent of its ability to ubiquitylate p27, as Skp2–NES fails to form a ubiquitin ligase2. Further studies should elucidate mechanisms by which cyto-plasmic Skp2 affects cell motility.

Taken together, these studies provide com-pelling evidence that Skp2 phosphorylation on Ser 72 has a central role in tumorigenesis. Skp2 phosphorylation seems to affect Skp2 localiza-tion and activity by several complementary mechanisms. The cluster of three phosphoryla-tion sites of different phylogenetic conservation located within a region of Skp2 required for Cdh1 binding and adjacent to the F-box sug-gests possible redundant functions that could explain the variable molecular consequences observed in response to phosphorylation. It is interesting that although mouse Skp2 lacks Ser 72, most molecular consequences of Akt phosphorylation are also observed in mice2 sug-gesting that the Akt–Skp2 axis is functionally conserved but may use distinct mechanisms.

1. Frescas, D. & Pagano, M. Nature Rev. Cancer 8, 438–449 (2008).

2. Lin, et al. Nature Cell Biol. 11, 420–432 (2009).3. Gao, et al. Nature Cell Biol. 11, 397–408 (2009).4. Chu, I. M., Hengst, L. & Slingerland, J. M. Nature Rev.

Cancer 8, 253–267 (2008).5. Hershko, D. D. Cancer 112, 1415–1424 (2008).6. Reichert, M., Saur, D., Hamacher, R., Schmid, R. M. &

Schneider G. Cancer Res. 67, 4149–4156 (2007).7. Barré, B. & Perkins, N. D. EMBO J. 26, 4841–4855

(2007).8. Jonason, J. H., Gavrilova, N., Wu, M., Zhang, H. & Sun,

H. Cell Cycle 6, 951–961 (2007).9. Schulman, B. A. et al. Nature 408, 381–386 (2000).10. Rodier, G., Coulombe, P., Tanguay, P. L., Boutonnet, C.

& Meloche, S. EMBO J. 27, 679–691 (2008).11. Bashir, T., Dorrello, N. V., Amador, V., Guardavaccaro,

D. & Pagano, M. Nature 428, 190–193 (2004).12. Wei, W. et al. Nature 428, 194–198 (2004).13. Zhang, H., Kobayashi, R., Galaktionov, K., Beach, D.

Cell 82, 915–925 (1995).14. Yam, C. H., Ng, R. W., Siu, W. Y., Lau, A. W. & Poon,

R. Y. Mol. Cell Biol. 19, 635–645 (1999).15. Zheng, et al. Nature 416, 703–709 (2002).16. Dephoure, N. et al. Proc. Natl Acad. Sci. USA 105,

10762–10767 (2008).

Targeting protein ubiquitylation: DDB1 takes its RinG offSarah Jackson and Yue Xiong

Ubiquitin e3 ligases of the RinG and HeCT families are distinct not only in their catalytic mechanisms but also in targeting substrates. now it seems that one heterodimeric complex can target substrates to both types of e3 ligase.

Protein ubiquitylation has a broad and criti-cal role in regulating a wide range of cellu-lar processes. The addition of Lys 48-linked polyubiquitin chains to specific substrate

proteins regulates timely degradation by the 26S proteasome. In addition, like other covalent modifications, ubiquitylation can modulate the function of a substrate by caus-ing a conformational change. Ubiquitylation begins with the ATP-dependent activation of ubiquitin by the E1 enzyme, and is followed by the subsequent transfer of ubiquitin to one of a small family of E2 ubiquitin-conjugating

enzymes; finally, an E3 ubiquitin ligase is responsible for recognizing a specific substrate and promoting ubiquitin ligation. More than 1,000 distinct E3 ligases are predicted to exist, either as individual proteins or multi-subunit complexes, in mammalian cells.

There are two major families of E3 ligases distinguished by their active domains: the HECT family (‘homologous to the E6-AP

Sarah Jackson and Yue Xiong are in the Department of Biochemistry and Biophysics, Lineberger Comprehensive Cancer Center, University of North Carolina at Chapel Hill, North Carolina 27599, USA.e-mail: [email protected]

nature cell biology volume 11 | number 4 | APrIl 2009 379 nature cell biology volume 11 | number 4 | APrIl 2009 379




carboxy terminus’) and the RING family (first recognized in the human ‘really interesting new gene product’)1,2. The HECT domain mediates interaction with the cognate E2 and, through an evolutionarily conserved cysteine residue, forms a thioester linkage with ubiquitin. Human cells contain as many as 28 HECT proteins and most, if not all, are believed to function as E3 ligases. Unlike the HECT domain, the RING domain promotes a direct transfer of ubiquitin from the E2 to the substrate without forming an intermediate with ubiquitin. Human cells express more than 450 RING proteins, and E3 ligase activity has been experimentally demonstrated for many of them. In addition, although not containing a RING domain themselves, members of the evolutionarily conserved cullin family can bind a small RING protein, either ROC1 or ROC2 (also known as Rbx). A remarkable feature of cullin proteins is that the amino-terminal sequence in each of the six classical human cul-lin family members interacts selectively with a different motif such as an F-box, a SOCS box, a BTB domain and a WD40 repeat. These common motifs are present in many proteins, suggesting the potential assembly of as many as 300–500 distinct cullin–RING ligase (CRL) complexes in vivo3, making cullins the largest subfamily of E3 ligases.

Not only do HECT and CRL E3 ligases use different catalytic mechanisms in catalysing the transfer of ubiquitin from E2 to the substrate, they are also thought to have unique means of assembly, regulation and substrate targeting. On page 409 of this issue, Maddika and Chen4 identify and characterize a novel E3 ligase that uses DYRK2 as a scaffold for the assembly

of a HECT E3 complex and a heterodimeric complex consisting of DDB1 and VPRBP for recruiting substrate. This finding is particularly unexpected because DYRK2 is a protein kinase and DDB1 is established as a key adaptor pro-tein for recruiting substrate to the Cul4–RING ligases (CRL4s)5–8.

DYRK2 is a member of evolutionarily con-served dual-specificity tyrosine (Y)-regulated kinases, whose function has been broadly linked to DNA repair, cell proliferation, differ-entiation and apoptosis. Maddika and Chen identify a novel DYRK2 complex that contains EDD, DDB1 and VPRBP. EDD (E3 identified by differential display) is a large protein contain-ing multiple domains linked to ubiquitylation, including an N-terminal ubiquitin associated (UBA) domain, a UBR box (a motif important for the targeting of N-end rule substrates) and a C-terminal HECT domain. No known sub-strate has previously been identified for EDD. DDB1 (damaged-DNA-binding protein) serves as a key linker to bridge a subset of WD40-containing proteins to Cul4–RING ligases5–8. As many as one-third of the 300 WD40 proteins found in human cells could interact with DDB1 (ref.s 5). VPRBP, a WD40-containing protein that binds DDB1, was initially identified as the human HIV Vpr-binding protein. The sig-nificance of VPRBP–Vpr interaction remains unclear, especially whether Vpr, like E6, hijacks a VPRBP complex or exploits normal substrate ubiquitylation to benefit HIV propagation. So far, only one candidate substrate, the cytoplas-mic localized neurofibromatosis type 2 (NF2) tumor suppressor gene product, Merlin, has been reported to be targeted by VPRBP to the DDB1–Cul4–ROC1 ligase for degradation9.

However, there are reasons to believe that VPRBP may target additional proteins, because VPRBP can bind to chromatin and is required for normal DNA replication, and genetic dis-ruption of VPRBP causes early embryonic lethality in mouse and various developmental defects in plants10,11.

In Caenorhabditis elegans, the DYRK2 homolog MBK-2 phosphorylates and regu-lates the meiotic protein, MEI-1/katanin, the catalytic subunit of the microtubule-sev-ering AAA ATPase complex. Maddika and Chen4 therefore tested whether mammalian katanin was a substrate for the newly identi-fied DYRK2 E3 complex, referred to as EDVP (EDD–DDB1–VPRBP). In vitro binding and in vivo ubiquitylation assays demonstrated that katanin associates with and is polyubiq-uitylated by the EDVP E3 ligase complex. VPRBP binds directly to, and is required for, bringing katanin to the EDVP E3 ligase; notably, no Cul4 or ROC1 is detected in the complex. Silencing individual components of EDVP, but not Cul4A and Cul4B, severely impaired katanin polyubiquitylation. Maddika and Chen show that DYRK2 acts as a scaffold to assemble the complex components, but this scaffold function does not rely on its kinase activity. However, phosphorylation by DYRK2 is required for subsequent katanin polyubiq-uitylation: coexpression of either a catalytically inactive DYRK2 or a triple phospho-mutant of katanin inhibits katanin polyubiquitylation. Supporting the physiological relevance of this ubiquitylation, ectopic expression of katanin causes mitotic defects (as determined by the increase in cells with 4N DNA content and positive for phopho-histone H3) that can be largely alleviated by co-expression with wild-type, but not kinase-dead, DYRK2. Knocking down either DYRK2 or EDD causes katanin accumulation and a similar increase in G2/M cells, which can be rescued by simultaneous silencing of katanin. Hence, the EDVP E3 com-plex is capable of phosphorylation and subse-quent ubiquitylation of its substrate.

This study raised two interesting ques-tions whose resolution may shed new light on mechanisms of ubiquitylation and substrate targeting. First, how does DYRK2-mediated phosphorylation of substrate katanin con-tribute to subsequent ubiquitylation by EDD? Substrate phosphorylation is known to have a key function in the initial recognition by some E3s, as best documented for several substrates whose phosphorylation triggers the binding

VprBP

ROC

substrate

DYRK2 Cul4A/B

VprBP

VPRBP

P PP

HE

CT

EDD

E 2

Katanin Merlin

Katanin

DDB1

DDB1

VPRBP VPRBP

DDB1

Merlin

E2E2

Ub Ub

UbUb

Ub Ub

Ub

UbUb

UbUb

Figure 1 DDB1–VPRBP targets substrates to distinct E3 ubiquitin ligase complexes. The DDB1–VPRBP heterodimer can target different substrate to a DYRK2–HECT or a Cul4–ROC1 E3 ligase complex. DYRK2 is required for assembly of the E3 complex and for phosphorylation of its substrate katanin, but not for the initial binding of katanin with VPRBP. Ub, ubiquitin.




with specific F-box proteins and subsequent ubiquitylation by the SCF/CRL1 complex. Unlike phosphorylation-dependent binding between substrate and the F-box, there is no evidence that DYRK2-mediated phosphoryla-tion is required for katanin to bind with VprBP-DDB1. However, the phospho-mutant katanin cannot be efficiently ubiquitinated. Similarly, the catalytic mutant DYRK2 does not seem to have any defect in assembling EDD, DDB1 and VPRBP but fails to promote katanin polyubiq-uitylation. Could phosphorylation have a func-tion in orienting the substrate towards or closer to the ubiquitin-linked catalytic Cys in the HECT domain of EDD? It has been deduced from structural analysis of several E3s that the distance between the active Cys residue either in the E2 bound to the RING finger or in the HECT domain is too far away for transfer of ubiquitin to the substrate. For example, the Cys in the active site of E2is 41 Å away from the active site in the HECT domain in E6AP, and 50 Å away from the nearest amino-acid F-box protein in the SCF/CRL1 complex12,13.

Second, how does the DDB1–VPRBP het-erodimer determine which substrate is targeted to which E3? Among the estimated 90-plus DWD (DDB1-binding WD40) proteins, VPRBP is unique in that it is a particularly large protein that is abundantly expressed in many cell types and, like DDB1, has an essential func-tion for cell growth and embryo development. Do these properties make the DDB1–VPRBP heterodimer a unique complex in recruiting different substrates to different E3 ligases? Are there other DWD proteins, in addition to VPRBP, that are also capable of shuttling between both families of E3 ligases? DYRK2 was not detected by several previous proteomic screens of proteins associated with DDB1 and VPRBP, suggesting that we may still be under-estimating the reach of adaptor proteins and substrate receptor complexes in targeting substrate proteins for ubiquitylation. We have already seen that individual F-box proteins can target multiple substrates to specific CRLs. For example, the SKP2 and β-TrCP F-box proteins have each been linked to the ubiquitylation of

nearly 30 proteins14. These current findings demonstrate even more versatility in targeting substrates for ubiquitylation than previously realized, and indicate the potential to expand the repertoire of specific protein substrates ubiquitylated by E3 ligases.

1. Huibregtse, J. M., Scheffner, M., Beaudenon, S. & Howley, P. M. Proc. Natl Acad. Sci. USA 92, 2563–2567 (1995).

2. Lovering, R. et al. Proc. Natl Acad. Sci. USA 90, 21112–22116 (1993).

3. Petroski, M. D. & Deshaies, R. J. Nature Rev. Mol. Cell Biol. 6, 9–20 (2005).

4. Maddika, S. & Chen, J. Nature Cell Biol. 11, 409–419 (2009).

5. He, Y. J., McCall, C. M., Hu, J., Zeng, Y. & Xiong, Y. Genes Dev. 20, 2949–2954 (2006).

6. Higa, L. A. et al. Nature Cell Biol. 8, 1277–1283 (2006).

7. Angers, S. et al. Nature 443, 590–593 (2006).8. Jin, J., Arias, E. E., Chen, J., Harper, J. W. & Walter,

J. C. Mol. Cell 23, 709–721 (2006).9. Huang, J. & Chen, J. Oncogene 27, 4056–4064

(2008).10. McCall, C. M. et al. Mol. Cell. Biol. 28, 5621–5633

(2008).11. Zhang, Y. et al. Plant Cell 20, 1437–1455 (2008).12. Huang, L. et al. Science 286, 1321–1326 (1999).13. Zheng, N. et al. Nature 416, 703–709. (2002).14. Frescas, D. & Pagano, M. Nature Rev. Cancer 8, 438–

449 (2008).

sOC: now also store-operated cyclaseJames W. Putney Jr

depletion of Ca2+ from intracellular stores has long been known to signal to and activate plasma membrane ‘store-operated’ channels. we now learn that store depletion also controls the formation of cyclic aMP (caMP) through the regulation of adenylyl cyclase (a-Cyclase). These findings substantially broaden the scope and biological significance of Ca2+ store-regulated signalling.

The generation of intracellular Ca2+ signals by hormones, neurotransmitters and other extra-cellular ligands represents a major mechanism for the regulation of rapid to long-term cellular responses. Typically, these Ca2+ signals com-prise a combination of intracellular discharge of Ca2+ from stores and influx of Ca2+ across the plasma membrane. Intracellular messengers, most typically inositol trisphosphate (InsP3), are responsible for intracellular Ca2+ release. Although there are several mechanisms under-lying the activation of plasma membrane Ca2+ channels, the most common involves signalling

from the depleted endoplasmic reticulum (ER) to the channels, a process long referred to as ‘capacitative’ or ‘store-operated’ Ca2+ entry1. On page 433 of this issue, Lefkimmiatis et al.2 pro-vide convincing evidence that the same store-operated pathway can also signal to and activate A-Cyclase, thus resulting in the formation of the second messenger cAMP.

The concept of store-operated Ca2+ entry is now over 20 years old1. However, it is only in the past few years that modern high-through-put genetic screening techniques have identi-fied two of the key molecular players in this pathway. Signalling from the ER to the plasma membrane is initiated by the Ca2+ sensor pro-teins STIM1 and STIM2. These proteins are single-pass membrane proteins, with Ca2+-binding EF-hand motifs directed to the lumen of the endoplasmic reticulum. Dissociation

of Ca2+ causes the proteins to aggregate and accumulate in regions just beneath the plasma membrane3. There, they communicate with proteins of the Orai (also known as CRACM) family (Orai1–3; refs 3, 5), resulting in chan-nel activation and the appearance of the highly Ca2+-selective current Icrac (calcium-release-activated calcium current) 1.

The original idea of store-operated calcium entry came from studies of the mechanism by which intracellular stores were replaced fol-lowing their release1. Initially, it was unclear whether this mode of entry represented a true signalling function, or a housekeeping role ensuring adequate ER Ca2+ levels for proper protein synthesis and folding6. The discovery of the signalling proteins STIM1 and STIM2 clearly indicates that STIM1-activated entry functions primarily as a signalling pathway7,8,

James W. Putney Jr is in the Laboratory of Signal Transduction, National Institute of Environmental Health Sciences–NIH, Department of Health and Human Services, PO Box 12233, Research Triangle Park, NC 2770, USA.e-mail: [email protected]





whereas STIM2 may have the more fundamen-tal role of Ca2+ store maintenance9.

The study by Lefkimmiatis et al.2 shows that Ca2+ store depletion can also activate A-Cyclase through the Ca2+ sensor STIM1, further supporting the view that store-operated signalling represents a general mechanism for coordinating intracellular and plasma membrane signalling events. But what is the physiological function of such a link? The dynamic interplay between Ca2+ and cAMP signalling has long been appreciated10. Activation of receptors linked to Ca2+ mobili-zation regulate A-Cyclases, and cAMP can in turn regulate Ca2+ signalling10. Indeed, inter-actions between these two major pathways can result in complex signalling patterns and may contribute to the wide variety of subtly distinct cellular response profiles11. Thus, it should not be surprising that one of the most general Ca2+ signalling mechanisms, the store-dependent activation of Ca2+ channels, also crosstalks directly with the cAMP pathway under certain conditions. It is noteworthy that an indirect link between store depletion and A-Cyclase has previously been established, as

Ca2+ entering the cell through store-operated Ca2+ channels seems to be specifically cou-pled to both positive and negative regulation of A-Cyclases12.

The mechanism by which STIM1 regulates A-Cyclase is unknown. Indeed, the mecha-nism by which STIM1 regulates Orai Ca2+ channels is also not entirely clear. There is evidence for direct protein–protein interac-tion between STIM1 and Orai, but there is also evidence for the involvement of other, as yet unidentified, proteins13. An unidentified ‘calcium influx factor’ has been proposed to mediate the actions of STIM1 on Orai chan-nels14: could such a factor also be involved in regulation of A-Cyclase? Other questions remain to be answered. For example, the more ‘conventional’ mechanism for activa-tion of A-Cyclases involves heterotrimeric G-proteins: are G-proteins required for acti-vation by store depletion? The synergism between store depletion and G-protein acti-vating agonists might suggest so. Regulation of A-Cyclases by G-proteins, or by Ca2+, can result in inhibition or activation, depend-ing on the nature of the G-protein, or on the

specific isoform of A-cyclase, respectively; therefore, another question is whether there will be instances where Ca2+ store depletion could result in A-Cyclase inhibition.

Significantly, the discovery of a ‘non-cal-cium’ signal regulated by Ca2+ store depletion and STIM1 raises the question of how many other, if any, signalling pathways might be sim-ilarly regulated. The idea of Ca2+ store deple-tion as a more general signal has been raised previously, on the basis of Ca2+ depletion from the ER inducing apoptosis independently of Ca2+ entry through store-operated chan-nels15. As pointed out by Lefkimmiatis et al.2, sustained loss of ER Ca2+ results in impaired protein synthesis and protein folding, ulti-mately culminating in a stereotypic stress response16. The ER stress response probably requires more extensive Ca2+ depletion than that which initiates STIM1 dependent signal-ling, but multiple signalling mechanisms may be involved. Hopefully continued research in this arena, perhaps involving newly described STIM1 gene knockout models17,18, will provide insights into these important questions.

1. Parekh, A. B. & Putney, J. W. Physiol. Rev. 85, 757–810 (2005).

2. Lefkimmiatis, K. et al. Store-operated cyclic AMP signaling mediated by STIM1. Nature Cell Biol. 11, 433–442 (2009).

3. Cahalan, M. D. et al. Cell Calcium 42, 133–144 (2007).

4. Gwack, Y., Feske, S., Srikanth, S., Hogan, P. G. & Rao, A. Cell Calcium 42, 145–156 (2007).

5. Vig, M. & Kinet, J. P. Cell Calcium 42, 157–162 (2007).

6. Verkhratsky, A. & Petersen, O. H. Eur. J. Pharmacol. 447, 141–154 (2002).

7. Putney, J. W. & Bird, G. S. J. Physiol. 586, 3055–3059 (2008).

8. Parekh, A. B. Cell Calcium 42, 111–121 (2007).9. Brandman, O., Liou, J., Park, W. S. & Meyer, T. Cell

131, 1327–1339 (2007).10. Rasmussen, H. Science 170, 404–412 (1970).11. Zaccolo, M. & Pozzan, T. Trends Neurosci. 26, 53–55

(2003).12. Martin, A. C. & Cooper, D. M. Biochem. Soc. Trans. 34,

480–483 (2006).13. Varnai, P., Toth, B., Toth, D. J., Hunyady, L. & Balla, T.

J. Biol. Chem. 282, 29678–29690 (2007).14. Csutora, P. et al. J. Biol. Chem. 283, 14524–14531

(2008).15. Bian, X., Hughes, F. M. Jr, Huang, Y., Cidlowski, J. A.

& Putney, J. W. Am. J. Physiol. 272, C1241–C1249 (1997).

16. Berridge, M. J. Cell Calcium 32, 235–249 (2002).17. Baba, Y. et al. Nature Immunol. 9, 81–88 (2008).18. Baba, Y. et al. Proc. Natl Acad. Sci. USA 103, 16704–

16709 (2006).

R1 R2

GsGq

GsPLC

Ag1 Ag2

InsP3

Ca2+

Ca2+

InsP3R

Orai

STI

M1

STI

M1

Plasmamembrane

ER

A-Cyclase

ATPcAMP Otherpathways?

?

Figure 1 Calcium store-operated cell signalling. Agonist ligands (for example, Ag1 or Ag2) activate their receptors (R1 or R2), which are coupled to either G protein Gq/ phospholipase C (PLC) to generate Ca2+ signals, or to G protein Gs to activate adenylyl cyclase (A-Cyclase). Phospholipase C and the subsequent formation of InsP3 cause discharge of intracellular Ca2+ from endoplasmic reticulum (ER) stores through the InsP3 receptor (InsP3R). The fall in ER Ca2+ results in Ca2+ dissociation from the amino terminus of the Ca2+ sensor STIM1, which in turn signals to store-operated plasma membrane Ca2+ channels (Orai). The study by Lefkimmiatis et al.2 shows that STIM1 can also activate A-Cyclase, either on its own or in synergy with other A-cyclase activators, raising the question as to whether other signalling pathways are similarly regulated.




research h ighl ightsBud14 — actin’ in formin displacementThe formin family of actin regulators catalyse actin polymerization and associate with grow-ing filament ends through their FH2 domains, protecting them from capping proteins which block actin subunit binding. Goode and col-leagues (Dev. Cell 16, 292–302; 2009) now identify Bud14 as a new formin-displacement factor in yeast. Bud14 was found in a screen for regulators of mother-cell actin cables and its depletion leads to formation of abnormally long and bent actin filaments. Genetically Bud14 was shown to function upstream of the formin Bnr1 and, although Bud14 has no affect on in vitro actin assembly on its own, it inhibits the activity of Bnr1. Bud14 binds the Bnr1 FH2 domain directly and prevents actin assembly in the presence of capping proteins, suggesting that it displaces the FH2 domain of Bnr1 from actin. In agreement with a role for Bud14 in the control of actin architecture, actin-dependent secretory vesicle transport is impaired in Bud14 depleted cells. A previously demonstrated role of Bud14 in dynein-dependent microtubule sliding along the cell cortex is shown here to be separable from its role in actin organization. How the two functions of Bud14 are coordi-nated are topics for the future, as is the question of whether a similar class of formin inhibitors exists in mammals. CKR

Nucleosome organization drives gene expression divergence

Low nucleosome occupancy of gene pro-moters correlates with higher gene expres-sion. Evolutionary changes in yeast species have now been linked to variations in nucle-osome occupancy by Segal and colleagues (Nature Genet., doi: 10.1038/ng.324; 2009). The authors compared the transcription program and nucleosome distribution of the aerobic human pathogen Candida albicans with those of the anaerobic Saccharomyces cerevisiae, in which expression of respira-tory genes is low under typical growth con-ditions. Using the large datasets available for both species and a computational approach to assess nucleosome occupancy at the pro-moters of protein-coding genes, they classify genes according to their expression relative to that of cytoplasmic ribosomal protein (CRP)-coding genes, which usually corre-lates with cellular growth. In both yeasts, expression of genes required for basal cel-lular growth correlated highly with that of CRPs and with low nucleosome occupancy. Conversely, genes involved in response to specific environmental conditions did not correlate with CRPs and were predicted to show high nucleosome occupancy.

Interestingly, genes required for respira-tion correlated with CRPs and low nucleo-some occupancy in the aerobic C. albicans but not in the anaerobic S. cerevisiae, sug-gesting that nucleosome occupancy was linked to diversity in terms of metabolism. These predictions were verified by mapping nucleosome positions in vivo and by recon-stituting nucleosomes on naked DNA from both species in vitro. The same correlations between gene expression and nucleosome occupancy were predicted in 12 additional yeast species, suggesting that phenotypic diversity is linked to nucleosome organiza-tion in promoters through changes in DNA sequence. NLB

Re-growing out of the niche

In plants, apical stem-cell niches sustain the indeterminate growth of roots and shoots. A study by Birnbaum and colleagues now reveals that plants are able to regenerate these organs in the absence of a functional stem-cell niche (Nature 457, 1150–1153; 2009). Following root-tip excision and removal of the niche quiescent centre, regeneration was analysed in Arabidopsis thaliana by tracking the re-establishment of the different cell types through time-lapse high resolution imaging of cell-identity markers and concomitant analysis of cell-type-specific transcriptional profiles. This demonstrated that, at regen-eration sites, cell identities were re-specified within hours of excision, and that fully func-tional specialized cells were restored before recovery of the stem-cell niche. Moreover, regeneration and functional specification of roots still occurred in plants with muta-tions that cause root growth defects due to impaired stem-cell niche maintenance. Further marker analysis indicated that the competence to regenerate might be a feature of differentiating cells sharing a common set of stem-cell-like properties. These proper-ties are therefore not restricted to niches, but rather widely dispersed in plant meristematic tissues, a characteristic that could explain the high regenerative capacity of plants. SG

Written by Nathalie Le Bot, Silvia Grisendi, Bernd Pulverer and Christina Karlsson Rosenthal

A complex DNA damage response complex The tumour suppressor and breast cancer susceptibility gene BRCA1 is engaged in several multiprotein complexes, and has key roles in the DNA damage response by regulating DNA repair, transcription and ubiquitylation. A complex containing Abra1/Abraxas/CCDC98, RAP80, BRCC36 and BRE/BRCC45 is implicated in the recruitment of BRCA1 to DNA double-stranded breaks through a damage signalling pathway involving the kinase ATM, the histone variant γ-H2AX, Mdc1, the ubiquitin ligase RNF8 and the conjugating enzyme Ubc13. Three groups independently identified a new component of this stable complex called MERIT40 or NBA1 through a shRNA screen (Wang et al.; Genes Dev., doi: 10.1101/glad.1739609; 2009) or affinity purification schemes (Feng et al.; Genes Dev., doi: 10.1101/glad.1770609; 2009 and Shao et al.; Genes Dev., doi: 10.1101/glad.1770309; 2009). All three papers show MERIT40/NBA1 regulates localization of complex components as well as BRCA1 to DNA breaks. The new component of the BRCA1 complex mediates resistance to ionizing radiation and it is essential for the G2/M DNA damage checkpoint. MERIT40/NBA1 is recruited by directly interacting with BRE. Indeed, MERIT40 and BRE are required to maintain stability of the complex and Abra1 seems to serve as a central organizing adaptor. The complex appears to interact with a spectrum of ubiquitin chains through four different ubiquitin-binding domains. Shao et al. also show MERIT40 is required for the known Lys 63 de-ubiquitylation activity of BRCC36, which is implicated in both the checkpoint and resistance to ionizing radiation. Interestingly, Wang et al. point out that a structural model of the complex resembles the 19S lid of the 26S proteasome. BP


ART ICLES

Two Beclin 1-binding proteins, Atg14L and Rubicon, reciprocally regulate autophagy at different stagesKohichi Matsunaga1,2, Tatsuya Saitoh3,4, Keisuke Tabata1, Hiroko Omori1, Takashi Satoh3,4, Naoki Kurotori1, Ikuko Maejima1, Kanae Shirahama-Noda1, Tohru Ichimura5, Toshiaki Isobe5, Shizuo Akira3,4, Takeshi Noda1 and Tamotsu Yoshimori1,6,7

Beclin 1, a protein essential for autophagy, binds to hVps34/Class III phosphatidylinositol-3-kinase and UVRAG. Here, we have identified two Beclin 1 associated proteins, Atg14L and Rubicon. Atg14L and UVRAG bind to Beclin 1 in a mutually exclusive manner, whereas Rubicon binds only to a subpopulation of UVRAG complexes; thus, three different Beclin 1 complexes exist. GFP–Atg14L localized to the isolation membrane and autophagosome, as well as to the ER and unknown puncta. Knockout of Atg14L in mouse ES cells caused a defect in autophagosome formation. GFP–Rubicon was localized at the endosome/lysosome. Knockdown of Rubicon caused enhancement of autophagy, especially at the maturation step, as well as enhancement of endocytic trafficking. These data suggest that the Beclin 1–hVps34 complex functions in two different steps of autophagy by altering the subunit composition.

Macroautophagy, hereafter referred to as autophagy, is an intracellular process in which cytoplasmic materials are transported by autophago-somes to lysosomes for degradation1–4. Autophagy contributes to sur-vival during starvation, cytoplasmic renewal, elimination of intracellular aggregate-prone proteins and pathogens, innate and acquired immunity, and context-dependent programmed cell death5–9.

Beclin 1 is a coiled-coil protein involved in the regulation of autophagy in mammalian cells10–13. Beclin 1 binds to hVps34/class III phosphatidyli-nositol-3-kinase (PI(3)K) through its evolutionarily conserved domain (ECD)14,15. The hVps34/PI(3)K generates phosphatidylinositol-3-phosphate (PI(3)P), which has important roles in several membrane trafficking path-ways, including the multivesicular body pathway, retrograde trafficking from endosomes to the Golgi and phagosome maturation16,17. PI(3)P is also involved in the regulation of autophagy: wortmannin, a potent inhibitor of PI(3)K, efficiently inhibits autophagy18,19. Beclin 1 diverts a subpopulation of hVps34 that is devoted to autophagy, but the mechanisms underlying this specificity remain unclear. In recent years, a number of proteins associated with Beclin 1 have been identified. UVRAG acts as an autophagy-associated protein by binding to Beclin 1 (ref. 20). UVRAG functions in the autophagy maturation process through its binding to Class C VPS, which is involved in the fusion of autophagosomes and lysosomes21. UVRAG also binds to another potential autophagy-associated protein, Bif-1 (ref. 22). Two other proteins that bind to Beclin 1, Ambra-1 and VMP1 may have positive roles in autophagosome formation23,24. In addition, Beclin 1 facilitates autophagic

cell death by regulated binding to the prototypic apoptosis inhibitor Bcl-2 (refs 5, 25–27). Beclin 1 and its associated proteins may also have anti-tumour activity, potentially by modulating autophagic cell death10,13,28. Indeed, Beclin 1+/– mice develop spontaneous tumours, suggesting that Beclin 1 is a haploinsufficient tumour supressor13,28.

In Saccharomyces cerevisiae, Atg6/Vps30, the orthologue of Beclin 1, conducts two distinctive cellular processes: autophagy and endosome-to-Golgi retrograde trafficking29,30. Each role is catalysed by a distinct protein complex: autophagy involves complex I, which consists of Atg14, Atg6/Vps30, Vps15 and Vps34, whereas retrograde trafficking involves complex II, which consists of Vps38, Atg6/Vps30, Vps15 and Vps34 (refs 29, 31, 32). Thus, all the subunits except for Atg14 and Vps38 are shared between the two complexes. However, counterparts for Atg14 and Vps38 have not been described in other organisms, leading to the assumption that this molecular architecture may apply only to yeast.

Here, we have used a highly sensitive method and mild purification con-ditions to identify five Beclin 1-interacting proteins, two of which have not been described previously. Analysis of these new proteins identified three distinct Beclin 1 complexes, which regulate autophagy at different steps.

RESULTSIdentification of Beclin 1 binding proteinsTo comprehensively identify Beclin 1 interacting protein(s), we used the tan-dem affinity purification approach based on the MEF tag (Myc–TEV–Flag),

1Department of Cellular Regulation, Research Institute for Microbial Diseases, Osaka University, 3‑1Yamadaoka, Suita, Osaka 565‑0871, Japan. 2Department of Genetics, The Graduate University for Advanced Studies, Mishima 455‑8540, Japan. 3Laboratory of Host Defense, WPI Immunology Frontier Research Center, Osaka University, 3‑1 Yamadaoka, Suita, Osaka 565‑0871, Japan. 4Department of Host Defense, Research Institute for Microbial Diseases. Osaka University, 3‑1Yamadaoka, Suita, Osaka 565‑0871, Japan. 5Department of Chemistry, Graduate School of Science, Tokyo Metropolitan University, Hachioji, Tokyo 192‑0397, Japan. 6CREST, Japan Science and Technology Agency, Kawaguchi‑Saitama 332‑0012, Japan.7Correspondence should be addressed to T.Y. (e‑mail: [email protected]‑u.ac.jp)

Received 3 September 2008; accepted 23 December 2008; published online 8 March 2009; DOI: 10.1038/ncb1846




A RT I C L E S

which contains Myc and Flag tandem epitope tags connected by a spacer sequence, including a TEV protease cleavage site33. We stably expressed human Beclin 1 fused to the MEF tag at its amino terminus in MCF7 cells

and recovered the protein in successive purification steps using an anti-Myc antibody, TEV protease, anti-Flag antibody and elution with a synthetic Flag peptide. By comparison with the control, we found seven bands specific for

Figure 1 Identification and analysis of Beclin 1‑binding proteins. (a) MCF7 cells stably expressing Myc–TEV–Flag–Beclin 1 (MEF–Beclin 1) and parental controls were lysed and subjected to MEF‑tag‑based purification. Proteins bound to MEF–Beclin 1 were isolated and detected by SDS–PAGE and silver staining. Seven bands specific for MEF–Beclin 1 are numbered. (b) Six bands were identified by nanoflow LC‑MS/MS and MASCOT software. (c) MCF7 cells were lysed and subjected to immunoprecipitation using the indicated antibodies. (d) Differential subcellular fractionation. Post‑nuclear supernatant of A549 cell lysates were spun at 20,000g for 10 min and the supernatant was further spun at 100,000g for 60 min. Equivalent quantities of each supernatant and

pellets from the same number of cells were subjected to western blotting with the indicated antibodies, including anti‑Sec61β (ER), cathepsin D (lysosome), EEA1 (early endosome), transferrin receptor (plasma membrane and recycling endosome) and GAPDH (cytosol). (e) Gel filtration analysis. The supernatant fraction of A549 cell (or Atg14L knockdown) lysates were subjected to a Superose 6 column and each fraction was immunoblotted with the indicated antibody (see Supplementary Information, Fig. S2e). Relative amounts of each fraction determined by densitometry were plotted. Vo, void fraction. The elution pattern of each protein was reproduced in several experiments. Full scans of the gel and blots in a, c and d are available in Supplementary Information, Fig. S6.

75

100

150

100

150

100

100

25

15

37

a

75

100

150

250

25

37

50

20

Control MEF–Beclin-1

No. 1No. 2No. 3No. 4

No. 5No. 6

No. 7

No. 1 hVps15 (p150)No. 2 Rubicon (KIAA0226)No. 3 Class III PI 3-kinase (hVps34)No. 4 UVRAGNo. 5 Atg14L (KIAA0831)No. 6 Beclin-1 (Flag–Beclin-1)No. 7 Not determined

Pre

-im

mun

e

Atg

14L

UV

RA

G

hVps34

Beclin-1

Immunoprecipitation

Rub

icon

Bec

lin-1

hVp

s34

Rubicon

UVRAG

Atg14L

1 2 3 4 5 6 7

Total

Antibody:

hVps15 (p150)

d

Atg14L

UVRAG

Rubicon

Beclin-1

hVps34

GAPDH

Transferrinreceptor

EEA1

Sec61β

Cathepsin D

20,0

00g

S P S P

100,

000g

PNS

Frac

tion/

tota

l (p

erce

ntag

e)

50

40

30

20

10

01(Vo) 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

669 440 232 158Standard Mr(K)

Beclin-1

Atg14L

UVRAG

RubiconRubiconin Atg14 knockdown

X

75

100

150

100

150

b

c

e

Mr(K)

Mr(K)

Mr(K)W

B:



A RT I C L E S

the MEF–Beclin 1 eluate (Fig. 1a). Analysis by direct nanoflow LC-MS/MS identified six proteins (Fig. 1b; Supplementary Information, Table S1), including Beclin 1 itself, UVRAG and hVps34, which were previously reported to bind Beclin 1 (refs 14,20), p150/hVps15, a regulatory subunit of hVps34 that had not previously been reported to associate with Beclin 1, and two new proteins, KIAA0831 and KIAA0226, which are ubiquitously expressed in human tissues34.

KIAA0831 is predicted to consist of 492 amino acids, with a rela-tive molecular mass of 55,300 (Mr 55.3K). It possesses three potential coiled-coil domains in its N-terminal half (Supplementary Information,

Fig. S1a, b). CLUSTAL analysis showed that a part of KIAA0831 has slight similarity to the yeast Atg14 protein29 (Supplementary Information, Fig. S1a). On the basis of the results discussed below and this similar-ity, we termed KIAA0831 Atg14L (Atg14-like protein). KIAA0226 is a protein of 972 amino acid residues with a calculated molecular mass of 108.6K (Supplementary Information, Fig. S1c). Its N-terminal region (residues 48–189) contains a RUN domain. These domains are not well understood, but some proteins possessing them participate in GTPase function35 (Supplementary Information, Fig. S1d). The KIAA0226 centre region contains a Ser-rich domain (residues 204–447) and a coiled-coil

Figure 2 The effects of Atg14L and Rubicon knockdown on GFP–LC3 dot formation. (a) A549 cells were infected with adenovirus harbouring shRNA targeting Atg14L, Rubicon, UVRAG or a control (luciferase). Cell lysates were subjected to immunoblotting with the antibodies indicated. The asterisk in the Rubicon blot represents a non‑specific cross‑reacting band. (b) Quantification of GFP‑positive dots per cell. Images were taken from cells shown in c and the number of GFP‑positive dots was determined. At

least 140 cells were counted per experiment and the values are mean ± s.d. of three independent experiments. (c) A monoclonal A549 cell line stably expressing GFP–LC3 was infected with adenovirus harbouring shRNA targeting Atg14L, Rubicon or control (luciferase). The cells were cultured under nutrient‑rich or starvation conditions. GFP was observed by confocal laser microscopy. Scale bar, 10 μm. The full scan of the blot in a is available in Supplementary Information, Fig. S6.

150

*

75

100

100

50

Control shRNA Rubicon shRNA

Nut

rient

-ric

hS

tarv

atio

n

GFP

–LC

3-p

ositi

ve d

ots

(dot

/cel

l)

shRNA:

Medium:0

20

40

60

80

100

120

Control

N S

Atg14L

N S

Rubicon

N S

Atg14L shRNA

a

Atg14L

UVRAG

Rubicon

α-tubulin

shRNA: Atg14

L

Contro

l

UVRAG

Rubico

n

b

c

Mr(K)



A RT I C L E S

domain (residues 509–551); the carboxy-terminal region contains a Cys-rich domain (residues 881–932) and it has no apparent homologue in S. cerevisiae. We termed KIAA0226 Rubicon (Run domain protein as Beclin-1 interacting and cystein-rich containing).

We generated specific antibodies against Atg14L, Rubicon and UVRAG, and tested their interactions with Beclin 1 by immunopre-cipitation and immunoblotting. MEF–Beclin 1 robustly pulled down all three endogenous proteins, confirming the identification experiments (Supplementary Information, Fig. S2a).

Identification of three distinct Beclin 1-containing complexesNext, to investigate the composition of endogenous Beclin 1 complexes, we performed immunoprecipitation of each endogenous protein. The anti-Atg14L antibody co-precipitated with Beclin 1, hVps34 and hVps15 but not with UVRAG or Rubicon (Fig. 1c). In contrast, Rubicon co-precipitated with UVRAG, Beclin 1, hVps34 and hVps15 but not with Atg14L. Similarly, UVRAG co-precipitated with Rubicon but not with Atg14L (Fig. 1c). These results suggest that Atg14L and Rubicon are present in different complexes that share Beclin 1, hVps34 and hVps15. UVRAG is also present in the Rubicon complex. It is unlikely that this result is an artefact caused by steric hindrance of the regions bound by these antibodies, as we obtained similar

results following transient expression of tagged proteins and immunopre-cipitation using anti-tag antibodies; Atg14L did not pull down UVRAG or Rubicon, and vice versa (Supplementary Information, Fig. S2b–d). Furthermore, these proteins behaved differently in subcellular fractiona-tion experiments: Rubicon and UVRAG were recovered in the 20,000g and 100,000g pellet fractions, as well as in the major cytosolic pool (Fig. 1d). Atg14L was partly recovered only in the 100,000g pellet faction, consistent with the idea that it is in a complex other than the one containing UVRAG and Rubicon. In gel filtration analysis of the soluble fraction, Rubicon peaked at about 650K (fraction 7), whereas UVRAG peaked at about 500K (fraction 9). The difference in peak size approximates the molecular weight of Rubicon (Fig. 1e; Supplementary Information, Fig. S2e), suggesting that only a subpopulation of UVRAG complexes harbour Rubicon. We expected that knockdown of UVRAG would downshift the peak of Rubicon; however, Rubicon became highly unstable in UVRAG knockdown cells and could not be clearly detected by immunoblotting (Fig. 2a). The Atg14L peak was similar to the Rubicon peak, but was slightly smaller (fraction 7–8; Fig. 1e). Knockdown of Atg14 (Fig. 2a) did not affect the distribution of Rubicon, indicating that Atg14L is not contained in the Rubicon complex (Fig. 1e). Beclin 1 was broadly distributed between the peaks of its three binding part-ners. From these results, we conclude that there exist three types of Beclin

15

50

50

50

75

50

5.5 kbp(WT)

1.9 kbp(KO)

LC3

N S

LC3-I

LC3-II

WT

– + – +

p62

N S N S N S

WT

Medium:

Medium:

Cell:

Cell:L-

14C

-val

ine

rele

ase

(per

cent

age)

Medium:

Cells: WT

N S W N S WN S W0

2

4

6

8

10

12

N S

E64d/pepstatin A:

WT

– + – +

Atg14L+/+ +/– –/–

a

Atg14L

WT

Atg5

–/–

Atg1

4L–/

–b c

d e

α-tubulin

α-tubulin

α-tubulin

Mr(K)Mr(K)

Mr(K)

Atg14L–/–

Atg14L–/– Atg14L–/–

Atg14L–/–

+Atg14L Atg5–/–

Atg14L–/– Atg5–/–

Figure 3 The effect of Atg14L knockout on protein degradation and LC3 turnover. (a) Southern blot analysis of genomic DNA isolated from wild‑type (WT, mAtg14L+/+) , mAtg14L+/– and knockout (KO, mAtg14L–/–) ES cells after digestion with PstI. Details of the probe are shown in Supplementary Information, Fig. S3. (b) Analysis of Atg14L protein in knockout cells. Lysates from WT, mAtg14L–/– ES cells and mAtg5–/– mouse ES cells were subjected to western blotting with the antibodies indicated. (c) The effect of Atg14L knockout on p62 turnover. WT, mAtg14L–/–, mAtg14L–/– + Atg14L, and mAtg5–/– ES cells were cultured in nutrient‑rich (N) or starvation medium (S) for 3 h. Cell lysates were subjected to western

blotting with the antibodies indicated. (d) Degradation of long‑lived proteins is reduced in Atg14L knockout cells. WT, mAtg14L–/–, and mAtg5–

/– ES cells were cultured in N or S medium or S medium with wortmannin (W, 100 nM) for 3 h and degradation of long‑lived proteins was measured. The data are the mean ± s.d. of three independent experiments. (e) The effect of Atg14L knockout on LC3 turnover. WT, mAtg14L–/–and mAtg5–/– ES cells were cultured in N or S medium for 3 h in the presence or absence of E64d (50 µg ml–1) and pepstatin A (50 µg ml–1). Cell lysates were subjected to western blotting with the antibodies indicated. Full scans of the blots in b, c and e are available in Supplementary Information, Fig. S6.



A RT I C L E S

1–hVps34–hVps15 complex: an Atg14L complex, a UVRAG complex and a Rubicon–UVRAG complex. We were able to detect direct interactions between Beclin 1 and Atg14L and between Beclin 1 and UVRAG by yeast two-hybrid analysis (Supplementary Information, Fig. S2f). Pulldown experiments using a deletion series of Beclin 1 protein showed that the coiled-coil region is required for its binding to both UVRAG and Atg14L (Supplementary Information, Fig. S2g, h). We also found that the coiled-coil region in Atg14L is necessary and sufficient for its binding to Beclin 1 (Supplementary Information, Fig. S2g, i). UVRAG also binds to Beclin 1 through its coiled-coil region20. These results indicate that UVRAG and Atg14L bind to the same region in Beclin 1 in a mutually exclusive manner that brings about the existence of different complexes. Rubicon, in turn, binds only to a subpopulation of UVRAG complexes to form the third type of Beclin 1–hVps34–hVps15 complex.

Atg14L deficiency causes defects in autophagosome formationUsing an adenovirus vector-based shRNA, we were able to efficiently knock down the expression of Atg14L in A549 cells (Fig. 2a). We analysed

autophagy in these cells by morphological observation of GFP–LC3, a marker of autophagosomes36 (Fig. 2b, c). In control cells stably express-ing GFP–LC3, the number of GFP–LC3 puncta was small in nutrient-rich conditions but was increased by starvation, indicating induction of autophagy (Fig. 2b, c). In contrast, the number of GFP–LC3 dots was markedly decreased in Atg14L knockdown cells under starvation condi-tions (Fig. 2b, c). This result suggests that Atg14L is required for autophagy. To test this hypothesis, we generated Atg14L knockout mouse ES cells using conventional knockout techniques (Supplementary Information, Fig. S3). The knockout was confirmed by Southern and western blot analyses (Fig. 3a, b). p62/SQSTM1 is a protein involved in the formation of ubiquitin-positive cytoplasmic inclusion bodies and is constitutively degraded by the autophagic machinery through specific binding to LC3 (refs 37,38). Therefore, the steady-state levels of p62 reflect the rate of autophagic degradation. Higher levels of p62 accumulated in Atg14L knockout cells than in control cells, and these levels were comparable to those observed in Atg5 knockout mouse ES cells, another line of mutant cells completely defective in autophagy (Fig. 3c)39. Accumulation of p62

Figure 4 The effect of Atg14L knockout on autophagosome formation. (a) The effect of Atg14L knockout (mAtg14–/–) on Atg16L dot formation. Wild‑type (WT) and mAtg14–/– ES cells were cultured in nutrient‑rich or starvation medium for 3 h. Cells were fixed and subjected to immunofluorescence with an anti‑Atg16L antibody. Images of Atg16L staining (grey‑scale) and Atg16L merged with Hoechst staining (colour) are shown. (b) The effect of Atg14L knockout on LC3 dot formation. WT and mAtg14L–/– ES cells were cultured in nutrient‑rich or starvation

medium for 3 h. Cells were fixed and subjected to immunofluorescence with anti‑LC3 antibody. Images of LC3 staining (grey‑scale) and LC3 merged with Hoechst staining (colour) are shown. Scale bars, 20 μm. (c) Quantification of the number of Atg16L dots in a. The data are the mean ± s.d. of three independent experiments. (d) Quantification of the number of LC3 dots in b. The data are the mean ± s.d. of three independent experiments. For each of the experiments in c and d at least 13 cells were counted.

0.8

a Nutrient rich Starvation

Nutrient rich Starvation

Atg16L

WT

Atg

14L–/

–W

TA

tg14

L–/–

LC3 LC3

Medium:

WT

N S

1.2

0.2

0.4

0.6

1.0

0N S

Atg14L–/–

Medium:

WT

N S N S

Atg14L–/–

4

LC3

dot

s (d

ot/c

ell)

Atg

16L

dot

s (d

ot/c

ell)

6

1

2

3

5

0

b

c

d

Hoechst/Atg16L Atg16L Hoechst/Atg16L

Hoechst/LC3 Hoechst/LC3



A RT I C L E S

could be suppressed by the expression of Atg14L, indicating that the defect is caused by the absence of Atg14L (Fig. 3c). Furthermore, we examined the starvation-induced bulk degradation of long-lived proteins, a stand-ard assay to monitor autophagy. In Atg14L knockout cells, degradation of long-lived proteins was significantly reduced under starvation conditions and was comparable to degradation in Atg5 knockout cells (Fig. 3d). These results suggest that autophagic protein degradation is impaired in Atg14L knockout cells. Lipidated LC3 (LC3-II), but not unlipidated LC3 (LC3-I), binds to autophagosomes and LC3 lipidation correlates with autophago-some formation36. In addition, in Atg14L knockout cells, the levels of LC3-II, a lipidated form of LC3, were markedly reduced, compared with control cells (Fig. 3e). LC3-II on the autophagosome inner membrane is

finally degraded on fusion of autophagosomes and lysosomes40. To assess autophagic flow, we added E64d and pepstatin A, inhibitors of the lyso-somal proteases that inhibit autophagic degradation of LC3, to the cells. Consistent with inhibition of autophagy, this treatment had little effect in increasing the level of LC3-II in Atg14L knockout cells (Fig. 3e).

Next, we examined Atg16L immunofluorescence in these cells. Atg16L is involved in autophagosome formation, and transiently associates with the surface of forming autophagosomes (isolation membranes); there-fore, dot structures observed by Atg16L immunofluorescence represent sites of autophagosome formation41. No formation of Atg16L-positive puncta was observed in Atg14L knockout cells, either under nutrient-rich or starvation conditions (Fig. 4a, c). Furthermore, in knockout cells,

100

Col

ocal

izat

ion

(per

cent

age) Medium:

40

60

80

120

0

20

(Atg

14L

x A

tg16

L)/A

tg16

L

(Atg

14L

x A

tg16

L)/A

tg14

L

(Atg

14L

x LC

3)/A

tg14

L

Starvation

Imm

uno

EM

: ant

i-G

FP

GFP–Atg14L Anti-calnexin

Expression

GFP–Atg14L Atg16L GFP–Atg14L/Atg16L

Nut

rient

-ric

hS

tarv

atio

n

Sta

rvat

ion

a GFP–Atg14L LC3

(Atg

14L

x LC

3)/L

C3

GFP GFP–Atg14L

GFP–Atg14L/LC3b

c

GFP–Atg14L/calnexind

e

Figure 5 Localization of GFP–Atg14L. (a) Immunofluorescence using an anti‑Atg16L antibody. A549 cells were transfected with adenovirus expressing GFP–Atg14L and cultured in nutrient‑rich or starvation medium for 4 h. Cells were fixed and stained with an anti‑Atg16L antibody. (b) Immunofluorescence with an anti‑LC3 antibody. A549 cells were transfected with adenovirus expressing GFP–Atg14L and cultured in starvation medium for 4 h. Cells were fixed and stained with an anti‑LC3 antibody. Scale bar, 10 μm. (c) Colocalization efficiency in a and b. The percentage of colocalization between Atg14L and Atg16L or LC3‑positive dot is shown. For example, (Atg14L×Atg16L)/Atg16L means the ratio of colocalization of Atg14L and Atg16L puncta to all of the

Atg16L‑positive puncta. For colocalization between Atg14L and Atg16L the data are mean ± s.d. of 25 cells. For colocalization between Atg14L and LC3 the data are mean ± s.d. of 20 cells. d) Immunofluorescence with an anti‑calnexin (ER) antibody. A549 cells transfected with adenovirus‑based GFP–Atg14L. After 48 h, the cells were fixed and immunostained with calnexin (ER).Scale bar, 10 μm. (e) Immuno‑EM analysis of GFP–Atg14L. A549 cells transfected with adenovirus‑vector based GFP or GFP–Atg14L were subjected to immuno‑EM analysis using an anti‑GFP antibody. Scale bar, 500 nm. The arrows indicate the ER membrane and the arrowheads indicate the autophagosome membrane.



A RT I C L E S

the starvation-induced increase of LC3 positive dots was completely suppressed, as observed in knockdown cells, and LC3 dots were mostly decreased even under nutrient-rich conditions (Fig. 4b, d). Overexpression of GFP–Atg14L in A549 cells caused a slight increase in the number of LC3 dots under nutrient-rich conditions (Supplementary Information, Fig. S4a). These results suggest that Atg14L is required for both basal and inducible autophagy.

To assess the localization of Atg14L, A549 cells were transfected either with adenovirus expressing N-terminally GFP-tagged human Atg14L or plasmid-based transient expression of Atg14L–GFP and Flag–Atg14L (Fig. 5a, b, d, e; Supplementary Information, Fig. S4a–e). All showed similar distribution; therefore, we conclude that the influence of the tags

on localization was negligible. Under nutrient-rich conditions, most of the GFP–Atg14L was dispersed. During a 4-h starvation period, GFP–Atg14L-positive punctate structures markedly increased in number. Almost half of these puncta overlapped with Atg16L, and most Atg16L-positive puncta overlapped with GFP–Atg14, indicating that a portion of GFP–Atg14L localizes to isolation membranes (Fig. 5a). GFP–Atg14L puncta also overlapped considerably with anti-LC3-positive dots, indicat-ing that GFP–Atg14L also localizes to autophagosomes (Fig. 5b). It should be noted, however, that almost half of the Atg14L-positive puncta did not colocalize with Atg16L and LC3 (Fig. 5c). Furthermore, these non-colocalizing dots were not coincident with either endosomal/lysosomal or Golgi markers (Supplementary Information, Fig. S4b). Some Atg14L

75

50

75

15

50

150

a

LC3LC3-ILC3-II

N

Control

E64d/Pepstatin A:

Rubicon

Medium:

Rubicon

S N S

+– +– +– +–

1 2 4(h) 1 2 4 1 2 4 1 2 4 1 2 4 1 2 4 1 2 4 1 2 4 Mr(K)

Mr(K)

Medium:

shRNA: Control Rubicon

N S W0

5

10

15

20

25

N S

Rubicon

Phosphop70S6K

p70S6K

α-tubulin

Medium: N S

Control

50

Atg

16L

dot

s (d

ot/c

ell) 40

30

20

10

0Medium:


N S

10

Medium:

Rub

icon

N S

20

5

15

0

GFP

–LC

3 p

ositi

ve d

ots

(dot

/cel

l)

shRNA:

N SN S

Atg1

4LR

ubic

on +

Atg

14L

N S

N S W

P = 1.3 X 10–16

b shRNA:

c P = 5.5 X 10–9 d e

shRNA:

α-tubulinL-

14C

-val

ine

rele

ase

(per

cent

age)

Figure 6 The effects of Rubicon knockdown on autophagy. (a) Double knockdown of Rubicon and Atg14L. A549 cells stably expressing GFP–LC3 were infected with adenovirus expressing either shRNA against Rubicon and Atg14L or control shRNA (luciferase). The image is available in Supplementary Information, Fig. S5a. The GFP–LC3 puncta were counted in nutrient‑rich (N) or starvation (S) conditions. The data are mean ± s.d. of three independent experiments; at least 23 cells were counted for each experiment. (b) A549 cells were infected with adenovirus harbouring shRNA for Rubicon or control (luciferase). The cells were cultured in N or S medium in the presence or absence of E64d (50 µg ml–1) and pepstatin A (50 µg ml–1) for 1–4 h. Cell lysates were subjected to western blotting with the antibodies indicated. (c) Quantification of Atg16L puncta in Rubicon knockdown cells. The

images are available in Supplementary Information, Fig. S5c. The data are mean ± s.d., n = 60 (60 cells, sample of Medium: N, shRNA: Control), n = 68 (68 cells, sample of Medium: S, shRNA: Control), n = 78 (78 cells, sample of Medium: N, shRNA: Rubicon), n = 83 (83 cells sample of Medium: S, shRNA: Rubicon). d) Bulk degradation activity of Rubicon knockdown cells. The bulk degradation activity in control and Rubicon knockdown cells in nutrient‑rich medium was measured. Wortmannin (W, 100 nM) was added as a negative control. The data are mean ± s.d. of three independent experiments. (e) A549 cells infected with adenovirus harbouring shRNA for Rubicon and control (luciferase) were cultured in N or S medium for 4 h and cell lysates were subjected to immunoblotting with the antibodies indicated. Full scans of the blots in b and e are available in Supplementary Information, Fig. S6.



A RT I C L E S

showed a reticular pattern and it colocalized with the ER markers calnexin and mStrawberry–msALDH(35) (Fig. 5d; Supplementary Information, Fig. S4e)42. Immuno-electron microscopy analysis supported the ER localization of Atg14L, together with its autophagic structure (Fig. 5e). Together, these results indicate that Atg14L is localized on the autophago-some, isolation membrane, ER and an unknown punctate structure, and that it is indispensable for autophagosome formation.

Rubicon deficiency increases LC3-positive punctaNext, we knocked down Rubicon in A549 cells (Fig. 2a). Strikingly, even in Rubicon knockdown cells cultured under nutrient-rich conditions, the number of GFP–LC3 puncta was increased, compared with starved control cells; starvation further enhanced the number of puncta (Fig. 2b, c). Double knockdown of Rubicon and Atg14L suppressed the number of GFP–LC3 puncta to the level of that observed with Atg14 single knockdown (Fig. 6a;

GFP-Rubicona

100

Expression

140

80

120

GFP

–LC

3-p

ositi

ve d

ots

(dot

/cel

l)

60

40

20

0

mS

traw

ber

ry

mS

traw

ber

ry-M

m_R

ubic

on

mS

traw

ber

ry-M

m_R

ubic

on∆

(393

–849

)

Rub

icon

KD

/nut

rient

ric

h

mS

traw

ber

ry

GFP–LC3 mStrawberry GFP–LC3 / mStrawberryb

GFP–LC3/organelle marker

anti-GM130

anti-EEA1

anti-Rab9

anti-LAMP-1

mS

traw

ber

ry–

Rub

icon

mS

traw

ber

ry–

Rub

icon

∆(39

3-84

9)

Figure 7 Localization of Rubicon. (a) A549 cells transfected with a GFP–Rubicon expression plasmid and were subjected to immunofluorescence with various organelle marker antibodies. Scale bar, 10 μm. (b) A549 cells stably expressing GFP–LC3 were subjected to Rubicon knockdown (KD) and then transfected with mStrawberry, mStrawberry–Rubicon or mStrawberry–Rubicon mutant lacking the Beclin 1 binding region

(Δ393–849). Cells with strong expression of mStrawberry–Rubicon showed an overexpression‑dependent dominant‑negative phenotype; therefore, we selected cells with modest expression. The number of GFP–LC3 puncta was counted and mean ± s.d. are shown. mStrawberry, n = 12 (12 cells); mStrawberry‑Mm‑Rubicon, n = 14 (14 cells); mStrawberry‑Mm‑RubiconΔ393–849, n = 13 (13 cells).



A RT I C L E S

Supplementary Information, Fig. S5a). This indicates that Atg14L is required for Rubicon knockdown-dependent GFP–LC3 puncta formation. However, western blot analysis showed that endogenous LC3 levels are reduced in Rubicon knockdown cells not transfected with GFP–LC3 (Fig. 6b). This

could not be attributed to decreased LC3 mRNA, as quantitative RT–PCR of LC3A/B/C mRNA showed no significant change between control and Rubicon knockdown cells (Supplementary Information, Fig. S5b). Moreover, the decrease in LC3-II in the cells was not due to inhibition of LC3 lipidation.

150

150

25

50

50

150

250

150

GFP-Rubicon

a GFP-LC3 LAMP-1 GFP-LC3 /LAMP-1

LAM

P-1

/GFP

-LC

3 (p

erce

ntag

e)

40

30

20

10

0


E64d/Pepstatin A: +– +–

Num

ber

of A

P o

r A

L p

er c

ell

15

5

20

10

0

shRNA: Control RubiconAP AL

shR

NA

Con

trol

Rub

icon

shR

NA

Con

trol

Rub

icon

Med

ium

: nut

rient

+ P

I

AP AL

P = 0.014

Medium:

EGFR

EGF (min): 0 120 240 0 120 240

GFPExpression:

GFP–Rubicon

GFP

EGF : (min)

shRNA: Control Rubicon0 15 60 120 180

EGFR

Rubicon

α-tubulin

α-tubulin

0 15 60 120 180

EG

FR/t

ubul

in (p

erce

ntag

e)

120100806040200

140160180200

(Time min)0 60 120 180

P = 3.6 X 10–10

b

P = 2.8 X 10–5

Nutrient-rich

c

d e

Mr(K)

Mr(K)

Atg14LBeclin-1hVps34hVps15

UVRAGBeclin-1hVps34hVps15

RubiconUVRAGBeclin-1hVps34hVps15

Fusion

Enhance?

Endocytictraffic

Autophagosomematuration

Endosome/lysosome

Autophagosomeformation

Figure 8 The role of Rubicon in autophagosome maturation and endocytic traffic. (a) The maturation of autophagosomes is enhanced in Rubicon knockdown cells. A549 cells stably expressing GFP–LC3, with Rubicon or control shRNA, were incubated in nutrient‑rich medium with or without the protease inhibitors (PI) E64d (50 µg ml–1) and pepstatin A (50 µg ml–1) treatment for 4 h. The cells were stained with an anti‑Lamp1 antibody (upper panels) and colocalization efficiency between GFP–LC3 and Lamp1 was counted (lower panel). The data are mean ± s.d. of 30 cells. (b). Electron microscopy analysis of Rubicon knockdown cells. Control (luciferase) or Rubicon shRNA A549 cells grown in nutrient‑rich medium was subjected to electron microscopy analysis (left panels). The arrows indicate autophagosomes and the arrowheads indicate autolysosomes. Scale bar, 1 μm. Autophagosomes and autolysosomes were counted in 21 cells, and the mean ± s.d. are shown (right panel). (c) EGFR degradation in Rubicon knockdown cells. Control or Rubicon shRNA A549 cells were

treated with EGF (200 ng ml–1) for the indicated periods and the lysates were subjected to western blotting with the antibodies indicated (left panel). The band intensity was measured in three independent experiments and the mean ± s.d. are shown (right panel). (d) The effect of Rubicon overexpression in the turnover of EGF receptor. A549 cells were transfected with advenovirus‑based GFP or GFP–Rubicon. After 24 h, EGF was added to cells expressing GFP–Rubicon or control GFP. At the times indicated, the cells were lysed and subjected to western blotting with the indicated antibodies. (e) Model of the function of three Beclin 1 containing complexes. The Atg14L complex functions in autophagosome formation. The UVRAG complex functions in autophagosome and endosome maturation. The Rubicon–UVRAG complex suppresses autophagosome and endosome maturation. Enhanced autophagosome maturation and/or endocytosis may lead to enhanced autophagosome formation. Full scans of the blots in c and d are available in Supplementary Information, Fig. S6.



A RT I C L E S

When Rubicon knockdown cells were treated with the protease inhibitors E64d and pepstatin A, LC3-II levels increased markedly to a level comparable to that in control cells (Fig. 6b). This strongly suggests that LC3-II is formed but then degraded, depending on autophagic flow to lysosomes. Consistent with this, in the Rubicon knockdown cells, even under nutrient-rich con-ditions, the number of endogenous Atg16L dots was markedly increased (Fig. 6c; Supplementary Information, Fig. S5c). Consistently, degradation of long-lived proteins in the Rubicon knockdown cells was much higher than in control cells (Fig. 6d). Interestingly, increased degradation of p62 was not observed (Supplementary Information, Fig. S5d). Autophagic degradation of p62 is thought to involve a specific recognition process; therefore, the increased autophagy in Rubicon knockdown cells may exclude p62-targeting autophagy. Collectively, these results indicate that loss of Rubicon enhances autophagy. Decreased LC3 levels can be explained by hyperinduction of autophagic degradation. We also monitored the phosphorylation of p70 S6 kinase, an index of the activity of mTOR (mammalian target of rapamy-cin), an autophagy suppressor. In Rubicon knockdown cells, mTOR activity was not significantly different from control cells (Fig. 6e). This means that Rubicon acts either downstream of mTOR or in an independent pathway. To discriminate between these possibilities, we examined the effect of Rheb, the upstream activator of mTOR43. Overexpression of Rheb and hyperactivation of mTOR suppressed GFP–LC3 dot formation caused by Rubicon knock-down (Supplementary Information, Fig. S5e). These data strongly suggest that the action point of Rubicon is independent of the mTOR-dependent autophagy induction pathway. Thus, Rubicon is negatively involved in autophagic processes, but not through changes in the mTOR pathway.

Rubicon knockdown enhances autophagosome maturation and endocytosisWe next examined the subcellular localization of Rubicon (Fig. 7a). As the antibody against Rubicon was not suitable for indirect immunofluo-rescence, we used transient expression of GFP–Rubicon, Rubicon–GFP and Flag–Rubicon. All tagged Rubicons colocalized with the late endo-some/lysosome markers Rab9 and LAMP1, and partially overlapped with the early endosome marker EEA1, but did not colocalize with the Golgi-marker GM130 (Figs 7a; Supplementary Information, Fig. S4f, g) or starvation-induced Atg16L and LC3 dots (Supplementary Information, Fig. S4h). GFP–UVRAG showed similar late-endosome/lysosome locali-zation patterns in addition to partial localization to the early endosome (Supplementary Information, Fig. S4i). Thus, the Rubicon–UVRAG–Beclin 1 complex localizes to the early and late-endosome/lysosomes.

Next we mapped the binding regions of each protein. We identi-fied a region in Rubicon that is sufficient for binding to the UVRAG–Beclin 1-containing complex (Supplementary Information, Fig. S5f, g). RubiconΔ393–849, a mutant that lacks the binding region, did not bind to the Beclin 1 complex (Supplementary Information, Fig. S5f, g). The RubiconΔ393–849 mutant also failed to suppress the autophagy enhance-ment phenotype in Rubicon knockdown cells, supporting the idea that Rubicon functions by binding to the UVRAG–Beclin 1 complex (Fig. 7b). RubiconΔ393–849 also failed to localize to endosome/lyso-somes, supporting the idea that Rubicon exists there as a complex with UVRAG–Beclin 1 (Fig. 7b).

This localization pattern prompted us to examine the maturation proc-ess of autophagy, which is the fusion of autophagosomes and the endo-some/lysosome. Following a 4-h lysosomal protease inhibitor treatment of Rubicon knockdown cells, a subpopulation of GFP–LC3 puncta became

LAMP-1 positive at a significantly higher percentage than observed dur-ing constitutive autophagy in control cells (Fig. 8a). Electron microscopy analysis of Rubicon knockdown cells showed that more autolysosomes accumulated than autophagosomes, and its ratio was much higher than constitutive autophagy (Fig. 8b). On the other hand, overexpression of Rubicon inhibited turnover of the LC3-II form, indicating that autophago-some maturation was severely impaired (Supplementary Information, Fig. S5h). Together, these results indicate that Rubicon is negatively involved in the autophagosome maturation process.

We further examined whether the endocytic pathway is affected in Rubicon knockdown cells. We found that lysosomal degradation of endocytosed EGF receptor was accelerated, compared with control cells (Fig. 8c). Moreover, the amount of transferrin receptor, which is usually recycled back to the plasma membrane after endocytic internalization, was markedly reduced in these mutants (Supplementary Information, Fig. S5i). In contrast, overexpression of GFP–Rubicon caused defects in the endocytic pathway: degradation of the EGF receptor after endocytic internalization was inhibited (Fig. 8d). Furthermore, EGF accumulated in an abnormally enlarged compartment distinct from the lysosome, which was labelled with fluid-phase endocytic marker dye before transfection with GFP–Rubicon (Supplementary Information, Fig. S5j). These results indicate Rubicon negatively regulates endocytic trafficking as well.

DISCUSSIONHere, we have identified two Beclin 1 interacting proteins, Atg14L and Rubicon. We have also shown that three distinct Beclin 1 complexes exist in cells: one contains Beclin 1, hVps34, hVps15 and Atg14L; the second contains Beclin 1, hVps34, hVps15 and UVRAG; and the third contains Beclin 1, hVps34, hVps15, UVRAG and Rubicon. The knockdown phe-notypes of Atg14L and Rubicon in A549 cells are different, demonstrating that Beclin 1 has multiple roles in autophagy through formation of differ-ent complexes (Fig. 8e).

In mouse ES cells, autophagic degradation was inhibited by knock-out of Atg14L; therefore, we conclude that Atg14L is necessary for the process. The similarity between Atg14L and yeast Atg14 is quite low; however, the structural properties seem to be conserved, particularly in the N-terminal coiled-coil domains, which are important for yeast Atg14 function44. The most plausible role of Atg14L is to divert Beclin 1–hVps34/Class III PI(3)K into an autophagic role. GFP–Atg14L localizes on the isolation membrane; therefore, it is possible that Atg14L directly or indirectly determines the site where the complex localizes. Beclin 1–GFP localizes on LC3 positive autophagosomes and/or isolation membranes24, and Atg14L may determine this localization. Furthermore, a subpopula-tion of Atg14L seems to localize on the ER membrane. ER has long been considered a candidate for the source of autophagosome membrane and recent paper reported that a specialized domain of ER has an important role in autophagosome formation by recruiting hVps34 (ref. 45). This point should be further examined in future studies.

We have characterized the Beclin 1-complex-associating protein Rubicon, whose knockdown results in a considerable increase in the number of autophagosomes/autolysosomes in cells. On the basis of the observation that the percentage of autolysosome is increased in those cells, we reasoned that Rubicon is negatively involved in the maturation of the autophagosome. Furthermore, endocytosis is facilitated in Rubicon knockdown cells. These results show an interesting similarity with a recent report that UVRAG overexpression provides the same effects in autophagosome maturation



A RT I C L E S

and endocytosis21. Thus, the Rubicon–UVRAG–Beclin 1–hVps34–hVps15 complex suppresses the process of autophagosome maturation and endocy-tosis, whereas the UVRAG complex seems to work in the opposite direction. This model is further supported by the observation that overexpression of Rubicon inhibits autophagosome maturation and endocytic traffick-ing; these phenotypes are also seen in UVRAG knockdown cells21. Because Rubicon localizes to endosome/lysosome, it may be directly involved in the regulation of membrane fusion processes of endosome/lysosome and autophagosome. This is a reasonable speculation as UVRAG has the capacity to bind to the Class C VPS complex, which is involved both in the fusion process of autophagosomes and in endocytosis21. These steps may be regulated through PI(3)P generated by the Beclin 1–hVps34 complex: in yeast, Vam7, a SNARE protein harbouring a PI(3)P-binding PX domain, is important in these steps, together with Class C VPS46. Furthermore, autophagosome formation is also enhanced in Rubicon knockdown cells as well as in UVRAG-overexpressing cells21. The rate-limiting step in overall autophagy is generally believed to be autophagosome formation; however, the maturation process may also represent another rate limiting step47. Although hVps34 is proposed to be the upstream activator of mTOR48, the knockdown data suggest that Rubicon acts independently of mTOR regulation. Thus, there may be a positive-feedback mechanism operating between the maturation process and the regulation of induction; further study may provide important clues about the molecular mechanisms of autophagy induction.

Further characterization of the separate Beclin 1-containing com-plexes could shed light on the complex mechanism of Beclin 1-depend-ent autophagy regulation. The concerted action of multiple accessory proteins in the Beclin 1–hVps34 axis may ‘tune’ autophagy in response to various physiological conditions.

Note added in proof: a related manuscript by Zhong et al. (Nature Cell Biol. 11, doi:10.1038/ncb1854; 2009) is also published in this issue. While our manuscript was under review, two studies49,50 reported the identifica-tion of the mammalian homologue of Atg14 protein.

METHODSMEF tag based protein purification and tandem mass spectrometry. The puri-fication procedure was essentially the same as reported previously33. Briefly, about 5 × 108 cells were lysed in 13 ml of lysis buffer. Cleared lysates were subjected to immunoprecipitation with anti-Myc antibody, cleavage by TEV protease, immu-noprecipitation with an anti-Flag antibody and Flag-peptide-dependent elution. The final eluate was separated by SDS–PAGE and visualized by silver staining. Specific bands were excised and digested in the gel with trypsin, and the resulting peptide mixtures were analysed by nanoflow LC-MS/MS at Tokyo Metropolitan University. All MS/MS spectra were searched against the non-redundant protein sequence database at the National Center for Biotechnology Information (NCBI) using the Mascot software (Matrix Science).

Adenoviral shRNA expression system. Oligonucleotide sequences for shRNA interference with Atg14L, UVRAG, or Rubicon expression are bp 383–405 of Atg14L (5´-GCAAGAUGAGGAUUGAACA-3´), bp 513–536 of Rubicon (5´-GAUCGAUGCGUCCAUGUUU-3´), bp 484–502 of UVRAG (5´-GCCAGACCGTCTTGATACA-3´) and GL2 luciferase (negative control, 5´-CGTACGCGGAATACTTCGA-3´) followed by a 9-nucleotide non-com-plementary spacer (TTCAAGAGA) and the reverse complement of the initial 19-nucleotide sequence. These dsDNA oligonucleotides were cloned into the pENTR/U6 vector (Invitrogen) and transferred into the pAd/PL vector (Invitrogen) by the Clonase LR recombination reaction (Invitrogen). Adenoviral production was performed in accordance with the manufacturer’s protocol. Sub-confluent cells in 35-mm dishes were infected with adenovirus and transferred to 60-mm dishes 24 h later. After an additional 24 h, the cells were infected again with the virus. The

medium was changed after an additional 24 h and after a further 24 h, the cells were transferred to 100-mm dishes. After an additional 48 h, the cells were re-plated at 50% confluency and experiments were performed 24 h later.

Long-lived protein degradation assay. Wild-type and Atg14 knockout cells were plated on 0.1% gelatine-coated 24-well plates and cultured in complete ES medium for 12 h. Cells were then incubated for 18 h in complete ES medium containing l-14C-valine (0.6 μCi ml–1) (Moravec). Cells were washed three times with complete ES medium and incubated for 4 h with complete ES medium containing unlabelled valine (10 mM). After three washes with complete ES medium, cells were incubated with complete ES medium or Earle’s balanced salt solution (EBSS) containing unlabelled valine (10 mM) and 0.1% BSA in the presence or absence of wortmannin (100 nM). After 3 h, the medium was precipitated in 10% TCA and TCA-soluble radioactivity was measured. Cells were lysed with RIPA buffer (150 mM NaCl, 50 mM Tris-HCl, 5 mM EDTA, 0.1% SDS, 1% TritonX-100, 1× protease inhibitor cocktail, 1 mM PMSF) and precipitated in 10% TCA. Precipitates were then washed once with acetone. Total cell radioactivity was measured after solubilization with 6 M urea. L-14C-valine release was estimated as a percentage of the radioactivity in the TCA-soluble material relative to the total cell radioactivity. For Rubicon knockdown cells, ES medium was replaced with DMEM and the final incubation was 4 h using the above method.

Electron microscopy. GFP–Atg14L was overexpressed in A549 cells cultured on a polystyrene coverslip, Cell Desk (Sumitomo Bakelite). Cells were starved for 4 h in EBSS, fixed with 4% paraformaldehyde for 1 h in 0.1M sodium-phosphate buffer (pH 7.4) and washed for 5 min three times in sodium-phosphate buffer. Cells were permeabilized and blocked for 30 min with 0.2% saponin, 10% BSA, 10% normal goat serum and 0.1% cold-water fish gelatin in the sodium-phosphate buffer. Cells were stained with anti GFP rabbit polyclonal antibodies (ab6556, Abcam) overnight at 4 °C, washed for 10 min six times in sodium-phosphate buffer containing 0.1% saponin, then stained for 2 h at room temperature with an anti rabbit IgG con-jugated to 1.4 nm gold particle (Nanogold Fab’ fragment of goat anti-rabbit IgG, Nanoprobes), washing for 10 min five times in sodium-phosphate buffer containing 0.1% saponin and for another 10 min without saponin. Cells were fixed for 10 min with 1% glutaraldehyde and washed for 5 min three times in sodium-phosphate same buffer to prevent secondary antibodies from uncoupling. Cells were treated with the gold enhancement mixture GoldEnhance-EM (Nanoprobes) to increase the size of gold particles and improve visualization by electron microscopy, and washed with distilled water. Cells were then post-fixed for 1 h with 1% osmium tetroxide and 1.5% potassium ferrocyanide in 0.1 M sodium-phosphate buffer (pH 7.4), dehydrated in a graded series of ethanol and embedded in Epon812 (TAAB). Ultra-thin (80 nm) sections of cells were stained with saturated uranyl acetate and Reynolds lead citrate solution. Electron micrographs were obtained with a JEM-1011 transmission electron microscope (JEOL). Rubicon knockdown cells on Cell Desk were fixed with 2% paraformaldehyde and 2.5% glutaraldehyde in 0.1 M sodium phosphate buffer overnight at 4 °C. After starvation, washed with 0.1 M sodium phosphate buffer and then post-fixed for 1 h with 1% osmium tetroxide and 1% potassium ferrocyanide in 0.1 M sodium pohosphate buffer. The following procedure was as described above.

Gel filtration. Gel filtration analysis was performed as described previously51. A549 cells were homogenized in two volumes of ice-cold homogenization buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1 mM PMSF, protease inhibitor cocktail; Roche) by repeatedly shearing 15 times through a 25-gauge needle mounted on a 1-ml syringe. The homogenate was subjected to low-speed centrifugation at 17,000g for 5 min and then ultracentrifuged at 100,000g for 60 min. The resulting supernatant was then applied to a Superose 6 column (GE Healthcare) and eluted at a flow rate of 0.5 ml min–1. Fractions (0.6 ml) were examined by western blotting.

Differential centrifugation. A549 cells were homogenized in two volumes of ice-cold homogenization buffer (20 mM Hepes-KOH (pH 7.4), 0.08 M sucrose 0.22 M mannitol, 1 mM PMSF, protease inhibitor cocktail; Roche), for centrifugation at 100,000g; homogenization buffer was supplemented with 1 mM KCl.) by repeat-edly shearing 15 times through a 25-gauge needle mounted on a 1 ml syringe. After 1,000g for 10 min, lysates were subjected to low-speed centrifugation at 20,000g for 10 min to generate a pellet fraction. The supernatant was further centrifuged at 100,000g for 60 min to generate supernatant and pellet fractions. Equivalent fractions were examined by immunoblotting.



A RT I C L E S

Degradation of the EGF receptor. A549 cells in DMEM and 10% FBS were trans-fected with adenovirus harbouring GFP–Rubicon or GFP; after 24 h, the medium was changed. After another 24 h, the medium was changed to DMEM without serum and the samples were incubated for 4 h. EGF (200 ng ml–1; Invitrogen) was added and the cells were retrieved at 0, 2 and 4 h. Control and Rubicon knockdown A549 cells were incubated in DMEM without serum for 2 h. EGF (200 ng ml–1) was added and the cells were retrieved at 0, 15, 60, 120, 180 min. These cell lysates were subjected to western blotting with anti-EGF receptor antibody.

Statistical analysis. Statistical analyses were performed using a two-tailed unpaired t-test. P values < 0.05 were considered statistically significant.

Note: Supplementary Information is available on the Nature Cell Biology website.

AcKNOwledgeMeNTSThe authors thank Beth Levine for Beclin 1 cDNA, Noboru Mizushima for anti-Atg16L antibody and Atg5 knockout mouse ES cells, Sumio Sugano and Yutaka Suzuki for UVRAG cDNA, Toshio Kitamura for PLAT-E cells and pMX-puro vector, Roger Y. Tesien for the plasmid encoding mStrawberry protein, Ryuichi Masaki for pEGFP–msALDH(35) vector, Yusuke Yamada for Beclin 1 deletion mutants cDNA construction, Asaya Nishi and Kunihiro Kawanishi for technical assistance. The work described in this report was supported in part by Special Coordination Funds for Promoting Science and Technology of the Ministry of Education, Culture, Sports, Science and Technology (MEXT) of Japan.

AuTHOr cONTrIbuTIONSK.M. performed most of the experiments; T. Saitoh, T. Satoh and S.A. generated Atg14–/– ES cells; K.T. performed the experiments shown in Fig. 8a, c and Supplementary Information, Fig. S5i; H.O. performed electron microscopy; N.K. performed the experiments shown in Supplementary Information, Fig. 3e and S5d; I.M. and K.S.N. provided technical support; T. Ichimura and T. Isobe provided MEF system techniques and mass spectrometry data analysis; K.M., T.N. and T.Y. analysed and discussed the data; T.N. and T. Y. wrote the manuscript; T. Y. supervised the project.

cOMpeTINg fINANcIAl INTereSTSThe authors declare no competing financial interests.


1. Levine, B. & Klionsky, D. J. Development by self‑digestion: molecular mechanisms and biological functions of autophagy. Dev. Cell 6, 463–477 (2004).

2. Dice, J. F. Chaperone‑mediated autophagy. Autophagy 3, 295–299 (2007).3. Muller, O. et al. Autophagic tubes: vacuolar invaginations involved in lateral membrane

sorting and inverse vesicle budding. J. Cell Biol. 151, 519–528 (2000).4. Mizushima, N. Autophagy: process and function. Genes Dev. 21, 2861–2873 (2007).5. Pattingre, S. et al. Bcl‑2 antiapoptotic proteins inhibit Beclin 1‑dependent autophagy.

Cell 122, 927–939 (2005).6. Paludan, C. et al. Endogenous MHC class II processing of a viral nuclear antigen after

autophagy. Science 307, 593–596 (2005).7. Nakagawa, I. et al. Autophagy defends cells against invading group A Streptococcus.

Science 306, 1037–1040 (2004).8. Hara, T. et al. Suppression of basal autophagy in neural cells causes neurodegenerative

disease in mice. Nature 441, 885–889 (2006).9. Komatsu, M. et al. Loss of autophagy in the central nervous system causes neurode‑

generation in mice. Nature 441, 880–884 (2006).10. Liang, X. H. et al. Induction of autophagy and inhibition of tumorigenesis by beclin 1.

Nature 402, 672–676 (1999).11. Tassa, A., Roux., M. P., Attaix, D. & Bechet, D. M. Class III phosphoinositide 3‑kinase–

Beclin1 complex mediates the amino acid‑dependent regulation of autophagy in C2C12 myotubes. Biochem. J. 376, 577–586 (2003).

12. Zeng, X., Overmeyer, J. H. & Maltese, W. A. Functional specificity of the mammalian Beclin–Vps34 PI 3‑kinase complex in macroautophagy versus endocytosis and lyso‑somal enzyme trafficking. J. Cell Sci. 119, 259–270 (2006).

13. Yue, Z., Jin, S., Yang, C., Levine, A. J. & Heintz, N. Beclin 1, an autophagy gene essential for early embryonic development, is a haploinsufficient tumor suppressor. Proc. Natl Acad. Sci. USA 100, 15077–15082 (2003).

14. Kihara, A., Kabeya, Y., Ohsumi, Y. & Yoshimori, T. Beclin–phosphatidylinositol 3‑kinase complex functions at the trans‑Golgi network. EMBO Rep. 2, 330–335 (2001).

15. Furuya, N., Yu, J., Byfield, M., Pattingre, S. & Levine, B. The evolutionarily conserved domain of Beclin 1 is required for Vps34 binding, autophagy and tumor suppressor function. Autophagy 1, 46–52 (2005).

16. Lindmo, K. & Stenmark, H. Regulation of membrane traffic by phosphoinositide 3‑kinases. J. Cell Sci. 119, 605–614 (2006).

17. Skanland, S. S., Walchli, S., Utskarpen, A., Wandinger‑Ness, A. & Sandvig, K. Phosphoinositide‑regulated retrograde transport of ricin: crosstalk between hVps34 and sorting nexins. Traffic 8, 297–309 (2007).

18. Petiot, A., Ogier‑Denis, E., Blommaart, E. F., Meijer, A. J. & Codogno, P. Distinct classes of phosphatidylinositol 3´‑kinases are involved in signaling pathways that control mac‑roautophagy in HT‑29 cells. J. Biol. Chem. 275, 992–998 (2000).

19. Blommaart, E. F., Krause, U., Schellens, J. P., Vreeling‑Sindelarova, H. & Meijer, A. J. The phosphatidylinositol 3‑kinase inhibitors wortmannin and LY294002 inhibit autophagy in isolated rat hepatocytes. Eur. J. Biochem. 243, 240–246 (1997).

20. Liang, C. et al. Autophagic and tumour suppressor activity of a novel Beclin1‑binding protein UVRAG. Nature Cell Biol. 8, 688–699 (2006).

21. Liang, C. et al. Beclin1‑binding UVRAG targets the class C Vps complex to coordinate autophagosome maturation and endocytic trafficking. Nature Cell Biol. 10, 776–787 (2008).

22. Takahashi, Y. et al. Bif‑1 interacts with Beclin 1 through UVRAG and regulates autophagy and tumorigenesis. Nature Cell Biol. 9, 1142–1151 (2007).

23. Fimia, G. M. et al. Ambra1 regulates autophagy and development of the nervous system. Nature 447, 1121–1125 (2007).

24. Ropolo, A. et al. The pancreatitis‑induced vacuole membrane Protein 1 triggers autophagy in mammalian cells. J. Biol. Chem. 282, 37124–37133 (2007).

25. Shimizu, S. et al. Role of Bcl‑2 family proteins in a non‑apoptotic programmed cell death dependent on autophagy genes. Nature Cell Biol. 6, 1221–1228 (2004).

26. Yu, L. et al. Regulation of an ATG7–beclin 1 program of autophagic cell death by caspase‑8. Science 304, 1500–1502 (2004).

27. Maiuri, M. C. et al. Functional and physical interaction between Bcl‑X(L) and a BH3‑like domain in Beclin‑1. EMBO J. 26, 2527–2539 (2007).

28. Qu, X. et al. Promotion of tumorigenesis by heterozygous disruption of the beclin 1 autophagy gene. J. Clin. Invest. 112, 1809–1820 (2003).

29. Kametaka, S., Okano, T., Ohsumi, M. & Ohsumi, Y. Apg14p and Apg6/Vps30p form a protein complex essential for autophagy in the yeast, Saccharomyces cerevisiae. J. Biol. Chem. 273, 22284–22291 (1998).

30. Seaman, M. N., Marcusson, Cereghino, J. L. & Emr, S. D. Endosome to Golgi retrieval of the vacuolar protein sorting receptor, Vps10p, requires the function of the VPS29, VPS30, and VPS35 gene products. J. Cell Biol. 137, 79–92 (1997).

31. Burda, P., Padilla, S. M., Sarkar, S. & Emr, S. D. Retromer function in endosome‑to‑Golgi retrograde transport is regulated by the yeast Vps34 PtdIns 3‑kinase. J. Cell Sci. 115, 3889–3900 (2002).

32. Kihara, A., Noda, T., Ishihara, N. & Ohsumi, Y. Two distinct Vps34 phosphatidyli‑nositol 3‑kinase complexes function in autophagy and carboxypeptidase Y sorting in Saccharomyces cerevisiae. J. Cell Biol. 152, 519–530 (2001).

33. Ichimura, T. et al. 14‑3‑3 proteins modulate the expression of epithelial Na+ chan‑nels by phosphorylation‑dependent interaction with Nedd4–2 ubiquitin ligase. J. Biol. Chem. 280, 13187–13194 (2005).

34. Kikuno, R., Nagase, T., Waki, M. & Ohara, O. HUGE: a database for human large proteins identified in the Kazusa cDNA sequencing project. Nucleic Acids Res. 30, 166–168 (2002).

35. Callebaut, I., de Gunzburg, J., Goud, B. & Mornon, J. P. RUN domains: a new family of domains involved in Ras‑like GTPase signaling. Trends Biochem. Sci. 26, 79–83 (2001).

36. Kabeya, Y. et al. LC3, a mammalian homologue of yeast Apg8p, is localized in autophagosome membranes after processing. EMBO J. 19, 5720–5728 (2000).

37. Komatsu, M. et al. Homeostatic levels of p62 control cytoplasmic inclusion body formation in autophagy‑deficient mice. Cell 131, 1149–1163 (2007).

38. Bjorkoy, G. et al. p62/SQSTM1 forms protein aggregates degraded by autophagy and has a protective effect on huntingtin‑induced cell death. J. Cell Biol. 171, 603–614 (2005).

39. Mizushima, N. et al. Dissection of autophagosome formation using Apg5‑deficient mouse embryonic stem cells. J. Cell Biol. 152, 657–668 (2001).

40. Tanida, I., Minematsu‑Ikeguchi, N., Ueno, T. & Kominami, E. Lysosomal turnover, but not a cellular level, of endogenous LC3 is a marker for autophagy. Autophagy 1, 84–91 (2005).

41. Mizushima, N. et al. Mouse Apg16L, a novel WD‑repeat protein, targets to the autophagic isolation membrane with the Apg12–Apg5 conjugate. J. Cell Sci. 116, 1679–1688 (2003).

42. Masaki, R., Kameyama, K. & Yamamoto, A. Post‑translational targeting of a tail‑anchored green fluorescent protein to the endolpasmic reticulum. J. Biochem. 134, 415–426 (2003).

43. Scott, R. C., Schuldiner, O. & Neufeld, T. P. Role and regulation of starvation‑induced autophagy in the Drosophila fat body. Dev. Cell 7, 167–178 (2004).

44. Obara, K., Sekito, T. & Ohsumi, Y. Assortment of phosphatidylinositol 3‑kinase com‑plexes — Atg14p directs association of complex I to the pre‑autophagosomal structure in Saccharomyces cerevisiae. Mol. Biol. Cell 17, 1527–1539 (2006).

45. Axe, E. L. et al. Autophagosome formation from membrane compartments enriched in phosphatidylinositol 3‑phosphate and dynamically connected to the endoplasmic reticulum. J. Cell Biol. 182, 685–701 (2008).

46. Sato, T. K., Darsow, T. & Emr, S. D. Vam7p, a SNAP‑25‑like molecule, and Vam3p, a syntaxin homolog, function together in yeast vacuolar protein trafficking. Mol. Cell Biol. 18, 5308–5319 (1998).

47. Kimura, S., Noda, T. & Yoshimori, T. Dynein‑dependent movement of autophagosomes mediates efficient encounters with lysosomes. Cell Struct. Funct. 33, 109–122 (2008).

48. Gulati, P. et al. Amino acids activate mTOR complex 1 via Ca2+/CaM signaling to hVps34. Cell. Metab. 7, 456–465 (2008).

49. Itakura, E. et al. Beclin 1 forms two distinct phosphatidylinositol‑3‑kinase complexes with mammalian Atg14 and UVRAG. Mol. Biol. Cell 19, 5360–5372 (2008).

50. Sun, Q. et al. Indentification of Barkor as a mammalian autophagy‑specific factor for Beclin 1 and class III phosphatidylinositol‑3‑kinase. Proc. Natl Acad. Sci USA 105, 19211–19216 (2008).

51. Fujita, N. et al. The Atg16L complex specifies the site of LC3 lipidation for membrane biogenesis in autophagy. Mol. Biol. Cell 19, 2092–2100 (2008).



A RT I C L E S

Phosphorylation by Akt1 promotes cytoplasmic localization of Skp2 and impairs APC–Cdh1-mediated Skp2 destructionDaming Gao1, Hiroyuki Inuzuka1, Alan Tseng1, Rebecca Y. Chin1, Alex Toker1 and Wenyi Wei1,2

Deregulated Skp2 function promotes cell transformation, and this is consistent with observations of Skp2 overexpression in many human cancers. However, the mechanisms underlying elevated Skp2 expression are still unknown. Here we show that the serine/threonine protein kinase Akt1, but not Akt2, directly controls Skp2 stability by a mechanism that involves degradation by the APC–Cdh1 ubiquitin ligase complex. We show further that Akt1 phosphorylates Skp2 at Ser 72, which is required to disrupt the interaction between Cdh1 and Skp2. In addition, we show that Ser 72 is localized within a putative nuclear localization sequence and that phosphorylation of Ser 72 by Akt leads to cytoplasmic translocation of Skp2. This finding expands our knowledge of how specific signalling kinase cascades influence proteolysis governed by APC–Cdh1 complexes, and provides evidence that elevated Akt activity and cytoplasmic Skp2 expression may be causative for cancer progression.

The SCF-Skp2 E3 ubiquitin ligase complex regulates the destruction of numerous cell-cycle regulators including p27, Foxo1 and p130 (ref. 1). Elevated Skp2 expression is frequently observed in many tumours, includ-ing breast and prostate carcinomas2,3. However, the molecular mecha-nisms underlying elevated Skp2 expression have not been fully explored. We and others have identified Cdh1 as the E3 ligase that promotes Skp2 destruction4,5. Compared with the frequency of Skp2 overexpression, loss of Cdh1 is not a frequent event in human cancer. In contrast, hyperacti-vation of the Akt pathway is considered a hallmark of many cancers and it has been reported that activation of the phosphatidylinositol-3-OH kinase (PI(3)K)/Akt pathway enhances p27 destruction6. This suggests that sustained Akt activity can influence Skp2 activity7,8.

The Akt family of kinases includes three closely related family members designated Akt1, Akt2 and Akt3 (ref. 9). Because most of the upstream regulators and downstream mediators of the Akt pathway are either onco-genes or tumour suppressors, it is not surprising to find that Akt activity is abnormally elevated in most human cancers10. Enhanced Akt signal-ling in tumour cells can suppress apoptosis by promoting the phospho-rylation and subsequent cytoplasmic localization of many downstream pro-apoptotic protein targets such as Bad11, Foxo1 (ref. 12) and Foxo3a13. Akt upregulation can also promote cell growth by inactivating the nega-tive cell-cycle regulators p21 (ref. 14) and p27 (refs 15–17). Most studies exploring a role for the Akt pathway in cell-cycle progression, survival and cancer progression have generally assumed that all three isoforms function in overlapping, redundant roles. However, recent studies have begun to suggest isoform-specific functions for Akt18–20.

Here we have evaluated the mechanism by which Akt controls Skp2 stability and the subcellular localization of Skp2. Our findings pro-vide a mechanistic explanation for elevated Skp2 expression as well as Skp2 cytoplasmic staining in tissues derived from advanced breast and prostate cancers21,22.

ReSulTSSkp2 expression is regulated by the PI(3)K/Akt pathwayRecent reports have suggested that the PI(3)K/Akt pathway regulates Skp2 expression levels by one or more unknown mechanisms6,23. To investigate the contribution of Akt signalling in Skp2 expression, we first treated HeLa and PC3 cells with the PI(3)K inhibitor LY294002, and found a time-dependent decrease in Skp2 protein levels concomi-tant with a robust inhibition of PI(3)K activity as revealed by the loss of phospho-Akt (pS473). However, the expression of Cdh1, the known E3 ligase of Skp2, was not affected by LY294002. In addition, the expression of other Cdh1 substrates such as cyclin A did not respond to inhibition of PI(3)K activity (Fig. 1a; Supplementary Information, Fig. S1a–c). Second, we used insulin-like growth factor-1 (IGF-1), which potently activates PI(3)K in all cell types, and observed increased Skp2 protein levels also concomitantly with enhanced phosphorylation by Akt (Supplementary Information, Fig. S1d, e). Figure 1b shows that specific depletion of Akt1, but not that of Akt2, markedly decreases Skp2 protein levels in HeLa cells and induces its downstream target p27. However, depletion of Akt1 did not change the expression of Cdh1 and other Cdh1 substrates that we examined (Fig. 1b). Similar results were also obtained in U2OS cells

1Department of Pathology, Beth Israel Deaconess Medical Center, Harvard Medical School, Boston, Massachusetts 02215, USA.2Correspondence should be addressed to W.W. (e-mail: [email protected])

Received 26 August 2008; accepted 3 December 2008; published online 8 March 2009; corrected online 18 March 2009; DOI:10.1038/ncb1847




A RT I C L E S

and SKBR3 cells (Supplementary Information, Fig. S1f, g). Conversely, inactivation of PTEN, which results in elevated Akt activity, leads to upregulation of Skp2 in both asynchronized and synchronized HeLa cells (Fig. 1c, d). This finding is further supported by the positive correlation between Skp2 expression and Akt activity in a panel of breast cancer cell lines (Fig. 1e). Furthermore, the suppression of Akt activity by LY294002 in both MDA-MB468 and SKBR3 cells leads to downregulation of Skp2 expression, providing further evidence that elevated Akt activity is one major cause of the observed upregulation of Skp2 in these two cell lines (Supplementary Information, Fig. S1b, c). Thus, in agreement with pre-vious reports6,8, the PI(3)K pathway regulates Skp2 expression; moreover, this occurs selectively through Akt1 signalling.

Our finding that inactivation of PTEN in mouse embryonic fibrob-lasts (MEFs) led to mild upregulation (1.5–1.7-fold) of mouse Skp2 levels

(Supplementary Information, Fig. S1h) is consistent with a previous report8. However, this is not likely to operate through the Akt pathway, because downregulation of either Akt1 or Akt2 by short hairpin RNA (shRNA) in MEFs did not affect mouse Skp2 protein levels (Supplementary Information, Fig. S1i). These results indicate that Akt signalling differs in mouse cells and human cells with respect to regulation of Skp2 expression.

The PTeN/PI(3)K/Akt pathway regulates both Skp2 transcription and Skp2 stabilityNext, we examined how Akt regulates Skp2 expression mechanistically. We found that, in agreement with previous reports24,25, inactivation of Akt1, but not that of Akt2, leads to a 40% decrease in Skp2 mRNA levels (Fig. 2a). This is possibly through either E2F1 or NF-κB pathways that are subjected to regulation by Akt1 (refs 24, 25). However, we observed

a c

0 3 6 9 12 24

b

0 4 8 12 16 20 0 4 8 12 16 20

MC

F10A

MC

F7

SK

BR

3M

DA

-MB

468d e

LY294002h

IB: anti-Skp2 (monoclonal Ab) IB: anti-Skp2 (monoclonal Ab)

IB: anti-Skp2 (monoclonal Ab)

IB: anti-PTEN

IB: anti-pS473 Akt

IB: anti-Cdh1

IB: anti-Cdc20

siRNA treatments

siRNA treatments

Nocodazole releaseh

IB: anti-Skp2 (monoclonal Ab)

IB: anti-Skp2

IB: anti-pS473 Akt

IB: anti-Akt

IB: anti-PTEN

IB: anti-cyclin E

IB: anti-p27

IB: anti-PTEN

siRNA treatments

IB: anti-pS473 Akt

IB: anti-AktIB: anti-Akt1

IB: anti-Akt2IB: anti-cyclin A

IB: anti-cyclin A

IB: anti-cyclin B

IB: anti-cyclin E

IB: anti-cyclin A

IB: anti-cyclin B

IB: anti-cyclin E

IB: anti-Plk-1

IB: anti-cyclin A

IB: anti-cyclin EIB: anti-p27

IB: anti-p27

IB: anti-Cdh1

IB: anti-Cdh1

IB: anti-Cdc20

IB: anti-Cdh1

IB: anti-cyclin E

IB: anti-Cdk2

IB: anti-Cdk2

IB: anti-pS807.pS811.Rb

IB: anti-Cdk2IB: anti-tubulin

IB: anti-tubulin

IB: anti-tubulin

IB: anti-tubulin

IB: anti-tubulin

Moc

kC

ontr

olLu

cife

rase

Akt

1A

kt2

Akt

1+2

Luci

fera

se

Luciferase PTEN-1

PTE

N-1

PTE

N-2

PTE

N-3

Figure 1 Human Skp2 protein levels are regulated by the PTEN/PI(3)K/Akt pathway. (a) Immunoblot (IB) analysis of HeLa cells treated with the PI(3)K inhibitor LY294002 (20 µM) for the indicated durations. Ab, antibody. (b, c) Immunoblot analysis of HeLa cells transfected with the indicated siRNA oligos. The control lane is scrambled E2F-1 siRNA. Luciferase, siRNA against firefly luciferase; PTEN-1, PTEN-2, PTEN-3, three independent

PTEN siRNA oligos. (d) Immunoblot analysis of HeLa cells transfected with the indicated siRNA oligos, after synchronization with nocodazole and release. (e) Immunoblot analysis of the indicated cell lines cultured in serum-free medium. Whole-cell lysates were isolated in the presence of phosphatase inhibitors. Full-length blots are provided in Supplementary Information, Fig. S11.



A RT I C L E S

a much greater (about eightfold) decrease in Skp2 protein abundance (Fig. 2b), arguing that Akt1 could also regulate Skp2 expression post-translationally. To address this possibility further, we assessed the half-life of endogenous Skp2 half-life after modulating the PTEN/Akt pathway. We found that inactivation of Akt1 by short interfering RNA (siRNA) shortened the half-life of endogenous Skp2, whereas depletion of PTEN, in a similar fashion to depletion of Cdh1, stabilized Skp2 (Fig. 2c, d; Supplementary Information, Fig. S2).

Akt1 interacts with and phosphorylates Skp2 at Ser 72Sequence analysis revealed that human Skp2 contains an Akt consen-sus phosphorylation site at Ser 72, which conforms to the optimal Akt motif RxRxxS/T. The motif surrounding Ser 72 is also conserved in Skp2 orthologues in all mammals except mouse (Fig 3a). We therefore reasoned that Skp2 is a substrate of Akt, whose phosphorylation may influence Skp2 stability. However, because mouse Skp2 lacks the Ser 72 site, it is likely that mouse Skp2 is not an optimal Akt substrate. This explains why a loss of Akt1 would not affect Skp2 expression in MEFs (Supplementary Information, Fig. S1i). However, the absence of Ser 72 in Xenopus and zebrafish suggests that the Akt/Skp2 regulatory pathway might be a relatively late event acquired during evolution.

Consistent with the hypothesis that Skp2 is a putative Akt substrate, an activated allele of Akt1 (myr-Akt1) interacts with Skp2 as detected by co-immunoprecipitation (Fig. 3b). Furthermore, when overexpressed

in 293T cells, Skp2 specifically interacts with endogenous Akt1 but not Akt2 (Fig. 3c, 3d). In support of this finding, using both in vivo co-immunoprecipitation (Fig. 3e) and in vitro glutathione S-transferase (GST) pulldown (Fig. 3f) assays, we were able to show that Skp2 inter-acts specifically with Akt1 but not Akt2. Moreover, we showed that the Skp2 constructs lacking the first 90 amino-acid residues failed to interact with both overexpressed Akt1 (Fig. 3g) and endogenous Akt1 (Fig. 3h). These results provide further evidence of the molecular mech-anism for the Akt1 isoform-specific regulation of Skp2.

To test whether Skp2 is a substrate for Akt, we first performed in vitro kinase assays. Figure 4a shows that Akt1 phosphorylates wild-type Skp2, but not the S72A mutant. Furthermore, we show that Akt could phospho-rylate Skp2 as efficiently as it does another known Akt substrate, Mdm2 (ref. 26) (Supplementary Information, Fig. S3a). We also examined Skp2 phosphorylation by using a substrate-directed phospho-specific antibody that recognizes the optimal Akt consensus phosphorylation motif27 (Fig. 4b). Because the Scansite program showed that there are additional suboptimal putative Akt phosphorylation sites present in human Skp2, including Thr 21, Ser 75 and Ser 157, to pinpoint the exact Akt phosphorylation site we performed mass spectrometric analysis of GST–WT (wild-type) Skp2 after incubation with myr-Akt. This analysis revealed that Ser 72 is the only phosphorylation event identified under these experimental conditions (Supplementary Information, Fig. S3b). Next, we found that expression of activated Akt significantly enhanced

a b

0 1 2 h4 7 0 1 2 4 7 0 1 2 4 7

c

0 1 2 4 7

d

0

–6

–5

–4

–3

–2

–1

0

1

0 2 4 6 8

Luc

Akt

PTEN

Cdh1

Rel

ativ

e S

kp2

mR

NA

leve

l

Rel

ativ

e S

kp2

pro

tein

leve

l

siRNA: Mock Luc Akt1 Akt2 siRNA: Mock Luc Akt1 Akt2

1.6

1.4

1.2

1.0

0.8

0.6

0.4

0.2

0

1.6

1.4

1.2

1.0

0.8

0.6

0.4

0.2

siRNA: Luciferase Akt1 PTEN Cdh1

CHX:

IB: anti-Skp2

IB: anti-Akt1

IB: anti-Cdh1

IB: anti-PTEN

IB: anti-tubulin

Ban

d in

tens

ity (L

og2)

CHX treatment (h)

Figure 2 The PTEN/PI(3)K/Akt pathway regulates both Skp2 transcription and Skp2 stability. (a) Real-time RT–PCR analysis to examine the relative Skp2 mRNA expression levels in HeLa cells transfected with the indicated siRNA oligonucleotides. Luc, siRNA against firefly luciferase. Results are shown as means ± s.d. for three independent sets of experiments. (b) Immunoblot analysis to examine the relative Skp2 protein levels in HeLa cells transfected with the indicated siRNA oligos. Results are shown as means ± s.d. for three independent sets of experiments. (c) HeLa cells

were transfected with indicated siRNA oligos. At 40 h after transfection, cells were treated with 20 µg ml−1 cycloheximide (CHX). At the indicated time points, whole-cell lysates were prepared and immunoblots were probed with indicated antibodies. (d) Quantification of the band intensities in c. Skp2 band intensity was normalized to tubulin, then normalized to the t=0 controls. Results are shown as means ± s.d. for three independent sets of experiments. Full-length blots are provided in Supplementary Information, Fig. S11.



A RT I C L E S

the phosphorylation of wild-type Skp2, whereas Skp2-S72A phospho-rylation was not detected (Fig. 4c). The reactivity of Skp2 with the Akt substrate antibody was reversed when the cell lysates were incubated with lambda phosphatase (Fig. 4d). In addition, Skp2 phosphorylation was detected by the Akt substrate antibody in IGF-1-stimulated cells; moreo-ver, phosphorylation was decreased in cells transduced with Skp2 siRNA or Akt1, but not Akt2, shRNA (Fig. 4e; Supplementary Information, Fig. S3f). Although both Akt1 and Akt2 phosphorylated Skp2 at relatively similar efficiencies in vitro (Supplementary Information, Fig. S3c), when overexpressed in 293T cells, Akt1 is more potent than Akt2 in phospho-rylating the Skp2 protein (Supplementary Information, Fig. S3d). These

results are therefore consistent with the notion that endogenous Akt1, but not Akt2, directly phosphorylates Skp2 in cells, and suggests that this may be causally linked to decreased Skp2 expression after Akt1 deple-tion (Fig. 1b). It was recently shown that, besides Akt, other AGC-family kinases such as serum and glucocorticoid-inducible kinase (SGK) could also phosphorylate p27, a well-known Akt substrate28. We found that Akt is the only kinase capable of phosphorylating Skp2 in vivo: both S6 kinase (S6K) and SGK failed to phosphorylate Skp2 (Fig. 4f).

High-stringency Akt sites are not found in mouse Skp2, although there are several suboptimal sites, including Thr 21, Ser 75 and Ser 133. This low stringency is consistent with the failure of the phospho-Akt substrate

– ++– ++

a b c

d

+++

+––+ ––

e f

WT

S72

A∆9

0

EV

WT

S72

A∆9

0

EV

h

+–+ +–++– +

+–+

––– –– –g

Human (63–80)Chimpanzee (63–80)Macaque (63–80) Dog (63–80) Cow (63–80) Horse (63–80) Pig (63–80) Rat (63–80)Mouse (63–80)Akt consensus14-3-3 motif

Flag–Skp2HA–myr-Akt

HA–myr-Akt2HA–myr-Akt1Flag–WT.Skp2

IP: anti-Flag

WCL

IP: anti-Flag

WCL

IB: anti-Flag

IB: anti-Flag

IP: anti-Flag

IB: anti-Flag

IB: anti-Flag

IB: anti-HA

IB: anti-HA

IB: anti-HA

IB: anti-Flag

IB: anti-HA

IB: anti-Flag

IB: anti-HA

IB: anti-HA

IP: a

nti-H

A

IB: anti-Cdk2

IB: anti-Cdk2

IB: anti-Cdk2

IB: anti-Cdk2

IB: anti-Cyclin A

IB: anti-LSD1

IB: IgG light chain

IB: anti-Akt1

IB: anti-Akt1

IB: anti-Flag

IP: anti-Flag

WCL

WCL

Transfection: EV

HA

–Skp

2

EV

HA

–Skp

2

WC

L

IP: a

nti-H

A

WC

L

Transfection: EV

HA

–Skp

2

EV

HA

–Skp

2

IB: anti-Cdk2

IB: anti-LSD1

IB: anti-Akt2

IB: anti-Akt2

GST proteins: GS

TG

ST–

WT.

Skp

2

GST–Skp2GST

1% in

put

GS

TG

ST–

WT.

Skp

21%

inp

ut

Akt1

Bound IN Bound IN

Akt2

Flag–∆90.Skp2Flag–WT.Skp2HA–myr-Akt1

IB: anti-HA (Akt1)

IB: anti-HA (Akt1)

Flag–Skp2 constructs

Figure 3 The Skp2 protein contains a canonical Akt phosphorylation site at Ser 72 and interacts with Akt1, but not with Akt2, in vivo. (a) Sequence alignment of the putative Akt phosphorylation site at Ser 72 in Skp2 from different species. (b) Immunoblot (IB) analysis of whole-cell lysates (WCL) and immunoprecipitates (IP) derived from 293T cells transfected with HA–myr-Akt and Flag–Skp2 constructs. (c, d) Immunoblot analysis of whole-cell lysates and immunoprecipitates derived from 293T cells transfected with HA–Skp2 or empty vector (EV). Cdk2 and cyclin A antibodies were used as positive controls to detect the interaction with Skp2, and the LSD1 antibody was used as a negative control. (e) Immunoblot (IB) analysis of whole-cell lysates and

immunoprecipitates derived from 293T cells transfected with the indicated HA–myr-Akt1 or HA–myr-Akt2 and Flag–Skp2 constructs. (f) Autoradiography of 35S-labelled Akt1 or Akt2 bound to the indicated GST fusion proteins. IN, input. (g) Immunoblot analysis of whole-cell lysates and immunoprecipitates derived from 293T cells transfected with the indicated HA–myr-Akt1 and Flag–Skp2 constructs. (h) Immunoblot analysis of whole-cell lysates and immunoprecipitates derived from 293T cells transfected with the indicated Flag–Skp2 construct or empty vector. Cdk2 antibody was used as a positive control to detect the interaction with Skp2. Full-length blots are provided in Supplementary Information, Fig. S11.



A RT I C L E S

antibody to recognize phosphorylation at any of these sites in mouse Skp2 (Supplementary Information, Fig. S3e). Furthermore, we observed only a very weak incorporation of γ-32P into a mouse GST–Skp2 fusion protein after incubation with myr-Akt in vitro (Fig. 4g).

Phosphorylation of the Skp2 Ser 72 site by Akt1 triggers subsequent phosphorylation of the Ser 75 site by casein kinase ITo gain a better understanding of the phosphorylation events of human Skp2 protein in vivo, we immunoprecipitated the ectopically expressed hae-magglutinin (HA)–Skp2 protein and analysed its phosphorylation status by mass spectroscopy. As illustrated in Fig. 5a, we identified phosphorylation of Ser 72, Ser 75 and the previously reported Ser 64 (ref. 29; Supplementary

Information, Fig. S4a, b). Sequence analysis revealed that Ser 75 is a putative casein kinase I (CKI) site. Consistent with this, the CKI ε isoform was found to associate specifically with Skp2 protein by mass spectroscopy (data not shown). This finding was further validated by immunoblot analysis of Skp2 immunoprecipitates (Fig. 5b). In addition, inhibition of the CKI kinase activ-ity with specific inhibitors resulted in a decrease in Skp2 expression level in both HeLa and U2OS cell lines (Fig. 5c), indicating that CKI is involved in regulating Skp2 stability by directly phosphorylating the Skp2 protein.

We reasoned that phosphorylation of Ser 72 might create a priming site that facilitates phosphorylation of Ser 75 by CKI, a mechanism that has been reported for phosphorylation of the Foxo family of transcrip-tion factors30 (Fig. 5d). Indeed, we found that replacement of Ser 72 by

–+ –+– – ++

– – ++

– –– –

– –

––

–

–

++ –+ –+a b

e

– + – + – + –+ –+– – ++

– – ++

– –– –

– –

––

–

–

++ –+ –+gf

dc –+ –+– – ++– +–+

–+ –+– – ++

– +–+

+–

+

+–

–– –++– –

HA–myr-Akt

HA–myr-Akt

HA–myr-Akt

GST–WT.Skp2

GST–Skp2

GST–S72A.Skp2GST–S72A.S75A.Skp2

–+ –+– – ++

– – ++

– –– –

– –

––

–

–

++ –+ –+HA–myr-AktGST–WT.Skp2

GST–S72A.Skp2GST–S72A.S75A.Skp2

32P-GST–Skp2 IB: anti-HA

IB: anti-GST

IB: anti-Akt substrate(RxRXXpS/pT)IB: anti-Akt substrate(R/KxR/KXXpS/pT)

Flag–WT.Skp2

IP: anti-Skp2IB: anti-Skp2

IB: anti-Skp2

IB: anti-Akt1IB: anti-Akt2

IB: anti-pS473.Akt

IB: anti-actin

Flag–S72A.Skp2

IP: anti-Flag

WCL

WCLWCL

IP: anti-Flag

WCL

IB: anti-Akt substrate

IB: anti-Flag IB: anti-Akt substrate

IB: anti-Flag

IB: anti-Flag

IB: anti-Flag

IB: anti-Flag

IP: anti-Flag

IB: anti-HA (Akt)

IB: anti-HA

IB: anti-HA

IB: anti-HA

IB: IgG heavy chain

IB: IgG heavy chain

IgG heavy chain

IB: anti-Flag

Flag–WT.Skp2Flag–S72A.Skp2

HA–Myr-Akt

HA–Myr.Akt

HA–Myr-Akt

GST–Skp2

λ-phosphatase

Control Akt1 Akt2 shRNAIGF-1IB: anti-Akt substrate

IB: 32P-GST–Skp2

GST–Mouse.WT.Skp2GST–Human.S72A.Skp2

GST–Human.WT.Skp2H

A–S

GK

.CA

HA

–S6K

.KD

HA

–S6K

.CA

HA

–Myr

.Akt

1E

V

anti-Aktsubstrate

Figure 4 Akt phosphorylates the human Skp2 protein at Ser 72. (a) Akt phosphorylates Skp2 in vitro at Ser 72. HA–myr-Akt was transfected into 293T cells, recovered by anti-HA immunoprecipitation and incubated with 5 µg of indicated GST–Skp2 in the presence of [γ-32P]ATP. The kinase reaction products were resolved by SDS–PAGE, and phosphorylation was detected by autoradiography. (b) Akt phosphorylates Skp2 in vitro at Ser 72. HA–myr-Akt was transfected into 293T cells, recovered by anti-HA immunoprecipitation and incubated with 5 µg of the indicated GST–Skp2 in the presence of unlabelled ATP. The kinase reaction products were resolved by SDS–PAGE, and phosphorylation was detected by the phospho-Akt substrate antibody that recognizes either the RXRXXpS/pT or the R/KXR/KXXpS/pT motif. IB, immunoblot. (c) Immunoblot analysis of whole-cell lysates (WCL) and immunoprecipitates (IP) derived from HeLa cells transfected with HA–myr-Akt and Flag–Skp2. WT, wild-type Skp2; S72A, Skp2 mutated at the Akt site. (d) Immunoblot analysis of whole-cell lysates

and immunoprecipitates derived from 293T cells transfected with HA–myr-Akt and the indicated Flag–Skp2 plasmids. Where indicated, the whole-cell lysates were treated with λ-phosphatase before immunoprecipitation. (e) Immunoblot analysis of whole-cell lysates and immunoprecipitates derived from HeLa cells infected with Akt1 and Akt2 lentiviral shRNA. Endogenous Skp2 was immunoprecipitated with anti-Skp2 and immunoblotted with the Akt substrate-directed phospho-antibody. (f) Immunoblot analysis of whole-cell lysates and immunoprecipitates derived from HeLa cells transfected with the indicated kinases and the Flag–Skp2 construct. (g) Mouse Skp2 protein is a poor Akt substrate. HA–myr-Akt was transfected into 293T cells, recovered by anti-HA immunoprecipitation and incubated with 5 µg of indicated GST–Skp2 in the presence of [γ-32P]ATP. The kinase reaction products were resolved by SDS–PAGE, and phosphorylation was detected by autoradiography. Full-length blots are provided in Supplementary Information, Fig. S11.



A RT I C L E S

Asp to mimic phosphorylation by Akt enhanced phosphorylation of this S72D-Skp2 mutant by CKI (Fig. 5e; Supplementary Information, Fig. S4c). In contrast, mutation of Ser 75 reduced phosphorylation, indicating that CKI-mediated phosphorylation of Skp2 occurs at Ser 75 (Fig. 5f). These data suggest that CKI may function as the Ser 75 kinase after a priming phosphorylation of Ser 72 by Akt1.

Overexpression of Akt protects Skp2 from Cdh1-mediated destructionOur previous studies suggested that a region of Skp2 between residues 46 and 90 (refs 4, 5), which contains both the Akt and CKI phosphoryla-tion sites, is both sufficient and required for interaction with Cdh1. We therefore further examined how Akt and/or CKI phosphorylation affects

a

p p p

Human Skp2 (59–80)

b

Transfection: EV

HA

–Skp

2

EV

HA

–Skp

2

IB: anti-casein kinase I δ

IB: anti-casein kinase I ε

IB: anti-cyclin A

IB: anti-HA

IB: anti-TSC1

IB: IgG light chain

IP: anti-HA WCL

c

DM

SO

D44

76, 5

0 µM

D44

76, 1

00 µ

MIC

261,

50

µMIC

261,

125

µM

DM

SO

D44

76, 5

0 µM

D44

76, 1

00 µ

MIC

261,

50

µMIC

261,

125

µM

HeLa U2OS

CKI inhibitors

Cell lines

IB: anti-Skp2

IB: anti-Akt1

IB: anti-Cdh1

IB: anti-cyclin A

IB: anti-tubulin

dHuman Foxo3a (305–323)

Human Foxo1a (309–327)

Human Foxo4 (249–267)

Human Skp2 (63–80)

Chimpanzee Skp2 (63–80)

Macaque Skp2 (63–80)

–+ –+

– – ++

– – ++

– –

– –

– –

–

–

–

–

+ + –+ –+

– – ++– ––––

–

–

–

––

–

–

+e

CKI

GST

GST–WT.Human.Skp2

GST–S72D.Human.Skp2

GST–WT.Mouse.Skp2

32P-GST–Skp2

Non-specific proteinco-purified with CKI

CKI

GST–Skp2GST

–+ –+

– – ++

– – ++

– –

– –

– –

–

–

–

–

+ + –+ –+

– – ++– ––––

–

–

–

––

–

–

+– –+ – –– –––

+

–

––

––

f

CKI

GSTGST–WT.Skp2

GST–S72D.Skp2

GST–S75D.Skp2

GST–S72D.S75D.Skp2

32P-GST–Skp2

Non-specific proteinco-purified with CKI

CKI

GST–Skp2GST

Figure 5 Phosphorylation of Skp2 at Ser 72 triggers the subsequent phosphorylation of Ser 75 by CKI. (a) In vivo Skp2 phosphorylation sites detected by mass spectroscopy. (b) Immunoblot (IB) analysis of whole-cell lysates (WCL) and immunoprecipitates (IP) derived from 293T cells transfected with HA–Skp2. The anti-cyclin A antibody was used as a positive control to detect the interaction with Skp2, and the TSC1 antibody was used as a negative control. (c) Immunoblot analysis of HeLa and U2OS cells treated with the CKI inhibitors D4476 and IC261 at the indicated concentrations for

12 h. DMSO, dimethylsulphoxide. (d) Schematic representation of the Skp2 protein sequence showing the Ser 75 and Ser 72 sites. The same sequential phosphorylation cascade has been described in the Foxo family of transcription factors. (e, f) CKI phosphorylates Skp2 in vitro at Ser 75. Purified CKI protein (from New England Biolabs) was incubated with 5 µg of indicated GST–Skp2 in the presence of γ-32P-ATP. The kinase reaction products were resolved by SDS–PAGE, and phosphorylation was detected by autoradiography. Full-length blots are provided in Supplementary Information, Fig. S11.



A RT I C L E S

the interaction between Cdh1 and Skp2. We find that replacement of both Ser 72 and Ser 75 by phospho-mimetic amino acids disrupted the interaction between Skp2 and Cdh1 as detected by both in vitro GST pull-down assays (Fig. 6a; Supplementary Information, Fig. S4d) and in vivo co-immunoprecipitation analysis (Supplementary Information, Fig. S4e).

In support of this finding, we further demonstrated that overexpression of activated Akt resulted in decreased interaction between Skp2 and Cdh1 in vivo (Fig. 6b). Next, we asked whether this leads to the stabilization of Skp2. In keeping with previous reports4,5, we found that expression of Cdh1 downregulated wild-type Skp2, whereas overexpression of activated Akt

a

– +–+– +

– +–+– +

b

c

–+ –+– – ++

d

– +–

+– +

0 4 8 12 16

e

0 4 8 12 16 0 4 8 12 16 0 4 8 12 16 0 4 8 12 16

5% in

put

GS

T

GS

T–W

T.S

kp2

GS

T–S

72D

.Skp

2

GS

T–S

72D

.S75

D.S

kp2

GS

T–S

75D

.Skp

2

GST proteins

Cdh1

Flag–WT.Skp2

Flag–WT.Skp2

siRNA

siRNA

Nocodazole release

IB: anti-Skp2 (monoclonal)

IB: anti-Skp2 (polyclonal)

IB: anti-p27 (short exposure)

IB: anti-p27 (long exposure)

IB: anti-Cdh1

IB: anti-Akt1

IB: anti-Akt2

IB: anti-cyclin B

IB: anti-Plk-1

IB: anti-Cdc20

IB: anti-cyclin A

IB: anti-cyclin E

IB: anti-Cdk2

IB: anti-tubulin

h

HA–Myr-Akt

IB: anti-Cdh1

IB: anti-Flag

IB: anti-Cdh1 IB: anti-Cdh1

IB: anti-Plk1

IB: anti-GFP

IB: anti-tubulin

IB: anti-GFP

IB: anti-tubulin

IB: anti-Flag

IB: anti-Flag

IB: anti-HA

IP: anti-Flag

WCL

IN Bound

GST.Skp2

GST

Flag–S72D.S75D.Skp2

WT

S72

D.S

75D

Flag.Skp2 Constructs

HA.Myr.AktHA.Cdh1

IB: anti-FlagIB: anti-HA (Akt)

IB: anti-HA (Cdh1)

Luc

Cd

h1

Luc Akt1 Cdh1+Akt1Akt2 Cdh1

Luc

Cd

h1

Figure 6 Phosphorylation of Skp2 by Akt1 protects Skp2 from Cdh1-mediated destruction. (a) Autoradiography of 35S-labelled Cdh1 bound to the indicated GST fusion proteins. IN, input. (b) Immunoblot analysis of whole-cell lysates (WCL) and immunoprecipitates (IP) derived from 293T cells transfected with the Flag–Skp2 construct in the presence or absence of HA–myr-Akt. (c) Immunobot analysis of HeLa cells transfected with the indicated Flag–Skp2 and HA–Cdh1 plasmids in the presence or absence of HA–myr-Akt. A plasmid encoding GFP was used as a negative

control for transfection efficiency. (d) Immunoblot analysis of HeLa cells transfected with the indicated Flag–Skp2 plasmids and siRNA oligonucleotidess. A plasmid encoding GFP was included as negative control for transfection efficiency. (e) Immunoblot analysis of HeLa cells transfected with the indicated siRNA oligonucleotides, synchronized by growth in nocodazole, and then released for the indicated periods. The Akt1 and Akt2 samples, and the Cdh1 and Cdh1+Akt1 samples were run on individual gels.



A RT I C L E S

abolished Cdh1-mediated Skp2 destruction. Phospho-mimetic mutants of Ser 72 and Ser 75 (S72D and S75D) were resistant to Cdh1-mediated destruction (Fig. 6c). Conversely, depletion of endogenous Cdh1 with

siRNA upregulated wild-type Skp2 but not the Skp2 S72D/S75D phospho-mimetic mutant, further supporting the idea that the S72D/S75D mutant is resistant to Cdh1-mediated destruction (Fig. 6d).

–1

–3

–2

0

0 2 4 6 8

0 1 2 4 7 0 1 2 4 7 0 1 2 4 7 CHX

IB: anti-Flag

IB: anti-GFP

IB: anti-tubulin

IB: anti-tubulin

a b

0 4 8 12 16 0 4 8 12 16 0 4 8 12 16 0 4 8 12 16 0 4 8 12 16 0 4 8 12 16 h

h

c d

0

0.2

0.4

0.6

.0.8

1.0

1.2

1.4

0 4 8 12 16

EV

+– +– +–

WT

WT

AA

AA

∆90

∆NLS

S72

A

S64

A

S64

D

+–

e

g

f

WT S72A

WTS72A

S72D.S75D

WT WT+Akt S72A S72A+Akt S72D.S75D S72D.S75D+Akt

S72D.S75D

CHX treatment (h)

Rel

ativ

e S

kp2

exp

ress

ion

leve

l

WTWT+AktS72AS72A+AktS72D/S75D

S72D/S75D+Akt

Nocodazole release (h)IB: anti-tubulin

IB: anti-GFP

IB: anti-HA (Akt)

IB: anti-geminin

IB: anti-Cdh1

IB: anti-cyclin B

IB: anti-Flag

Nocodazole release

IP: anti-Flag

WCL

IP: anti-Flag

WCL

IP: anti-Flag

WCL

Con

trol

Cyc

lin A

Cd

k2A

kt1

Flag–Skp2 Constructs

HA–Myr-Akt

IB: anti-Akt substrate

IB: anti-Flag

IB: anti-Flag

IB: anti-Flag

IB: anti-FlagIB: anti-cyclin A

IB: anti-cyclin A

IB: anti-cyclin A

IB: anti-Cdk2

IB: anti-Cdk2

IB: anti-Cdk2

IB: anti-Cdh1

IB: anti-Cdh1

IB: anti-Akt1

IB: anti-pS473.Akt

IB: anti-Flag

IB: anti-Akt substrateIB: anti-Akt substrate

Flag–Skp2 constructssiRNA

IB: anti-HA

IB: anti-Flag

Ban

d in

tens

ity (L

og2)

Figure 7 Phosphorylation of Skp2 at Ser 72 by Akt affects Skp2 protein stability. (a) HeLa cells were transfected with the indicated Flag–Skp2 plasmids. At 20 h after transfection, cells were split into 60-mm dishes, and after a further 20 h they were treated with 20 µg ml−1 cycloheximide (CHX). At the indicated time points, whole-cell lysates were prepared and immunoblots (IB) were probed with the indicated antibodies. (b) Quantification of the band intensities in a. Skp2 band intensity was normalized to GFP, then normalized to the t = 0 controls. Results are shown as means ± s.d. for three independent sets of experiments. (c) Immunoblot analysis of HeLa cells transfected with limiting amounts of Flag–Skp2 (wild-type; S72A; S72D/S75D) plasmids with or without HA–myr-Akt, along with a green fluorescent protein (GFP) as transfection control. HeLa cells were synchronized in M phase

with nocodazole, and then released into G1 for the indicated time periods. (d) Quantification of the band intensities in c. Skp2 band intensity was normalized to tubulin, then normalized to the t = 0 control of wild-type Skp2. (e) Immunoblot analysis of whole-cell lysates (WCL) and immunoprecipitates (IP) derived from HeLa cells transfected with the Flag–Skp2 construct together with the indicated siRNA oligos. (f) Immunoblot analysis of whole-cell lysates (WCL) and immunoprecipitates (IP) derived from 293T cells transfected with the indicated Flag–Skp2 constructs. (g) Immunoblot analysis of whole-cell lysates and immunoprecipitates derived from HeLa cells transfected with the indicated Flag–Skp2 constructs in the presence or absence of HA–myr-Akt. Full-length blots are provided in Supplementary Information, Fig. S11.



A RT I C L E S

Cdh1 activity is required for Akt1-dependent regulation of Skp2 levelsAlthough Cdh1 is the only E3 ligase identified so far that targets Skp2, Akt1-mediated Skp2 regulation could occur through either Cdh1-dependent or Cdh1-independent mechanisms. To further test the contribution of Cdh1 in Akt-dependent regulation of Skp2, we used siRNA. Figure 6e shows that depletion of Akt1, but not that of Akt2, leads to a decrease in Skp2 protein levels and subsequent accumulation of p27. Conversely, depletion of Cdh1 enhanced upregulation of Skp2 and downregulation of p27. Skp2 levels were restored to normal control levels when both Cdh1 and Akt1 were depleted concomitantly. This result suggests that Cdh1 is required for the ability of Akt1 to regulate Skp2. Thus, in normal cycling cells, the ability of Akt1 to phosphor-ylate Skp2 at Ser 72 protects Skp2 from Cdh1-mediated degradation,

such that loss of Akt1 leads to enhanced Skp2 degradation primarily through the Cdh1-dependent destruction pathway. Also consistent with this was our finding that in T98G cells released from serum star-vation, Skp2 expression showed up earlier than most APC–Cdh1 sub-strates (data not shown)31,32.

Phosphorylation of Skp2 at Ser 72 by Akt affects Skp2 protein stabilityTime-course experiments in cells expressing Skp2 mutants and treated with the protein synthesis inhibitor cycloheximide revealed that the effects of Akt on Skp2 protein levels are due to alterations in the half-life of the Skp2 protein (Fig. 7a, b). To further investigate the effects of Akt1 activity on Skp2 protein levels, various Skp2 constructs were transfected with or without activated myr-Akt and their expression

a

b

c

d

+–C N C N

+–C N C N

+–C N C N

Cell linesLY294002

IB: Anti-Skp2

IB: Anti-SP1

IB: Anti-IKKα

e

GS

T

5% in

put

GS

T–W

T.S

kp2

GS

T–∆N

LS.S

kp2

GS

T–S

72D

.Skp

2G

ST–

S72

D.S

75D

.Skp

2G

ST–

S75

D.S

kp2

f g h

Akt consensusHuman p27 (148–166) Human p21 (136–153)Human Foxo1 (247–264)Human Skp2 (63–80)Rat Skp2 (63–80) SV40 T NLS (126–132)

WT.Skp2

WT.Skp2

S72A.Skp2

WT.Skp2+ HA–Myr-Akt

S72D.Skp2

S72A.Skp2+ HA–Myr-Akt

S72D.Skp2+ HA–Myr-Akt

DMSO

LY294002

∆NLS.Skp2

Anti-Flag DAPI Merge

DAPI MergeAnti-Skp2 Anti-HA

Anti-Skp2 DAPI Merge

Percentage of cells in the cytoplasm



7.4%

64.2%

19.1%

96.5%

10.9%

47.7%

5.5%

8.1%

34.8%

35.3%

HeLa MDA-468 SKBR-3

GST proteins: GS

T

5% in

put

GS

T–W

T.S

kp2

GS

T–∆N

LS.S

kp2

GS

T–S

72D

.Skp

2G

ST–

S72

D.S

75D

.Skp

2G

ST–

S75

D.S

kp2

GST proteins: GS

T

5% in

put

GS

T–W

T.S

kp2

GS

T–∆N

LS.S

kp2

GS

T–S

72D

.Skp

2G

ST–

S72

D.S

75D

.Skp

2G

ST–

S75

D.S

kp2

GST proteins:

IN Bound IN Bound IN Bound

Importin α1

GST–Skp2GST

Importin α5

GST–Skp2GST

Importin α7

GST–Skp2GST

Figure 8 Akt phosphorylation of Skp2 promotes its cytoplasmic translocation. (a) Sequence alignment of Skp2 with p27, p21 and Foxo1 NLS. (b) Immunofluorescence and 4,6-diamidino-2-phenylindole (DAPI) staining of HeLa cells transfected with Flag-tagged wild-type or ∆NLS Skp2 plasmids. Scale bars, 20 µm. (c) Immunofluorescence and DAPI staining of HeLa cells transfected with indicated Flag–Skp2 in the presence or absence of HA–myr-Akt constructs. Scale bars, 20 µm.

(d) Immunofluorescence and DAPI staining of SKBR3 cells treated with LY294002 (or dimethylsulphoxide (DMSO) as negative control) for 12 h. Scale bars, 20 µm. (e) Immunoblot analysis of nuclear (N) and cytoplasmic (C) fraction of HeLa, MDA-MB468 and SKBR3 cells treated with LY294002 (or DMSO as a negative control) for 12 h. (f–h) Autoradiography of 35S-labelled importin α1 (f), importin α5 (g) and importin α7 (h) bound to the indicated GST fusion proteins. IN, input.



A RT I C L E S

levels were monitored during cell-cycle progression. As shown in Fig. 7c, d, wild-type Skp2 protein levels decline rapidly in early G1 when APC–Cdh1 is active, and expression of activated Akt delayed the rate of degradation. The S72A mutant was degraded significantly more rapidly than wild-type Skp2. Conversely, the S72D/S75D mutant was degraded with slower kinetics than wild-type Skp2. Taken together, these data demonstrate that phosphorylation by Akt influences the destruction of Skp2 governed by APC–Cdh1.

As mentioned above, using mass spectrometry we found that in addition to phosphorylation at Ser 72 and Ser 75, the Ser 64 site of Skp2 is phospho-rylated in vivo (Fig. 5a), and the cyclin A–Cdk2 complex was implicated as the kinase responsible29. In agreement with a recent report32, we found that phosphorylation of Skp2 by cyclin A–Cdk2 at Ser 64 also stabilizes Skp2 (Supplementary Information, Fig. S5a, b). However, phosphoryla-tion of Ser 64 did not affect the interaction between Skp2 and Cdh1 in vitro (Supplementary Information, Fig. S5c) or in vivo (Supplementary Information, Fig. S4e), indicating that cyclin A–Cdk2 affects Skp2 protein stability through a different mechanism from that of Akt.

Because both cyclin A–Cdk2 and Akt affect Skp2 stability, we next sought to investigate the potential connection between these two kinases. We found that neither overexpression of HA–myr-Akt nor depletion of Akt significantly affects cyclin A or Cdk2 expression levels or their kinase activity (Supplementary Information, Fig. S6a, b). In contrast, inactivation of either cyclin A and Cdk2 leads to a significant decrease in Akt activity, as illustrated by the decrease in pS473 Akt signals (Supplementary Information, Fig. S6c, d) and the decreased efficiency in phosphorylating Skp2 (Fig. 7e). This indicates that cyclin A–Cdk2 could execute their function partly through activation of the Akt kinase. However, Akt can efficiently phosphorylate a Skp2 mutant (AAAA.Skp2) that fails to interact with the cyclin A–Cdk2 complex33 (Fig. 7f) as well as Skp2 mutants in which Ser 64 is replaced with either Ala or Asp (S64A.Skp2 and S64D.Skp2) (Fig. 7g). This indicates that phosphorylation of Skp2 by Akt is independent of the phosphorylation event occurring at Ser 64 of Skp2.

Akt phosphorylation of Skp2 promotes its cytoplasmic translocationAkt has been reported to have a major function in the cellular localization of many of its substrates, including p21, p27 and Foxo1 (refs 14–17, 34). Amino-acid alignment of human Skp2 with known Akt substrates revealed that Skp2 also contains a putative nuclear localization sequence (NLS) and that Ser 72 is located within this NLS (Fig. 8a). Immunofluorescence exper-iments revealed that the bulk of cellular Skp2 is localized in the nucleus in dividing cells normally. However, a mutant Skp2 whose putative NLS had been deleted (∆NLS.Skp2) was localized predominantly in the cytoplasm (Fig. 8b; Supplementary Information, Fig. S8a, b). Previous studies have also indicated that, at least for p21, p27 and Foxo, phosphorylation of serine and/or threonine residues adjacent to the NLS leads to a mask-ing of the NLS mediated by the binding of 14-3-3. Similarly, we and the Pandolfi group (page XXX of this issue) found that the interaction between 14-3-3 and Skp2 was enhanced by activated myr-Akt (Supplementary Information, Fig. S7a, b). Furthermore, we found that the ability to inter-act with 14-3-3 is greatly decreased in the S72A and ∆NLS Skp2 mutants and that the Skp2 mutant lacking the first 90 amino-acid residues failed to interact with 14-3-3 at all (Supplementary Information, Fig. S7c, d). It has been reported previously that phosphorylation of BAD at Ser 136 by

Akt promotes its interaction with 14-3-3, disrupting its interaction with Bcl-XL35. However, we found that blockage of Akt-induced 14-3-3 interac-tion with Skp2 by the R18 peptide36 does not affect the interaction between Skp2 and Cdh1 (Supplementary Information, Fig. S7e, f).

There are also documented examples in which phosphorylation of ser-ine or threonine residues near or within the NLS reduces the interaction between the NLS and the importin protein complex, thus affecting nuclear import37,38. Using both immunofluorescence microscopy and cellular fractionation, we found that expression of activated myr-Akt promotes the cytoplasmic localization of wild-type Skp2 (Fig. 8c; Supplementary Information, Fig. S7g). In contrast, the non-phosphorylatable Skp2.S72A mutant is restricted to the nuclear compartment in the presence of activated Akt. Conversely, a significant fraction of the phospho-mimetic Skp2.S72D, Skp2.S72D.S75A and Skp2.S72D.S75D mutants is located in the cytoplasm, and their localization is unaffected by the expression of activated Akt (Fig. 8c; Supplementary Information, Fig. S9). Furthermore, we found that there is a significant pool of cytoplasmic Skp2 in SKBR3 cells, which harbour elevated Akt activity (Fig. 1e), and that inhibition of Akt activity by LY294002 results in the translocation of Skp2 into the nucleus (Fig. 8d, e). We found that the Skp2 protein specifically interacts with the nuclear import receptors importin α5 and α7 but not α1, and that phosphorylation of the Ser 72 site by Akt is sufficient to disrupt the interaction between Skp2 and importin α5 and α7 (Fig. 8f–h). Because the importin complex has a critical function in transporting proteins into the nucleus, we reasoned that the dissociation of Skp2 from the importin complex would retain Skp2 in the cytoplasm. Because the Ser 64 site is very close to the putative NLS, we also investigated the potential effects of Ser 64 phosphorylation on Skp2 cellular localization. In agreement with a previous report32, we found that the phosphorylation status of Ser 64 did not affect the cellular localization of Skp2 (Supplementary Information, Fig. S8c). We further demonstrated that whereas phosphorylation of Skp2 by Akt at Ser 72 abolished the interaction between Skp2 and the importin complex, phosphorylation of Ser 64 was not sufficient to disrupt the interaction between Skp2 and the importin complex (Supplementary Information, Fig. S10a–c).

DISCuSSIONThe data presented above provide evidence for a novel mechanism by which Akt1-mediated phosphorylation of Skp2 at Ser 72 protects Skp2 from Cdh1-mediated destruction through disruption of the interac-tion between Skp2 and its E3 ligase Cdh1, as well as by inducing the cytoplasmic translocation of Skp2. The Ser 72 phosphorylation site on human Skp2 is not present in the mouse sequence. Similar inter-species differences have been reported for other Akt substrates, including p27 (ref. 17) and caspase-9 (ref. 39). However, the Ser 72 site is conserved in most large mammals (Fig. 3a). It is plausible that for larger animals with a longer life span than mice, cell-cycle control is more stringent, illustrated by the additional layer of Akt regulation on Skp2 stability.

For most SCF–F-box complexes, the regulation of substrate recog-nition occurs at the level of the substrate, whereas the interaction of Cdh1 and Cdc20 with their substrates does not usually require any post-translational modifications40. Our finding provides another unique mechanism for the selective degradation of Cdh1 downstream targets. This protective mechanism mediated by the Akt pathway is very similar to that of the Cdk2–cyclin E complex, which protects Cdc6 from Cdh1-mediated destruction41.



A RT I C L E S

The Akt pathway functions to promote both cell survival and cell growth by inactivating many of its downstream substrates9. For p27, p21 and Foxo proteins, phosphorylation by Akt triggers the recruitment of 14-3-3, which results in masking the NLS followed by cytoplasmic trans-location37. We also observed an enhanced interaction of 14-3-3 with Skp2 in cells expressing activated Akt. Moreover, phosphorylation of Skp2 by Akt at Ser 72 greatly decreases the interaction between Skp2 and importin. It is possible that both of these mechanisms contribute to the cytoplasmic translocation of Skp2 after phosphorylation by Akt42,43. Thus our results offer a molecular mechanism for the cytoplasmic localization of Skp2, which has been observed in many clinical tumour samples and is cor-related with aggressive malignancy and poor diagnosis3,21,22,44.

Our data point to Akt isoform specificity in the regulation of Skp2 protein stability (Fig. 4). Furthermore, we demonstrated that when overexpressed in 293T cells, human Skp2 interacts specifically with endogenous Akt1 but not with Akt2 (Fig. 3c, d), although the precise mechanism by which Akt1 can, whereas Akt2 cannot, signal to Skp2 has yet to be defined.

Taken together, our results provide new insight into how Akt activity could influence the Skp2/p27 pathway, which is a known hotspot for mutations in human cancer. On one level, our findings provide a mecha-nism by which Akt influences cell-cycle progression. On another level, we offer a new mechanism by which Akt affects the order of degradation of specific APC–Cdh1 substrates. Ultimately, these data may provide the rationale for the development of specific Akt1 inhibitors as efficient anti-cancer drugs.

Note added in proof: a related manuscript by Lin et al. (Nature Cell Biol. 11, doi:10.1038/ncb1849; 2009) is also published in this issue.

MeTHODSPlasmids. Flag–Skp2, HA–Skp2 and HA–myr-Akt1 plasmids were described previ-ously5,45. HA–myr-Akt2 plasmid was purchased from Addgene. The first 90 residues of human Skp2 protein were fused in frame with the GST protein to create the pGEX.WT.human.Skp2 construct. Mouse Skp2 cDNA was amplified from a mouse cDNA library (a gift from Ronald DePinho) with Pfu polymerase (Stratagene). Full-length mouse Skp2 cDNA was subcloned into the pCMV–Flag vector (Sigma) to create the Flag–mouse.Skp2 construct, and the first 90 residues of mouse Skp2 were fused in frame with the GST protein to create the pGEX.WT.mouse.Skp2 construct. Skp2 mutants were generated with the QuikChange XL Site-Directed Mutagenesis Kit (Stratagene). The HA–Cdh1 construct was obtained from Peter Jackson. The HA–S6K.CA and HA–S6K.KD constructs were obtained from John Blenis. The HA–SGK.CA construct was a gift from Suzanne Conzen. The importin α1, importin α5 and impor-tin α7 plasmids were obtained from the DF/HCC DNA Resource Core.

Antibodies and reagents. Anti-Akt antibody (9272), anti-Akt2 antibody (5B5), anti-phospho-Akt antibody (4051) and anti-phospho-Akt substrate (9614) were purchased from Cell Signaling. Anti-p27 antibody (SC-528), polyclonal anti-HA antibody (SC-805), polyclonal anti-Skp2 antibody (SC-7164), anti-cyclin A anti-body (SC-751), anti-cyclin B antibody (SC-245), anti-Cdc20 antibody (SC-8358), anti-CKI δ antibody (SC-6473), anti-14-3-3β antibody (SC-629), anti-Geminin antibody (SC-13015), anti-Plk1 antibody (SC-17783), anti-cyclin E antibody (SC-247), anti-SP1 antibody (SC-59), anti-IKK-α antibody (SC-7184) and anti-TSC1 antibody (SC-13013) were purchased from Santa Cruz. Anti-tubulin antibody (T-5168), polyclonal anti-Flag antibody (F2425), monoclonal anti-Flag antibody (F-3165), peroxidase-conjugated anti-mouse secondary antibody (A4416) and peroxidase-conjugated anti-rabbit secondary antibody (A4914) were purchased from Sigma. Monoclonal anti-HA antibody (MMS-101P) was purchased from Covance. Anti-(green fluorescent protein) (anti-GFP) antibody (632380), mono-clonal anti-Skp2 antibody (32-3400) and polyclonal anti-Cdh1 antibody (34-2000) were purchased from Invitrogen. Monoclonal anti-Cdh1 (CC43) was purchased from Oncogene. Anti-CKI ε antibody (AP7403a) was purchased from Abgent.

Polyclonal anti-Akt1 isoform-specific antibody was produced in house by immunizing rabbits with a synthetic peptide (VDSERRPHFPQFSYSASGTA). Oligofectamine, Lipofectamine and Plus reagent were purchased from Invitrogen. Recombinant human IGF-1 was purchased from R&D Systems.

siRNAs. Human Akt1 siRNA oligonucleotide (sense, 5´-GAGUUUGAGUACCUGAAGCUGUU-3´) and human Akt2 siRNA oli-gonucleotide (sense, 5´-GCGUGGUGAAUACAUCAAGACUU-3´) have been validated previously20 and were purchased from Dharmacon, or sequences were cloned into the pLKO lentiviral expression system and virus was gener-ated in 293T cells for infection, as described20. Mouse Akt1 and mouse Akt2 siRNA oligonucleotides were validated by Laura Benjamin’s laboratory (per-sonal communication), and sequences were cloned into the pLKO lentiviral expression system. PTEN-1 (sense, 5´-AGGCACAAGAGGCCCUAGA-3´), PTEN-2 (sense, 5´-AAGAGGAUGGAUUCGACUUAG-3´) and PTEN-3 (sense, 5´-AUCGUUAGCAGAAACAAAAGG-3´) have been validated previously6,46 and were purchased from Dharmacon. Luciferase GL2 siRNA oligo was purchased from Dharmacon, and the Cdh1 siRNA oligo has been described previously5. Cdk2, cyclin E and cyclin A siRNA oligos have been described previously47. As described previously, siRNA oligos were transfected into subconfluent cells with Oligofectamine or Lipofectamine 2000 (Invitrogen) in accordance with the manu-facturer’s instructions5,48.

Cell culture and cell synchronization. Cell culture, including synchronization and transfection, have been described previously5. Where indicated, the PI(3)K inhibitor LY294002 (Sigma) or cycloheximide (Sigma) were added to the cell culture media. INK4a−/− mouse embryonic fibroblasts (MEFs) and INK4a−/−.PTENloxp/loxp MEFs were a gift from Ronald DePinho.

Immunoblots and immunoprecipitation. Cells were lysed in EBC (50 mM Tris-HCl pH 8.0, 120 mM NaCl, 0.5% Nonidet P40) buffer supplemented with protease inhibitors (Complete Mini; Roche) and phosphatase inhibitors (phos-phatase inhibitor cocktail set I and II; Calbiochem). The protein concentrations of the lysates were measured with the Bio-Rad protein assay reagent on a Beckman Coulter DU-800 machine. The lysates were then resolved by SDS–PAGE and immunoblotted with the indicated antibodies. For immunoprecipitation, 800 µg of lysates were incubated with the appropriate antibody (1–2 µg) for 3–4 h at 4 °C followed by incubation for 1 h with Protein A-Sepharose beads (GE Healthcare). Immune complexes were washed five times with NETN buffer (20 mM Tris-HCl pH 8.0, 100 mM NaCl, 1 mM EDTA, 0.5% Nonidet P40) before being resolved by SDS–PAGE and immunoblotted with the indicated antibodies. Quantification of the immunoblot band intensity was performed with ImageJ software.

Skp2 binding assays. Binding to immobilized GST proteins was performed as described previously5.

Cellular fractionation. The NE-PER kit (Pierce) was used to perform cellular fractionation in accordance with the manufacturer’s instructions. Buffers were supplemented with both protease inhibitor (Roche) and phosphatase inhibitors (Calbiochem).

Real-time RT–PCR analysis. RNA was extracted with a Qiagen RNeasy mini kit, and the reverse transcription reaction was performed with the ABI Taqman Reverse Transcriptional Reagents (N808-0234). After mixing the resulting tem-plate with Skp2 (Hs00180634-m1) or glyceraldehyde-3-phosphate dehydrogenase (GAPDH; Hs99999905-m1) primers and ABI Taqman Fast Universal PCR Master Mix (4352042), the real-time reverse transcriptase (RT)–PCR reaction was per-formed with the ABI-7500 Fast Real-time PCR system.

Indirect immunofluorescence microscopy. Cells grown on coverslips were fixed in 4% paraformaldehyde and permeabilized with 0.2% Triton X-100. The cells were stained with polyclonal anti-HA antibody (Santa Cruz) and monoclonal anti-Skp2 antibody (Invitrogen) in blocking buffer (3% BSA in PBS) for 30 min, and then rinsed and incubated with secondary Alexa Fluor 594-conjugated anti-mouse antibody and Alexa Fluor 488-conjugated anti-rabbit antibody (Invitrogen) for 1 h. Cells were then rinsed with PBS, stained with 4,6-diamidino-2-phenylindole (DAPI) and mounted. The slides were examined with a fluorescence microscope (Eclipse TE300; Nikon) and digital image analysis software (IPLab; Scanalytics).



A RT I C L E S

Protein degradation analysis. Cells were transfected with a plasmid encoding a Flag-tagged version of the protein of interest along with a plasmid encoding GFP as a negative control. For half-life studies, cycloheximide (20 µg ml−1; Sigma) was added to the medium 40 h after transfection. At various time points thereafter, cells were lysed and protein abundances were measured by immunoblot analysis. Where indicated, HA–Cdh1 and/or HA–myr-Akt constructs were co-transfected into the cells to examine their effects on the abundance of the protein of interest.

In vitro kinase assay. 293T cells were transfected with HA–myr-Akt. After 48 h, Akt was immunoprecipitated with HA-matrix (Roche). It was then incubated with 5 µg of GST–Skp2 proteins (wild type or S72A mutant) in the presence of 5 µCi of [γ-32P]ATP and 20 µM unlabelled ATP in the Akt kinase reaction buffer for 15–30 min. The reaction was stopped by the addition of SDS-containing lysis buffer, resolved by SDS–PAGE and detected by autoradiography. CKI was pur-chased from New England Biolabs. The CKI in vitro kinase assays were performed in accordance with the manufacturer’s instructions. The Cdk2 kinase assay was performed as described previously47.


ACknoWleDGemenTsWe thank William Kaelin Jr, Lewis Cantley, Roya Khosravi-Far and Susan Glueck for critical reading of the manuscript; James DeCaprio, Christoph Geisen, Ronald DePinho, Laura Benjamin, Suzanne Conzen, John Blenis and Peter Jackson for providing reagents; Ross Tomaino for his kind assistance on the mass spectrum analysis; Isaac Robinovitz for his technical support on the fluorescence microscopy; Pier Paolo Pandolfi for sharing unpublished data; and members of the Wei and Toker laboratories for useful discussions. W.W. is a Leukemia and Lymphoma Society Special Fellow, Kimmel Scholar and V Scholar. This work was supported in part by the Harvard Medical School Milton Fund (W.W.) and the Emerald Foundation, and by grants from the National Institutes of Health (W.G.K., CA076120; A.T., CA122099) and the Susan G. Komen Breast Cancer Foundation (R.Y.C., 0706963).

AuTHoR ConTRIbuTIonsD.G. and H.I. performed most of the experiments with the assistance of A.Tseng. R.Y.C. performed the shAkt1 and shAkt2 experiment to examine its effects on Skp2 Ser 72 phosphorylation. W.W. and A.T. designed the experiments. W.W. supervised the study. W.W. wrote the paper with the assistance of A.T. All authors commented on the manuscript.

CompeTInG fInAnCIAl InTeResTsThe authors declare no competing financial interests.


1. Cardozo, T. & Pagano, M. The SCF ubiquitin ligase: insights into a molecular machine. Nature Rev. Mol. Cell Biol. 5, 739–751 (2004).

2. Gstaiger, M. et al. Skp2 is oncogenic and overexpressed in human cancers. Proc. Natl Acad. Sci. USA 98, 5043–5048 (2001).

3. Signoretti, S. et al. Oncogenic role of the ubiquitin ligase subunit Skp2 in human breast cancer. J. Clin. Invest. 110, 633–641 (2002).

4. Bashir, T., Dorrello, N. V., Amador, V., Guardavaccaro, D. & Pagano, M. Control of the SCF(Skp2-Cks1) ubiquitin ligase by the APC/C(Cdh1) ubiquitin ligase. Nature 428, 190–193 (2004).

5. Wei, W. et al. Degradation of the SCF component Skp2 in cell-cycle phase G1 by the anaphase-promoting complex. Nature 428, 194–198 (2004).

6. van Duijn, P. W. & Trapman, J. PI3K/Akt signaling regulates p27kip1 expression via Skp2 in PC3 and DU145 prostate cancer cells, but is not a major factor in p27kip1 regulation in LNCaP and PC346 cells. Prostate 66, 749–760 (2006).

7. Andreu, E. J. et al. BCR-ABL induces the expression of Skp2 through the PI3K pathway to promote p27Kip1 degradation and proliferation of chronic myelogenous leukemia cells. Cancer Res. 65, 3264–3272 (2005).

8. Mamillapalli, R. et al. PTEN regulates the ubiquitin-dependent degradation of the CDK inhibitor p27KIP1 through the ubiquitin E3 ligase SCFSKP2. Curr. Biol. 11, 263–267 (2001).

9. Woodgett, J. R. Recent advances in the protein kinase B signaling pathway. Curr. Opin. Cell Biol. 17, 150–157 (2005).

10. Parsons, R. Human cancer, PTEN and the PI-3 kinase pathway. Semin. Cell Dev. Biol. 15, 171–176 (2004).

11. Datta, S. R. et al. Akt phosphorylation of BAD couples survival signals to the cell-intrinsic death machinery. Cell 91, 231–241 (1997).

12. Zhao, X. et al. Multiple elements regulate nuclear/cytoplasmic shuttling of FOXO1: characterization of phosphorylation- and 14-3-3-dependent and -independent mecha-nisms. Biochem. J. 378, 839–849 (2004).

13. Brunet, A. et al. Akt promotes cell survival by phosphorylating and inhibiting a Forkhead transcription factor. Cell 96, 857–868 (1999).

14. Zhou, B. P. et al. Cytoplasmic localization of p21Cip1/WAF1 by Akt-induced phospho-rylation in HER-2/neu-overexpressing cells. Nature Cell Biol. 3, 245–252 (2001).

15. Viglietto, G. et al. Cytoplasmic relocalization and inhibition of the cyclin-dependent kinase inhibitor p27Kip1 by PKB/Akt-mediated phosphorylation in breast cancer. Nature Med. 8, 1136–1144 (2002).

16. Liang, J. et al. PKB/Akt phosphorylates p27, impairs nuclear import of p27 and opposes p27-mediated G1 arrest. Nature Med. 8, 1153–1160 (2002).

17. Shin, I. et al. PKB/Akt mediates cell-cycle progression by phosphorylation of p27Kip1 at threonine 157 and modulation of its cellular localization. Nature Med. 8, 1145–1152 (2002).

18. Whiteman, E. L., Cho, H. & Birnbaum, M. J. Role of Akt/protein kinase B in metabolism. Trends Endocrinol. Metab. 13, 444–451 (2002).

19. Heron-Milhavet, L. et al. Only Akt1 is required for proliferation, while Akt2 promotes cell cycle exit through p21 binding. Mol. Cell. Biol. 26, 8267–8280 (2006).

20. Irie, H. Y. et al. Distinct roles of Akt1 and Akt2 in regulating cell migration and epithe-lial–mesenchymal transition. J. Cell Biol. 171, 1023–1034 (2005).

21. Drobnjak, M. et al. Altered expression of p27 and Skp2 proteins in prostate cancer of African-American patients. Clin. Cancer Res. 9, 2613–2619 (2003).

22. Radke, S., Pirkmaier, A. & Germain, D. Differential expression of the F-box proteins Skp2 and Skp2B in breast cancer. Oncogene 24, 3448–3458 (2005).

23. Shapira, M., Kakiashvili, E., Rosenberg, T. & Hershko, D. D. The mTOR inhibitor rapamycin down-regulates the expression of the ubiquitin ligase subunit Skp2 in breast cancer cells. Breast Cancer Res. 8, R46 (2006).

24. Reichert, M., Saur, D., Hamacher, R., Schmid, R. M. & Schneider, G. Phosphoinositide-3-kinase signaling controls S-phase kinase-associated protein 2 transcription via E2F1 in pancreatic ductal adenocarcinoma cells. Cancer Res. 67, 4149–4156 (2007).

25. Barre, B. & Perkins, N. D. A cell cycle regulatory network controlling NF-κB subunit activity and function. EMBO J. 26, 4841–4855 (2007).

26. Zhou, B. P. et al. HER-2/neu induces p53 ubiquitination via Akt-mediated MDM2 phosphorylation. Nature Cell Biol. 3, 973–982 (2001).

27. Zhang, H. et al. Phosphoprotein analysis using antibodies broadly reactive against phosphorylated motifs. J. Biol. Chem. 277, 39379–39387 (2002).

28. Hong, F. et al. mTOR-raptor binds and activates SGK1 to regulate p27 phosphorylation. Mol. Cell 30, 701–711 (2008).

29. Yam, C. H., Ng, R. W., Siu, W. Y., Lau, A. W. & Poon, R. Y. Regulation of cyclin A-Cdk2 by SCF component Skp1 and F-box protein Skp2. Mol. Cell. Biol. 19, 635–645 (1999).

30. Tran, H., Brunet, A., Griffith, E. C. & Greenberg, M. E. The many forks in FOXO’s road. Sci. STKE 2003, RE5 (2003).

31. Carrano, A. C., Eytan, E., Hershko, A. & Pagano, M. SKP2 is required for ubiquitin-medi-ated degradation of the CDK inhibitor p27. Nature Cell Biol. 1, 193–199 (1999).

32. Rodier, G., Coulombe, P., Tanguay, P. L., Boutonnet, C. & Meloche, S. Phosphorylation of Skp2 regulated by CDK2 and Cdc14B protects it from degradation by APCCdh1 in G1 phase. EMBO J. 27, 679–691 (2008).

33. Ji, P. et al. Skp2 contains a novel cyclin A binding domain that directly protects cyclin A from inhibition by p27Kip1. J. Biol. Chem. 281, 24058–24069 (2006).

34. Vogt, P. K., Jiang, H. & Aoki, M. Triple layer control: phosphorylation, acetylation and ubiquitination of FOXO proteins. Cell Cycle 4, 908–913 (2005).

35. Chiang, C. W. et al. Protein phosphatase 2A dephosphorylation of phosphoserine 112 plays the gatekeeper role for BAD-mediated apoptosis. Mol. Cell. Biol. 23, 6350–6362 (2003).

36. Wang, B. et al. Isolation of high-affinity peptide antagonists of 14-3-3 proteins by phage display. Biochemistry 38, 12499–12504 (1999).

37. Poon, I. K. & Jans, D. A. Regulation of nuclear transport: central role in development and transformation? Traffic 6, 173–186 (2005).

38. Fabbro, M. & Henderson, B. R. Regulation of tumor suppressors by nuclear-cytoplasmic shuttling. Exp. Cell Res. 282, 59–69 (2003).

39. Fujita, E. et al. Akt phosphorylation site found in human caspase-9 is absent in mouse caspase-9. Biochem. Biophys. Res. Commun. 264, 550–555 (1999).

40. Harper, J. W., Burton, J. L. & Solomon, M. J. The anaphase-promoting complex: it’s not just for mitosis any more. Genes Dev. 16, 2179–2206 (2002).

41. Mailand, N. & Diffley, J. F. CDKs promote DNA replication origin licensing in human cells by protecting Cdc6 from APC/C-dependent proteolysis. Cell 122, 915–926 (2005).

42. Hennekes, H., Peter, M., Weber, K. & Nigg, E. A. Phosphorylation on protein kinase C sites inhibits nuclear import of lamin B2. J. Cell Biol. 120, 1293–1304 (1993).

43. Zhang, F., White, R. L. & Neufeld, K. L. Cell density and phosphorylation control the subcellular localization of adenomatous polyposis coli protein. Mol. Cell. Biol. 21, 8143–8156 (2001).

44. Dowen, S. E., Scott, A., Mukherjee, G. & Stanley, M. A. Overexpression of Skp2 in carcinoma of the cervix does not correlate inversely with p27 expression. Int. J. Cancer 105, 326–330 (2003).

45. Yoeli-Lerner, M. et al. Akt blocks breast cancer cell motility and invasion through the transcription factor NFAT. Mol. Cell 20, 539–550 (2005).

46. Hamada, K. et al. The PTEN/PI3K pathway governs normal vascular development and tumor angiogenesis. Genes Dev. 19, 2054–2065 (2005).

47. Zhang, G. J. et al. Bioluminescent imaging of Cdk2 inhibition in vivo. Nature Med. 10, 643–648 (2004).

48. Wei, W., Jin, J., Schlisio, S., Harper, J. W. & Kaelin, W. G. Jr. The v-Jun point mutation allows c-Jun to escape GSK3-dependent recognition and destruction by the Fbw7 ubiquitin ligase. Cancer Cell 8, 25–33 (2005).






A RT I C L E S

Protein kinase DYRK2 is a scaffold that facilitates assembly of an E3 ligaseSubbareddy Maddika1 and Junjie Chen1,2

Protein kinases have central functions in various cellular signal transduction pathways through their substrate phosphorylation. Here we show that a protein kinase, DYRK2, has unexpected role as a scaffold for an E3 ubiquitin ligase complex. DYRK2 associates with an E3 ligase complex containing EDD, DDB1 and VPRBP proteins (EDVP complex). Strikingly, DYRK2 serves as a scaffold for the EDVP complex, because small-interfering-RNA-mediated depletion of DYRK2 disrupts the formation of the EDD–DDB1–VPRBP complex. Although the kinase activity of DYRK2 is dispensable for its ability to mediate EDVP complex formation, it is required for the phosphorylation and subsequent degradation of its downstream substrate, katanin p60. Collectively, our results reveal a new type of E3-ubiquitin ligase complex in humans that depends on a protein kinase for complex formation as well as for the subsequent phosphorylation, ubiquitylation and degradation of their substrates.

DYRK2 is a member of an evolutionarily conserved family of dual-specificity tyrosine-phosphorylation-regulated kinases (DYRKs) that belongs to the CMGC group of protein kinases1,2. During protein syn-thesis, DYRK2 autophosphorylates a tyrosine residue in its own activa-tion loop. Once autophosphorylated at this tyrosine residue, DYRK2 loses its tyrosine kinase activity and functions only as a serine/threonine kinase3. DYRK2 phosphorylates a very limited number of substrates such as NFAT4, eIFB5, Glycogen synthase6, Oma-1 (ref. 7), MEI-1 (ref. 8) and the chromatin remodelling factors SNR1 and TRX9, thus regulating cal-cium signalling, protein synthesis, glucose metabolism, developmental processes and gene expression. Recently, DYRK2 has also been suggested to function in the DNA damage signalling pathway by phosphorylating p53 at serine 46 in the nucleus and promoting cellular apoptosis after genotoxic stress10. In addition to its role in cellular responses and devel-opmental processes, DYRK2 is a potential oncogene11, because DYRK2 amplification and overexpression have been reported in adenocarcino-mas of the oesophagus and lung12. However, the exact mechanism of DYRK2 in tumorigenesis remains to be clarified.

RESultSDYRK2 associates with EDVP E3 ligase complexIn an attempt to elucidate DYRK2 function further, we established 293T derivative cell line stably expressing a triple-epitope (S-protein, Flag and streptavidin-binding peptide)-tagged version of DYRK2 (SFB–DYRK2). Tandem affinity purification with streptavadin-agarose beads and S-protein-agarose beads followed by mass spectrometry analysis allowed us to discover several proteins that interacted with DYRK2 (Fig. 1a and Supplementary Information, Table S1). Among them we repeatedly

identified EDD, DDB1 and VPRBP as major DYRK2-associated pro-teins (Fig. 1a). EDD (also known as UBR5, hHYD or KIAA0896) is an E3 ligase with a distinct amino-terminal UBA domain, a UBR box and a carboxy-terminal HECT domain that mediates ubiquitin-dependent protein degradation13,14. EDD is likely to be involved in tumorigene-sis, because an allelic imbalance at the EDD locus has been reported in several cancers15,16. DDB1 (DNA-damage binding protein 1)17 is an adaptor subunit of the Cul4–Roc1 E3 ligase complex18 that mediates the ubiquitin-dependent degradation of various substrates including Cdt1, p21Cip1/WAF1 and c-Jun. VPRBP (also known as DCAF1)19,20, a WD40-domain-containing protein, is a substrate recognition subunit of the DDB1–Cul4A–Roc1 complex.

By transient overexpression of SFB–DYRK2 in 293T cells, we con-firmed the interaction of DYRK2 in vivo with EDD, DDB1 and VPRBP (Fig. 1b). Although DDB1 and VPRBP have been discovered recently as key components in the Cul4–Roc1 E3 ligase complex18,21,22, surprisingly we did not identify either Cul4 or Roc1 in our purification. Indeed, we could not detect any interaction of overexpressed DYRK2 with either Cul4A or Roc1 (Fig. 1b), confirming that Cul4 and Roc1 are not compo-nents of this novel complex, which contains DYRK2, EDD, DDB1 and VPRBP. We further confirmed the existence of this complex in vivo by demonstrating that endogenous DYRK2 co-immunoprecipitated with EDD, DDB1 and VPRBP (Fig. 1c). In contrast, Cul4A–Roc1 components were not seen in EDD immunoprecipitates (Fig. 1c). However, neither EDD nor DYRK2 was seen in Cul4A immunoprecipitates, supporting the notion that the presence of the EDVP complex is independent of the Cul4A–Roc1 complex (Fig. 1c). The interactions between EDD, DDB1 and VPRBP with DYRK2 are specific, because we could only observe

1Department of Therapeutic Radiology, Yale University School of Medicine, PO Box 208040, New Haven, Connecticut 06520, USA.2Correspondence should be addressed to J.C. (e-mail: [email protected])

Received 22 October 2008; accepted 4 December 2008; published online 15 March 2009; DOI:10.1038/ncb1848




A RT I C L E S

these associations in cells transfected with control small interfering RNA (siRNA) but not in cells after transfection with DYRK2-specific siRNA (Supplementary Information, Fig. S1a). In addition, exogenously

expressed Myc–EDD interacted only with Flag–DYRK2 but not with another DYRK family member, DYRK1B (Supplementary Information, Fig. S1b, c), underlining the specificity of the interaction between DYRK2

Anti-EDD

Ant

i-Fl

ag

Ant

i-Ig

G

a b

c

SFB–DYRK2

75

50

250

EDD 53

DDB1 20

DYRK2 17

VPRBP 1326

Anti-VPRBP

Anti-DDB1

Anti-Cul4A

Anti-Flag(DYRK2)

Anti-EDD

Anti-VPRBP

Anti-DDB1

Anti-Cul4A

Anti-GST

GS

T

GST pulldown

GS

T–D

YR

K2

Ant

i-D

DB

1

Ant

i-V

PR

BP

Anti-EDD

Anti-VPRBP Anti-DDB1

Anti-DYRK2 Anti-DYRK2

Anti-DYRK2

d

Anti-Roc1

Anti-Roc1

Anti-Cul4A

Ant

i-C

ul4A

Anti-Cul4A

Anti-DYRK2

Anti-EDD

Anti-Roc1

Mar

ker

DYR

K2 c

ompl

ex

Mr(K)

DYRK2 complex proteins

Protein Peptide no.

Inp

ut (5

%)

IP

IP IP

IPIP

Ant

i-Ig

G

Ant

i-Ig

GA

nti-

IgG

Ant

i-E

DD

Inp

ut (5

%)

Inp

ut (5

%)

Inp

ut (5

%)

Ant

i-Ig

G

Inp

ut (5

%)

Inp

ut (5

%)

Figure 1 Identification of EDD–DDB1–VPRBP as DYRK2-associated proteins. (a) Tandem affinity purification of DYRK2-containing protein complexes was conducted with 293T cells stably expressing triple-tagged DYRK2. Associated proteins were separated by SDS–PAGE and revealed by staining with Coomassie blue. The proteins and the number of peptides identified by mass spectrometry analysis are shown in the table at the right and also in Supplementary Information, Table S1. (b) Immunoprecipitation (IP) with control IgG or anti-Flag (DYRK2) antibody were performed with extracts prepared from 293T derivative cells stably expressing Flag-tagged DYRK2. The presence of EDD, DDB1, VPRBP, Cul4A or Roc1 in these immunoprecipitates was evaluated by

immunoblotting with their respective antibodies. (c) Reverse co-immunoprecipitation experiments were performed with anti-EDD, anti-Cul4A, anti-DDB1 and anti-VPRBP antibodies and the associated endogenous DYRK2 and other indicated proteins was identified by western blotting with their respective antibodies. (d) A GST pulldown assay was performed with immobilized control GST or GST–DYRK2 fusion proteins on agarose beads, followed by incubation with extracts prepared from 293T cells. The interaction of EDD, DDB1, VPRBP or Cul4A with DYRK2 was assessed by immunoblotting with their respective antibodies. Uncropped images of blots are shown in Supplementary Information, Fig. S4.



A RT I C L E S

and EDD. Bacterially expressed glutathione S-transferase (GST)-tagged DYRK2 pulled down EDD, DDB1 and VPRBP, but not Cul4A from cell extracts (Fig. 1d), again arguing that DYRK2 forms a distinct complex with EDD, DDB1 and VPRBP. The formation of a DYRK2–EDD–DDB1–VPRBP complex might not be strictly regulated by the cell cycle, because we observed an interaction of DYRK2 with the other components of the complex independently of cell cycle phases, although the levels of EDD and VPRBP interacting with DYRK2 vary and are proportional to their protein levels at specific phases of the cell cycle (Supplementary Information, Fig. S2c).

EDD is a known HECT-domain-containing E3 ligase that regulates the ubiquitin-dependent degradation of its substrates14. We named this E3 ligase complex containing EDD, DDB1 and VPRBP proteins the EDVP complex, to distinguish it from the previously identified Cul4–Roc1–DDB1–VPRBP E3 ligase complex. To assess the significance of the interaction between DYRK2 and this new EDVP E3 ligase complex, we checked DYRK2 protein levels in cells depleted of EDD, DDB1 or

VPRBP. We found no difference between the DYRK2 protein levels in any of the knockdown cells in comparison with control-siRNA-trans-fected cells (Supplementary Information, Fig. S2a). We also checked the protein levels of DYRK2, EDD and VPRBP at different phases of the cell cycle. Whereas the levels of EDD and VPRBP fluctuated during the cell cycle, DYRK2 levels remained constant (Supplementary Information, Fig. S2b), thus ruling out the possibility that DYRK2 may be a target for proteasomal degradation mediated by the EDVP ligase complex.

DYRK2 acts as an adaptor in the EDVP complexSeveral Kelch-motif-containing proteins were shown previously to act as E3 ligase adaptors for specific substrates23–25. A preliminary analysis of the DYRK2 protein sequence revealed the presence of a Kelch motif (amino-acid residues 390–433) within its protein kinase catalytic domain (Supplementary Information, Fig. S3), so we next investigated whether DYRK2 functions as a molecular adaptor in the EDVP ligase complex. We depleted DYRK2 with siRNA and checked for complex formation

a c

IP

Anti- EDD

Anti-DDB1

Control siRNA DYRK2 siRNA

Anti-EDD

Anti-VPRBP

Anti-VPRBP

Anti-DYRK2

Anti-Cul4A

b

Anti-EDD

Anti-EDD

DYRK2 WT DYRK2 KD

Anti-Flag(DYRK2)

0

15

30

Mea

n C

.P.M

. (x

1,00

0)

Con

trol

DY

RK

2 si

RN

A

DY

RK

2 si

RN

A +

SiR

-DY

RK

2 W

T

DY

RK

2 si

RN

A +

SiR

-DY

RK

2 K

D

Anti-actin

Anti-DYRK2

Con

trol

DYRK2 siRNA

SiR

-DY

RK

2 W

T

SiR

-DY

RK

2 K

D

–

Ant

i-D

DB

1

Ant

i-Ig

G

Inp

ut (5

%)

IP

IP IP

IP IP

Ant

i-D

DB

1

Ant

i-Ig

G

Inp

ut (5

%)

Ant

i-V

PR

BP

Ant

i-Ig

GA

nti-

IgG

Inp

ut (5

%)

Inp

ut (5

%)

Ant

i-V

PR

BP

Ant

i-Ig

G

Inp

ut (5

%)

Ant

i-E

DD

Ant

i-Ig

G

Inp

ut (5

%)

Ant

i-E

DD

IP IP

IP IP

Ant

i-D

DB

1

Ant

i-Ig

G

Inp

ut (5

%)

Ant

i-D

DB

1

Ant

i-Ig

G

Inp

ut (5

%)

Ant

i-Ig

G

Inp

ut (5

%)

Ant

i-Ig

G

Inp

ut (5

%)

Ant

i-V

PR

BP

Ant

i-V

PR

BP

Figure 2 DYRK2 functions as an adaptor in the EDVP E3 ligase complex. (a) HeLa cells were transfected with control siRNA or DYRK2-specific siRNA and immunoprecipitation (IP) was performed with anti-DDB1 (top panel), anti-VPRBP (middle panel) or anti-EDD (bottom panel) antibodies. The presence of associated proteins in the immunoprecipitated complexes was assessed by immunoblotting with antibodies as indicated. (b) HeLa cells transfected with DYRK2 specific siRNA were retransfected with either siRNA-resistant wild-type DYRK2 (SiR-DYRK2 WT) or kinase-dead DYRK2 (SiR-DYRK2 KD). The expression of endogenous DYRK2 and the transfected siRNA-resistant DYRK2 was assessed by immunoblotting with anti-DYRK2 antibody.

Actin was used as a loading control. The graph represents DYRK2 kinase activity measured after performing an in vitro kinase assay with DYRK2 immunoprecipitates prepared from the indicated cell lysates, with Woodtide peptide as a substrate (means ± s.d., n = 3). (c) HeLa cells transfected with DYRK2-specific siRNA were retransfected with either siRNA-resistant wild-type (WT) DYRK2 or kinase-dead (KD) DYRK2. Lysates prepared from these cells were used to immunoprecipitate DDB1 or VPRBP with their respective antibodies. The associated EDD in these immunoprecipitates was assessed by immunoblotting with anti-EDD antibody. Uncropped images of blots are shown in Supplementary Information, Fig. S4.



A RT I C L E S

of EDD, DDB1 and VPRBP. The interaction of EDD with DDB1 and VPRBP was seen only in the presence of intact DYRK2, whereas the knockdown of DYRK2 led to a loss of the interaction between EDD

and DDB1 or VPRBP (Fig. 2a). Neither the interaction of DDB1 with VPRBP nor the association of Cul4A with DDB1–VPRBP was affected by the absence of DYRK2 (Fig. 2a). These experiments suggest that

IP

Anti-EDD

Anti-VPRBP

Anti-DYRK2

Anti-Flag(katanin p60)

a

IP: Anti-katanin WB: Ubiquitin

Anti-VPRBP

Anti-Cul4A

Anti-EDD

Anti-DYRK2

Anti-actin

Anti-DDB1

Anti-Cul4B

Anti-Cul3

Anti-katanin

Con

trol

siR

NA

DY

RK

2 si

RN

A

ED

D s

iRN

A

DD

B1

siR

NA

VP

RB

P s

iRN

A

Cul

4A s

iRN

A +

Cul

4B s

iRN

A

Cul

3 si

RN

A

e

Con

trol

siR

NA

VP

RB

P s

iRN

A

Anti-VPRBP

Anti-EDD

Anti-DDB1

Anti-katanin

Anti-EDD

Anti-DDB1

IP: Anti-katanin

Input

b

Con

trol

EDD siRNA

SiR

-ED

D W

T

SiR

-ED

D C

/A

–

Anti-EDD

Anti-katanin

d

Pulldown: Anti-MBPWB: Anti-GST (katanin)

MB

P–E

DD

MB

P–D

DB

1

MB

P–V

PR

BP

WB: Anti-GST(katanin)

WB: Anti-MBP Input

c

Ant

i-Ig

G

Ant

i-Fl

ag

Inp

ut (5

%)

IP: Anti-kataninWB: Ubiquitin

Figure 3 Katanin p60 is the ubiquitylation substrate for EDVP E3 ligase complex. (a) Control (IgG) or anti-Flag immunoprecipitates (IP) were prepared from 293T cells transfected with plasmid encoding a triple-tagged katanin. Western blotting (WB) was performed with the indicated antibodies to show a specific interaction between the DYRK2–EDVP complex and katanin p60. (b) Bacterially expressed recombinant MBP-tagged EDD, DDB1 or VPRBP bound to amylase-Sepharose beads was incubated with recombinant GST–katanin, and the association of katanin was detected by western blotting with anti-GST antibody. The expression of MBP fusion proteins was detected by anti-MBP antibody. (c) HeLa cells were transfected with either control siRNA or VPRBP siRNA, and the association of EDD and DDB1 with katanin was assessed by immunoblotting with their respective antibodies after immunoprecipitation with anti-katanin antibody. (d) HeLa

cells were transfected with different siRNAs as indicated. Cell lysates prepared after 5 h of treatment with 10 µM MG132 were subjected to immunoprecipitation with anti-katanin antibodies. The ubiquitylated katanin was detected with anti-ubiquitin antibody. The protein expression and the specificity of different siRNAs were confirmed by immunoblotting of cell extracts with antibodies as indicated. (e) HeLa cells transfected with EDD-specific siRNA were retransfected with either siRNA-resistant wild-type EDD (SiR-EDD WT) or catalytically inactive EDD (SiR-EDD C/A). Ubiquitylation of katanin was assessed by immunoblotting with anti-ubiquitin antibody after immunoprecipitation with anti-katanin antibody. The expression of endogenous EDD and the transfected siRNA resistant EDD was assessed by immunoblotting with anti-EDD antibody. Uncropped images of blots are shown in Supplementary Information, Fig. S4.



A RT I C L E S

DYRK2 functions as a scaffold, required for the specific recruitment of EDD to DDB1–VPRBP, thus forming a novel EDVP E3 ligase complex. Surprisingly, DYRK2 kinase activity is dispensable as an scaffold for this E3 ligase complex, since the transfection of either siRNA-resistant wild-type DYRK2 or kinase-inactive DYRK2 (each of which was validated by both expression and kinase activity with a synthetic peptide; Fig. 2b), into DYRK2-depleted cells was able to restore the association of EDD with the DDB1–VPRBP complex (Fig. 2c).

EDVP–DYRK2 complex regulates katanin p60 ubiquitylationWe next examined the likely substrates of this new E3 ligase complex. Previously, MBK-2, the Caenorhabditis elegans homologue of mam-malian DYRK2, was shown to phosphorylate and regulate MEI-1 during meiotic maturation in C. elegans8. MEI-1 (a C. elegans homo-logue of katanin p60) is an AAA-ATPase that associates with MEI-2 and functions as a microtubule-severing enzyme. When the C. elegans embryo enters the first mitotic division after exiting from the meiosis, MBK2 phosphorylates MEI-1, which is then degraded through a ubiq-uitin-dependent mechanism by binding to MEL-26, a BTB-domain-containing substrate adaptor protein complexed with Cul3 (ref. 24). The ubiquitin-mediated degradation of MEI-1/katanin was further regulated by a series of neddylation and deneddylation of Cul3 medi-ated by COP9/signalosome26. In higher eukaryotes, it is not yet known whether DYRK2 regulates katanin p60. We therefore first tested and showed that katanin p60 readily associated with the DYRK2–EDVP complex in vivo (Fig. 3a). To identify the direct katanin-binding subunit

of the EDD–DDB1–VPRBP complex, we performed an in vitro binding assay using bacterially expressed recombinant maltose-binding protein (MBP)-tagged EDD, DDB1 and VPRBP along with GST-tagged katanin. Recombinant katanin directly binds VPRBP but not EDD and DDB1 in vitro (Fig. 3b). In addition, katanin interacts with EDD and DDB1 only in the presence of intact VPRBP but not in VPRBP knockdown cells, thus confirming VPRBP as the substrate-binding receptor subunit in the EDVP complex (Fig. 3c). We further examined whether the associ-ated katanin p60 is a substrate of the DYRK2–EDVP E3 ligase complex. We evaluated endogenous katanin ubiquitylation in cells transfected with either control siRNA or siRNAs specific for different components of the DYRK2–EDVP complex in the presence of MG132, a proteaso-mal inhibitor. Katanin p60 was polyubiquitylated in the presence of the intact DYRK2–EDVP, but its ubiquitylation was severely decreased by the depletion of DYRK2, EDD, DDB1 or VPRBP (Fig. 3d). In contrast, katanin polyubiquitylation was unaffected in cells transfected with siRNAs against Cul4A and Cul4B. Previously, it was shown that Cul3/MEL26 also has a function in the degradation of MEI-1 (a C. elegans homologue of katanin p60) during meiotic maturation of C. elegans24,26. We therefore tested whether a similar mechanism for katanin regulation occurs in humans. Interestingly, knocking down Cul3 does lead to a modest decrease in katanin polyubiquitylation, although the severity of this decrease is not comparable with those observed in cells with knockdown of subunits of the EDVP complex (Fig. 3d). It is therefore likely that the EDVP complex has a primary function, whereas Cul3 is of secondary importance in promoting katanin polyubiquitylation in

a b

Anti-VPRBP

Anti-Cul4A

Anti-EDD

Anti-DYRK2

Anti-actin

Anti-DDB1

Anti-Cul4B

Anti-Cul3

Anti-katanin

Con

trol

siR

NA

DY

RK

2 si

RN

A

ED

D s

iRN

A

DD

B1

siR

NA

VP

RB

P s

iRN

A

Cul

4A s

iRN

A +

Cul

4B s

iRN

A

Cul

3 si

RN

A

HeLa control

DYRK2 siRNA

EDD siRNA

VPRBP siRNA

SFB–DYRK2

SFB–VPRBP

SFB–DDB1

CHX (h) 0 1 3 7 9 11

Myc–katanin

WB: Anti-katanin

WB: Anti-APC2

WB: Anti-actin

Con

trol

siR

NA

AP

C2

siR

NA

c

Ant

i-M

yc (k

atan

in)

Figure 4 EDVP E3 ligase complex regulates katanin p60 protein levels. (a) HeLa cells were transfected with control siRNA or siRNAs against DYRK2, EDD, DDB1, VPRBP, Cul4A/Cul4B or Cul3. The protein levels of katanin were assessed by immunoblotting with anti-katanin antibody, and the efficiency of different siRNAs was shown by immunoblotting with the indicated antibodies. (b) HeLa cells transiently expressing Myc-tagged katanin were either transfected with siRNAs against DYRK2, EDD or VPRBP or with plasmids encoding SFB-tagged DYRK2, VPRBP or

DDB1. At 12 h after transfection, cells were treated with cycloheximide (CHX) and collected at the indicated times afterwards. The protein levels of katanin were determined by anti-Myc immunoblotting. (c) Cells transfected with either control siRNA or APC2 siRNA were lysed, and the expression of katanin and APC2 was detected by western blotting (WB) with their respective antibodies. Actin is used as a loading control. Uncropped images of blots are shown in Supplementary Information, Fig. S4.



A RT I C L E S

human cells. We also investigated whether EDD is the functional E3 ligase in the EDVP complex. Knockdown of EDD with siRNA severely affected katanin polyubiquitylation. This defect in katanin ubiquityla-tion was fully rescued by the expression of siRNA-resistant wild-type EDD, but not by the expression of a catalytically inactive HECT-domain mutant of EDD (Fig. 3e). EDD is therefore the catalytic subunit in this E3 ligase complex.

Polyubiquitylation of katanin by the DYRK2–EDVP complex is likely to be required for katanin degradation, as knockdown of DYRK2, EDD, DDB1 or VPRBP, but not that of Cul4A and Cul4B, increased the steady-state lev-els of katanin protein (Fig. 4a). Knockdown of Cul3 also resulted in a small increase in katanin protein levels, again suggesting a secondary role of the Cul3 complex in katanin degradation. In addition, in a cycloheximide chase experiment, Myc-tagged katanin was stabilized in cells depleted of DYRK2, EDD or VPRBP in comparison with cells transfected with control siRNAs. In sharp contrast, overexpression of DYRK2, DDB1 or VPRBP along with katanin led to diminished katanin stability (Fig. 4b). Taken together, these data suggest that katanin is a substrate of the DYRK2–EDVP E3 ligase complex. Because anaphase-promoting complex (APC) was also shown to work with MBK-2 and regulate the degradation of C. elegans katanin p60 in a Cul3 redundant pathway8, we knocked down APC2, a critical subunit of APC, in human cells. However, we did not observe any change in katanin protein levels (Fig. 4c) and therefore concluded that APC may not be involved in katanin degradation in humans.

DYRK2-mediated phosphorylation is required for katanin p60 degradationMBK2/DYRK2 is known to phosphorylate the katanin homologue MEI-1 in C. elegans27. We next investigated whether DYRK2 would phospho-rylate katanin and be required for katanin ubiquitylation by the EDVP complex. In vitro kinase assays revealed that immunoprecipitated wild-type DYRK2, but not kinase-inactive DYRK2, could phosphorylate bacterially expressed GST–katanin (Fig. 5a). We also tested whether DYRK2 in the EDVP complex could phosphorylate katanin by per-forming immunoprecipitation of the EDVP complex followed by an in vitro kinase assay. Immunoprecipitates of EDD, DDB1 and VPRBP, but not of Cul4A, showed intrinsic kinase activity towards katanin (Fig. 5b), suggesting that DYRK2 in the EDVP complex is still capable of phos-phorylating its substrate. Katanin contains several consensus DYRK2 phosphorylation sites28 (Fig. 5c). We mutated katanin at these serine or threonine residues individually or in combination and examined whether any of these residues were potential DYRK2 phosphorylation sites in vitro. Serine 42, serine 109 and threonine 133 are likely to be the major DYRK2 phosphorylation sites, because single mutations at these sites showed decreased phosphorylation by DYRK2, and the triple mutant showed almost no DYRK2-mediated phosphorylation (Fig. 5d). Furthermore, we detected the presence of phosphoserine and phosphothreonine resi-dues in the immunoprecipitated wild-type katanin (Fig. 5e), indicating that katanin is phosphorylated in vivo. However, these phosphorylations

32P

Anti-GST(katanin)

32P

Anti-GST(katanin)

DY

RK

2 W

T

GS

T–ka

tani

n

GS

T–ka

tani

n+

DY

RK

2 W

T

GS

T–ka

tani

n+

DY

RK

2 K

D

a c

d

WT

S42

A

S81

A

S10

9A

S42

A/S

109A

T133

A

S42

A/T

133A

S10

9A/T

133A

S42

A/S

109A

/T13

3A

Ant

i-Ig

G

e f

InputAnti-Myc(katanin)

IgH

IgH

pSer

pThr

Myc

-kat

anin

WT

Myc

-kat

anin

AAA

IP

Ant

i-ka

tani

n

DYRK2 consensus: R/K x x (x) S/T P/V

Katanin p60 Ser 42: K Y L V S V

Katanin p60 Ser 109: R R P S P

Katanin p60 Thr 81: K L D S T P

Katanin p60 Thr 133: R P S T T V

b

32P

GST–katanin

Ant

i-E

DD

Ant

i-D

DB

1

Ant

i-V

PR

BP

Ant

i-D

YR

K2

Ant

i-C

ul4A

Ant

i-Ig

G

WB: Anti-pSer

WB: Anti-pThr

InputAnti-katanin

IP: Anti-Myc

WB: Anti-pSer

IP: Anti-MycWB: Anti-pThr

Figure 5 DYRK2 phosphorylates katanin. (a) An in vitro kinase assay was performed with a bacterially expressed GST–katanin and immunoprecipitated wild-type or kinase-inactive DYRK2. (b) An in vitro kinase assay was performed with a bacterially expressed GST–katanin with immunoprecipitates prepared by using antibodies against EDD, DDB1, VPRBP, DYRK2 and Cul4A. (c) The alignment of potential katanin phosphorylation sites with DYRK2 consensus sequence is presented. Bold lettering indicates the phosphorylated residue. (d) In vitro DYRK2 kinase assays were conducted with different bacterially expressed GST–katanin phosphorylation-site mutants as

indicated. (e) The in vivo phosphorylation of katanin was detected by immunoblotting (WB) with anti-phosphoserine-specific or anti-phosphothreonine-specific antibodies after immunoprecipitation (IP) with control IgG or katanin antibodies. (f) The in vivo phosphorylation of wild-type katanin and the triple phospho-mutant of katanin (AAA) was assessed by immunoblotting with phosphoserine-specific or phosphothreonine-specific antibodies after anti-Myc immunoprecipitation of extracts prepared from 293T cells expressing Myc-tagged wild-type or mutant katanin. IgH indicates IgG heavy chain. Uncropped images of blots are shown in Supplementary Information, Fig. S4.



A RT I C L E S

were greatly diminished in the triple phospho-mutant (AAA mutant) of katanin (Fig. 5f), suggesting that these residues are indeed major phos-phorylation sites in vivo.

We further investigated whether the DYRK2-mediated phosphoryla-tion of katanin is required for its ubiquitylation in vivo. Ubiquitylation of Myc-tagged wild-type katanin was easily detected, whereas the

ubiquitylation of the AAA mutant of katanin was severely diminished (Fig. 6a). To further support the idea that the intact DYRK2–EDVP complex mediates the ubiquitylation of phosphorylated katanin, we performed in vitro reconstitution assays with GST–katanin as the ubiq-uitylation substrate. As shown in Fig. 6b, only the intact EDVP complex containing wild-type EDD (but not mutant EDD) resulted in robust

DDB1

VPRBP

Katanin

DYRK2

EDD WT

EDD C/A

E2

E1

Anti-Ub

Anti-GST(katanin)

Ub + + ++++++– + ++++++

+ – ++++++

– – –––+++

+ + +++–––

+ + ++–++–

+ + ++++++

+ + ++++++

+ + +–+++–

Anti-Ub

Anti-GST(katanin)

DYRK2 KDK

atan

in W

TK

atan

in A

AA

Anti-Myc(katanin)

Anti-Flag(DYRK2)

Anti-actin

Anti-Myc(katanin)

Anti-Flag(DYRK2)

Anti-actin

DYRK2 WT

DYRK2 KDDYRK2 WT

a

Anti-Myc (katanin p60)

Anti-Flag (VPRBP)

Myc–katanin WT + –

– +Myc–katanin AAA

Flag–VPRBPHA–Ub

dc

b

+ ++ +

IP: Anti-Myc(katanin p60)

WB: Anti-HA(Ub)

Katanin WT

Katanin AAA

DYRK2

EDD WT

EDD C/A

E2

E1

Ub + + ++++++

– + ++–––+

– – –––++–

– – +––+–+

– – +++–––

+ + ––

++

+ + ++++++ ++

+ + ++++++ ++

––

++

+–

++

–––+++

µg3210.50.25–3210.50.25–

µg3210.50.25–3210.50.25–

Figure 6 DYRK2 kinase activity is required for the regulation of katanin degradation. (a) Myc-tagged wild-type (WT) katanin or the triple phospho-mutant of katanin (AAA) was expressed in HeLa cells along with Flag–VPRBP and HA-tagged ubiquitin (HA–Ub). Levels of katanin ubiquitylation were evaluated by anti-HA immunoblotting (WB) after immunoprecipitation (IP) of katanin from the cell extracts. (b) In vitro reconstitution experiments were performed with GST–katanin as a substrate in the presence of recombinant ubiquitin, E1 (UBE1), E2 (UbcH5), MBP-tagged EDD, EDD C/A, DDB1, VPRBP and DYRK2 in various combinations as indicated. Ubiquitylated species of katanin and GST–katanin were detected by immunoblotting with anti-ubiquitin and anti-GST antibodies,

respectively. (c) In vitro reconstitution experiments were performed as in b, using either wild-type GST–katanin or the AAA mutant of katanin as a substrate in the presence of various recombinant proteins as indicated. Ubiquitylated species of katanin and GST–katanin were detected by immunoblotting with anti-ubiquitin and anti-GST antibodies, respectively. (d) The effect of DYRK2 kinase activity and katanin phosphorylation on the regulation of katanin protein levels was assessed by transient transfection experiments. 293T cells were transfected with the indicated expression vectors for DYRK2 and katanin, and the protein levels were estimated by immunoblotting 24 h after transfection. Uncropped images of blots are shown in Supplementary Information, Fig. S4.



A RT I C L E S

katanin polyubiquitylation. Previous phosphorylation of katanin is essential for katanin ubiquitylation: only wild-type katanin, but not its AAA mutant, could be readily ubiquitylated by the EDVP complex in vitro (Fig. 6c). These data suggest that DYRK2-dependent katanin phos-phorylation is a prerequisite for katanin ubiquitylation, indicating that

DYRK2 kinase activity is critical for the function of the DYRK2–EDVP E3 ligase complex. The phosphorylation-dependent degradation of katanin was substantiated by co-transfection experiments. Co-transfection of wild-type DYRK2, but not kinase-inactive DYRK2, along with katanin decreased the steady-state levels of katanin protein (Fig. 6d). In contrast,

Anti-DYRK2

Anti-EDD

Anti-katanin

Con

trol

Kat

anin

DY

RK

2 si

RN

A

ED

D s

iRN

A

DY

RK

2 si

RN

A +

kat

anin

siR

NA

ED

D s

iRN

A +

kat

anin

siR

NA

0

5

10

15

Kat

anin

WT

Kat

anin

AA

A

Kat

anin

AA

A

+ D

YR

K2

Kat

anin

WT

+ D

YR

K2

Kat

anin

AA

A +

DY

RK

2 K

D

Kat

anin

WT

+ D

YR

K2

KD

Con

trolP

hosp

ho-H

3-p

ositi

ve c

ells

(per

cent

age)

HeLa Katanin HeLa Control DYRK2 siRNA EDD siRNA DYRK2 siRNA +katanin siRNA

EDD siRNAkatanin siRNA

2N 2N 2N 2N 2N 2N4N 4N 4N 4N 4N 4N

a

b c

d e

p pp

(Ub)n

DYRK2

EDD

DDB1 VPRBPKatanin

G1: 62% S: 16%

G2/M: 21.9% >4N: 0.1%

G1: 40% G1: 33% G1: 26% G1: 63% G1: 60%S: 10% S: 12% S: 15% S: 8% S: 10%

G2/M: 43% G2/M: 48% G2/M: 53% G2/M: 25% G2/M: 27%>4N: 7% >4N: 7% >4N: 6% >4N: 4% >4N: 3%

0

25

50

Kat

anin

WT

Kat

anin

AA

A

Kat

anin

AA

A+

DY

RK

2

Kat

anin

WT

+ D

YR

K2

Kat

anin

AA

A+

DY

RK

2 K

D

Kat

anin

WT

+ D

YR

K2

KD

Con

trol

G2/

M-a

rres

ted

cel

ls (p

erce

ntag

e)

Cou

nts

DYRK2–EDD E3 ligase complex

G2/M progression

Figure 7 DYRK2 regulates mitotic progression by means of its adaptor and kinase function. (a) HeLa cells were transfected with plasmids encoding katanin or with different siRNAs as indicated, and the cell cycle profiles were determined by propidium iodide staining followed by flow cytometric analysis. (b) The protein levels of katanin, DYRK2 and EDD in HeLa cells transfected with plasmids encoding katanin or different siRNA combinations were determined by western blotting with the indicated antibodies. (c) HeLa cells were transfected with plasmids encoding wild-type katanin or the triple phospho-mutant of katanin (AAA) along with plasmids encoding wild-type

(WT) or kinase-dead (KD) DYRK2. The percentage of cells in G2/M was determined by fluorescence-activated cell sorting analysis. Data are presented as mean ± SD for three different experiments. (d) The percentages of mitotic cells as measured by positive phospho-H3 staining were determined in HeLa cells transfected with the indicated constructs. Data are presented as mean ± SD for three different experiments. (e) Model for a novel DYRK2–EDD E3 ligase complex demonstrates that DYRK2 functions as both an adaptor and a kinase and regulates G2/M cell cycle progression. Uncropped images of blots are shown in Supplementary Information, Fig. S4.



A RT I C L E S

the protein levels of the AAA katanin mutant remained largely unaffected (Fig. 6d), suggesting that DYRK2-mediated phosphorylation is a priming event required for katanin ubiquitylation and degradation.

EDVP–DYRK2 complex controls mitotic transitionPrevious studies have reported that katanin is a microtubule AAA-ATPase that is important in mitosis29,30. Katanin is required for sev-ering microtubules at the mitotic spindles when disassembly of the microtubules is required for the segregation of sister chromatids during anaphase. Both DYRK2 (ref. 7) and EDD31 have also been suggested to function during mitosis. To establish a functional link between the DYRK2–EDVP ligase complex and katanin degradation, we checked the cell cycle profile of HeLa cells transiently transfected with katanin. Overexpression of katanin led to the accumulation of a 4N population and polyploid (>4N) cells (Fig. 7a). Similarly, siRNA-mediated downreg-ulation of either DYRK2 or EDD, which led to an upregulation of katanin (Fig. 7b), also resulted in the accumulation of cells with a 4N DNA con-tent (Fig. 7a). This abnormal accumulation of 4N cells after depletion of DRYK2 or EDD could be rescued by the simultaneous depletion of katanin by siRNA (Fig. 7a). In addition, DYRK2-mediated katanin phos-phorylation is required for proper cell cycle progression: co-expression of DYRK2 with the katanin AAA mutant but not wild-type katanin led to an increase in cells with a 4N DNA content (Fig. 7c). This increase in the 4N population is attributed to defective mitotic progression, because we observed an increased number of phospho-H3-positive cells when wild-type or non-phosphorylatable katanin was overexpressed (Fig. 7d). Co-expression of wild-type DYRK2 but not the kinase-dead version with wild-type katanin decreased the number of mitotic cells to normal levels, whereas co-expression of DYRK2 with the katanin AAA mutant failed to rescue this mitotic defect (Fig. 7d). Collectively, these results suggest that an active DYRK2–EDVP ligase complex regulates mitotic transition through the modulation of katanin protein levels.

DiSCuSSioNProtein kinases regulate a variety of biological processes, including cell proliferation, apoptosis, development and tumorigenesis by phosphor-ylating their respective downstream substrates32. In this study we have uncovered a novel role for protein kinase; for example, it functions as an assembly factor for an E3 ubiquitin ligase complex. In particular, we have shown that DYRK2 has dual roles in this E3 ligase complex. Not only is it required for the assembly of the complex, but it also phosphorylates its substrate and primes the substrate for degradation. Phosphorylation-dependent protein degradation is a common mechanism for regulat-ing protein stability in a cell-cycle-dependent or stimulus-dependent manner. This often occurs as a two-step process33, in which initially a kinase phosphorylates the substrate. Once phosphorylated, the sub-strate is recognized by F-box-containing or BC-box substrate receptor proteins and targeted to E3 ligase complexes for degradation. Here we provide evidence for the integral presence of a kinase, DYRK2, within the EDD–DDB1–VPRBP (EDVP) E3 ligase complex, which merges the functional properties of a protein kinase and an E3 ligase into a single unit that can recognize, phosphorylate and degrade substrates in con-cert. It is currently unclear how DYRK2-mediated phosphorylation of katanin promotes its ubiquitylation, but it is possible that phosphoryla-tion of katanin leads to a conformational change that exposes some of the substrate residues for efficient ubiquitylation by EDD. Our functional

studies further suggest that the DYRK2–EDVP E3 ligase complex has a crucial function in regulating normal mitotic progression. Because overexpression of both DYRK2 (refs 11, 12) and EDD15,16 is frequently reported in cancers, it is tempting to speculate that aberrant mitosis and altered cell cycle progression through the hyperactivation of the DYRK2–EDVP E3 ligase complex might be a key mechanism in promot-ing neoplastic transformation. Future studies with animal models will reveal the role of this DYRK2–EDVP complex in cancer development and progression.

MEtHoDSPlasmids. Full-length DYRK2, DYRK1B, EDD, DDB1, VPRBP, Cul4A, katanin and DYRK2-KD were cloned into a S-protein–Flag–SBP (streptavidin-bind-ing protein) triple tagged destination vector with the Gateway cloning system (Invitrogen). Full-length EDD, katanin and katanin AAA mutant were also cloned to a Myc-tagged destination vector. GST-tagged DYRK2, MBP-tagged EDD, cata-lytically inactive EDD (EDD C/A), DDB1 and VPRBP bacterial expression vectors were generated by transferring their coding sequences into destination vectors with the Gateway system. Various deletion and point mutants for katanin p60 and the kinase-dead version of DYRK2 were generated by PCR-based site-directed mutagenesis. Wild-type katanin and mutants of katanin were also cloned to a GST-tagged vector. Constructs of Myc-tagged ubiquitin and haemagglutinin (HA)-tagged ubiquitin were used in ubiquitylation assays in vivo. siRNA-resistant wild-type DYRK2, kinase-dead DYRK2, wild-type EDD and EDD C/A mutant constructs were generated by introducing silent mutations into their respective triple-tagged vectors by using site-directed mutagenesis; the constructs were veri-fied by sequencing.

Antibodies. Rabbit anti-katanin antibodies were raised by immunizing rabbits with GST–katanin p60 fusion protein (residues 30–240). Antisera were affinity-purified with the AminoLink Plus Immobilization and Purification Kit (Pierce). Anti-DYRK2 (Abcam), anti-EDD, anti-DDB1, anti-VPRBP, anti-Cul4A, anti-Cul3 (all from Bethyl Laboratories), anti-FBX22 (Novus Biologicals), anti-PBK (Cell Signaling Technology), anti-Flag, anti-(maltose-binding protein); Clone 17, anti-actin, anti-Cul4B (Sigma), anti-Roc1 (Invitrogen), anti-GST, anti-Myc; Clone 9E10 (Santa Cruz Biotechnologies), anti-phosphoserine H3, anti-phosphothreonine (Cell Signaling Technology) and anti-phosphoserine, anti-ubiquitin (Millipore) antibodies were used in this study.

Tandem affinity purification. 293T cells were transfected with S-protein-Flag-SBP triple-tagged DYRK2; three weeks later, puromycin-resistant colonies were selected and screened for DYRK2 expression. The DYRK2-positive stable cells were then maintained in RPMI medium supplemented with fetal bovine serum and 2 µg ml−1 puromycin. The SFB–DYRK2 stable cells were lysed with NETN buffer (20 mM Tris-HCl pH 8.0, 100 mM NaCl, 1 mM EDTA, 0.5% Nonidet P40) containing 50 mM β-glycerophosphate, 10 mM NaF, 1 µg ml−1 pepstatin A and 1 µg ml−1 aprotinin on ice for 20 min. After removal of cell debris by cen-trifugation, crude cell lysates were incubated with streptavidin-Sepharose beads (Amersham Biosciences) for 1 h at 4 °C. The bound proteins were washed three times with NETN buffer and then eluted twice with 2 mg ml−1 biotin (Sigma) for 30 min at 4 °C. The eluates were incubated with S-protein-agarose beads (Novagen) for 1 h at 4 °C and then washed three times with NETN buffer. The proteins bound to S-protein-agarose beads were resolved by SDS–PAGE and revealed by staining with Coomassie blue. The identities of eluted proteins were revealed by mass spectrometry analysis performed at the Taplin Biological Mass Spectrometry Facility, Harvard University.

Cell transfections, immunoprecipitation and immunoblotting. 293T cells or HeLa cells were transfected with various plasmids by using Lipofectamine (Invitrogen) in accordance with the manufacturer’s protocol. For immunopre-cipitation assays, cells were lysed with NETN buffer as described above. The whole-cell lysates obtained by centrifugation were incubated with 2 µg of speci-fied antibody bound to either protein A-Sepharose or protein G-Sepharose beads (Amersham Biosciences) for 1 h at 4 °C. The immunocomplexes were then washed with NETN buffer four times and subjected to SDS–PAGE. Immunoblotting was performed with standard protocols.



A RT I C L E S

GST pulldown and in vitro binding assays. Bacterially expressed GST–DYRK2 or control GST bound to glutathione-Sepharose beads (Amersham) was incu-bated with 293T cell lysates for 1 h at 4 °C, and the washed complexes were eluted by boiling in SDS sample buffer and then separated by SDS–PAGE; the interac-tions were analysed by western blotting. For in vitro binding assays, bacterially expressed MBP–EDD, MBP–DDB1 or MBP–VPRBP bound to amylase-Sepharose beads were incubated with bacterially purified GST–katanin for 1 h at 4 °C; the washed complexes were eluted by boiling in SDS sample buffer and separated by SDS–PAGE, and the interactions were analysed by western blotting with the indicated antibodies.

RNA interference. Control siRNA and the smart-pool siRNAs against DDB1, Cul4A, Cul4B, Cul3, APC2 and katanin and the on-target plus individual siRNAs against DYRK2, EDD and VPRBP were purchased from Dharmacon Inc. Transfection was performed twice, 30 h apart, with 200 nM siRNA using Oligofectamine reagent in accordance with the manufacturer’s protocol (Invitrogen). The following target sequences were used.

Individual siRNA sequences: DYRK2 siRNA, 5´-GGUGCUAUCA-CAUCUAUAU-3´;

EDD siRNA, 5´-CAACUUAGAUCUCCUGAAA-3´; DDB1 siRNA, 5´-ACACUUUGGUGCUCUCUU-3´; VPRBP siRNA, 5´-GAUGGCGGAUGC-UUUGAUA-3´.

Pooled siRNA sequences: Cul4A siRNA, 5´-GCACAGAUCCUUCCGUUUA-3´, 5´-GAACAGCGAUCGUAAUCAA-3´, 5´-GCAUGUGGAUUCAAAGUUA-3´ and 5´-GCGAGUACAUCAAGACUUU-3´; Cul4B siRNA, 5´-GCUAUU-GGCCGACAUAUGU-3´, 5´-CAGAAGUCAUUAAUUGCUA-3´, 5´-CAAA-CGGCCUAGCCAAAUC-3´ and 5´-CGGAAAGAGUGCAUCUGUA-3´; Cul3 siRNA, 5´-CCGAACAUCUCAUAAAUAA-3´, 5´-GAGAAGAUGUACU-AAAUUC-3´, 5´-GAGAUCAAGUUGUACGUUA-3´ and 5´-GCGGAA AGGA-GAAGUCGUA-3´; APC2 siRNA, 5´-GAGAUGAUCCAGCGUCUGU-3´, 5´-GACAUCAUCACCCUCUAUA-3´, 5´-GAUCGUAUCUACAACAUGC-3´ and 5´-GAGAAGAAGUCCACACUAU-3´; katanin siRNA, 5´-GGGAGGA-GCUAUUACGAAU-3´, 5´-GCUGUUCGUUGUCGUGAAA-3´, 5´-GGAU-CAUGCUAACUCGAGA-3´ and 5´-CAUUGAAAGAUACGAGAAA-3´.

In vitro kinase assay. Wild-type DYRK2 and kinase-inactive DYRK2, which were expressed in 293T cells, were immunoprecipitated with Flag-agarose beads and used as a kinase source. GST-tagged katanin and its mutant proteins, expressed in Escherichia coli strain BL21, were purified with glutathione-Sepharose beads and used as substrates. The kinase (DYRK2) and substrates (katanin) were incubated in kinase assay buffer (10 mM HEPES pH 7.5, 50 mM NaCl, 10 nM MgCl2, 10 mM MnCl2, 1 mM EGTA, 1 mM dithiothreitol, 5 µM ATP, 10 mM NaF, 50 mM glycerophosphate) along with 10 µCi of [γ-32P] ATP) for 30 min at 30 °C. Reactions were stopped by the addition of SDS sample buffer. Then samples were boiled for 5 min at 100 °C followed by SDS–PAGE and autoradiography. The validation of siRNA-resistant wild-type DYRK2 and kinase-dead DYRK2 was performed by DYRK2 immunoprecipitation followed by a kinase assay with a synthetic Woodtide peptide (KKISGRLSPIMTEQ) as a substrate (purchased from Millipore).

In vivo ubiquitylation assay. HeLa cells were transfected with various combina-tions of plasmids as indicated in Figs 2d and 3f along with either Myc-tagged ubiquitin or HA-tagged ubiquitin. At 24 h after transfection, cells were treated with MG132 (4 µΜ) for 6 h and the whole-cell extracts prepared by lysis in NETN buffer were subjected to immunoprecipitation of the substrate protein. Analysis of ubiquitylation was performed by immunoblotting with either anti-Myc or anti-HA antibodies.

In vitro reconstitution assay. The reactions were performed at 30 °C for 15 min in 25 µl of ubiquitylation reaction buffer (40 mM Tris-HCl pH 7.6, 2 mM dithio-threitol, 5 mM MgCl2, 0.1 M NaCl, 2 mM ATP) containing the following compo-nents: 100 µΜ ubiquitin, 20 nM E1 (UBE1) and 100 nM UbcH5b (all from Boston Biochem). Various combinations of EDVP E3 ligase components (25 ng of EDD or EDD C/A, plus DDB1, VPRBP and DYRK2) as indicated were added to the reaction. Either wild-type GST–katanin or the AAA mutant bound to glutathione-Sepharose beads was used as a substrate in the reaction. After ubiquitylation reaction, the glutathione-Sepharose beads were washed five times with NETN

buffer and boiled with SDS–PAGE loading buffer; the ubiquitylation of katanin was monitored by western blotting with anti-ubiquitin antibody.

Cell cycle analysis. HeLa cells transfected with the desired expression vectors and siRNA were harvested, washed with phosphate-buffered saline and fixed with ice-cold 70% ethanol for at least 1 h. Cells were washed twice in PBS and treated for 30 min at 37 °C with RNase A (5 µg ml−1) and propidium iodide (50 µg ml−1), then analysed on a FACScan flow cytometer (Becton Dickinson). The percentage of cells in different cell cycle phases was calculated with Flowjo analysis software.

Immunofluorescence staining. Cells grown on coverslips were fixed with 3% paraformaldehyde solution in PBS containing 50 mM sucrose at 25 °C for 15 min. After permeabilization at room temperature with 0.5% Triton X-100 buffer con-taining 20 mM HEPES pH 7.4, 50 mM NaCl, 3 mM MgCl2 and 300 mM sucrose for 5 min, cells were incubated with a primary phosphoserine H3 antibody at 37 °C for 20 min. After being washed with PBS, cells were incubated with rhod-amine-conjugated secondary antibody at 37 °C for 20 min. Nuclei were counter-stained with DAPI (4,6-diamidino-2-phenylindole). After a final wash with PBS, coverslips were mounted in glycerine containing p-phenylenediamine.


ACknowledgeMentSWe thank Jamie Wood for critical reading of the manuscript and for providing valuable suggestions. We thank Amanda Russell for providing EDD expression vectors. This work was supported in part by grants from the National Institutes of Health (to J.C.). J.C. is a recipient of an Era of Hope Scholars award from the Department of Defense and is a member of Mayo Clinic Breast SPORE programme.

Author ContributionSS.M. performed all the experiments. S.M. and J.C. designed the experiments, analysed the data and wrote the manuscript.

CoMpeting finAnCiAl intereStSThe authors declare that they have no competing financial interests.

Published online at http://www.nature.com/naturecellbiology Reprints and permissions information is available online at http://npg.nature.com/reprintsandpermissions/

1. Becker, W. et al. Sequence characteristics, subcellular localization, and substrate specificity of DYRK-related kinases, a novel family of dual specificity protein kinases. J. Biol. Chem. 273, 25893–25902 (1998).

2. Kannan, N. & Neuwald, A. F. Evolutionary constraints associated with functional specifi-city of the CMGC protein kinases MAPK, CDK, GSK, SRPK, DYRK, and CK2α. Protein Sci. 13, 2059–2077 (2004).

3. Lochhead, P. A., Sibbet, G., Morrice, N. & Cleghon, V. Activation-loop autophosphoryla-tion is mediated by a novel transitional intermediate form of DYRKs. Cell 121, 925–936 (2005).

4. Gwack, Y. et al. A genome-wide Drosophila RNAi screen identifies DYRK-family kinases as regulators of NFAT. Nature 441, 646–650 (2006).

5. Woods, Y. L. et al. The kinase DYRK phosphorylates protein-synthesis initiation factor eIF2Bε at Ser539 and the microtubule-associated protein tau at Thr212: potential role for DYRK as a glycogen synthase kinase 3-priming kinase. Biochem. J. 355, 609–615 (2001).

6. Skurat, A. V. & Dietrich, A. D. Phosphorylation of Ser640 in muscle glycogen synthase by DYRK family protein kinases. J. Biol. Chem. 279, 2490–2498 (2004).

7. Nishi, Y. & Lin, R. DYRK2 and GSK-3 phosphorylate and promote the timely degradation of OMA-1, a key regulator of the oocyte-to-embryo transition in C. elegans. Dev. Biol. 288, 139–149 (2005).

8. Lu, C. & Mains, P. E. The C. elegans anaphase promoting complex and MBK-2/DYRK kinase act redundantly with CUL-3/MEL-26 ubiquitin ligase to degrade MEI-1 micro-tubule-severing activity after meiosis. Dev. Biol. 302, 438–447 (2007).

9. Kinstrie, R., Lochhead, P. A., Sibbet, G., Morrice, N. & Cleghon, V. dDYRK2 and Minibrain interact with the chromatin remodelling factors SNR1 and TRX. Biochem. J. 398, 45–54 (2006).

10. Taira, N., Nihira, K., Yamaguchi, T., Miki, Y. & Yoshida, K. DYRK2 is targeted to the nucleus and controls p53 via Ser46 phosphorylation in the apoptotic response to DNA damage. Mol. Cell 25, 725–738 (2007).

11. Gorringe, K. L., Boussioutas, A. & Bowtell, D. D. Novel regions of chromosomal ampli-fication at 6p21, 5p13, and 12q14 in gastric cancer identified by array comparative genomic hybridization. Genes Chromosomes Cancer 42, 247–259 (2005).

12. Miller, C. T. et al. Amplification and overexpression of the dual-specificity tyrosine-(Y)-phosphorylation regulated kinase 2 (DYRK2) gene in esophageal and lung adenocarci-nomas. Cancer Res. 63, 4136–4143 (2003).



http://www.nature.com/naturecellbiology



A RT I C L E S

13. Callaghan, M. J. et al. Identification of a human HECT family protein with homology to the Drosophila tumor suppressor gene hyperplastic discs. Oncogene 17, 3479–3491 (1998).

14. Honda, Y. et al. Cooperation of HECT-domain ubiquitin ligase hHYD and DNA topoi-somerase II-binding protein for DNA damage response. J. Biol. Chem. 277, 3599–3605 (2002).

15. Clancy, J. L. et al. EDD, the human orthologue of the hyperplastic discs tumour sup-pressor gene, is amplified and overexpressed in cancer. Oncogene 22, 5070–5081 (2003).

16. O’Brien, P. M. et al. The E3 ubiquitin ligase EDD is an adverse prognostic factor for serous epithelial ovarian cancer and modulates cisplatin resistance in vitro. Br. J. Cancer 98, 1085–1093 (2008).

17. Chu, G. & Chang, E. Xeroderma pigmentosum group E cells lack a nuclear factor that binds to damaged DNA. Science 242, 564–567 (1988).

18. Angers, S. et al. Molecular architecture and assembly of the DDB1–CUL4A ubiquitin ligase machinery. Nature 443, 590–593 (2006).

19. Lee, J. & Zhou, P. DCAFs, the missing link of the CUL4-DDB1 ubiquitin ligase. Mol. Cell 26, 775–780 (2007).

20. Jin, J., Arias, E. E., Chen, J., Harper, J. W. & Walter, J. C. A family of diverse Cul4-Ddb1-interacting proteins includes Cdt2, which is required for S phase destruction of the replication factor Cdt1. Mol. Cell 23, 709–721 (2006).

21. He, Y. J., McCall, C. M., Hu, J., Zeng, Y. & Xiong, Y. DDB1 functions as a linker to recruit receptor WD40 proteins to CUL4–ROC1 ubiquitin ligases. Genes Dev. 20, 2949–2954 (2006).

22. Higa, L. A. et al. CUL4–DDB1 ubiquitin ligase interacts with multiple WD40-repeat proteins and regulates histone methylation. Nature Cell Biol. 8, 1277–1283 (2006).

23. Cullinan, S. B., Gordan, J. D., Jin, J., Harper, J. W. & Diehl, J. A. The Keap1-BTB pro-tein is an adaptor that bridges Nrf2 to a Cul3-based E3 ligase: oxidative stress sensing by a Cul3-Keap1 ligase. Mol. Cell. Biol. 24, 8477–8486 (2004).

24. Xu, L. et al. BTB proteins are substrate-specific adaptors in an SCF-like modular ubiquitin ligase containing CUL-3. Nature 425, 316–321 (2003).

25. Salinas, G. D. et al. Actinfilin is a Cul3 substrate adaptor, linking GluR6 kainate recep-tor subunits to the ubiquitin–proteasome pathway. J. Biol. Chem. 281, 40164–40173 (2006).

26. Pintard, L. et al. Neddylation and deneddylation of CUL-3 is required to target MEI-1/Katanin for degradation at the meiosis-to-mitosis transition in C. elegans. Curr. Biol. 13, 911–921 (2003).

27. Stitzel, M. L., Pellettieri, J. & Seydoux, G. The C. elegans DYRK Kinase MBK-2 marks oocyte proteins for degradation in response to meiotic maturation. Curr. Biol. 16, 56–62 (2006).

28. Campbell, L. E. & Proud, C. G. Differing substrate specificities of members of the DYRK family of arginine-directed protein kinases. FEBS Lett. 510, 31–36 (2002).

29. McNally, F. J. & Thomas, S. Katanin is responsible for the M-phase microtubule-severing activity in Xenopus eggs. Mol. Biol. Cell 9, 1847–1861 (1998).

30. McNally, K., Audhya, A., Oegema, K. & McNally, F. J. Katanin controls mitotic and meiotic spindle length. J. Cell Biol. 175, 881–891 (2006).

31. Munoz, M. A. et al. The E3 ubiquitin ligase EDD regulates S-phase and G2/M DNA damage checkpoints. Cell Cycle 6, 3070–3077 (2007).

32. Manning, G., Whyte, D. B., Martinez, R., Hunter, T. & Sudarsanam, S. The protein kinase complement of the human genome. Science 298, 1912–1934 (2002).

33. Hunter, T. The age of crosstalk: phosphorylation, ubiquitylation, and beyond. Mol. Cell 28, 730–738 (2007).



A RT I C L E S

Phosphorylation-dependent regulation of cytosolic localization and oncogenic function of Skp2 by Akt/PKBHui-Kuan Lin1,3,8, Guocan Wang1,2, Zhenbang Chen1,2,7, Julie Teruya-Feldstein1, Yan Liu4, Chia-Hsin Chan3, Wei-Lei Yang3, Hediye Erdjument-Bromage5, Keiichi I. Nakayama6, Stephen Nimer4, Paul Tempst5 and Pier Paolo Pandolfi1,2,8

Skp2 is an F-box protein that forms the SCF complex with Skp1 and Cullin-1 to constitute an E3 ligase for ubiquitylation. Ubiquitylation and degradation of p27 are critical for Skp2-mediated entry to the cell cycle, and overexpression and cytosolic accumulation of Skp2 have been clearly associated with tumorigenesis, although the functional significance of the latter is still unknown. Here we show that Akt/protein kinase B (PKB) interacts with and directly phosphorylates Skp2. We find that Skp2 phosphorylation by Akt triggers SCF complex formation and E3 ligase activity. A phosphorylation-defective Skp2 mutant is drastically impaired in its ability to promote cell proliferation and tumorigenesis. Furthermore, we show that Akt-mediated phosphorylation triggers 14-3-3β-dependent Skp2 relocalization to the cytosol, and we attribute a specific role to cytosolic Skp2 in the positive regulation of cell migration. Finally, we demonstrate that high levels of activation of Akt correlate with the cytosolic accumulation of Skp2 in human cancer specimens. Our results therefore define a novel proto-oncogenic Akt/PKB-dependent signalling pathway.

The ubiquitin–proteasome system regulates the cell cycle through the control of protein ubiquitylation and degradation1,2. One of the key ubiquitin ligases (E3 ligase) in this process is the Skp1/Cul-1/F-box (SCF) complex, which consists of Skp1, Cullin-1 (Cul-1) and RBX1, as well as an F-box protein, all required for its E3 ubiquitin ligase activity. Disruption of this complex severely ablates its enzymatic activity1,2.

Skp2 (S-phase kinase associated protein-2) is an SCF F-box protein and is responsible for substrate recognition1,2. It binds to p27 and tar-gets it for ubiquitylation and degradation3–5. Overexpression of Skp2 induces cell cycle entry, and the degradation of p27 is required for Skp2-mediated cell cycle progression6,7. Skp2 deficiency shows elevated p27 protein levels and a profound impairment in proliferation accompanied by nuclear enlargement, polypoidy and centrosome overduplication8,9. Overexpression of Skp2 is frequently observed in human cancers of diverse histology, whereas in most human cancers a lower level of p27 is an adverse prognostic marker1,2. Skp2 cooperates with H-RasG12V to transform primary rodent fibroblasts10. Overexpression of Skp2 in the T-cell compartment cooperates with N-Ras to induce T-cell lym-phomas11, and prostate-specific expression of Skp2 leads to prostatic intraepithelial neoplasia12. These observations suggest that Skp2 over-expression may contribute to tumorigenesis.

Although substantial advances have been made in understanding the mechanisms that control its levels of expression, the molecular mecha-nisms by which Skp2 activity within the SCF complex and its subcel-lular localization are regulated are currently unknown. This is of further relevance because, in human cancer, Skp2 is frequently found aberrantly localized in the cytosol. Here we show that phosphorylation of Skp2 by Akt/PKB constitutes a molecular switch that critically controls the formation, localization and function of the Skp2-SCF complex.

RESUltSAkt/PKB interacts with and phosphorylates Skp2Skp2 is phosphorylated during G1/S transition1,2,13. Mitogens such as epidermal growth factor (EGF) can also lead to Skp2 phosphoryla-tion14. However, the functional relevance of this phosphorylation event is unclear and the kinases that execute it are still unknown. Because EGF can activate both the phosphatidylinositol-3-OH kinase (PI(3)K)/Akt and the mitogen-activated protein kinase (MAPK) pathways, we speculated that Skp2 might be the phosphorylation target of one of these pathways. We therefore tested whether Akt/PKB might be a Skp2 kinase. Skp2 was found to interact with Akt1 in reciprocal co-immuno-precipitation experiments (Fig. 1a–c). Interaction between endogenous

1Cancer Biology and Genetics Program, Department of Pathology, Sloan-Kettering Institute, Memorial Sloan-Kettering Cancer Center, 1275 York Avenue, New York, New York 10021, USA. 2Cancer Genetics Program, Beth Israel Deaconess Cancer Center and Department of Medicine and Pathology, Beth Israel Deaconess Medical Center, Harvard Medical School, 330 Brookline Avenue, Boston, Massachusetts 02215, USA. 3Department of Molecular and Cellular Oncology, The University of Texas M. D. Anderson Cancer Center, Houston, Texas 77030, USA. 4Molecular Pharmacology and Chemistry Program, 5Molecular Biology Program, Sloan-Kettering Institute, Memorial Sloan-Kettering Cancer Center, 1275 York Avenue, New York, New York 10021, USA. 6Department of Molecular and Cellular Biology, Medical Institute of Bioregulation, Kyushu University, Fukuoka, Fukuoka 812-8582, Japan. 7Present address: Department of Cancer Biology, Meharry Medical College, 1005 Dr D. B. Todd Jr Boulevard, Nashville, Tennessee 37067-3599, USA.8Correspondence should be addressed to P.P.P. or H.-K. L. (e-mail: [email protected]; [email protected])

Received 26 August 2008; accepted 15 December 2008; published online 8 March 2009; DOI:10.1038/ncb1849





A RT I C L E S

Skp2 and Akt1 was detected on insulin-like growth factor-1 (IGF-1) stimulation, whereas the interaction was abolished by the PI(3)K inhibi-tor LY294002, suggesting that Akt activity may favour the formation of the Akt/Skp2 complex (Fig. 1d). In support of this notion, we found that an Akt1-kinase-dead mutant (K179A) interacted with exogenous Skp2 much less effectively than the constitutively active Akt1 (data not shown). In glutathione S-transferase (GST) pulldown assays, Akt1 was able to interact with Skp2 directly (Fig. 1f).

We next determined whether Skp2 was a substrate for Akt1 in vitro. Skp2 was readily phosphorylated by recombinant active Akt1 (Fig. 2a). Phosphorylation of Skp2 by Akt1 was comparable to phosphorylation of TSC2, a well-known Akt substrate, by Akt1 (Supplementary Information, Fig. S1b)15–18. The Scansite program19 (http://scansite.mit.edu) identifies Skp2 Ser 72 (S72) within an Akt consensus site (RXRXXS/T, where X is any amino acid) identified at medium stringency, which is conserved from rat to human (Fig. 2b). To determine whether S72 is a site for Akt-mediated Skp2 phosphorylation, we mutated this residue from serine to alanine (S72A) and used this Skp2 mutant in kinase assays in vitro. Indeed, Akt-mediated phosphorylation of Skp2 S72A was markedly reduced (Fig. 2c), even though Skp2 S72A still interacted with Akt as efficiently as wild-type (WT) Skp2 (Fig. 1e). Similarly, in vivo phospho-labelling experiments and western blot analysis with a phospho-Akt

substrate antibody revealed that a constitutively active Akt1 (Mri-Akt) phosphorylated WT Skp2 but not Skp2 S72A (Fig. 2d, e).

To verify whether S72 is phosphorylated by Akt, we next performed mass spectrometry analysis. Skp2 phosphorylated in vivo (Skp2-P) and unphosphorylated Skp2 (Skp2-C) was isolated from 293T cells, digested with trypsin and analysed by matrix-assisted laser desorption ionization– reflectron time-of-flight (MALDI–reTOF) mass spectrometry (Fig. 2f; Supplementary Information, Fig. S1a). Peptide patterns were then com-pared for differences. One m/z peak, at 1,671.88 atomic mass units (amu), was observed in the spectra of ‘Skp2-P’ that was absent from those of ‘Skp2-C’. The m/z value mapped to the predicted, monophosphorylated fragment of the Skp2 sequence (LKS72KGS75DKDFVIVR) with monoi-sotopic (12C) mass discrepancies of less than 40 p.p.m. Next, the same peptide was selectively retrieved by metal-affinity chromatography20 and reanalysed by MALDI–TOF/TOF tandem mass spectrometry (MS/MS) sequencing. The presence of unique fragment ions confirmed the identity and the monophosphorylation state (characteristic loss of 98 amu) for the peptide and allowed us to narrow down the site of phosphorylation to either S72 or S75. In contrast, in similar experimental conditions, the phosphorylated peptide at 1,671.88 amu was not identified when analys-ing a Skp2 S72A mutant (designed ‘S72A-P’), whereas the non-phosphor-ylated mutant peptide was detected, strongly suggesting that S72 is the site

Akt1

IP: XP

XP-Skp2

– +

IP: XP IP: XP Input

+ –

+ –

+

+ – +

+– +

+

Mri-Akt + +

ba c

IP: IgG Skp2

Akt

XP-Skp2

– + +

IgG

GST

GST

–Akt

GST

GST

–Akt

20%

inpu

t

GST–Akt

GST

d

XP-Skp2

IP: HA

XP-Skp2

Akt1

Akt1

XP-Skp2

HA–Mri-Akt

+ – +

+– +

+ ++ +

XP-Skp2

HA–Mri-Akt

Cell lysates Cell lysates

Cell lysates Cell lysates

XP-Skp2

Akt1

XP-Skp2

Akt1

e f

– – +

pAkt

Akt

XP-Skp2

Akt

Akt

Vect

or

XP-S

kp2

XP-S

72A

XP-S

72D

IP: XP

XP-Skp2

Mri-Akt

50

50

Mr(K) Mr(K)

Mr(K)

Mr(K)

Akt1

XP-Skp2

XP-Skp2

Skp2

IGF-1LY

50

50

75

50

50

50

50

25

75

37

100

35S-Skp2

Figure 1 Skp2 interacts with Akt. (a, b) 293T cells were transfected with the indicated plasmids and harvested for co-immunoprecipitation experiments. IP, immunoprecipitation; XP, Xpress; HA, haemagglutinin. (c) Exogenous Skp2 interacts with endogenous Akt1. 293T cells were transfected with the indicated plasmids and harvested for co-immunoprecipitation experiments. (d) Endogenous Skp2 interacts with endogenous Akt1. 293T cells were serum-starved for 1 day, treated with IGF (100 ng ml−1) in the presence or absence of LY294002 (20 µM) for 6 h, and harvested for co-immunoprecipitation

experiments and western blot analysis. (e) Analysis of the interaction between Akt and WT Skp2 or Skp2 mutants. 293T cells were transfected with XP-Skp2, XP-Skp2 S72A (XP-S72A) or XP-Skp2 S72D (XP-S72D) along with Mri-Akt were immunoprecipitated with XP antibody, washed, and subjected to western blot analysis. (f) Akt1 interacts with Skp2 in vitro. GST–Akt1 proteins were incubated overnight with in vitro translated 35S-Skp2 at 4 °C, washed with PBS (lanes 2 and 3) or NETN buffer (lanes 3 and 4), and subjected to 8% SDS–PAGE. Uncropped images of blots are shown in Supplementary Information, Fig. S16.



A RT I C L E S

for Akt-mediated Skp2 phosphorylation in vivo. Similar results were also obtained when using Skp2 phosphorylated in vitro by recombinant Akt1 (data not shown, and Supplementary Information, Fig. S1c, d).

Using a phospho (S72)-Skp2 specific antibody that we generated, we found that Mri-Akt could phosphorylate Skp2, but not Skp2 S72A (Fig. 2g). In PC-3 prostate cancer cells, where Akt is constitutively

a b

– + – +

– + – +

P P R K R L K

P P R K R L K

P P R K R L K

P P R K R L K

P P R K R L K

P P R K R V K

K G S D

K G S D

K G N D

K G N D

K G S D

K G S D

h

– +

S

S

S

S

S

S

G

R X R X X /T

– +

– +

– + + +

ji

c

d

g

i

– + +

Mr(K)

32P-GFP–Skp2

Rec AktGFP–32P-Skp2

GFP–Skp2

GFP–32P-Skp2

p72-Skp2

Skp2

Skp2 S72A

pAkt

GFP–Skp2

pAkt

Mri-Akt

Mri-Akt

Rec Akt

75 Akt concensus motif:

Human:

Monkey:

Dog:

Cow:

Rat:

Mouse:

WT S72A

WT S72A WT S72A

IP: GFP

Cell lysates

IP: Skp2 IP: Skp2

p72-Skp2

Skp2

pAkt

p72-Skp2

Skp2

pAkt

pS6

S6

Rapa

Akt

LY

– + – +

e

P-Skp2

Skp2

pAkt

Mri-Akt

Cell lysates

150

75

50

37

25

15

f

1 2 3Mr(K)

GFP–Skp2IgG

– +LY

p72-Skp2

Skp2

pAkt

p72-Skp2

Skp2

pAkt

β-Actin

WB: p72-Skp2

WB: Skp2

Cell lysates

IP: Skp2

DM

SO

DM

SO

WN

U01

26EGF

Figure 2 Akt/PKB phosphorylates Skp2 at S72 in vitro and in vivo. (a) Skp2 is phosphorylated by recombinant active Akt1 (Rec Akt) in vitro. GFP–Skp2 was immunoprecipitated from 293T, incubated with recombinant active Akt for 30 min, and subjected to SDS–PAGE analysis. (b) Skp2 contains a conserved Akt consensus site in rat, cow, dog, monkey and human, but not in mouse. (c) Skp2, but not Skp2 S72A, is phosphorylated by Rec Akt in vitro. GFP–Skp2 or GFP–Skp2 S72A was immunoprecipitated from 293T cells, incubated with recombinant active Akt for 30 min, and subjected to SDS–PAGE analysis. (d, e) Skp2, but not Skp2 S72A, is phosphorylated by Akt in vivo. Skp2 phosphorylation was determined by phospho-labelling (d) and Akt phospho-substrate antibody (e). (f) Akt induces Skp2 phosphorylation at S72 in vivo. GFP–Skp2 or GFP–Skp2 S72A was isolated from 293T cells transfected with GFP–Skp2 alone (lane 1, Skp2-C), GFP–Skp2 and Mri-Akt (lane 2, Skp2-P) or GFP–Skp2 S72A and Mri-Akt (lane 3, S72A-P), and analysed by MALDI–reTOF mass spectrometry (see Methods). Mass

spectrometry analysis revealed that the LKS72KGS75DFVIVR peptide from GFP–Skp2, but not the A72KGS75DFVIVR peptide from GFP–Skp2 S72A, was phosphorylated at S72 by Akt in vivo. (g) Akt induces Skp2 phosphorylation at S72 in vivo. 293T cells were transfected with the indicated plasmids, and harvested for western blot analysis using an anti-phospho (S72)-Skp2 antibody. (h, i) Endogenous Skp2 is phosphorylated by Akt at S72. PC-3 prostate cancer cells treated for 5 h with 20 µM LY294002 (LY, h) or 10 nM rapamycin (Rapa, i) were harvested for immunoprecipitation (IP), followed by western blot analysis with an anti-phospho-(S72)-Skp2 antibody. (j) EGF induces Skp2 phosphorylation at S72 by means of Akt. 293T cells were serum-starved (0.1% FBS) for 2 days, treated with EGF (100 ng ml−1) in the presence or absence of wortmannin (100 nM, WN) or U0126 (20 µM) for 3 h, and harvested for western blot analysis with an anti-phospho-(S72)-Skp2 antibody. DMSO, dimethylsulphoxide. Uncropped images of blots are shown in Supplementary Information, Fig. S16.



A RT I C L E S

b

d e

g

f

0

10

20

30

40

c

250150100

75

50

37

a

– + + + + + +

0

100

200

300

0 2 4 6 8

MockSkp2S72AS72D

MockSkp2S72AS72D

Time (days)

0

200

400

600

800

1,000

1,200

1 2 3 4 5 6Time (weeks)

* ** *

* *

* ** *

*

* *

** *

Mock Skp2 S72A S72D

**

Cel

l num

ber

(×10

4 )

p27

Skp2

β-Actin

Moc

k

Skp2

S72A

S72D

Moc

k

Skp2

S72A

S72D

Skp2

+LY

Skp2

+WN

Mock Skp2 S72A S72D Skp2+LY Skp2+WN

Tum

our

volu

me

(mm

3 )

Cel

ls in

S p

hase

(per

cent

age)

Brd

UD

AP

IMr(K) M

ock

Moc

kSk

p2S7

2AS7

2D

Moc

k

Skp2

S72A

S72D

Skp2

+LY

Skp2

+WN

p27-(UB)nIP: Nickel

columnWB: p27

Total cell lysates

(His)6-UB

p27

Skp2

p27

Skp2

β-Actin

1.0 0.3 0.8 0.3

Figure 3 Phosphorylation of Skp2 is required for Skp2 E3 ligase activity and function. (a) Skp2 and the Skp2 S72D mutant, but not Skp2 S72A, promote ubiquitylation of endogenous p27. 293T cells were transfected with the indicated plasmids, treated with vehicle, 20 µM LY294002 (LY) or 100 nM wortmannin (WN) along with 10 µM MG132 for 6 h and harvested for an in vivo ubiquitylation assay (see Methods). (b) Downregulation of p27 protein expression by Skp2 and Skp2 S72D, but not Skp2 S72A (S72A). 293T cells were transfected with the indicated plasmids for 48 h and harvested for western blot analysis. The numbers represent the relative intensity of p27 normalized by β-actin. (c) Loss of Skp2 phosphorylation at Ser72 significantly compromises an Skp2-mediated increase in S-phase cells. 293T cells were transfected with the indicated plasmids in a serum-starved condition (0.1% FBS) for 24 h, refreshed with 10% FBS for 16 h, incubated with 20 µM bromodeoxyuridine (BrdU) for 1 h, and harvested for quantification of bromodeoxyuridine incorporation. Cells (200–300) were scored and a representative result is shown as means ± s.d. from three independent experiments. Two asterisks,

P < 0.01 (Student’s t-test, n = 3). WN, wortmannin; DAPI, 4,6-diamidino-2-phenylindole. Scale bar, 50 µm. (d) S72 phosphorylation is critical for Skp2-mediated cell proliferation. LNCaP prostate cancer cells mock-transfected or stably transfected with Skp2, Skp2 S72A or Skp2 S72D were plated in 12 wells for cell growth analysis. Results are presented as means ± s.d. from a representative experiment performed in triplicate. Asterisk, P < 0.05, **P < 0.01 (Student’s t-test, n = 3). (e) p27 protein levels are regulated by Skp2 and Skp2 S72D, but not Skp2 S72A. p27 protein expression in LNCaP cells stably transfected with Mock, Skp2, Skp2 S72A, or Skp2 S72D. (f, g) S72 phosphorylation is critical for Skp2-mediated tumorigenesis. LNCaP cells mock-transfected or stably transfected with Skp2, Skp2 S72A or Skp2 S72D were injected into nude mice (n = 5 for each group) and followed up for tumorigenesis (see Methods). Pictures in g were taken 4 weeks after injection. Results are presented as means ± s.d. from two independent experiments. Asterisk, P < 0.05; two asterisks, P < 0.01 (Student’s t-test). Uncropped images of blots are shown in Supplementary Information, Fig. S16.



A RT I C L E S

active, endogenous Skp2 was phosphorylated at S72 (Fig. 2h), whereas LY294002 drastically decreased its phosphorylation (Fig. 2h).

Because the p70 S6 kinase (S6K) is activated by Akt, we determined whether S6K could mediate Akt-induced phosphorylation of Skp2 S72.

Rapamycin, a well-known potent inhibitor of mTOR, abrogated S6K activity towards S6, whereas phosphorylation of Skp2 S72 was unaffected (Fig. 2i), suggesting that S6K is not involved in the Akt-dependent phos-phorylation of Skp2 S72.

50

25

37

50

50

a b

– + + +

Inpu

t

c

– – + –

IP: I

gG

IP: I

gG

IP: S

kp1

IP: S

kp1

His

dep

letio

nSk

p1 d

eple

tion

Inpu

tM

ock

XP-S

kp2

XP-S

72A

XP-S

72D

– – – +

– – +– + –

– + –

– – +

– ++ +

1

f

g

– + – +

1 5.4 0.9 1.5

h i

– + – – + – – + – – + –

– – + – – + – – + –– +

d e

Skp1

Cul1

Skp2

Skp1

Skp1

Cul1

Cul1

XP-Skp2

XP-Skp2

14-3-3β

Mri-Akt

Cul1

Cul1

His-Skp2

IP: Nickel column

1.0 2.8 0.4

1.0 1.2 0.41.0 0.2 1.9

1.0 0.7 1.2

1.0 1.4 0.0 1.0 1.02.1

Cell lysatesCell lysates

Cell lysatesCul1

Skp1

His-Skp2

Mri-Akt

Mri-Akt

Mri-Akt

LY

WN

Skp2

Skp1

Skp1

pAkt

IP: XP

p72-Skp2

Cul-1

LY

Skp1

p72-Skp2

Cul-1

Skp2

Skp1lysates

IP

His Skp1

IP: XP

Skp1/Skp2

Skp1

Cul-1

Skp1

Cul-1Skp1

Cul-1

Flag–Skp2 Flag–Skp2

Flag–Skp2

Flag–∆N-Skp2 Flag–∆N-Skp2

Flag–∆N-Skp2

Lamin B1

α-Tubulin

p72–Skp2

XP-Skp2

XP-Skp2CIP

Cul-1/Skp2

Input IP: Flag

0.2

1 0.6

GST GST–Skp1

XP-Skp2 XP-72A

Mri-Akt

Skp2 Input

GST

GST–Skp1

Skp2

Input IP: Flag

Cytosolic Nuclear Cytosolic Nuclear

Figure 4 Akt-mediated Skp2 phosphorylation regulates SCF complex formation. (a) Akt promotes the interaction of exogenous Skp2 with endogenous Skp1 and Cul-1. 293T cells transfected with the indicated plasmids were serum-starved, treated with or without 100 nM wortmannin for 6 h, and harvested for pulldown with a nickel column. (b) Endogenous SCF complex formation is positively regulated by Akt. 293T cells were transfected with the indicated plasmids, serum-starved, treated with 20 µM LY294002 for 6 h, and harvested for co-immunoprecipitation experiments. (c) S72 phosphorylation is critical for SCF complex formation. 293T cells transfected with indicated plasmids and Mri-Akt were serum-starved and harvested for co-immunoprecipitation experiments. (d) The Skp2-SCF complex contains phosphorylated Skp2 S72. 293T cells were mock-transfected or transfected with Mri-Akt, serum-starved, treated with 20 µM LY294002, and harvested for co-immunoprecipitation experiments. (e) Skp1 depletion decreases the amount of phospho-(S72)-Skp2 proteins in the cell lysates. Total cell lysates from 293T transfected with Skp2 and Mri-Akt were immunoprecipitated with anti-His antibody or Skp1 antibody, washed, and subjected to western blot analysis. (f) Treatment with CIP disrupts

the Skp1 and Skp2 interaction. Cell lysates from 293T cells transfected with XP-Skp2 were immunoprecipitated with an XP antibody, treated with CIP for 1 h, washed, and subjected to western blot analysis with an anti-phospho-(S72)-Skp2 antibody. (g) Phosphorylation of Skp2 S72 regulates Skp2 and Skp1 interaction in vitro. GST–Skp1 bound to glutathione-agarose beads was incubated with cell lysates from 293T cells transfected with the indicated plasmids in the serum-free medium, washed, and subjected to western blot analysis. (h) Flag–∆N-Skp2 forms a SCF complex much more efficiently than the WT Skp2. 293T cells transfected with Flag–Skp2 (residues 1–424) or Flag–∆N-Skp2 (residues 91–424) were lysed by RIPA buffer containing 1% Nonidet P40, immunoprecipitated with anti-Flag antibody and washed, followed by western blot analysis. (i) Flag–∆N-Skp2 participates much more efficiently in SCF complex formation than its WT counterpart in both cytosol and nucleus. Cytosolic and nuclear fractions prepared from 293T cells transfected with Flag–Skp2 or Flag–∆N-Skp2 were immunoprecipitated and washed, followed by western blot analysis. Uncropped images of blots are shown in Supplementary Information, Fig. S16.



A RT I C L E S

In addition, treatment with EGF induced the activation of Akt and phosphorylation of Skp2 S72, which was inhibited by wortmannin but not by the MAP kinase/ERK kinase 1 (MEK1) inhibitor U0126, sug-gesting that the EGF-mediated phosphorylation of Skp2 S72 occurs

through an Akt-dependent pathway independently of MAPK–p90 ribosomal S6 kinase (RSK) activation (Fig. 2j). These results provide strong support for the direct phosphorylation of Skp2 S72 by Akt/PKB both in vitro and in vivo.

c

a

e

b

10

0

20

30

40

10

0

20

30

40

0 1 3 6 1 3 6 0 1 3 6 1 3 6IGF-1 (h)

IGF-1 (h)p72-Skp2

Skp2

pAkt

Skp2

pAkt

0 0.5 1 2 3 64.5 0 0.5 1 2 3 64.5

e f

0

0.40.81.2

1.8

+ –+–+ +––

+ –+–+ +––

– + + + – + + +

d

– – LY WN LY WN– –

75

50

50

50

50

Skp2 Mri-Akt DAPI Merge Skp2 DAPI Merge

Moc

k

Veh

icle

IGF-

1IG

F-1+

LYLY

Moc

kM

ock

Mri-

Akt

Mri-

AktXP

-Skp

2X

P-S

72D

XP

-S72

A

Cel

ls w

ith c

ytos

olic

S

kp2

(per

cent

age)

Cel

ls w

ith c

ytos

olic

S

kp2

(per

cent

age)

XP-Skp2 XP-S72A XP-S72A

MockMri-Akt

Vehicle IGF-1 IGF-1+LY LY

Vehicle IGF-1 IGF-1+LY LY

Nuclear Cytosolic

Skp2

Sp3

pAkt

Skp2

pAkt

α-TubulinSkp2

α-Tubulin

α-Tubulin

Lamin B1

Lamin B1

pAkt

IGF-1

IGF-1LY

Skp

2 C

/N r

atio

p72-Skp2

Cytosolic Nuclear

Cytosolic Nuclear

Mr(K)

DMSO WNDMSO WN DMSO WN

Figure 5 Skp2 phosphorylation by Akt regulates the cytosolic localization of Skp2. (a) S72 phosphorylation mediates Akt-induced cytosolic localization of Skp2. 293T cells were transfected with the indicated plasmids, and cells were harvested for immunofluorescence analysis. Cells (100–200) were scored; a representative result is shown from three independent experiments. (b, c) Activation of the PI(3)K/Akt pathway promotes the cytosolic localization of endogenous Skp2. 293T cells were serum-starved (0.1% FBS) overnight, treated with vehicle or IGF-1 in the presence or absence of LY294002 or wortmannin for 6 h, and harvested for immunofluorescence (b) or fractionation, followed by western blot analysis (c). Scale bars, 10 µm. (d) Phospho-S72 Skp2 localizes in the cytosol. 293T cells were serum-starved overnight, treated

with vehicle or IGF-1 in the presence or absence of LY294002 or wortmannin for 6 h, and harvested for fractionation, followed by western blot analysis with a phospho-(S72)-Skp2-specific antibody. (e) Kinetics of phosphorylation of Skp2 S72. 293T cells were serum-starved for 2 days, treated with IGF (100 ng ml−1) in the presence or absence of wortmannin for the indicated durations and harvested for western blot analysis with an anti-phospho-(S72)-Skp2 antibody. (f) Kinetics of cytosolic localization of Skp2. 293T cells were serum-starved (0.1% FBS) for 2 days, treated with IGF in the presence or absence of wortmannin for the indicated durations and harvested for fractionation and western blot analysis. Uncropped images of blots are shown in Supplementary Information, Fig. S16.



A RT I C L E S

S72 phosphorylation is critical for the biochemical and proto-oncogenic functions of Skp2We next determined whether phosphorylation of Skp2 S72 modulates its E3 ligase activity towards p27. WT Skp2 and the phosphomimetic Skp2 S72D mutant readily promoted the ubiquitylation of endogenous and exogenous p27, whereas Skp2 S72A was profoundly impaired in this function (Fig. 3a; Supplementary Information, Fig. S2a). Although phosphorylation of p27 at T187 is proposed as an important event for the interaction of p27 with the Skp2-SCF complex1, we found, surpris-ingly, that Skp2-mediated ubiquitylation of WT p27 and p27 T187A

was comparable (Supplementary Information, Fig. S3a). Similarly, the interaction of Skp2 with WT p27 and p27 T187A was also comparable (Supplementary Information, Fig. S3b), which is consistent with the results from the Roberts group21 showing that the interaction of p27 with Skp2 regulated by p27 T187 phosphorylation is dependent on the cell type.

Whereas WT Skp2 and Skp2 S72D promoted the degradation of p27, Skp2 S72A markedly decreased this activity (Fig. 3b; Supplementary Information, Fig. S2b). It should be noted that within this time frame there was no significant change in cell-cycle distribution, suggesting

75

37

50

25

75

ca b

Inpu

t

+ ++–+ –––– +––

+––

– ––– +

0

10

20

30

40

d

– ––– –

+–––+

1.0 0.2 0.0

1.0 1.2 0.3 0.1 0.1

0.9

– + + +– –

f+ – +

e

Skp2

14-3-3

14-3-3-β

His-Skp2

14-3-3-β

Skp2

His-Skp2Mri-Akt

WNLY

WN LYPTEN

IP:

IgG

IP:

IgG

IP:

IgG

IP: 1

4-3-

3β

IP: 1

4-3-

3γ

IP: 1

4-3-

3θ

pAkt

Akt

IGF-1

IP: Nickel column IP: Skp2

Cell lysates

Cell lysates

GFP–35S-Skp2

GST–14-3-3β

IP: 14-3-3β

GST

Rec Akt

Input GST

GST

–14-

3-3β

XP-S

kp2

XP-S

72A

XP-S

72D

Moc

k

Mr(K)

XP-Skp2

14-3-3β

XP-Skp2

14-3-3β

Skp2 Mri-Akt DAPI Overlay siRNA

Lamin A/C

14-3-3β

14-3-3γ

14-3-3β 14-3-3γ

Moc

kM

ri-A

kt

Cel

ls w

ith c

ytos

olic

Skp

2 (p

erce

ntag

e)

Mock

Mri-Akt

siRNA Lamin A/C

Figure 6 14-3-3β interacts with Skp2 and is essential for Akt-mediated Skp2 relocalization in the cytosol. (a) Skp2 preferentially interacts with the 14-3-3β isoform. 293T cell lysates were immunoprecipitated with IgG, 14-3-3β, 14-3-3γ or 14-3-3θ antibody, washed, and harvested for western blot analysis. (b) The PI(3)K–Akt pathway positively regulates the interaction of Skp2 with 14-3-3β. 293T cells were transfected with the indicated plasmids, and cells were harvested for co-immunoprecipitation experiments. (c) IGF-1 promotes the interaction between endogenous Skp2 and 14-3-3β. 293T cells were serum-starved for 1 day, treated with IGF in the presence or absence of LY294002 or wortmannin (WN) for 6 h, and harvested for co-immunoprecipitation experiments and western blot analysis. (d) 14-3-3β interacts with Skp2 directly in vitro. GST–14-3-3β

bound to glutathione-agarose beads were incubated with in vitro translated GFP–35S-Skp2, which was preincubated for 30 min with recombinant Akt kinase at 30 °C, washed and subjected to 8% SDS–PAGE, followed by autoradiography. (e) Phosphorylation-dependent regulation of Skp2 interaction with 14-3-3β. 293T cells were transfected with plasmids as indicated, and cells were harvested for co-immunoprecipitation. (f) 14-3-3β, but not 14-3-3γ, mediates Akt-dependent cytosolic localization of Skp2. 293T cells were transfected with small interfering RNA (siRNA) and plasmids as indicated, and harvested for immunofluorescence analysis. Cells (100–200) were scored and a representative result from three independent experiments is shown. Scale bar, 10 µm. Uncropped images of blots are shown in Supplementary Information, Fig. S16.



A RT I C L E S

that p27 degradation is not caused by variation in cell cycle status (Supplementary Information, Fig. S2c). The inability of Skp2 S72A to promote the ubiquitylation and degradation of p27 was also not due to its inability to interact with p27 or a decrease in its expression level (Fig. 3b; Supplementary Information, Fig. S2d). Furthermore, LY294002 or wortmannin suppressed Skp2-mediated p27 ubiquitylation (Fig. 3a). These results suggest that Akt activity and phosphorylation of Skp2 S72 promote Skp2 E3 ligase activity.

Skp2 overexpression is known to promote cell proliferation by triggering p27 degradation6,7. We therefore determined whether S72 phosphorylation regulates Skp2-mediated cell proliferation. Both Skp2 and the S72D mutant increased the number of cells in S phase, whereas Skp2 S72A was markedly impaired in this activity (Fig. 3c; Supplementary Information, Fig. S3c). LY294002 or wortmannin suppressed the Skp2-mediated increase in the number of S-phase cells (Fig. 3c).

0

40

80

120

160

200

100

b

0 WT Skp2–/–

a

040

80

120160

WT

c d

1.0 0.3 1.2

75

50

* * * * * * *

* *

* * * *

pBabe

S72A

Skp2

S72D

WT Skp2–/–

WT Skp2–/–

pBabe Skp2 S72A S72D Mr(K)

Skp2

GRP78

Cel

l num

ber

/fie

ldC

ell n

umb

er/f

ield

Cel

l num

ber

/fie

ld

MSCV MSCV

Skp2–/–

Skp2–/–MSCV-Skp2-NES

Skp2-NES DAPI Overlay

p27

Skp2

β-Actin

Moc

k

Skp2

Skp2

-NES

MSC

VM

SCV-

Skp2

-NES

Figure 7 Cytosolic Skp2 positively regulates cell migration. (a) Skp2 and Skp2 S72D, but not Skp2 S72A, promotes cell migrations. Wild-type MEFs were infected with pBabe, pBabe-Skp2, pBabe-Skp2 S72A or pBabe-Skp2 S72D selected by puromycin (2 µg ml−1) for 3 days, and plated for cell migration assays (Methods). Migrated cells were counted from three random fields, and means ± s.d. were calculated. GRP78 (glucose regulated protein 78) served as a loading control. Representative results from two independent experiments are shown. Two asterisks, P < 0.01; three asterisks, P < 0.001 (Student’s t-test, n = 3). Scale bar, 50 µm. (b) Loss of Skp2 impairs cell migration. Skp2 WT and Skp2−/− primary MEFs were plated for cell migration assays. Migrated cells were counted from three random fields, and means ± s.d.

were calculated. Representative results from three independent experiments are shown. Two asterisks, P < 0.01 (Student’s t-test, n = 3). Scale bar, 50 µm. (c) Cytosolic Skp2 restores cell migration defect in Skp2−/− MEFs. Skp2 WT and Skp2−/− primary MEFs were infected with the indicated viral constructs for 2 days, selected with puromycin for 3 days, and plated for cell migration assays. Scale bar, 50 µm. (d) Skp2-NES shows cytosolic localization and does not induce p27 degradation. 293T cells were transfected with XP-Skp2-NES for 2 days and harvested for immunofluorescence and western blot analysis. Scale bar, 10 µm. Note that the lanes in d are from the same gel; the line indicates that the gel was cut. Uncropped images of blots shown in d are available in Supplementary Information, Fig. S16.



A RT I C L E S

To study the role of Skp2 and Skp2 mutants in tumorigenesis, we gen-erated LNCaP prostate cancer cells with stable overexpression of Skp2 or Skp2 mutants for tumorigenic potential. Overexpression of WT and Skp2 S72D, but not Skp2 S72A, markedly enhanced LNCaP prolifera-tion and decreased p27 protein levels (Fig. 3d, e). Whereas Skp2 and Skp2 S72D profoundly promoted tumorigenesis in comparison with the mock control, Skp2 S72A completely lost this activity (Fig. 3f, g).

Skp2 phosphorylation regulates SCF complex formation and activityBecause the integrity of Skp2-SCF complex is critical for Skp2 E3 ligase activity, we next determined whether activation of the PI(3)K/Akt pathway could regulate the Skp2-SCF complex. Mri-Akt enhanced the interaction of exogenous Skp2 with endogenous Skp1 and Cul-1, whereas wortmannin suppressed the formation of the Skp2–Skp1 complex (Fig. 4a). In this cell culture condition, we did not observe an appreciable change in S-phase cell numbers by either the activation or inhibition of Akt (Supplementary Information, Fig. S4).

Similarly, suppression of the PI(3)K/Akt pathway profoundly impaired endogenous Skp2-SCF complex formation, whereas we observed a slight increase in the formation of this complex after activation of Akt (Fig. 4b). In agreement with these findings, WT Skp2 and Skp2 S72D readily com-plexed with endogenous Skp1 and Cul-1, whereas Skp2 S72A signifi-cantly decreased its interaction with Skp1 and Cul-1 (Fig. 4c).

The Skp2-SCF complex isolated from cells transfected with WT Skp2 or its mutants was used in an in vitro p27 ubiquitylation assay. The Skp2-SCF complex readily promoted p27 ubiquitylation (Supplementary Information, Fig. S5a): the complex failed to induce p27 ubiquitylation unless p27 had been phosphorylated by the cyc-lin A–Cdk2 complex before this event (Supplementary Information, Fig. S5a). The ability of Skp2 S72A to complex with Skp1 and Cul-1 and promote p27 ubiquitylation was markedly impaired when compared with that of WT Skp2 and the Skp2 S72D mutant (Supplementary Information, Fig. S5a). Furthermore, Mri-Akt sig-nificantly increased Skp2-SCF complex formation and p27 ubiquit-ylation, whereas LY294002 decreased this effect (Supplementary Information, Fig. S5b). We further showed that the SCF complex con-tained Skp2 phosphorylated on S72 (Fig. 4d, e), whereas calf intesti-nal alkaline phosphatase (CIP) induced Skp2 S72 dephosphorylation and disrupted the interaction between Skp2 and Skp1 (Fig. 4f). In addition, the lysates depleted in Skp1 decreased the phosphoryla-tion of Skp2 S72 and Cul-1 (Fig. 4e). GST pulldown assays showed that whereas Mri-Akt, which induced phosphorylation of Skp2 S72 (Fig. 2d, g), greatly increased Skp1 and Skp2 interaction in vitro, Skp2 S72A markedly decreased its interaction with Skp1 (Fig. 4g). These results strongly suggest that Akt positively regulates Skp2-SCF complex formation, therefore enhancing its E3 ligase activity through phosphorylation of Skp2 S72.

Akt-mediated phosphorylation induces cytosolic localization of Skp2Skp2 normally resides in the nucleus; however, cytosolic relocalization of Skp2 has been observed in several human cancers22–25. However, the molecular mechanisms underlying this phenomenon and its biological consequence are currently unknown. Loss or mutation of the tumour suppressor PTEN is frequently observed in human cancers, resulting in

marked activation of Akt26,27. Akt/PKB may represent a potential candi-date for triggering the relocalization of Skp2 to the cytosol, given that phosphorylation of Akt substrates by Akt often results in a change in protein localization28,29. Immunofluorescence revealed that although Skp2 was localized in the nucleus, Mri-Akt induced the relocalization of Skp2 to the cytosol in about 40% of cells. In contrast, this effect was markedly

c

020406080

100

020406080

100120

010203040

506070

a

b

Skp2 pAkt PTEN

Skp2 pAkt PTEN

Nor

mal

Nor

ma/

PIN

Can

cer

Can

cer

Skp2 nuclear/negativeSkp2 cytoplasmic

Skp2 nuclearSkp2 cytoplasmic

P < 0.001 P = 0.006

P = 0.03

P < 0.006 P < 0.008

Skp

2 lo

caliz

atio

n (p

erce

ntag

e)M

etas

tasi

s (p

erce

ntag

e)

Skp2 localization

pAkt lowpAkt high

PTEN low PTEN high PTEN low PTEN high

pAkt lowpAkt high

Prostate Colon

Figure 8 Cytosolic Skp2 correlates with activation of the Akt kinase, PTEN loss, and metastasis in human cancer specimens. (a, b) Cytosolic Skp2 localization correlates with high pAkt levels and PTEN loss in colon cancer (a) and prostatic intraepithelial neoplasia (PIN) (b). Colon cancer TMAs (84 cases) and prostate TMAs (101 cases) were stained for Skp2, pAkt and PTEN (Methods). Representative examples are shown. Scale bar, 50 µm. (c) Correlation of cytosolic Skp2 with high pAkt levels, PTEN loss and metastasis in colon and prostate cancers.



A RT I C L E S

reduced with Skp2 S72A (Fig. 5a; Supplementary Information, Fig. S6a). Skp2 S72D relocalized to the cytosol even without Mri-Akt, but the addi-tion of Mri-Akt did not enhance the accumulation of S72D in the cytosol (Fig. 5a; Supplementary Information, Fig. S6a; and data not shown). We further demonstrated that Akt activation on IGF-1 stimulation or Mri-Akt overexpression induced the relocalization of endogenous Skp2 to the cytosol, as determined by immunofluorescence and fractionation/western blot analyses (Fig. 5b, c, f; Supplementary Information, Fig. S6b). We found that phosphorylation of Skp2 S72 was detected predominantly in the cytosol and not in the nucleus (Fig. 5d).

We next determined the kinetics of phosphorylation of Skp2 S72 and cytosolic Skp2 localization after stimulation with IGF-1. IGF-1 induced the activation of Akt and the phosphorylation of Skp2 S72 at 30 min, which lasted for several hours (Fig. 5e). Strikingly, wortmannin mark-edly suppressed IGF-1-induced phosphorylation of Skp2 S72 (Fig. 5e). Cytosolic accumulation of Skp2 occurred as early as 1 h after IGF-1 treat-ment, and was significantly blocked by wortmannin (Fig. 5f). Within this time frame there was no significant change in cell cycle distribution after treatment with IGF-1 (Supplementary Information, Fig. S7). We obtained similar results in IMR90 cells. At 1 h after treatment with IGF, up to 75–80% of cells showed accumulation of Skp2 in the cytosol, although the cell cycle distribution was minimally affected (Supplementary Information, Fig. S8a, b, d). Once again, treatment with IGF led to the activation of Akt accompanied by phosphorylation of Skp2 S72, which was abrogated by wortmannin (Supplementary Information, Fig. S8c). Our results therefore suggest that Akt induces the phosphorylation of Skp2 S72, in turn resulting in the cytosolic relocalization of Skp2.

14-3-3β interacts with Skp2 and is required for Akt-mediated cytosolic relocalization of Skp2Akt phosphorylation is known to trigger the nuclear export of proteins. We therefore tested whether phosphorylation of Skp2 S72 would prevent its nuclear localization. In fact, the Skp2 sequence PRKRLKS (where S is at residue 72) was identified as a putative nuclear localization sequence (NLS) (Supplementary Information, Fig. S9a). We fused this putative NLS and its mutant sequences (S72A or S72D) to green fluorescent pro-tein (GFP) and examined their ability to promote the nuclear import of GFP. We found that neither mutant promoted the nuclear localization of GFP, whereas the NLS from simian virus 40 large T antigen fused to GFP did so readily (Supplementary Information, Fig. S9a). These results suggested that the PRKRLKS sequence from Skp2 might not function sufficiently as an NLS and that the Akt-dependent cytosolic accumula-tion of Skp2 may be due to enhanced nuclear export. In full agreement with this hypothesis, leptomycin B (LMB), which blocks the nuclear export of proteins, significantly prevented the Akt-mediated cytosolic localization of Skp2 (Supplementary Information, Fig. S9b). Although the PRKRLKS sequence from Skp2 is not functionally sufficient as a NLS in our assays, it is required for the nuclear localization of Skp2: it has been demonstrated that a Skp2 mutant devoid of this sequence primarily localizes to the cytoplasm30.

The 14-3-3 protein controls a variety of biological processes through the regulation of partner protein localization31. It binds to phospho-serine/phosphothreonine-containing motifs in a sequence-specific manner31. Because several Akt substrates interact with 14-3-3 after Akt-mediated phosphorylation, which in turn results in a change of protein localization32–35, we examined whether 14-3-3 mediates the cytosolic

localization of Skp2 after the activation of Akt. Co-immunoprecipitation experiments revealed that endogenous Skp2 interacted with endog-enous 14-3-3β but not with 14-3-3γ and 14-3-3θ (Fig. 6a), although we observed a weak interaction between exogenous Skp2 and 14-3-3γ (data not shown). Suppression of the PI(3)K/Akt pathway by LY294002 or PTEN significantly abolished this interaction (Fig. 6b). We further showed that the interaction of Skp2 with 14-3-3β was enhanced by IGF-1 through the PI(3)K/Akt-dependent pathway (Fig. 6c). GST pulldown assays revealed that 14-3-3β interacted with Skp2 directly and that Akt profoundly enhanced this interaction (Fig. 6d).

Whereas Skp2 and Skp2 S72D interacted effectively with 14-3-3β, Skp2 S72A compromised this interaction (Fig. 6e), suggesting that phos-phorylation of Skp2 S72 is critical for the interaction of Skp2 with 14-3-3β. We next determined whether 14-3-3β is required for the Akt-mediated cytosolic relocalization of Skp2. The 14-3-3β and 14-3-3γ small interfer-ing RNAs profoundly suppressed the expression of 14-3-3β and 14-3-3γ proteins, respectively (Supplementary Information, Fig. S10a). In par-ticular, silencing 14-3-3β expression inhibited the Akt-mediated cytosolic relocalization of Skp2 (Fig. 6f; Supplementary Information, Fig. S10b). Similarly, silencing 14-3-3β expression also decreased the cytosolic localization of Skp2 S72D (Supplementary Information, Fig. S10c). As expected, silencing 14-3-3γ expression failed to affect the Akt-mediated cytosolic relocalization of Skp2 (Fig. 6f). These results suggest that 14-3-3β is required for Skp2 cytosolic relocalization induced by Akt.

Cytosolic Skp2 mediates cell migrationSkp2 overexpression is frequently observed in advanced human can-cers. Skp2 may have a function in tumour invasion and metastasis. We examined whether Skp2 has a function in cell migration, which underlies these steps in tumour progression. Overexpression of Skp2 and Skp2 S72D in murine embryonic fibroblasts (MEFs) markedly enhanced cell migration, whereas Skp2 S72A did not (Fig. 7a). We next determined whether Skp2 is required for cell migration. Skp2−/− MEFs showed a profound defect in cell migration (Fig. 7b, c). Similarly, in wound scratch assays, loss of Skp2 significantly delayed wound closure (Supplementary Information, Fig. S11a). Because advanced cancers often possess cytosolic Skp2 (refs 22–25), we hypothesized that Skp2 relocali-zation to the cytosol would favour cell motility. To test this possibility, we created a Skp2–NES (nucleus export signal) construct in which the NES was fused to the carboxy terminus of Skp2. Immunofluorescence reveals that Skp2–NES was predominantly localized in the cytosol (Fig. 7d, upper panel). Restoration of Skp2–NES rescued the migration defect in Skp2−/− MEFs (Fig. 7c; Supplementary Information, Fig. S11a). This function of Skp2–NES is probably independent of its ability to pro-mote p27 degradation, because Skp2–NES did not form the Skp2-SCF complex efficiently and failed to induce downregulation of p27 (Fig. 7d; Supplementary Information, Fig. S11b).

Cytosolic Skp2 correlates with activated Akt levels, PtEN loss and metastasis in human cancersBecause Akt induced Skp2 cytosolic relocalization, we next determined whether Skp2 localization correlates with activated Akt (pAkt) and PTEN levels in cancer specimens. We immunoassayed Skp2, pAkt and PTEN in colon and prostate tumour microarrays (TMAs). Of 84 cases of colonic adenocarcinoma, 16 cases showed cytoplasmic Skp2 staining in infiltrating carcinoma cells, in contrast with normal colonic mucosa



A RT I C L E S

from the same whole section of tissue or from control tissue cores on the TMA; normal colonic mucosa showed scattered nuclear staining in the crypts and glands (Fig. 8a). Sequential sections showed high levels of pAkt (Fig. 8a, c; Supplementary Information, Fig. S12a; P < 0.003) and low levels of PTEN (Fig. 8a, c; Supplementary Information, Fig. S12a; P < 0.008) in these tumours. The correlation between cytoplasmic Skp2 and high levels of pAkt was also highly significant in prostatic adeno-carcinomas, and these cases also showed low PTEN levels (Fig. 8b, c; Supplementary Information, Fig. S12b; P < 0.001). In agreement with the role of cytosolic Skp2 in cell migration, cytosolic Skp2 strongly correlates with lymph node metastasis in colon cancers (Fig. 8c; Supplementary Information, Fig. S12a, P = 0.03). These results show that cytosolic Skp2 strongly correlates with pAkt levels, PTEN loss and metastasis in human cancers, further implying that Skp2 cytosolic relocalization may have a function in promoting tumour metastasis.

Evolutionary conservation of the functional and biochemical cross-talk of Akt-Skp2Skp2 S72 is conserved in many organisms from rat to primates (Fig. 2b). However, this residue is not conserved in mouse Skp2 (mSkp2), raising the question of when and how this regulatory pathway evolved. We there-fore tested the consequences of Akt activation in murine cells. We found that both Akt-mediated mSkp2 phosphorylation and the functional con-sequences of this event still occur in murine cells. Although Akt did trigger mSkp2 phosphorylation in vivo and even in vitro (Supplementary Information, Fig. 13a–c), it phosphorylated mSkp2 in vitro much less efficiently than the human Skp2 protein (Supplementary Information, Fig. 13a), demonstrating that mSkp2 is a poor direct substrate of Akt. However, SCF complex formation with mSkp2 was also enhanced by Mri-Akt (Supplementary Information, Fig. 13c). Akt-mediated cytosolic localization of mSkp2 was also observed (Supplementary Information, Fig. 13d). In agreement with these findings, in Pten−/− MEFs, in which Akt is constitutively active, mSkp2 accumulated in the cytosol and relo-calized in the nucleus after treatment with LY294002 (Supplementary Information, Fig. 13e). These results therefore support the notion that the Akt–Skp2 pathway is evolutionarily conserved even though Skp2 may not be a direct target of Akt in murine cells.

DiSCUSSiONWe show that phosphorylation of Skp2 S72 by Akt is required for Skp2 E3 ligase activity and for tumorigenesis. Phosphorylation of Skp2 S72 provides a molecular switch that orchestrates the formation of the SCF complex. The proto-oncogenic PI(3)K/Akt pathway can therefore regulate Skp2 protein levels through the enhancement of Skp2 transcription36 and stability (see ref. 30), its subcellular localiza-tion, and, importantly, the function of the SCF complex through Skp2 phosphorylation. Taken together, these findings identify PI(3)K/Akt signalling as a major regulatory pathway of Skp2 function at multiple levels. Very recently, Skp2 was shown to be phosphorylated by Cdk2 at S64; phosphorylation at this residue also regulates Skp2 stability37. However, we found that phosphorylation of Skp2 at S64 by Cdk2 was not required for formation of the Skp2-SCF complex (Supplementary Information, Fig. S14a).

It should be noted that varying Akt activity, by varying culture condi-tions, can affect the ability of Skp2 and Skp2 S72A to form a Skp2-SCF complex. This is because Skp2 S72A showed only a minor change in SCF

complex formation compared with WT Skp2 in 0.1% fetal bovine serum (FBS), in which Akt activity was low, whereas Skp2 S72A was impaired in forming a SCF complex on Mri-Akt overexpression in comparison with WT Skp2 (Supplementary Information, Fig. S15a).

In addition, the Scansite program also predicts Skp2 T21 as a pos-sible additional Akt phosphorylation site at low stringency (data not shown). We did indeed find that Akt-mediated phosphorylation of the Skp2 T21A mutant was partly decreased in vitro. Moreover, the ability of the Skp2 T21A mutant to form a SCF complex was also decreased (Supplementary Information, Fig. S15b, c). These results suggest that T21 site on Skp2 may be an additional site for the effect of Akt on for-mation of the Skp2-SCF complex, at least in vitro. However, because the stringency of this site as assessed by a Scansite analysis is low, it remains to be determined whether this site is at all relevant in vivo.

The crystal structure of the Skp2-SCF complex has been resolved38. However, the amino-terminal Skp2 tail (residues 1–108, encompassing the S72 sites) was omitted in these structural studies, suggesting that the N-terminal Skp2 tail is not required for the formation of the SCF complex itself. By contrast, it may contain an inhibitory domain that, in the context of the full-length protein, hinders the Skp2 interaction interface for Skp1 binding. Akt-mediated phosphorylation of Skp2 S72 may induce its conformational change, in turn allowing the interaction of Skp2 with Skp1. In support of this notion, a Skp2 mutant (residues 91–424) devoid of the N-terminal tail forms a SCF complex much more efficiently than WT Skp2 in the cytosol and the nucleus (Fig. 4h, i).

Skp2 normally localizes to the nucleus and aberrantly localizes to the cytosol during cancer progression by mechanisms that are as yet unknown22–25. We provide convincing evidence that cytosolic Skp2 local-ization is regulated by Akt through the phosphorylation of Skp2 S72. Although Skp2 is also phosphorylated by Cdk2 at S64 (ref. 37), phos-phorylation of Skp2 at S64 does not regulate the cytosolic localization of Skp2 (Supplementary Information, Fig. S14b). Our results showing that the cytosolic localization of Skp2 strongly correlates with pAkt level, PTEN loss and cancer metastasis suggest that cytosolic Skp2 may be important in tumour metastasis. In agreement with this notion, we find that Skp2 has an essential and dose-dependent function in cell migration, whereas a cytosolic Skp2 mutant restores the migration defects in Skp2−/− MEFs. In conclusion, we show that Akt-mediated Skp2 phosphorylation is critical for formation of the SCF complex, cytosolic localization of Skp2, and biological functions of Skp2, thus providing the molecular basis and the biological relevance underlying the aberrant cytosolic localization of Skp2 in tumorigenesis.

Note added in proof: a related manuscript by Gao et al. (Nature Cell Biol. 11, doi:10.1038/ncb1847; 2009) is also published in this issue.

MEtHODSMice, cell culture and reagents. MEFs from wild-type and Skp2−/− mice were pre-pared as described8,39. 293T, PC-3, COS-1 and IMR90 cells (from M. Pagano) were cultured in DMEM medium containing 10% FBS, and LNCaP cells were grown in RPMI medium containing 10% FBS. To clone Xp-Skp2, Skp2 was amplified by the polymerase chain reaction (PCR) using pcDNA3-Skp2 as a template and inserted into a pcDNA4/HisMax-TOPO vector (Invitrogen). To clone GFP–Skp2, Skp2 was amplified by PCR using pcDNA3-Skp2 as a template and inserted into pcDNA3.1/NT-GFP–TOPO vector (Invitrogen). pcDNA3-Skp2 S72A, pcDNA3-Skp2 S72D, Xp-Skp2 S64A, Xp-Skp2 S64D, Xp-Skp2 S72A, Xp-Skp2 S72D and GFP–Skp2 S72A constructs were generated with a site-directed mutagenesis kit (Stratagene) in accordance with the manufacturer’s standard procedure. The



A RT I C L E S

Skp2-NES construct was generated by fusing the nuclear export signal (NES) (ELALKLAGLDINKTE) to the C terminus of Skp2. MSCV-PIG-Skp2-NES was generated by subcloning Skp2-NES into the Xho1 site of the MSCV-PIG vec-tor. pBabe-Skp2, pBabe-Skp2 S72A and pBabe-Skp2 S72D were generated by subcloning Skp2, Skp2 S72A and Skp2 S72D into the EcoRI site of the pBabe vector. GFP–PRKRLKS (from Skp2), GFP–PRKRLKA (from Skp2 S72A), GFP–PRKRLKD (from Skp2 S72D) and GFP–PKKKRKVD (from SV40 large T anti-gen) constructs were generated by introducing the peptide encoding sequence in frame at the 3´ end of the GFP coding sequence in a pcDNA3.1/NT-GFP–TOPO vector (Invitrogen). His6-ubiquitin, haemagglutinin-tagged p27 and pGEX-4X1-Skp1 constructs were gifts from D. Bohmann, M. Pagano and P. Jackson, respec-tively. To clone Xp-mSkp2, the mouse Skp2 complementary DNA was amplified by RT–PCR from MEFs and was inserted into pcDNA4/HisMax-TOPO vector. Flag-Skp2 and Flag–∆N-Skp2 were from W. Wei. The p27-T187A construct was from M. H. Lee. IGF-1, EGF, MG132, U0126, rapamycin and LY294002 were obtained from Calbiochem. Wortmannin was purchased from Sigma. GST–Akt1 and recombinant active Akt1 were obtained from Cell Signaling.

Immunoprecipitation, immunoblotting and immunofluorescence. Immunoprecipitation (IP), immunoblotting (IB) and immunofluorescence (IF) were performed as described39,40. For protein–protein interaction, cells were lysed in E1A lysis buffer (250 mM NaCl, 50 mM HEPES pH 7.5, 0.1% Nonidet P40, 5 mM EDTA, protease inhibitor cocktail (Roche)). The following antibodies were used for IP, IB or IF: anti-Skp2 (IP: 1:250; IB: 1:1,000; IF: 1:100; Zymed), anti-Skp2 (IB 1:1,000; Santa Cruz), anti-phospho-(S473)-Akt (IB: 1:1,000; Cell Signaling), anti-Akt1 (IB: 1:1,000; Cell Signaling), anti-phospho-Akt substrate antibody (IB: 1:1,000; Cell Signaling), anti-Skp1 (IP: 1:200; IB: 1:1000; BD Transduction Lab), anti-14-3-3β (IP: 1:100; IB: 1:2,000; Santa Cruz), anti-14-3-3γ (IB: 1:500; Santa Cruz), phospho-(S72)-Skp2 antibody (IB: 1:1,000; Pocono Rabbit Farm and Laboratory Inc.), anti-Xpress (IP and IF: 1:500; IB: 1:5,000; Invitrogen), anti-haemagglutinin (IB and IF, 1:1,000; Covance, Upstate), anti-α-tubulin (IB: 1:1,000; Sigma), anti-β-actin (IB: 1:1,000; Sigma), anti-GFP (IP: 1:500; IB: 1:1,000; BD Clontech), anti-p27 (IP: 1:100; IB: 1:1,000; BD Transduction Lab) and anti-E2F1 (IB: 1:500; Santa Cruz).

In vitro and in vivo phosphorylation assay. The in vitro phosphorylation assay was performed as described41. In brief, GFP–Skp2 or GFP–Skp2 S72A was immu-noprecipitated with GFP antibody from 293T cells cultured in DMEM medium containing 0.5% FBS. Immune complexes were washed three times in RIPA lysis buffer (150 mM NaCl, 10 mM Tris-HCl pH 7.5, 1% Nonidet P40, 0.5% deoxy-cholate, 0.1% SDS, protease inhibitor cocktail (Roche)), then washed twice in 1 × kinase buffer (25 mM Tris-HCl pH 7.5, 5 mM β-glycerophosphate, 2 mM dithiothreitol, 0.1 mM Na3VO4, 10 mM MgCl2, 2 µM unlabelled ATP), and incubated with 1 µl recombinant active Akt1 kinase and 2 µCi of [γ-32P]ATP in 50 µl total reaction buffer for 30 min at 30 °C. Reactions were stopped by wash-ing twice in kinase buffer and boiling in 2 × SDS loading buffer. Proteins were resolved by 8% SDS–PAGE and transferred to the nitrocellulose membrane, and 32P incorporation was detected by autoradiography. For in vivo labelling experi-ments, 293T cells were transfected with the indicated plasmids for 24 h, and the medium was changed to phosphate-free DMEM with 0.25% dialysed FBS containing 200 µCi ml−1 ortho-32PO4 for 4 h. Cells were lysed by RIPA buffer for immunoprecipitation and the Skp2 immune complex was subjected to 8% SDS–PAGE, followed by autoradiography.

Identification of phosphorylation sites by mass spectrometry. Gel-resolved proteins from in vitro and in vivo phosphorylation reactions were digested with trypsin and batch purified on a reverse-phase (RP) micro-tip; an aliquot was ana-lysed by MALDI–reTOF mass spectrometry (MS; UltraFlex TOF/TOF; Bruker Daltonics, Bremen, Germany) for peptide mass fingerprinting, as described41,42. This served to confirm the identity of the proteins and to locate possible differ-ences between the tryptic peptide maps of the phosphorylated and unphosphor-ylated forms. The remainder of the RP-eluted digest mixtures was then subjected to immobilized gallium(III) affinity chromatography for selective capture of phos-phopeptides, followed by elution with phosphate buffer, desalting over an RP tip, and a second round of MALDI-reTOF MS20. Peak m/z values were matched to the protein sequence, allowing for the likely presence of one or more phosphate groups. Mass spectrometric sequencing of the putative phosphopeptides was then

performed by MALDI-TOF/TOF MS/MS analysis with the UltraFlex instrument in ‘LIFT’ mode. Fragment ion spectra were inspected for the characteristic partial loss of 98 Da, indicating the presence of phosphoserine or phosphothreonine, and for the a´´, b´´ and y´´ ions to compare with the computer-generated frag-ment ion series of the predicted tryptic peptides to locate the exact position of phosphoamino acids.

Ubiquitylation assay in vitro and in vivo. In vitro p27 ubiquitylation assays were performed essentially as described4. In brief, the Skp2-SCF complex was immu-noprecipitated from 293T cells and mixed with in vitro-translated 35S-p27 that had previously been incubated with cyclin E/Cdk2 for in vitro phosphorylation along with methylated ubiquitin and ubiquitin aldehyde for 60 min at 30 °C. The reaction was stopped with 2 × SDS sample buffer and run on polyacrylamide gels. In vivo ubiquitylation assays were performed as described43. In brief, 293T cells were transfected with the indicated plasmids for 24 h, treated with vehicle, 20 µM LY294002 or 100 nM wortmannin together with 20 µM MG132 for 6 h, and lysed in denaturing buffer (6 M guanidine-HCl, 0.1 M Na2HPO4/NaH2PO4, 10 mM imidazole). The cell extracts were then incubated with nickel beads for 3 h, washed, and subjected to western blot analysis.

Cell growth and in vivo tumorigenesis assay. Bromodeoxyuridine incorpora-tion assays were performed with the In Situ Cell proliferation Kit (Fluos; Roche). LNCaP stable transfectant cells were generated as described44. In brief, LNCaP cells were transfected with pcDNA3, pcDNA3-Skp2 or pcDNA3-Skp2 S72A and selected with G418 (300 µg ml−1). The individual clone was picked and verified by western blot analysis (data not shown). Two positive clones were pooled and used for western blot analysis, cell growth, and the in vivo tumorigenesis assay. For the cell growth assay, 2 × 104 cells were seeded in 12 wells in triplicate, harvested, and stained with trypan blue on different days. Viable cells were counted directly under the microscope. For in vivo tumorigenesis assays, LNCaP stable cells (106) mixed with matrigel (1:1) were injected subcutaneously into the left flank of six-week-old athymic male nude mice (NCRNU-M; Taconic Farms Inc.). Tumour size was measured weekly with a calliper, and tumour volume was determined with the standard formula: L × W2 × 0.52, where L is the longest diameter and W the shortest diameter.


ACKNoWLEdGEmENTSWe thank D. Bohmann, P. Jackson, W. Wei, M. Pagano and M. H. Lee for reagents. We are also grateful to M. Asherov and I. Linkov in the Immunohistochemistry Pathology Core Laboratory, T. Matos for immunohistochemistry technical assistance, P. Bonner for data management, L. Lacomis for help with mass spectrometry, X. H. Zhu for technical advice, and S. Clohessy for flow cytometry analysis. We also thank M. C. Hung and L. Cantley for insightful comments and suggestions, and W. Wei for discussion and for sharing experimental results. Special thanks are extended to B. Carver and L. DiSantis for editing and critical reading of the manuscript, as well as to all the members of the Pandolfi laboratory for comments and discussion. This work was supported by NIH grants RO1 CA-71692 and CA-74031 to P.P.P. and by M. D. Anderson Cancer Center Trust Scholar funds to H.K.L. The Microchemistry & Proteomics Core is supported by NIH grant P30 CA-08748.

AuTHor CoNTrIBuTIoNSH.K.L. and P.P.P. designed the experiments and wrote the manuscript. H.K.L., G.W. Z.C., Y.L., C.H.C. and W.L.Y. performed the experiments. J.T. performed the immunohistochemistry and analysed the data. K.I.N. provided the Skp2−/− mice. S.N. provided valuable suggestions. H.E. and P.T. performed the mass spectrometry analysis.

ComPETING FINANCIAL INTErESTSThe authors declare no competing financial interests.


1. Bloom, J. & Pagano, M. Deregulated degradation of the cdk inhibitor p27 and malignant transformation. Semin. Cancer Biol. 13, 41–47 (2003).

2. Nakayama, K. I. & Nakayama, K. Regulation of the cell cycle by SCF-type ubiquitin ligases. Semin. Cell Dev. Biol. 16, 323–333 (2005).

3. Amati, B. & Vlach, J. Kip1 meets SKP2: new links in cell-cycle control. Nature Cell Biol. 1, E91–E93 (1999).






A RT I C L E S

4. Carrano, A. C., Eytan, E., Hershko, A. & Pagano, M. SKP2 is required for ubiqui-tin-mediated degradation of the CDK inhibitor p27. Nature Cell Biol. 1, 193–199 (1999).

5. Tsvetkov, L. M., Yeh, K. H., Lee, S. J., Sun, H. & Zhang, H. p27Kip1 ubiquitination and degradation is regulated by the SCFSkp2 complex through phosphorylated Thr187 in p27. Curr. Biol. 9, 661–664 (1999).

6. Kossatz, U. et al. Skp2-dependent degradation of p27kip1 is essential for cell cycle progression. Genes Dev. 18, 2602–2607 (2004).

7. Sutterlüty, H. et al. p45SKP2 promotes p27Kip1 degradation and induces S phase in quiescent cells. Nature Cell Biol. 1, 207–214 (1999).

8. Nakayama, K. et al. Targeted disruption of Skp2 results in accumulation of cyclin E and p27Kip1, polyploidy and centrosome overduplication. EMBO J. 19, 2069–2081 (2000).

9. Nakayama, K. et al. Skp2-mediated degradation of p27 regulates progression into mitosis. Dev. Cell 6, 661–672 (2004).

10. Gstaiger, M. et al. Skp2 is oncogenic and overexpressed in human cancers. Proc. Natl Acad. Sci. USA 98, 5043–5048 (2001).

11. Latres, E. et al. Role of the F-box protein Skp2 in lymphomagenesis. Proc. Natl Acad. Sci. USA 98, 2515–2520 (2001).

12. Shim, E. H. et al. Expression of the F-box protein SKP2 induces hyperplasia, dyspla-sia, and low-grade carcinoma in the mouse prostate. Cancer Res. 63, 1583–1588 (2003).

13. Zhang, H., Kobayashi, R., Galaktionov, K. & Beach, D. p19Skp1 and p45Skp2 are essential elements of the cyclin A-CDK2 S phase kinase. Cell 82, 915–925 (1995).

14. Bilodeau, M. et al. Skp2 induction and phosphorylation is associated with the late G1 phase of proliferating rat hepatocytes. FEBS Lett. 452, 247–253 (1999).

15. Dan, H. C. et al. Phosphatidylinositol 3-kinase/Akt pathway regulates tuberous scle-rosis tumor suppressor complex by phosphorylation of tuberin. J. Biol. Chem. 277, 35364–35370 (2002).

16. Inoki, K., Li, Y., Zhu, T., Wu, J. & Guan, K. L. TSC2 is phosphorylated and inhibited by Akt and suppresses mTOR signalling. Nature Cell Biol. 4, 648–657 (2002).

17. Manning, B. D., Tee, A. R., Logsdon, M. N., Blenis, J. & Cantley, L. C. Identification of the tuberous sclerosis complex-2 tumor suppressor gene product tuberin as a target of the phosphoinositide 3-kinase/akt pathway. Mol. Cell 10, 151–162 (2002).

18. Potter, C. J., Pedraza, L. G. & Xu, T. Akt regulates growth by directly phosphorylating Tsc2. Nature Cell Biol. 4, 658–665 (2002).

19. Yaffe, M. B. et al. A motif-based profile scanning approach for genome-wide prediction of signaling pathways. Nature Biotechnol. 19, 348–353 (2001).

20. Posewitz, M. C. & Tempst, P. Immobilized gallium(III) affinity chromatography of phos-phopeptides. Anal. Chem. 71, 2883–2892 (1999).

21. Malek, N. P. et al. A mouse knock-in model exposes sequential proteolytic pathways that regulate p27Kip1 in G1 and S phase. Nature 413, 323–327 (2001).

22. Drobnjak, M. et al. Altered expression of p27 and Skp2 proteins in prostate cancer of African-American patients. Clin. Cancer Res. 9, 2613–2619 (2003).

23. Li, Q., Murphy, M., Ross, J., Sheehan, C. & Carlson, J. A. Skp2 and p27kip1 expression in melanocytic nevi and melanoma: an inverse relationship. J. Cutan. Pathol. 31, 633–642 (2004).

24. Lim, M. S. et al. Expression of Skp2, a p27Kip1 ubiquitin ligase, in malignant lymphoma: correlation with p27Kip1 and proliferation index. Blood 100, 2950–2956 (2002).

25. Radke, S., Pirkmaier, A. & Germain, D. Differential expression of the F-box proteins Skp2 and Skp2B in breast cancer. Oncogene 24, 3448–3458 (2005).

26. Cantley, L. C. & Neel, B. G. New insights into tumor suppression: PTEN suppresses tumor formation by restraining the phosphoinositide 3-kinase/AKT pathway. Proc. Natl Acad. Sci. USA 96, 4240–4245 (1999).

27. Di Cristofano, A. & Pandolfi, P. P. The multiple roles of PTEN in tumor suppression. Cell 100, 387–390 (2000).

28. Brazil, D. P., Park, J. & Hemmings, B. A. PKB binding proteins. Getting in on the Akt. Cell 111, 293–303 (2002).

29. Datta, S. R., Brunet, A. & Greenberg, M. E. Cellular survival: a play in three Akts. Genes Dev. 13, 2905–2927 (1999).

30 Gao, D. et al. Phosphorylation by Akt1 promotes Skp2 cytoplasmic localization and impairs APC/Cdh1-mediated Skp2 destruction. Nature Cell Biol. 11, doi:10.1038/ncb1847 (2009).

31. Hermeking, H. The 14-3-3 cancer connection. Nature Rev. Cancer 3, 931–943 (2003).

32. Brunet, A. et al. Akt promotes cell survival by phosphorylating and inhibiting a Forkhead transcription factor. Cell 96, 857–868 (1999).

33. Brunet, A. et al. 14-3-3 transits to the nucleus and participates in dynamic nucleocy-toplasmic transport. J. Cell Biol. 156, 817–828 (2002).

34. Fujita, N., Sato, S., Katayama, K. & Tsuruo, T. Akt-dependent phosphorylation of p27Kip1 promotes binding to 14-3-3 and cytoplasmic localization. J. Biol. Chem. 277, 28706–28713 (2002).

35. Sekimoto, T., Fukumoto, M. & Yoneda, Y. 14-3-3 suppresses the nuclear localization of threonine 157-phosphorylated p27Kip1. EMBO J. 23, 1934–1942 (2004).

36. Mamillapalli, R. et al. PTEN regulates the ubiquitin-dependent degradation of the CDK inhibitor p27KIP1 through the ubiquitin E3 ligase SCFSKP2. Curr Biol 11, 263–267 (2001).

37. Rodier, G., Coulombe, P., Tanguay, P. L., Boutonnet, C. & Meloche, S. Phosphorylation of Skp2 regulated by CDK2 and Cdc14B protects it from degradation by APCCdh1 in G1 phase. EMBO J. 27, 679–691 (2008).

38. Zheng, N. et al. Structure of the Cul1–Rbx1–Skp1–F boxSkp2 SCF ubiquitin ligase complex. Nature 416, 703–709 (2002).

39. Lin, H. K., Bergmann, S. & Pandolfi, P. P. Cytoplasmic PML function in TGF-β signal-ling. Nature 431, 205–211 (2004).

40. Lin, H. K., Wang, L., Hu, Y. C., Altuwaijri, S. & Chang, C. Phosphorylation-dependent ubiquitylation and degradation of androgen receptor by Akt require Mdm2 E3 ligase. EMBO J. 21, 4037–4048 (2002).

41. Ma, L., Chen, Z., Erdjument-Bromage, H., Tempst, P. & Pandolfi, P. P. Phosphorylation and functional inactivation of TSC2 by Erk: implications for tuberous sclerosis and cancer pathogenesis. Cell 121, 179–193 (2005).

42. Sebastiaan Winkler, G. et al. Isolation and mass spectrometry of transcription factor complexes. Methods 26, 260–269 (2002).

43. Kim, S. Y., Herbst, A., Tworkowski, K. A., Salghetti, S. E. & Tansey, W. P. Skp2 regulates Myc protein stability and activity. Mol. Cell 11, 1177–1188 (2003).

44. Lin, H. K., Yeh, S., Kang, H. Y. & Chang, C. Akt suppresses androgen-induced apoptosis by phosphorylating and inhibiting androgen receptor. Proc. Natl Acad. Sci. USA 98, 7200–7205 (2001).



A RT I C L E S

Store-operated cyclic AMP signalling mediated by STIM1Konstantinos Lefkimmiatis1, Meera Srikanthan1, Isabella Maiellaro1,3, Mary Pat Moyer2, Silvana Curci1 and Aldebaran M. Hofer1,4

Depletion of Ca2+ from the endoplasmic reticulum (ER) results in activation of plasma membrane Ca2+ entry channels. This ‘store-operated’ process requires translocation of a transmembrane ER Ca2+ sensor protein, stromal interaction molecule 1 (STIM1), to sites closely apposed to Ca2+ channels at the cell surface. However, it is not known whether a reduction in Ca2+ stores is coupled to other signalling pathways by this mechanism. We found that lowering the concentration of free Ca2+ in the ER, independently of the cytosolic Ca2+ concentration, also led to recruitment of adenylyl cyclases. This resulted in enhanced cAMP accumulation and PKA activation, measured using FRET-based cAMP indicators. Translocation of STIM1 was required for efficient coupling of ER Ca2+ depletion to adenylyl cyclase activity. We propose the existence of a pathway (store-operated cAMP signalling or SOcAMPS) in which the content of internal Ca2+ stores is directly connected to cAMP signalling through a process that involves STIM1.

Ca2+ and cAMP are the universal intracellular currency of G‑protein‑coupled receptor (GPCR) signalling, exchanging information derived from many, if not most, extracellular ligands into intracellular sig‑nalling events1. These two fundamental second messenger pathways govern a vast array of cellular functions, but also regulate one another on multiple levels.

Agonist‑induced Ca2+ signalling events typically consist of two phases: release of Ca2+ from internal stores (mainly the ER), which leads to a second phase of sustained Ca2+ entry across the plasma membrane2,3, known as store‑operated or capacitative Ca2+ entry4. Recently the ER transmembrane protein STIM1 has been identified as the molecular sensor that couples reduction in intraluminal ER concentration of Ca2+ to the activation of influx pathways for Ca2+ in the plasma membrane5–8. Available evidence indicates that store depletion induces oligomeriza‑tion of STIM1, followed by translocation from the ER to sites adjacent to the plasma membrane9. This allows activation of store‑operated Ca2+ entry pathways (in particular, members of the recently identified Orai family10–12), and Ca2+ influx.

Changes in cytosolic Ca2+ levels are known to either enhance or depress cAMP production through the various Ca2+‑sensitive isoforms of adenylyl cyclase (AC)13. Of the nine transmembrane AC isoforms described so far, AC1 and AC8 are the major Ca2+‑activated isoforms, whereas AC5 and AC6 are subject to direct inhibition by physiological Ca2+ in the cytosol. The Ca2+‑inhibitable ACs are particularly suscepti‑ble to inhibition by Ca2+ entering through store‑operated entry path‑ways13. Ca2+‑dependent regulation of the enzymes that degrade cAMP, the cAMP phosphodiesterases (PDEs) can also significantly influence levels of the second messenger14,15.

Here we provide evidence for a form of store‑operated signalling that couples the Ca2+ content of the ER directly to cAMP production by ACs, independently of changes in cytosolic Ca2+ levels. This mechanism coexists with the more conventional regulation of ACs by cytosolic Ca2+ signalling events, but the physiological circumstances leading to the initiation of this process may be different. We further show that activation of this pathway after Ca2+ depletion from the ER requires expression of STIM1 and re‑localization and clustering of this ER protein in punctae near the plasma membrane. STIM1 may therefore fulfil broader functions than previously appreciated, serving as an organizer for a plasma membrane signalling complex that incorporates not only Ca2+ entry channels, but also ACs.

RESULTSChelation of ER Ca2+ enhances cAMP production independent of cytosolic Ca2+

Findings obtained during the course of an earlier study16 led us to consider the somewhat unorthodox possibility that Ca2+ concentration in the ER ([Ca2+]ER) could control cAMP production directly. To test this idea, we used a low‑affinity membrane‑permeant Ca2+ buffer, TPEN (N,N,N’,N’‑Tetrakis‑(2‑pyridylmethyl)‑ethylenediamine)17, to lower free (but not total) Ca2+ within stores in a reversible and concentration‑dependent manner18,19. TPEN (1 mM) rapidly reduced [Ca2+]ER in NCM460 cells (a colonic epithelial cell line derived from normal human colon20), as meas‑ured using a compartmentalized low‑affinity Ca2+ indicator mag‑fura‑2 (ref. 21; Fig. 1a). Washout of TPEN promptly restored [Ca2+]ER to resting levels, even in the absence of extracellular Ca2+. As expected, addition of TPEN (Kd for Ca2+ ~120 μM) in Ca2+‑free solutions did not cause detect‑able alterations in cytosolic Ca2+ levels (Fig. 1a, lower panel).

1VA Boston Healthcare System and the Department of Surgery, Brigham and Women’s Hospital and Harvard Medical School, 1400 VFW Parkway, West Roxbury, Massachusetts 02132, USA. 2INCELL Corporation LLC, San Antonio, Texas, 78249, USA. 3Current address: Dipartimento di Fisiologia Generale ed Ambientale, Universitá degli Studi di Bari, Bari, Italy 70126.4Correspondence should be addressed to A.M.H. (e-mail: [email protected])

Received 15 September 2008; accepted 5 December 2008; published online 15 March 2009; DOI: 10.1038/ncb1850




A RT I C L E S

Chelation of [Ca2+]ER with TPEN had the anticipated effects on re‑localization of YFP‑labelled STIM1 (provided by Tobias Meyer9) or mCherry–STIM1, as monitored using TIRF microscopy. It is now well‑established that when [Ca2+]ER is lowered, labelled STIM relocates from the bulk ER and coalesces into larger immobile aggregates, or punctae, just under the plasma membrane22. In the absence of extracellular Ca2+, the TIRF signal increased after TPEN treatment (indicating the appear‑ance of labelled STIM1 at the plasma membrane) and this was rapidly reversed when TPEN was washed out (Fig. 1b).

The action of TPEN on intracellular cAMP was next assessed in real‑time using a ratiometric FRET‑based cAMP sensor Epac H30

(a gift from Kees Jalink and colleagues23). Acute addition of TPEN in Ca2+‑free solutions caused the 480 nm/535 nm FRET ratio of the cAMP sensor to increase (Fig. 1c), consistent with an elevation in intracellular cAMP. This response was markedly hastened in the pres‑ence of the AC activator forskolin (3.55 ± 0.45‑fold faster, 47 cells in 6 experiments; P < 0.0001). The action of TPEN on the FRET ratio was concentration‑dependent, with a good correlation between the predicted extent of reduction in [Ca2+]ER

18,19 and cAMP production (Supplementary Information, Fig. S1).

We further tested the action of TPEN on signalling downstream of cAMP using an improved FRET‑based probe for PKA phosphorylation,

IonoTPENTPEN

mag

-fur

a-2

ratio

5 min

0.5

0.6

0.7

0.8

5 min

TPENTPEN

Fsk

c

Control TPEN Fsk Fsk + TPEN

WashTPEN

Lono

0

0

100

150

50

200

250

300

350

400

TPEN

+Fsk

TPEN

1 min

Em

issi

on r

atio

(480

/535

)

1.4

1.5

1.6

1.7

Fsk, 50 µm+IBMXTPEN

TPEN

5 min

a b

Em

issi

on r

atio

(535

/480

)

PKAphosphorylation

d

e

Per

cent

age

of T

PE

N s

lop

e

bc

d

Internal store

TPEN + Ca2+ TPEN Ca2+trapped

Plasma membrane

TPEN Mag-fura-2

0.05

0 Ca2+

0 Ca2+

0 Ca2+

0 Ca2+

0 Ca2+

TPEN

TPEN

TPEN

Iono

IonoFsk, 2 µm

Fsk, 2 µm

Fura-2

5 min

5 min

Ca2+

(nm

)

1200

1000

800

600

400

200

0

400

200

TIRF intensity

Fluo

resc

ence

inte

nsity

Epac-H30

Fsk, 1 µM

AKAR3

a”

a’

Figure 1 Chelation of Ca2+ within internal stores of NCM460 cells induces STIM1 translocation, cAMP signalling and PKA phosphorylation. (a) 345 nm/385 nm excitation ratio of compartmentalized mag-fura-2 in intact cells bathed in Ca2+-free solutions (black trace, top upper panel) showing a rapidly reversible effect of TPEN (1 mM) on intraluminal Ca2+. Addition of ionomycin (Iono, 5 μM) caused even further loss of stored Ca2+. Data are mean ± s.e.m. of 9 cells. Cytosolic Ca2+ levels (measured by fura-2, red trace, lower panel) did not change during TPEN treatment in Ca2+-free solutions (Fsk, forskolin). The large Ca2+ peak elicited by Iono is shown for comparison. Data are mean ± s.e.m. of 36 cells. (b) Time course of YFP–STIM1 translocation to the cell surface following treatment with TPEN (1 mM) as measured using TIRF microscopy (upper panel; n = 16 cells, 6 experiments). The lower panels show TIRF images of punctae in cells with the indicated treatments corresponding to those shown in the upper panel. Scale bar, 15 μm. (c) NCM460 cells expressing Epac-based cAMP

sensor. Addition of TPEN (1 mM) in Ca2+-free solution by itself caused an increase in the 480 nm/535 nm FRET emission ratio but this response was markedly potentiated in the presence of Fsk (2 μM). Low concentrations of TPEN (10–40 μM) did not affect the FRET ratio, making a role for heavy metals in this process unlikely. The inset shows summary data (mean ± s.e.m. of 47 cells in 6 experiments) for the relative rate of TPEN responses in the presence and absence of forskolin. (d) 535 nm/480 nm FRET emission ratio of the PKA phosphorylation sensor, AKAR3, after treatment with TPEN (representative of 10 cells in 7 independent experiments). At the end of most experiments, cells were stimulated with saturating concentrations of the PDE inhibitor isobutylmethyl xanthine (IBMX, 1 mM) and Fsk (25–50 μM) to establish the maximum ratio response. (e) Overlay of time courses of TPEN responses in the presence of forskolin from experiments in a–d for mag-fura-2 (a´), fura-2 (a´´), YFP–STIM1 TIRF (b), Epac-H30 (c) and AKAR-3 (d).



A RT I C L E S

AKAR3 (A‑kinase activity reporter 3; provided by Jin Zhang24). TPEN caused a reversible increase in the 535 nm/480 nm FRET emission ratio of AKAR3 in NCM460 cells, consistent with an increase in PKA phos‑phorylation. An overlay showing the temporal correlation of the various TPEN responses (Fig. 1e) indicates that lowering [Ca2+]ER (mag‑fura‑2) and translocation of STIM1 (TIRF) precede cAMP production (Epac H30), and that these events occur well before PKA‑dependent phospho‑rylation (AKAR3).

Release of Ca2+ stores using diverse strategies enhances cAMP signallingDepletion of stored Ca2+ using ionomycin in Ca2+‑free solutions pro‑duced a small but consistent increase in cAMP (Fig. 2a). The subse‑quent response to forskolin was markedly potentiated after ionomycin

pre‑treatment, compared with forskolin alone (283 ± 17%; n = 63 cells in 5 experiments, P < 0.0001). Two consecutive control challenges with forskolin did not produce significantly different responses (data not shown; amplitude of second, compared with the first response, 102 ± 5%, n = 11 cells in 4 experiments). Parallel experiments with fura‑2 showed that the spike in cytosolic Ca2+ caused by release from stores had subsided in less than about 3 min. In contrast, the increase in cAMP in response to store‑emptying persisted even 60 min after addition of ionomycin (data not shown).

Addition of ionomycin to NCM460 cells pre‑stimulated with forskolin in Ca2+‑free solutions also resulted in a marked and persistent increase in cAMP (345 ± 18% of the initial forskolin response, n = 289 cells in 49 experiments, P < 0.0001; Fig. 2b). Although our Ca2+‑free solutions typi‑cally contained EGTA (25 μM) to buffer contaminating Ca2+, this result

b c

d

a

e f

+ IBMX

0.7

0.8

0.9

1.0

1.1

1.2Fsk

Fsk

3 min

E

mis

sion

rat

io (4

80/5

35)

050

100150200250300 ***

Iono0 Ca2+

Fsk, 50 µM

Fura-2ratio

0.3

0.2

Per

cent

age

Fsk

cont

rol

Fsk+

IonoFs

k

30

50

70

90*

10

*

Intr

acel

lula

r cA

MP

(100

% b

ound

) cAMP Immunoassay

Bas

al

Fsk

Fsk

Ion

oIo

no

Fsk

5 min

Fsk FskIono

0.5

0.6

0.7

0.8

0

100

200

300

0 Ca2+

Per

cent

age

Fsk

cont

rol

IonoFs

k

Em

issi

on r

atio

(480

/535

)

Fura-2ratio

0.4

0.2

5 min

Iono

0.8

0.9

1.0

1.1

1.2

1.3

0.3

5 min 0.2

Fsk

Con

trol

BA

PTA

-AM

0 Ca2+

+ IBMXFsk, 50 µM

Fura-2ratio

Em

issi

on r

atio

(480

/535

)

BAPTA-AM Pretreated

Iono

. res

pon

se

400

300

200

100

0

1.6

1.7

1.8

1.9

2

3 min

AKAR3

IonoFsk

0 Ca2+

+ IBMXFsk, 50 µM

Em

issi

on r

atio

(535

/480

)

PKAphosphorylation

0

1

2

3

NT

Fsk

Fsk

+ Io

no

Fsk

+ IB

MX

ICER

*** **

mR

NA

ICE

R/C

RE

M (m

ultip

le o

f NT

cont

rol)

Figure 2 Depletion of internal Ca2+ stores with ionomycin enhances cAMP signalling, PKA phosphorylation and expression of ICER in NCM460 cells. (a, b) The blue traces shows 480 nm/535 nm FRET emission ratio of the Epac H30 cAMP sensor in cells treated with ionomycin (Iono, 5 μM) followed by forskolin (Fsk, 2 μM) (a) or Fsk followed by Iono (b) in Ca2+-free solutions. At the end of most experiments, cells were stimulated with saturating concentrations of IBMX (1 mM) and Fsk (25–50 μM) to establish the maximum ratio response. Insets: summary data (mean ± s.e.m. of 63 cells, 5 experiments (a) and 289 cells, 49 experiments (b) of the amplitude of Iono-induced cAMP response in the presence of Fsk, compared with Fsk alone. Similar results were obtained using an Epac-based cAMP sensor targeted to the nucleus (14 cells, 4 experiments; data not shown). Red traces show the time course of the intracellular Ca2+ spike measured in parallel under same conditions using fura-2. (c) NCM460 cells pre-treated with the high-affinity Ca2+ chelator BAPTA-AM (20 μM, 40 min). Ionomycin-induced enhancement of the FRET signal (blue trace, average of 8 cells ± s.e.m. ) was not significantly

different between BAPTA-AM-loaded or control cells (summary data in inset mean ± s.e.m. of 24 cell, 5 experiments for the BAPTA-loaded condition). Parallel measurements with fura-2 (red trace) showed that ionomycin-induced Ca2+ changes were undetectable after BAPTA loading (24 cells, 5 experiments). (d) Quantitative competitive immunoassay for cAMP. ‘Basal’ indicates absorbance reading obtained in the absence of any treatment. A 10-min treatment with Fsk (5 μM) in Ca2+-free solutions caused a significant (P < 0.02, compared with basal) increase in cAMP. Fsk→Iono, pre-treatment with Fsk followed by Fsk and Iono (5 min each, Ca2+-free solutions; protocol similar to Fig. 2b). Iono→Fsk, (5 min each; protocol as in Fig. 2a). Data are mean ± s.e.m. of 6 independent experiments with duplicate measurements for each condition. (e) 535 nm/480 nm FRET emission ratio of the PKA phosphorylation sensor, AKAR3. Data are representative of 9 cells, 5 experiments. (f) Real-time PCR analysis of ICER expression (a cAMP-dependent gene). NT indicates not treated. Data are mean ± s.e.m. of 4 experiments performed in quadruplicate. *P < 0.01; **P < 0.002; ***P < 0.001; ****P < 0.0001.



A RT I C L E S

did not change using higher concentrations of EGTA (1–5 mM), making it unlikely that low levels of Ca2+ entry were responsible for the observed effects (data not shown; n = 98 cells, in 7 experiments). However, the response was prevented in cells pre‑treated with the SERCA inhibitor thapsigargin (100 nM for 60 min, data not shown; n = 58 cells in 5 experi‑ments), showing a dependence of this process on thapsigargin‑sensitive stores.

The increase in cAMP following ionomycin was independent of changes in cytosolic Ca2+ levels, as the response in cells pre‑treated with BAPTA‑AM (a high‑affinity cell‑permeant Ca2+ buffer that prevents the cytosolic Ca2+ spike;) was not significantly different from control (312 ± 35% of the initial forskolin response; n = 24 cells in 5 experiments, P < 0.0001; Fig. 2c). Parallel measurements with fura‑2 indicated that the intracellular Ca2+ spike was not detectable under these conditions. We cannot exclude the existence of cytosolic ‘nanodomains’ of Ca2+

that escape buffering by BAPTA, but our data did show that a nominal increase in cytosolic Ca2+ levels elicited by very low concentrations of a Ca2+‑mobilizing agonist did not appreciably alter cAMP (Supplementary Information, Fig. S2).

Similar data were obtained when other native agonists working through GPCRs were used to increase intracellular cAMP in place of forskolin (10 nM prostaglandin E2, n = 27 cells in 6 experiments; 5 nM vasoactive intestinal peptide, n = 9 cells in 2 experiments or 10 nM iso‑prenaline, n = 13 cells in 2 experiments; data not shown). This phe‑nomenon was also observed using different means of releasing Ca2+ stores, including treatment with Ca2+‑mobilizing agonists (ATP, carba‑chol or the bile acid deoxycholic acid), either alone or in combination with the SERCA inhibitor, tert‑butylhydroquinone (tBHQ, 10 μM; see Supplementary Information, Figs S2, S3 for examples).

A slow increase in cAMP was also revealed following passive store depletion by long‑term (40–60 min) incubation in Ca2+‑free solutions containing EGTA (5 mM; data not shown; n = 172 cells, 10 experiment) or acute addition of high concentrations of BAPTA‑AM (40 μM; n = 74 cells, 4 experiments). Moreover, this activity was present to varying degrees in other cell types (HEK293, HT‑29, CaCo‑2 colon carcinoma cells and HepG2 hepatocytes, but with no response in HeLa or H1299 human lung carcinoma cells; see Supplementary Information, Fig. S3 for details). The most robust responses were observed in NCM460 cells.

The ability of ionomycin to elevate cAMP, as measured with Epac H30, was confirmed using a cAMP immunoassay (Fig. 2d) and the PKA phosphorylation sensor AKAR3 (Fig. 2e). Real‑time PCR data (Fig. 2f) showed that the cAMP generated by store depletion was translated into downstream changes in the expression of a PKA‑dependent, cAMP‑specific gene, ICER25.

The preceding experiments were conducted in Ca2+‑free solutions to minimize confounding effects of sustained Ca2+ entry on cAMP metabolism. In the presence of physiological levels of extracellular Ca2+, cAMP decreased initially, but increased in all cells about 30 min after the addition of ionomycin (Fig. 3a). Additional cAMP was generated when Ca2+ was removed from the bath. This suppression of the signal was probably caused by Ca2+‑dependent inhibition of AC5 and AC6 (see below). Similarly, TPEN‑induced store depletion also exerted its effects on intracellular cAMP when Ca2+ was present extracellularly18 (Fig. 3b). Collectively these results also indicate that a decrease in cytosolic Ca2+ (that is, the undershoot19 observed when stores are released in the absence of external Ca2+), is unlikely to account for our observations.

Store-operated cAMP signalling is due to activation of ACs and not inhibition of PDEsThe increased levels of intracellular cAMP following store depletion could be caused by either activation of ACs or inhibition of PDEs. Ionomycin (Fig. 4a) or TPEN treatments (Fig. 4b) were still able to elicit an increase in the cAMP FRET ratio in the presence of a broad‑spectrum PDE inhibitor, isobutylmethyl xanthine (IBMX, 1 mM), indicating that PDEs are probably not involved in this phenomenon.

Our RT–PCR analysis of AC expression in NCM460 cells showed the presence of mRNA transcripts for AC3‑7 and AC9 but not AC1, AC2 or AC8 (Supplementary Information, Fig. S4a). Pre‑incubation with a general AC inhibitor, SQ22536, prevented ionomycin‑stimulated cAMP elevation (Fig. 4c; see Supplementary Information, Fig. S4b for the cor‑responding control). We also observed partial (39.2 ± 3.5%) inhibition

TPEN545 µM

3min

480/

535

Fura-2 ratio

b

1.0

1.1 Fsk, 2 µM

Fsk, 2 µM

a + Extracellular Ca2+

lono 0 Ca2+

Em

issi

on r

atio

(480

/535

)

Fsk, 50 µM+IBMX

5 min

3 min

0.1

cAMP

Fsk, 50 µM+IBMX

Ca2+

Iono

0.1

0 mM Ca2+1.75 mM0 mM Ca2+

TPEN,2 mM

Figure 3 Effects of store depletion on intracellular cAMP in the presence of extracellular Ca2+ in NCM460 cells. (a) In the presence of a physiological concentration of extracellular Ca2+ (1 mM), the response to ionomycin (Iono, 5 μM) was suppressed, but the net effect was a significant increase in the FRET ratio of the Epac H30 probe (representative of 49 cells, 5 experiments). Removal of Ca2+ from the bath in three additional experiments (39 cells) caused an even further increase in the ratio (mean ± s.e.m. 18 cells). (b) Effect of TPEN in the presence of extracellular Ca2+. Solutions containing TPEN (2 mM) and Ca2+ (1.75 mM) had a calculated free TPEN concentration (the membrane-permeant form) of 0.545 mM and a calculated free Ca2+ concentration of 0.3 mM. Direct measurements of [Ca2+]ER have shown that this buffered solution causes a significant drop in free [Ca2+]ER (ref. 18). TPEN (0.545 mM) in Ca2+-free solution caused a similar elevation in the FRET ratio. Data are representative of 33 NCM460 cells, 5 experiments and 30 HEK293cells, 2 experiments (data not shown). Parallel control measurements with fura-2 (red trace) showed that cytosolic Ca2+ did not change during the experimental manoeuvers (mean ± s.e.m. of 33 cells). The response to Iono at end of the fura-2 experiment is shown for comparison.



A RT I C L E S

of the ionomycin response by acute treatment with NKY80, an AC5‑specific, AC5‑specific inhibitor that is non‑competitive with respect to forskolin (Fig. 4d). Collectively these results suggest that AC5 and at least one other AC isoform (susceptible to SQ22536 but not NKY80) are required for this phenomenon.

Participation of STIM1 in store-operated cAMP signallingSo far we have provided evidence for a store‑operated signal transduction pathway that links [Ca2+]ER to cAMP production by ACs, independently of the conventional regulation of ACs by cytosolic Ca2+. We have named this phenomenon SOcAMPS for store‑operated cAMP signalling. We considered STIM1as a potential sensor that may couple [Ca2+]ER to cAMP production, as it has a Ca2+‑sensing EF‑hand domain situated within the ER lumen.

To evaluate the participation of STIM1 in SOcAMPS, we used two inde‑pendent gene‑silencing approaches. Knockdown of STIM1 using a short

hairpin RNA (shRNA) produced a marked decrease in the store‑depend‑ent cAMP signal (Fig. 5a), compared with the control (Fig. 5b). Parallel experiments with fura‑2 showed attenuation of store‑operated Ca2+ entry in STIM1 knockdown cells, as reported previously6–8 (Fig. 5c), providing functional evidence for the efficiency of our shRNA knockdown strategy.

The data described above were confirmed using a short interfering RNA (siRNA) against STIM1, which effectively reduced STIM1 protein expression (Fig. 5d). The amplitude of the ionomycin response was sig‑nificantly decreased when compared with control or scramble siRNA populations (Fig. 5d, summary data), further indicating that STIM1 expression was necessary for store‑operated cAMP signalling.

As the NKY80 data shown in Fig. 4d implicate AC5 in SOcAMPS, we attempted to co‑immunoprecipitate AC5 and STIM1 before and after release of internal Ca2+ stores. Lack of suitable reagents prevented us from immunoprecipitating AC5; however, the converse experiment

ba

c d

5 min

0.8

0.9

1.0

1.1

1.2

IBMX0 Ca2+

Iono Fsk,50 µM

Em

issi

on r

atio

(480

/535

)

0.5

0.6

0.7

0.8

5 min

IBMX

0 Ca2+

Fsk, 50 µM

Em

issi

on r

atio

(480

/535

)

TPEN

0.5

0.6

0.7

0.8

0.9

5 min0

100

150

50

200*

0 Ca2+

Iono

Fsk, 5 µM

NKY80 50 µM

Em

issi

on r

atio

(480

/535

)

Iono

res

pon

se

0.5

0.6

0.7

0.8

0.9

5 min

100

0

200

300

400

0 Ca2+

Iono Fsk, 50 µM+IBMX

Em

issi

on r

atio

(480

/535

)

SQ22536

Rinse SQ22536

Iono

res

pon

se

***

Figure 4 Store depletion-induced cAMP signalling results from activation of ACs and not inhibition of PDEs. (a) Treatment of NCM460 cells expressing the Epac H30 cAMP sensor with IBMX (1 mM) did not prevent the ionomycin (Iono, 5 μM)-induced increase in the FRET ratio (mean of 3 cells; data are representative of 47 cells, 6 experiments.). (b) IBMX (1 mM) did not prevent the increase in cAMP elicited by TPEN (1 mM). However, the rapid reversal of the FRET ratio normally observed after TPEN washout was blocked in the presence of IBMX, showing that this recovery was dependent on PDE activity (mean of 13 cells, representative

of 32 cells in 3 experiments). (c) The response to Iono was significantly inhibited in cells pre-treated for 60 min with the AC antagonist SQ22536 (250 μM, black bar), but recovered when the inhibitor was washed out (blue bar). Trace shows mean of 11 cells, representative of 101 cells in 8 experiments. A control experiment shows that cAMP did not change when SQ22536 was washed out in the absence of Iono (Supplementary Information, Fig. S4). (d) NKY80, an inhibitor specific for AC5, inhibited the response to Iono by 39% (mean of 10 cells, representative of 44 cells in 4 experiments). *P < 0.02; ***P < 0.0001.



A RT I C L E S

(pulldown of STIM1 and probing for AC5) did not show an interaction between these two proteins (Supplementary Information, Fig. S4c).

Overexpression of mCherry–STIM1 did not significantly affect SOcAMPS (Supplementary Information, Fig. S5a). However, the response was significantly attenuated by overexpression of STIM1D76A, a well‑characterized EF‑hand mutant of STIM1 (ref. 6; see Supplementary Information, Fig. S5b). STIM1D76A is known to con‑stitutively cluster near the plasma membrane in the absence of store depletion, where it activates store‑operated Ca2+ influx pathways. As

expected, mCherry‑tagged STIM1D76A was pre‑localized to punctae at rest in NCM460 cells (data not shown) and caused a significant basal influx of Ca2+ (Supplementary Information, Fig. S5c). The find‑ing that the presence of defective STIM did not completely abolish the cAMP response may be explained by the existence of substantial levels of endogenous STIM1, which, in spite of some hindrance from the mutant form already residing in punctae, could still translocate to and communicate with the cAMP‑generating machinery (Supplementary Information, Fig. S5d).

3 min

a

0

100

80

60

40

20

120

0

100

80

60

40

20

120

0.9

1.1

1.3Iono Iono

b

3 min N.S. N.S.

GADPDH

STIM1

GAPDH

Not

Tre

ated

Scr

amb

le

STI

M1

siR

NA

STI

M1

siR

NA

NCM WT 30nM Si 48h

0

100

150

50

200

250

300

350

Con

trol

Scr

amb

le

d

3 min0.9

0.4

0.6

0.8 2mM Ca2+

STIM1

RT-PCR

c

STIM1 shRNA Scramble

STIM1 shRNA

0 Ca2+

0 Ca2+

0 Ca2+

Fsk,50 µM

Fsk,50 µM

Fsk, 2 µM Fsk, 2 µM

Em

issi

on r

atio

(480

/535

)

0.9

1.1

1.3

Em

issi

on r

atio

(480

/535

)

Control

mCherry (+)

Control

mCherry (+)

Control

mCherry (+)

cAMP cAMP

Per

cent

age

iono

myc

in c

ontr

ol

Per

cent

age

iono

myc

in c

ontr

ol

Con

trol

Ampl

itude

Slop

e

Con

trol

Ampl

itude

Slop

e

**

***

100µM ATP + tBHQ

Fura

-2 r

atio

Ca2+

Scr

amb

le

STI

M1

shR

NA

STIM1 siRNA

Iono

myc

in r

esp

onse

*****

Figure 5 STIM1 expression is required for store-operated cAMP signalling. (a) NCM460 cells stably expressing the Epac-based cAMP sensor transfected simultaneously with shRNAs against STIM1 and plasmid encoding mCherry48, the latter used to identify transfected cells. Isolated mCherry-positive cells (red traces) were compared with control cells in the same microscope field (grey trace, Epac sensor only; mean of 10 cells). The amplitude and speed of the ionomycin-induced cAMP responses were significantly reduced in mCherry-positive cells (amplitude, 38.3 ± 3.7% of control, ***P < 0.0001, rate (slope, 59.9 ± 6.9% of control, **P < 0.02; paired data from 66 control and 16 mCherry-positive cells, n = 13 experiments). (b) Control experiments: in cells co-transfected with a non-effective shRNA (scramble) plus mCherry. Neither the amplitude nor rate of the ionomycin-induced increase was significantly different from control (grey trace; mean of 14 control cells, 74 control and 13 mCherry-positive cells, n = 9 experiments). (c) Functional evidence for STIM1 knockdown. Capacitative Ca2+ entry measured using fura-2 was decreased in STIM1 shRNA-expressing cells (identified with mCherry co-expression), as reported previously6,7,49. The initial peak phase of Ca2+ release was not

significantly different in control cells (black trace; mean of 9 cells), compared with neighbouring STIM1 shRNA-transfected cells (co-transfected with mCherry; red trace) in the same microscope field. The entry phase following Ca2+ re-addition, however, was significantly reduced (amplitude, 42.4±9.6% of control in same field, P < 0.02; rate of increase, 40.8 ± 11.7% of control in same field, P < 0.05; 71 control cells, 7 mCherry-positive cells, n = 5 experiments). Scramble shRNA had no effect on release or entry phase (data not shown; 134 control, 10 mCherry-positive cells, 9 experiments). Inset: RT–PCR shows reduced expression of STIM1 (representative of 3 independent experiments). (d) An alternative siRNA-based approach was effective in knocking down STIM1 protein, as measured by western blotting (left panel; n = 4 experiments). As we could not co-transfect mCherry to identify siRNA-positive cells with this construct, we performed imaging experiments as in a and b, and averaged the population responses of all cells. Summary of pooled responses showed significant attenuation of the ionomycin response. Data are mean ± s.e.m. of 127 cells, 8 experiments for scramble; 135 cells, 14 experiments for STIM1 siRNA. **P < 0.02, ***P < 0.0001.



A RT I C L E S

Pharmacological inhibition of STIM1 translocation blocks SOcAMPSA recent report showed that ML‑9, a non‑specific myosin light chain kinase (MLCK) inhibitor, prevented or rapidly reversed store‑operated Ca2+ influx and re‑localization of STIM1. This group concluded, how‑ever, that ML‑9 was probably not exerting its effects on STIM1 dynamics by inhibiting MLCK26.

Pre‑treatment with ML‑9 significantly (P < 0.0001) inhibited ionomy‑cin‑induced SOcAMPS (Fig. 6a); on average the amplitude was reduced by 88% and 65% in the presence of 100 and 200 μM, respectively, of ML‑9. When the drug was washed out, cAMP recovered rapidly. Similar data were obtained using TPEN (Fig. 6a). Acute treatment with ML‑9 also quickly reversed the cAMP elevation evoked by ionomycin in the absence of forskolin (Fig. 6b; n = 169 cells, 12 experiments). Parallel measure‑ments using TIRF microscopy in NCM460 cells expressing YFP–STIM1 or mCherry–STIM1 showed that ML‑9 rapidly reversed the formation of STIM1 punctae elicited by treatment with ionomycin (Fig. 6c; n = 14 cells, 3 experiments) or TPEN (data not shown; n = 10 cells, 2 experiments). Pre‑treatment of cells with ML‑9 also prevented the redistribution of STIM1 ordinarily stimulated by ionomycin (n = 11 cells, 4 experiments) or TPEN (data not shown, n = 8 cells, 2 experiments). Notably, a sub‑stantial number of STIM1 punctae were detected in resting NCM460 cells (Fig. 6c, lower panel) and in resting ML‑9 treated cells (irrespec‑tive of expression levels of the STIM1 construct). As ML‑9 seemed to be blocking the forward movement of STIM1 cycling towards the plasma

membrane, it is possible that some component of the SOcAMPS exists in the absence of gross relocalization of STIM1 proteins, mediated by complexes already situated in or under the membrane. This may explain why ML‑9 was unable to completely inhibit SOcAMPS in experiments such as those of Fig. 6a and b.

We further observed that pre‑treatment with high concentrations of 2‑APB (50 μM), a non‑specific inhibitor of store‑operated Ca2+ entry, reversibly blocked SOcAMPS, as did ML‑9 (data not shown; n = 15 cells, 3 experiments.). Recently, it was reported that 2‑APB, like ML‑9, reverses STIM1 punctae formation27,26. We confirmed these findings using TIRF microscopy in NCM460 cells, where we observed that 2‑APB (50 μM) rapidly reversed (n = 11 cells, 4 experiments) or prevented (n = 9 cells, 3 experiments) relocalization of STIM1–YFP. The preceding results with ML‑9 and 2‑APB are consistent with a requirement for STIM1 transloca‑tion in the initiation of the SOcAMPS phenomenon.

DISCUSSIONDepletion of internal Ca2+ stores is a pervasive physiological conse‑quence of hormone‑stimulated Ca2+ signalling events28. Dysregulation of ER Ca2+ homeostasis also occurs pathologically after exposure to toxic insults, ischaemia and trauma, and is possibly linked to several neu‑rodegenerative disorders, including amyotrophic lateral sclerosis and Alzheimer’s, Parkinson’s and Huntington’s diseases. Prolonged loss of stored Ca2+ has catastrophic consequences for the cell, inducing ER stress responses and apoptosis29,30.

a

0.6

0.7

0.8

0.9

3 min

Per

cent

age

Iono

+M

L-9

Iono

+ M

L-9

Iono

TPEN

+ M

L-9

TPEN

Per

cent

age

TPE

N+

ML-

9

3 min.0.5

0.6

0.7

0.8

.0.5

b

0

20

40

60

80

100

120

***

Iono

lono

+ M

L-9

0 Ca2+ 0 Ca2+ 0 Ca2+

Fsk, 2 µM

Iono

Iono Iono ML-9Fsk,50 µM +

IBMX

Fsk,50 µM +

IBMX 250 µM

0

160

240

320

400

3 min

I

IIIII

IV

I II III IV

c

ML-9 200µM

ML-9

Em

issi

on r

atio

(480

/535

)

Em

issi

on r

atio

(480

/535

)

200

100

0

200

100

0

******

Per

cent

age

Iono

TPEN

Fluo

resc

ence

inte

nsity

1 mM

Control TPEN Iono ML-9

Figure 6 Translocation of STIM protein is necessary to initiate store-operated cAMP signalling. (a) Treatment of Epac sensor-expressing NCM460 cells with ML-9 (100–200 μM) significantly inhibited ionomycin-induced elevation of the FRET ratio. In the presence of ML-9 (100 μM), the ratio change after ionomycin (Iono, 5 μM) was 129.77 ± 12.4% of the initial forskolin (Fsk) response (mean ± s.e.m. of 24 cells in 3 experiments). In the presence of ML-9 (200 μM), the ratio change after Iono was 185.66 ± 18.85% of the initial Fsk response (mean ± s.e.m. of 56 cells in 5 experiments). The trace shows the mean of 3 cells. This was significantly different (***P < 0.0001 for both concentrations) from the response in the absence of ML-9 (345±18% of the initial forskolin response; 285 cells in 48 experiments.). Pre-treatment of NCM460 cells with ML-9 also reversibly attenuated the response to

TPEN using a similar protocol (summary data from 95 cells, 9 experiments, ***P < 0.0001). (b) Acute addition of ML-9 (200 μM) to cells stimulated with Iono in the absence of Fsk also caused a significant inhibition of the response (data from 169 cells in 12 experiments; ***P < 0.0001). (c) TIRF microscopy of NCM460 cells expressing YFP–STIM1 fusion protein. Chelating Ca2+ in the ER lumen with TPEN (250 μM or 1 mM) resulted in a concentration-dependent redistribution of STIM1 into discrete punctae, monitored as an increase in TIRF intensity. After a wash period, the same cell was challenged with Iono, resulting in profound re-localization of YFP–STIM1. Addition of ML-9 (200 μM) rapidly reversed the formation of these aggregates. Note the presence of small numbers of aggregates in the resting state. Scale bar, 10 μm.



A RT I C L E S

Here we describe a cellular mechanism that sensitizes the cAMP sig‑nalling machinery in the face of persistent loss of ER Ca2+. Lowering [Ca2+]ER using several independent strategies (Ca2+ ionophores, chelators, SERCA pump inhibitors, native agonists and passive depletion with high concentrations of EGTA) led to augmentation of agonist‑ or forskolin‑induced cAMP generation and a measurable increase in cAMP in rest‑ing cells (Figs 1–3). Although it is indisputable that cytosolic Ca2+ can influence cellular cAMP through regulation of ACs and PDEs, results using BAPTA‑loaded cells, acute addition of Ca2+ chelators (TPEN or BAPTA‑AM) or passive depletion with EGTA indicate that this effect was independent of Ca2+ in the cytosol. By all the criteria that have been used to define capacitative Ca2+ entry as a store‑operated process4, we show here that SOcAMPS is also a store‑dependent signalling pathway.

Taken together, the results using shRNA and siRNA for STIM1, the EF‑hand mutant of STIM1D76A and the STIM translocation inhibitors ML‑9 and 2‑APB, suggest that this ER transmembrane protein can serve as a sensor linking intraluminal Ca2+ levels to cAMP production. STIM1 is known to interact physically with Ca2+ entry channels of the Orai (or CRACM) family10,11,31. Orai1 co‑clusters with STIMs following store depletion32. Sophisticated chemical cross‑linking studies suggest that Orai1 is situated in a larger macromolecular complex when STIM1 approaches the plasma membrane after store depletion33. This could indicate the existence of other potential binding partners for STIM1. In fact, STIM1 has also recently been shown to regulate other types of store‑operated channels, namely members of the TRPC family34–38. Our data suggest still another regulatory role for STIM1, which is per‑haps working as a scaffolding protein that facilitates the recruitment and activation of ACs, which are also known to form complexes, for example AC5/AC2 heterodimers39. Although not entirely conclusive, our preliminary attempts to co‑immunoprecipitate AC5 and STIM1 (Supplementary Information, Fig. S4c) suggested that these two proteins were not bound to one another. The nature of the interactions between STIM1 and ACs, and the identity of specific AC isoforms involved has yet to be elucidated. However, our data raise the possibility that STIM1 links yet other membrane proteins (channels, transporters, signalling proteins) to Ca2+ store depletion.

Most of our studies were conducted using Ca2+‑free solutions to min‑imize the actions of Ca2+ entry on Ca2+‑sensitve ACs or PDEs13,14. Ca2+ influx had a clear inhibitory effect on cAMP signalling in NCM460 cells, possibly through the contribution of Ca2+‑inhibitable isoforms AC5 and AC6 (although a recent study40 also showed that STIM1 clus‑tering is inhibited by cytosolic Ca2+). Nevertheless, the main effect of store depletion in the presence of physiological external Ca2+ was an overall increase in cAMP (Fig. 3). This indicates that SOcAMPS was the predominant activity following prolonged store emptying, even in the presence of Ca2+.

As a decrease in [Ca2+]ER activates capacitative Ca2+ entry in nearly all cell types, the question arises as to why cells should possess an inde‑pendent store‑operated mechanism that seems to be redundant with the more conventional regulation of ACs occurring through cytosolic Ca2+. However, it is quite likely that physiological or pathological conditions exist in which persistent ER Ca2+ store depletion becomes uncoupled from capacitative Ca2+ influx. Elevation of Ca2+ levels caused by Ca2+ entry are frequently compensated over the long‑term by other Ca2+ homeostatic mechanisms, such as upregulation of Ca2+ pumping. Influx of Ca2+ can also be terminated by other signalling pathways (for example,

PKC, nitric oxide), modulated according to mitochondrial energy status and can differ during development4,28. It should also be noted that not all cells express Ca2+‑activated ACs (in fact, AC1 and AC8 do not seem to be expressed in NCM460; Supplementary Information, Fig. S4a).

Our preliminary survey of different human cell lines indicates that SOcAMPS is manifested to varying degrees in some cell types (NCM460, HEK293, HepG2, HT‑29) but not others (HeLa, H1299). The physiological circumstances in which this pathway is functional and its tissue distribution in vivo remain to be determined. However, our data may explain how certain cAMP‑dependent processes (for example, fluid and enzyme secretion, gene transcription) become syn‑ergized by concurrent activation of GPCRs coupled to cAMP and Ca2+ signalling cascades41,42. It may also account for previous reports of cel‑lular processes apparently activated by store depletion in the absence of cytosolic Ca2+ increases43–45.

Apart from serving as a reservoir of Ca2+ for intracellular signalling purposes, Ca2+ within the ER also fulfils several biochemical roles (for example, protein folding, glycosylation and processing). Prolonged dis‑ruption of ER Ca2+ homeostasis initiates a complex cascade of reactions known collectively as the ER stress response46. The SOcAMPS pathway provides a unique avenue for cross‑talk between ER Ca2+ and the cAMP second messenger system. Our data using AKAR3 and real‑time PCR (Fig. 2) showed that the increase in cAMP following store depletion was sufficient to activate downstream signalling through PKA and expression of the cAMP‑dependent gene, ICER. Because SOcAMPS was strongly acti‑vated following persistent depletion of [Ca2+]ER, the idea that this pathway may participate in ER stress responses, including the initiation of specific programs of gene expression, remains a distinct possibility.

METHODSReagents. Fura‑2‑AM, mag‑fura‑2‑AM, BAPTA‑AM and TPEN were obtained from Molecular Probes and Invitrogen. ML‑9 was from Calbiochem. All other reagents were from Sigma unless otherwise noted.

Cell culture and transfection. NCM460 cells20 were obtained by a licensing agree‑ment from INCELL Corporation, LLC and grown in M3:10 medium (INCELL) according to the supplier’s recommendations. CaCo‑2 cells, HEK293 and HeLa cells (ATCC) were grown in DMEM + 10% FCS according to the supplier’s recommendations. HT‑29 cells (ATCC) were grown in McCoy’s 5A modified medium containing 10% fetal calf serum (FCS). Pure populations of HEK293 and NCM460 cells stably expressing the Epac‑based cAMP sensor23 were gener‑ated by repeat sorting using a fluorescence activated cell sorter (FACS; BIDMC Flow Cytometry Core). Unless otherwise noted, all other constructs (AKAR3, mCherry–STIM1, etc.) were transiently transfected using Effectene transfection reagent (Qiagen) in cells grown on glass coverslips, as described previously16 and used 24–48 h later.

Ratio imaging. Real‑time FRET and fura‑2 ratio imaging experiments were per‑formed as described previously16 using fluorescence ratio imaging systems built around Nikon TE200 and Nikon TE2000‑U microscopes. Metafluor software (Molecular Devices) was used to control filter wheels (Sutter Instruments) placed in the excitation and emission path, and to acquire ratio data. Glass coverslips with cells were mounted in a home‑built flow through perfusion chamber, and cells imaged using a ×40 or ×63 oil immersion objective. Cells were bathed in a HEPES‑buffered Ringer’s solution containing (in mM): 125 NaCl, 25 HEPES, 10 Glucose, 5 K2HPO4, 1 MgSO4 and 1 CaCl2, pH, 7.40. For Ca2+‑free solutions (0 Ca2+), the CaCl2 was omitted and EGTA (25 μM) added to buffer contaminating Ca2+.

The 480 nm/535 nm FRET emission ratios from the Epac‑based cAMP sensor were acquired every 5–10 s, as were the 340 nm/385 nm excitation ratios (510 nm emission) for fura‑2 and mag‑fura‑2. PKA phosphorylation was measured using the 535 nm/480 nm FRET emission ratio (440 nm excitation) of AKAR3. Of note, AKAR3 relies on a gain of FRET between CFP and a circularly permuted Venus



A RT I C L E S

(cp Venus 172) to signify an increase in phosphorylation activity of PKA, whereas loss of FRET between the CFP/YFP FRET pair of the Epac H30 cAMP sensor indicates an elevation in cAMP. The fact that the FRET ratio results from these diverse sensors were consistent with each other lends further credibility to the idea that cAMP levels were increased when intraluminal [Ca2+]ER was reduced with TPEN or ionomycin. The fluorescence of mCherry (excitation 585 nm, 610 nm emission) did not interfere with any of the preceding fluorescence measurements. NCM460 cells showed robust transfer of second messengers through gap junc‑tions (K.L. and A.M.H., unpublished observations). Therefore, in experiments in which cells transfected with STIM constructs or shRNAs were compared with neighbouring control cells in the same field (for example, Fig. 5), great care was taken to select only those single cells not connected to other cells.

TPEN‑containing solutions were freshly prepared by dissolving crystals directly into Ringer’s with vigorous stirring for several hours. TPEN binds heavy metals (for example, Zn2+) with extremely high affinity. However, low concentrations of the chelator (20–40 μM), which are expected to completely bind up all cellular heavy metals17, had virtually no effect on the mag‑fura‑2 ratio, fura‑2 ratio, TIRF intensity or cAMP FRET ratio (see Supplementary Information, Fig. 3c for an example), consistent with the effects of TPEN working on compartments contain‑ing relatively high Ca2+, and not heavy metals.

Total internal reflection fluorescence (TIRF) microscopy. NCM460 cells expressing YFP–STIM1 or mCherry–STIM1 were imaged in real‑time in TIRF mode using a white‑light TIRF accessory from Nikon (T‑FL‑WTIRF) and a 1.49 or 1.45 N.A. ×60 Plan Apo TIRF oil‑immersion objective mounted on a Nikon TE2000‑U microscope. Metafluor software (Molecular Devices) was used to con‑trol hardware and acquire single wavelength fluorescence data (5–10‑s interval) as for other imaging experiments.

Knockdown of STIM1. Two independent knockdown strategies based on either siRNA or shRNA were used. For the first, coverslips of subconfluent NCM460 cells stably expressing the Epac‑based cAMP sensor H30 were transfected with synthetic siRNA (30 nM) against STIM1 (Ambion) using a high efficiency transfection reagent, Lipofectamine 2000 (Invitrogen). For the second, cover‑slips were transfected using Effectene transfection reagent (Qiagen) with pRS plasmids encoding four independent 29mer shRNAs against STIM1 (21 ng ml–1 for each plasmid; HuSH from OriGene Technologies). This latter approach had poor transfection efficiency (~5%) in NCM460 cells. This was overcome by co‑transfection with 8.4 ng ml–1 of plasmid encoding the red fluorescent protein mCherry to facilitate identification of transfected cells during imaging experi‑ments. A non‑effective shRNA against EGFP (84 ng ml–1; OriGene) non‑effective siRNA (30 nM; Ambion) was used for the respective control experiments. Forty‑eight hours after transfection, imaging experiments were performed. Responses from the siRNA‑transfected cells were pooled and averaged. Responses from individual shRNA‑transfected cells (identified by mCherry expression) were compared with control cells in the same microscope field.

STIM1 shRNA and siRNA sequences (from Fig. 5): HuSH 29mer shRNA con‑structs against STIM1, 5´‑GATGATGCCAATGGTGATGTGGATGTGGA‑3´; 5 ´ ‑ A C A G T G A A A C A C A G C A C C T T C C A T G G T G A ‑ 3 ´ ; 5´‑CTGCTGGTTTGCCTATATCCAGAACCGTT‑3´; 5´‑CTGAGCAGAGTC‑TGCATGACCTTCAGGAA‑3´; HuSH shRNA negative control (scramble), 5´‑TGACCACCCTGACCTACGGCGTGCAGTGCTCAAG; AGGCACTGC‑ACGCCGTAGGTCAGGGTGGTCA‑3´; Ambion Silencer Select STIM1 siRNA: sense sequence, 5´‑GCCUAUAUCCAGAACCGUUtt‑3´; antisense sequence, 5´‑AACGGUUCUGGAUAUAGGCaa‑3´; Ambion Silencer Select negative control no. 1 siRNA: sense sequence, 5´‑UAACGACGCGACGACGUAAtt‑3´; antisense sequence, 5´‑UUACGUCGUCGCGUCGUUAtt‑3´.

Real-time PCR experiments. Equal numbers of NCM460 cells were seeded in 6‑well Petri dishes. Twenty‑four hours after plating, SOcAMPS was induced in Ca2+‑free solutions with ionomycin (5 μM) in the presence of forskolin (2 μM) for 20 min. After 20 min. cells were rinsed four times with Ringer’s solution con‑taining EGTA (25 μM). After incubation in EGTA‑Ringer’s solution (2 h, total) the cells were lysed and RNA was obtained. Single‑strand complementary DNA (cDNA) was synthesized from mRNA (1 μg) using random hexamers. We chose to follow the expression of ICER (inducible cAMP early repressor). Expression of ICER is not affected by non‑cAMP signals, making this gene an ideal marker of

cAMP‑dependent transcriptional activation47. Total ICER/CREM mRNA levels for each treatment were obtained using multiplex real‑time RT–PCR using a TaqMan‑based assay (Applied Biosystems no. Hs01590456_m1) and normal‑ized to the endogenous 18S rRNA levels (the latter amplified using a standard primer set, Applied Biosystems no. 4319413E). ICER/CREM mRNA values for each treatment were calculated from four individual samples generated in 4 independent experiments.


ACKnowLedgeMentSWe are grateful to the following individuals for their kind gifts of plasmids: Kees Jalink for the Epac sensor, Roger Tsien for mCherry, Jin Zhang for AKAR3, Marie Dziadek and Lorna Johnstone for STIM1 and STIM2, Masaru Okabe for pCX–EGFP, and Tobias Meyer for YFP–STIM1, YFP–STIM1D76A, and YFP–STIM2 constructs. We thank Jessica Roy for technical assistance and Drs. Dheeraj Pelluru, Eberhard Frömter and Raj K. Goyal for helpful comments. This study was supported by a Merit Review award from the Department of Veteran’s Affairs (A.M.H.) and by an NIH Center grant from the Harvard Digestive Diseases Center (to A.M.H.). K.L. is the recipient of an American Heart Association Postdoctoral Fellowship award..

AutHor ContrIbutIonSK.L., I.M. and A.M.H. performed experiments; K.L., A.M.H., and S.C. designed experiments, and along with M.S., analyzed data. M.P.M. provided reagents. A.M.H. wrote the manuscript.

CoMPetIng fInAnCIAL IntereStSM.P.M. holds partial ownership of INCELL Corporation, which sells the M3:10 tissue culture medium used to grow NCM460 cells. The other authors declare no competing financial interests.


1. Neves, S. R., Ram, P. T. & Iyengar, R. G protein pathways. Science 296, 1636–1639 (2002).

2. Luik, R. M. & Lewis, R. S. New insights into the molecular mechanisms of store-operated Ca2+ signaling in T cells. Trends Mol. Med. 13, 103–107 (2007).

3. Putney, J. W., Jr New molecular players in capacitative Ca2+ entry. J. Cell Sci. 120, 1959–1965 (2007).

4. Parekh, A. B. & Putney, J. W., Jr Store-operated calcium channels. Physiol. Rev. 85, 757–810 (2005).

5. Dziadek, M. A. & Johnstone, L. S. Biochemical properties and cellular localisation of STIM proteins. Cell Calcium 42, 123–132 (2007).

6. Liou, J. et al. STIM is a Ca2+ sensor essential for Ca2+-store-depletion-triggered Ca2+ influx. Curr. Biol. 15, 1235–1241 (2005).

7. Zhang, S. L. et al. STIM1 is a Ca2+ sensor that activates CRAC channels and migrates from the Ca2+ store to the plasma membrane. Nature 437, 902–905 (2005).

8. Roos, J. et al. STIM1, an essential and conserved component of store-operated Ca2+ channel function. J. Cell Biol. 169, 435–445 (2005).

9. Liou, J., Fivaz, M., Inoue, T. & Meyer, T. Live-cell imaging reveals sequential oligomeri-zation and local plasma membrane targeting of stromal interaction molecule 1 after Ca2+ store depletion. Proc. Natl Acad. Sci. USA 104, 9301–9306 (2007).

10. Prakriya, M. et al. Orai1 is an essential pore subunit of the CRAC channel. Nature 443, 230–233 (2006).

11. Vig, M. et al. CRACM1 is a plasma membrane protein essential for store-operated Ca2+ entry. Science 312, 1220–1223 (2006).

12. Yeromin, A. V. et al. Molecular identification of the CRAC channel by altered ion selec-tivity in a mutant of Orai. Nature 443, 226–229 (2006).

13. Willoughby, D. & Cooper, D. M. Organization and Ca2+ regulation of adenylyl cyclases in cAMP microdomains. Physiol. Rev. 87, 965–1010 (2007).

14. Goraya, T. A. & Cooper, D. M. Ca2+-calmodulin-dependent phosphodiesterase (PDE1): current perspectives. Cell Signal. 17, 789–797 (2005).

15. Houslay, M. D. & Milligan, G. Tailoring cAMP-signalling responses through isoform multiplicity. Trends Biochem. Sci. 22, 217–224 (1997).

16. Gerbino, A. et al. Termination of cAMP signals by Ca2+ and G(α)i via extracellular Ca2+ sensors: a link to intracellular Ca2+ oscillations. J. Cell Biol. 171, 303–312 (2005).

17. Arslan, P., Di Virgilio, F., Beltrame, M., Tsien, R. Y. & Pozzan, T. Cytosolic Ca2+ home-ostasis in Ehrlich and Yoshida carcinomas. A new, membrane-permeant chelator of heavy metals reveals that these ascites tumor cell lines have normal cytosolic free Ca2+. J. Biol. Chem. 260, 2719–2727 (1985).

18. Caroppo, R. et al. A reassessment of the effects of luminal [Ca2+] on inositol 1, 4, 5-trisphosphate-induced Ca2+ release from internal stores. J. Biol. Chem. 278, 39503–39508 (2003).

19. Hofer, A. M., Fasolato, C. & Pozzan, T. Capacitative Ca2+ entry is closely linked to the filling state of internal Ca2+ stores: a study using simultaneous measurements of ICRAC and intraluminal [Ca2+]. J. Cell Biol. 140, 325–334 (1998).






A RT I C L E S

20. Moyer, M. P., Manzano, L. A., Merriman, R. L., Stauffer, J. S. & Tanzer, L. R. NCM460, a normal human colon mucosal epithelial cell line. In Vitro Cell Dev. Biol. Anim. 32, 315–317 (1996).

21. Hofer, A. M. & Machen, T. E. Technique for in situ measurement of calcium in intra-cellular inositol 1, 4, 5-trisphosphate-sensitive stores using the fluorescent indicator mag-fura-2. Proc. Natl Acad. Sci. USA 90, 2598–2602 (1993).

22. Wu, M. M., Buchanan, J., Luik, R. M. & Lewis, R. S. Ca2+ store depletion causes STIM1 to accumulate in ER regions closely associated with the plasma membrane. J. Cell Biol. 174, 803–813 (2006).

23. Ponsioen, B. et al. Detecting cAMP-induced Epac activation by fluorescence reso-nance energy transfer: Epac as a novel cAMP indicator. EMBO Rep. 5, 1176–1180 (2004).

24. Allen, M. D. & Zhang, J. Subcellular dynamics of protein kinase A activity visualized by FRET-based reporters. Biochem. Biophys. Res. Commun. 348, 716–721 (2006).

25. Sassone-Corsi, P. Transcription factors responsive to cAMP. Annu. Rev. Cell Dev. Biol 11, 355–377 (1995).

26. Smyth, J. T., Dehaven, W. I., Bird, G. S. & Putney, J. W., Jr. Ca2+-store-dependent and -independent reversal of Stim1 localization and function. J. Cell Sci. 121, 762–772 (2008).

27. Peinelt, C., Lis, A., Beck, A., Fleig, A. & Penner, R. 2-APB directly facilitates and indirectly inhibits STIM1-dependent gating of CRAC channels. J. Physiol. (2008).

28. Rizzuto, R. & Pozzan, T. Microdomains of intracellular Ca2+: molecular determinants and functional consequences. Physiol. Rev. 86, 369–408 (2006).

29. Paschen, W. & Mengesdorf, T. Endoplasmic reticulum stress response and neurode-generation. Cell Calcium 38, 409–415 (2005).

30. Burdakov, D., Petersen, O. H. & Verkhratsky, A. Intraluminal calcium as a primary regulator of endoplasmic reticulum function. Cell Calcium 38, 303–310 (2005).

31. Gwack, Y. et al. Biochemical and functional characterization of Orai proteins. J. Biol. Chem. 282, 16232–16243 (2007).

32. Xu, P. et al. Aggregation of STIM1 underneath the plasma membrane induces clustering of Orai1. Biochem. Biophys. Res. Commun. 350, 969–976 (2006).

33. Varnai, P., Toth, B., Toth, D. J., Hunyady, L. & Balla, T. Visualization and manipulation of plasma membrane-endoplasmic reticulum contact sites indicates the presence of additional molecular components within the STIM1–Orai1 complex. J. Biol. Chem. 282, 29678–29690 (2007).

34. Liao, Y. et al. Functional interactions among Orai1, TRPCs, and STIM1 suggest a STIM-regulated heteromeric Orai/TRPC model for SOCE/Icrac channels. Proc. Natl Acad. Sci. USA 105, 2895–2900 (2008).

35. Cheng, K. T., Liu, X., Ong, H. L. & Ambudkar, I. S. Functional requirement for Orai1 in store-operated TRPC1/STIM1 channels. J. Biol. Chem. (2008).

36. Huang, G. N. et al. STIM1 carboxy terminus activates native SOC, I(crac) and TRPC1 channels. Nature Cell Biol. 8, 1003–1010 (2006).

37. Yuan, J. P., Zeng, W., Huang, G. N., Worley, P. F. & Muallem, S. STIM1 heteromultimer-izes TRPC channels to determine their function as store-operated channels. Nature Cell Biol. 9, 636–645 (2007).

38. Ong, H. L. et al. Dynamic assembly of TRPC1–STIM1–Orai1 ternary complex is involved in store-operated calcium influx. Evidence for similarities in store-operated and cal-cium release-activated calcium channel components. J. Biol. Chem. 282, 9105–9116 (2007).

39. Baragli, A., Grieco, M. L., Trieu, P., Villeneuve, L. R. & Hebert, T. E. Heterodimers of adenylyl cyclases 2 and 5 show enhanced functional responses in the presence of Gα s. Cell Signal. 20, 480–492 (2008).

40. Malli, R., Naghdi, S., Romanin, C. & Graier, W. F. Cytosolic Ca2+ prevents the subplas-malemmal clustering of STIM1: an intrinsic mechanism to avoid Ca2+ overload. J. Cell Sci. 121, 3133–3139 (2008).

41. Selbie, L. A. & Hill, S. J. G protein-coupled-receptor cross-talk: the fine-tuning of multiple receptor-signalling pathways. Trends Pharmacol. Sci. 19, 87–93 (1998).

42. Urushidani, T. & Forte, J. G. Signal transduction and activation of acid secretion in the parietal cell. J. Membr. Biol. 159, 99–111 (1997).

43. Maloney, J. A. et al. Activation of ERK by Ca2+ store depletion in rat liver epithelial cells. Am. J. Physiol. 276, C221–230 (1999).

44. Strayer, D. S., Hoek, J. B., Thomas, A. P. & White, M. K. Cellular activation by Ca2+ release from stores in the endoplasmic reticulum but not by increased free Ca2+ in the cytosol. Biochem. J. 344, 39–46 (1999).

45. Sargeant, P., Farndale, R. W. & Sage, S. O. Calcium store depletion in dimethyl BAPTA-loaded human platelets increases protein tyrosine phosphorylation in the absence of a rise in cytosolic calcium. Exp. Physiol. 79, 269–272 (1994).

46. Yoshida, H. ER stress and diseases. FEBS J. 274, 630–658 (2007).47. Mayr, B. & Montminy, M. Transcriptional regulation by the phosphorylation-dependent

factor CREB. Nature Rev. Mol. Cell Biol. 2, 599–609 (2001).48. Shaner, N. C. et al. Improved monomeric red, orange and yellow fluorescent proteins

derived from Discosoma sp. red fluorescent protein. Nature Biotechnol. 22, 1567–1572 (2004).

49. Williams, R. T. et al. Identification and characterization of the STIM (stromal interaction molecule) gene family: coding for a novel class of transmembrane proteins. Biochem. J. 357, 673–685 (2001).



LETTERS

Myosin IIIa boosts elongation of stereocilia by transporting espin 1 to the plus ends of actin filamentsFelipe T. Salles1,6, Raymond C. Merritt, Jr1,2,6, Uri Manor1, Gerard W. Dougherty1,5, Aurea D. Sousa1, Judy E. Moore3, Christopher M.Yengo3, Andréa C. Dosé4 and Bechara Kachar1,7

Two proteins implicated in inherited deafness, myosin IIIa1, a plus-end-directed motor2, and espin3–6, an actin-bundling protein containing the actin-monomer-binding motif WH2, have been shown to influence the length of mechanosensory stereocilia7,8. Here we report that espin 1, an ankyrin repeat-containing isoform of espin6, colocalizes with myosin IIIa at stereocilia tips and interacts with a unique conserved domain of myosin IIIa. We show that combined overexpression of these proteins causes greater elongation of stereocilia, compared with overexpression of either myosin IIIa alone or espin 1 alone. When these two proteins were co-expressed in the fibroblast-like COS-7 cell line they induced a tenfold elongation of filopodia. This extraordinary filopodia elongation results from the transport of espin 1 to the plus ends of F-actin by myosin IIIa and depends on espin 1 WH2 activity. This study provides the basis for understanding the role of myosin IIIa and espin 1 in regulating stereocilia length, and presents a physiological example where myosins can boost elongation of actin protrusions by transporting actin regulatory factors to the plus ends of actin filaments.

Stereocilia, the prominent actin protrusions on the apical surfaces of sensory hair cells, emerge early during development and their lengths are maintained at fixed heights for the lifetime of the organism. The bundle of parallel actin filaments that make up the core of each stereocilium is continually renewed, with the entire actin bundle constantly assem-bled at the tip, treadmilling downward and disassembling at the base9–11. Given that stereocilia can be up to 100 μm in length, it is likely that some form of regulated transport is necessary to localize components of the actin polymerization machinery to the plus end of actin filaments. Although several myosins have been shown to alter stereocilia length and shape, depending on their expression levels9,12–16, the mechanisms by which these motors or their binding partners regulate actin dynamics and stereocilia length remain unclear.

Using antibodies specific for the ankyrin repeat domain (ARD) of espin 1 (Supplementary Information, Fig. S1), we show localization at stereocilia tips with a tip-to-base gradient distribution (Fig. 1) similar to that previ-ously described for myosin IIIa. Immunofluorescence of espin 1 was more intense in the longer stereocilia where the characteristic tip-to-base fluo-rescence intensity gradient also had a longer decay length (Fig. 1g, h). In contrast to other espin isoforms, which are present inside the actin core and along the entire stereocilia length8, espin 1 was excluded from the actin cores and formed a thimble-like distribution at the tips of stereocilia (Fig. 1). Immunofluorescence in developing hair cells of the rat organ of Corti showed that espin 1 can be detected at the tips of stereocilia during their elongation and maturation phases (Fig. 1k–m). To confirm the locali-zation of espin 1 at the tips of stereocilia, we overexpressed GFP–espin 1 in organotypic cultures of hair cells. Transfected hair cells show that GFP–espin 1 localizes at the tips of stereocilia, showing a tip-to-base gradient of intensity (Fig. 2) comparable to the immunolocalization (Fig. 1). These localization patterns — tip-to-base gradients and thimble-like distributions7, as well as the temporal expression pattern17 — closely match those of myosin IIIa. We hypothesized that targeting espin 1 to stereocilia tips, which is the site of actin polymerization9,11, influences actin polymerization and stereocilia elonga-tion. Analysis of the heights of stereocilia of cochlear and vestibular hair cells transfected with GFP–espin 1 showed that stereocilia were elongated when espin 1 was overexpressed (Fig. 2), consistent with our hypothesis.

The striking similarity of the tip-to-base gradient localization of both espin 1 and myosin IIIa prompted us to investigate whether myosin IIIa helps localize espin 1 to the tips of stereocilia and whether they have a com-bined role in the regulation of stereocilia length. We compared stereocilia length in hair cells transfected with espin alone, myosin IIIa alone and with a combination of both plasmids (Fig. 2). Hair cells transfected with myosin IIIa and espin 1 showed an increase in stereocilia length, higher than the combined increase observed for myosin IIIa alone and espin 1 alone (Fig. 2). It is important to note that any analysis of lengthening due to overexpression of myosin IIIa and espin 1 in hair cells must take into

1Laboratory of Cell Structure and Dynamics, National Institute on Deafness and Other Communication Disorders, National Institutes of Health, Bethesda, MD 20892, USA. 2Department of Biology, University of Maryland, College Park, MD 20742, USA. 3Department of Biology, University of North Carolina at Charlotte, Charlotte, NC 28223, USA. 4Department of Molecular and Cell Biology, University of California, Berkeley, CA 94720, USA. 5Current address: Consorzio Mario Negri Sud, Department of Cell Biology and Oncology, Santa Maria Imbaro, Chieti, Italy 66030.6These authors contributed equally to the work.7Correspondence should be addressed to B.K (e-mail: [email protected])

Received 13 November 2008; accepted 19 January 2009; published online 15 March 2009; DOI: 10.1038/ncb1851




L E T T E R S

account intrinsic limitations due to natural variations in stereocilia length and, importantly, the fact that stereocilia are already quite elongated and express substantial amounts of endogenous espin 1 and myosin IIIa.

We tested whether myosin IIIa can effectively interact with and trans-port espin 1 in COS-7 cells, using filopodia as a model to study actin pro-trusions. Myosin IIIa and espin 1 are not naturally expressed at detectable

m

l

k

j

i

S

M

T

µm64200

0.2

0.4

0.6

0.8

1rpih

S

M

T

g

f

e

d

Pan-espin

Actin

Merge

c

Myosin IIIa

Actin

Merge

b

Espin 1

Actin

Merge

a

Figure 1 Espin 1 distribution in stereocilia is similar to myosin IIIa distribution. (a–c) Confocal images show that espin 1 (green, a) localized at the tips of stereocilia in rat cochlear hair cells at postnatal day (P) 6 matches the localization observed for myosin IIIa (green, b). In contrast, pan-espin labelling (green) is seen along the entire stereocilia (c). Scale bar, 5 μm. (d–f) Immunogold labelling shows espin 1 localized at the tip of stereocilia around the actin core (d, e), whereas labelling with a pan-espin antibody shows localization throughout the entire stereocilia actin cores (f) in plunge frozen vestibular (d, f) or directly frozen cochlear (d) tissues obtained from adult rats. Scale bars, 200 nm. (g) Espin 1 (green) localization in guinea pig vestibular stereocilia reveals a tip-to-base gradient that is more extended in longer stereocilia (inset T) than medium (inset M) and short (inset S) stereocilia. (h) Measurements of

green (espin 1) and red (actin) relative pixel intensity (rpi) of fluorescence along the stereocilia for each rectangular inset in g confirm this tip-to-base concentration gradient of espin 1 immunofluorescence. (i, j) Longitudinal (i) and cross-section (j) images reveal a thimble-like distribution of espin 1 (green) at the tips of stereocilia in P8 rat cochlear hair cells. (k–m) Immunofluorescence of rat cochlear hair cells at different developmental time-points shows that espin 1 targets stereocilia tips as early as P0 (k) and undergoes progressive compartmentalization at the tips during P2 (l) and P4 (m), reaching a peak of intensity at about P6 as shown in b. Scale bars, 5 μm. Antibodies: espin 1 (PB539), myosin IIIa (PB638) and pan-espin (PB127). In all immunofluorescence images, the immunolabelling was visualized using Alexa-488-conjugated secondary antibody and F-actin (red) is visualized using Alexa 568-phalloidin.



L E T T E R S

levels in COS-7 cells7. Myosin IIIa has been shown to induce filopodial actin protrusions and localize to their tips in cultured cells particularly well when its kinase domain is removed (myosin IIIaΔK)7,18, suggesting that the kinase could serve to downregulate the functional activity of myosin IIIa. We examined the distribution of co-expressed mCherry–ARD of espin 1 with GFP-tagged myosin IIIaΔK (Fig. 3). We also co-expressed mCherry–ARD with GFP-tagged myosin X and GFP-tagged myosin XVa. As all of these myosins accumulate at the tips of filopodia13,16,19, they pro-vide a well-defined spatial compartment where any potential interaction can be clearly visualized. Co-transfections showed that mCherry–ARD is efficiently targeted to the tips of filopodia initiated by myosin IIIaΔK, but not by myosins X or XVa (Fig. 3a), demonstrating a specific colo-calization of ARD with myosin IIIa. Live imaging of COS-7 cells trans-fected with GFP–myosin IIIaΔK and mCherry–ARD showed dynamic colocalization at the filopodia tips from the early steps of their initiation

and elongation (Fig. 3b; Supplementary Information, Movie 1). Live imaging also showed matching forward and rearward intra-filopodial movements of the GFP–myosin IIIaΔK and mCherry–ARD fluorescence puncta while maintaining steady-state tip-to-base distributions (Fig. 3c; Supplementary Information, Movies 2, 3), similar to the distributions of myosin IIIa and espin 1 observed in stereocilia (Figs 1, 2). The intensity profiles for mCherry–ARD and GFP–myosin IIIaΔK within each frame of the video (Fig. 3c) are closely correlated (cross-correlation = ~0.990), sup-porting the view that these two proteins traffic together in the filopodia. The interaction between espin 1 ARD and myosin IIIaΔK was confirmed with a GST pulldown assay (Fig. 3e). The dynamic localization of espin 1 ARD at filopodia tips when co-transfected with myosin IIIa, but not with myosin X or with myosin XVa, along with our GST pulldown assay results, led us to hypothesize that myosin IIIa transports espin 1 to the tips of stereocilia.

0

0.5

1

1.5

2

2.5

Espin 1 +Myosin IIIa

Espin 1 Myosin IIIa

lg

GFPActin

fGFP–myosin IIIa + espin 1

GFPActin

eGFP–myosin IIIa + espin 1

GFPActin

dGFP–myosin IIIa

GFPActin

cGFP–myosin IIIa

Espin 1

Actin

µm2.52.01.510.500

0.4

0.6

0.8

1

rpi

baGFP–espin 1

GFPActin

Figure 2 Espin 1 alone or when overexpressed with myosin IIIa elongates stereocilia. (a) GFP–espin 1 localizes to stereocilia tips in a tip-to-base gradient distribution in transfected cultured organ of Corti hair cells. (b) High magnification close-up view (upper panel) and measurement of the relative pixel intensity (rpi, lower panel) of GFP–espin 1 and actin (Alexa 568-phalloidin) fluorescence along the distal portion of the stereocilia shown in the rectangular inset in a matches the tip-to-base concentration-gradient observed for endogenous espin 1 (Fig. 1h). (c–f) Organ of Corti (c) and vestibular (d) hair cells transfected with GFP–myosin IIIa show tip localization similar to that of espin 1. Co-transfection of vestibular hair cells with GFP–myosin IIIa (arrow, green) and untagged espin 1 together (e, f, images of two different cells) produce longer stereocilia than in

hair cells transfected with GFP–espin 1 alone (a) or with GFP–myosin IIIa alone (c, d). (g) The average ratios of stereocilia length (l) between transfected (HT) and neighbouring non-transfected (HNT) cells, l = HT

HNT .

GFP–espin 1 alone = 1.5 ± 0.67, n = 19 (~50% increase); GFP–myosin IIIa alone = 1.1 ± 0.14, n = 16 (~10% increase); and GFP–myosin IIIa and espin 1 = 2.3 ± 0.69, n = 14, (~130% increase). Note that a value of l = 1 (indicated by the dotted line in the graph) corresponds to a zero percent increase in length. Data are mean ± s.d.; the mean value for the hair cells co-transfected with espin1 and myosin IIIa was significantly higher than that of the cells transfected with espin 1 alone (P = 0.002, ANOVA); the mean bundle heights for hair cells transfected with myosin IIIa alone were not significantly higher than the controls (P = 0.149, ANOVA). Scale bars, 5 μm.



L E T T E R S

We next asked which specific region of myosin IIIa is involved in the interaction with espin 1. Myosin IIIa has two conserved tail homology domains, designated as 3THDI and 3THDII20. We first co-transfected COS-7 cells with espin 1 and GFP–myosin IIIa and showed that the two

proteins colocalize at actin bundles as well as at filopodia tips (Fig. 4a, b). This pattern of colocalization was abolished when we used a GFP–myosin IIIa construct that lacks the portion of the tail containing both 3THDI and 3THDII (GFP–myoIIIaΔ32; Fig. 4a, b; Supplementary Information,

Anti-GFPAnti-GST

30-

60-

80-

120-

220-

GS

T

GS

T–A

RD

Sup

erna

tant

Hom

ogen

ate

GS

T

GS

T–A

RD

GFP–myosin IIIa∆KInpute

motor

3TH

DII

3TH

DI

GST–ARDd

µm 210

mCherry-ARD

0

0.5

1rpi

Merge

mCherry–ARD

GFP–myosin IIIa∆K



c

b

Mer

gem

Che

rry–

AR

D

GFP–myosin XVaGFP–myosin XGFP–myosin IIIa∆K a

Mr(K)

Figure 3 Espin 1 interacts with myosin IIIa through its ankyrin repeats domain (ARD) in transfected COS-7 cells. (a) mCherry-tagged espin 1 ARD shows filopodia tip localization when co-transfected with GFP–myoIIIaΔK but failed to localize at the filopodia tips when co-transfected with GFP–myosin X or GFP–myosin XVa. Scale bar, 5 μm. (b) Selected video images from a time-lapse video (Supplementary Information, Movie 1) show co-transport of GFP–myoIIIaΔK and mCherry–ARD during the formation and extension of filopodia in COS-7 cells. Scale bar, 2.5 μm. (c) Single representative frame from a time-lapse video of a single filopodium (Supplementary Information, Movie 2) graph with the normalized relative pixel intensity of the fluorescence along the filopodium show co-distribution (cross-correlation of the green and red intensity

values = 0.990 for this frame; the average cross-correlation for six randomly selected frames from the same video was 0.99) of GFP–myoIIIaΔK and mCherry–ARD forming a tip-to-base gradient and colocalization of the green and red fluorescence puncta (arrowheads) corresponding to the co-transport of GFP–myoIIIaΔK and mCherry–ARD trafficking into and out of the tip of the filopodia. Scale bar, 0.5 μm. (d) Schematic representation of the constructs analysed in this figure. GST, glutathione S-transferase; GFP, green fluorescence protein; myosin IIIaΔK, myosin IIIa with a deletion of it kinase domain; 3THDI and 3THDII, myosin IIIa tail homology domains 1 and 2, respectively. (e) Western blots of GST pulldowns show that purified GST–ARD co-precipitates with GFP–myosin IIIaΔK, but not with GST alone. Precipitates were detected using anti-GST and anti-GFP antibodies.



L E T T E R S

Table S1, Fig. S2). We next co-transfected COS-7 cells with espin 1 and with a GFP-tagged tail portion lacking 3THDII (GFP–tailΔ3THDII; Fig. 4a, b) and narrowed down the region of interaction to the 3THDI and its immediate flanking regions. We observed colocalization of espin 1 with GFP–myosin IIIa tail that contained only the 3THDI domain (GFP–3THDI; Fig. 4a, b), but not with regions of only the myosin IIIa tail immediately amino-terminal (pre3THDI) or carboxy-terminal (post3THDI) to the 3THDI domain (Fig. 4a, b). This suggests that the 3THDI domain is necessary for the myosin IIIa–espin 1 interaction. Together these data suggest that espin 1 and myosin IIIa interact specifi-cally through their ARD and 3THDI domains, respectively. We verified this interaction in vitro using a GST pulldown assay and showed that GST–ARD binds to GFP–3THDI, but not to the pre3THDI or post-3THDI regions (Fig. 4c).

The fact that stereocilia length can be influenced by either espin 1 (refs 3, 8) or myosin IIIa7, along with the observation that they both localize to the same compartment at stereocilia tips and interact biochemically, suggests a combined functional role for the myosin IIIa–espin 1 complex in the elongation of stereocilia F-actin. We dis-covered that COS-7 cells co-transfected with myosin IIIaΔK and espin 1 (Fig. 5a–c) show filopodial actin protrusions that can be up to ten times longer (mean length = 14.3 ± 9.1 μm; number of cells, nc = 18; number of filopodia, nf = 56) than those transfected with myosin IIIaΔK alone (1.7 ± 0.83 μm, nc = 12, nf = 49), or with espin 1 alone (1.3 ± 0.28 μm, nc = 13, nf = 104). Mean lengths of filopodia of COS-7 cells transfected with an empty GFP vector was 1.26 ± 0.7 (nc = 10, nf = 59). The syn-ergistic effect between myosin IIIa and espin 1 is specific for myosin IIIa, as elongation was not enhanced when espin 1 was co-expressed with either myosin X (2.40 ± 1.50 μm, nc = 16, nf = 165) or myosin XVa (2.08 ± 1.63 μm, nc = 15, nf = 134).

We used myosin IIIa without the kinase domain to observe the behav-iour of the dephosphorylated and more functionally active myosin. To exclude the possibility that deletion of the kinase domain produces aber-rant behaviour, we developed a kinase-dead construct, myosin IIIaK50R (Supplementary Information, Table S1). This construct allowed us to examine the role of autophosphorylation in the regulation of motor function, which in turn enabled us to investigate the role of myosin IIIa motor function in espin 1 tip-localization activity. We have deter-mined that inactivation of the myosin IIIa kinase in a myosin IIIa 2IQ construct reduces KATPase, yet it does not affect maximal ATPase activ-ity (Supplementary Information, Table S2 and Fig. S3). We next evalu-ated the role of kinase activity in myosin IIIa tip localization in COS-7 cells using GFP-tagged constructs. Full-length myosin IIIaK50R local-ized more efficiently to the tips of filopodia in COS-7 cells (39% at tips nc = 137) than wild-type (5% at tips nc = 200), although not as strikingly as myosin IIIaΔΚ (93% at tips n = 105). Furthermore, co-expression of myosin IIIaK50R and espin 1 (Fig. 5e) produced longer filopodia (mean length = 5.93 ± 3.10 μm, nc = 15, n f = 89) than those produced by co-expression of wild-type myosin IIIa and espin 1 (3.7 ± 3.2 μm, nc = 15, nf = 63; Fig. 5d), but not as long as those seen with myosin IIIaΔK and espin 1 co-expression. These data show that myosin IIIa motor ATPase activity parallels the ability of myosin IIIa to localize to filopodia tips and to elongate filopodia when co-expressed with espin 1.

Interestingly, espin 1 co-expressed with a myosin IIIaΔK lacking the tail domain downstream of exon 32 (myosin IIIaΔK,33,34; Supplementary Information, Table S1, Fig. S2) resulted in slightly shorter filopodia

(10.0 ± 4.74 μm, n = 64; Fig. 5f) than co-expression with myosin IIIaΔK. Using COS-7 cell co-expression and GST pulldown assays, we con-firmed that the upstream portion of 3THDI (3THDIΔ33, Supplementary Information, Fig. S4) binds to espin 1. The 3THDII of myosin IIIa has

Anti-GFPAnti-GSTAnti-GFPAnti-GSTAnti-GFPAnti-GST

30-

40-

60-80-

GS

TG

ST–

AR

D

Sup

erna

tant

Hom

ogen

ate

GS

T

GS

T–A

RD

GS

TG

ST–

AR

D

Sup

erna

tant

Hom

ogen

ate

GS

T

GS

T–A

RD

GS

TG

ST–

AR

D

Sup

erna

tant

Hom

ogen

ate

GS

T

GS

T–A

RD

GFP–post3THDIInputGFP–3THDIInputGFP–pre3THDIInputc

post3THDI

3THDI

pre3THDI

Myosin IIIa

MergeAnti-espin 1 (PB539)GFPb

espin 1ABMWH2ARD

GFP–post3THDIGFP–3THDI

GFP–pre3THDIGFP–tail 3THDII

GFP–myosin IIIa∆32

GFP–myosin IIIa3THDIa

Mr(K)

Tail∆3THDII

Myosin IIIa∆32

Figure 4 Myosin IIIa interacts with espin 1 through its 3THDI domain. (a) Schematic representation of the espin 1 and myosin IIIa constructs analysed in this figure. ABM, actin binding module; WH2, Wiskott-Aldrich homology domain 2; GFP–myosin IIIaΔ32, myosin IIIa lacking exon 32 which causes a frame shift rendering the protein without the 3THDI and 3THDII domains. (b) Co-expression of untagged espin 1 shows that GFP–myosin IIIa, GFP–tailΔ3THDII, and GFP–3THDI (green) colocalize with espin 1 (red) along actin filament bundles. In contrast, GFP–myosin IIIaΔ32, GFP–pre-, and post3THDI are dispersed in the cytoplasm, despite the presence of espin 1 bundles. Scale bar, 5 μm. (c) Western blots of GST pulldowns confirm that the 3THDI region of myosin IIIa is necessary and sufficient for binding to espin 1-ARD, as pre3THDI and post3THDI show no binding to GST–ARD. Precipitates were detected using anti-GST and anti-GFP antibodies.



L E T T E R S

been shown previously to be an actin-binding site18. Previous studies reported that myosin IIIa lacking the 3THDII actin-binding domain does not localize to filopodia tips7,18, but here we show that when co-expressed with espin 1, myosin IIIa goes to the tip and promotes filopo-dia elongation (Fig. 5f). It seems that the association with espin 1, which does have actin-binding sites, compensates for the missing actin-binding site in the myosin IIIa without the 3THDII domain.

Co-expression of espin 1 and myosin IIIa results in enhanced localiza-tion of espin 1 at filopodia tips (Supplementary Information, Fig. S5). When myosin IIIaΔK was co-expressed with espin 1 lacking the ARD domain, both espin tip localization and filopodia elongation were abol-ished (Fig. 5g). These results show that the actin crosslinking activity of espin 1 is not solely responsible for the enhanced elongation of filopodia or stereocilia observed in our experiments. We conclude that espin 1

promotes enhanced elongation of filopodia only when transported to the polymerization end of actin filaments by myosin IIIa. The finding that espin 1 elongates filopodia only when localized to the F-actin plus ends by myosin IIIa suggests that WH2-dependent polymerization activity is involved in elongation. We tested this hypothesis by substituting the first two of three highly conserved Leu residues of the espin 1 WH2 motif (L655A, L656A), which have been shown to be essential for its actin-monomer-binding activity21,22. In COS-7 cells co-transfected with the WH2-mutated espin 1 construct (espin 1mWH2) and myosin IIIaΔK (Fig. 5h), the average length of filopodia (2.65 ± 1.50 μm, nc = 10, nf = 75) remained comparable to the protrusions induced by myosin IIIaΔK alone. The lack of enhanced elongation despite the colocalization of espin 1mWH2 and myosin IIIaΔK at the tips of filopodia (Fig. 5h) demonstrates that the WH2 motif is essential for the effect of espin 1 in elongation.

a b c d e f hg

30

20

15

10

5

0GFPPB539

h

GFP–myosin IIIa∆K + espin 1mWH2

GFPPB127

gGFP–myosin IIIa∆K + espin 1∆ARD

GFPActin

fGFP–myosin IIIa∆K,33,34 + espin 1

GFPActin

e

GFP–myosin IIIaK50R + espin 1

GFPActin

d

GFP–myosin IIIa + espin 1

µm43210

0.5

1.0rpiGFP

Actinc

GFPActin

bGFP–espin 1

GFPActin

aGFP–myosin IIIa∆K + espin 1GFP–myosin IIIa∆K

Figure 5 Myosin IIIa and espin 1 synergistically elongate filopodia in COS-7 cells through espin 1 WH2 activity. (a–c) Overexpression of either GFP–myosin IIIaΔK (a) or GFP–espin 1 (b) resulted in formation of short filopodia (mean lengths = 1.7 ± 0.83 μm and 1.3 ± 0.28 μm, respectively). In contrast, co-expression of GFP–myosin IIIaΔK (green) and espin 1 (c) had a synergistic effect that generated extremely long filopodia (14.3 ± 9.1 μm). F-actin (red) was visualized using Alexa 568-phalloidin. Graph (inset) of the relative pixel intensity (rpi) of the GFP–myosin IIIaΔK distribution in the single filopodium indicated by the rectangle in c shows the characteristic tip-to-base decaying gradient. (d) Co-expression of full-length GFP–myosin IIIa and espin 1 produced a more limited tip localization of these proteins and elongation of filopodia (3.7 ± 3.2 μm) when compared with co-expression of GFP–myosin IIIaΔK and espin 1. (e) The enhanced elongation phenotype was restored to a limited extent (5.93 ± 3.10 μm) when COS-7 cells were co-transfected

instead with GFP–myosin IIIaK50R and espin 1. (f) The enhanced elongation phenotype was similar to c when the cell was co-transfected with GFP–myosin IIIaΔK,33,34 and espin 1 (10.02 ± 4.7 μm). (g) Co-expression of GFP–myoIIIaΔK and espin 1 lacking ARD (espin 1ΔARD, labelled with the pan espin antibody, red) failed to elongate filopodia (2.05 ± 1.8 μm) and to show tip localization of espin 1ΔARD (inset). (h) Co-expression of GFP–myosin IIIaΔK and espin 1 with a mutated WH2 domain (espin 1mWH2, labelled with an anti-espin 1 antibody, red) failed to elongate filopodia (2.65 ± 1.5 μm) despite the fact that espin 1mWH2 localized to the tip and formed the tip-to-base gradient matching the distribution of the GFP–myoIIIaΔK (inset). Scale bars, 2.5 μm. Measurements of filopodia lengths for each of the combinations shown in the panels above are presented as box-plots; upper and lower whiskers represent the range, the top and bottom of the box represent the upper and lower 25th percentile, and the filled squares represent the mean values.



L E T T E R S

The steady-state distribution of myosin IIIa in a tip-to-base gradient is probably dynamically maintained. The length of myosin IIIa distri-bution should be inversely proportional to the net velocity of myosin towards the tip23, which will be slower for faster treadmilling actin cores (that is, in longer stereocilia9 and filopodia24). This prediction is also consistent with our observation that wild-type myosin IIIa, which has relatively low activity, has decreased tip localization in the filopodia, compared with the more active kinase mutant forms of myosin IIIa used in our experiments (Fig. 5). However, in stereocilia where the actin treadmilling is much slower, the wild-type myosin IIIa self-localizes effectively to the tip7 (Fig. 2). Similarly, the observed steady-state tip-to-base gradient distribution of espin 1 is not compatible with a model where espin 1 passively diffuses and binds to myosin IIIa resident at the tip, as this would result in a homogenous distribution along the entire length of the stereocilia with no detectable concentration gradient at steady-state. The gradient distribution of espin 1 at steady-state is remi-niscent of a myosin VI-driven gradient for the stereocilia membrane protein PTPRQ, and is best explained by a model that includes bind-ing, directed transport and diffusion of myosins and their cargo25. A more detailed consideration of this dynamic process that also accounts for actin treadmilling and plus-end directed motors predicts a similar distribution, which can be several microns long for longer stereocilia23. Thus, we favour a model where myosin IIIa–espin 1 complexes are dynamically associated with the treadmilling actin core. This model suggests that espin 1 is transported to the tips of stereocilia by myosin IIIa, where it remains bound to the surface of the actin core for a period of time. Interestingly, abolishing or reducing myosin IIIa kinase activity enhances the affinity of myosin IIIa for actin, providing further evi-dence that the kinase domain has a role in regulating myosin IIIa motor kinetics and actin-binding properties26,27. While the myosin IIIa–espin 1 complex is tightly bound to actin, it travels back towards the base of the stereocilia along with the treadmilling actin core. In support of this model, live video imaging in transfected COS-7 cells shows fluores-cent puncta of GFP–myosin IIIaΔK and mCherry–ARD (Supplementary Information, Movie 4) that move rearwards at rates matching the rates reported for actin treadmilling in filopodia (~0.5 μm min–1)24. We sug-gest that these puncta are stably bound to the surface of the treadmilling actin filament bundle.

Notably, the stereocilia tips are also the site of mechanoelectrical trans-duction (MET)28, that the myosin IIIa developmental expression level is correlated with maturation of MET in stereocilia17, and that myosin IIIa has been shown to transport components of the photoreceptor trans-duction machinery in Drosophila29,30. We cannot exclude the possibility that the localization and dynamics of the myosin IIIa–espin 1 complex are also affected by interactions with other proteins at the stereocilia tip. Furthermore, ankyrin repeats have been shown to be promiscuous bind-ers of membrane proteins31. It is possible that the turnover and dynamic localization of the espin 1–myosin IIIa complex are influenced by interac-tions with components of the MET machinery, and vice-versa.

METHODSAntibodies. Affinity-purified polyclonal antibodies (PB538 and PB539) were developed in rabbits immunized with a synthetic peptide (Princeton Biomolecules) corresponding to the amino acid sequence (LDALPVHHAARSGKLHCLR) of the first ankyrin repeat of mouse espin 1. A similarly raised antibody specific for a region conserved in all isoforms of espin (pan-espin, PB127; ref. 8) and anti-myosin IIIa (PB638; ref. 7) antibodies have been previously described.

Immuofluorescence and microscopy. After CO2 anaesthesia, rats, mice and guinea pigs were euthanized in accordance with National Institutes of Health (NIH) guidelines, and their temporal bones fixed by immersion in 4% paraformaldehyde in phosphate buffered saline (PBS; pH 7.4) for 2 h at room temperature. Sensory tissue was dissected in PBS, permeabilized with 0.5% Triton X-100 for 30 min and blocked overnight at 4 °C with 4% bovine serum albumin in PBS. Tissues were then incubated with primary antibody for 2 h, rinsed with PBS, stained with Alexa Fluor 488-conjugated secondary antibody (Molecular Probes) for 1 h, counterstained with Alexa Fluor 568 phalloidin (0.001 U μl–1; Molecular Probes) and mounted using Prolong Antifade (Molecular Probes). Fluorescence confocal images were obtained with a Nikon microscope equipped with a ×100 1.45 numerical aperture (NA) objective and a spinning disk confocal unit (PerkinElmer).

Electron microscopy. Rat organ of Corti or vestibular tissues were either rap-idly frozen by contact with a liquid nitrogen cooled metal block in a LifeCell (The Woodlands) freezing apparatus or fixed, glycerinated and plunge-frozen in Freon 22 cooled in liquid nitrogen before freeze-substitution in 1.5% uranyl acetate in absolute methanol at –90 °C. Freeze-substituted tissues were infiltrated with Lowicryl HM-20 resin (Electron Microscopy Sciences) at –45 °C, polymer-ized with UV light, thin-sectioned and labelled with immunogold. Samples were viewed and photographed with a Zeiss 922 electron microscope. As control for the immunogold labelling, we also used the antibody PB288, which is unrelated to espin or to myosin IIIa (Supplementary Information, Fig. 1d).

Expression plasmids. Espin 1 (NCBI accession number NM_031475) in pSPORT1 vector was obtained from imaGenes and PCR-cloned into pEGFP-C2 (Clontech) and pcDNA3.1(–) (Invitrogen) through EcoRI and KpnI sites. The site-directed Leu to Ala mutations in the WH2 motif of espin 1 were generated using a GeneTailor site-directed mutagenesis kit (Invitrogen). The ARD between amino acid positions 16 and 363 was PCR-amplified using the mouse espin 1 template (NM_207687) and subcloned in-frame into the mCherry-C1 (Clontech) expression vector through XhoI and EcoRI sites, and pDEST 15 GST expression vector (Invitrogen) using the Gateway LR Clonase cloning method (Invitrogen). The GFP-tagged expression plasmids used were espin 1ΔARD (a gift from James Bartles, Northwestern University, Chicago IL), myosin X (a gift from Richard Cheney, UNC, Chapel Hill NC), myosin XVa (a gift from Thomas Friedman, NIDCD/NIH), as well as full-length and deletion constructs of myosin IIIa that were generated in our laboratories. Myosin IIIa 2IQΔK constructs were generated as described previously26,34. Myosin IIIa 2IQK50R and myosin IIIaK50R constructs were generated by performing site-directed mutagenesis on the myosin IIIa 2IQ and myosin IIIa full-length constructs, respectively. All expression plasmids were sequence verified. Further details about the clones used are available in Supplementary Information, Table S1 and Fig. S3.

Cultures and transfection of COS‑7 cells. COS-7 (ATCC) cells were plated on coverslips and maintained at 37 °C in DMEM with 10% FBS. Cultures were transfected using GeneJuice transfect reagent (Novagen) and incubated for 24 h. Time-lapse videos of live cells were acquired at the maximum nominal laser power and camera gain allowed by the confocal microscope. Samples were also fixed for 20 min in 4% paraformaldehyde in PBS, permeabilized for 30 min in 0.5% Triton X-100 in PBS, and counterstained or processed for immunofluorescence as described above.

Culture and transfection of rat inner ear tissue. Organ of Corti and vestibular tissues were dissected from postnatal day 0–4 rats and attached to coverslips previ-ously coated with Cell-Tak (150 μg μl–1; BD Biosciences). Cultures were maintained in DMEM/F12 (Invitrogen) with 5–7% fetal bovine serum (FBS) and ampicillin (1.5 μg ml–1; Sigma) and kept at 37 °C and 5% CO2. For transfections, 50 μg of DNA was precipitated onto 25 mg of 1 μm gold particles and loaded into the Helios Gene Gun cartridges (BioRad). Tissue explants were transfected with the gene gun set at 95 psi of helium and maintained in culture for 18–48 h. Samples were fixed and counterstained for confocal microscope viewing as described above. The efficiency of transfection ranged from 0–9 hair cells per explant.

Image analysis. Image analysis was performed with ImageJ software (NIH). To esti-mate the relative increase in stereocilia length, we compared the heights of the tallest row of well-preserved stereocilia of cochlear and vestibular hair cells transfected



L E T T E R S

(Ht) with the average height of all their respective neighbouring (usually between 3–5) non-transfected cells (Hnt) within the field of view of our camera/confocal setup (30 × 45 μm). The average ratio of stereocilia length was calculated as 1 = HT

HNT.

ANOVA was performed using MATLAB (Mathworks). Cross-correlation analysis for the intensity plot in Fig. 3 was performed using Microsoft Excel.

Western blotting. 100-mm dishes of transfected semi-confluent COS-7 cells were rinsed in PBS and scraped in 160 μl of lysis buffer: PBS, 1% Triton-X, 5 mM DTT, 1 mM Pefabloc, 5 μg ml–1 pepstatin A, 5 μg ml–1 leupeptin, 2 mM EDTA, 0.2 mM PMSF and 1% mammalian protease inhibitor cocktail (Sigma). After addition of 1× loading sample buffer and dithiothreitol (Invitrogen), samples were boiled and 10 μl of lysates were loaded in 4–12% Bis-Tris minigel (Invitrogen). Western blots were incubated overnight at 4 °C with 4 μg ml–1 of the primary antibody. Horseradish peroxidase-conjugated goat anti-rabbit antibodies (Santa Cruz) and ECL chemilu-minescence system (Amersham Biosciences) were used for detection.

GST pulldown assays. Protein expressions of glutathione S-transferase (GST) alone or fused to ARD (GST–ARD) were optimized under l-arabinose induc-tion in BL21-AI bacteria (Invitrogen). GST proteins were purified from bacterial extracts using glutathione–Sepharose 4B beads according to the manufacturer’s instructions (Amersham Biosciences). GFP–myosin IIIaΔK, – pre3THDI, –3THDI and –post3THDI proteins were extracted from 24-h COS-7 transfectants by brief sonication and 20-min ultracentrifugation at 145,000g in ice-cold lysis buffer (1% Triton X-100, 5 mM DTT, 150 mM NaCl, 50 mM Tris pH 7.4, 2 mM EDTA, 3 mM Pefabloc SC, 1× Pefabloc (Roche) and 1× mammalian protease inhibitor cocktail (Sigma)). To test for myosin IIIa interactions, the same amount of GST–ARD or GST alone was bound to 4B beads for 1 h at 4 °C followed by incubation with the GFP-tagged myosin IIIa fragment in CLB for 2 h. The beads were then washed four times with lysis buffer. Bound proteins were separated by electrophoresis on NuPAGE Bis-Tris 4-12% gels (Invitrogen) and analysed by western blotting using rabbit polyclonal anti-GFP and anti-GST antibodies (Invitrogen).

ATPase assays. The steady-state actin-activated ATPase activity of baculovirus-expressed myosin IIIa 2IQK50R and wild-type were assessed in an NADH-coupled assay26,32. ATPase activity of fully phosphorylated myosin IIIa was compared with unphosphorylated myosin IIIa after a 60-min incubation at room temperature in the presence and absence of 200 μM ATP, respectively. The kinase activity of the myosin IIIa 2IQ constructs was assayed using 32P-ATP or western blotting with anti-phosphothreonine antibodies26,32.


ACKnoWlEDGEMEnTSWe thank Chi W. Pak for discussions and for the suggestion of mutations in the WH2 motif, Mark Schneider and Saeeda Latham for initial help with experiments and for discussions related to this work, Martin Horak for advice on cloning procedures, and Ronald Petralia for comments on the manuscript. This work was supported by NIDCD, DIR, NIH and in part by NIH grants to A.C.D. (no. EY003575) and to C.M.Y. (no. EY016419).

AUThoR ConTRiBUTionSR.C.M. designed probes and experiments, performed the GST pulldowns, cell culture, immunocytochemistry and transfections; F.T.S. performed the dissections, cell and organotypic cultures, immunohistochemistry, transfections and confocal imaging; G.W.D. and A.C.D. designed probes, performed transfections and contributed to the experimental design; U.M. performed image and statistical analyses; A.D.S. characterized antibody and DNA probes; C.M.Y. and J.E.M. generated myosin IIIa kinase dead cDNA, purified and performed kinase and motor activity assays; B.K. performed electron microscopy, designed experiments and analysed the results together with all the other authors. All authors discussed and helped to prepare the manuscript.

CoMpETinG FinAnCiAl inTERESTSThe authors declare no competing financial interests.


1. Walsh, T. et al. From flies’ eyes to our ears: mutations in a human class III myosin cause progressive non-syndromic hearing loss DFNB30. Proc. Natl Acad. Sci. USA 99, 7518–7523 (2002).

2. Komaba, S., Inoue, A., Maruta, S., Hosoya, H. & Ikebe, M. Determination of human myosin III as a motor protein having a protein kinase activity. J. Biol. Chem. 278, 21352–21360 (2003).

3. Zheng, L. et al. The deaf jerker mouse has a mutation in the gene encoding the espin actin-bundling proteins of hair cell stereocilia and lacks espins. Cell 102, 377–385 (2000).

4. Donaudy, F. et al. Espin gene (ESPN) mutations associated with autosomal dominant hearing loss cause defects in microvillar elongation or organisation. J. Med. Genet. 43, 157–161 (2006).

5. Naz, S. et al. Mutations of ESPN cause autosomal recessive deafness and vestibular dysfunction. J. Med. Genet. 41, 591–595 (2004).

6. Sekerkova, G., Zheng, L., Loomis, P. A., Mugnaini, E. & Bartles, J. R. Espins and the actin cytoskeleton of hair cell stereocilia and sensory cell microvilli. Cell. Mol. Life Sci. 63, 2329–2341 (2006).

7. Schneider, M. E. et al. A new compartment at stereocilia tips defined by spatial and temporal patterns of myosin IIIa expression. J. Neurosci. 26, 10243–10252 (2006).

8. Rzadzinska, A., Schneider, M., Noben-Trauth, K., Bartles, J. R. & Kachar, B. Balanced levels of Espin are critical for stereociliary growth and length maintenance. Cell. Motil. Cytoskeleton 62, 157–165 (2005).

9. Rzadzinska, A. K., Schneider, M. E., Davies, C., Riordan, G. P. & Kachar, B. An actin molecular treadmill and myosins maintain stereocilia functional architecture and self-renewal. J. Cell Biol. 164, 887–897 (2004).

10. Lin, H. W., Schneider, M. E. & Kachar, B. When size matters: the dynamic regulation of stereocilia lengths. Curr. Opin. Cell Biol. 17, 55–61 (2005).

11. Schneider, M. E., Belyantseva, I. A., Azevedo, R. B. & Kachar, B. Rapid renewal of auditory hair bundles. Nature 418, 837–838 (2002).

12. Manor, U. & Kachar, B. Dynamic length regulation of sensory stereocilia. Semin. Cell Dev. Biol. 19, 502–510 (2008).

13. Belyantseva, I. A. et al. Myosin-XVa is required for tip localization of whirlin and dif-ferential elongation of hair-cell stereocilia. Nature Cell Biol. 7, 148–156 (2005).

14. Prosser, H. M., Rzadzinska, A. K., Steel, K. P. & Bradley, A. Mosaic complementation demonstrates a regulatory role for myosin VIIa in actin dynamics of stereocilia. Mol. Cell Biol. 28, 1702–1712 (2008).

15. Tokuo, H., Mabuchi, K. & Ikebe, M. The motor activity of myosin-X promotes actin fiber convergence at the cell periphery to initiate filopodia formation. J. Cell Biol. 179, 229–238 (2007).

16. Tokuo, H. & Ikebe, M. Myosin X transports Mena/VASP to the tip of filopodia. Biochem. Biophys. Res. Commun. 319, 214–220 (2004).

17. Waguespack, J., Salles, F. T., Kachar, B. & Ricci, A. J. Stepwise morphological and functional maturation of mechanotransduction in rat outer hair cells. J. Neurosci. 27, 13890–13902 (2007).

18. Les Erickson, F., Corsa, A. C., Dose, A. C. & Burnside, B. Localization of a class III myosin to filopodia tips in transfected HeLa cells requires an actin-binding site in its tail domain. Mol. Biol. Cell 14, 4173–4180 (2003).

19. Bohil, A. B., Robertson, B. W. & Cheney, R. E. Myosin-X is a molecular motor that func-tions in filopodia formation. Proc. Natl Acad. Sci. USA 103, 12411–12416 (2006).

20. Dose, A. C. et al. Myo3A, one of two class III myosin genes expressed in vertebrate retina, is localized to the calycal processes of rod and cone photoreceptors and is expressed in the sacculus. Mol. Biol. Cell 14, 1058–1073 (2003).

21. Quinlan, M. E., Heuser, J. E., Kerkhoff, E. & Mullins, R. D. Drosophila Spire is an actin nucleation factor. Nature 433, 382–388 (2005).

22. Loomis, P. A. et al. Targeted wild-type and jerker espins reveal a novel, WH2-domain-dependent way to make actin bundles in cells. J. Cell Sci. 119, 1655–1665 (2006).

23. Naoz, M., Manor, U., Sakaguchi, H., Kachar, B. & Gov, N. Protein localization by actin treadmilling and molecular motors regulates stereocilia shape and treadmilling rate. Biophys. J. (2008).

24. Mallavarapu, A. & Mitchison, T. Regulated actin cytoskeleton assembly at filopodium tips controls their extension and retraction. J. Cell Biol. 146, 1097–1106 (1999).

25. Sakaguchi, H. et al. Dynamic compartmentalization of protein tyrosine phosphatase receptor Q. at the proximal end of stereocilia: Implication of myosin VI-based transport. Cell. Motil. Cytoskeleton (2008).

26. Dose, A. C. et al. The kinase domain alters the kinetic properties of the myosin IIIA motor. Biochemistry 47, 2485–2496 (2008).

27. Kambara, T., Komaba, S. & Ikebe, M. Human myosin III is a motor having an extremely high affinity for actin. J. Biol. Chem. 281, 37291–37301 (2006).

28. Ricci, A. J., Kachar, B., Gale, J. & Van Netten, S. M. Mechano-electrical transduction: new insights into old ideas. J. Membr. Biol. 209, 71–88 (2006).

29. Porter, J. A., Yu, M., Doberstein, S. K., Pollard, T. D. & Montell, C. Dependence of calmodulin localization in the retina on the NINAC unconventional myosin. Science 262, 1038–1042 (1993).

30. Wes, P. D. et al. Termination of phototransduction requires binding of the NINAC myosin III and the PDZ protein INAD. Nature Neurosci. 2, 447–453 (1999).

31. Mosavi, L. K., Cammett, T. J., Desrosiers, D. C. & Peng, Z. Y. The ankyrin repeat as molecular architecture for protein recognition. Protein Sci. 13, 1435–1448 (2004).

32. Dose, A. C., Ananthanarayanan, S., Moore, J. E., Burnside, B. & Yengo, C. M. Kinetic mechanism of human myosin IIIA. J. Biol. Chem. 282, 216–231 (2007).






l e t t e r s

p53-cofactor JMY is a multifunctional actin nucleation factorJ. Bradley Zuchero1, Amanda S. Coutts2, Margot E. Quinlan3, Nicholas B. La Thangue2 and R. Dyche Mullins1,4

Many cellular structures are assembled from networks of actin filaments, and the architecture of these networks depends on the mechanism by which the filaments are formed. Several classes of proteins are known to assemble new filaments, including the Arp2/3 complex, which creates branched filament networks, and Spire, which creates unbranched filaments1,2. We find that JMY, a vertebrate protein first identified as a transcriptional co-activator of p53, combines these two nucleating activities by both activating Arp2/3 and assembling filaments directly using a Spire-like mechanism. Increased levels of JMY expression enhance motility, whereas loss of JMY slows cell migration. When slowly migrating HL-60 cells are differentiated into highly motile neutrophil-like cells, JMY moves from the nucleus to the cytoplasm and is concentrated at the leading edge. Thus, JMY represents a new class of multifunctional actin assembly factor whose activity is regulated, at least in part, by sequestration in the nucleus.

By searching genome databases for sequences related to the Wiskott–Aldrich syndrome protein (WASp) Homology 2 (WH2) domain we discovered a potential Arp2/3-activating sequence, WWWCA, in the vertebrate protein JMY (Fig. 1a). This sequence is composed of three tandem repeats of the actin-monomer-binding WH2 domain (WWW), a central domain (C) that binds actin and Arp2/3, and an Arp2/3-binding acidic domain (A). These sequence elements, first identified in WASp-family proteins1,3, collaborate in activating Arp2/3. The identification of these elements in JMY was surprising, because JMY localizes primarily to the nucleus and was originally discovered as a binding partner of p300, a co-activator for many transcription factors, including the tumour suppressor p53 (ref. 4). In fibroblasts JMY accumulates in the nucleus in response to DNA damage, where it enhances the p53-dependent tran-scription of pro-apoptotic genes4,5.

To determine whether JMY also has a function in the assembly of the actin cytoskeleton, we tested the effect of JMY expression on actin organi-zation in vivo. Overexpression of JMY in human U2OS cells induces the formation of elongated actin filament structures that co-localize with

JMY (Fig. 1b), in a similar manner to overexpression of WASp-family proteins and the actin nucleation factor Spire2,6. Truncation mutants dem-onstrate that the WH2 cluster is required for this effect but, curiously, the Arp2/3-binding CA domain is not (Supplementary Information, Fig. S1). Expression of the carboxy-terminal region of JMY fused to green fluo-rescent protein (GFP–PWWWCA) produces a linear, dose-dependent increase in the cellular concentration of filamentous actin (Fig. 1c) as judged by correlating Alexa Fluor 568-phalloidin staining with GFP fluorescence (n = 457; Fig. 1d). In contrast, expressing GFP alone has no significant effect on cellular levels of filamentous actin (n = 458; Fig. 1d; Supplementary Information, Fig. S1).

To study how JMY affects actin assembly in vitro, we first verified that all three putative WH2 domains (Wa, Wb and Wc, from amino ter-minus to carboxy terminus) bind monomeric actin (Supplementary Information, Fig. S2). JMY activates Arp2/3 in vitro, as determined by pyrene-actin polymerization assays using a C-terminal fragment of JMY (WWWCA) (Fig. 1e; t1/2 (JMY + actin + Arp2/3) = 42.2 ± 2.3 s, versus t1/2 (actin + Arp2/3) = 1,193 ± 28 s (mean ± s.e.m.). JMY WWWCA also induces rapid actin polymerization in the absence of Arp2/3 (Fig. 1e; t1/2 = 188.5 ± 4.4 s). The JMY-dependent increase in polymerization rate is dose-dependent, both with and without Arp2/3 (Fig. 1f). This result is quite surprising and distinguishes JMY from all other proteins known to activate Arp2/3. N-WASp WWCA, for example, activates Arp2/3 to a similar extent as JMY but does not accelerate polymerization in the absence of Arp2/3 (Fig. 1e).

By itself JMY could accelerate actin assembly by nucleating new filaments, by increasing the rate of filament elongation, or by sever-ing existing filaments to create new barbed ends2,7. JMY does not affect the elongation of preformed filaments, arguing that it neither accelerates elongation nor severs, but rather nucleates new filaments (Supplementary Information, Fig. S3a–c). Indeed, we observe more than 12-fold more filaments in the presence of JMY WWWCA compared with actin alone (Fig. 2a, b). In the presence of Arp2/3, JMY WWWCA produces branched filaments, which is consistent with nucleation by Arp2/3 (Fig. 2c, left panel), whereas, by itself, JMY nucleates unbranched filaments (Fig. 2a). In assembly reactions, filament length is inversely

1Cellular and Molecular Pharmacology, University of California, San Francisco, California 94158, USA. 2Laboratory of Cancer Biology, Division of Medical Sciences, University of Oxford, Oxford OX3 9DU, UK. 3Chemistry and Biochemistry, University of California, Los Angeles, California 90095, USA.4Correspondence should be addressed to R.D.M. (e-mail: [email protected])

Received 3 December 2008; accepted 29 January 2009; published online 15 March 2009; DOI:10.1038/ncb1852




l e t t e r s

aWa Wb Wc C AProCoiled-coilNT

b

WH2 Central Acidic

Merge HA–JMY (green)Phalloidin (red) DAPI (blue)

Concentration JMY WWWCA(µM)

t 1/2 (

s) WWWCA

WWWCA+ Arp2/3

0

200

400

600

800

1,000

1,200

0.001 0.01 0.1 1

NW WWCA

WWWCA

WWWCA + Arp2/3

NW WWCA + Arp2/3Actin

Arp2/3

Time (s)

Fluo

resc

ence

(a.u

.)

–0.2

0.0

0.2

0.4

0.6

0.8

1.0

1.2

0 200 400 600 800 1,000

Arp2/3-interacting

c

d e f

Merge GFP–PWWWCA (green)Phalloidin (red) DAPI (blue)

GFP intensity (a.u.)

Ale

xa 5

68 in

tens

ity (a

.u.)

GFP–PWWWCA

20

40

60

80

0 20 40 60 80 100 120

GFP alone

Figure 1 JMY nucleates actin filaments and activates the Arp2/3 complex. (a) Domain structure of JMY. The C terminus of JMY is homologous to activators of Arp2/3. A poly-proline (P) domain28 is followed by three tandem actin-monomer-binding WH2 domains (Wa, Wb and Wc), a central domain (C) binding actin and Arp2/3, and an Arp2/3-binding acidic domain (A). Alignment shows individual WH2 domains of JMY, and compares the sequences of the WCA regions of JMY, Scar and N-WASp. JMY WCA is 28% identical to N-WASp WCA (ClustalW), and residues putatively involved in binding actin and Arp2/3 (refs 29, 30) are 100% conserved between all available JMY sequences. (b) Expression of haemagglutinin (HA)-tagged JMY (top panels; revealed by indirect immunofluorescence of HA, green) in U2OS cells induces the formation of filamentous actin structures (revealed with Alexa Fluor 568-phalloidin, red). These elongated actin structures co-localize with JMY and are not seen in untransfected cells (bottom panels). Nuclei were revealed with DAPI (blue). Scale bar, 10 µm. (c) Expression of GFP–PWWWCA (green) in U2OS cells increases cellular F-actin content (Alexa Fluor

568-phalloidin, red). (d) Quantification of the increase in F-actin induced by GFP–PWWWCA expression. Phalloidin intensity was plotted as a function of GFP–PWWWCA intensity and shows a linear increase in F-actin content with increased expression of GFP–PWWWCA (n = 457). In contrast, expressing GFP alone has only a minor effect on the red intensity detected, probably due to a small amount of bleed-through (n = 458). a.u., arbitrary units. Error bars indicate s.e.m. (e) Pyrene-actin polymerization assays show that JMY WWWCA both activates Arp2/3 and nucleates actin in the absence of Arp2/3 (red and blue traces). N-WASp (NW) WWCA activates Arp2/3 (green), but does not nucleate actin on its own (grey). (f) Intrinsic nucleation and activation of Arp2/3 by WWWCA are dose-dependent. Pyrene-actin polymerization assays were conducted in the absence (blue) or presence (red) of Arp2/3, with increasing concentrations of JMY WWWCA. Time to half-maximal polymerization (t1/2) was plotted as a function of WWWCA concentration. Pyrene-actin polymerization assays were performed in 1 × KMEI and contained 2 µM actin, 167 nM JMY or N-WASp, and 2.5 nM Arp2/3, where noted.



l e t t e r s

proportional to the rate of nucleation, and filaments made in the pres-ence of JMY are shorter than those made with actin alone. Arp2/3-nucleated filaments are even shorter (Fig. 2d). JMY produces twice as many branches per micrometre as Scar1/WAVE1 (Fig. 2e), which is con-sistent with JMY inducing a faster rate of Arp2/3 activation than Scar.

Some proteins that cap barbed ends in vivo (for example capping pro-tein) nucleate filaments that elongate from their pointed end in vitro8. To determine whether JMY caps barbed ends, we made filaments and depolymerized them in the presence of JMY WWCA. JMY does not inhibit disassembly, arguing that it does not cap barbed ends but instead promotes polymerization by nucleating new filaments that elongate from their barbed ends (Supplementary Information, Fig. S3d, e). Direct nucleation seems to explain the unexpected ability of JMY to induce actin assembly in vivo in the absence of the Arp2/3-binding CA domain (Supplementary Information, Fig. S1).

We proposed that, in a similar manner to Spire2, JMY uses its tandem WH2 domains to nucleate new filaments. We tested the activity of con-structs consisting of either all three WH2 domains (JMY WWW) or the two C-terminal WH2 domains (JMY WbWc). JMY WWW nucleates actin as well as WWWCA, indicating that the CA region has no func-tion in direct nucleation (Fig. 2f). Moreover, addition of Arp2/3 to JMY WWW does not further increase the rate of polymerization (Fig. 2f). As with Spire, the two C-terminal WH2 domains (JMY WbWc) are sufficient for nucleation, but all three JMY WH2 domains are required for maximal activity (Supplementary Information, Fig. S3h, i).

In contrast, a fragment composed of a single WH2 domain (Wc) and the CA domains (JMY WCA) activates Arp2/3 but does not nucleate filaments on its own (Fig. 2g). JMY contains a conserved tryptophan residue known to be important for Arp2/3 binding in all WASp-family proteins9. Replacing this tryptophan residue with alanine in JMY decreases the activation of Arp2/3 without affecting intrinsic nucleation activity (Supplementary Information, Fig. S3j). These results suggest that nucleation and Arp2/3 activation are sepa-rable activities, and that JMY activates Arp2/3 by the same mechanism as WASp-family proteins.

Similarities in the sequences and activities of JMY and Spire suggested that the two proteins nucleate filaments by a common mechanism. We compared the kinetics of nucleation by JMY and Spire over a range of actin concentrations. At all concentrations of actin tested, JMY and Spire had nearly identical kinetics (Supplementary Information, Fig. S3k, l). In addition, similarly to Spire, high concentrations of JMY sequester actin monomers (Supplementary Information, Fig. S3m, n). Unlike Spire, however, JMY WWCA does not prevent the dissociation of mono-mers from the pointed end of filaments (Supplementary Information, Fig. S3f, g). It is possible that the full-length JMY interacts with filament ends, but this is not required for nucleation.

Spire nucleates actin by stitching monomers together with tandem WH2 domains and a novel actin-binding motif (previously called Linker 3; ref. 2) that we designate the monomer binding linker (MBL). Spire–MBL, a short (about 15-residue) sequence connecting the third and fourth WH2 domains, is sufficient to promote weak nucleation2. We compared the sequences of Spire–MBL with the region between the two C-terminal WH2 domains of JMY (Wb and Wc) and N-WASp. Most residues conserved between Spire homologues are also conserved in JMY, but not in N-WASp (Fig. 3a). Another WH2-containing nucleation factor, Cordon Bleu (Cobl), is thought to operate by a different mechanism10;

consistent with this was the fact that we saw no conservation between the linker regions of Cobl and Spire (Fig. 3a).

Do JMY and Spire nucleate actin by the same mechanism? We replaced the JMY linker with a set of glycine-serine repeats (JMY WbWc(gs5)) and compared it with a similar Spire mutant (Spire CD(gs5)). JMY WbWc(gs5) has a modest defect in nucleation, whereas replacing the linker in Spire has a more pronounced effect (Fig. 3b, d; Supplementary Information, Fig. S3o). Interestingly, inserting either the JMY–MBL or Spire–MBL between the N-WASp WH2 domains (NW WJW or WSW) converted N-WASp into a nucleator (Fig. 3c, d; Supplementary Information, Fig. S3p), whereas inserting a flexible linker (NW WW(gs5)) did not (Fig. 3b, d). NW WJW and WSW do not nucleate as well as JMY or Spire. Thus, nucleation by JMY and Spire requires unique properties of both the linker (MBL) and the WH2 domains (Fig. 3e).

To investigate its role in actin assembly in vivo, we determined the localization of JMY in multiple cell types. JMY is primarily nuclear in mouse embryonic fibroblasts, B16-F10 mouse melanoma cells, NIH 3T3 cells and primary rat neurons (Supplementary Information, Fig. S4, and data not shown). In ruffling B16-F10 cells, we observed a small fraction of JMY co-localized with actin filaments at the leading edge (Supplementary Information, Fig. S4b). In highly motile, primary human neutrophils, JMY was almost entirely excluded from the nucleus and co-localized with filaments at the leading edge (Fig. 4d). JMY does not bind actin filaments in vitro, so this localization does not simply reflect the direct interaction of JMY with filaments (Supplementary Information, Fig. S4d).

Our localization studies suggest that the presence of JMY at the lead-ing edge correlates with motility. To test this further, we investigated JMY expression and localization in HL-60 cells, which can exist as non-motile, relatively undifferentiated cells or be induced to differentiate into highly motile cells11. In undifferentiated HL-60 cells, JMY is primarily nuclear and does not co-localize with filaments (primarily nuclear in 91% of cells, n = 246; Fig. 4a; Supplementary Information, Fig. 5). The addition of 1.3% dimethylsulphoxide (DMSO) to the culture medium induces differentiation into highly motile cells that polarize and undergo chemotaxis. During differentiation, JMY localization shifts markedly, becoming almost entirely cytoplasmic. This shift occurs about two days before cells are competent to polarize in response to formyl pep-tide chemoattractant (fMLP; Supplementary Information, Fig. S5c). In differentiated cells, JMY co-localizes with filamentous actin in the cell cortex (cytoplasmic in 94%, nuclear in 6%, n = 212; Fig. 4b), and on addition of fMLP, JMY localizes strongly to the leading edge, where it overlaps with filamentous actin (88% of polarized cells, n = 311; Fig. 4c; Supplementary Information, Fig. S5d). A comparison of JMY staining with the location of soluble GFP in polarized cells shows that enrich-ment of JMY at the leading edge is not a volume artefact (Supplementary Information, Fig. S5e–g). These data suggest that translocation of JMY to the cytosol has a function in building the leading edge. Interestingly, expression of JMY’s nuclear binding partner p300 disappears when HL-60 cells are differentiated, suggesting that the nuclear (p300/p53-dependent transcription) and cytoplasmic (actin nucleation) roles of JMY are regulated by separate pathways (Fig. 4e).

To examine the role of JMY in cell motility, we stably expressed GFP–JMY and GFP–JMY∆CA in U2OS cells, grew monolayers, scratched them, and monitored wound healing over time12. Cells expressing GFP–JMY migrated 17% faster than wild-type cells (n = 3;



l e t t e r s

f g

a

JMY WCA + Arp2/3

JMY WCA

Actin

Scar WCA + Arp2/3

Time (s)

Fluo

resc

ence

(a.u

.)

0 200 400 600 800 1,000–0.2

0

0.2

0.4

0.6

0.8

1.0

1.2

–0.2

0.2

0.4

0.6

0.8

1.0

1.2

0 200 400 600 800 1,000

JMY WWW

JMY WWW + Arp2/3

Actin

Time (s)

Fluo

resc

ence

(a.u

.) JMY WWWCA

NW WW

JMY (6’)

JMY + Arp2/3 Scar + Arp2/3 Actin alone

Actin alone+JMY WWWCA

At

t9/10

of J

MY

At

t9/10

of r

eact

ion

b

c

d e

JMY

Actin

Ave

rage

num

ber

of

fila

men

ts p

er fi

eld

0

10

20

30

40

50

0

4

8

12

Ave

rage

fila

men

t le

ngth

(µm

)

0

0.02

0.04

0.06

JMY

Actin

JMY

+R

Scar

+R

JMY

Actin

JMY

+R

Scar

+R

Bra

nchi

ng d

ensi

ty(b

ranc

hes/

µm)

Figure 2 Mechanistic dissection of JMY. (a) JMY nucleates unbranched filaments and increases the number of filaments over actin alone. Filaments made in the presence (left) or absence (right) of 167 nM JMY WWWCA were fixed with Alexa Fluor 488 phalloidin and Latrunculin B at 6 min (t9/10 of JMY WWWCA reaction) before dilution and spotting onto poly-(l-lysine) coverslips2,24. Scale bars, 5 µm. (b) Quantification of filaments per field in images from a demonstrates that JMY nucleates new filaments (JMY, 42.7 ± 4.2 filaments per micrometre, n = 35 fields; actin alone, 3.2 ± 0.5 filaments per micrometre, n = 30 fields). Error bars indicate s.e.m. (c) Filaments prepared as in a, in the presence of JMY plus Arp2/3 (left), Scar plus Arp2/3 (centre) or actin alone (right). The concentration of filaments was kept constant by arresting reactions at their individual t9/10 values (see Methods). Filaments nucleated in the presence of JMY and Arp2/3 are branched, which is consistent with JMY activating Arp2/3, as are filaments made in the presence of Scar

and Arp2/3. The shorter, more abundant filaments seen here are due to Arp2/3 nucleating actin more rapidly than intrinsic nucleation by JMY. (d) Quantification of filament length at t9/10. The rate of nucleation is inversely proportional to the length of filament (rate: JMY+Arp2/3 > JMY > Scar+Arp2/3 >> actin alone). Error bars indicate s.e.m. (e) Quantification of filament branching in each condition. n > 300 filaments per condition. (f) Tandem WH2 domains from JMY are sufficient for actin nucleation. Actin polymerization is as fast with WWW as it is with WWWCA. WWW lacks the Arp2/3-binding CA domain, so adding Arp2/3 to WWW does not accelerate polymerization over WWW alone. N-WASp WW does not nucleate actin. (g) JMY WCA is sufficient to activate Arp2/3 but does not nucleate actin. The rate of actin polymerization in the presence of JMY WCA and Arp2/3 is similar to reactions containing Scar WCA and Arp2/3. In the absence of Arp2/3, JMY WCA has no effect on actin polymerization. Experimental conditions were as in Fig. 1.



l e t t e r s

P < 0.003; Fig. 5a, b; Supplementary Information, Fig. S6). In contrast, cells expressing GFP–JMY∆CA migrated at the same rate as wild-type cells (GFP–JMY, 43.8 ± 1.6 µm h−1; GFP–JMY∆CA, 38.5 ± 1.7 µm h−1; wild-type, 37.5 ± 1.0 µm h−1), suggesting that JMY requires its Arp2/3-activating activity to enhance motility.

We next knocked down JMY expression in cultured U2OS and HEK 293 cells by RNA-mediated interference (RNAi). Western blotting shows that RNAi was efficient (Fig. 5e; Supplementary Information, Fig. S6). In both cell lines, knockdown of JMY expression signifi-cantly slows the rate of wound healing (Fig. 5c, d; Supplementary Information, Fig. S6). In the first 6 h, cells treated with JMY small interfering RNA (siRNA) moved 38.0% more slowly than control U2OS cells (n = 4; P < 0.05). By 24 h, control cells migrated fully into the wound, whereas cells treated with JMY siRNA migrated only 86% of the distance (Fig. 5a, b; n = 4). Control HEK 293 cells migrated 92%

of the distance by 24 h, in contrast with 60% for cells treated with JMY siRNA (Supplementary Information, Fig. S6; n = 7).

Defects in wound healing could be caused by (1) reduced actin assem-bly, (2) altered transcription, (3) reduced cell division or increased apoptosis, or (4) off-target effects of the siRNA. Consistent with the first explanation was our observation that that JMY knockdown cells contained 11.4 ± 0.4% (n = 396, P < 0.03) less filamentous actin than wild-type cells (Fig. 5f). Loss of either p53 or p300, both of which are required for all JMY’s known transcriptional effects, leads to increased cell migration13–15, which is inconsistent with transcriptional effects on motility. In addition, less than 5% of cells close to the wound divide in the first 6 h of the assay, making it unlikely that cell division contrib-utes to the observed defect. Knockdown of JMY was shown to decrease apoptosis5, ruling out the third explanation. To address possible off-target effects, we used three different siRNAs, from different regions of

a

b dc

e

N-WASp Poly Pro WH2WH2 C A

Spire WH2WH2WH2WH2

JMY Poly Pro WH2cWH2b C A

Nucleates actinActivates

the Arp2/3 complex

WH2a

Time (s)

Fluo

resc

ence

(a.u

.)

–0.2

0.0

0.2

0.4

0.6

0.8

1.0

1.2

0 200 400 600 800 1,000

Spire CD

Spire CD(gs5)

JMY WbWc

WbWc(gs5)

NW WW(gs5)Actin

–0.2

0.0

0.2

0.4

0.6

0.8

1.0

1.2

0 200 400 600 800 1,000

Time (s)

Fluo

resc

ence

(a.u

.)

NW WJWJMY WbWc

NW WSW

NW WWActin

t 1/2

(s)

Spire

CD

Spire

CD

(gs5

)

JMY

WbW

c

JMY

WbW

c(gs

5)

NW

WW

(gs5

)

Actin

NW

WSW

NW

WJW

0

200

400

600

800

NW

WW

Hs JMYCf JMYXt JMYDr JMY

Dm SpireMm Spire-1Rn Spire-1Hs Spire-1Xl Spir

Mm N-WASpMm WHAMMMm Cobl L1Mm Cobl L2

NW WJW

NW WSW

Mm JMY

Figure 3 JMY nucleates actin by the same mechanism as Spire. (a) The region between JMY Wb and Wc is homologous to the short actin nucleation motif from Spire (monomer-binding linker, MBL). The figure shows an alignment between Spire–MBL and the same region of JMY (mJMY residues 903–917) with homologous residues coloured grey. The MBL sequence is not homologous to the analogous position in N-WASp, WHAMM31 or Cordon Bleu10. The position of the glycine-serine repeats in b is underlined, and the sequences of N-WASp gain-of-function mutations (NW WJW and WSW) in c are shown. Organisms: Mm, Mus musculus; Hs, Homo sapiens; Cf, Canis familiaris; Xt, Xenopus tropicalis; Dr, Danio rerio; Dm, Drosophila melanogaster; Rn, Rattus norvegicus; Xl, Xenopus laevis. (b) JMY–MBL is important for actin nucleation. Replacing the MBL in JMY WbWc and Spire CD (the two C-terminal WH2 domains in Spire) with a flexible linker

of glycine-serine repeats (gs5) causes a nucleation defect in both JMY and Spire. The analogous N-WASp mutant does not promote nucleation, but instead inhibits spontaneous polymerization. (c) Gain of function. Replacing the linker region between the WH2 domains of N-WASp WW with JMY–MBL or Spire–MBL (NW WJW or WSW) converts N-WASp into a weak actin nucleator. This shows that JMY–MBL and Spire–MBL are sufficient for nucleation. Mutated residues of WJW and WSW are shown in a. Reactions in b and c contained 4 µM actin. Experimental conditions were as in Fig. 1. (d) t1/2 values of reactions from b and c. Reactions were repeated more than three times each. Error bars indicate s.e.m. (e) Actin nucleation and activation of Arp2/3 are distinct activities of JMY that overlap spatially. Tandem JMY WH2 domains and the MBL (star) nucleate actin, similarly to Spire, and JMY WCA activates Arp2/3, similarly to N-WASp and Scar.



l e t t e r s

the human JMY gene, to knock down expression in HEK 293 cells. All three caused migration defects (Supplementary Information, Fig. S6g), whereas a control siRNA had no effect.

We next tested whether JMY localizes to the leading edge of U2OS cells during migration in response to a wound, as seen in highly motile HL-60 cells. We fixed and stained U2OS cells 15 min after wounding, as well as cells in subconfluent and non-wounded cultures. Although most of JMY was nuclear in these cells, we also observed both endogenous JMY and GFP–JMY co-localized with filamentous actin at the leading edge (Fig. 5g; Supplementary Information, Fig. S4c), both in cells adja-cent to a wound and in ruffling edges of subconfluent cells (data not shown). Thus, even in cells in which JMY is predominantly nuclear, a fraction of the protein can influence actin assembly at the leading edge and promote migration.

JMY combines an unusual set of activities. The combination of tran-scriptional co-activation and actin nucleation activities suggests that JMY might mediate cellular decisions involving apoptosis and migra-tion. Alternatively, JMY might be a chimaera whose transcriptional and

cytoskeletal functions are more or less independent. More intriguing is JMY’s combination of Arp2/3-dependent and Arp2/3-independent nucleation activities. A cell switching from a resting to a motile state must remodel its cytoskeleton extensively16. The leading edge of most migrating cells is characterized by a highly branched dendritic arbour of actin filaments nucleated by Arp2/3, and one requirement for generating such a network is the pre-existence of ‘mother’ filaments for Arp2/3 to bind and branch from17,18. Although resting cells contain some actin fila-ments (such as cortical actin or stress fibres), recent work suggests that not all filaments can serve as efficient substrates for Arp2/3-dependent nucleation19,20. We propose that JMY contributes to cell motility by nucle-ating filaments to jump-start Arp2/3-dependent nucleation and branch-ing. By first nucleating new mother filaments and then activating Arp2/3 to branch off these filaments, JMY could promote the rapid formation of a branched actin network. It is also possible that JMY has evolved to promote actin polymerization in different cellular contexts, either in the cytoplasm in the presence of Arp2/3, or in the nucleus in the absence of Arp2/3 (see model in Fig. 5h). For example, it would be interesting

a

b

c

Merge Phalloidin (red) JMY (green) DAPI (blue)

d

Und

iffer

entia

ted

Diff

eren

tiate

dD

iffer

entia

ted

+ fM

LPN

eutr

ophi

l + fM

LP

e

DUJMY

p300

120

300Mr(K)

Figure 4 JMY localizes to the leading edge of motile cells. (a–c) Redistribution of JMY from the nucleus to the leading edge in HL-60 cells. (a) JMY is primarily nuclear in undifferentiated HL-60 cells. (b) After differentiation into motile cells by being cultured in 1.3% DMSO for 5–7 days, JMY co-localizes with filamentous actin in the cytoplasm. (c) Differentiated HL-60 cells were polarized by exposure to 100 nM fMLP, a chemoattractant. JMY is distributed throughout the cytoplasm, where it co-localizes strongly with filamentous actin at the leading edge.

Cells were fixed and stained with Alexa Fluor 568-phalloidin (red), anti-JMY (green) and DAPI (blue). Scale bars, 10 µm. (d) Human primary neutrophils were obtained by finger pinprick26, stimulated with 20 nM fMLP, and fixed and stained as above. JMY (green) co-localizes with filamentous actin (red) at the leading edge. Scale bar, 5 µm. (e) Western blots of JMY and its binding partner p300 in undifferentiated (U) and differentiated (D) HL-60 cells. p300 is expressed in undifferentiated, but not differentiated, HL-60 cells.



l e t t e r s

a b

e g

0.0

0.2

0.4

0.6

0.8

1.0

Wou

nd s

ize

(mm

)

siRNA: Control JMY

Mer

geP

hallo

idin

(red

)

siRNA

Control

JMY

JMY

Actin

siRNA: C J

c d

0

20

40

60

80

100

568-

Pha

lloid

in in

tens

ity (a

.u.)

non-RNAi JMY RNAi

*

WT

GFP–JMY

0

10

20

30

40

50

Mig

ratio

n ra

te (µ

m h

–1)

WT GFP–∆CAGFP–JMY

**Time: 0 2 4 6 h

Time: 0 3 6 24 h

0 3 6 24 h

f

JMY

(gre

en)

DA

PI

(blu

e)

h

JMY JMY

Barbed(+) end

Pointed(–) end

JMY JMY

Nucleation of actinfilament by JMY

Branching by Arp2/3

Arp2/3complex

WH2 domains

Monomer-binding linker

Central domainAcidic domain

Actinmonomers

Model 1

Nucleus:nucleating

actin

Cytoplasm:activating the

Arp2/3 complex

Model 2

120

42

Mr(K)

Figure 5 JMY contributes to cell motility. (a, b) Expressing GFP–JMY in U2OS cells significantly increases their motility in wound-healing assays. (a) Stable lines of GFP–JMY and GFP–JMY∆CA were wounded by being scraped with micropipette tips. Images were acquired at 0, 2, 4, 6 and 12 h after wounding. Scale bars, 100 µm. (b) The migration rate from 0–6 h was averaged from a minimum of four replicates on each of three days. GFP–JMY expression induces cells to migrate 16.6% faster than wild-type cells (n = 3, P < 0.003). Cells expressing a truncation of JMY lacking the Arp2/3-interacting CA domain migrate at the same rate as wild-type cells. Error bars indicate s.e.m. (c–e) Wound-healing assays in U2OS cells indicate that knocking down JMY by RNAi impairs cell migration. Cells were transfected with JMY or control non-targeting 2 (Dharmacon) siRNA (C) and wounded (red dashed line) as in a. Images were taken at the same position 0, 3, 6 and 24 h after wounding. (d) Wound size at each time point for all conditions. (n = 4). Error bars indicate s.e.m. (e) Western blots show RNAi efficiency in U2OS cells. (f) Knocking down

JMY expression decreases cellular levels of F-actin. HEK 293 cells were transfected with a vector encoding both GFP and a JMY-specific shRNA25, then fixed and stained with Alexa Fluor 568-phalloidin (568-phalloidin). The average phalloidin intensity in GFP-negative (non-RNAi-treated) and GFP-positive (RNAi-treated) cells is plotted (n = 396, P < 0.03; see Methods). Error bars indicate s.e.m. (g) JMY localizes to the leading edge of U2OS cells. Cells were grown as above, then fixed and stained for JMY 15 min after wounding. JMY is primarily nuclear, but it is also enriched at the leading edge (indirect immunofluorescence, green) where it co-localizes with a subset of actin filaments (Alexa Fluor 568-phalloidin, red). Inset: leading edge, contrast enhanced to show actin filaments and JMY. Scale bar, 10 µm. (h) Models of in vivo role of JMY. In model 1, JMY nucleates filaments that then serve as substrates for dendritic nucleation by Arp2/3. In model 2, JMY evolved to nucleate actin in different cellular contexts. In the nucleus it nucleates unbranched filaments, and in the cytoplasm it both nucleates filaments and activates Arp2/3.



l e t t e r s

to test whether inactivation of JMY’s function in actin dynamics affects its role as a p300-dependent transcriptional co-activator for p53. Future work is aimed at testing these models.

MeTHodSMolecular biology and biochemistry. Constructs were cloned from full-length mouse JMY4, fly Spire2 and rat N-WASp, with standard techniques. Primer sequences are available from the authors on request. All constructs were sequenced to ensure that no mutations were introduced during cloning. JMY fragments were expressed as glutathione S-transferase (GST) fusions in Escherichia coli and puri-fied with a combination of glutathione and cation chromatography. Except for indi-vidual WH2 domains (Supplementary Information, Fig. S2), the GST was removed to prevent dimerization of the recombinant protein, which results in a marked increase in the rate of both intrinsic nucleation and activation of Arp2/3 (data not shown). To improve the reproducibility of fluorimetry reactions, we mutated the non-conserved cysteine C978 to serine. This mutation does not change the kinetics of activation of Arp2/3 (Supplementary Information, Fig. S6h, i) but results in a peptide that is more stable than the wild type. JMY concentrations were calcu-lated form predicted molar extinction coefficients for JMY peptides that contained tryptophan residues (ProtParam), or by quantitative SDS–PAGE and staining with Sypro-Red (Invitrogen).

Actin polymerization assays. Actin was purified from Acanthamoeba castellani as described21, labelled with pyrene iodoacetamide as described22, and stored on ice. Arp2/3 was purified from Acanthamoeba as described23 and flash-frozen with 10% glycerol. For all assays, Arp2/3 was thawed daily and diluted with 1 mg ml−1 BSA in buffer A (0.2 mM ATP, 0.5 mM tris(2-carboxyethyl)phosphine (TCEP), 0.1 mM CaCl2, 0.02% w/v sodium azide, 2 mM Tris-HCl pH 8.0 at 4 °C). Actin polymeriza-tion assays were performed in 1 × KMEI (50 mM KCl, 1 mM MgCl2, 1 mM EGTA, 10 mM imidazole pH 7.0). Ca2+-actin was converted into Mg2+-actin by incubating actin in 50 mM MgCl2, 0.2 mM EGTA for 2 min before adding 10 × KMEI and test components. Pyrene fluorescence was measured with an ISS PCI/K2 fluorimeter. Unless otherwise noted, polymerization reactions contained 2 µM actin (labelled with 5% pyrene), 2.5 nM Arp2/3 and 167 nM JMY or N-WASp. JMY proteins were diluted with 10 mg ml−1 BSA in buffer A to prevent loss of activity. To normalize fluorimetry data, we subtracted the offset from zero then divided by the plateau value of actin alone, so as to not mask the effects of sequestration by JMY. For half-time calculations, reactions were normalized by dividing by their own plateau and solving for time at half-maximal fluorescence (0.5 arbitrary units).

To reveal actin filaments we polymerized 2 µM actin under the same condi-tions as those used for fluorimetry, then arrested reactions at the times indicated with Alexa Fluor 488 phalloidin and Latrunculin B. This technique preserves the ratio of monomeric actin to filamentous actin at the moment of quenching24. To keep the concentration of F-actin constant we arrested reactions at their indi-vidual t9/10 (time at nine-tenths-maximal fluoresence, 0.9 arbitrary units) values: 1 min 40 s for JMY + Arp2/3, 6 min 20 s for Scar + Arp2/3, and 37 min 0 s for actin alone. Filaments were diluted to low nanomolar concentrations and spotted onto coverslips coated with poly-(l-lysine) (Sigma), using wide-bore pipette tips to minimize shearing.

Cell culture. U2OS (ATCC) and HEK 293 cells were cultured in DMEM medium supplemented with 10% FBS, 2 mM l-glutamine, non-essential amino acids, and penicillin–streptomycin (UCSF Cell Culture facility). For transfection, cells were seeded onto glass coverslips and transfected with haemagglutinin-tagged JMY4 using GeneJuice (Merck), or with GFP–JMY and GFP–RNAi constructs using Lipofectamine LTX (Invitrogen), in accordance with the manufacturer’s protocol. HL-60 cells were cultured as described25, in RPMI-1640 with 25 mM HEPES, 2.0 g l−1 NaHCO3, 10% FBS, 1% antibiotic–antimycotic (Fisher), and were pas-saged every three or four days to a density of 2 × 105 cells ml−1. Differentiation was in complete medium containing 1.3% DMSO (Hybrimax; Sigma). All cells were grown at 37 °C with 5% CO2. Primary human neutrophils were obtained by finger pinprick as described26.

Immunofluorescence of HL-60 cells was performed as described25. In brief, flamed coverslips were treated with 200 µg ml−1 bovine plasma fibronectin (Sigma) in PBS, washed with PBS, and blocked with 1.8% low-endotoxin BSA (Sigma) in modified Hanks buffered saline solution (mHBSS: 150 mM NaCl, 4 mM KCl, 1 mM MgCl2, 10 mM glucose, 20 mM HEPES pH 7.4 at 22 °C) for 5 min before

adhering cells. Cells were pelleted and resuspended in BSA/mHBSS, adhered to coverslips for 30–60 min at 37 °C, and then washed to remove unbound cells. Stimulation or mock stimulation was in BSA/mHBSS with 100 nM (HL-60s) or 20 nM (human neutrophils) fMLP, or DMSO (carrier), for 5 min at 22 °C. Cells were fixed in 3.2% formaldehyde in cytoskeletal buffer (138 mM KCl, 3 mM MgCl2, 2 mM EGTA, 320 mM sucrose, 10 mM HEPES pH 7.2 at 22 °C). For immunofluorescence of U2OS cells, cells were plated on flamed, fibronectin-coated coverslips, and fixed for 30 min in 3.2% formaldehyde in PBS. Cells were then permeabilized with 0.1% Triton X-100 in PBS with 1.4 U ml−1 Alexa Fluor 568-phalloidin (Invitrogen) to stabilize and reveal filaments. For JMY immunolo-calization, rabbit polyclonal JMY antibody 1289 (ref. 5) or anti-haemagglutinin antibody HA11 (Babco) was used at 1:500 dilution. Alexa Fluor 488-labelled goat anti-rabbit secondary antobody (Invitrogen) was used at 1:500 dilution. 4,6-Diamidino-2-phenylindole (DAPI; Sigma) was used at 0.5 µg ml−1. Samples were mounted with fluorescent mounting medium (DakoCytomation).

Epifluorescence and wide-field images were acquired on a Nikon TE300 inverted microscope equipped with a Hamamatsu C4742-98 cooled charge-coupled device camera, with Simple PCI software (Compix), using 100× and 60× 1.4 numerical aperture (NA) Plan Apo objectives (Nikon), or with a 10× 0.6 NA phase objective, using MicroManager software27. For quantification of F-actin levels in cells expressing GFP–PWWWCA or GFP, micrographs were acquired with an IX Micro automated microscope, using identical illumination condi-tions. Cells were identified and outlined by using ImageJ software, background was subtracted, and the average intensities in the red and green channels were measured. We used ImageJ (National Institutes of Health) and Adobe Photoshop (Adobe) for image analysis and contrast adjustment.

For wound-healing experiments, cells were plated on marked coverslips at the same density, scratch-wounded with a micropipette tip, washed to removed detached cells, and then given fresh medium and kept at 37 °C during image acquisition. Stable lines of U2OS cells stably expressing GFP–JMY and variants were selected in 500 µg ml−1 G418 for at least three weeks, then enriched by fluo-rescence-activated cell sorting (FACS) (UCSF Flow Cytometry Core). siRNA was used at a final concentration of 25 nM (hJMY siRNA from Santa Cruz and control non-targeting 2 from Dharmacon) and transfected into cells with Oligofectamine or Dharmafect 1 (both in accordance with the manufacturer’s protocols), and cells were grown for 72 h before experiments. To discriminate between cells treated with RNAi and cells not treated with RNAi, pL-UGIH plasmid (American Type Culture Collection) was modified to contain a human JMY-specific hairpin based on the sequence of JMY-specific siRNA-3 by the method of ref. 25. HEK 293 cells transiently transfected with this construct were grown for seven days before fixa-tion or immunoblotting.

Standard methods were used for western blotting, using 1:500 1289 or 1:1,000 L-16 (Santa Cruz Biotechnology) JMY primary antibodies and 1:5,000 horse-radish peroxidase (HRP)-labelled anti-rabbit secondary antibodies (Jackson ImmunoResearch). p300-CT primary antibody (Millipore) was used at 1:500, goat anti-human glyceraldehyde-3-phosphate dehydrogenase (Santa Cruz) was used at a dilution of 1:10,000, and mouse anti-human actin (Sigma) was used at 1:20,000. HRP-labelled secondary antibodies (Dako) were used at 1:10,000 dilu-tion, and enhanced chemiluminescence reagent (SuperSignal West Pico; Pierce) was used in accordance with the manufacturer’s instructions. All error values are s.e.m. We used two-tailed unpaired t-tests, assuming unequal variance, to calculate P values (Microsoft Excel).


ACkNowLEDgEMENTSWe thank A. Kelly and R. Manlove for sequence analysis; O. Akin for help with Matlab scripts and advice; C. Campbell and H. Bourne for critical reading of the manuscript; and O. Weiner, S. Wilson, P. Temkin, E. Oh, C. Vizcarra, S. Cai, M. D’Ambrosio, K. Campellone, and members of the Mullins laboratory for reagents and helpful discussions. This work was supported by grants from the National Institutes of Health, the American Heart Association Predoctoral Fellowship (J.B.Z.), Medical Research Council funding (A.S.C.), and the Burroughs-Wellcome Fund Career Award in the Biomedical Sciences Fellowship (M.E.Q.).

AuThoR CoNTRiBuTioNSJ.B.Z., A.S.C. and M.E.Q. conducted the experiments and analysed the results. J.B.Z., A.S.C., M.E.Q., N.B.T. and R.D.M. conceived the experiments and wrote the manuscript.



l e t t e r s

CoMpETiNg fiNANCiAL iNTERESTSThe authors declare that they have no competing financial interests.

Published online at http://www.nature.com/naturecellbiology reprints and permissions information is available online at http://npg.nature.com/reprintsandpermissions/

1. Welch, M. & Mullins, R. Cellular control of actin nucleation. Annu. Rev. Cell Dev. Biol. 18, 247–288 (2002).

2. Quinlan, M. E., Heuser, J. E., Kerkhoff, E. & Mullins, R. D. Drosophila Spire is an actin nucleation factor. Nature 433, 382–388 (2005).

3. Marchand, J. B., Kaiser, D. A., Pollard, T. D. & Higgs, H. N. Interaction of WASP/Scar proteins with actin and vertebrate Arp2/3 complex. Nature Cell Biol. 3, 76–82 (2001).

4. Shikama, N. et al. A novel cofactor for p300 that regulates the p53 response. Mol. Cell 4, 365–376 (1999).

5. Coutts, A., Boulahbel, H., Graham, A. & La Thangue, N. Mdm2 targets the p53 tran-scription cofactor JMY for degradation. EMBO Rep. 8, 84–90 (2006).

6. Symons, M. et al. Wiskott–Aldrich syndrome protein, a novel effector for the GTPase CDC42Hs, is implicated in actin polymerization. Cell 84, 723–734 (1996).

7. Pollard, T. & Borisy, G. Cellular motility driven by assembly and disassembly of actin filaments. Cell 112, 453–465 (2003).

8. Cooper, J. & Pollard, T. Effect of capping protein on the kinetics of actin polymerization. Biochemistry 24, 793–799 (1985).

9. Pan, F., Egile, C., Lipkin, T. & Li, R. ARPC1/Arc40 mediates the interaction of the actin-related protein 2 and 3 complex with Wiskott–Aldrich syndrome protein family activators. J. Biol. Chem. 279, 54629–54636 (2004).

10. Ahuja, R. et al. Cordon-Bleu is an actin nucleation factor and controls neuronal mor-phology. Cell 131, 337–350 (2007).

11. Collins, S., Ruscetti, F., Gallagher, R. & Gallo, R. Normal functional characteristics of cultured human promyelocytic leukemia cells (HL-60) after induction of differentiation by dimethylsulfoxide. J. Exp. Med. 149, 969–974 (1979).

12. Kowalski, J. R. et al. Cortactin regulates cell migration through activation of N-WASP. J. Cell Sci. 118, 79–87 (2005).

13. Sablina, A., Chumakov, P. & Kopnin, B. Tumor suppressor p53 and its homologue p73α affect cell migration. J. Biol. Chem. 278, 27362–27371 (2003).

14. Gadea, G., de Toledo, M., Anguille, C. & Roux, P. Loss of p53 promotes RhoA-ROCK-dependent cell migration and invasion in 3D matrices. J. Cell Biol. 178, 23–30 (2007).

15. Krubasik, D. et al. Absence of p300 induces cellular phenotypic changes characteristic of epithelial to mesenchyme transition. Br. J. Cancer 94, 1326–1332 (2006).

16. Southwick, F. S., Dabiri, G. A., Paschetto, M. & Zigmond, S. H. Polymorphonuclear leukocyte adherence induces actin polymerization by a transduction pathway which differs from that used by chemoattractants. J. Cell Biol. 109, 1561–1569 (1989).

17. Mullins, R., Heuser, J. & Pollard, T. The interaction of Arp2/3 complex with actin: nucleation, high affinity pointed end capping, and formation of branching networks of filaments. Proc. Natl Acad. Sci. USA 95, 6181–6186 (1998).

18. Blanchoin, L. et al. Direct observation of dendritic actin filament networks nucleated by Arp2/3 complex and WASP/Scar proteins. Nature 404, 1007–1011 (2000).

19. Blanchoin, L., Pollard, T. D. & Hitchcock-DeGregori, S. E. Inhibition of the Arp2/3 complex-nucleated actin polymerization and branch formation by tropomyosin. Curr. Biol. 11, 1300–1304 (2001).

20. Shao, D., Forge, A., Munro, P. & Bailly, M. Arp2/3 complex-mediated actin polymeri-sation occurs on specific pre-existing networks in cells and requires spatial restric-tion to sustain functional lamellipod extension. Cell Motil. Cytoskel. 63, 395–414 (2006).

21. Gordon, D. J., Eisenberg, E. & Korn, E. D. Characterization of cytoplasmic actin isolated from Acanthamoeba castellanii by a new method. J. Biol. Chem. 251, 4778–4786 (1976).

22. Cooper, J. A., Walker, S. B. & Pollard, T. D. Pyrene actin: documentation of the validity of a sensitive assay for actin polymerization. J. Muscle Res. Cell Motil. 4, 253–262 (1983).

23. Dayel, M. J., Holleran, E. A. & Mullins, R. D. Arp2/3 complex requires hydrolyzable ATP for nucleation of new actin filaments. Proc. Natl Acad. Sci. USA 98, 14871–14876 (2001).

24. Akin, O. & Mullins, R. D. Capping protein increases the rate of actin-based motility by promoting filament nucleation by the Arp2/3 complex. Cell 133, 841–851 (2008).

25. Weiner, O. et al. Hem-1 complexes are essential for Rac activation, actin polymerization, and myosin regulation during neutrophil chemotaxis. PLoS Biol. 4, e38 (2006).

26. Cassimeris, L., McNeill, H. & Zigmond, S. Chemoattractant-stimulated polymorphonu-clear leukocytes contain two populations of actin filaments that differ in their spatial distributions and relative stabilities. J. Cell Biol. 110, 1067–1075 (1990).

27. Stuurman, N., Amodaj N., Vale, R. D. Micro-Manager: open source software for light microscope imaging. Microsc. Today 15, 42–43 (2007).

28. Petrella, E., Machesky, L., Kaiser, D. & Pollard, T. Structural requirements and ther-modynamics of the interaction of proline peptides with profilin. Biochemistry 35, 16535–16543 (1996).

29. Kelly, A., Kranitz, H., Dötsch, V. & Mullins, R. Actin binding to the central domain of WASP/Scar proteins plays a critical role in the activation of the Arp2/3 complex. J. Biol. Chem. 281, 10589–10597 (2006).

30. Kreishman-Deitrick, M. et al. NMR analyses of the activation of the Arp2/3 com-plex by neuronal Wiskott–Aldrich syndrome protein. Biochemistry 44, 15247–15256 (2005).

31. Campellone, K. G., Webb, N. J., Znameroski, E. A. & Welch, M. D. WHAMM is an Arp2/3 complex activator that binds microtubules and functions in ER to Golgi transport. Cell 134, 148–161 (2008).






l e t t e r s

Characterization of the interface between normal and transformed epithelial cellsCatherine Hogan1, Sophie Dupré-Crochet1,7, Mark Norman1, Mihoko Kajita1, Carola Zimmermann1, Andrew E. Pelling3, Eugenia Piddini4, Luis Alberto Baena-López4, Jean-Paul Vincent4, Yoshifumi Itoh5, Hiroshi Hosoya6, Franck Pichaud1,2 and Yasuyuki Fujita1,2,8

In most cancers, transformation begins in a single cell in an epithelial cell sheet1–3. However, it is not known what happens at the interface between non-transformed (normal) and transformed cells once the initial transformation has occurred. Using Madin-Darby canine kidney (MDCK) epithelial cells that express constitutively active, oncogenic Ras (RasV12) in a tetracycline-inducible system, we investigated the cellular processes arising at the interface between normal and transformed cells. We show that two independent phenomena occur in a non-cell-autonomous manner: when surrounded by normal cells, RasV12 cells are either apically extruded from the monolayer, or form dynamic basal protrusions and invade the basal matrix. Neither apical extrusion nor basal protrusion formation is observed when RasV12 cells are surrounded by other RasV12 cells. We show that Cdc42 and ROCK (also known as Rho kinase) have vital roles in these processes. We also demonstrate that E‑cadherin knockdown in normal cells surrounding RasV12 cells reduces the frequency of apical extrusion, while promoting basal protrusion formation and invasion. These results indicate that RasV12-transformed cells are able to recognize differences between normal and transformed cells, and consequently leave epithelial sheets either apically or basally, in a cell-context-dependent manner.

Cell transformation arises from the activation of oncoproteins and/or inactivation of tumour suppressor proteins1. During the initial stage of carcinogenesis, transformation occurs in a single cell in an epithelial monolayer2,3. However, it remains unclear what happens at the interface between normal and transformed cells during this process. For exam-ple, do surrounding normal cells recognize the transformation that has occurred in their neighbour? What is the fate of the transformed cell when surrounded by normal cells? To address these questions, we

established MDCK epithelial cells expressing GFP-tagged RasV12 in a tet-racycline-inducible manner (MDCK-pTR GFP–RasV12; Supplementary Information, Fig. S1a, b). After addition of tetracycline to cells plated at low density, expression of GFP–RasV12 induced cell scattering and downregulation of cell–cell contacts (Supplementary Information, Fig. S1c)4. In contrast, when cells were plated at high density, expression of GFP–RasV12 did not induce loss of cell–cell adhesion, and cells retained an epithelial morphology (Supplementary Information, Fig. S1d). In all subsequent experiments, we cultured cells at high density.

To examine the fate of a single RasV12 cell in a monolayer of normal MDCK cells, MDCK-pTR GFP–RasV12 cells were labelled with fluo-rescent dye (CMPTX) and mixed with normal MDCK cells at a ratio of 1:100. The combination of cells was then cultured on a collagen matrix in the absence of tetracycline until a monolayer was formed. Subsequently, GFP–RasV12 expression was induced with tetracycline, and the fate of RasV12 cells surrounded by normal cells was observed using time-lapse microscopy. Between 13 and 25 h after tetracycline addition, RasV12 cells were frequently extruded from the apical surface of the monolayer (84% of tracked cells, n = 74; Fig. 1a; Supplementary Information, Movie 1). We observed that RasV12 cells often underwent cell division before extrusion; however, we also observed extrusion of single RasV12 cells (Supplementary Information, Fig. S2a). After extrusion, RasV12 cells continued to proliferate and form multicellular aggregates (Fig. 1a, c; Supplementary Information, Movie 1). These multicellular structures attached loosely to the underlying normal cells and dynamically moved over them (Supplementary Information, Fig. S2b and Movie 1). Extruded RasV12 cells did not stain with ethidium dye (data not shown), indicating that they remained alive. Treatment with the K+ channel inhibitor 4-aminopyridine (4-AP), which blocks the early stage of apoptosis in epithelial cells5, did not appreciably reduce the frequency of extrusion (Fig. 1d), further indicating that extrusion of RasV12 cells occurs in a cell death-independent manner. Cells expressing

1MRC Laboratory for Molecular Cell Biology and Cell Biology Unit, 2Department of Cell and Developmental Biology, University College London, Gower Street, London, WC1E 6BT, UK. 3London Centre for Nanotechnology and Centre for Nanomedicine, 17–19 Gordon Street, London, WC1H 0AH, UK. 4Division of Developmental Neurobiology, National Institute for Medical Research, London, NW7 1AA, UK. 5Imperial College London, The Kennedy Institute of Rheumatology, 1 Aspenlea Road, London, W6 8LH, UK. 6Department of Biological Science, Graduate School of Science, Hiroshima University, Higashi-Hiroshima 739–8526, Japan. 7Current address: INSERM UMR 757, Université Paris-sud, Bat 443, 91405 Orsay cedex, France. 8Correspondence should be addressed to Y.F. (e-mail: [email protected])

Received 30 October 2008; accepted 12 December 2008; published online 15 March 2009; DOI: 10.1038/ncb1853




l e t t e r s

GFP–Rap1V12, a constitutively active form of Ras-related non-oncogenic small GTPase6, were not extruded but remained in the monolayer of normal cells (Supplementary Information, Fig. S2c, d and Movie 2). In the absence of tetracycline, we did not observe extrusion of RasV12 cells (data not shown). Moreover, fluorescently labelled RasV12 cells were not extruded when mixed with non-labelled RasV12 cells (Fig. 1b, c; Supplementary Information, Movie 3), suggesting that the extrusion of RasV12 cells depends on their interaction with normal cells. Similarly, fluorescently labelled normal cells were not extruded from a monolayer of normal cells (data not shown). Taken together, these data indicate that the extrusion of RasV12 cells occurs in a non-cell-autonomous man-ner only when they are surrounded by normal cells. To examine the physiological relevance of this observation, RasV12 was expressed in a mosaic manner in wing imaginal discs of Drosophila melanogaster. RasV12-expressing cells were apically extruded from normal epithelial cell sheets (14%, n = 85; Fig. 1e, left panels), showing that apical extru-sion of RasV12-expressing cells occurs in vivo. Cells expressing RasN17, a constitutively inactive form of Ras, or wild-type Ras were not extruded and remained within the disc epithelium (0%, n = 28 or 0%, n = 19;

Fig. 1e, right panels; Supplementary Information, Fig. S3a). When RasV12 was expressed in the entire epithelium, the monolayer became irregu-larly folded but we did not observe apical extrusion (Supplementary Information, Fig. S3b)7.

To understand the molecular mechanism of apical extrusion, we analysed RasV12-expressing MDCK cells that were not yet extruded and remained in the monolayer of normal MDCK cells. We found that the height of RasV12 cells along the apicobasal axis was significantly higher than that of the surrounding normal cells (Fig. 2a, b). When RasV12 cells alone formed a monolayer, the cell height was comparable to that of normal cells (Fig. 2b). These results indicate that the height of RasV12 cells increases in a cell-context-dependent manner. We also observed that phosphorylation of myosin light chain was enhanced in RasV12 cells with increased height (Fig. 2c), but not in those with similar height to surrounding normal cells (data not shown), suggesting a correlation between increased cell height and activation of myosin. Furthermore, we found that F-actin accumulated at cell–cell contacts between RasV12 cells surrounded by normal cells (Fig. 2a, d), but not when surrounded by RasV12 cells (Fig. 2d). F-actin did not accumulate between normal

d

Control + 4-AP

GFP

–Ras

V12

cel

ls e

xtru

ded

(per

cent

age)

0102030405060708090

100

a CMPTX

10’ 1 h 4 h 7 h

13 h 24 h

GFP–RasV12

24 h20 h

b

1 h 4 h

13 h

CMPTX

10’

20 h 24 h

7 h

24 h

GFP–RasV12

e

xz

xz

LacZ LacZ

Merge

RasV12 RasN17

Merge

Apical

Apical

Basal

Basal

c

xy

xy

xz

+ Tet (48 h)– Tet

xz

GFP–RasV12:MDCK1:100

GFP–RasV12

Figure 1 Epithelial cells expressing RasV12 are apically extruded from surrounding normal epithelium in a non-cell-autonomous manner. (a, b) RasV12-expressing cells were extruded from a monolayer of normal cells, but not from a monolayer of RasV12-expressing cells. MDCK-pTR GFP–RasV12 cells were combined with normal MDCK cells (a) or MDCK-pTR GFP–RasV12 cells (b) at a ratio of 1:100 and cultured on type-I collagen gels, followed by tetracycline treatment. Images were extracted from a representative time-lapse analysis. Red arrows indicate fluorescently labelled RasV12 cells. (c) Confocal images of xz and xy sections of GFP–RasV12 cells combined with normal MDCK cells (upper panels) or

with GFP–RasV12 cells (lower panels) on collagen gels. Cells were fixed after 48 h incubation with (+ Tet, 2 μg ml–1) or without (– Tet) tetracycline and stained with anti-gp135 antibody (cyan) and Hoechst (blue). (d) Quantification of time-lapse analyses of RasV12 cells extruded from a monolayer of normal cells 24 h after tetracycline addition in the absence (control) or presence of 4-AP. Data are mean ± s.d. of three independent experiments (n = 98 cells control, n = 101 cells 4-AP). (e) Drosophila wing imaginal disc epithelium co-expressing lacZ and RasV12 or RasN17. Cells were stained with anti-β-galactosidase antibody (green) and Hoechst (blue). Scale bars, 20 μm (a–c, e).



l e t t e r s

and RasV12 cells (Fig. 2a, d). Collectively, these results suggest that RasV12 cells recognize that they are surrounded by normal cells and modulate their shape and cytoskeleton accordingly. E-cadherin-based cell–cell

contacts between RasV12 cells were not disrupted when these cells were in the monolayer of normal cells (Supplementary Information, Fig. S4a), or after they were apically extruded (data not shown), suggesting that

a b

GFP–RasV12

F-actin

Merge

GFP–RasV12 cells aloneGFP–RasV12

MDCK GFP–RasV12:MDCK1:100

Cel

l hei

ght

(pix

els)

0

10

20

30

40

50

60

70 ******

c d

F-ac

tin a

ccum

ulat

ion

(pix

el in

tens

ity)

0

5000

10000

15000

20000

25000

30000

***

M:MR:R R:M


R:R

GFP–RasV12

cells alone

GFP–RasV12

p-MLC

Merge

e f< 25% extrusion

> 75% extrusion

DM

SO

Y276

32

U01

26

LY29

4002

Ble

bbis

tatin

Cyt

o D

0

10

20

30

40

50

60

70

80

90

100

Ap

ical

ext

rusi

on (p

erce

ntag

e)

***

**

**

*

***

******

+ U0126

+ DMSO

+ Blebbistatin

+ Cyto D

Figure 2. Molecular mechanism for apical extrusion of RasV12-expressing cells from a monolayer of normal cells. (a) Confocal images of xz sections of MDCK-pTR GFP–RasV12 cells in a monolayer of normal MDCK cells. Twenty-four hours after tetracycline addition, cells were stained with TRITC-phalloidin (red) and Hoechst (blue). (b) Quantification of cell height. Grey bar, MDCK cells (n = 155); light blue bar, GFP–RasV12 cells surrounded by MDCK cells (n = 103); dark blue bar, GFP–RasV12 cells only (n = 177). Data are mean ± s.d.; ***P < 0.0001; n, cells. A total of 40–85 cells from three independent experiments were analysed. (c) Enhanced phosphorylation of myosin light chain (p-MLC) in GFP–RasV12 cells surrounded by normal cells. Twenty hours after tetracycline addition, cells were stained with anti-phospho-MLC antibody (red). (d) Quantification of F-actin accumulation at cell–cell contacts. R:R, between RasV12 cells (with normal cells n = 38, RasV12 cells alone n = 64); R:M, between RasV12 and normal MDCK cells

(n = 24); M:M, between normal MDCK cells (n = 52). Data are mean ± s.d.; ***P < 0.0001; n, cell–cell contacts. (e) GFP–RasV12 cells in a monolayer of normal MDCK cells after treatment with inhibitors. Twenty-four hours after tetracycline addition in the presence of U0126, blebbistatin or cytochalasin D (Cyto D), cells were stained with TRITC-phalloidin (red) and Hoechst (blue). Merged images are shown. (f) Quantification of frequency of apical extrusion of GFP–RasV12 cells from a monolayer of MDCK cells in the presence of various inhibitors. In each experiment, approximately 20 groups of RasV12-expressing cells were counted. White bar, more than 75% of cells in a group of RasV12 cells are apically extruded. Black bar, less than 25% of cells are extruded. Data are mean ± s.d. of at least four independent experiments ;***P < 0.0001, **P < 0.005, *P < 0.05; n = 123, 57, 76 , 55, 52 and 57 groups of cells, for each inhibitor, respectively. Scale bars, 20 μm (a, c, e).



l e t t e r s

deregulation of E-cadherin in RasV12 cells is not involved in apical extru-sion. To further investigate the molecular mechanism of apical extrusion of RasV12 cells, we examined the effect of various inhibitors that modulate cell signalling and/or the cytoskeleton. Among the signalling pathways downstream of active Ras, the mitogen-activated protein kinase (MAPK) and phosphatidylinositol-3-kinase (PI(3)K) pathways are required for Ras-induced transformation8. Addition of U0126, an inhibitor of MAPK kinases, completely suppressed apical extrusion of RasV12 cells, whereas the PI(3)K inhibitor LY294002 had no effect (Fig. 2e, f), suggesting that the MAPK pathway is involved in apical extrusion. As PI(3)K has been shown to have a crucial role in lamellipodia formation9, it is unlikely that apical extrusion is due to increased migration of RasV12 cells. Indeed, we never observed lamellipodia or membrane ruffling of RasV12 cells during or after extrusion (data not shown). Blebbistatin and cytochalasin D, which inhibit myosin-II activity and actin polymerization respectively, significantly suppressed apical extrusion (Fig. 2e, f), as did the ROCK inhibitor Y27632 (Fig. 2f). Blebbistatin also suppressed the increase in cell height of RasV12 cells surrounded by normal cells (data not shown). Taken together, these data indicate that myosin-II activity and actin polymerization are required for apical extrusion of RasV12 cells.

Although most RasV12 cells were apically extruded when surrounded by normal cells, we also observed RasV12 cells that were not extruded.

We examined the fate of these non-extruded cells, and found that non-extruded RasV12 cells formed large protrusions that dynamically extended and retracted, often over distances of several cell diameters (Fig. 3a; Supplementary Information, Movie 4). Confocal microscopy analysis, showed that these protrusions extended beneath the neighbouring nor-mal cells (Fig. 3b, lower panels). In contrast, MDCK cells expressing GFP did not produce large protrusions (Fig. 3b, upper panels). Apical extrusion and basal protrusion formation of RasV12 cells were observed in the same monolayer of normal cells (Supplementary Information, Fig. S4b). When expression of RasV12 was induced in a group of cells within a monolayer of normal cells, major protrusions (> 10 μm) were frequently formed at the interface between RasV12 and normal cells (Fig. 3c, white arrows, 3d), but were rarely observed between RasV12 cells (Fig. 3c, d). This indicates that protrusion formation also occurs in a non-cell-autonomous fashion. Both F-actin and microtubules were found within the protrusions (Supplementary Information, Fig. S4c), suggesting a role for the cytoskeleton in the dynamic movement of pro-trusions. Although blebbistatin inhibited apical extrusion (Fig. 2f), it did not affect protrusion formation (Fig. 3e). In contrast, LY294002 inhib-ited protrusion formation (Fig. 3e), but not apical extrusion (Fig. 2f). Therefore, apical extrusion and basal protrusion formation are regulated, at least partially, by distinct molecular mechanisms. We also found that

d

xyBasal

xz

xz

xyBasal

MDCKGFP Merge

GFP–RasV12 MergeMDCK

b

Major protrusionsMinor protrusions

Between RasV12

and normal cellsBetween RasV12

cells

Cel

l–ce

ll co

ntac

ts w

ith

pro

trus

ions

(per

cent

age)

0

10

20

30

40

50

60 ***

DM

SO

Y276

32

Ble

bbis

tatin

U01

26

LY29

4002

Cyt

o D

Pro

trus

ion

form

atio

n(r

atio

rel

ativ

e to

DM

SO

)

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

**

**

e gf

xz

xy xy

xz

a

c

GFP–RasV12 E-cadherin

GFP–RasV12

Figure 3. Non-extruded GFP–RasV12 cells produce dynamic basal protrusions beneath the neighbouring MDCK cells. (a) Images extracted from a time-lapse analysis of a GFP–RasV12 cell (red arrow) surrounded by normal cells. Images were captured at 50 min intervals. White arrows indicate protrusions. (b) Confocal images of MDCK cells expressing GFP (upper panels) or GFP–RasV12 (lower panels) in a monolayer of normal cells. Red bars in the xy-labelled panels denote the cross-sections represented in xz labelled panels. Surrounding normal cells were labelled with CMPTX (red). (c) Confocal images of GFP–RasV12 cells in a monolayer of normal cells stained with anti-E-cadherin antibody (red, right panel). White arrows indicate protrusions. (d) Quantification of protrusion formation at cell–cell

contacts between RasV12 and normal cells (n = 230) or between RasV12 cells (n = 180). Data are mean ± s.d.; ***P < 0.0001; n, cell–cell contacts. (e) Quantification of protrusion formation of GFP–RasV12 cells in a monolayer of normal cells in the presence of various inhibitors. Values are expressed as a ratio relative to the DMSO control, and represent mean ± s.e.m.; *P < 0.05 **P < 0.005; n = 320, 170, 300, 382, 362 and 40 groups of cells, for each inhibitor, respectively. (f, g) Confocal images of GFP– RasV12 cells in a monolayer of normal MDCK cells on collagen gels. After (f) 48 h or (g) 120 h of tetracycline incubation, cells were stained with anti-gp135 antibody (cyan) and Hoechst (blue). White arrows and arrowhead indicate protrusions. Scale bars, 20 μm (a–c, f, g).



l e t t e r s

adherens junctions are specifically disrupted at the interface between RasV12 and normal cells where protrusions are formed (Supplementary Discussion and Supplementary Information, Fig. S5). RasV12 cells also formed major protrusions when cultured with normal cells on a col-lagen matrix (Fig. 3f, white arrows and arrowhead). Such protrusions extended beneath neighbouring normal cells and into the collagen. When cells were cultured for longer periods (120 h), non-extruded RasV12 cells invaded the collagen and proliferated underneath the nor-mal cells (Fig. 3g). This occurred non-cell-autonomously, as in a mon-olayer of only RasV12 cells, RasV12 cells did not form major protrusions nor invade the collagen (0%, n = 25 independent experiments; Fig. 1c, lower right panels). Basal extrusion of RasV12 cells was also observed in vivo in the wing imaginal disc epithelium of Drosophila (40%, n = 85; Supplementary Information, Fig. S3c).

The molecular mechanisms of apical extrusion and basal protrusion for-mation of RasV12 cells were further studied using transient expression sys-tems. First, we confirmed that transient transfection of MDCK cells with GFP–RasV12 induced apical extrusion and basal protrusions comparable

to those observed in MDCK-pTR GFP–RasV12 cells (Fig. 4a). Expression of dominant-active Raf (RafCAAX) alone induced neither apical extrusion nor basal protrusions (Fig. 4a; Supplementary Information, Fig. S6a). Co-expression of dominant-negative Raf markedly suppressed RasV12-induced MAPK activation and apical extrusion (Fig. 4b; Supplementary Information, Fig. S6a). These results suggest that the MAPK pathway is required, but not sufficient, to promote apical extrusion of RasV12 cells. Expression of dominant-active PI(3)K (p110CAAX) alone promoted nei-ther apical extrusion nor basal protrusion formation (Fig. 4a), whereas addition of LY294002 significantly suppressed basal protrusion formation (Fig. 3e), suggesting that the PI(3)K pathway is required, but not sufficient, to induce basal protrusions. These results suggest an involvement of other Ras targets in these processes10.

ROCK induces phosphorylation of myosin light chain and activates myosin-II (Supplementary Information, Fig. S7a)11. Co-expression of dominant-negative ROCK (ROCKDN) significantly inhibited RasV12-induced phosphorylation of myosin light chain (Supplementary Information, Fig. S7b, c). It also decreased apical extrusion while

d e


MDCK GFP–RasV12

GFP–RasV12

cells alone

GFP–RasV12

GS

T–W

AS

P-C

RIB

(pix

el in

tens

ity)

0

1000

2000

3000

4000

5000

6000

7000

8000 *****+ GST + GST–WASP-CRIB


GFP–RasV12

Merge

GST

Merge

GST

GFP–RasV12

GFP–RasV12

Merge

GST GST

Merge

GFP–RasV12 GFP–RasV12

a

ProtrusionApical extrusion

Cel

ls +

phe

noty

pes

(per

cent

age)

Cel

ls +

phe

noty

pes

(per

cent

age)

Cel

ls +

phe

noty

pes

(per

cent

age)

0

5

10

15

20

25

30

35 ***

GFP GFP–RasV12

GFP–RasV12

GFP

RafCAAX p110CAAX

+

b


0

5

10

15

20

25

30

35 ***

GFP

GFP

–Ras

V12

GFPRafS621A

+GFP–RasV12

RafS621A

+GFP

c


+GFP–RasV12

Cd

c42N

17

RO

CK

DN

Cd

c42N

17

RO

CK

DN

+GFP

0

10

20

30

40

50

60

70

80***

**

Figure 4. Molecular mechanisms for apical extrusion and basal protrusion formation of RasV12 cells in a monolayer of normal cells. (a, b) Involvement of MAPK and PI(3)K pathways. Indicated proteins were transiently expressed in MDCK cells, and transfected cells that were apically extruded or formed basal protrusions were quantified. GFP or GFP–RasV12 was expressed with or without dominant-active Raf (RafCAAX) or PI(3)K (p110CAAX, a). GFP or GFP–RasV12 was expressed with or without dominant negative Raf (RafS621A, b). For each experiment 50–150 cells were counted; ***P < 0.0001. (c) Involvement of Cdc42 and ROCK. GFP or GFP–RasV12 was expressed with or without constitutively inactive Cdc42 (Cdc42N17) or dominant-negative ROCK (ROCKDN) in MDCK cells, and the transfected cells that were apically extruded or formed basal protrusions were quantified. For

each experiment 50–115 cells were counted; **P < 0.005 ***P < 0.0001. White bar, cells that are apically extruded; black bar, cells that form major protrusions (a–c). (d, e) Non-cell-autonomous activation of Cdc42 in RasV12 cells. Confocal images of MDCK-pTR GFP–RasV12 cells (d) in a monolayer of normal MDCK cells (upper panels) or MDCK-pTR GFP–RasV12 cells (lower panels). Eight hours after tetracycline addition, cells were fixed and incubated with GST (left panels) or GST–WASP-CRIB proteins (right panels), followed by immunostaining with anti-GST antibody (cyan) and Hoechst (blue). Scale bar, 20 μm. WASP-CRIB staining was quantified (e). Data are mean ± s.e.m.; **P < 0.005 ***P < 0.0001; n = 55, 50 and 65 cells for each condition, respectively. Data are mean ± s.e.m. from more than five independent experiments in a, b and c.



l e t t e r s

increasing basal protrusion formation (Fig. 4c), suggesting an impor-tant role of ROCK in these processes. We also examined whether Rho GTPases, crucial regulators for cytoskeletons and cell morphology12,13, have a role in these processes. Co-expression of constitutively inac-tive Cdc42 (Cdc42N17) with RasV12 strongly suppressed apical extru-sion and enhanced basal protrusion formation (Fig. 4c), whereas co-expression of constitutively inactive Rac had no effect on these proc-esses (Supplementary Information, Fig. S7d), indicating that Cdc42 is another crucial regulator in these phenomena. Expression of Cdc42N17 or ROCKDN significantly suppressed the effect of RasV12 on cell height and intercellular F-actin accumulation (Supplementary Information, Fig. S7e, f). However, the effect of ROCKDN on cell height was much weaker than that of Cdc42N17, and vice versa, on F-actin accumulation, suggesting that the effects of these two proteins are mediated, at least partially, by different downstream pathways. Using a recombinant pro-tein for the Cdc42-binding domain (CRIB) of WASP (Supplementary Information, Fig. S6b), we found that the amount of GTP-bound active

Cdc42 was substantially increased in RasV12 cells surrounded by nor-mal cells (Fig. 4d and e). In contrast, it was significantly reduced in RasV12 cells surrounded by RasV12 cells (Fig. 4d, e)14. Taken together, these results indicate that Cdc42 is activated in a non-cell-autonomous manner in RasV12 cells surrounded by normal cells, which profoundly influences the fate of RasV12 cells.

Finally, we induced expression of E‑cadherin shRNA in the surround-ing normal cells, and examined the effect of knockdown of E-cadherin protein on the fate of RasV12 cells. When surrounded by E-cadherin-deficient cells, RasV12 cells produced large basal protrusions more fre-quently (Fig. 5a, c) and were more often basally delaminated from the monolayer (Fig. 5b, c) than when surrounded by normal cells (Fig. 5c). In contrast, apical extrusion of RasV12 cells was greatly reduced when they were surrounded by E-cadherin-deficient cells (Fig. 5c). No sig-nificant basal delamination or protrusion formation was observed when normal cells were surrounded by E-cadherin-deficient cells (Fig. 5c). These results suggest that the fate of RasV12 cells can be influenced by

a c

b

Basal delaminationProtrusion

Apical extrusion

GFP–RasV12 MDCK

+ E-cadherin shRNA1:100

GFP–RasV12

+ MDCK1:100

Cel

ls +

phe

noty

pes

(per

cent

age)

0

10

20

30

40

50

60

70

***

***

XZ

XY

XZ

d (A) RasV12

Basal protrusionsInvasion

(C)(B)Apical cellextrusion

Cdc42ROCK in

RasV12 cells

E-cadherin-basedcell–cell adhesions

in surrounding normal cells

Figure 5. E-cadherin-based intercellular adhesions of surrounding normal cells can influence the fate of RasV12 cells. (a, b) Confocal images of MDCK pTR GFP–RasV12 cells combined with MDCK cells expressing E-cadherin shRNA, on collagen gels. After 24 h of tetracycline addition, cells were stained with TRITC-phalloidin (red) and Hoechst (blue). Arrowheads and arrow indicate the basal protrusions and a basally delaminated RasV12 cell, respectively. Scale bars represent 20 μm. (c) Quantification of basal protrusion formation (black bar), basal delamination (grey bar) or apical extrusion (white bar) of GFP–RasV12 or MDCK cells surrounded by E-cadherin shRNA expressing MDCK (n = 301 and 396, respectively) or normal MDCK (n = 536) cells. Values represent

mean ± s.d.; *P < 0.05 **P < 0.005 ***P < 0.0005; n, cells. (d) Model showing that (A) when expression of RasV12 is induced in a single epithelial cell (green) within a monolayer of normal cells, two independent phenomena can occur in a non-cell-autonomous fashion. RasV12-expressing cells are either (B) apically extruded from the monolayer or (C) form basal protrusions beneath the surrounding neighbours. After apical extrusion, RasV12 cells are viable and form multicellular aggregates above the monolayer of normal cells. RasV12 cells that form basal protrusions frequently invade collagen. The fate of RasV12 cells is influenced by the activity of Cdc42 and ROCK in RasV12 cells and by E-cadherin-based cell–cell adhesions in the surrounding normal cells.



l e t t e r s

E-cadherin-based intercellular adhesions of surrounding normal cells. It is possible that basal invasion by Ras-transformed cells may be pro-moted under pathological conditions where E-cadherin-based cell–cell contacts are not properly formed in the surrounding epithelium, such as chronic inflammation or infection (see Supplementary Information Discussion on the molecular mechanisms of these processes and Fig. S8 for phenomena induced by other oncogenic stimuli).

In Drosophila, it has been reported that the interaction between nor-mal and transformed epithelial cells can cause several cellular processes, such as cell extrusion and apoptosis, that occur in a non-cell-autono-mous fashion15–18. However, the molecular mechanism whereby normal and transformed cells recognize differences between each other is not clearly understood. Our data also suggest that RasV12 cells recognize that they are surrounded by normal cells. There are two possible molecular mechanisms of cell recognition: cells detect differences either in physical properties of other cells or in the composition of molecules (for example, lipids and proteins) in plasma membranes. We have found that RasV12 cells have higher membrane elasticity and cell viscosity than normal cells (Supplementary discussion and Supplementary Information, Fig. S9). It remains to be determined whether these physical properties are impor-tant in initial cell recognition machineries and/or whether there are unidentified molecules that are involved in cell-recognition between normal and transformed cells.

In summary, when RasV12 is expressed in single cells in an epithe-lial monolayer, two independent phenomena can occur in a non-cell-autonomous manner (Fig. 5d). RasV12-expressing cells are either apically extruded from the monolayer, or form basal protrusions leading to basal invasion into the matrix. The fate of RasV12 cells is influenced by the activity of Cdc42 and ROCK in RasV12 cells and by E-cadherin-based cell–cell adhesions in the surrounding normal cells. Thus, RasV12 cells leave epithelial sheets either apically or basally in a cell-context-depend-ent manner. Our data and previous reports suggest that several cellular processes can occur at the interface between normal and transformed cells in Drosophila and vertebrates15–21. In future studies, it needs to be further clarified whether these processes indeed occur at the initial step of human carcinogenesis. These investigations may lead to a new preven-tive or therapeutic treatment for cancer.

METHODSAntibodies and materials. Rat anti-E-cadherin (ECCD2) and rabbit anti-ZO-1 antibodies were from Zymed. Mouse anti-β-catenin, mouse anti-E-cadherin and mouse anti-phospho-tyrosine antibodies were from BD Biosciences. Rat and mouse anti-E-cadherin antibodies were used for immunofluorescence and western blotting, respectively. Rat anti-α-tubulin and mouse anti-GFP antibod-ies were from Abcam and Roche Diagnostics, respectively. Rabbit anti-phospho-myosin light chain 2 (p-MLC; Thr 18/Ser 19) antibody was from Cell Signaling Technology, mouse anti-Pan-Ras antibody was from Calbiochem, Anti-GAPDH antibody was from Chemicon International and mouse anti-gp135 antibody was provided by G. K. Ojakian (SUNY Downstate Medical Center, NY) and used as an apical membrane marker. Mouse anti-β-galactosidase antibody was from Promega. Mouse anti-phospho-MAPK and rabbit anti-GST antibod-ies were from Sigma-Aldrich and Santa Cruz, respectively. Alexa-568- and Alexa-647-conjugated anti-rat, anti-mouse and anti-rabbit antibodies were from Invitrogen. All primary antibodies were used at a dilution of 1:100 for immunofluorescence, except anti-p-MLC antibody, which was used at 1:25, and anti-phospho-MAPK antibody, which was used at 1:50. All secondary antibodies were used at 1:200. TRITC-phalloidin (Sigma-Aldrich) was used at 1.5 μg ml–1. Hoechst (Invitrogen) was used at 1:5,000. For western blotting, antibodies were used as follows: anti-Pan-Ras at 1:1,000, anti-GFP and anti-E-cadherin at 1:2,000 and anti-GAPDH at 1:5,000.

The following inhibitors were used: 4-AP (2 mM, Sigma-Aldrich), Ethidium homodimer-1 (Ethd-1, 1 μM, Invitrogen), (S)-(–)-blebbistatin (50 μM, Toronto Research Chemicals), U0126 (10 μM, Promega), LY294002 and Y27632 (10 μM, Calbiochem) and cytochalasin D (100 nM, Sigma-Aldrich). All inhibitors were added at the beginning of the experiments for 16–24 h incubation. DMSO (Sigma-Aldrich) was added at a dilution of 1:1,000 as a control.

Time-lapse microscopy. MDCK cells stably expressing GFP–RasV12 or GFP–Rap1V12 were fluorescently labelled using CellTracker dye CMPTX (Invitrogen) according to the manufacturer’s instructions. Labelled cells were then trypsinized and combined with MDCK cells at a ratio of 1:100. Cells were plated at a density of 1 × 106 cells per well in 35 mm glass-bottom culture dishes (MatTek Corporation) for all experiments. Mixed cells were incubated for 8 h at 37 oC before being transferred to a tetracycline-containing medium. Where indicated, cells were analysed in the presence of 4-AP. To obtain time-lapse images, we used a Zeiss Axiovert 200 M microscope with a Ludl Electronic Products Biopoint Controller and a Hamamatsu C4742-95 Orca camera (Hamamatsu). Images were captured and analysed using Openlab or Volocity software (Improvision).

Collagen assays. Type-I collagen was obtained from R&D Systems (Cultrex Rat Collagen I) or Nitta Gelatin (Nitta Cellmatrix type 1-A), and was neutralized on ice to a final concentration of 2 mg ml–1 according to the manufacturer’s instruc-tions. Glass coverslips in 6-well culture dishes were coated with 1 ml of neutralized collagen and allowed to solidify for 30 min at 37 oC. For time-lapse experi-ments, 35 mm glass-bottom dishes were pre-coated with a thin layer (approxi-mately 200 μl) of neutralized collagen. For each assay, between 5 x105 and 1 x106 cells were plated per well onto the collagen gel. RasV12 cells were combined with MDCK cells at a ratio of 1:100. After incubation for 8–16 h at 37 oC, tetracycline was added to induce RasV12 expression. A control (absence of tetracycline) was also included for each experiment. Cells were incubated for the indicated times, and tetracycline was replaced every 48 h.

For immunostaining, cells on collagen were fixed with 4% paraformaldehyde/PBS for 15 min at 37 oC, washed three times in PBS and permeabilized in 0.5% Triton X-100/PBS for 20 min. Cells were then washed once in PBS followed by three 10-min washes in glycine wash buffer (7 mM Na2HPO4, 3.5 mM NaH2PO4, 130 mM NaCl and 100 mM glycine). Cells were blocked for 1 h in blocking buffer (7 mM Na2HPO4, 3.5 mM NaH2PO4, 130 mM NaCl, 0.2% Triton X-100, 0.05% Tween-20, 10% FCS, 0.02% BSA and 7.7 mM NaN3), and incubated with primary antibodies for 16 h at 4 oC. This was followed by three 10-min washes with gentle agitation in blocking buffer before incubation with Alexa-568- and Alexa-647-conjugated secondary antibodies and/or TRITC-phalloidin for 2–3 h at room temperature. Cells were then washed three times in blocking buffer and were incubated with Hoechst/PBS for 2 min, followed by washing in PBS and mounting onto Mowiol on a glass slide.

Cdc42 activation assay. The level of active GTP-bound Cdc42 in cells was deter-mined using a GST fusion protein containing the Cdc42-interacting domain of WASP (WASP-CRIB)22. GST or GST–WASP-CRIB protein was produced as described previously22,23. After fixation, permeabilization and blocking as described above, cells were incubated with 100 ng ml–1 of GST or GST–WASP-CRIB protein in blocking buffer for 1 h at room temperature, followed by three 10-min washes with gentle agitation in blocking buffer before incubation with anti-GST antibody for 16 h at 4 oC. After three 10-min washes with gentle agita-tion in blocking buffer, cells were incubated with Alexa-647-conjugated second-ary antibody for 2 h at room temperature. Cells were then washed three times in blocking buffer, incubated with Hoechst/PBS for 2 min and mounted onto Mowiol on a glass slide.

Generation and analyses of Ras overexpressing clones in Drosophila wing imaginal discs. Experimental genotypes were: RasV12: hs-FLP/+; tub>>Gal4 UAS–lacZ/UAS–rasV12..RasN17: hs-FLP/UAS–rasN17; tub>>Gal4 UAS–lacZ/+. Raswt: hs-FLP/+; tub>>Gal4 UAS–lacZ/UAS–raswt. Clones were induced by heat-shocking larvae three days after egg laying (AEL) for 30 min at 37 °C. Larvae were dissected 68–72 h after clone induction and fixed in 4% paraformaldehyde/PBS. Dissected hemilarvae were permeabilized in 0.05% TritonX-100/PBS (PBT) and incubated for 30 min in 4% FCS/PBT. Samples were incubated with a primary antibody (mouse anti β-galactosidase) for 16 h at 4 oC, and then with Alexa-488-conjugated secondary antibody in 4% FCS/PBT for 2 h at room temperature. After



l e t t e r s

each incubation with an antibody, samples were extensively washed in PBT. Nuclei were counterstained with Hoechst/PBS for 15 min at room temperature. Imaging was performed on a confocal Leica SPE acquiring xz sections every 0.75–1.0 μmin. To induce the expression of RasV12 in the entire wing disc, we drove the expression of UAS–RasV12 under the control of nubbin–Gal4 at 25 oC.

Data analyses. All statistical analyses were carried out using Student’s t tests. Detailed statistical analyses are described in Supplementary Information Methods.


ACKNowLEDgEMENtSWe thank G. K. Ojakian for the anti-gp135 antibody, A. Hall, A. Lloyd, R. Y. Tsien, S. Lowe and E. Sahai for constructs, and A. Vaughan for technical assistance with microscopes. We also thank Y. Morishita for discussion on physical forces at cell–cell adhesions. S.D-C. was supported by a FEBS Long Term Fellowship. A.E.P. acknowledges the Interdisciplinary Research Collaboration (IRC) in Nanotechnology (Cambridge, EPSRC UK) and the Dr Mortimer and Theresa Sackler Trust for financial support. This work is supported by MRC funding of the Cell Biology Unit.

AutHor CoNtrIButIoNSC.H. designed the experiments and generated most of the data; S.D-C. established stable MDCK cell lines and performed statistical analyses (Fig. 1d); M.N. analysed clonal expression of RasV12, RasN17 and RasWT in Drosophila wing imaginal discs (Fig. 1e and Supplementary Information, Fig. S3); M.K. performed western blot analyses (Supplementary Information, Fig. S1b), immunofluorescence studies (Supplementary Information, Fig. S4a) and established stable MDCK cell lines; C.Z. performed western blot and time-lapse analyses (Supplementary Information, Figs S2a, d and S5e), and established stable MDCK cell lines; A.E.P. performed AFM experiments and analyses (Supplementary Information, Fig. S9); E.P., L.A.B-L. and J-P.V. analysed clonal expression of RasV12, RasN17 and RasWT in Drosophila wing imaginal discs (Fig. 1e and Supplementary Information, Fig. S3); Y.I. provided technical expertise on use of collagen; H.H. provided technical expertise on myosin-II; F.P. assisted with Drosophila experiments; Y.F. conceived and designed the study and acted as principal investigator.

CoMPEtINg FINANCIAL INtErEStSThe authors declare no competing financial interests.

Published online at http://www.nature.com/naturecellbiology/ reprints and permissions information is available online at http://npg.nature.com/reprintsandpermissions/

1. Hanahan, D. & Weinberg, R. A. The hallmarks of cancer. Cell 100, 57–70 (2000).2. Fialkow, P. J. Clonal origin of human tumors. Biochim. Biophys. Acta. 458, 283–321

(1976).

3. Nowell, P. C. The clonal evolution of tumor cell populations. Science 194, 23–28 (1976).

4. Schoenenberger, C. A., Zuk, A., Kendall, D. & Matlin, K. S. Multilayering and loss of apical polarity in MDCK cells transformed with viral K-ras. J. Cell Biol. 112, 873–889 (1991).

5. Rosenblatt, J., Raff, M. C. & Cramer, L. P. An epithelial cell destined for apoptosis signals its neighbors to extrude it by an actin- and myosin-dependent mechanism. Curr. Biol. 11, 1847–1857 (2001).

6. Bos, J. L. Ras-like GTPases. Biochim. Biophys. Acta. 1333, M19–31 (1997).7. Baena-Lopez, L. A., Pastor-Pareja, J. C. & Resino, J. Wg and Egfr signalling antagonise

the development of the peripodial epithelium in Drosophila wing discs. Development 130, 6497–6506 (2003).

8. Rodriguez-Viciana, P. et al. Role of phosphoinositide 3-OH kinase in cell transformation and control of the actin cytoskeleton by Ras. Cell 89, 457–467 (1997).

9. Cantrell, D. A. Phosphoinositide 3-kinase signalling pathways. J. Cell Sci. 114, 1439–1445 (2001).

10. Karnoub, A. E. & Weinberg, R. A. Ras oncogenes: split personalities. Nature Rev. Mol. Cell Biol. 9, 517–531 (2008).

11. Kimura, K. et al. Regulation of myosin phosphatase by Rho and Rho-associated kinase (Rho-kinase). Science 273, 245–248 (1996).

12. Jaffe, A. B. & Hall, A. Rho GTPases: biochemistry and biology. Annu. Rev. Cell Dev. Biol. 21, 247–269 (2005).

13. Vega, F. M. & Ridley, A. J. Rho GTPases in cancer cell biology. FEBS Lett. 582, 2093–2101 (2008).

14. Sahai, E., Olson, M. F. & Marshall, C. J. Cross-talk between Ras and Rho signalling pathways in transformation favours proliferation and increased motility. EMBO J. 20, 755–766 (2001).

15. de la Cova, C., Abril, M., Bellosta, P., Gallant, P. & Johnston, L. A. Drosophila Myc regulates organ size by inducing cell competition. Cell 117, 107–116 (2004).

16. Moreno, E. & Basler, K. dMyc transforms cells into super-competitors. Cell 117, 117–129 (2004).

17. Vidal, M., Larson, D. E. & Cagan, R. L. Csk-deficient boundary cells are eliminated from normal Drosophila epithelia by exclusion, migration, and apoptosis. Dev. Cell 10, 33–44 (2006).

18. Brumby, A. M. & Richardson, H. E. scribble mutants cooperate with oncogenic Ras or Notch to cause neoplastic overgrowth in Drosophila. EMBO J. 22, 5769–5779 (2003).

19. Stoker, M. G., Shearer, M. & O’Neill, C. Growth inhibition of polyoma-transformed cells by contact with static normal fibroblasts. J. Cell Sci. 1, 297–310 (1966).

20. Bignami, M., Rosa, S., La Rocca, S. A., Falcone, G. & Tato, F. Differential influence of adjacent normal cells on the proliferation of mammalian cells transformed by the viral oncogenes myc, ras and src. Oncogene 2, 509–514 (1988).

21. Alexander, D. B. et al. Normal cells control the growth of neighboring transformed cells independent of gap junctional communication and SRC activity. Cancer Res. 64, 1347–1358 (2004).

22. Wells, C. M., Walmsley, M., Ooi, S., Tybulewicz, V. & Ridley, A. J. Rac1-deficient mac-rophages exhibit defects in cell spreading and membrane ruffling but not migration. J. Cell Sci. 117, 1259–1268 (2004).

23. Hogan, C. et al. Rap1 regulates the formation of E-cadherin-based cell-cell contacts. Mol. Cell. Biol. 24, 6690–6700 (2004).






L E T T E R S

Distinct regulation of autophagic activity by Atg14L and Rubicon associated with Beclin 1–phosphatidylinositol-3-kinase complexYun Zhong2,5, Qing Jun Wang1,3,5, Xianting Li1, Ying Yan4, Jonathan M. Backer4, Brian T. Chait3, Nathaniel Heintz2 and Zhenyu Yue1,6

Beclin 1, a mammalian autophagy protein that has been implicated in development, tumour suppression, neurodegeneration and cell death, exists in a complex with Vps34, the class III phosphatidylinositol-3-kinase (PI(3)K) that mediates multiple vesicle-trafficking processes including endocytosis and autophagy. However, the precise role of the Beclin 1–Vps34 complex in autophagy regulation remains to be elucidated. Combining mouse genetics and biochemistry, we have identified a large in vivo Beclin 1 complex containing the known proteins Vps34, p150/Vps15 and UVRAG, as well as two newly identified proteins, Atg14L (yeast Atg14-like) and Rubicon (RUN domain and cysteine-rich domain containing, Beclin 1-interacting protein). Characterization of the new proteins revealed that Atg14L enhances Vps34 lipid kinase activity and upregulates autophagy, whereas Rubicon reduces Vps34 activity and downregulates autophagy. We show that Beclin 1 and Atg14L synergistically promote the formation of double-membraned organelles that are associated with Atg5 and Atg12, whereas forced expression of Rubicon results in aberrant late endosomal/lysosomal structures and impaired autophagosome maturation. We hypothesize that by forming distinct protein complexes, Beclin 1 and its binding proteins orchestrate the precise function of the class III PI(3)K in regulating autophagy at multiple steps.

Macroautophagy (herein referred to as autophagy) is a regulated process by which a portion of the cytoplasm is sequestered and delivered to lysosomes for degradation. Currently, the autophagic process in mammals is poorly understood. Identification and characterization of mammalian autophagy proteins are crucial to elucidate details of mammalian autophagy. Beclin 1 (encoded by Becn1, the orthologue of yeast ATG6/Vps30), is one of the earliest characterized mammalian autophagy proteins1. Initially identi‑fied as a Bcl‑2‑binding protein2, Beclin 1 has been shown both in vitro

and in vivo to participate in autophagy regulation and to have important roles in development3, tumorigenesis1,3–5 and neurodegeneration6–8. As with yeast Atg6, Beclin 1 forms a complex with Vps34/class III PI(3)K9–11. Saccharomyces cerevisiae has at least two Atg6/Vps34 protein complexes: one containing Atg14 and participating in autophagy, and the other con‑taining Vps38 and functioning in non‑autophagic pathways10. However, there is no concrete evidence for multiple Beclin 1–Vps34 complexes or multiple functions associated with Beclin 1–Vps34 in mammals11.

To determine the mechanism by which the Beclin 1–Vps34 interac‑tion regulates autophagy, we combined mouse genetics and biochemistry to identify Beclin 1‑associated protein complexes in vivo. We geneti‑cally modified mice to functionally replace endogenous Beclin 1 with an enhanced green fluorescent protein‑tagged Beclin 1 protein (Beclin 1–EGFP; Fig. 1a; Supplementary Information, Fig. S1a–b). In these mice (Becn1–/–;Becn1–EGFP/+), only the Beclin 1–EGFP fusion protein, but not endogenous Beclin 1 was detected by an anti‑Beclin 1 antibody (Fig. 1a). These mice were born at the expected Mendelian ratio (Supplementary Information, Fig. S1c), survived postnatally and were phenotypically nor‑mal at the adult stage, suggesting a full ‘rescue’ of the embryonic lethality of Becn1–/– mice by a functional Becn1–EGFP transgene.

Using these ‘rescued’ mice, we isolated Beclin 1–EGFP protein com‑plexes by affinity purification from liver, brain (Fig. 1b) and thymus (data not shown), and identified their components using mass‑spec‑trometry (Supplementary Information, Fig. S2). The rescued mice, but not the control mice, were associated with at least six readily detect‑able protein bands common to both liver and brain (Fig. 1b). These bands include Beclin 1–EGFP (no. 5; molecular mass of about 90,000), three previously reported Beclin 1‑binding proteins p150/Vps15 (no. 1; ref. 12), Vps34 (no. 3; refs 9–11) and UVRAG (no. 4; ref. 13), and two newly identified proteins (nos 2 and 6). The first of these two proteins (no. 6, ~60K, gi|27369860) contains 492 amino acids and has a conserved SMC (structural maintenance of chromosomes) motif

1Departments of Neurology and Neuroscience, Mount Sinai School of Medicine, New York, NY 10029, USA. 2Laboratory of Molecular Biology, Howard Hughes Medical Institute, Rockefeller University, New York, NY 10065, USA. 3Laboratory of Mass Spectrometry and Gaseous Ion Chemistry, Rockefeller University, New York, NY 10065, USA. 4Department of Molecular Pharmacology, Albert Einstein College of Medicine, 1300 Morris Park Avenue, Bronx, NY 10461, USA.5These authors contributed equally to the work.6Correspondence should be addressed to Z.Y. (e-mail: [email protected])

Received 17 November 2008; accepted 3 February 2009; published online 8 March 2009; DOI: 10.1038/ncb1854




L E T T E R S

or two coiled‑coil domains (CCD; amino acids 75–95 and 148–178) near the amino terminus (Fig. 1c). Interestingly, the sequence of this protein shows modest similarity to yeast Atg14 (overall 15% identity; Supplementary Information, Fig. S3a). Thus, we named this pro‑tein Atg14L for yeast Atg14‑like. The second protein (no. 2, ~124K, gi|45708948) contains 941 amino acids and has a conserved RUN domain (amino acids 49–190) near the amino terminus, a cysteine‑rich domain (amino acids 837–890) near the carboxy terminus and a CCD (amino acids 488–508) in the central region (Fig. 1c). Thus, we named this protein Rubicon for RUN domain, a cysteine‑rich domain contain‑ing, Beclin 1‑interacting protein. No sequence homology was observed between Rubicon and Vps38 or Atg14 (data not shown). Notably, the protein levels of affinity‑purified Beclin 1–EGFP, p150/Vps15, Vps34 and UVRAG were comparable and reproducibly higher than those of Atg14L and Rubicon (Fig. 1b), suggesting a stable ‘core’ Beclin 1–Vps34 complex consisting of Beclin 1, Vps34, p150 and UVRAG. Additionally, we did not detect the previously identified Beclin 1‑associated proteins,

such as nPIST6, Bcl‑2 (ref. 2), Ambra‑1 (ref. 14) or Bif1 (ref. 15), raising the possibility that their interactions with Beclin 1 may be relatively unstable, transient or occur only under specific conditions.

We next examined the specific binding of Atg14L or Rubicon to Beclin 1 in transfected mammalian cells. We showed that Flag‑ or EGFP‑tagged Atg14L or Rubicon co‑immunoprecipitated with endogenous Beclin 1 (Supplementary Information, Fig. S3b–c) and Vps34 (Supplementary Information, Fig. S3d–e). We also constructed a series of deletion mutants to analyse the sequence domains required for Beclin 1–Atg14L/Rubicon associations (Supplementary Information, Fig. S3f). We found that although the CCD of Beclin 1 is sufficient for binding Atg14L, the CCD and evolutionarily conserved domain (ECD) of Beclin 1 are necessary for binding Rubicon (Supplementary Information, Fig. S3g–h). Furthermore, both CCD domains of Atg14L are required for efficient binding of Atg14L to Beclin 1 and Vps34 (Supplementary Information, Fig. S3i), whereas the central region of Rubicon, which contains the CCD, is important for the binding of Rubicon to Beclin 1 and Vps34 (Supplementary Information,

Cys-richRubicon

Atg14L

SMCa b

Con

trol

Res

cued

Con

trol

Res

cued

Brain

2 43

220160120100

90807060

50

40

30

2520

1510

12*345

6*

d

Beclin 1–EGFP

Beclin 1

c

e

1 2

f

IP

IPWCL

WCL

IP

IP

WCL

WCL

IPWCLWCL

IPWCLWCL

pEG

FP–N

3 +

Flag

–Rub

icon

pEG

FP–N

3 +

Flag

–Atg

14L

Rub

icon

–EG

FP +

Fla

g–At

g14L

Atg1

4L–E

GFP

+ F

lag–

Rub

icon

Anti-Flag

Anti-GFP

Anti-Beclin 1

Anti-actin

Anti-Flag

Anti-GFP

Anti-Beclin 1

Anti-actin

IP: Anti-GFPIP: Anti-GFP

Atg14L

Rubicon

Vps34

Beclin 1

80

60

60

50

50

120100

120100

80

8080

Mr(K)

670K 158K

Size of complexes

1

Mr(K)

Mr(K)

1 49

1 75 95 148 178 492

190 488 508 837 890 941

CCD1 CCD2

RUN CCD

Liver

51 52 5354 55 56 57 58 59 60 61 62 63 643738 3940 41 42 43 44 45 46 47 48 49 50

Bec

n1+/

+B

ecn1

+/– ; B

ecn1

–EG

FP/+

Bec

n1–/

– ; Bec

n1–E

GFP

/+

75

50

Figure 1 Identification of Beclin 1-interaction proteins from Becn1–/–;Becn1–EGFP/+ mice. (a) Western blot analysis showing the replacement of endogenous Beclin 1 with Beclin 1–EGFP in Becn1–/–;Becn1–EGFP/+ mice, as detected by an anti-Beclin 1 antibody. (b) Coomassie-stained SDS–PAGE showing the Beclin 1-interacting proteins immuno-isolated from brain and liver of the ‘rescued’ mice (lanes 2 and 4) and of control Becn1+/– littermates (lanes 1 and 3), using an anti-GFP antibody. Proteins in the gel bands were extracted and identified by mass spectrometry as Vps15/p150 (band 1), Vps34/class III PI(3)K (band 3), UVRAG (band 4), Beclin 1–EGFP (band 5), Atg14L (band 6, asterisk, gi|27369860) and Rubicon (band 2, asterisk, gi|45708948). UVRAG levels varied with different affinity-purification conditions, suggesting an unstable association of UVRAG with the complex. (c) Schematic representations of the domain structures of Atg14L and Rubicon. Atg14L contains two coiled-coil domains (CCD1 and CCD2), which are also homologous with the SMC domain (structural maintenance of chromosomes). Rubicon contains an N-terminal RUN (for RPIP8, UNC-14 and NESCA) domain,

a C-terminal Cys-rich domain and a central CCD domain. (d) Western blot analysis of Atg14L, Rubicon, Vps34 and Beclin 1 in gel filtration fractions from wild-type mouse liver extract showed co-elution of these proteins in fractions 38–45. Atg14L was also eluted in later fractions 51–56. The fractions for the peak elution of thyroglobulin (670K) and γ-globulin (158K) are indicated by arrows. Control siRNA-transfected NIH 3T3 cell lysate was loaded as a positive control (labelled as 1) for the migration position of the Atg14L protein on SDS–PAGE; Atg14L siRNA-transfected NIH 3T3 cell lysate was loaded as a negative control (labelled as 2). (e, f) Co-immunoprecipitation confirmed protein–protein interaction between Atg14L and Rubicon. HEK 293 cells were co-transfected with Atg14L–EGFP and Flag–Rubicon (e) or Rubicon–EGFP and Flag–Atg14L (f). Cell lysates were used for immunoprecipitation with an anti-GFP antibody and the resulting immunoprecipitates were blotted with an anti-Flag antibody. Our results show immunoprecipitation of Rubicon by Atg14L (e) and vice versa (f). WCL, whole cell lysate; IP, immunoprecipitated. See Supplementary Information, Fig. S6 for full scans of blots in a, d, e and f.



L E T T E R S

a

Atg14L

p62Beclin1

LC3 ILC3 IIActin

Atg14L

p62

Beclin1

LC3 ILC3 IIActin

b

c d

*

0.90

Untreated Starved Starved +3MA

2.73

1.71

0.83 0.771.00

Myc–Vps34–Vps15

Myc–Vps34–Vps15

Beclin 1–EGFP

+ + + +– – + + Beclin 1–EGFP

Anti-Myc (IP)

PI(3)P

*

2.5

e

Atg14L–EGFP

Atg14L–As-Red

Beclin 1–AsRed Merge

Merge

*

*

*

*

N

MM M M

M

*

N

M

M MM

M

M

1 4

*

EGFP–Atg12 Beclin 1–Myc

Atg14L–As-Red MergeEGFP–Atg12 Beclin 1–Myc

f

*

g

Con

trol

siR

NA

Bec

lin s

iRN

A

Atg

14L

siR

NA

Con

trol

siR

NA

Bec

lin s

iRN

A

Atg

14L

siR

NA

No starvation 2 h starvationFl

ag

Flag

–Atg

14L

Flag

Flag

-Atg

14L

Flag

Flag

-Atg

14L

Flag

Flag

–Atg

14L

Deg

rad

atio

n of

long

-liv

edp

rote

ins

(nor

mal

ized

)

PI(3

)P fo

ld c

hang

e

3

2

1

0

3

2

1

0

Control siRNA

Atg14L siRNA

++++

++––

2 3

Figure 2 Atg14L positively regulates autophagy, and Beclin 1 and Atg14L synergistically promote double-membrane formation. (a) Beclin 1 or Atg14L siRNA reduced Atg14L levels and increased p62/SQSTM1 and LC3 II levels under normal and nutrient-starvation conditions in NIH 3T3 cells. (b) Compared with control siRNA, Atg14L siRNA decreased long-lived protein degradation in NIH 3T3 cells under normal (*P = 0.007) and starvation (*P = 5 × 10–6) conditions (one-tailed Student’s t-test with equal variances, n = 4). This difference was diminished when the starved cells were treated with 3-methyladenine (3MA, 10 mM), a PI(3)K inhibitor. (c) Vps34 kinase assay. HEK 293T cells were co-transfected with Myc–Vps34–Vps15 and Flag–Atg14L or Flag vector, either in the absence or in the presence of Beclin 1–EGFP. Myc–Vps34–Vps15 was immunoprecipitated by anti-Myc antibody for the in vitro kinase assay. The resulting radioactive PI(3)P was separated by thin-layer chromatography (TLC), quantified and normalized against the amount of immunoprecipitated Myc-tagged Vps34 as measured by western blot (upper panel). The quantified results (lower panel) show that overexpressing Atg14L significantly upregulated Vps34 kinase activity by 2.5-fold, but only when Beclin

1 was also overexpressed (*P = 0.04, one-tailed Student’s t-test with unequal variances, n = 5). (d) Colocalization of co-expressed Atg14L–EGFP (green) and Beclin 1–AsRed (red) in punctate structures in transiently transfected HeLa cells. Scale bar, 10 μm. (e) Electron microscopy images show large structures (asterisks) that are often enwrapped with double membranes in the HEK 293T cells co-transfected with Atg14L–EGFP and Beclin 1–AsRed: concentric membrane ‘rings’ (panel 1); two large structures (3–5 μm in diameter, panel 2) containing material with high electron density (inset, enwrapping double membranes); numerous autophagosomes (arrows, panel 3) in the cytoplasm; immuno-electron microscopy image of a Atg14L–Beclin 1 structure (labelled with anti-GFP antibody and developed by DAB, panel 4) enwrapped with concentric membrane ‘rings’. M, mitochondria; N, nucleus. Scale bar, 500 nm. (f, g) EGFP–Atg12 (f) or EGFP–Atg5 (g) (green) was colocalized with the large structures (arrows) that were labelled by Atg14L–AsRed (red) and Beclin 1–Myc (blue) in transfected HeLa cells. Some of these structures seemed to be ‘ring’-shaped (yellow arrows and inset). Scale bar, 10 μm. See Supplementary Information, Fig. S6 for full scans of blots in a and c.



L E T T E R S

Fig. S3j). Interestingly, the RUN or Cys‑rich domain of Rubicon seemed to inhibit the binding of Rubicon to Beclin 1 and Vps34 (Supplementary Information, Fig. S3j).

We then characterized the composition of the Beclin 1 complexes. Using anti‑Atg14L and anti‑Rubicon antibodies, we found that Atg14L co‑immunoprecipitated with Vps34 (Fig. S4a) and Beclin 1 (data not shown) but not with Rubicon (Supplementary Information, Fig. S4a). Rubicon co‑immunoprecipitated with Vps34 and UVRAG, but not with

Atg14L (Supplementary Information, Fig. S4a). Therefore, Atg14L and Rubicon seem to exist in separate Beclin 1 complexes.

We performed gel filtration experiments with tissue extracts pre‑pared from either wild‑type (Fig. 1d) or ‘rescued’ (Supplementary Information, Fig. S4b) mouse liver. For each sample, eighty fractions of the eluent were collected and analysed by immunoblotting. We found that the endogenous Vps34, Beclin 1 (or Beclin 1–EGFP), Atg14L and Rubicon proteins were primarily co‑eluted in fractions 38–45 (Fig. 1d;

*0.58

a b c

Rubicon

p62/SQSTM1

d

p62/SQSTM1Actin

Actin

EG

FP

Rub

icon

–EG

FP

p62/SQSTM1

EG

FP

Rub

icon

–EG

FP

293 stable cells

293T transienttransfection

e

*

No starvation

Flag

Flag

–Rub

icon

Flag

Flag

–Rub

icon

Myc–Vps34–Vps15Beclin 1–EGFP

Anti-Myc (IP)

PI(3)P

PI(3

)P fo

ld c

hang

e

+ + + +– – + +

LC3 ILC3 IIActin

Con

trol

siR

NA

Rub

icon

siR

NA

Con

trol

siR

NA

Rub

icon

siR

NA

2 h starvationNo starvation

2 h starvationNo starvation

Myc–Vps34–Vps15Beclin 1–EGFP

Flag

Flag

–Rub

icon

Flag

Flag

–Rub

icon

Flag

Flag

–Rub

icon

+ + + +– – + +

1.0

0.5

0.0

50

40

30

20

10

0R

ed-o

nly

pun

cta

(per

cent

age)

MergemCherryGFP

MergemCherryGFP

Flag

Flag–Rubicon

Figure 3 Rubicon is a negative regulator of autophagy. (a) Rubicon siRNA treatment of the NIH 3T3 cells led to decreased levels of p62 and LC3 II under both normal and nutrient-starvation conditions. (b) Overexpression of Rubicon resulted in increased levels of p62 under both normal and nutrient-starved conditions in HEK 293 cells either stably expressing (upper rows) or transiently transfected with (lower rows) Rubicon–EGFP. The control cells were either stably expressing or transiently transfected with the EGFP–N3 vector. (c) Vps34 kinase activity. HEK 293T cells were co-transfected with Myc–Vps34–Vps15 and Flag–Rubicon or Flag vector, either in the absence or in the presence of Beclin 1–EGFP. Myc–Vps34–Vps15 was immunoprecipitated by an anti-Myc antibody and used for the in vitro kinase assay. The resulting radioactive PI(3)P was separated by TLC, quantified and normalized against the amount of immunoprecipitated Myc-tagged Vps34, as measured by western blotting (upper panel). The quantified results (lower panel) show that overexpressing Rubicon significantly downregulated the Vps34 kinase activity to 0.58-fold, but only without Beclin 1 overexpression (*P = 0.04, one-tailed Student’s t-test with unequal variances, n = 4). (d) Effect of overexpressing Flag–Rubicon on autophagosome acidification, as

monitored by mCherry–GFP–LC3 fluorescence. HeLa cells were transiently co-transfected with mCherry–GFP–LC3 and Flag–Rubicon (or control Flag vector). Cells co-expressing mCherry–GFP–LC3 and control Flag vector contained many red-only puncta along with yellow (indicating the presence of both red and green) puncta, suggesting the presence of both autolysosomes and nascent autophagosomes (upper panel). In contrast, cells co-expressing mCherry–GFP–LC3 and Flag–Rubicon contained primarily yellow or white puncta, suggesting the presence of only nascent autophagosomes (lower panel, white arrows). Notably, some cells, which were co-transfected with mCherry–GFP–LC3 and Flag–Rubicon but expressed high levels of mCherry–GFP–LC3 and undetectable levels of FLAG–Rubicon, contained many red-only puncta (lower panel, yellow arrows). (e) Quantification of the results in d show that overexpressing Flag–Rubicon markedly reduced the percentage of red-only puncta (mCherry–LC3) from 39% in the control Flag vector-transfected cells to 2% in the Flag–Rubicon-transfected cells (*P = 2 × 10–26, one-tailed Student’s t-test with unequal variances, n = 30), indicating that overexpression of Rubicon blocks autophagosome acidification or maturation. See Supplementary Information, Fig. S6 for full scans of blots in a–c.



L E T T E R S

Supplementary Information, Fig. S4b), suggesting that these fractions contain a major Beclin 1–Vps34 complex (>700K) that includes both Atg14L and Rubicon. We also performed gel filtration experiments with cell lysates prepared from stable cell lines expressing either Atg14L–EGFP

(Supplementary Information, Fig. S4c) or Rubicon–EGFP (Supplementary Information, Fig. S4d). Again, endogenous Beclin 1 was co‑eluted with Atg14L–EGFP (Supplementary Information, Fig. S4c) and Rubicon–EGFP (Supplementary Information, Fig. S4d). Interestingly, starvation of these

a

b

c

d

Rubicon–EGFP Lamp1 Merge

Rubicon–EGFP LBPA Merge

M

M

Anti-GFP DAB and Anti-Lamp1 Au-enhanced

Anti-GFPNo antibody

1 2 3 4

5 6 7

1 2 3 4

Figure 4 Overexpressing Rubicon causes aberrant expansion of late endosomes/lysosomes. (a) Colocalization of Rubicon–EGFP-associated structures with the late endosome/lysosome marker Lamp1 (arrows) in HeLa cells transfected with Rubicon–EGFP. Note that some of the Rubicon–EGFP-associated structures show a ‘ring’ shape (yellow arrows). Scale bar, 10 μm. (b) Partial colocalization of Rubicon–EGFP-associated structures with the MVB marker LBPA (arrows) in HeLa cells transfected with Rubicon–EGFP. Scale bar, 10 μm. (c) Representative ultrastructural images show aberrant expansion of late endosomal/lysosomal structures in HEK 293T cells overexpressing Rubicon–EGFP. These abnormal organelles are large in

size, with high (orange arrows) or low (black arrows) electron density. Some enclose small vesicles (purple arrows) and some resemble the MVB (blue arrows). Scale bars, 500 nm. (d) Representative ultrastructural images show late endosome/lysosome-like structures that are labelled with anti-GFP gold particles (panels 3, 4) in HEK 293T cells transiently transfected with Rubicon–EGFP. These structures are enwrapped by double membranes (panel 4 inset) and co-labelled by anti-GFP (developed by DAB) and anti-Lamp1 (gold enhanced) (panels 5–7) antibodies. Note that mitochondria are mostly negative for Rubicon–EGFP (panel 4). The negative controls are without antibody (panels 1–2). M, mitochondria. Scale bars, 200 nm.



L E T T E R S

a

b

c

Rubicon–EGFP Rubicon–EGFP

Rubicon∆RUN–EGFPRubicon–EGFP Rubicon∆C–EGFP Rubicon∆RUN∆C–EGFP

Merge

Merge Mergee Flag–Beclin1–CEBeclin1–AsRedd

g

Inhibition of autophagosomematuration

Beclin1

Actin

Contro

l siR

NABec

lin 1

siRNA

Control siRNA Beclin 1 siRNA

fRubicon–AsRedp40 (phox)-PX–EGFP

No

trea

tmen

tW

ortm

anni

n

Autophagosomebiogenesis

Atg14L–Beclin 1 containing complex Rubicon-containing complexBeclin 1–Vps34–Vps15–UVRAG–Atg14L core complex is associated with Rubicon

Beclin1 Beclin1

Vps34 Vps34UVRAG

Rubicon

Vps34

Rubicon

Vps15

Atg14L Atg14L

Figure 5 Overexpressed Rubicon is localized on PI(3)P-enriched structures in a Beclin 1-independent manner. (a) Local sequence alignment between the C-terminal Cys-rich domain of Rubicon and FYVE domains of several known FYVE-containing proteins. Rubicon does not possess the key consensus sequences of a typical FYVE domain, that is, N-terminal WxxD, central R[R/K]HHCR and C-terminal RVC (indicated by red bars). (b) Colocalization of the PI(3)P-enriched lipid domain marker p40 (phox)-PX–EGFP (green) and Rubicon–AsRed (red) on large punctate structures (arrows) in the co-transfected HeLa cells (upper panels). Treatment with the PI(3)K inhibitor wortmannin (75 nM) for 1 h caused disappearance of the PI(3)P-enriched lipid domains, whereas the Rubicon–AsRed-positive structures were maintained (lower panels). Scale bars, 10 μm. (c) Subcellular localization of transiently transfected Rubicon–EGFP, RubiconΔRUN–EGFP, RubiconΔC–EGFP or RubiconΔRUNΔC–EGFP in HeLa cells. In contrast to punctate Rubicon–EGFP and RubiconΔRUN–EGFP, RubiconΔC–EGFP and RubiconΔRUNΔC–EGFP were dispersed in the cytoplasm. ΔRUN, RUN domain deletion; ΔC, cysteine-rich domain deletion. Scale bars, 10 μm. (d, e) Absence of full-length Beclin 1 (d) or Beclin 1-CE mutant (e) (red) on the Rubicon–EGFP-positive structures (green) in

HEK 293 cells stably expressing Rubicon–EGFP. These cells were transiently transfected with either Beclin 1–AsRed (d) or Flag–Beclin 1-CE (e; that is, the Flag-tagged Beclin 1 mutant containing both CCD and ECD, which mediate the Beclin 1–Rubicon interaction as shown in Supplementary Information, Fig. S3h). Scale bars, 10 μm. (f) The formation of the Rubicon–EGFP-positive structures was not affected by siRNA knockdown of Beclin 1 in HEK 293 cells stably expressing Rubicon–EGFP. (g) A model for the Beclin 1–Vps34 protein complexes and their functions. Note that this model is not intended to propose a direct binary interaction. In this model, a core Beclin 1 complex is composed of Vps34/PI(3)K, p150/Vps15, Beclin 1, UVRAG and probably substoichiometric Atg14L (indicated by the tight binding and functional connection between Atg14L and Beclin 1). Under physiological conditions, a large Beclin 1–Vps34 complex is formed, including the core complex and Rubicon. This large complex may be reduced to form smaller complexes, such as an Atg14L–Beclin 1-containing complex and a Rubicon-containing complex. These smaller complexes may be the functional units participating in autophagy regulation through modulating the Vps34 lipid kinase activity. See Supplementary Information, Fig. S6 for full scans of blots in f.



L E T T E R S

stable cells did not affect the elution profiles of Atg14L–EGFP, Rubicon–EGFP and Beclin 1 (Supplementary Information, Fig. S4c–d).

To test the possibility that Atg14L and Rubicon are present in separate protein complexes while co‑eluted, we added an anti‑Rubicon antibody to the tissue extract before the gel filtration run and immunoblotted the resulting fractions with an anti‑Atg14L antibody. Our data show that Atg14L was co‑eluted with the anti‑Rubicon antibody (Supplementary Information, Fig. S4e), indicating that this antibody binds to the Atg14L‑containing complex, suggesting that Rubicon is present in this complex.

Moreover, we observed mutual co‑immunoprecipitation of Atg14L and Rubicon from transfected cells, further supporting that Atg14L and Rubicon can be present in the same protein complex (Fig. 1e, f); UVRAG was also co‑immunoprecipitated with Atg14L or Rubicon, and the inter‑action between UVRAG and Rubicon was significantly enhanced in the presence of Beclin 1 (Supplementary Information, Fig. S4f–h).

From these results, we conclude that Atg14L, Rubicon, UVRAG, Beclin 1, p150/Vps15 and Vps34 can form a major Beclin 1–Vps34 complex in vivo. However, Atg14L was also eluted in the later fractions (51–56) containing Beclin 1 or Beclin 1–EGFP (but not Rubicon; Fig. 1d; Supplementary Information, Fig. S4b), suggesting that Atg14L is also associated with a smaller Beclin 1 complex without Rubicon.

We performed several assays to determine the role of Atg14L in autophagy. First, we knocked down Atg14L expression in cultured cells by RNA inter‑ference (RNAi) using a short interfering RNA (siRNA) and analysed the levels of LC3II, a lipid‑conjugated form of LC3 that is normally localized on autophagosomes, by immunoblotting16–18. As with Beclin 1 siRNA, Atg14L siRNA resulted in increased levels of LC3II, compared with con‑trol siRNA (Fig. 2a, left). Second, we examined levels of p62/SQSTM1, a known autophagy substrate that normally accumulates when autophagy is impaired19–21. Again, like Beclin 1 siRNA, Atg14L siRNA resulted in increased p62/SQSTM1 levels (Fig. 2a). The increase in p62/SQSTM1 and LC3II levels resulting from Beclin 1 or Atg14L siRNA treatment was also significant after starvation (Fig. 2a). Therefore, knockdown of Atg14L or Beclin 1 impaired the autophagy‑mediated clearance of p62/SQSTM1 and LC3II.

Third, we knocked down Atg14L expression in MLE12 cells that stably expressed GFP–LC3. In control siRNA‑treated cells, many small GFP–LC3 puncta were observed, indicating that basal levels of autophagosomes were present (Supplementary Information, Fig. S5a). In contrast, Atg14L siRNA transfection resulted in accumulation of large GFP–LC3 puncta (Supplementary Information, Fig. S5a). These large GFP–LC3 puncta were colocalized with p62/SQSTM1 (Supplementary Information, Fig. S5b) and ubiquitin (Supplementary Information, Fig. S5c), indicating that these are ubiquitylated protein inclusions, as previously shown in Atg5- or Atg7‑deficient mouse‑tissues22,23. Ultrastructural analysis showed that Atg14L siRNA transfection resulted in a reduced number of autophagosomes (data not shown). These analy‑ses suggest that reduced Atg14L expression abolishes autophagosome formation and increases the levels of ubiquitylated proteins.

Fourth, under nutrient‑rich conditions, Atg14L siRNA treatment caused a slight decrease in the rate of degradation of long‑lived pro‑teins (~10%, P = 0.007) when compared with control siRNA treatment. However, this rate was markedly reduced upon nutrient withdrawal (~37%, P = 5 × 10–6) ; this effect of Atg14L siRNA was diminished in the presence of 3‑methyladenine, an inhibitor of autophagy (Fig. 2b).

Fifth, we investigated whether Atg14L modulates Vps34 kinase activity using a kinase assay that included Vps15/p150 (ref. 24). Our results show that

co‑expression of Flag–Atg14L with Myc–Vps34–Vps15 plasmids resulted in an increase in Vps34 activity that was 2.5‑fold higher than that caused by co‑expression of control Flag with Myc–Vps34–Vps15 plasmids (Fig. 2c). Interestingly, Atg14L‑mediated stimulation of Vps34 activity occurred only when co‑expressing Beclin 1. This result suggests that overexpression of Atg14L enhances Vps34 activity in a Beclin 1‑dependent manner.

Consistent with a previous report for Beclin 1–EGFP transgenic tis‑sues25, we found that Atg14L–EGFP or Beclin 1–EGFP stably expressed in cells was primarily diffuse in the cytoplasm (Supplementary Information, Fig. S5d). However, co‑expression of Atg14L–EGFP and Beclin 1–AsRed resulted in their colocalization on punctate structures (Fig. 2d). Electron microscopy analysis of these Atg14L–EGFP and Beclin 1–AsRed co‑trans‑fected cells showed many large ‘organelles’ (~3–5 μm; Fig. 2e). Some of these structures show concentric ‘rings’ with double membranes (Fig. 2e1, asterisks); many are large vacuole‑like structures filled with materials of high electron density (Fig. 2e2, asterisks) and enwrapped with double‑membranes (Fig. 2e2, insets) that are readily distinguishable from typical aggresomes or protein aggregates (usually not associated with limiting membranes). These structures are positive for Atg14L–EGFP, as shown by immuno‑electron microscopy (Fig. 2e4). We also observed an increased in the number of autophagosomes in these transfected cells (Fig. 2e3).

We next studied the nature of these Beclin 1–Atg14L‑resident structures. We found that these structures were negative for Golgi (Supplementary Information, Fig. S5e) or ER (Supplementary Information, Fig. S5f) mark‑ers. In contrast, they were colocalized with GFP–LC3 (Supplementary Information, Fig. S5g), suggesting that these Beclin 1–Atg14L struc‑tures probably recruit LC3. Moreover, they were colocalized with co‑expressed EGFP–Atg12 (Fig. 2f) or EGFP–Atg5 (Fig. 2g), suggesting that these Beclin 1–Atg14L structures may be involved in the early steps of autophagosome biosynthesis by recruiting Atg12 and Atg5.

To investigate the role of Rubicon in autophagy, we knocked down endogenous Rubicon protein levels. In contrast to Atg14L or Beclin 1 siRNA, Rubicon siRNA caused reduced steady‑state levels of LC3II and p62/SQSTM under normal or nutrient‑starvation conditions (Fig. 3a), suggesting that knockdown of Rubicon promotes autophagic activity. Conversely, in cells stably or transiently transfected with Rubicon–EGFP, the p62/SQSTM1 protein levels were markedly enhanced, compared with those in cells transfected with EGFP (Fig. 3b), suggesting that over‑expression of Rubicon inhibits autophagy.

To examine whether Rubicon also modulates Vps34 lipid kinase activity, we performed the lipid kinase assay described above. Our results show that co‑expression of Flag–Rubicon with Myc–Vps34–Vps15 markedly reduced Vps34 activity, but only in the absence of Beclin 1–EGFP overexpression (Fig. 3c). This result suggests that overexpression of Rubicon inhibits Vps34 kinase activity and that this effect does not require Beclin 1.

In previous studies, mCherry–GFP–LC3 was used to examine autophagosome maturation, for example, autophagosome acidifica‑tion following fusion with late endosomes/lysosomes26. We found that cells co‑expressing mCherry–GFP–LC3 and Flag–Rubicon contained primarily yellow fluorescent mCherry–GFP–LC3 puncta (immature autophagosomes), whereas cells expressing only mCherry–GFP–LC3 or co‑expressing mCherry–GFP–LC3 and control vector Flag contained considerable numbers of red fluorescent mCherry–GFP–LC3 puncta (mature autophagosomes; Fig. 3d, e). This suggests that overexpres‑sion of Rubicon may block autophagy by inhibiting autophagosome maturation.



L E T T E R S

Interestingly, Rubicon–EGFP (or Flag–Rubicon, data not shown) expres‑sion showed punctate subcellular localization (Fig. 4a). The Rubicon–EGFP puncta, some of which were ‘ring’‑shaped, were occasionally labelled with the early endosomal marker EEA1 (Supplementary Information, Fig. S5h) and were primarily colocalized with the late endosomal/lysosomal marker Lamp1 (Fig. 4a). Moreover, some of the Rubicon–EGFP puncta were positively stained with an antibody against lysobisphosphatidic acid (LBPA; Fig. 4b), an unusual eukaryotic lipid found only in the multi‑vesicular body (MVB)27, suggesting that some of the Rubicon–EGFP structures may be related to the MVB28. Electron microscopy analysis of Rubicon–EGFP‑transfected cells showed many abnormal, large, vacuole‑like structures (1–5 μm in diameter; Fig. 4c). Some of these structures contained high electron density molecules (Fig. 4c1–2), characteristic of late endosomes/lysosomes; some had relatively less content with overall low electron density, which may represent enlarged early‑stage endosomes (Fig. 4c3). Notably, some seemed to enclose numer‑ous small vesicles of multiple layers (Fig. 4c2, 4), whereas others resembled the MVB28 (Fig. 4c1–2). Through immuno‑electron microscopy using anti‑GFP gold particles, we observed that these vacuole‑like structures in the Rubicon–EGFP‑transfected cells were positive for Rubicon–EGFP. Moreover, Rubicon–EGFP was associated with the limiting membranes of these particular structures (Fig. 4d3, 4). Therefore, these structures cor‑responded to the fluorescent Rubicon–EGFP puncta (Fig. 4a). In addition, our immuno‑electron microscopy result confirmed the colocalization of Rubicon–EGFP and Lamp1 at the ultrastructural level (Fig. 4d5–7).

Bioinformatic analysis revealed that the cysteine‑rich domain of Rubicon shares sequence homology with the FYVE domain (Fig. 5a), a well‑charac‑terized motif specific for phosphatidylinositol‑3‑phosphate (PI(3)P) bind‑ing29. When examined experimentally, Rubicon–EGFP was not pulled down by PI(3)P‑conjugated sepharose beads, in contrast to the control PI(3)P‑binding protein 2×FYVE–EGFP (data not shown). However, co‑expressed Rubicon‑AsRed and p40 (phox)‑PX–EGFP, another reporter for PI(3)P binding, showed extensive colocalization (Fig. 5b), suggesting that the Rubicon‑associated structures are enriched in PI(3)P. Moreover, wortmanin, an inhibitor of Vps34 kinase, effectively dispersed the p40 (phox)‑PX–EGFP puncta but not the Rubicon–AsRed structures (Fig. 5b), suggesting that the maintenance of these Rubicon‑associated structures does not depend on PI(3)P. Furthermore, through immunofluorescence imaging of Rubicon truncation mutants, we found that the cysteine‑rich domain of Rubicon is required for the formation of the Rubicon‑positive structures that are enriched in PI(3)P and associated with the aberrant endosomes/lysosomes (Fig. 5c). Finally, we found that Beclin 1–AsRed or Flag–Beclin 1‑CE was excluded from Rubicon–EGFP puncta (Fig. 5d, e); Beclin 1 siRNA did not affect the formation of the Rubicon–EGFP puncta (Fig. 5f). Therefore, the formation of these Rubicon‑associated, late endo‑somal/lysosomal structures are Beclin 1‑independent.

In summary, our study has identified Atg14L and Rubicon, two com‑ponents in the Beclin 1–Vps34 protein complexes, and reveals their dis‑tinct roles in regulating autophagy and Vps34 kinase activity. We show that Atg14L and Rubicon may regulate autophagy by modulating Vps34 activity. Our study also suggests the existence of multiple Beclin 1 protein complexes that are engaged in distinct functions in autophagy regulation (Fig. 5g). The significance of these distinct Beclin 1 complexes remains to be fully elucidated. The dynamic change in protein composition between different functional Beclin 1–Vps34 complexes may have a central role in mediating the Beclin 1–Vps34 activity, which governs multiple cellular events, including autophagy.

Note added in proof: a related manuscript by Matsunaga et al. (Nature Cell Biol. 11, doi:10.1038/ncb1846; 2009) is also published in this issue.

METHODSReagents, antibodies and microscopy. See Supplementary Information for details.

Mouse genetics. Becn1–EGFP/+ mice were generated using BAC mouse transgenics30. Becn1–EGFP/+ transgenic and Becn1+/– mice were genetically crossed to generate the ‘rescued’ mice in which both endogenous Becn1 alleles are deleted and only Becn1–EGFP transgene is expressed. See Supplementary Information for further details.

Affinity purification and mass spectrometry. Affinity purification of Beclin 1 interacting proteins and mass spectrometric identification of these proteins were carried out as described previously19 with slight modifications. See Supplementary Information for details.

Plasmid constructs and stable cell lines. Total RNA was extracted from postnatal day 12 mouse whole brain using RNeasy mini kit (Qiagen). Full‑length cDNA was synthesized with Omniscript RT kit (Qiagen) and used as templates for PCR ampli‑fications with KOD HiFi DNA polymerase (Novagen). Becn1 was cloned into EcoRI and BamHI sites of pEGFP–N3 and pAs‑Red vectors (Clonetech). Atg14L was cloned into EcoRI and BamHI sites of pEGFP–N3, pAsRed and pCMV–Flag vectors (Sigma). Rubicon was cloned into HindIII and BamHI sites of pEGFP–N3, pAs‑Red and pCMV–Flag2 vectors. UVRAG was cloned into XhoI and BamHI sites of the pEGFP–N3 vector. Single or combinations of Becn1 domains were cloned into EcoRI and BamHI sites of the pCMV–Flag2 vector. Truncated Atg14L mutants were cloned into EcoRI and BamHI sites of the pEGFP–N3 vector. Truncated Rubicon mutants were cloned into KpnI and BamHI sites of the pEGFP–N3 vector. EGFP–Atg12 and EGFP–Atg5 constructs were provided by X. Jiang (Sloan‑Kettering Memorial Cancer Center, New York, NY). Myc–hVps34–hVps15–V5–His/pVITRO2 plasmid was described previously24. HEK 293 stable cells stably transfected with pEGFP–N3 vector, Beclin 1–EGFP, Atg14L–EGFP or Rubicon–EGFP were generated as described in the Supplementary Information.

Cell culture. Human embryonic kidney (HEK) 293 and 293T, Hela and NIH 3T3 cells were maintained in Dulbecco’s modified Eagle’s medium (DMEM) sup‑plemented with 10% fetal bovine serum (FBS) and 1% penicillin‑streptomycin (Invitrogen). MLE12 cells were provided by C. Münz (Rockefeller University, New York, NY) and maintained in DMEM/F12 medium (ATCC) supplemented with insulin (0.005 mg ml–1), transferrin (0.01 mg ml–1), sodium selenite (30 nM), hydro‑cortisone (10 nM), β‑estradiol (10 nM), HEPES (10 mM), l‑glutamine (2 mM), 2% FBS and 1% penicillin‑streptomycin. Transient DNA transfection was per‑formed using a standard calcium phosphate precipitation procedure, FuGene 6 or Lipofectamine 2000 kit, following the manufacturer’s protocol (Invitrogen and Roche Diagnostics). Transfection of NIH 3T3 cells and GFP–LC3 MLE12 stable cells with siRNA was performed with a Lipofectamine RNAi MAX kit following the reverse transfection protocol provided by the manufacturer (Invitrogen). The sequences of siRNA are: Beclin 1, CAGUUUGGCACAAUCAAUA; Atg14L, UUUGCGUUC‑AGUUUCCUCACUGCGC; Rubicon, GCCUUCAGUCUAUGCCACA.

In vitro protein immunoprecipitation. DNA plasmids were transfected into HEK 293T cells. For co‑immunoprecipitation experiments, two or three plasmids were transfected simultaneously in equal amounts. Cells were lysed in immuno‑precipitation lysis buffer (20 mM HEPES, pH 7.4, 1 mM MgCl2, 0.25 mM CaCl2, 0.2% Triton X‑100, 150 mM NaCl, EDTA‑free protease inhibitor cocktail (PIC, 1 tablet per 10 ml), 200 μg ml–1 phenylmethylsulphonyl fluoride (PMSF), 4 μg ml–1 pepstatin and DNase I). For immunoprecipiration with GFP, anti‑Atg14L or Rubicon antibodies, dynabeads M‑270 E‑proxy (Invitrogen) were conjugated with each antibody, and incubated with cell lysates at 4 °C for 2 h. After the beads were washed five times in immunoprecipitation lysis buffer, proteins were eluted by incubating beads in elution buffer (0.5 mM EDTA, pH 8 and 0.5 M NH3•H2O) at room temperature for 20 min, frozen in liquid nitrogen and dried in a vacuum speed centrifuge. For Flag‑tagged protein immunoprecipitation, anti‑Flag M2 affinity resin (Sigma) was used according to the manufacturer’s protocol.

Vps34 kinase assay. Vps34 kinase assay was performed as described previously31. Myc–hVps34–hVps15–V5–His/pVITRO2 plasmid was transfected into HEK 293T cells in combination with other Flag‑ or EGFP‑tagged plasmids. Cells were



L E T T E R S

lysed in 1% Nonidet P‑40 lysis buffer (20 mM Tris/pH 7.5, 137 mM NaCl, 1 mM MgC12, 1 mM CaC12, 100 mM NaF, 10 mM sodium pyrophosphate, 100 μM Na3VO4, 10% glycerol, 0.35 mg ml–1 PMSF, protease and phosphatase inhibitor cocktails). Immunoprecipitation was performed with anti‑Myc affinity gel beads, according to the manufacturer’s protocol (Sigma). Beads (associated with puri‑fied proteins) were washed three times in lysis buffer, followed by three washes in washing buffer (100 mM Tris‑HCl/pH7.4 and 500 mM LiCl) and two washes in reaction buffer (10 mM Tri‑HCl/pH7.4, 100 mM NaCl and 1 mM EDTA). Beads were resuspended in 60 μl of reaction buffer and MnCl2 (10 μ1 of 100 mM) and sonicated phosphatidylinositol (10 μl of 2 μg μl–1) were added. The reaction was started by the addition of ATP (10 μl of 440 μM) containing γ‑32P‑ATP (10 μCi), and beads were incubated for 10 min at room temperature. The reaction was ter‑minated by adding HCl (20 μl of 8 M), and the organic phase was extracted with 160 μl chloroform:methanol (1:1). Extracted phospholipid products were resolved by TLC using a coated silica gel and a solvent composed of chloroform:methanol:H2O:ammonium hydroxide (v/v/v/v, 9:7:1.7:0.3), followed by visualization with Typhoon 9400 Variable Imager (GE Healthcare Biosciences).

Gel filtration. Liver and brain extracts from both Becn1+/– and Becn1–/–;Becn1–GFP/+ mice (4 months of age) were prepared as described in the Supplementary Information. Cell extracts were prepared as described previously19. Tissue and cell extracts were diluted with equal volumes of 2× pull‑out buffer (1× containing 20 mM HEPES, pH 7.4, 1 mM MgCl2, PIC, 100 μg ml–1 PMSF, 2 μg ml–1 pepstatin, 0.2% triton X‑100 and 150 mM NaCl) and incubated for 15 min at 4 °C. The sam‑ples were then subject to ultracentrifugation at 100,000g and the resulting super‑natants were used for gel filtration experiments. A Superdex 200 HR10/30 column (Pharmacia) was equilibrated with 2‑bed volumes of filtered running buffer (1× pull‑out buffer without PIC and Triton‑X‑100). The column was calibrated using Biorad gel filtration calibrant mixtures that are composed of thyroglobulin (670K), γ‑globulin (158K), ovalbumin (44K), myoglobin (17K) and vitamin B12 (1,350). A spike of these calibrants (10 μl) was also added to each sample (240 μl) as internal calibrants. Both calibrants and samples were run at a flow rate of 0.2 ml min–1. For each run, 2‑bed volumes of running buffer were used to elute the sample and a total of 80 fractions were collected 25–29 min after starting the runs and at a rate of 1 fraction min–1. Two bed volumes of running buffer were used to wash the column at the same flow rate in between two consecutive runs.

Long-lived protein degradation assay. Long‑lived protein degradation was assessed as described previously32. In brief, NIH 3T3 cells were transfected with either control or Atg14L siRNA and plated in 12‑well plates. After 48 h, DMEM was changed to leucine‑free medium supplemented with 3H‑l‑leucine (1 μCi ml–1). After being pulse‑labelled for 24 h, cells were washed three times and cultured in DMEM supplemented with excess unlabelled leucine (5 mM) for 16 h to chase out short‑lived proteins. Cells were then washed three times and further cultured for 4 h in DMEM, Earle’s Balanced salt solution (EBSS), or EBSS supplemented with 10 mM 3‑methyladenine, all containing unlabelled leucine (5 mM). Both medium and cell lysates were subject to trichloroacetic acid (TCA) precipitation. Long‑lived protein degradation was calculated as the ratio of TCA‑soluble medium to TCA‑precipitated cell lysate radioactivity.

Statistical analysis. Statistical analyses were carried out as described previously33.


ACkNoWLedgeMeNTsThis work was supported by National Institutes of Health Grants RNS055683A (Z.Y.), RR00862 and RR022220 (B.T.C.) and by the Howard Hughes Medical Institute (N.H.). We thank A. Tolkovsky for GFP–LC3 HeLa cells, C. Münz for GFP–LC3 MLE12 cells, T. Johansen for mCherry–EGFP–LC3 plasmid and X. Jiang for EGFP–Atg12 and EGFP–Atg5 constructs. We thank W. Yang for help with cell culture and generation of stable cells, and H. Shio, E. Sphicas and A. North in the Bio‑Imaging Resource Center at The Rockefeller University for help with microscopy.

AuTHor CoNTriBuTioNs Z.Y. and Q.J.W. conceived the project. Z.Y. coordinated all efforts in the study; Z.Y., Q.J.W., Y.Z., N.H. and B.T.C. planned the project; Y.Z. and Q.J.W. performed most of the assays; X.L. assisted with p40 (phox)‑PX–EGFP localization and Vps34 kinase analyses; Y.Y. and J.M.B. developed the Myc–Vps34–Vps15 constructs and Vps34 kinase assay protocol; Z.Y., Q.J.W. and Y.Z. wrote the paper.

CoMpeTiNg fiNANCiAL iNTeresTsThe authors declare no competing financial interests.


1. Liang, X. H. et al. Induction of autophagy and inhibition of tumorigenesis by beclin 1. Nature 402, 672–676 (1999).

2. Liang, X. H. et al. Protection against fatal Sindbis virus encephalitis by beclin, a novel Bcl-2-interacting protein. J. Virol. 72, 8586–8596 (1998).

3. Yue, Z., Jin, S., Yang, C., Levine, A. J. & Heintz, N. Beclin 1, an autophagy gene essential for early embryonic development, is a haploinsufficient tumor suppressor. Proc. Natl Acad. Sci. USA 100, 15077–15082 (2003).

4. Qu, X. et al. Promotion of tumorigenesis by heterozygous disruption of the beclin 1 autophagy gene. J. Clin. Invest. 112, 1809–1820 (2003).

5. Edinger, A. L. & Thompson, C. B. Defective autophagy leads to cancer. Cancer Cell 4, 422 (2003).

6. Yue, Z. et al. A novel protein complex linking the δ2 glutamate receptor and autophagy: implications for neurodegeneration in lurcher mice. Neuron 35, 921–933 (2002).

7. Shibata, M. et al. Regulation of intracellular accumulation of mutant Huntingtin by Beclin 1. J. Biol. Chem. 281, 14474–14485 (2006).

8. Pacheco, C. D., Kunkel, R. & Lieberman, A. P. Autophagy in Niemann-Pick C disease is dependent upon Beclin-1 and responsive to lipid trafficking defects. Hum. Mol. Genet. 16, 1495–1503 (2007).

9. Kihara, A., Kabeya, Y., Ohsumi, Y. & Yoshimori, T. Beclin-phosphatidylinositol 3-kinase complex functions at the trans-Golgi network. EMBO Rep. 2, 330–335 (2001).

10. Kihara, A., Noda, T., Ishihara, N. & Ohsumi, Y. Two distinct Vps34 phosphatidyli-nositol 3-kinase complexes function in autophagy and carboxypeptidase Y sorting in Saccharomyces cerevisiae. J. Cell Biol. 152, 519–530 (2001).

11. Zeng, X., Overmeyer, J. H. & Maltese, W. A. Functional specificity of the mammalian Beclin-Vps34 PI 3-kinase complex in macroautophagy versus endocytosis and lyso-somal enzyme trafficking. J. Cell Sci. 119, 259–270 (2006).

12. Stack, J. H., DeWald, D. B., Takegawa, K. & Emr, S. D. Vesicle-mediated protein transport: regulatory interactions between the Vps15 protein kinase and the Vps34 PtdIns 3-kinase essential for protein sorting to the vacuole in yeast. J. Cell Biol. 129, 321–334 (1995).

13. Liang, C. et al. Autophagic and tumour suppressor activity of a novel Beclin1-binding protein UVRAG. Nature Cell Biol. 8, 688–699 (2006).

14. Fimia, G. M. et al. Ambra1 regulates autophagy and development of the nervous system. Nature 447, 1121–1125 (2007).

15. Takahashi, Y. et al. Bif-1 interacts with Beclin 1 through UVRAG and regulates autophagy and tumorigenesis. Nature Cell Biol. 9, 1142–1151 (2007).

16. Kabeya, Y. et al. LC3, a mammalian homologue of yeast Apg8p, is localized in autophagosome membranes after processing. EMBO J. 19, 5720–5728 (2000).

17. Ichimura, Y. et al. A ubiquitin-like system mediates protein lipidation. Nature 408, 488–492 (2000).

18. Kabeya, Y. et al. LC3, GABARAP and GATE16 localize to autophagosomal membrane depending on form-II formation. J. Cell Sci. 117, 2805–2812 (2004).

19. Wang, Q. J. et al. Induction of autophagy in axonal dystrophy and degeneration. J. Neurosci. 26, 8057–8068 (2006).

20. Yue, Z. Regulation of neuronal autophagy in axon: implication of autophagy in axonal function and dysfunction/degeneration. Autophagy 3, 139–141 (2007).

21. Komatsu, M. et al. Homeostatic levels of p62 control cytoplasmic inclusion body formation in autophagy-deficient mice. Cell 131, 1149–1163 (2007).

22. Hara, T. et al. Suppression of basal autophagy in neural cells causes neurodegenerative disease in mice. Nature 441, 885–889 (2006).

23. Komatsu, M. et al. Loss of autophagy in the central nervous system causes neurode-generation in mice. Nature 441, 880–884 (2006).

24. Yan, Y., Flinn, R. J., Wu, H., Schnur, R. S. & Backer, J. M. hVps15 but not calcium/calmodulin is required for the activity and regulation of hVps34 in mammalian cells. Biochem. J. 417, 747–755 (2009).

25. Arsov, I. et al. BAC-mediated transgenic expression of fluorescent autophagic protein Beclin 1 reveals a role for Beclin 1 in lymphocyte development. Cell Death Differ. 15, 1385–1395 (2008).

26. Pankiv, S. et al. p62/SQSTM1 binds directly to Atg8/LC3 to facilitatedegradation of ubiq-uitinated protein aggregates by autophagy. J. Biol. Chem. 282, 24131–24145 (2007).

27. Kobayashi, T. et al. A lipid associated with the antiphospholipid syndrome regulates endosome structure and function. Nature 392, 193–197 (1998).

28. Katzmann, D. J., Odorizzi, G. & Emr, S. D. Receptor downregulation and multivesicular-body sorting. Nature Rev. Mol. Cell Biol. 3, 893–905 (2002).

29. Stenmark, H., Aasland, R. & Driscoll, P. C. The phosphatidylinositol 3-phosphate-bind-ing FYVE finger. FEBS Lett. 513, 77–84 (2002).

30. Gong, S., Yang, X. W., Li, C. & Heintz, N. Highly efficient modification of bacterial artificial chromosomes (BACs) using novel shuttle vectors containing the R6Kgamma origin of replication. Genome Res. 12, 1992–1998 (2002).

31. Miled, N. et al. Mechanism of two classes of cancer mutations in the phosphoinositide 3-kinase catalytic subunit. Science 317, 239–242 (2007).

32. Gronostajski, R. M. & Pardee, A. B. Protein degradation in 3T3 cells and tumorigenic transformed 3T3 cells. J. Cell Physiol. 119, 127–132 (1984).

33. Komatsu, M. et al. Essential role for autophagy protein Atg7 in the maintenance of axonal homeostasis and the prevention of axonal degeneration. Proc. Natl Acad. Sci. USA 104, 14489–14494 (2007).






L E T T E R S

A mechanism for chromosome segregation sensing by the NoCut checkpointManuel Mendoza1,3, Caren Norden1,4, Kathrin Durrer1, Harald Rauter1, Frank Uhlmann2 and Yves Barral1,5

In Saccharomyces cerevisiae1 and HeLa cells2, the NoCut checkpoint, which involves the chromosome passenger kinase Aurora B, delays the completion of cytokinesis in response to anaphase defects. However, how NoCut monitors anaphase progression has not been clear. Here, we show that retention of chromatin in the plane of cleavage is sufficient to trigger NoCut, provided that Aurora/Ipl1 localizes properly to the spindle midzone, and that the ADA histone acetyltransferase complex is intact. Furthermore, forcing Aurora onto chromatin was sufficient to activate NoCut independently of anaphase defects. These findings provide the first evidence that NoCut is triggered by the interaction of acetylated chromatin with the passenger complex at the spindle midzone.

During mitosis, the spindle assembly checkpoint delays the onset of ana-phase until all chromosomes are properly attached to spindle microtubules3,4. Subsequently, sister-chromatids separate and the cleavage furrow starts to pinch the cell5,6. However, cell cleavage, or abscission, does not occur until all chromatids are retracted from the cleavage plane1,2. How cells coordinate abscission and chromosome segregation is poorly understood.

In S. cerevisiae, inactivation of spindle midzone components, such as Ase1 or Ndc10, causes premature spindle breakage and compromises chromosome segregation. In these cells, furrow ingression proceeds prop-erly but abscission is delayed by the NoCut pathway, preventing damage of stranded chromosomes by cytokinesis1. The chromosome passenger complex (CPC), consisting of the Aurora kinase Ipl1 and its regulator INCENP/Sli15 (refs 7, 8), acts at the top of the NoCut pathway. It con-veys the NoCut signal by targeting the NoCut effectors Boi1 and Boi2, two anillin-like proteins, to the site of cleavage at the bud neck1. CPC-dependent transfer of Boi1 and Boi2 to the bud neck occurs in every ana-phase, but is relieved at the onset of cytokinesis, after proper completion of anaphase. In cells with spindle defects, Boi1 and Boi2 remain at the bud neck and abscission is delayed. However, the exact event triggering the NoCut response is unknown. Interestingly, NoCut function is con-served in HeLa cells, where human Aurora B delays abscission in cells with chromosome bridges2. To investigate how Aurora kinase monitors

completion of anaphase in S. cerevisiae, we first assessed whether NoCut responds to lagging chromatin in the absence of spindle defects.

The topoisomerase II (Top2) mutant top2‑4 fails to decatenate sister chromatids at the restrictive temperature (30°C), causing chromatin to lag over the spindle midzone9 (Supplementary Information, Fig. S1). Top2 inactivation did not damage the spindle; at 30°C, localization of the mid-zone reporters Ase1–GFP and Ipl1–3GFP to anaphase spindles was similar in wild-type and top2‑4 mutant cells (Fig. 1a), which showed no increase in broken spindles. Using the pleckstrin homology domain of phospholipase C fused to GFP (PH–GFP) to visualize the plasma membrane, we next tested whether lagging chromatin affected cytokinesis progression. We inspected the plasma membrane at the bud neck of anaphase and telophase cells, that is, with one spindle-pole body (SPB, visualized using Spc42–CFP as a reporter) in the bud, and scored whether neck membranes were open, contracted or resolved into two separate membranes (Fig. 1b). The fraction of cells with an open bud neck was comparable between top2‑4 (39 ± 5%, mean ± s.d.) and wild-type cells (39 ±2 %), indicating that the onset of furrowing was not delayed. In contrast, the pre-abscission index (the ratio of cells with contracted versus resolved plasma membrane) was increased almost fourfold in the top2‑4 mutant cells, compared with wild-type cells (Fig. 1c, P < 0.005), but not in the top2‑4 boi1∆ boi2∆ triple-mutant cells (Fig. 1c). Thus, cells with defective Top2 form normal spindles yet trigger NoCut and delay abscission.

To test whether lagging chromatin on its own triggers NoCut, we next examined whether failure to separate chromosome arms also affected cytokinesis. Expression of non-cleavable cohesin Scc1RRDD blocks nuclear division in the first cycle after release from G1 phase10, yet these cells assemble a robust spindle midzone, as shown by the accumulation of Ase1–GFP and Ipl1–3GFP (Fig. 1d). These cells failed to complete sep-tation and separate from their buds, and became bi-budded (Fig. 1e). These bi-budded cells were not resolved on cell wall digestion by zymo-lyase (Fig. 1f, P < 0.001, compared with the wild-type). In contrast, most of the GAL:SCC1RRDD;boi1∆;boi2∆ mutant cells completed cyto-kinesis similarly to wild-type cells (Fig. 1e, f, P < 0.002, compared with GAL:SCC1RRDD). Unlike the GAL:SCC1RRDD single mutant, these triple-mutant cells developed a penetrant ‘cut’ phenotype, with nuclei cleaved

1Institute of Biochemistry, Biology Department, ETH Zurich, 8093 Zurich, Switzerland. 2Chromosome Segregation Laboratory, Cancer Research UK London Research Institute, Lincoln’s Inn Fields Laboratories, 44 Lincoln’s Inn Fields, London WC2A 3PX, United Kingdom. 3Current addresses: Centre for Genomic Regulation (CRG), C/ Dr. Aiguader, 88, 08003 Barcelona, Spain; 4Physiology Development and Neuroscience, Cambridge University, Downing St, Cambridge CB2 3DY, UK5Corresponding should be addressed to Y.B. (e-mail: [email protected])

Received 10 November 2008; accepted 18 February 2009; published online 8 March 2009; DOI: 10.1038/ncb1855




L E T T E R S

into unequal masses (Fig. 1e). Thus, in the absence of midzone damage, cells unable to separate chromosome arms during anaphase triggered NoCut and cytokinesis was aborted.

Non-cleavable cohesin impaired cytokinesis, but preventing chro-mosome separation by inactivating separase/Esp1 did not. Four hours after release from G1 at the restrictive temperature (37°C), the esp1‑1 cells became bi-budded. However, cell wall digestion efficiently resolved these aggregates, showing that cytokinesis was complete (Fig. 2a). Accordingly, calcofluor staining established that they had completed septation (Fig. 2b). Because the esp1‑1 mutant is defective in chromo-some segregation but not in cytokinesis, cell division should lead to DNA damage in this mutant. Indeed, foci of the DNA damage reporter protein Ddc1–GFP11 accumulated in esp1‑1 cells during the late stages of division (Fig. 2c). Preventing cytokinesis using the cdc12‑6 septin mutation sup-pressed the accumulation of such damage (esp1‑1;cdc12‑6, Fig. 2c). Thus, separase inactivation led to a cut phenotype; cytokinesis proceeded in the absence of chromosome segregation and damaged DNA.

Together, these results suggest that Esp1 contributed to the coordination of cytokinesis with chromosome segregation. On release from G1 arrest at the restrictive temperature (37°C), ndc10‑1 mutant cells prematurely break their spindle and cytokinesis fails, leading to the accumulation of bi-budded cells that are not resolved on cell wall digestion1 (Fig. 2d, P < 0.01, compared with wild-type). In contrast, ndc10‑1;esp1‑1 double-mutant cells were effectively resolved into unbudded and single-budded

cells by zymolyase treatment. Furthermore, plasma membrane and cell wall imaging showed that these cells properly completed abscission and septation, unlike the ndc10‑1 single-mutant cells (Fig. 2e). Thus, separase function was required to inhibit abscission in cells with a fragile spindle. We conclude that NoCut function depends on separase.

Early in yeast anaphase, separase cleaves cohesin to resolve sister-chro-matid cohesion12, and promotes activation of the protein phosphatase Cdc14 (ref 13, 14). This latter function is independent of the proteo-lytic activity of separase; protease-dead Esp1C1531A successfully mediates Cdc14 activation, but not cohesin cleavage15. Remarkably, expression of Esp1C1531A impaired cytokinesis in esp1‑1 single- and ndc10‑1 esp1‑1 double-mutant cells, as shown by calcofluor staining and zymolyase digestion (Fig. 3a, b). In contrast, Esp1C1531A expression did not impair cytokinesis in wild-type cells. Thus, the non-catalytic function of sep-arase is required for cytokinesis inhibition in response to both spindle and chromosome segregation defects.

In addition to separase, the FEAR network, including the nucleolar proteins Bns1 and Spo12, and the spindle-associated factor Slk19, is also required for Cdc14 activation in early anaphase14. Zymolyase digestion and imaging of the septum and plasma membrane (Fig. 3c, d) showed that, in contrast to the ndc10‑1 single-mutant, the ndc10‑1;bns1∆;spo12∆ triple-mutant completed cytokinesis efficiently within 4 h. Inactivation of Slk19 in the ndc10‑1 background also restored membrane resolu-tion and septation (Fig. 3d). Deletion of BNS1 and SPO12 or SLK19 did

cba

e f

Bi-

bud

ded

cel

ls a

fter

zym

olya

se

dig

estio

n (p

erce

ntag

e)

0

25

50

SCC1-RRDD SCC1-RRDD

boi1∆ boi2∆WT

* *

top

2-4

Ipl1–3GFPSpc72–GFP

Scc

1-RR

DD

WT

d

Spc42–CFP pPH–GFP

Open Contracted Resolved

GAL–SCC1-RRDD

GAL–SCC1-RRDD

boi1∆ boi2∆

DA

PI /

pha

seC

alco

fluor

WT top2-4 top2-4boi1∆ boi2∆

0

0.5

1.0

Pre

-ab

scis

sion

ind

ex

*

Ase1–GFPSpc42–CFP

Ipl1–3GFPSpc72–GFP

Ase1–GFPSpc42–CFP

Figure 1 Defects in chromosome segregation trigger NoCut-dependent inhibition of cytokinesis in the absence of midzone damage. (a) Localization of the central spindle components Ase1 and Ipl1 in wild-type (WT) and top2‑4 mutants. (b) Configuration of the plasma membrane (PH–GFP) in anaphase and post-anaphase cells with one SPB (Spc42–CFP, red arrows) segregated into the bud. White symbols indicate open (arrowhead), contracted (arrow) or resolved (asterisk) bud neck membranes. (c) Quantification of the pre-abscission index (fraction of cells with contracted/resolved membranes) in cells of the indicated strains. Except when indicated otherwise, in this and following graphs statistically significant differences

(P < 0.02) are highlighted with an asterisk. Data are mean ± s.d., n = 3 (d) Localization of Ase1 and Ipl1 in cells expressing the non-cleavable cohesin Scc1RRDD. (e) DAPI/phase and calcofluor white cell wall staining of BOI1 BOI2 and boi1 boi2 cells expressing Scc1RRDD under the control of the GAL promoter (GAL–SCC1RRDD). Arrowheads point to open bud necks and the arrow points to a completed septum. (f) Fraction of bi-budded cells in the indicated strains after septum digestion with zymolyase. In a–c, WT and top2‑4 cells were grown at 30°C for 4 h. Strains expressing Scc1RRDD and cells in d–f were released from a G1 arrest in galactose medium at 37°C and examined after 90 min (d) or 4 h (e, f). Scale bars, 1 μm.



L E T T E R S

not suppress the spindle fragility of the ndc10‑1 cells (Supplementary Information, Fig. S2). Similarly, bns1∆ spo12∆ mutant cells expressing Scc1RRDD underwent two rounds of budding within 4 h without cytoki-nesis failure, and did not accumulate as zymolyase-resistant bi-budded cells (Fig. 3c). Thus, the NoCut response to fragile spindles and sister chromatid resolution defects depended on FEAR function.

Because Cdc14 mediates both mitotic exit and cytokinesis onset16, we could not directly test whether it has a role in inhibiting cytokinesis completion. Instead, we assessed whether dephosphorylation of a known Cdc14 target contributes to NoCut function. Interestingly, Cdc14 regu-lates the CPC by dephosphorylating INCENP/Sli15 at the onset of ana-phase. Sli15 dephosphorylation depends on FEAR and targets CPC to the central spindle17. To investigate whether NoCut function requires Sli15

dephosphorylation, we tested whether constitutively dephosphorylated Sli15, Sli15-6A18, could restore NoCut in FEAR-defective cells. In contrast to esp1‑1 and ndc10‑1 esp1‑1 mutant cells, cell wall digestion and imaging of septa and plasma membrane all indicated that cytokinesis was aborted in many esp1‑1 SLI15‑6A (P < 0.01, compared with esp1‑1) and ndc10‑1 esp1‑1 SLI15‑6A (P < 0.0001, compared with ndc10‑1 esp1‑1) mutant cells (Fig. 3e, f). Thus, non-phosphorylatable Sli15 restored NoCut function in FEAR defective cells. It also restored Ipl1–3GFP recruitment to the spindle in the esp1‑1 and ndc10‑1 esp1‑1 cells, as predicted17 (Fig. 3g and data not shown), as well as Ase1 localization in ndc10‑1 esp1‑1 mutant cells (Fig. 3i). Thus, Cdc14-dependent regulation of CPC accounted for most, if not all, FEAR requirement in NoCut. Importantly, expression of Sli15-6A caused cytokinesis to be aborted only in the esp1‑1 and ndc10‑1

Bi-

bud

ded

cel

ls (p

erce

ntag

e)

Time after G1 release (min)

Frac

tion

of c

ells

with

D

dc1

–GFP

foci

(per

cent

age)

0

20

40

0 100 200 300

WTesp1-1esp1-1 cdc12-6

a b

c

Dd

c1–G

FP

esp1-1 esp1-1cdc12-6

WT esp1-1

Cal

coflu

or

Untreated Zymolyase

0

25

50

WT esp1-1

d

Bi-

bud

ded

cel

ls a

fter

zym

olya

se

dig

estio

n (p

erce

ntag

e)

0

30

60

WT esp1-1 ndc10-1 ndc10-1esp1-1

*

ndc1

0-1

PH–GFPCalcofluornd

c10-

1 es

p1-

1e

Figure 2 Separase is required for the NoCut response. (a) Wild-type (WT) and esp1‑1 cells were released from G1 arrest at 37°C for 4 h and the fraction of multi-budded cells was determined before and after digestion of the cell wall with zymolyase. Data are mean ± s.d., n = 3. (b) Calcofluor white staining of the cell wall and septum in WT and esp1‑1 mutant cells. Open bud necks are marked with arrowheads and arrows point to completed septa. (c) Time-course of Ddc1–GFP foci formation (left panel). Data shown are from one representative experiment. Nuclear foci of Ddc1–GFP (arrows), corresponding to DNA double-strand breaks, were observed in esp1‑1 cells

but not in esp1‑1 cdc12‑6 cells (right panel). (d) Fraction of bi-budded cells in the indicated strains following septum digestion with zymolyase. Data are mean ± s.d., n = 3. (e) Status of the division septum (stained with calcofluor) and plasma membrane (PH–GFP) in cells of the indicated strains. Arrowheads point to open septa or open bud neck membranes; arrows point to complete septa or resolved membranes; asterisks mark contracted membranes. Cells of the indicated strains were arrested in G1 with α-factor, released in fresh medium at 37°C and analysed every 30 min (c) or after 4 h. Scale bars, 1 μm.



L E T T E R S

esp1‑1 cells, but not in the wild-type. Therefore, recruitment of CPC to the midzone is required for NoCut to sense spindle and sister chromatid resolution defects, but does not trigger NoCut on its own.

Independently, while screening for NoCut genes (H.R. and Y.B., unpub-lished observations) we identified chromatin components, including Ahc1,

a scaffolding element of the ADA histone acetyltransferase19. In support of Ahc1 functioning in NoCut, ndc10‑1 ahc1∆ cells failed to inhibit abscis-sion, as shown by the reduction in the number of multi-budded cells fol-lowing zymolyase treatment (Fig. 4a). Further, the double mutant ahc1∆ ase1∆ showed reduced viability when compared with either single mutant

ndc1

0-1

esp

1-1

GA

L-E

sp1C

1531

A

esp

1-1

GA

L-E

sp1C

1531

A

Calcofluora b

Bi-

bud

ded

cel

ls a

fter

zym

olya

se d

iges

tion

(per

cent

age)

0

30

60

WT

esp

1-1

esp

1-1

Esp

1C15

31A

ndc1

0-1

esp

1-1

ndc1

0-1

esp

1-1

Esp

1C15

31A

Esp

1C15

31A

ndc10-1bns1∆ spo12∆

d ndc10-1slk19∆

c

0

30

60

WT

SC

C1-R

RD

D

bns

1∆ s

po1

2∆

ndc1

0-1

ndc1

0-1

bns

1∆ s

po1

2∆

bns

1∆ s

po1

2∆

SC

C1-R

RD

D

*

*

**

0

25

50

WT

esp

1-1

ndc1

0-1

esp

1-1

ndc1

0-1

esp

1-1

Sli1

5-6A

esp

1-1

Sli1

5-6A

Sli1

5-6A

Bi-

bud

ded

cel

ls a

fter

zym

olya

se d

iges

tion

(per

cent

age)

Bi-

bud

ded

cel

ls a

fter

zym

olya

se d

iges

tion

(per

cent

age)

* *

esp1-1 esp1-1 Sli15-6A

SC

C1-R

RD

D

WT

esp

1-1

esp

1-1

Sli1

5-6ALa

rge-

bud

ded

cel

ls w

ith Ip

l1

at t

he m

idzo

ne (p

erce

ntag

e)

0

25

50

esp1-1 ndc10-1esp1-1esp1-1ndc10-1

SLI15-6A

Ase

1–G

FP

Sp

c42-

CFP

Ipl1

–3G

FPS

pc7

2-G

FP

Cal

coflu

or

ndc10-1 esp1-1 Sli15-6A

PH

–GFP

Cal

coflu

orP

H–G

FP

WT (meta)

e f

h

g

i

Figure 3 FEAR is required for the NoCut response through Ipl1 targeting to spindle microtubules. (a, c, e) Fraction of bi-budded cells after septum digestion with zymolyase. (b, d, f) Status of the division septum (stained with calcofluor) and of the plasma membrane (PH–GFP) in cells of the indicated strains. Arrowheads point to open septa or open bud neck membranes and arrows point to complete septa and contracted membranes; asterisks mark resolved membranes. (g–i) Recruitment of

Ipl1–3GFP and Ase1–GFP to the spindle midzone in large-budded cells of the indicated strains after 2 h at 37°C. A small-budded, metaphase wild-type cell is shown in g for comparison; SCC1RRDD midzones are shown in Fig. 1d. In all panels, cells were arrested in G1 with α-factor and then released in fresh galactose (for Scc1RRDD and Esp1C1531A induction) or glucose medium at 37°C for 4 h before processing. Scale bars, 1 μm.



L E T T E R S

(Fig. 4b). Because ase1∆ cells rely on NoCut to prevent chromosome dam-age1, this observation suggests that Ahc1 is required to delay abscission in cells with anaphase-spindle defects. Accordingly, although the pre-abscis-sion index of ase1∆ cells was increased as reported1, this abscission delay was suppressed in the ahc1∆ ase1∆ cells (Fig. 4c). Furthermore, localiza-tion of the NoCut effector Boi2 to the cleavage site was perturbed in ahc1∆ mutant cells (Fig. 4d, e). Therefore, the chromatin component Ahc1 was required for proper NoCut function in cells with anaphase defects.

These observations raise the possibility that chromatin is directly involved in NoCut sensing. As chromatin is a potent activator of Aurora B in vitro20, the CPC on the midzone may signal the presence of chromatin. We rational-ized that if this is correct, preventing CPC segregation away from chromatin should trigger NoCut independently of chromosome segregation defects or spindle damage. To test this possibility, Ipl1 was fused to the Tet repressor and YFP (Ipl1–TetR–YFP) to tether it to chromatin in cells carrying Tet operator (TetO) repeats. Expression of Ipl1–TetR–YFP did not affect growth of wild-type cells and fully complemented the ipl1‑321 mutant (Fig. 5a). Furthermore, like wild-type Ipl1, the fusion protein localized to two dots located between the two spindle poles of metaphase cells, probably the kine-tochores, and to the spindle midzone during anaphase (Fig. 5b). In cells containing TetO repeats, one or two supplementary YFP foci were observed in metaphase and anaphase nuclei, respectively (Fig. 5c). Thus, a fraction of Ipl1–TetR–YFP successfully attached to chromatin in these cells.

We determined the pre-abscission index of cells co-expressing Ipl1–TetR–YFP, Spc42–CFP and PH–GFP, and containing TetO arrays on the sub-telomeric region of chromosome XII (tetO:TEL12R) or near the centromere of chromosome IV (tetO:TRP1). The fraction of cells with an open bud neck was comparable in TetO:TEL12R cells carrying an empty plasmid or expressing Ipl1–TetR–YFP (control, 41.5 ± 6.1%; Ipl1–TetR–YFP, 37.8 ± 6.3%). In contrast, the pre-abscission index was increased 2.5-fold in IPL1–TetO:TEL12R cells expressing the Ipl1–TetR–YFP

fusion, compared with TetO cells carrying an empty plasmid (Fig. 5d, P < 0.005;). This increase was independent of the position of the TetO array on the chromosome, as Ipl1–TetR elicited a comparable effect in tetO:TRP1 and tetO:TEL12R (Supplementary Information, Fig. S3). As these effects were observed in the presence of wild-type, endogenous Ipl1, the Ipl1–TetR construct acted dominantly.

The inhibition of abscission caused by Ipl1–TetR required Ipl1 kinase activity; TetR fused to kinase-dead Ipl1 (Ipl1D227A)21 did not delay abscission. Furthermore, Ipl1–TetR did not inhibit abscission of IPL1;boi1∆;boi2∆;tetO:TEL12R cells (Fig. 5e). Thus, Ipl1 tethering to chromatin led to a NoCut-dependent abscission delay. This delay was not due to spindle stabilization or destabilization by Ipl1–TetR, as the frac-tion of cells undergoing anaphase was not altered by Ipl1–TetR expres-sion (Supplementary Information, Fig. S4). Furthermore, Ipl1–TetR fully bypassed the requirement for FEAR and ADA function in NoCut acti-vation; Ipl1–TetR–YFP sucessfully activated NoCut in slk19∆ TetO and ahc1∆ TetO cells (Fig. 5e). Thus, although simultaneous occurrence of Ipl1 on the spindle midzone and lagging chromatin around it is normally required to trigger NoCut, these requirements were bypassed when Ipl1 was forced onto chromatin. In addition, these results indicate that Ahc1 contributes to NoCut upstream of Ipl1 function. Clustering of aurora-B can lead to its activation, as observed when two CPC complexes are brought together with specific antibodies20. However, the effect of ‘clustering-medi-ated’ activation of Ipl1 is probably limited, as TetR-mediated dimerization22 caused only a mild abscission delay on its own, that is, in cells lacking TetO sequences (Fig. 5d, P < 0.008). We conclude that tethering Ipl1 to chroma-tin mimics the events required for the activation of the NoCut response in cells with chromosome segregation and spindle defects.

In summary, our data show that 1) NoCut is triggered by the presence of unsegregated chromatin lagging over the spindle midzone, even in the absence of spindle defects; 2) NoCut function requires targeting of the CPC

0

30

60

WT ndc10-1 ndc10-1ahc1∆B

i-b

udd

ed c

ells

aft

er z

ymol

yase

d

iges

tion

(per

cent

age)

0

0.5

1.0

WT ase1∆ ase1∆ahc1∆

Pre

-ab

scis

sion

ind

exa b

c

0

40

80

Bud neck Bud cortex Mislocalized

WTahc1∆

Boi

2–G

FP lo

caliz

atio

n in

ana

pha

se

and

tel

opha

se (P

erce

ntag

e)

Bud neck Bud cortex Mislocalized

Boi2–GFP Spc42–CFPd e

WT

ahc1∆

ase1∆

ahc1∆ase1∆

25ºC 37ºC

wt wt ahc1 ahc1

Figure 4 The ADA histone acetyltransferase component Ahc1 is required for the NoCut response. (a) Fraction of bi-budded cells following septum digestion with zymolyase. Cells were released from G1 arrest at 37°C for 4 h before processing. (b) Threefold serial dilutions of cells of the indicated strains were plated on YPD and grown for 2–3 days at the indicated

temperatures. (c) Quantification of the pre-abscission index (fraction of cells with contracted/resolved membranes) in cells of the indicated strains expressing Spc42–CFP and PH–GFP. (d, e) Localization of Boi2–GFP (green) in wild-type (WT) and ahc1∆ cells expressing Spc42–CFP (red). In c–e, cells were grown in minimal medium at 23°C. Scale bars, 1 μm.



L E T T E R S

to the central spindle during anaphase, indicating that the NoCut signal is generated at this location and time; 3) The chromatin component Ahc1 contributes upstream of Ipl1 to proper NoCut activation. As Ahc1 is a core component of the histone acetylation complex ADA, acetylation events might trigger or be required for Ipl1 to detect anaphase defects. Remarkably,

ahc1∆ mutant cells not only failed to inhibit abscission in response to ana-phase defects, but also showed a high incidence of DNA damage23. At least some of this damage may result from NoCut inactivation1. 4) Holding Ipl1 in contact with chromatin throughout anaphase triggers the NoCut response independently of anaphase defects, Ahc1 function, FEAR and

a b

c d

Spc42–CFP Ipl1–TetR–YFP

PH-GFP Merge

Met

apha

seA

nap

hase

WT

ipl1

-321

Emptyvector

GAL:Ipl1–TetR–YFP

GAL:Ipl1–TetR–YFP

GAL:Ipl1–YFP

Emptyvector

Spc42–CFP GAL:Ipl1–TetR–YFP PH–GFP

e

Chromatin in midzoneIpl1 active

No abscission

Chromatin away from midzoneIpl1 inactiveAbscission

Ipl1Sli15

Chromatin

CPC

0

1.0

2.0

Emptyvector

Ipl1–TetR

Emptyvector

Ipl1–TetR

WT TetO:TEL12R

boi1∆boi2∆

Pre

-ab

scis

sion

ind

ex

*

*

0

0.5

1.0

IPL1–TetR

IPL1D227A–TetR

ahc1∆slk19∆WT WT

Ipl1-TetR

Pre

-ab

scis

sion

ind

exin

Tet

O:T

EL1

2R (n

orm

aliz

ed)

f

Sli15

MTs

Chromatin

CPC

Ipl1Sli15

P

Cdc14

Ipl1?

PP

Ac Ac

Ac Ac Ac

AcAc

Ac

Spc42–CFP TetO:TEL12R GAL:Ipl1–TetR–YFP PH–GFP

Open Contracted Resolved

Ipl1

–Tet

R–Y

FPP

H–G

FPS

PC

42-C

FP(p

roje

ctio

n)P

H–G

FP(s

ingl

e se

ctio

n)

Figure 5 Tethering of Ipl1 to chromosome arms triggers the NoCut response. (a) Threefold serial dilutions of wild-type (WT) or ipl1‑321 cells bearing control, Ipl1–YFP- or Ipl1–TetR–YFP-encoding plasmids were plated on galactose medium and grown for 3 days at 35°C. (b, c) Localization of Ipl1–TetR–YFP (green) and Spc42–CFP (red) in wild-type (b) or TetO strains expressing the membrane marker PH–GFP (green; c). Arrowhead, nuclear focus possibly corresponding to kinetochores in the metaphase cell; arrow, spindle midzone in the anaphase cell (b). Green arrows, Ipl1 foci at chromosomal TetO:TEL12R arrays in anaphase and telophase cells; white symbols indicate open (arrowhead), contracted (arrow) or resolved (asterisk) bud neck membranes (c). Scale bars, 1 μm. (d, e) Pre-abscission index

(fraction of cells with contracted/resolved membranes) in cells of the indicated strains expressing Ipl1–TetR fusions. (f) A model for how Aurora monitors chromatin segregation during anaphase. Upper panel: in early anaphase, separase and FEAR-dependent activation of Cdc14 targets the CPC (the Ipl1 and Sli15 subunits are depicted in red and blue, respectively) to spindle midzone microtubules (MTs; in orange). We speculate that Ipl1 is activated there through interaction with chromatin-associated factors (in green; chromosomes are depicted in purple), which require Ahc1 function to interact with midzone-bound CPC. As a result, abscission is inhibited. Lower panel: on completion of chromosome segregation, the CPC is no longer activated by chromatin and the NoCut signal is turned off, thus abscission ensues.



L E T T E R S

proper localization of Ipl1 to the spindle midzone. Together, these data suggest that the CPC acts as a sensor that activates NoCut in response to the presence of chromatin around the spindle midzone (Fig. 5f). This model explains why abscission is delayed in response to situations as distinct as the presence of chromosome bridges and premature spindle breakage.

Studies in Schizosaccharomyces pombe identified the cut mutants, which block anaphase yet proceed through cytokinesis, cutting the undivided nucleus24. These findings suggest that cytokinesis is not coordinated with chromosome segregation in eukaryotes. However, the cut phenotype may be attributed to NoCut delaying rather than preventing cytokinesis, as we observed in the top2‑4 mutant. Indeed, these cells do eventually complete cytokinesis, which causes DNA damage25 and cell death. Alternatively, cut mutants may impair both chromosome segregation and NoCut. Indeed, the S.cerevisiae separase mutant esp1‑1 also develops a cut phenotype, but as we show, this is because separase is required for mounting the NoCut response. Thus, NoCut seems to be conserved in evolution. The cut gene collection will probably provide an excellent resource for further dissec-tion of the NoCut checkpoint.

METHODSStrains and plasmids. All yeast strains are derivatives of S288C. SPC72–GFP and ASE1–GFP strains have been described previously26. Gene deletion strains were obtained from EUROSCARF27 or were generated by PCR-based gene disruption. The Boi2–GFP plasmid has been described previously1. tetO strains were gifts from Luis Aragón (MRC, London) and Duncan Clarke (University of Minnesota, Minneapolis, MN). Ipl1 constructs were cloned into pRS416 (ref. 28) and the kinase dead Ipl1 allele was generated by site-directed mutagenesis (QuickChange, Stratagene). The pleckstrin homology domain of Rattus norvegicus phospholipase C ∂1 fused to GFP (PH–GFP), a gift from Scott Emr (Howard Hughes Medical Institute, Bethesda, MD), was expressed from pRS426-based plasmids29. The Sli15‑6A allele is identical to that described previously17 except in the choice of one phosphorylation site: Ser 335, 427, 437, 448, 462 and Thr 474 were mutated to Ala.

Growth conditions and staining procedures. Cells were grown in rich medium (YPD) at room temperature, unless indicated otherwise. For synchronization experiments, cells were arrested with α-factor (10 μg ml–1; Sigma) for 2–4 h at 22°C, washed twice in fresh medium and released at the restrictive temperature (for temperature-sensitive mutants) or at 22°C. For galactose induction, cells were arrested with α-factor in YP + 2% raffinose, released in YP + raffinose + 2% galactose and examined after 4–5 h. Expression of Ipl1 constructs was induced by addition of galactose (2%) to exponentially growing cultures in selective raffinose medium at 22°C. Cells were examined 4–5 h after galactose addition (8 h for ahc1∆). For DAPI staining, cells were fixed for 30 min in 70% ethanol, washed in PBS and resuspended in PBS containing 1 μg ml–1 DAPI. For calcofluor staining and septum digestion, cells were fixed with 3.7% formaldehyde for 30 min and washed twice with PBS. Calcofluor (Sigma) was used at 0.01 mg ml–1 in PBS.

Microscopy. Imaging was performed on Olympus BX50 and Leica AF7000 fluores-cence microscopy systems equipped with a piezo motor, as described previously30. Spindle images are maximum projections of z-stacks; plasma membrane status was evaluated on single, non-confocal z axis slices (9 stacks spaced 300 nm).

Cytokinesis assays. The frequency of multi-budded (cytokinesis-defective) cells was evaluated by light microscopy after 30 min digestion with zymolyase (2 mg ml–1) in 1 M sorbitol at 22°C. Pre-abscission indices of cells expressing PH–GFP and Spc42–CFP were calculated as (fraction of cells with contracted bud neck membranes)/(fraction of cells with resolved membranes). Only cells in which one SPB had entered the bud were considered for analysis. In all cases, the fraction of cells with open bud necks did not change significantly (P > 0.05). Results of abscission and zymolyase assays are expressed as the mean ± s.d. of at least three independent experiments. More than 100 cells were counted for each condition.

Statistical analysis. Unpaired two-tailed t-tests allowing for unequal variance were used (Microsoft Excel).


ACKNowleDgeMeNtsWe are grateful to Patrick Steigemann, Daniel Gerlich, Patrick Meraldi, Hemmo Meyer and all members of the Barral lab for fruitful discussions and critical reading of the manuscript. Thanks to Dominik Theler, Trinidad Sanmartin and Joelle Sasse for technical assistance, Orna Cohen-Fix for sharing reagents, and the ETH Light Microscopy Center for their invaluable support. This work was supported by an SNF Grant to Y.B. (2-77542-04).

CoMpetiNg FiNANCiAl iNteRestsThe authors declare no competing financial interests.


1. Norden, C. et al. The NoCut pathway links completion of cytokinesis to spindle midzone function to prevent chromosome breakage. Cell 125, 85–98 (2006).

2. Steigemann, P. et al. Aurora B-mediated abscission checkpoint protects against tetra-ploidization. Cell 136, 473–484 (2009).

3. Musacchio, A. & Salmon, E. D. The spindle-assembly checkpoint in space and time. Nature Rev. Mol. Cell Biol. 8, 379–393 (2007).

4. May, K. M. & Hardwick, K. G. The spindle checkpoint. J. Cell Sci. 119, 4139–4142 (2006).

5. Barr, F. A. & Gruneberg, U. Cytokinesis: placing and making the final cut. Cell 131, 847–860 (2007).

6. Eggert, U. S., Mitchison, T. J. & Field, C. M. Animal cytokinesis: from parts list to mechanisms. Annu. Rev. Biochem. 75, 543–566 (2006).

7. Ruchaud, S., Carmena, M. & Earnshaw, W. C. Chromosomal passengers: conducting cell division. Nature Rev. Mol. Cell Biol. 8, 798–812 (2007).

8. Vader, G., Medema, R. H. & Lens, S. M. The chromosomal passenger complex: guiding Aurora-B through mitosis. J. Cell Biol. 173, 833–837 (2006).

9. Holm, C., Goto, T., Wang, J. C. & Botstein, D. DNA topoisomerase II is required at the time of mitosis in yeast. Cell 41, 553–563 (1985).

10. Uhlmann, F., Lottspeich, F. & Nasmyth, K. Sister-chromatid separation at anaphase onset is promoted by cleavage of the cohesin subunit Scc1. Nature 400, 37–42 (1999).

11. Melo, J. A., Cohen, J. & Toczyski, D. P. Two checkpoint complexes are independently recruited to sites of DNA damage in vivo. Genes Dev. 15, 2809–2821 (2001).

12. Uhlmann, F., Wernic, D., Poupart, M. A., Koonin, E. V. & Nasmyth, K. Cleavage of cohesin by the CD clan protease separin triggers anaphase in yeast. Cell 103, 375–386 (2000).

13. Queralt, E., Lehane, C., Novak, B. & Uhlmann, F. Downregulation of PP2A(Cdc55) phosphatase by separase initiates mitotic exit in budding yeast. Cell 125, 719–732 (2006).

14. Stegmeier, F., Visintin, R. & Amon, A. Separase, polo kinase, the kinetochore protein Slk19, and Spo12 function in a network that controls Cdc14 localization during early anaphase. Cell 108, 207–220 (2002).

15. Sullivan, M. & Uhlmann, F. A non-proteolytic function of separase links the onset of anaphase to mitotic exit. Nature Cell Biol. 5, 249–254 (2003).

16. Stegmeier, F. & Amon, A. Closing mitosis: the functions of the Cdc14 phosphatase and its regulation. Annu. Rev. Genet. 38, 203–232 (2004).

17. Pereira, G. & Schiebel, E. Separase regulates INCENP-Aurora B anaphase spindle function through Cdc14. Science 302, 2120–2124 (2003).

18. Higuchi, T. & Uhlmann, F. Stabilization of microtubule dynamics at anaphase onset promotes chromosome segregation. Nature 433, 171–176 (2005).

19. Eberharter, A. et al. The ADA complex is a distinct histone acetyltransferase complex in Saccharomyces cerevisiae. Mol. Cell. Biol. 19, 6621–6631 (1999).

20. Kelly, A. E. et al. Chromosomal enrichment and activation of the aurora B pathway are coupled to spatially regulate spindle assembly. Dev. Cell 12, 31–43 (2007).

21. King, E. M., Rachidi, N., Morrice, N., Hardwick, K. G. & Stark, M. J. Ipl1p-dependent phosphorylation of Mad3p is required for the spindle checkpoint response to lack of tension at kinetochores. Genes Dev. 21, 1163–1168 (2007).

22. Ramos, J. L. et al. The TetR family of transcriptional repressors. Microbiol Mol. Biol. Rev. 69, 326–356 (2005).

23. Alvaro, D., Lisby, M. & Rothstein, R. Genome-wide analysis of Rad52 foci reveals diverse mechanisms impacting recombination. PLoS Genet. 3, e228 (2007).

24. Yanagida, M. Fission yeast cut mutations revisited: control of anaphase. Trends Cell Biol. 8, 144–149 (1998).

25. Baxter, J. & Diffley, J. F. Topoisomerase II inactivation prevents the completion of DNA replication in budding yeast. Mol. Cell 30, 790–802 (2008).

26. Huh, W. K. et al. Global analysis of protein localization in budding yeast. Nature 425, 686–691 (2003).

27. Winzeler, E. A. et al. Functional characterization of the S. cerevisiae genome by gene deletion and parallel analysis. Science 285, 901–906 (1999).

28. Sikorski, R. S. & Hieter, P. A system of shuttle vectors and yeast host strains designed for efficient manipulation of DNA in Saccharomyces cerevisiae. Genetics 122, 19–27 (1989).

29. Christianson, T. W., Sikorski, R. S., Dante, M., Shero, J. H. & Hieter, P. Multifunctional yeast high-copy-number shuttle vectors. Gene 110, 119–122 (1992).

30. Kusch, J., Meyer, A., Snyder, M. P. & Barral, Y. Microtubule capture by the cleav-age apparatus is required for proper spindle positioning in yeast. Genes Dev. 16, 1627–1639 (2002).






L E T T E R S

Modularity of MAP kinases allows deformation of their signalling pathwaysAreez Mody1,2, Joan Weiner1 and Sharad Ramanathan1,2,3,4,5

Eukaryotic protein kinase pathways have both grown in number and changed their network architecture during evolution. We wondered if there are pivotal proteins in these pathways that have been repeatedly responsible for forming new connections through evolution, thus changing the topology of the network; and if so, whether the underlying properties of these proteins could be exploited to re-engineer and rewire these pathways. We addressed these questions in the context of the mitogen-activated protein kinase (MAPK) pathways. MAPK proteins were found to have repeatedly acquired new specificities and interaction partners during evolution, suggesting that these proteins are pivotal in the kinase network. Using the MAPKs Fus3 and Hog1 of the Saccharomyces cerevisiae mating and hyper-osmolar pathways, respectively, we show that these pivotal proteins can be re-designed to achieve a wide variety of changes in the input-output properties of the MAPK network. Through an analysis of our experimental results and of the sequence and structure of these proteins, we show that rewiring of the network is possible due to the underlying modular design of the MAPKs. We discuss the implications of our findings on the radiation of MAPKs through evolution and on how these proteins achieve their specificity.

Kinase pathways form one example of protein-protein interaction net-works1. These networks have grown, changing their topology and con-stituent proteins, during evolution2. Whereas previous analyses have focused on the conservation properties of nodes in these networks3, it is the ability of certain nodes to consistently change connections that could allow the network topology to be malleable. Identifying such piv-otal nodes in particular signalling networks may enable us to re-engineer these biological circuits with ease.

Here, we study the network of MAPK pathways that transduce differ-ent signals and regulate stress response and growth. We find that MAPK proteins are the pivotal nodes that allow their pathways to acquire new components and connections. Through a combination of sequence analysis and experiments, we show that we can continuously change

the topology of the MAPK network and its signal processing capabilities by redesigning the MAPK proteins.

For our experiments, we chose two of the five MAPK pathways in S. cer-evisiae4 that are involved in mating and responding to high osmolarity (Fig. 1a). These two pathways involve the MAPK proteins Fus3 and Hog1. Since another MAPK Kss1 can partially substitute for Fus3 (ref. 5), fus3Δ kss1Δ cells are sterile and hog1Δ cells die after hyper-osmolar shock6.

To study the evolution of the MAPK pathways, we identified ortho-logues of their component proteins in fifteen yeast, six animal and one plant species (Figs 1b, 2; Supplementary Information Fig. S1a). From the fission yeast Schizosaccharomyces pombe7 to the higher eukaryotes, the number of MAPKs has increased from three to at least fifteen. As the MAPK family grows, many MAPK interaction partners appear de novo in different lineages. Except for Pbs2, orthologues of all the interaction partners of Fus3 and Hog1 (Fig. 1a) are found only in the yeast lineage, indicating that the specificities of these MAPKs for their partners devel-oped uniquely in yeast (Figs 1b, 2). This includes scaffolding proteins that nucleate these multi-protein complexes aiding MAPK specificities with their appropriate upstream and downstream partners8,9. Closely related MAPK paralogues and orthologues show distinct specificities, indicating that new interactions can be acquired with little sequence divergence. For example, in S. cerevisiae the paralogues Kss1 and Fus3 can be activated by Ste7, but only Fus3 can activate Far1 causing cell-cycle arrest10. Strikingly, we also found that orthologues had switched their specificity from one MAPK in the yeast to another MAPK in the animals (Fig. 2d), providing a clear example of network topology that has changed during evolution.

Despite this plasticity, the MAPKs have been more invariant through evolution than other members of their pathways (Fig. 1b). Which molec-ular properties have allowed the MAPKs to remain so conserved, while still having a pivotal role in allowing their pathways to rewire and acquire new components during evolution?

The MAPKs are small, compact globular proteins. As with their linear protein sequences, their three-dimensional structures are very similar to one another (Fig. 3). To understand how these proteins achieve their specificities, we first reviewed previous biochemical11,12, structural8,13,14 and peptide10,15 analyses of the roles of certain residues in Fus3 and

1FAS Center for Systems Biology, Harvard University, Cambridge, MA 02138, USA. 2Department of Molecular and Cellular Biology, Harvard University, Cambridge, MA 02138, USA. 3School of Engineering and Applied Sciences, Harvard University, Cambridge, MA 02138, USA. 4Harvard Stem Cell Institute, Havard University, Cambridge MA02138, USA.5Correspondence should be addressed to S.R. (e-mail: [email protected])

Received 24 October 2008; accepted 8 January 2009; published online 22 March 2009; DOI: 10.1038/ncb1856




L E T T E R S

Hog1 (Fig. 3c, d; Supplementary Information, Section 2). We found that, other than Asp 112 and His 113 in Fus3, these residues cannot function alone as specificity determinants because they are conserved in other MAPK paralogues and their orthologues in other species.

We extracted functional information on residues using a multiple sequence alignment consisting of orthologues of four S. cerevisiae MAPKs (16 Kss1, 15 Fus3, 15 Hog1 and 15 Slt2). The variable residues were almost exclusively on the surface of the proteins (Fig. 3a). Of these residues, we

Ste7

Fus3

Pbs2

Hog1

Ste

5

Hyper-osmolar shockPheromone

pFUS1 pSTL1

MAPKK

MAPK

a

Dig1 Dig2

Ste12

Msn2

Msn4

Hot1 Msn1Far1 Smp1 Sko1

C. alb

icansY

. lipolytica

S. p

omb

e

D. hansenii

A. gossyp

ii

A. gossypiiD. hansenii

C. albicansY. lipolytica

S. pombe

K. lactisS. kluyveri

K. waltii

C. glab

rataK

. lactisS

. kluyveri

K. w

altii

C. glabrata

H. sap

iens

S. cerevisae

S. p

aradoxus

S. m

ikataeS

. kudriavzevii

S. b

ayanus

S. castelli

H. sapiensS. cerevisae

S. paradoxusS. mikatae

S. kudriavzeviiS. bayanus

S. castelli

M. m

usculus

M. musculus

C. elegans

D. rerio

R. novegius

C. elegansD. rerio

R. novegius

A. thaliana

D. m

elanogastor

A. thalianaD. melanogastor

Hog1

0.24

0.15 0.34 0.19 0.17 0.08

Ste5

Pbs20.09

Fus3

Ste70.08

0.04

0.2 0.22 0.28

0.12

Hyper-osmolar shock

Pheromone

b

0.11

100

0

Am

ino

acid

iden

tity

(per

cent

age)

0.03

Ste12Msn2/4 Hot1Sko1Smp1 Msn1Far1 Dig1 Dig2

Figure 1 Conservation of downstream components in the pheromone and hyper-osmolar glycerol pathways in S. cerevisiae. (a) The pheromone signal results in phosphorylation of the MAPKK Ste7, which phosphorylates the MAPK Fus3 aided by a scaffolding protein Ste5 (ref. 23). Phosphorylated Fus3 in turn activates the transcription factor Ste12 by relieving its repression by Dig1 and Dig2 (ref. 24), and activates Far1 leading to cell-cycle arrest. Activated Ste12 initiates the transcription of mating-specific genes including FUS1. The stimulation of cells with an osmolyte (for example, sorbitol) results in phosphorylation of the MAPKK Pbs2, which phosphorylates the MAPK Hog1, which in turn activates several transcription factors (including Hot1). This initiates the transcription of several genes, including STL1. (b) Matrices corresponding to each protein in a encode

the pair-wise percentage identity between the amino-acid sequences of all orthologues of that protein. The matrix for Hog1 is enlarged and labelled. The outlined matrix element in the Hog1 matrix shows the percentage identity between the Hog1 orthologues in A.gossypii and S. castelli. As the matrices are symmetrical, only half of each is shown. Beside each matrix is the substitution ratio (N/S) measuring the evolution rate of each protein within the sensu strictu species. The MAPKs Fus3 and Hog1 are the most conserved elements of the pathways, as indicated by the predominance of red in their matrices and the low N/S values. Pbs2 is found in all the species, whereas Ste7 (a Pbs2 duplicate) is present only in yeast, as indicated by the completely black columns and rows in its matrix, corresponding to the higher eukaryotes. The remaining proteins in a only have orthologues in yeast.



L E T T E R S

computationally identified a subset putatively responsible for the differ-ences in specificities of Fus3 and Hog1 (Supplementary Information, Figs S2, S3a and Methods). Our sequence analysis suggests two charac-teristics of MAPKs: first, they are structurally robust to changes in most of their surface residues; second, distinct groups of surface residues stand

out on each MAPK as being responsible for specific interactions. The lat-ter finding was unexpected as it may be thought that structurally similar proteins of the same family will use the same surface regions to interact with their corresponding upstream or downstream partners. For instance, we found a group of residues (Pro 80, Phe 83, Glu 84 and Trp 348) that is

S. cerevisiae

S. paradoxus

S. mikatae

S. kudriavzevii

S. bayanus

S. castelli

C. glabrata

K. waltii

S. kluyverii

K. lactis

A. gossypii

D. hansenii

C. albicans

Y. lipolytica

S. pombe

H. sapiens

M. musculus

R. norvegius

D. rerio

D. melanogaster

C. elegans

A. thaliana

S0

S1

S9

S2

S3

S4

Sensu strictu

Gene pairsparental orthologue, duplicate

ZAP1,MSN4

PBS2, STE7

FUS3,KSS1

SLT2,SMK1

FAR1,STE5

HOT1, MSN1

DIG1, DIG2

RLM1,SMP1

MSN4, MSN2

Speciationevent

Newgene

Speciationevent

S2

FAR1

STE12

S1

S1

SKO1

S2HOT1

Fus3

Pbs2

Hog1 Slt2

Rlm1

ERK1/2

MEK1/2

ERK5

MEF2A

p38αMAPK

MAPKK

TF

Yeast Human, mouse and rat

ba

c

d

DIG1 S3

S2

S0

S0

S2

S2

S3

S9

S9

S9

Figure 2 Evolutionary history of MAPKs and their interacting partners. (a) The phylogenetic tree of 15 yeast species and 7 higher eukaryote species used in deducing gene duplication events. Each speciation event (S0,S1,S2,S3,S4…S9) leads to the introduction of a new or duplicate gene in the upper branch. (b) MAPK-interacting partners from Fig. 1a, which are introduced de novo into the yeast lineage, are shown alongside the speciation event that created them. (c) Gene duplications accounting for all the remaining interacting partners not in b. Many of these have the new genes in b as their parent. (d) The only two MAPK-interacting partners, Pbs2 and Rlm1, found in all the species have changed their specificities towards

MAPKs during evolution. In the yeast, the orthologue of the downstream transcription factor Rlm1 is activated by the orthologue of Slt2 (the MAPK of the hypo-osmolar pathway). In the higher eukaryotes, however, (for example human, mouse and rat), the Rlm1 orthologue (MEF2A) is activated by the Hog1 orthologue (p38α, refs 18, 25). A similar change occurs upstream. In yeast, the Pbs2 orthologue activates the Hog1 orthologue, whereas in human, mouse and rat the Pbs2 orthologue (MEK1/2) activates the Fus3 orthologue (ERK1/2, ref. 13). Thus despite being highly conserved, the MAPKs show flexibility in acquiring and swapping interaction partners through evolution.



L E T T E R S

conserved in Fus3, but absent in Hog1 (Fig. 3b). Other residue positions are well conserved in Hog1 but show great variability in Fus3, signifying neutral drift. Thus distinct patches may be important for the specific inter-actions of each MAPK with other members of its pathway (Supplementary Information, Figs S2, S3a). If MAPK plasticity stemmed from the underly-ing flexibility of these patches, then synthetic proteins containing different combinations of these patches should be able to deform the signalling pathways. A chimaera has been generated16 that, when expressed in mam-malian cell lines, directs a stress signal into mitotic output (Supplementary Information, Section 2), suggesting that such deformations may indeed be possible.

To investigate this possibility, we constructed proteins that contained residues from Fus3 and Hog1 in various combinations. We divided both

Fus3 and Hog1 into six segments (A, B, C, D, E and F for Fus3 and a, b, c, d, e and f for Hog1) linked by regions conserved in all the MAPKs (blue, Fig. 3c, d), and joined the genes coding for the 64 possible proteins com-posed of these segments (Supplementary Information, Methods Fig. S2). This achieved a combinatorial redistribution of the surface patches, while also ensuring that internal contacts along the sequence were preserved. The genes were driven by a FUS3 promoter. To test for pheromone out-put in response to either pheromone or osmolyte (sorbitol) exposure, the 64 hybrids on both low-copy (showing an expression within threefold of the native chromosomal gene; data not shown) and high-copy (show-ing an expression within eightfold of the native chromosomal gene; data not shown) plasmids were transformed into a haploid fus3Δ kss1Δ strain (Supplementary Information, Table S1). A fluorescent (GFP) reporter was

3.36

a

c

Ent

rop

y (b

its)

0

A

B

C

Fus3

Fus3

D314

D317Q107

D112H113

Y312

P315

G20

G22

C119

E

D

a

bc

d

f

e

D313

D316

A111N115

G30 G32

P314I116

Q120

E160D161

F

d

P80F83E84

Fus3

W348

ATP bindingpocket

Y182

T180

Y182

T180

Docking groove

Activation loop

Negatively chargedpatch

Hog1 (p38 M. musculus)

b

Figure 3 Sequence analysis of MAPKs Fus3 and Hog1. (a) The degree of variability, as measured from entropy, in a multiple sequence alignment of four MAPKs and their orthologues from 22 species (Supplementary Information, Data Analysis) is shown on the structure of Fus3. High entropy (red) positions are constrained to the surface, whereas low entropy (blue) positions are in the interior. (b) Residues on Fus3 deduced computationally as putatively being responsible for the differences in Fus3 and Hog1 specificities are shown highlighted on its steric surface. The protein structure on the left has the same orientation as that in a, and the structure on the right has been rotated 180o about the vertical axis. The residues are coloured according to the segment they belong to, with the same colours as used in the six segments in c and d. Some residues, such as P80, F83, E84 and W348

shown on the structure on the right, distinguish themselves by being present on Fus3 and not on Hog1. (c, d) The segments A/a, B/b, C/c, D/d, E/e and F/f in Fus3/Hog1 used to build the hybrid kinases are shown on the structures of Fus3 and p38α (mouse Hog1 orthologue), respectively. Five regions linking these segments, shown in blue, were chosen on the basis of their strict conservation in all the MAPKs in the multiple sequence alignment used in a. Specific residues previously identified in the literature are highlighted on both structures (Supplementary Information, Section 2). Most of these residues are common to both of the MAPKs and their orthologues from other species; however, they are shown highlighted on one or the other to reflect the MAPK in which they were identified. Thus, except for D112 and H113, these residues cannot be important for specificity.



L E T T E R S

placed under control of the native FUS1 promoter, a faithful reporter of the pheromone pathway (Supplementary Information, Fig. S3b, c). We measured the fold change of fluorescent protein after 2 h of pheromone or sorbitol exposure (Fig. 4; Supplementary Information, Fig. S5a). The strains expressing the hybrids were further assayed for their ability to mate (Fig. 4; Supplementary Information, Fig. S4a and Methods) and arrest their growth on pheromone exposure (Supplementary Table 2). To test for osmo-lar output in response to either input, hybrids on both low- and high-copy plasmids were transformed into a haploid hog1Δ strain with a fluorescent reporter placed under the native STL1 promoter, a faithful reporter of the osmolar pathway (Supplementary Information, Fig. S3b, d). We measured the fold change of fluorescent protein under the STL1 promoter after 2 h of pheromone or sorbitol exposure (Fig. 4; Supplementary Information, Fig. S5b). The strains with the hybrids were assayed for their ability to grow under high osmolarity conditions (Fig. 4; Supplementary Information, Fig. S4b, c). Hybrids were also assayed in a fus3Δ kss1Δ hog1Δ strain with distinct fluorescent reporters (YFP and CFP) under the FUS1 and STL1 promoters. Without the hybrids, this strain does not show any response to pheromone or sorbitol (Fig. S3e). We measured the time-series of the response for hybrids on low-copy plasmids (Supplementary Information, Figs S6, S7a), and when integrated under the FUS3 promoter in the chro-mosome (Supplementary Information, Fig. S7b).

Several hybrids show FUS1 promoter (pFUS1) activity in response to either stimulus (Fig. 4a). Another nine (ABcdeF, ABCDeF, aBCdEF, abC-DeF, AbCDEF, abCDEf, abCdEF, aBcdEF, AbcDEF) do so when expressed from high-copy plasmids (Supplementary Information, Figs S5a, S8). All low-copy hybrids that showed pFUS1 activity upon pheromone exposure also rescued the cells ability to mate (Fig. 4a). Unstimulated cells carrying ABcdEF and ABcdeF had constitutively active pFUS1 showing about 50% and 20%, respectively, of the fluorescence shown by pheromone stimu-lated wild-type cells. (Supplementary Information, Fig. S8 and Section 3). This is noteworthy, as producing phospho-mimicking mutations in the activation loop does not constitutively activate MAPKs as it does for the activating MAPK Kinases (MAPKK)17, 18, 19.

One hybrid expressed from a low-copy plasmid, and several hybrids expressed from a high-copy plasmid (Fig. 4a; Supplementary Information, Fig. S5a), also showed pFUS1 activity in response to sorbitol, in fus3Δ kss1Δ cells. Such cross-wiring could occur by either a direct or an indi-rect mechanism. A direct mechanism would involve the hybrid MAPK being directly activated by Pbs2 and in turn activating Ste12. There are two possible indirect mechanisms: the first is where the hybrid inhibits native Hog1 or Pbs2 while also performing the function of Fus3. This is because strains in which the osmolar pathway is interrupted at the level of Hog1 or Pbs2 (either by deletion or inhibition of these proteins) pro-miscuously channel the hyper-osmolar signal into a pheromone output (Supplementary Information, Fig. S3d). This promiscuous channelling happens upstream of the MAPK and is abolished by deleting Ste7 (ref. 20). The second indirect mechanism is one in which Hog1 fails to inhibit the hybrid, allowing the osmolar signal to leak through the pheromone pathway, again through Ste7. To discriminate between the direct and indirect mechanisms of cross-wiring, we assayed the hybrid MAPKs in a ste7Δ hog1Δ strain. Without the hybrids, a sorbitol stimulus does not get cross-wired into pheromone output in this strain (Information, Fig. S9a). Three hybrids (ABcdEF, ABcdeF and aBcdeF) showed pFUS1 activity on hyper-osmolar shock in the ste7Δ hog1Δ strain, suggesting that they were activated directly by Pbs2 (Fig. 4a; Supplementary Information,

0 0 2

abCdef (98.99)

abcdeF (81.31)

abCdeF (80.30)

aBcdeF(56.06)

5

a

b

Hybrid(percentage Hog1)

ABcDEF (98.99)

ABCdEF (91.41)

ABcdEF*(90.40)

ABcdEF*(90.40)

ABcdEF*(62.12)

aBCDEF (81.81)

aBcDEF (80.80)

aBcdeF (43.93)

ABCDEF (100)

Pheromone SorbitolHybrid (percentage Fus3)

10 200

Pheromone Sorbitol

0 10

aBcdEF (27.78)

Mating

Growth on 1M sorbitol

AbCdEf (47.47)

0 5 0 3

High copy plasmid

abcdef (100)

Low copy plasmid

pFUS1 fold change in a fus3 kss1 strain background.

Low copy plasmid

pSTL1 fold change in a hog1 strain background.

Figure 4 Several hybrid MAPKs function in vivo to faithfully transduce and cross-wire the pheromone and hyper-osmolar signals. (a) The mean fold change in pFUS1 activity (upper panel, measured by a fluorescent reporter) in ~100 induvidual fus3Δ kss1Δ haploid cells containing the hybrid MAPKs expressed from a low-copy plasmid, after stimulation for 2 h by 0.6 μM pheromone or 1 M sorbitol. The percentage of residues in the hybrid that belong to Fus3 and differ from Hog1 is shown in parentheses. Serial dilutions show the mating efficiency of fus3Δ kss1Δ MATa haploids containing the indicated hybrids. Error bars reflect the intrinsically large range of response in the cells and not the measurement of uncertainty in a single cell. One hybrid expressed in low-copy plasmid (upper panel), and two hybrids expressed in high-copy plasmid (lower panel) mediate a cross-wiring whereby sorbitol evokes pFUS1 activity. The persistence of this cross-wiring after deleting STE7 is indicated by ‘Δ’. Hybrids showing constitutive pFUS1 activity are marked with an asterisk. Sorbitol-induced pFUS1 activity of aBcdeF was measured in fus3Δkss1Δhog1Δ cells. (b) The mean fold change in pSTL1 activity measuring an osmolar response, in ~100 single hog1Δ haploid cells containing the hybrid MAPKs and stimulated in the same manner as in a. The axes for fold change are scaled differently for the pheromone and sorbitol stimuli. Serial dilutions show the efficient growth of hog1Δ strains carrying the hybrid on plates containing 1 M sorbitol. Two hybrids mediated a cross-wiring showing pSTL1 activity on pheromone exposure (Supplementary Information, Data Analysis). The pheromone induced pSTL1 activity of aBcdEF was measured in fus3Δ kss1Δ cells.



L E T T E R S

Fig. S9b). For the remaining hybrids, cross-activation was abolished in the ste7Δ hog1Δ strain, implying that the cross-wiring occurred indirectly (Supplementary Information, Fig. S9c).

When expressed from a low-copy plasmid in fus3Δ kss1Δ hog1Δ cells, only the hybrid aBcdeF is capable of cross-wiring a sorbitol input into a pheromone output. Cells carrying aBcdeF are insensitive to pheromone

aBcdeFSorbitolPheromone

AbCdEf

c

a

b

DIC

DIC

aBcdeF

SorbitolPheromone

AbCdEfSorbitolPheromone SorbitolPheromone

ABcdEFSorbitolPheromone

ABcdEFSorbitolPheromone

aBcdeF

Sorbitol

AbCdEf

SorbitolPheromone

ABcdEF

SorbitolPheromone

pFUS1pFUS1

abcdef

Hog1

abcdeF abCdef

aBcdeF abCdeF

abCDeF AbCdEfABcdeFabcDEF aBcdEF abCdEF

aBCdEFaBcDEF ABcdEF

aBCDEF ABCDeF ABCdEFABcDEF

ABCDEF

Fus3

AbCDEF

AbcDEF

abCDEf

pFUS1 signal

pSTL1 signal and growth on sorbitol

Growth on sorbitol

d

Low-copy plasmid response

pSTL1

pFUS1Pheromone Sorbitol

pFU

S1

pSTL1

pS

TL1

pSTL1

Figure 5 Modular design allows some hybrid MAPKs to be activated by either input and others capable of activating either output. The protein space between extant Fus3 and Hog1 is rich in function. (a, b) Differential interference contrast (DIC) and fluorescence images of cells containing three hybrids: aBcdeF and AbCdEf on a low-copy plasmid and ABcdEF on a high-copy plasmid. Cells were stimulated with pheromone (left) and sorbitol (right). Green fluorescence (a) shows pFUS1 activity in fus3Δ kss1Δ cells (which, except for aBcdeF, have an intact osmolar pathway) mediated by the hybrid on pheromone or sorbitol stimulation. Red fluorescence (b) shows the pSTL1 activity in hog1Δ cells (which have an intact pheromone pathway) mediated by the hybrid on stimulation by pheromone or sorbitol. (c) aBcdeF interacts upstream only with the MAPKK Pbs2 of the osmolar pathway, but can interact downstream with transcription factors of both pathways (Ste12, Hot1). Thus sufficiently non-overlapping groups of residues on the original MAPKs must be responsible for these downstream interactions. Analogously, AbCdEf interacts downstream only with Hot1 but interacts upstream with the

MAPKKs Ste7 and Pbs2. Therefore sufficiently non-overlapping groups of residues are also responsible for these upstream interactions. (d) Functional properties of each hybrid are summarized schematically in a 2 × 2 matrix. The hybrids are arranged from Hog1 (top) to Fus3 (bottom). Hybrids in each successive row below Hog1 have one more segment taken from Fus3. The hybrids discussed in a, b and c are underlined. Each element of the matrix denotes the hybrids ability to mediate one of four distinct input-output characteristics in a cell. Top left (green) indicates pheromone response to pheromone stimulus. Top right (green) indicates pheromone response to sorbitol stimulus. Bottom left (red) indicates sorbitol response to pheromone stimulus. Bottom right (red) indicates sorbitol response to sorbitol stimulus. The bottom right is grey for three hybrids rescuing growth on 1 M sorbitol plates without showing a transcriptional response (Supplementary Information, Fig. S4b, c). Our sampling of the intervening MAPK protein space between Hog1 and Fus3 reveals several continuous paths via functioning intermediates.



L E T T E R S

(Fig. 5a, b), implying that this hybrid is directly activated by Pbs2 but not by Ste7. The cross-wiring persisted in the ste7Δ hog1Δ strain. The cross-wiring only occurred in the absence of native HOG1 when aBcdeF was expressed from a low-copy plasmid. Over-expression from a high-copy plasmid rescued this cross-talk in the presence of native HOG1, suggesting either competitive binding to Pbs2 between aBcdeF and Hog1 or direct inhibition of aBcdeF by Hog1.

Several hybrids, when expressed from a low-copy plasmid, show STL1 promoter (pSTL1) activity in response to either input (Fig. 4b; Supplementary Information, Fig. S9d). The hybrid AbCdEf, which faithfully transduces the sorbitol signal almost as efficiently as wild-type Hog1, also cross-wires the pathways by showing pSTL1 activity in response to pheromone expo-sure (Fig. 5a, b). Another hybrid, aBcdEF, achieves similar cross-wiring but is insensitive to sorbitol itself (Supplementary Information, Fig. S9d). The cross-wiring by aBcdEF occurred only in the absence of native Fus3 and Kss1, suggesting a competitive binding to Ste7 between aBcdEF and Fus3 or Kss1. All low-copy hybrids that showed pSTL1 activity on sorbitol exposure also rescued the ability of the cells to grow under high osmolar conditions (Fig. 4b). Interestingly, certain hybrids rescued the growth of the cells on a high osmolar medium (Supplementary Information, Fig. S4b, c), but did not mediate reporter activity in the cell (Fig. 4b; Supplementary Information, Fig. S5b).

Besides demonstrating the flexibility of MAPKs, our results allow us to draw some conclusions about the modularity and specificity of these proteins.

We find that the MAPK proteins are modular. One in three hybrids were functional, some with as many of their distinguishing residues taken from Fus3 as from Hog1. This strongly suggests that the common conserved resi-dues at the core are sufficient for folding, and the variable exterior residues, where most of the evolution has occurred, control specificity. However, despite their having similar structures and symmetrical roles in their respec-tive pathways, the patches of surface residues used by the paralogues Fus3 and Hog1 to interact with their up and downstream factors seem substan-tially different. This is highlighted by the hybrids (Fig. 5a, b, c ) that are able to interact with both upstream partners, such as ABcdEF (which is one of three hybrids in which sorbitol evoked a pheromone response through Pbs2), or both downstream partners (aBcdeF). The hybrid aBcdeF yields both an osmolar and pheromone response to an osmolar input, but is non-responsive to pheromone (Fig. 5a), implying that it can be activated by Pbs2 alone but can activate both Ste12 and Hot1. As every segment is either from Fus3 or Hog1, the regions on Fus3 and Hog1 that interact with Ste12 and Hot1 must be considerably non-overlapping. Not only do we find that these MAPKs show a modular design, we also note that this design is implemented differently in the two MAPKs in spite of their great structural similarity. Thus, conformational change, which commonly underlies enzyme promis-cuity in other synthetic proteins21, is unlikely to be the mechanism by which our hybrid kinases achieved their various cross-wirings and specificities.

We can uncover how MAPK proteins achieve specificity. Our results indicate that the segment BEF is required in conjunction with the seg-ments A or D for pheromone input to invoke pheromone output (Fig. 4a). However, as aBcdeF activates the pheromone pathway only on osmolar input, and not on pheromone input, it appears that BF alone is sufficient for pheromone output and that E is important for pheromone input. The segment BF contains all the residues associated with the ‘docking domain’ (Fig. 3c; Supplementary Information, Section 2). These were identified and studied mainly in the context of Fus3 binding to Ste7 and to phosphatases10.

However, that BF appears sufficient for pheromone output suggests it is more important in vivo for the Fus3–Ste12 interaction, whereas E is more important for the Fus3–Ste7 interaction. The patch of residues (Pro 80, Phe 83, Glu 84 and Trp 348) identified by our sequence analysis to be unique to Fus3 is contained in BF. The hybrid aBCDEF transduces the phe-romone signal to induce high pFUS1 activity but fails to mediate cell-cycle arrest (Supplementary Table 2). Consistently, the strains carrying aBCDEF have a low mating efficiency (Supplementary Information, Fig. S4a). This implicates segment A in Fus3–Far1 interaction. Segment d is necessary for transducing a high osmolar signal into any output (Fig. 4a b). In Fus3, D has a disordered structure and most of its residues undergo neutral drift. This again illustrates how members of this family have specialized different residue patches on their surfaces to achieve analogous specificities rather than refining the residues on some common catalytic loop22.

Our data show that these proteins can find new specificities with rela-tively few changes in their sequence (as few as seven point-mutations). They can retain their original function while acquiring new interac-tion partners. This might be why they have increased in number, hav-ing repeatedly found new functions after duplication (Supplementary Information, Fig. S10). Indeed our results suggest promiscuous inter-mediates through which duplicated enzymes could evolve, en route to their new specificities22. We have also gained insights into the specificity determinants of extant proteins. Exploration building on our hybrids can complement traditional biochemical techniques, which are challenged by the transient nature of kinase–substrate interactions. The implication for synthetic design is that pivotal proteins such as these may serve as the best templates, or scaffolds, with which to design new specificities to create new connections in existing pathways.

METHODS Orthology and duplication. To construct the phylogeny and search for duplications involving a gene extant in S. cerevisiae, we found the best reciprocal BLAST hit in each species. The orthology relationship was established by confirming that orthologues in species sharing the same nearest common ancestor with S. cerevisiae had similar distances from the S. cerevisiae protein, and by noting the progressive increase in this distance for orthologues in species whose nearest common ancestor was further away (Supplementary Information, Fig. 1a). For the genes, putative orthologues were found in most species in one branch of a speciation event and not in the other, thereby making it obvious if, and in which, species this gene was lost. A simple indicator of the possibility of a duplication event was two proteins in S. cerevisiae having separate orthologues in all species on one side of a speciation event, but BLASTed to the same protein in species on the other side. Rather than multiple gene-loss events, a single duplication at the speciation event is the most parsimonious explanation. Again, the duplication was verified by viewing the protein sequences and BLAST scores in rela-tion to the species tree (see Supplementary Information for more details).

Computational identification of variable and putative specificity residues in MAPKs. To identify variable residues, the entropy at a given position j (Fig. 3a ; Supplementary Information, Fig. S1c) was computed from the multiple sequence alignment using S(j)=– fi(j) log(fi(j))Σ

20

i=1. Here, i indexes the 20 types of amino acid, and fi(j)

is the normalized frequency with which amino acid i occurs at position j.To identify the specificity patches of Fus3 and Hog1 in S. cerevisiae, we per-

formed a multiple alignment of Fus3 and Hog1 orthologues from the yeast species. Higher eukaryotic orthologues were deliberately excluded because the MAPK specificities in this lineage are different (Fig. 2d). Fus3 (Hog1) orthologue residues appearing directly above or below each other at a given position in the alignment were scored pair-wise by the BLOSUM50 matrix, and the sum of these scores was taken as indicative of the level of conservation at that position in Fus3 (Hog1). The linear ordering of the orthologues, on which this number depends, was taken to be that best representing the evolutionary relationships of the underly-ing phylogenetic tree. A plot of the score versus positions on Fus3 (Supplementary Information, Fig. 1b; sorted by increasing scores) shows a conspicuous plateau at



L E T T E R S

a score of 42, indicating a natural cut-off point above which a position is declared to be ‘well conserved’. For a position to qualify as putatively endowing specificity to Fus3 it also had to be different at the corresponding position in Hog1. This difference was quantified by summing the pair-wise BLOSUM50 scores for the Fus3 and Hog1 residues taken from orthologues belonging to the same species. A negative score was taken to signify a difference.

Yeast strains, plasmids and growth conditions. See Supplementary Table 1 for the list of strains and plasmids used in this study. Cells transformed with plasmids were grown overnight at 30 oC in a selective medium of SC-URA, re-inoculated into fresh medium for 8 h of exponential growth and then incubated with phe-romone (0.6 μM) and sorbitol (1 M). The time-dependence of the reporter activity was measured using fluorescence microscopy (Fig. 4 reports the mean fluores-cence after 2 h; Supplementary Information, Figs S6, S7a).

Construction of hybrid MAPKs. See Supplementary Information, Fig. S2 for the amino-acid sequences of the six segments A/a, B/b, C/c, D/d, E/e and F/f in Fus3/Hog1. The base-pair sequences of the hybrids were designed ab initio as follows: Codons for the common conserved regions separating the segments were selected to create unique restriction sites recognized by five different restriction enzymes. Achieving this in the region separating E/e and F/f, where the amino-acid sequences of Fus3 and Hog1 are not identical, effectively introduced two mutations in the amino-acid sequence of Fus3: Gln 287, Arg 288 to Glu 287, Lys 288, forcing them to agree with the corresponding residues of Hog1. The substitution of Gln to Glu is seen in the closely related sensu strictu yeast Saccharomyces kudriavzevii, and the substitu-tion Arg to Lys is extremely conservative as both Arg and Lys are positively charged. All other codons in the genes were optimized for expression in S. cerevisiae.

Microscopy. Cells were observed using a Zeiss 200M fluorescent microscope and Orca-II-ER and Hamamatsu EM-CCD cameras. Objectives used were: ×100 with N.A. = 1.45 and ×63 with N.A. = 1.4. Emission of CFP was visualized at 470 nm (30-nm bandwidth) on excitation at 430 nm (25-nm bandwidth). Emission from YFP was collected at 535 nm (30-nm bandwidth) on excitation at 500 nm (20-nm bandwidth). Emission from GFP was collected at 528 nm (38-nm bandwidth) on excitation at 490 nm (20-nm bandwidth).

Mating assay. fus3Δ kss1Δ MATa strains containing the hybrid MAPKs were grown in a selective medium to exponential phase, and then plated on YPD simul-taneously with a MATα strain, also grown to exponential phase. By measuring the cell density of the two cultures before plating, an equal number of cells of each mating type was plated. After 36 h, cells scraped off the resulting lawn were inoculated in YPD for 2 h before plating a threefold serial dilution of 10 μl spots on plates selected for diploids (Supplementary Information, Fig. 4a). As only the MATa strains were autotrophic for the amino acids histidine and tryptophan, and only the MATα strain contained a drug-resistant gene (resistant to Kanamycin), plates lacking histidine and tryptophan and containing the drug were used to select for diploids. Again, the cell density was carefully measured to ensure that the same initial density of cells was used for plating the first spot.

Image processing. The mean fold changes in fluorescence (Fig. 4) were computed by dividing the increase in fluorescence levels averaged over ~100 stimulated cells by the mean fluorescence levels averaged over unstimulated cells. The pheromone-induced pSTL1 fold change in cells carrying AbCdEf and aBcdEF was computed (Fig. 4b) using data from a selected fraction of the cells imaged, as approximately 50% of the cells carrying AbCdEf and about 5% of the cells carrying aBcdEF showed a response.


AcknoWledgeMentSWe thank B. Stern for critical readings and suggestions, H. Dohlman, A. Drummond, M. DePristo, A. Murray, D. Huse, D. Fisher, M. McClean, M. Thomson and I. Nachman for discussions and comments, and R.e Hellmiss for help with figures. Work was supported by an NIH grant (2P50GM068763).

AuthoR contRibutionSA. M. and S. R. conceived and planned the project and wrote the manuscript. A.M. designed and implemented the computational aspects of the project, performed the microscopy and analysed the data. J.W. and A.M. performed the strain and plasmid construction.

coMpeting finAnciAl inteReStSThe authors declare no competing financial interests.


1. Manning, G., Plowman, G. D., Hunter, T. & Sudarsanam, S. Evolution of protein kinase signaling from yeast to man. Trends Biochem. Sci. 27, 514–520 (2002).

2. Stelling, J., Sauer, U., Szallasi, Z., Doyle, F. J. & Doyle, J. Robustness of cellular func-tions. Cell 118, 675–685 (2004).

3. Wuchty, S., Oltvai, Z. N. & Barabasi, A. L. Evolutionary conservation of motif constitu-ents in the yeast protein interaction network. Nature Genet. 35, 176–179 (2003).

4. Banuett, F. Signalling in the yeasts: an informational cascade with links to the filamen-tous fungi. Microbiol. Mol. Biol. Rev. 62, 249–274 (1998).

5. Sabbagh, W. Jr, Flatauer, L. J., Bardwell, A. J. & Bardwell, L. Specificity of MAP kinase signaling in yeast differentiation involves transient versus sustained MAPK activation. Mol. Cell 8, 683–691 (2001).

6. Giaever, G. et al. Functional profiling of the Saccharomyces cerevisiae genome. Nature 418, 387–391 (2002).

7. Sipiczki, M. Where does fission yeast sit on the tree of life? Genome Biol. 1, R1011 (2000).

8. Bashor, C. J., Helman, N. C., Yan, S. & Lim, W. A. Using engineered scaffold interactions to reshape MAP kinase pathway signaling dynamics. Science 319, 1539–1543 (2008).

9. Bhattacharyya, R. P. et al. The Ste5 scaffold allosterically modulates signaling output of the yeast mating pathway. Science 311, 822–826 (2006).

10. Remenyi, A., Good, M. C., Bhattacharyya, R. P. & Lim, W. A. The role of docking inter-actions in mediating signaling input, output, and discrimination in the yeast MAPK network. Mol. Cell 20, 951–962 (2005).

11. Bardwell, L., Cook, J. G., Chang, E. C., Cairns, B. R. & Thorner, J. Signaling in the yeast pheromone response pathway: specific and high-affinity interaction of the mitogen-activated protein (MAP) kinases Kss1 and Fus3 with the upstream MAP kinase kinase Ste7. Mol. Cell. Biol. 16, 3637–3650 (1996).

12. Tanoue, T., Adachi, M., Moriguchi, T. & Nishida, E. A conserved docking motif in MAP kinases common to substrates, activators and regulators. Nature Cell Biol. 2, 110–116 (2000).

13. Chang, C. I., Xu, B. E., Akella, R., Cobb, M. H. & Goldsmith, E. J. Crystal structures of MAP kinase p38 complexed to the docking sites on its nuclear substrate MEF2A and activator MKK3b. Mol. Cell 9, 1241–1249 (2002).

14. Goldsmith, E. J., Cobb, M. H. & Chang, C. I. Structure of MAPKs. Methods Mol. Biol. 250, 127–144 (2004).

15. Bardwell, A. J., Abdollahi, M. & Bardwell, L. Docking sites on mitogen-activated protein kinase (MAPK) kinases, MAPK phosphatases and the Elk-1 transcription factor compete for MAPK binding and are crucial for enzymic activity. Biochem. J. 370, 1077–1085 (2003).

16. Brunet, A. & Pouyssegur, J. Identification of MAP kinase domains by redirecting stress signals into growth factor responses. Science 272, 1652–1655 (1996).

17. Mansour, S. J., Candia, J. M., Gloor, K. K. & Ahn, N. G. Constitutively active mitogen-activated protein kinase kinase 1 (MAPKK1) and MAPKK2 mediate similar transcrip-tional and morphological responses. Cell Growth Differ. 7, 243–250 (1996).

18. Zhang, J., Zhang, F., Ebert, D., Cobb, M. H. & Goldsmith, E. J. Activity of the MAP kinase ERK2 is controlled by a flexible surface loop. Structure 3, 299–307 (1995).

19. Diskin, R., Askari, N., Capone, R., Engelberg, D. & Livnah, O. Active mutants of the human p38alpha mitogen-activated protein kinase. J. Biol. Chem. 279, 47040–47049 (2004).

20. O’Rourke, S. M. & Herskowitz, I. The Hog1 MAPK prevents cross talk between the HOG and pheromone response MAPK pathways in Saccharomyces cerevisiae. Genes Dev. 12, 2874–2886 (1998).

21. Khersonsky, O., Roodveldt, C. & Tawfik, D. S. Enzyme promiscuity: evolutionary and mechanistic aspects. Curr. Opin. Chem. Biol. 10, 498–508 (2006).

22. Glasner, M. E., Gerlt, J. A. & Babbitt, P. C. Evolution of enzyme superfamilies. Curr. Opin. Chem. Biol. 10, 492–497 (2006).

23. Choi, K. Y., Satterberg, B., Lyons, D. M. & Elion, E. A. Ste5 tethers multiple protein kinases in the MAP kinase cascade required for mating in S. cerevisiae. Cell 78, 499–512 (1994).

24. Gelli, A. Rst1 and Rst2 are required for the a/alpha diploid cell type in yeast. Mol. Microbiol. 46, 845–854 (2002).

25. Yang, S. H., Galanis, A. & Sharrocks, A. D. Targeting of p38 mitogen-activated protein kinases to MEF2 transcription factors. Mol. Cell. Biol. 19, 4028–4038 (1999).






L E T T E R S

STAT3 inhibition of gluconeogenesis is downregulated by SirT1Yongzhan Nie1,7, Derek M. Erion4, Zhenglong Yuan5, Marcelo Dietrich1, Gerald I. Shulman4, Tamas L. Horvath1,2,3,8 and Qian Gao1,6,8

The fasting-activated longevity protein sirtuin 1 (SirT1, ref. 1) promotes gluconeogenesis in part, by increasing transcription of the key gluconeogenic genes pepck1 and g6pase2,3, through deacetylating PGC-1α and FOXO1 (ref. 4). In contrast, signal transducer and activator of transcription 3 (STAT3) inhibits glucose production by suppressing expression of these genes5,6. It is not known whether the inhibition of gluconeogenesis by STAT3 is controlled by metabolic regulation. Here we show that STAT3 phosphorylation and function in the liver were tightly regulated by the nutritional status of an animal, through SirT1-mediated deacetylation of key STAT3 lysine sites. The importance of the SirT1–STAT3 pathway in the regulation of gluconeogenesis was verified in STAT3-deficient mice in which the dynamic regulation of gluconeogenic genes by nutritional status was disrupted. Our results reveal a new nutrient sensing pathway through which SirT1 suppresses the inhibitory effect of STAT3, while activating the stimulatory effect of PGC-1α and FOXO1 on gluconeogenesis, thus ensuring maximal activation of gluconeogenic gene transcription. The connection between acetylation and phosphorylation of STAT3 implies that STAT3 may have an important role in other cellular processes that involve SirT1.

The transcription factor STAT3 participates in various critical cel-lular processes7. In the liver, STAT3 is known to suppress expression of the transcriptional co-activator of gluconeogenesis PGC-1α and to suppress gluconeogenic genes. Ectopic expression of STAT3 in lep-tin receptor mutant (lepr–/–) mice reduces PGC‑1α transcription and reverses diabetes. This effect of STAT3 is abolished when Tyr 705 is mutated to a phenylalanine (Y705F; ref. 5). Other protein modifica-tions of STAT3, such as acetylation, have recently been reported8–10. However, the functional significance of STAT3 acetylation remains ill-defined, and its relationship with STAT3 tyrosine phosphorylation remains unclear.

We hypothesized that STAT3 acetylation regulates physiological processes by mediating changes in the STAT3 phosphorylation status. We found that STAT3 acetylation was decreased after a 24-h fast and increased after feeding in the livers of C57BL/6J mice. STAT3 tyrosine phosphorylation directly correlated with the level of STAT3 acetylation, indicating that the reversible acetyl-modifications are functionally rel-evant (Fig. 1a). The observation that both acetylation and phospho-rylation of STAT3 were evident in fed, but dramatically reduced in fasted, animals suggests that STAT3 acetylation and phosphorylation are actively downregulated on fasting. Overall, these observations sig-nify that cellular metabolic status regulates STAT3 function in the liver, presumably owing to the sensitivity of the liver to the overall nutritional status of the organism11.

SirT1 can be induced to promote gluconeogenesis11,12 under con-ditions of fasting. Therefore, we asked whether SirT1 affects fed/fast-regulated STAT3 acetylation and phosphorylation. We injected the SirT1 inhibitor EX527, which has shown increased potency and specificity for SirT1 (ref. 13), into C57Bl/6J mice. EX527 increased acetylation and phosphorylation of hepatic STAT3 (Fig. 1c). These results are similar to those seen in animals treated with nicotinamide (Supplementary Information, Fig S1a), a less specific SirT1 inhibitor. EX527 also increased the acetylation of p53, a known SirT1 substrate, suggesting that a reduction of SirT1 function was achieved with EX527 treatment (Supplementary Information, Fig. S4b)1. In addition to EX527, we used an antisense oligionucleotide (ASO)14,15 to knockdown hepatic SirT1 on a chronic basis. SirT1 ASO induced significant STAT3 acetylation and tyrosine phosphorylation (Fig. 1c). Together, these results support the idea that SirT1 is critically involved in hepatic STAT3 regulation.

Next, we studied the effect of nicotinamide and resveratrol (a SirT1 acti-vator) on STAT3 acetylation and phosphorylation in an SV40-transformed mouse hepatic cell line, previously used in gluconeogenic studies16,17. In cells treated with nicotinamide (0.3 mM and 3 mM), the levels of STAT3 acetylation and phosphorylation increased in a dose-dependent man-ner and were independent of the kinase JAK2 (Fig. 1d) and class I and

Departments of 1Comparative Medicine, 2Obstetrics, Gynecology and Reproductive Sciences, and 3Neurobiology, 4Howard Hughes Medical Institute, Yale University School of Medicine, New Haven, CT 06520, USA. 5Department of Surgery, Brown University Medical School-Rhode Island Hospital, Providence, RI 02903, USA. 6Nanjing University School of Medicine, Jiangsu Province, 210093, China. 7Current Address: State Key Laboratory of Cancer Biology and Xijing Hospital of Digestive Diseases, Fourth Military Medical University, Xi’an, Shaanxi, 710032, China. 8Correspondence should be addressed to T.L.H. or Q. G. (e-mail: [email protected]; [email protected])

Received 17 November 2008; accepted 13 January 2009; published online 22 March 2009; DOI: 10.1038/ncb1857





L E T T E R S

II HDACs (other cells were treated with trichostatin A, TSA; Fig. 1e). Conversely, resveratrol18 (0.2 μM) reduced STAT3 acetylation and phos-phorylation in the cultured hepatic cells (Fig. 1f). To determine whether deacetylation of STAT3 requires SirT1, we used SirT1 knockout (KO) and wild-type mouse embryonic fibroblasts19 (MEFs). First, we found that lev-els of STAT3 acetylation and phosophrylation were constitutively higher in the SirT1 KO MEFs than in the wild-type MEFs (Fig. 1g). Second, treat-ment with EX527 increased levels of STAT3 acetylation and phosphoryla-tion in wild-type MEFs, but not in SirT1 KO MEFs (Fig. 1g). Moreover, resveratrol decreased STAT3 acetylation and phosphorylation in wild-type

MEFs, but not SirT1 KO MEFs (Fig. 1h). Together, these results further indicate that deacetylation of STAT3 is dependent on SirT1.

To investigate STAT3 as a SirT1 substrate, we studied the role of ectopically expressed SirT1. We found that the level of STAT3 acetylation was greatly reduced by cotransfection of human (h) SirT1, in HEK293T cells (Fig. 2a). Moreover, the introduced hSirT1 was able to potently suppress p300/CBP-induced STAT3 acetylation, suggesting that both factors affected the same set of lysine residues of STAT3 (Fig. 2a). The deacetylation of STAT3 by hSirT1 was more effective than that by HDAC1, HDAC 3 (Fig. 2a) or HDAC2 (data not shown), which were previously implicated in STAT3 deacetylation10,20.

0

0.5

1.0

1.5

2.0

Fed Fasted Re-fed HFD

Rel

ativ

e fo

lds

Rel

ativ

e fo

lds

Rel

ativ

e fo

lds

Rel

ativ

e fo

lds

Rel

ativ

e fo

lds

Rel

ativ

e fo

lds

Rel

ativ

e fo

lds

0

0.5

1.0

1.5

2.0

Fed Fasted Re-fed HFD

0

0.5

1.0

1.5

2.0

Vehicle EX527

0

0.5

1.0

1.5

2.0

Vehicle EX527

0

0.5

1.0

1.5

Sirt1 ASO Control ASO

0

1.0

2.0


0

1.0

2.0

3.0


0

1.0

2.0

WT KO WTEX527

KOEX527

WT KO WTEX527

KOEX527

0

2

4

a– – – – + + + +

b

c

0 3

d

TSA (µM) TSA (µM)0

e

0 6.0

f0 200 (nM)

+ + + +

g

WT WTRes

KO KORes

h

Acetyl-K

Acetyl-K

A-K685

A-STAT3

A-STAT3

A-STAT3

A-STAT3

P-STAT3

P-STAT3

P-STAT3

P-STAT3

T-STAT3

T-STAT3

β-actin

IP STAT3 Input Fed Fast Re-fed HFD

A-STAT3

P-STAT3

T-STAT3

p-JAK2

p-JAK2

SirT1

β-actin

β-actin

β-actin

β-actin β-actin

EX527 *

*

*

*

*

*

*

**

**

SirT1

SirT1 SirT1

SirT1

A-STAT3

P-STAT3

A-STAT3

P-STAT3

A-STAT3

P-STAT3

T-STAT3

A-STAT3

P-STAT3

T-STAT3

T-STAT3

β-actin

Control Control

NAM (mM) 0.3 1.0 6.0

Ace

tyl l

ysin

e

Res

SirT1

A-STAT3

P-STAT3

T-STAT3

SirT1

A-STAT3

P-STAT3

T-STAT3

SirT1

– – – – – –EX527

SirT1KO

SirT1KOWTWT

Figure 1 SirT1 is involved in regulating STAT3 acetylation. (a) STAT3 acetylation and phosphorylation changed in mouse livers under different nutritional conditions. Male C57Bl/6J mice were fed (normal laboratory chow), fasted (24 h), re-fed (24 h, normal laboratory chow after fasting) or fed with a high-fat diet (HFD, 48 h). The levels of STAT3 acetylation and phosphorylation in livers were determined by immunoprecipitation and/or western blot analysis. (b) EX527 induced acetylation and phosphorylation of STAT3. Fasted male C57Bl/6J mice were injected with EX527 (i.p. 10 mg per kg body weight) 6 h before being killed. (c) SirT1 ASO induced hepatic STAT3 acetylation and phosphorylation. C57Bl/6J mice were injected with ASO (i.p. 10 per kg body weight) five times in a 20-day period. (d–f) SV40-transformed mouse hepatic

cells were treated with different doses of nicotinamide (NAM, d), trichostatin A (TSA, e) and resveratrol (Res, f). STAT3 acetylation levels were significantly increased by nicotinamide, but reduced by resveratrol. No change was observed with TSA. The significant increase of total protein acetylation shown by the pan-anti-acetylated lysine antibody indicated TSA was effective (f, left panel). (g, h) Decetylation of STAT3 is dependent on SirT1 in vitro. EX527 (10 μM for 6 h) increased STAT3 acetylation and phosphorylation in wild-type MEFs, but not in SirT1 KO MEFs (g). Data are mean ± s.e.m. of three repeated experiments, n = 2 cells. Resveratrol (100 nM for 6 h) decreased STAT3 acetylation and phosphorylation in wild-type MEFs, but not in SirT1 KO MEFs (h). Data are mean ± s.e.m., n = 5 mice * P < 0.05, ** P < 0.01 in a, b and c.



L E T T E R S

To test whether SirT1 and STAT3 formed a complex in cells, HEK293T cells were co-transfected with the wild-type genes of hSirT1 and hSTAT3. Exogenous (Fig. 2b) and endogenous (Fig. 2c) hSirT1 proteins were detected in immunoprecipitation products, suggesting that STAT3 and hSirT1 formed complexes. Next, endog-enous STAT3 was co-precipitated by endogenous hSirT1 (Fig. 2d). To determine whether this physical interaction between hSirT1 and STAT3 occurs in vivo, we examined their association in mouse liver. SirT1 was observed in the STAT3 immunoprecipitation products, suggesting an interaction between endogenous STAT3 and SirT1 in the liver of fasted animals (Fig. 2e), and less so in the liver of fed animals (Fig. 2f).

To determine the region(s) in STAT3 responsible for SirT1 binding, a set of five c-Myc tagged STAT3 deletion mutants (T1 to T5) were generated that were designed to test each of the domains in STAT3. The DNA binding and linker domain (from amino acid 330–590) of STAT3 (ref. 21) was found to be the key region involved in the STAT3– SirT1 interaction (Fig. 2g). The various truncated STAT3s that did not contain the DNA binding and linker domain failed to form complexes with SirT1, whereas the truncated STAT3s that contained this domain pulled down SirT1 (Fig. 2g). Moreover, the lat-ter had greatly reduced acetylation and phosphorylation on cotransfection with SirT1 (data not shown), suggesting that a direct interaction between the STAT3 DNA binding and linker domain and SirT1 is required for the enzymatic function of SiT1 on STAT3.

0

2

4

6

8

Ctl HDAC1 HDAC3 SirT1

Rel

ativ

e fo

lds

of L

uc

Rel

ativ

e fo

lds

of L

uc0

1

2

3

4

Ctl HDAC1 HDAC3 SirT1

–+–+++––

+++++++

a

+

e

IP

d

IPIP

b

IP

c

f g

T1T2T3T4T5

Myc

T1 T2 T3 T4 T5 FULLIP: c-Myc

330509

MycMycMyc

MycMyc

anti-Myc

h i

j kWT H1 H3

****

SirT1 anti-SirT1 anti-SirT1

anti-SirT1

anti-SirT1

anti-SirT1

SirT1

SirT1

SirT1SirT1

SirT1SirT1

SirT1 H1 H1 H3 H3

STAT3

STAT3

anti-STAT3 anti-STAT3

anti-STAT3

anti-STAT3

A-STAT3

A-STAT3

A-STAT3

A-STAT3

P-STAT3

P-STAT3

P-STAT3

IP STAT3T-STAT3

T-STAT3

A-STAT3

P-STAT3

T-STAT3

T-STAT3

T-STAT3

P300

Flag–H1

Flag–H3

Fast Fed Input

– +

– ++ +

β-actin

A-STAT3

P-STAT3

T-STAT3

β-actin

Inpu

t

mIg

G

anti-

c-m

yc

Inpu

t

Inpu

t

mIg

G

r-IgGan

ti-c-

myc

anti-

SirT

1an

ti-Si

rT1

Inpu

t

r-IgG

anti-

Stat

3an

ti-St

at3

ND

0 130 320 465 585 688 770NH2

CCD DBD LD SH2D TADCOOH

668664

586

p100

p75

p50

p37

p25

FULL

Input

β-actin β-actin

Flag–H1Flag–H3

SirT1

Flag–H1Flag–H3

SirT1

SirT1 WT H1 H3 SirT1

Figure 2 STAT3 phosphorylation and transactivation were downregulated by SirT1. (a) SirT1 deacetylates STAT3 in cultured cells. The effect of p300, SirT1 or HDACs on STAT3 acetylation was measured in transfected HEK293T cells (H1, HDCA1; H3, HDCA3). (b–e) SirT1 and STAT3 form complexes in vitro and in vivo. HEK293T cells were transfected with SirT1 and STAT3, or STAT3 alone (b, c). The physical interactions between exogenous STAT3 and exogenous (b) or endogenous (c) SirT1 were detected. In HEK293T cells, endogenous SirT1 and endogenous STAT3 were co-precipitated (d). STAT3 and SirT1 were co-precipitated by a STAT3 antibody in the livers of wild-type male mice (n = 4, e). (f) The physical interaction between STAT3 and SirT1 was enhanced in fasting livers. (g) Except for full-length STAT3, SirT1 was only precipitated by truncated STAT3-T2 and -T3, suggesting that both the DNA binding and the linker domains of STAT3 are involved in the interaction of STAT3 and SirT1. ND, N-terminal domain; CCD, coil-coil domain; DBD, DNA binding domain;

LD, linker domain; SH2D, SH2 domain and TAD, transactivation domain. (h, i) SirT1-mediated deacetylation of STAT3 affects Y705-STAT3 phosphorylation. HEK293T cells were transfected with SirT1 and STAT3 (0.25 μg per well of each, h), or SirT1 alone (i), in 12-well-plates. (i) Each well was loaded with 100 μg of total protein to visually present the signals of endogenous A-STAT3 and P-STAT3. (j, k) The SirT1-mediated deacetylation of STAT3 affected STAT3 function (WT, wild-type). A2780 cells were treated with IL-6 (40 ng ml–1) for 12 h. A relatively low level of the SirT1, HDAC1 and HDAC3 plasmids (0.05 μg per well) were either transfected with STAT3 (0.1 μg per well) or untreated (control). The effects on either exogenous (20 μg per well, j) or endogenous (60 μg per well, k) STAT3 acetylation and phosphorylation were determined. STAT3 transactivation activities were detected by a STAT3 specific luciferase reporter (Luc). Data are mean ± s.e.m. of three repeated experiments, ** P < 0.01 in J and k.



L E T T E R S

We next analysed whether the acetylation state of STAT3 affects STAT3 phosphorylation and transactivation function. After hSirT1 overexpres-sion, the phosphorylation of exogenous (Fig. 2h) and endogenous (Fig. 2i)

STAT3 was again downregulated with high efficiency. Low levels of exog-enous hSirT1 DNA (0.05 μg) resulted in a reduction of exogenous and endogenous STAT3 acetylation and phosphorylation (Fig. 2j, k), whereas

0.0

1.0

2.0

3.0

3.5

2.5

1.5

0.5

WT

K67

9R

K68

5R

K70

7R

K70

9R

K67

9-78

5R

K70

7-70

9R

4K/R

Y70

5F

Vect

or

WT

K67

9R

K68

5R

K70

7R

K70

9R

K67

9-78

5R

K70

7-70

9R

4K/R

Y70

5F

Vect

or

Rel

ativ

e fo

lds

of L

uc

0

4

8

12

16

20

STAT3 STAT3+P300

Rel

ativ

e fo

lds

of lu

c

WT

4K/R

Y705F

0

12

3

4

56

7

8

WT 4K/R Y705FR

elat

ive

fold

s of

Luc

0

20

40

60

80

100

WT 4K/R Y705F

Nuc

lear

loca

lizat

ion

(per

cent

age)

+– – +

a c

d

e

f

g

+ +WT 4K/R WT 4K/R

h i

SirT1

A-STAT3

P-STAT3

T-STAT3

A-STAT3

P-STAT3

T-STAT3

β-actin

A-STAT3

P-STAT3

T-STAT3

β-actin– –IL-6

WT

WT

K685R

K685

R

K679

R

K707

RK7

09R

2K/R

679-

685

2K/R

707-

709

WT K685R

4K/R

Human DIPKEEAFGKYCRPESQEHPEADPGSAAPYLKTKFMouse DIPKEEAFGKYCRPESQEHPEADPGSAAPYLKTKFRat DIPKEEAFGKYCRPESQEHPEADPGSAAPYLKTKFBovin DIPKEDAFGKYCRPESQEHPEADPGSAAPYLKTKF

685679 707 7090

hSTAT3 PDIPKEEAFGKY-CRP-ESQEHPEADPGSAAPYLKTKFICVTPhSTAT1 PNIDKDHAFGKYYSRPKEAPEPMELDGPKGTGYIKTELISVSEhSTAT2 PRIPRDEAFGCYYQEKVN------LQERR-K-YLKHRLIVVSNhSTAT4 PDIPKDKAFGKHYSSQP--CEVSRPTERGDKGYVPSVFIPISThSTAT5 PDRPKDEVFSKYYTPV---------LAKAVDGYVKPQIKQVVPhSTAT6 PKKPKDEAFRSHYKP--------EQMGKDGRGYVSTTIKMTVE

707709

βH

0b

a-K8

7

a-K4

9 Helix Bundle DNA binding

100 200 300 400 500 600 700

SH2

a-K6

85p-

Y705

705

705679 685

Disordered

Control

IFN-γ

**

*0 µm4 µm

8 µm16 µm

WT 4K/R Y705FEX527

VehicleIL-6

Vehicle IL-6

*

**

*

Figure 3 Critical novel acetylation sites regulate STAT3 phosphorylation and transactivation. (a) HEK293 cells were transfected with wild-type (WT) or K685R- STAT3 plasmids alone, or co-transfected with SirT1. STAT3 acetylation was determined. (b) A schematic representation of lysine acetylation sites identified in STAT3. Green lines represent the sites previously reported. Red lines represent a further three acetylation sites, identified by tandem mass spectrometry (Supplementary Information, Fig. S2b). (c, d) Mutations at four C-terminal lysine residues (4K/R) abolished STAT3 phosphorylation. A2780 cells were transfected with different K/R STAT3 mutants (c). Cells transfected with 4K/R or wild-type STAT3 were stimulated with 40 ng ml–1) for 6 h (d). (e–g) The 4K/R mutation abolished STAT3 transactivation function. A2780 cells were cotransfected with different K/R-STAT3 mutants (0.15 μg), hSirT1 (0.1 μg), STAT3-

dependent luciferase reporter (0.1 μg) and an internal control reporter pRL-TK plasmid (0.01 μg, e). Cells were treated with IL-6 (40 ng ml–1) or left untreated. The transactivation function of STAT3 was assayed. The effect of p300 (0.35 μg per well, f) and various EX527 doses (g) on activating the transactivation function of wild-type or 4K/R- STAT3 were determined. (h, i) The 4K/R mutant disrupted STAT3 nuclear localization. STAT3 KO hepatoma cells were infected with retroviruses (pbabe-6Myc–STAT3 WT, Y705F and 4K/R) and stimulated with IFN-γ (50 ng ml–1) for 2 h. (i) Primary hepatocytes were prepared from fasted animals and cultured in low nutrient media for 12 h before treatment with EX527 (10 μM) for 6 h or IL-6 (40ng ml–1) for 1 h. The translocation of STAT3 was determined by immunofluorescence microscopy staining. Data are mean ± s.e.m. of three repeated experiments, * P < 0.005, ** P < 0.01 in e,f and g. Scale bars, 10 μm in h, 50 μm in i.



L E T T E R S

transfection with HDAC1 and HDAC 3 had a limited effect (Fig. 2j, k). Consistently, an assay of STAT3-dependent luciferase reporter10 (p4x IRF–Luc) revealed that SirT1, but not HDAC1 or HDAC3, suppressed the transactivation function of STAT3 (Fig. 2j, k). However, at increased doses of plasmid DNA, HDAC3 appreciably reduced STAT3 activity (Supplementary Information, Fig. S2a); this may contribute to downregu-lation of STAT3 acetylation and phosphorylation under certain conditions and in selected cells8,20. From these results, we conclude that SirT1 spe-cifically and effectively deacetylated STAT3 in cultured cells and in vivo, and that this modification is coupled with a downregulation of STAT3 phosphorylation and transactivation.

The direct interaction between SirT1 and STAT3, and its effect on STAT3 phosphorylation and function, suggest that the state of acetyla-tion of STAT3 may directly regulate its phosphorylation. Acetylation in a limited number of lysine residues in STAT3 was reported8–10; however, the coupling of acetylation and phosphorylation through these sites was not established, suggesting more lysine acetylation sites are involved. To identify these, particularly in the carboxy-terminal region, which is cru-cial for STAT3 phosphorylation, we initially examined acetylation in the K685R-STAT3 mutant: Lys 685 was reported to be the only acetylation site at the C terminus of STAT3 (ref. 10). We used an anti-acetyl-STAT3 antibody and, although it detected a minor reduction in acetylation, sig-nificant signal was still detected (Fig. 3a). This signal was downregulated by SirT1, suggesting that other acetylation sites exist. Three new lysine-acetylation residues, K679, 707 and 709, were identified by tandem mass spectrometry analysis (Supplementary Information, Fig. S2). These sites are evolutionally conserved among mammalian STATs, (Fig. 3b) and located either in the end β-sheet structure of the SH2 domain21 (βH), or in the disordered signal stretch immediately after the SH2 domain. SH2 is known to specifically bind to phospho–tyrosine peptides22,23; therefore, it is critical to tyrosine signalling. Notably, the three new lysine residues are all in the vicinity of the Y705 of STAT3 (Fig. 3b), indicating that these sites may be pertinent to STAT3 phosphorylation7,24.

Next, a systematic site-mutagenesis of these lysine sites of interest was performed. Mutation of all four lysine to arginine abolished the acetyla-tion signals of STAT3 (Fig. 3c). Single or double lysine-to-arginine muta-tions had a limited effect on STAT3 phosphorylation, whereas changes of all four lysines to arganine (4K/R) largely abolished STAT3 phosphoryla-tion. This effect is specific to the C-terminal acetylation sites, as muta-tions of the amino-terminal acetylation sites (K49/87R) did not affect the phosphorylation of STAT3 (Supplementary Information, Fig. S3a, b). These results imply that multiple lysine-acetylation sites adjacent to Y705 are vital for STAT3 phosphorylation. In contrast, the dominant-negative Y705F mutation did not affect the acetylation of STAT3 (Supplementary Information, Fig. S3c). More importantly, the 4K/R-STAT3 mutant was no longer sensitive to stimulation by IL-6 (interleukin) and IFN-γ (interferon-γ) which otherwise acetylated and phosphorylated STAT3. This suggests that the newly identified lysine acetylation sites are critical for cytokine stimulation of STAT3 (Fig. 3d).

To test whether the mutations of these acetylation sites affect STAT3 transcriptional function, STAT3-dependent luciferase reporter assays were performed in HEK293T cells and human ovarian cancer A2780 cells25. Consistent with the phosphorylation results, the 4K/R muta-tion largely abolished IL-6-induced STAT3 activation and had a dom-inant-negative-like effect, comparable to that of the Y705F mutation (Fig. 3e); whereas individual mutations had limited effects on STAT3

transactivation. Moreover, 4K/R-STAT3 was not capable of activating a known endogenous STAT3 target (hAGT; ref. 20) in HepG2 cells (Supplementary Information, Fig. S3d). Finally, ectopic expression of p300 (ref. 10) and treatment with EX527, which increased STAT3 acetylation and phosphorylation in vivo13, failed to activate 4K/R-STAT3, whereas the same treatments increased the transactivation of wild-type STAT3 (Fig. 3f, g). These results suggest that the acetylation of the clus-ter of C-terminal lysine sites is up or downregulated by p300 or SirT1, respectively, and is crucial for STAT3 phosphorylation and transactiva-tion. Next, we asked if the 4K/R mutations affect STAT3 localization, as phosphorylation is crucial for STAT3 nuclear translocation7. We intro-duced wild-type-, 4K/R- or Y705F-STAT3 into liver-STAT3 knockout (STAT3 LKO) hepatoma cells using a retrovirus (pbabe-6xcMyc–STAT3 containing wild-type-, Y705F- and 4K/R- STAT3) to test STAT3 nuclear translocation by immunofluorescence microscopy staining. Similarly to Y705F-STAT3, the mutant 4K/R-STAT3 significantly disrupted the nuclear translocation of STAT3 (Fig. 3h). This acetylation-related STAT3 localization was further supported by experiments in primary hepato-cytes treated with EX527 (Fig. 3i). In addition, we found that the effi-ciency of 4K/R-STAT3 dimerization was greatly reduced (Supplementary Information, Fig. S3e).

If SirT1 promotes gluconeogenesis, in part through the suppression of STAT3, the knockout of STAT3 in liver should mimic the effect of SirT1 and increase gluconeogenesis independent of SirT1 activity. In normal chow fed STAT3 LKO mice, we observed a significant upregulation of pepck1 and g6pase, but not of the control genes (cytochrome-c and gk; Fig. 4a–c). However, the further upregulation of pepck1 and g6pase genes on fasting26 was limited (Fig. 4a). The lack of repression of gluconeogenic gene expression after feeding, was due to the absence of STAT3 bound to the promoter region of pepck1, as shown by chromatin immunoprecip-tiation (ChIP) assay (Supplementary Information, Fig. S3f).

Next, we asked whether STAT3 is involved in the SirT1-mediated induction of hepatic gluconeogenesis. EX527 was used to inhibit SirT1 activity in STAT3 LKO mice and littermate controls after a 24-h fast. As reported previously6, STAT3 LKO mice maintained normal glucose lev-els under such conditions, despite increasing gluconeogenic gene expres-sion and plasma insulin levels (Fig. 4a, e). EX527 (i.p. 10 mg per kg body weight) reduced plasma glucose levels in fasting wild-type mice, whereas the plasma glucose levels of STAT3 LKO mice were unchanged (Fig. 4d), suggesting that STAT3 deficiency disrupted the ability of EX527 to lower fasting glucose levels independent of insulin concentration (Fig. 4e). Consistent with the reduction of plasma glucose levels in wild-type ani-mals after EX527 treatment, the expression of the gluconeogenic genes was reduced in the livers of wild-type mice but was blunted in STAT3 LKO mice (Fig 4f). Next, we studied the effect of SirT1 knockdown on glucose homeostasis in the STAT3 LKO model. The levels of hepatic STAT3 acetylation and phosphorylation were significantly increased with SirT1 ASO in wild-type mice (Fig. 1d; Supplementary Information, Fig. S4a). Moreover, SirT1 ASO decreased hepatic gluconeogenic gene transcription (pepck1, g6pase and fabpase) in wild-type, but not STAT3 LKO, mice (data not shown).

Finally, a glucose tolerance test (Fig. 4g), a pyruvate tolerance test (Fig. 4h) and a glucagon-stimulation test (Fig. 4i) were conducted to evalu-ate various aspects of glucose homeostasis in wild-type and STAT3 LKO mice, with or without SirT1 knockdown. An overall reduction in glucose production was indicated in the SirT1 ASO-treated control animals. SirT1



L E T T E R S

ASO treatment of STAT3 LKO mice had little effect on the parameters measured by all three tests (Fig. 4g–i). These data led us to conclude that SirT1 promotes gluconeogenesis, in part, by suppressing the inhibitory

effect of STAT3 on the expression of gluconeogenic genes. To test the effect of hepatic insulin resistance on the interaction of SirT1 with STAT3, all groups of mice were fed with a high-fat diet (Supplementary Information,

0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

FF KO

Rel

ativ

e fo

lds

of m

RN

A

FedFasted

FedFasted

FedFasted

FedFasted

0

1

2

3

4

5

6

FF KO

Rel

ativ

e fo

lds

of m

RN

A

0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

FF KO

Rel

ativ

e fo

lds

of m

RN

A

0

1

2

3

4

5

FF KO

Rel

ativ

e fo

lds

of m

RN

A

02468

10121416

0 h 0.5 h 1.5 h 3 h 6 h

Blo

od g

luco

se (g

litr

es–1

)

FFvehicle

FFEX527

KOvehicle

KOEX527

0

4

8

12

16

20

FF + Vehicle FF + EX527 KO + vehicle KO + EX527S

era

insu

lin (µ

M m

l–1)

Before EX527 After EX527

00.51.01.52.02.53.03.54.0

FF KO

Rel

ativ

e fo

lds

of m

RN

A

Fasted Fasted+EX527Fed

Fasted Fasted+EX527Fed

0.0

1.0

2.0

3.0

4.0

FF KO

Rel

ativ

e fo

lds

of m

RN

A

0

10

20

30

40

50

60

0 0.5 h 1.5 h 3 h

Glu

cose

(g li

tres

–1)

FF-control ASOFF-SirT1 ASO

0

20

40

60

80

0 0.5 h 1.5 h 3 h

Glu

cose

(g li

tres

–1)

KO-control ASO

KO-SirT1 ASO

0

5

10

15

20

25

0 0.5 h 1 h 2.5 h

Glu

cose

(g li

tres

–1)


0

5

10

15

20

25

30

30

0 0.5 h 1 h 2.5 h

Glu

cose

(g li

tres

–1)

KO-control ASOKO-sirT1 ASO

0

4

8

12

16

20

0 0.5 h 1 h 2 h

Glu

cose

(g li

tres

–1)


0

5

10

15

20

25

0 0.5 h 1 h 2 h

Glu

cose

(g li

tres

–1)

KO-control ASOKO-SirT1 ASO

5

7.5

10

12

15

17

20

Fed Fast

Glu

cose

(g li

tres

–1)

WT-SirT1 ASO

WT-control ASO

KO-SirT1 ASO

KO-control ASO

d

f g

h i

j

e

ba cpepck1

pepck1

g6pase

fbpase

Cytochrome-c gk

WT

WTWT STAT3 LKO

STAT3 LKOSTAT3 LKO

**

*

*

**

** *

** ** **

**

*

**

**

**

**

*Fasted Fasted+EX527Fed

0

1.0

2.0

3.0

4.0

5.0

6.0

FF KO

Rel

ativ

e fo

lds

of m

RN

A

g6pase

Figure 4 Liver-STAT3 deficiency disrupted fasting/SirT1 controlled gluconeogenesis. (a) Liver-STAT3 deficiency mimicked the effect of SirT1 in promoting gluconeogenic gene expression, independently of nutritional status and SirT1 activity. The mRNA levels of pck1 and g6pase in livers were determined in fed and fasted STAT3f/ f (FF) and STAT3 LKO (KO) mice using qrtPCR analysis (mean ± s.e.m., n = 5 livers). (b, c) Expression of the mitochondrial gene cytochrome-c as a control (b), and the glycolytic gene gk (c), were not altered by the absence of STAT3. Mean ± s.e.m., *P < 0.05, n = 5 livers in b and c. (d) STAT3 deficiency disrupted fasting/SirT1-induced hypoglycemia. The STAT3f/f and STAT3 LKO mice were fasted overnight, and then injected with EX527 (i.p. 10 mg per kg body weight). Plasma glucose levels were determined at 0, 30, 90, 180 and 360 min. (e) The same treatment did not alter the plasma insulin levels, which were determined before and after treatment with EX527. In d and e,

data are mean ± s.e.m., n = 6 mice, ** P < 0.01. (f) STAT3 deficiency in the livers disrupted fasting/SirT1 induced gluconeogenic gene expression. The transcripts of the key gluconeogenic genes pepck1, g6pase and fbpase in the livers were detected by qrt-PCR (mean ± s.e.m., n = 5 * P < 0.05, ** P < 0.01). (g–i) The liver STAT3 deficiency impaired SirT1 controlled liver glucose production as assessed by a glucose tolerance test (GTT, g), pyruvate tolerance test (PTT, h) and glucagon-stimulation test (GST, Glucagon, i). Data are mean ± s.e.m., n = 6 mice in g–i. (j) STAT3 LKO-impaired SirT1 knockdown induced a reduction in glucose in animals on a high-fat diet. STAT3f/f and STAT3 LKO mice were fed a high-fat diet for two-and-a-half weeks. SirT1 ASO or control ASO was administered five times during a two week period at a dose of i.p. 10 mg per kg body weight. The levels of plasma glucose levels were measured under conditions of feeding and overnight fasting ( mean ± s.e.m., n = 7 mice, * P = 0.022).



L E T T E R S

Fig. S5a). Blood glucose levels on fasting were decreased in SirT1 ASO-treated wild-type mice, but unchanged in SirT1-ASO treated STAT3 LKO mice (Fig. 4j; Fig. S5b). This observation indicates that the loss of hepatic STAT3 is critical for SirT1 function. In addition, our results are consistent with the SirT1 LKO model12, but differ from the SirT1 transgenic mouse model.3,27. However, gluconeogenic gene expression was increased in isolated primary SirT1 transgenic hepatocytes treated with cAMP and deprived of insulin treatment, suggesting that insulin-mediated inhibition of gluconeogenesis may be independent of the SirT1 pathway3.

If the C-terminal cluster of acetylation lysine sites are crucial for STAT3 transactivation, the C-terminal 4K/R mutant should disrupt the suppressive effect of STAT3 on hepatic glucose production. First, we found that the glucose production in STAT3 LKO primary hepa-tocytes was increased, compared with that in wild-type cells (Fig. 5a).

To further test the inhibitory effect of STAT3 on glucose production in hepatocytes, we promoted cellular glucose production in these cells, either by introducing PGC-1α or by treating cells with dexamethasone (50 nM) and 8-bromo-cAMP (2 μM; refs 2, 26). Equivalent amounts of STAT3 total proteins were detected in STAT3–/– cells in which wild-type STAT3, 4K/R- or Y705F-STAT3 had been re-introduced (Fig. 5b). Wild-type STAT3 suppressed hepatic glucose production (Fig. 5c, d), whereas, 4K/R- or Y705F-STAT3 had little effect. Consistent with the data on glucose production, we observed a strong reduction in expres-sion of the gluconeogenic pepck1 transcripts by wild-type-STAT3, but not by 4K/R-STAT3, in STAT3 LKO hepatocytes ectopically expressing PGC-1α or treated with dexamethasone/cAMP (Fig. 5e, f).

To determine whether the change in hepatic glucose production by re-introduced wild-type-STAT3 (but not 4K/R-STAT3) is mediated

0

50

100

150

200

Basal PGC-1a DEX-cAMP

Glu

cose

(µM

) FF KO

0

0.2

0.4

0.6

0.8

1.0

1.2

Ad-WT Ad-Y705F Ad-4/KR Ad-GFP

Rel

ativ

e fo

lds

of g

luco

se

Rel

ativ

e fo

lds

of g

luco

se

0

0.2

0.4

0.6

0.8

1.0

1.2

Ad-WT Ad-Y705F Ad-4/KR Ad-GFP

0

0.2

0.4

0.6

0.8

1.0

1.2

Ad-WT Ad-Y705F Ad-4/KR VectorR

elat

ive

fold

s of

mR

NA

Rel

ativ

e fo

lds

of m

RN

AR

elat

ive

fold

s of

mR

NA

0

0.2

0.4

0.6

0.8

1.0

1.2

Ad-WT Ad-Y705F Ad-4/KR Vector

0

10

20

30

40

Control GFP SirT1

Rel

ativ

e ra

tion

of g

luco

sep

rod

uctio

n

WT4/KR

DNGFP

WT4/KRDNGFP

0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

Control GFP SirT1

a b c

e

h

f

g

d

T-STAT3

GFP

GFP4K/RY705FWT

Ad-GFP Ad-SirT1W

T

4/K

R

Y70

5F

GFP

WT

4/K

R

Y70

5F

GFP

*

****

* *

* ** *

* ** *

**

*

pepck1 (PGC-1α) pepck1 (Dexa/cAMP)

pepck1

SirT1

CREB

PGC-1α

PGC-1α

Gluconeogenesis

STAT3

Foxo HNF4αPEPCKG6Pase

PYRUVATE/NAD+/NADH

SirT1

STAT3

GFP

Figure 5 The 4K/R mutant STAT3 is defective in suppressing hepatic gluconeogenesis. (a) Basal-, PGC-1α- or dexa/cAMP-stimulated glucose production in primary hepatocytes was significantly increased in the absence of STAT3 (* P < 0.05; FF, STAT3f/f; KO, STAT3 LKO). (b) Wild-type- (WT), 4K/R- and Y705F-STAT3 were introduced into primary hepatocytes by adenoviruses (Ad), and an equal amount of each STAT3 protein was detected. (c, d) Wild-type STAT3, but not 4K/R- or Y705F- STAT3, effectively inhibited the promotion of glucose production by either PGC-1α (c) or dexa/cAMP (d), in primary hepatocytes (** P < 0.01). (e, f) Similarly, the wild-type-STAT3, but not 4K/R- or Y705F-STAT3, inhibited the expression

of pepck1, when promoted by PGC-1α (e) or dexa/cAMP (f, * P < 0.05). (g) Ectopic SirT1 disrupted the effect of wild-type-STAT3 on suppressing glucose production. STAT3 KO primary hepatocytes were co-infected by adenoviruses, with adeno-SirT1, adeno-wild-type-STAT3, mutant-STAT3 or GFP control (** P < 0.01). The middle panel shows the level of protein expression of SirT1 and STAT3s; the right panel shows the level pepck1 mRNA. (h) A schematic representation of the nutrient sensing pathway through which SirT1 regulates hepatic gluconeogenesis by both suppressing STAT3 and activating PGC-1α /Foxo1. Data are mean ± s.e.m. of three repeated experiments in a–g.



L E T T E R S

by SirT1, we co-infected STAT3 LKO primary hepatocytes with both adeno-SirT1 and adeno-STAT3s (wild type, 4K/R, Y705F and GFP). Ectopic SirT1 blunted the suppression of glucose production by wild-type-STAT3; however, this effect was limited in 4K/R-STAT3- or Y705F-STAT3-expressing cells (Fig. 5g). These results further support the idea that hepatic glucose production is dependent on SirT1-mediated down-regulation of STAT3.

Our findings reveal a new molecular mechanism whereby SirT1 sup-presses the inhibitory effect of STAT3 on gluconeogenesis, while activat-ing PGC-1α and Foxo1 (ref. 4) to stimulate gluconeogenesis in the liver, in response to nutrient signals. This dual function of SirT1 has a central role in preventing concurrent activation of the two counter-mechanisms of gluconeogenesis regulation (Fig. 5h). These findings have implica-tions for defining the basic pathways of energy homeostasis, diabetes and lifespan.

METHODSAnimals. All the mice used in this study were on a C57BL/6J background. Twelve-week-old male wild-type mice (Jackson Laboratory) were fed ad libitum, fasted for 24 h, fasted and re-fed for 24 h or fed on a high-fat diet (45 Kcal% fat) for 48 h. STAT3 LKO mice were generated by crossing STAT3 f/f28 and Alb-Cre transgenic mice, B6.Cg-Tg (Alb-cre) 21Mgn/J (Jackson Laboratory). Various treatments were applied depending on the experiment, including: fasting, feeding, re-feeding, feeding with a high-fat diet and administration of various chemical compounds. EX527 (6-Chloro-2, 3, 4, 9-tetrahydro-1H-carbazole-1-carboxamide; Tocris; i.p.10 mg per kg body weight) and nicotinamide (Sigma; i.p. 50 mg per kg body weight or 150 mg per kg body weight ) were introduced for 6 h. Animals were killed and the liver, white adipose, muscles, kidney, heart, brain and spleen tis-sues were subjected to western blotting or quantitative real-time PCR (qrt-PCR) analysis. The sera were collected through tail-vein puncture at different times to test glucose, insulin and liver function (Alanine Aminotransferase, ALT; assays performed in the Yale Mouse Metabolic Phenotyping Center). All procedures were performed in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals, and under the approval of the Yale Medical School Animal Care and Use Committee.

SirT1 knockdown (ASO). Control and STAT3 LKO mice (2–4 months old) were divided into control ASO and SirT1 ASO groups. The mice received food ab libitum and were housed on a 12-h dark/light cycle. SirT1 ASO 5’-ATACCATTCTTTGGTCTAGA-3’ (ASO # 384856) or control ASO (ASO# 141923; ISIS Pharmaceuticals) was administered five times during a 2–3 week period (i.p. 10 mg per kg body weight). ASO solutions were sterilized through a 0.44-μm filter before injection. The ASO targets the 3’UTR of SirT1 mRNA and has no significant crossreactivity to other sirtuin family members. The control ASO had the same chemistry as SirT1 ASO, and had a scrambled oligonucleotide sequence. Both ASOs were prepared in normal saline.

Measurement of glucose metabolism. The glucose tolerance test (glucose 2 g per kg body weight i.p.), pyruvate tolerance test (Pyruvate 2 g per kg body weight i.p.) and glucagon-stimulation test (Glucagon 200 μg per kg body weight i.p.) were conducted to evaluate various aspects of glucose production and metabolism.

Constructs. The pcDNA3-6×Myc–mSTAT3 and its K685R mutant expression vectors were from the laboratory of Y. E. Chin10. pcDNA3-6×cMyc-mSTAT3 K49R, K87R, K49-87R, K679R, K685R, K707R, K709R, K685-679R, K707-709R, K679-685-707-709R and Y705F were derived from the wild-type pcDNA3-Myc–mSTAT3 vector using a site-mutagenesis kit (Stratagene). The deletion mutants of STAT3 were constructed through PCR, and the schematic map of deletions is shown in Fig. 2g. The sequences of the oligonucleotide primers are: STAT3-F, 5’-cgaattcc ATG GCT CAG TGG AAC CAG-3’; S134-R, 5’-cctcgag tca AGC TGT TGG GTG GTT GG-3’; S323-R, 5’-cctcgagtcaCAC CAC GAA GGC ACT CTT-3’; S590-R, 5’-cctcgagtcaGCT GAT GAA ACC CAT GAT G-3’; S668-R, 5’-cctcgagtcaTGG AGA CAC CAG GAT GTT G-3’; S770-R, 5’-cctcgag TCA CAT GGG GGA GGT AGC-3’; S586-F, 5’-cgaattcc ATG GGT TTC ATC AGC AAG G-3’; S664-F, 5’-cgaattccATC CTG GTG TCT CCA CTT G-3’.

Retroviruses pBaBe-6×cMyc-STAT3 (wild-type, Y705F, K49-87R and K679-685-707-709R) were subcloned from pcDNA3-xcMyc-STAT3s, by switching restriction endonuclease (SalI). All constructs were verified by nucleotide sequencing in Yale KECK facilities. pTOPO-hSirT1 and its inac-tivated form H363Y were from Wei Gu (Columbia University, New York). HA–p300 constructs were provided Tsi-Pang Yao (Duke University, Carolina). Flag–HDAC1, Flag–HDAC2 and Flag–HDAC3 constructs were originally provided by Edward Seto (H. Lee Moffitt Cancer Center and Research Institute, Florida). P4xIRF-1–Luc is a STAT3 specific Luciferase reporter. APRE–luciferase reporter29 was a gift from David Levy (New York University School of Medicine). A PRE-TK plasmid was used as an internal control for transfection efficiency.

Cell culture, plasmid transfection, drug treatment and primary hypatocytes. A2780 cells (Sigma) were cultured in RPMI 1640 medium with 10% fetal bovine serum (FBS) supplemented with glutamine (2 mM). The SV40-immobilized mouse hepatocytes (from Domenico Accili, Columbia University, New York) were maintained in modified Eagle’s medium containing 10% FBS. SirT1 KO MEFs and wild-type MEFs were gifts from Leonard Guarente (Harvard University, Massachusetts). MEFs and HEK293T cells were maintained in Dulbecco’s modi-fied Eagle’s medium containing 10% fetal bovine serum. Cells were transiently transfected using LipofectAMINE 2000 (Invitrogen), according to the manufac-turer’s protocol. Cells were collected and washed with cold PBS for experiments and cells received one of the following treatments: (i) 200 nM Resveratrol (Sigma) for 4 h; (ii) 0.3–3 mM Nicotinamide (Sigma) for 4 h; (iii) 1–6 μM TSA (Sigma) for 6–12 h; (iv) 40 ng ml–1 IL-6 (Sigma) for 6–12 h; (v) 50 ng ml–1 IFN-γ (Sigma) for 2–6 h; (vi) 10 mM EX527 (Tocris) or (vii) 50 nM Dexamathasone (Sigma) + 2 μM 8-bromo-cAMP (Sigma).

Mouse primary hepatocytes were prepared in the Hepatocyte Isolation Core of the Yale Liver Center (Yale University School of Medicine, New Haven). Details are shown in supplementary methods.

Immunoprecipitation. Protein–protein interactions in cells were also analysed by co-immunoprecipitation. The details of these experiments have been described previously30. Ectopic Myc–STAT3 or SirT1 were expressed in HEK293T cells by transient transfection with PcDNA3cMyc–STAT3 or PcDNA3.1/V5-his-SirT1. The interaction of ectopic STAT3 with ectopic SirT1, or ectopic STAT3 with endogenous SirT1, was tested using an anti-c-Myc antibody. Co-IP of endog-enous STAT3 and SirT1 from HEK293T cells and liver tissues was performed using an anti-STAT3 antibody (Santa Cruz). Cell pellets and mouse liver tissues were sonicated in modified IP buffer, and pre-cleaned with normal rabbit or mouse IgG-conjugated nProtein A Sepharose 4 Fast Flow beads (Amersham) for 2 h at 4 °C. The pre-cleaned lysates were then mixed with c-Myc, the anti-STAT3 antibody and normal rabbit or mouse IgG (negative control)-conjugated nProtein A Sepharose 4 Fast Flow beads. Immunoprecipitation products were separated by SDS–PAGE, and blotted using a SirT1 antibody (Millipore), and β-actin (Sigma), which was used as an internal control.

Identification of STAT3 acetylation sites by mass spectrometry. The STAT3 complexes were immunoprecipitated, separated by SDS–PAGE and stained with SYPRO Ruby (Bio-Rad). The visible bands were excised. Gel pieces were subjected to a modified in-gel trypsin digestion procedure, and the peptides were sub-jected to liquid chromatography-electrospray ionization-tandem mass spectrom-etry (LC-ESI-MS/MS) analysis (Taplin Biological Mass Spectrometry Facility). The most abundant ions were obtained, and the MS/MS spectra was directly searched against the non-redundant protein database of the National Center for Biotechnology Information with the SEQUEST database search algorithm.

ChIP assays. The ChIP Assay Kit and protocol (Upstate Biotechnology) were used. P200 (region -226 to -24 of the mouse C/EBPδ promoter containing STAT3 binding sites), was used for STAT3 positive control primers31, see Supplementary Information, Methods for details.

Statistical analysis. Results are expressed as the mean ± s.e.m., and statistical analysis was performed by one-way or two-way ANOVA analysis of variance and Student’s t-test. A P < 0.05 was considered significant.




L E T T E R S

AckNowLEDGEMENTSWe thank M. Shanabrough for her technical support and careful revision of this manuscript. Some constructs were obtained from Y. E. Chin, W. Gu, P. Yao, E. Seto and D. Levy. SirT1, PGC-1α adenovirus was a gift from P. P. Puigserver . SirT1 KO MEFs and wild-type MEFs were a gift from L. P. Guarente. Part of this work was supported by an ADA grant to Q. G. (1-08-RA-54) and NIH grants to T. L. H. (DK-08000 and DK-060711), G. I. S. (DK-40936 and DK-076169) and J.L.B. (DK-P30-34989). The preparation of primary hepatocytes was performed in the Liver Center of Yale University School of Medicine. SirT1 ASO and Control ASO were provided by ISIS Pharmaceuticals, Inc.

AuTHor coNTrIbuTIoNSY.N., T.L.H. and Q.G. designed, executed and analysed most of the experiments and wrote the paper. D.M.E. contributed to the execution of the SirT1-ASO animal experiments and edited the paper. Z.Y. contributed to construction of trunicated STAT3 plasmids. M.D. designed , performed and analysed the animal experiements with EX527 treatement. G.I.S. provided critical models and analysed the data of the animal experiments.

coMpETING fINANcIAL INTErESTSThe authors declare no competing financial interests.


1. Luo, J. et al. Negative control of p53 by Sir2α promotes cell survival under stress. Cell 107, 137–148 (2001).

2. Rodgers, J. T. et al. Nutrient control of glucose homeostasis through a complex of PGC-1α and SIRT1. Nature 434, 113–118 (2005).

3. Banks, A. S. et al. SirT1 gain of function increases energy efficiency and prevents diabetes in mice. Cell Metab. 8, 333–341 (2008).

4. Frescas, D., Valenti, L. & Accili, D. Nuclear trapping of the forkhead transcription factor FoxO1 via Sirt-dependent deacetylation promotes expression of glucogenetic genes. J. Biol. Chem. 280, 20589–20595 (2005).

5. Inoue, H. et al. Role of STAT-3 in regulation of hepatic gluconeogenic genes and car-bohydrate metabolism in vivo. Nature Med. 10, 168–174 (2004).

6. Inoue, H. et al. Role of hepatic STAT3 in brain-insulin action on hepatic glucose produc-tion. Cell Metab. 3, 267–275 (2006).

7. Zhong, Z., Wen, Z. & Darnell, J. E. Jr Stat3: a STAT family member activated by tyrosine phosphorylation in response to epidermal growth factor and interleukin-6. Science 264, 95–98 (1994).

8. Ray, S., Boldogh, I. & Brasier, A. R. STAT3 NH2-terminal acetylation is activated by the hepatic acute-phase response and required for IL-6 induction of angiotensinogen. Gastroenterology 129, 1616–1632 (2005).

9. Wang, R., Cherukuri, P. & Luo, J. Activation of Stat3 sequence-specific DNA binding and transcription by p300/CREB-binding protein-mediated acetylation. J. Biol. Chem. 280, 11528–11534 (2005).

10. Yuan, Z. L., Guan, Y. J., Chatterjee, D. & Chin, Y. E. Stat3 dimerization regulated by reversible acetylation of a single lysine residue. Science 307, 269–273 (2005).

11. Rodgers, J. T., Lerin, C., Gerhart-Hines, Z. & Puigserver, P. Metabolic adaptations through the PGC-1α and SIRT1 pathways. FEBS Lett. 582, 46–53 (2008).

12. Chen, D. et al. Tissue-specific regulation of SIRT1 by calorie restriction. Genes Dev. 22, 1753–1757 (2008).

13. Solomon, J. M. et al. Inhibition of SIRT1 catalytic activity increases p53 acetylation but does not alter cell survival following DNA damage. Mol. Cell. Biol. 26, 28–38 (2006).

14. Chan, J. H., Lim, S. & Wong, W. S. Antisense oligonucleotides: from design to thera-peutic application. Clin. Exp. Pharmacol. Physiol. 33, 533–540 (2006).

15. Savage, D. B. et al. Reversal of diet-induced hepatic steatosis and hepatic insulin resistance by antisense oligonucleotide inhibitors of acetyl-CoA carboxylases 1 and 2. J. Clin. Invest. 116, 817–824 (2006).

16. Puigserver, P. et al. Insulin-regulated hepatic gluconeogenesis through FOXO1-PGC-1α interaction. Nature 423, 550–555 (2003).

17. Nakae, J., Park, B. C. & Accili, D. Insulin stimulates phosphorylation of the forkhead transcription factor FKHR on serine 253 through a Wortmannin-sensitive pathway. J. Biol. Chem. 274, 15982–15985 (1999).

18. Howitz, K. T. et al. Small molecule activators of sirtuins extend Saccharomyces cerevi-siae lifespan. Nature 425, 191–196 (2003).

19. Li, X. et al. SIRT1 deacetylates and positively regulates the nuclear receptor LXR. Mol. Cell 28, 91–106 (2007).

20. Ray, S., Lee, C., Hou, T., Boldogh, I. & Brasier, A. R. Requirement of histone deacety-lase1 (HDAC1) in signal transducer and activator of transcription 3 (STAT3) nucleocy-toplasmic distribution. Nucleic Acids Res.(2008).

21. Becker, S., Groner, B. & Muller, C. W. Three-dimensional structure of the Stat3β homodimer bound to DNA. Nature 394, 145–151 (1998).

22. Koch, C. A., Anderson, D., Moran, M. F., Ellis, C. & Pawson, T. SH2 and SH3 domains: elements that control interactions of cytoplasmic signaling proteins. Science 252, 668–674 (1991).

23. Pawson, T. & Gish, G. D. SH2 and SH3 domains: from structure to function. Cell 71, 359–362 (1992).

24. Minami, M. et al. STAT3 activation is a critical step in gp130-mediated terminal dif-ferentiation and growth arrest of a myeloid cell line. Proc. Natl Acad. Sci. USA 93, 3963–3966 (1996).

25. Maloney, A. et al. Gene and protein expression profiling of human ovarian cancer cells treated with the heat shock protein 90 inhibitor 17-allylamino-17-demethoxygeldan-amycin. Cancer Res. 67, 3239–3253 (2007).

26. Yoon, J. C. et al. Control of hepatic gluconeogenesis through the transcriptional coac-tivator PGC-1. Nature 413, 131–138 (2001).






l e t t e r s

Absence of nucleolar disruption after impairment of 40s ribosome biogenesis reveals an rpl11-translation-dependent mechanism of p53 inductionStefano Fumagalli1,8, Alessandro Di Cara2, Arti Neb-Gulati1, Francois Natt3, Sandy Schwemberger4, Jonathan Hall3, George F. Babcock4,5, Rosa Bernardi6, Pier Paolo Pandolfi7 and George Thomas1,8

Impaired ribosome biogenesis is attributed to nucleolar disruption and diffusion of a subset of 60S ribosomal proteins, particularly ribosomal protein (rp)L11, into the nucleoplasm, where they inhibit MDM2, leading to p53 induction and cell-cycle arrest1–4. Previously, we demonstrated that deletion of the 40S rpS6 gene in mouse liver prevents hepatocytes from re-entering the cell cycle after partial hepatectomy5. Here, we show that this response leads to an increase in p53, which is recapitulated in culture by rpS6-siRNA treatment and rescued by the simultaneous depletion of p53. However, disruption of biogenesis of 40S ribosomes had no effect on nucleolar integrity, although p53 induction was mediated by rpL11, leading to the finding that the cell selectively upregulates the translation of mRNAs with a polypyrimidine tract at their 5´-transcriptional start site (5´-TOP mRNAs), including that encoding rpL11, on impairment of 40S ribosome biogenesis. Increased 5´-TOP mRNA translation takes place despite continued 60S ribosome biogenesis and a decrease in global translation. Thus, in proliferative human disorders involving hypomorphic mutations in 40S ribosomal proteins6,7, specific targeting of rpL11 upregulation would spare other stress pathways that mediate the potential benefits of p53 induction8.

Conditional deletion of the rpS6 gene in mouse liver abrogates 40S ribos-ome biogenesis and prevents hepatocytes from re-entering the cell cycle after partial hepatectomy5. To understand the mechanism responsible for this response, mRNA microarray analyses were performed after hepatec-tomy on livers of rpS6fl/flMxCre+ mice, after rpS6 gene deletion, and control rpS6fl/flMxCre− mice. Consistent with the effects on cell-cycle progression, hepatocytes from rpS6fl/flMxCre+ mice, in contrast with those from rpS6fl/

flMxCre− mice, failed to induce the expression of genes required for DNA

synthesis and cell-cycle progression, including those encoding MCM3, MCM5, CDC6 and CDC20 (Supplementary Information, Fig. S1a; Supplementary Information, Table S1). In parallel, p53 target genes involved in cell-cycle inhibition, including those encoding p21, Bax and MDM2, were upregulated in livers of rpS6fl/flMxCre+ mice (Fig. 1a; Supplementary Information, Fig. S1b). These data are consistent with reports implicating p53 in cell-cycle arrest and apoptosis in response to the deletion of one rpS6 allele in tissues other than the liver9–11. Indeed, analysis of p53 in livers of rpS6fl/flMxCre+ versus rpS6fl/flMxCre− mice showed a significant increase in p53 levels (Fig. 1b). These data indicate that deletion of rpS6 leads to p53 induction, preventing hepatocyte cell-cycle progression.

Given the difficulties in dissecting cell-cycle checkpoints in vivo, we deter-mined whether these responses could be recapitulated by depleting rpS6 in human A549 cells, which are wild-type for p53 (ref. 12). Transfection of two distinct rpS6-siRNAs led to a sharp decrease in rpS6 transcript levels in comparison with untransfected cells or cells transfected with a non-silencing siRNA (NSsiRNA) (Fig. 1c). In contrast, the effect on rpS6 protein levels was modest because of the long half-life of ribosomes5. Such treatment also resulted in a decrease in free 40S subunits and an increase in free 60S subunits (Fig. 1d), which was paralleled by a decrease in the incorpora-tion of 3H-uridine into nascent 18S, but not 28S, rRNA (Supplementary Information, Fig. S1c). Moreover, as reported for deletion of rpS6, rpS6-siRNA treatment led to the accumulation of a 34S rRNA (ref. 5), which we found, by employing specific probes, to be the 34S precursor of 18S rRNA (Supplementary Information, Fig. S1d). Finally, rpS6-siRNA treatment resulted in an induction of p53 and p21 (Fig. 1e), a 60% decrease in incor-poration of bromodeoxyuridine (BrdU) into DNA (Fig. 1f) and an accumu-lation of cells in G1 (Supplementary Information, Fig. S1e); all these effects were reversed by co-transfection of p53-siRNA (Fig. 1e, f; Supplementary Information, Fig. S1e). Thus, depletion of rpS6 largely phenocopies the responses observed in mouse liver on rpS6 deletion.

1Department of Cancer and Cell Biology, Genome Research Institute, University of Cincinnati, 2180 East Galbraith Road, Cincinnati, Ohio 45237, USA. 2Friedrich Miescher Institute for Biomedical Research, Maulbeerstrasse 66, PO Box 2543, CH-4058 Basel, Switzerland. 3Novartis Institutes for BioMedical Research Basel, Novartis Pharma AG, Lichtstrasse 35, CH-4002 Basel, Switzerland. 4Department of Surgery, University of Cincinnati, 231 Albert Sabin Way, P.O. Box 670558, Cincinnati, Ohio 45267-0558, USA. 5Cincinnati Shriner’s Hospital for Children, 3229 Burnet Avenue, Cincinnati, Ohio 45229-3095, USA. 6San Raffaele, Institute Via Olgettina, 5820132 Milano, Italy. 7Division of Genetics, Department of Medicine, Beth Israel Deaconess Medical Center 330 Brookline Avenue, Boston, Massachusetts 02215, USA.8Correspondence should be addressed to S.F. or G.T. (e-mail: [email protected]; [email protected])

Received 16 October 2008; accepted 15 December 2008; published online 15 March 2009; DOI:10.1038/ncb1858.





l e t t e r s

Induction of p53 in response to inhibition of ribosome biogenesis is attributed to nucleolar disruption13, which releases 60S rpL11 into the nucleoplasm, where it binds and inhibits MDM2 (refs 1, 4). To test this possibility, we analysed the nucleolar status of cells treated with NSsiRNA, actinomycin D, or rpS6-siRNA. RpS6 immunostaining of NSsiRNA-treated A549 cells revealed diffuse cytoplasmic, but discrete nucleolar, staining (Fig. 2a), presumably representing mature and nascent 40S ribosomes, respectively. Nucleolar rpS6 staining co-localized with that of nuclear high-mobility group-like protein 2 (NHP2), which localizes to dense fibrillar centres14. Treatment with actinomycin D led to a loss of nucleolar rpS6 staining and dispersion of NHP2 into nuclear clusters, which is consistent with its ability to inhibit rRNA transcription and to disrupt nucleoli13. In rpS6-siRNA-depleted cells, nucleolar rpS6 staining was absent, whereas it was largely unaffected in mature ribosomes of the cytoplasm (Fig. 2a). However, rpS6 depletion did not disrupt nucleoli, as demonstrated by the maintenance of NHP2 staining (Fig. 2a). Similarly, the localization of nucle-olin, which is enriched at the periphery of nucleoli, was unaffected by rpS6 depletion, whereas it was drastically altered by actinomycin D (Fig. 2b). Consistent with rpS6 depletion having no effect on the integrity of the nucleolus (Fig. 2), it did not affect the localization of nascent 60S ribos-omal proteins, as judged by immunostaining of 60S rpL7a (Supplementary Information, Fig. S2). Thus, nucleolar disruption is not a prerequisite for p53 induction after inhibition of 40S ribosome biogenesis13.

Because rpS6 depletion neither blunts 60S ribosome biogenesis (Fig. 1d; Supplementary Information, Fig. S1c) nor causes nucleolar disruption (Fig. 2), it seemed unlikely that p53 induction would be mediated by rpL11 (refs 1, 4). However, the induction of p53 and p21 by rpS6 depletion was abolished by the simultaneous depletion of rpL11 (Fig. 3a). As with rpS6 depletion, depletion of rpL11 had only a modest effect on protein levels (data not shown), although it clearly decreased mRNA levels (Fig. 3b). Moreover, rpL11 depletion in rpS6-siRNA-treated cells rescued S-phase entry (Fig. 3c). To ascertain the specificity of this response, we also analysed cells depleted of 40S rpS23 and 60S rpL7a. As judged from polysome profiles and incorporation of 3H-uridine into nascent rRNA, depletion of rpS23 or rpL7a resulted in the inhibition of 40S or 60S ribosome biogenesis, respectively (Fig. 3d; Supplementary Information, Fig. S3a). In a similar manner to rpS6 depletion (Fig. 1e), a decrease in either protein led to p53 and p21 induction, which was suppressed by the simultaneous depletion of rpL11 (Fig. 3e). Equivalent results were obtained with a second rpL11-siRNA targeting a distinct region of the transcript (Supplementary Information, Fig. 3b, c). These findings indicate that rpL11-mediated p53 induction is a general response to inhibition of 40S or 60S ribosome biogenesis.

If p53 induction, after rpS6 depletion, is dependent on rpL11, this would require increased amounts of rpL11 protein in the face of 60S ribosome biogenesis. Treatment with either rpS6-siRNA or rpL7a-siRNA leads to

rpS6fl/flMxCre– rpS6fl/flMxCre+

0.0

0.5

1.0

1.5

2.0

2.5

2 3 41

b

1

2

3

4

5

00 20 30 4010

Hours after hepatectomy

a

NS NS + rp

S6

p53

p21

siRNAs: rpS6

+ p53

rpS6

e

0

20

40

60

80

100

120

NS NS+ rpS6

rpS6+ p53

siRNAs:

f

rpS6 protein

rpS6 mRNA

rpS6

siR

NA2

rpS6

siR

NA1

NSs

iRN

A

–

c

80S

40S

60Sd

A26

0

Nor

mal

ized

exp

ress

ion

(arb

itrar

y un

its)

p53

exp

ress

ion

(arb

itrar

y un

its)

Rel

ativ

e p

erce

ntag

e of

Brd

U-

pos

itive

cel

ls

Figure 1 Depletion of rpS6 leads to p53-dependent cell-cycle arrest. (a) Expression kinetics of p21 mRNA after partial hepatectomy in the livers of rpS6fl/flMxCre− (filled diamonds) and rpS6fl/flMxCre+ (open squares) mice, as measured by hybridization of DNA microarrays. Results are shown as means ± s.e.m. for three independent experiments. (b) Protein levels of p53 in the livers of rpS6fl/flMxCre− and rpS6fl/flMxCre+ mice at 30 h after partial hepatectomy were measured with an ELISA kit (see Methods). For each genotype the measurements from two different animals are shown. (c) Northern blot (upper panel) and western blot (lower panel) showing the levels of rpS6 mRNA and rpS6 protein, respectively, in non-transfected A549 cells (lane 1), or A549 cells transfected for 24 h with the indicated

siRNAs. (d) Polysome profiles from extracts of A549 cells treated for 24 h with either NSsiRNA (grey line) or rpS6 siRNA2 (black line). The positions of 40S and 60S native subunits and 80S monosomes are indicated. (e) Western blots showing the levels of rpS6, p53 and p21 proteins in cells transfected for 24 h with NSsiRNA or rpS6 siRNA2 in combination with either NSsiRNA or p53 siRNA. (f) Quantification of BrdU incorporation in A549 cells transfected for 24 h with the indicated siRNAs. For each treatment, the values are normalized to values for cells transfected with NSsiRNA. Results are shown as means ± s.e.m. for three independent experiments. Uncropped images of blots are shown in Supplementary Information, Fig. S5.



l e t t e r s

p53 induction, in contrast with the NSsiRNA-treated control cells (Fig. 4a). However, whereas depletion of rpS6 led to an increase in total cellular L11 protein, depletion of rpL7a had the opposite effect (Fig. 4a), which is consistent with the decrease in total native 60S ribosomes (Fig. 3d). The prediction from these findings is that there is more ribosome-free rpL11 in rpS6-siRNA-treated cells than in NSsiRNA-treated cells. To examine this possibility, the distributions of rpL11 and β-actin, the lat-ter as a loading control, were examined on western blots after centrifu-gation of cellular extracts on shallow sucrose gradients12, in which large polysomes pellet to the bottom of the tube. In samples from NSsiRNA-treated cells, most rpL11 protein sedimented with native 60S ribosomes and polysomes, although ribosome-free rpL11 could be detected at the top of the gradient (Supplementary Information, Fig. S4a). Similarly, in samples from rpS6-siRNA-treated cells, most rpL11 was associated with free 60S ribosomes and polysomes, although the amount that sedimented with native 60S ribosomes was significantly increased (Supplementary Information, Fig. S4a), which is consistent with the increase in native 60S ribosomes (Fig. 1d). However, despite continued 60S ribosome biogenesis, the amount of free rpL11 at the top of the gradient was clearly increased over that detected in NSsiRNA-treated cells, with comparable levels of cytosolic β-actin present in each sample (Supplementary Information, Fig. S4a). An increase in ribosome-free rpL11 was also observed in rpL7a-depleted cells (data not shown), most probably due to inhibition of 60S ribosome biogenesis (Fig. 3d). In rpS6-depleted and rpL7a-depleted cells, we found an increase in the amount of rpL11 that co-immunoprecipitated with enhanced levels of p53-induced MDM2, which were comparable to those observed in cells treated with actinomycin D (Fig. 4b). Thus, inhi-bition of 40S ribosome biogenesis leads to an increase in ribosome-free rpL11, which can bind and inhibit MDM2.

Several mechanisms could explain the increase in ribosome-free rpL11 protein levels after disruption of 40S ribosome biogenesis. However, because ribosomal proteins have long half-lives and rpL11 mRNA lev-els are not significantly changed in rpS6-siRNA-treated cells from those

in NSsiRNA-treated cells (Fig. 3b; Supplementary Information, Fig. 3c), we focused on translation. Consistent with rpS6-siRNA-treated cells and rpL7a-siRNA-treated cells having fewer polysomes than NSsiRNA-treated cells (Figs 1d and 3d), β-actin mRNA was distributed to smaller polysomes in these cells (Fig. 4c). Similarly, more rpL11 transcripts were found in the non-polysome fraction in cells treated with rpL7a-siRNA than in those treated with NSsiRNA (Fig. 4d). However, despite com-parable inhibitory effects of rpS6-siRNA and rpL7a-siRNA on β-actin mRNA translation (Fig. 4c) and global protein synthesis (Fig. 5b), most of the rpL11 transcripts were recruited to actively translating polysomes in rpS6-siRNA-treated cells (Fig. 4d). Similarly, depletion of rpS23 or rpS7 resulted in inhibition of 40S ribosome biogenesis and the recruitment of rpL11 mRNAs to polysomes, whereas rpL23 depletion led to inhibition of 60S ribosome biogenesis and the accumulation of rpL11 mRNAs in the non-polysome fraction (Fig. 4e, and data not shown). During liver regeneration, the translation of mRNAs encoding ribosomal proteins is upregulated, whereas in non-proliferating liver they reside largely in the non-polysome portion of the gradient15. Consistent with this finding was our observation, in the regenerating livers of rpS6fl/flMxCre− mice, that most of the rpL11 mRNA was recruited to polysomes, with a small amount present in the non-polysome fraction (Supplementary Information, Fig. S4b). However, in the non-regenerating livers of rpS6fl/flMxCre+ mice, we found that all rpL11 mRNA shifted to polysomes, despite a decrease in the mean polysome size of rpL11 mRNA transcripts (Supplementary Information, Fig. S4b). Thus, translational upregulation of rpL11 mRNA seems to be a general response to inhibition of 40S ribosome biogenesis.

The increased rpL11 translation induced by depletion of 40S ribosomal proteins was unexpected, because decreased ribosomal protein mRNA translation is associated with global inhibition of protein synthesis16. Given that translational upregulation of rpL11 is responsible for the increase in p53 levels in cells in which 40S, but not 60S, ribosome biogenesis has been disrupted, this response should be more sensitive to global transla-tional inhibition. We found that cycloheximide, a general translational

NSsiRNA

NSsiRNA

rpS6siRNA

Anti-rpS6 Anti-rpS6 Anti-NHP2 Mergea b Anti-nucleolin Merge

ActD

rpS6siRNA

ActD

Figure 2 Depletion of rpS6 does not result in nucleolar disruption. (a, b) A549 cells treated with either NSsiRNA or rpS6 siRNA2 for 24 h or with 5 ng ml−1 actinomycin D (ActD) for 8 h were fixed and stained with anti-rpS6 and

anti-NHP2 (a) or anti-rpS6 and anti-nucleolin (b) antibodies and analysed by confocal microscopy with a Plan-Apochromat 100×/1.4 numerical aperture Oil Dic objective (Zeiss; see Methods). Scale bar, 10 µm.



l e t t e r s

inhibitor, blunted p53 expression in a dose-dependent manner in either rpS6-siRNA-transfected or rpL7a-siRNA-transfected cells, with the effect more pronounced in rpS6-depleted cells (Fig. 5a). Depletion of rpS6 or rpL7a decreased the incorporation of [35S]methionine into nascent protein by 50%, and in both cases further treatment with 0.1 µg ml−1 cyclohex-imide decreased translation a further 50% (Fig. 5b). However, treatment with cycloheximide impaired p53 accumulation to a larger extent in cells transfected with rpS6-siRNA than in those transfected with rpL7a-siRNA (Fig. 5c; Supplementary Information, Fig. S4c). These findings agree with rpL11 translational upregulation being responsible for p53 induction in cells with impaired 40S ribosome biogenesis.

Ribosomal protein mRNA translation is negatively controlled by a 5´-TOP (refs 17, 18): replacement of pyrimidines with purines at this site leads to their constitutive translation19,20. The fact that other 5´-TOP mRNAs, including those encoding rpS8, rpS16 and rpL26, were also upregulated in rpS6-depleted cells (Supplementary Information, Fig. S4d) suggested that selective translation of rpL11 is mediated by the 5´-TOP. We tested this possibility by transiently expressing two reporter con-structs containing the coding region of the mRNA encoding human

growth hormone (hGH) fused to either the first 29 nucleotides of the rpS16 5´-untranslated region, including the 5´-TOP (wtrpS16-hGH), or to a mutant in which five of the eight 5´-TOP pyrimidines were replaced by purines (Cm5rpS16-hGH; refs 19, 20). RpS6 depletion, in compari-son with controls, led to a decrease in the number of 40S ribosomes, a decrease in the mean polysome size and an increase in the number of free 60S ribosomes (Fig. 5d). This also resulted in the almost complete shift of rpL11 transcripts onto actively translating polysomes, despite decreased global translation, as reflected by the decreased mean polysome size of rpL11 mRNA (Fig. 5d). In a similar manner to rpL11, wtrpS16-hGH transcripts, in NSsiRNA-treated cells, were distributed between non-polysomes and polysomes (Fig. 5e), whereas the Cm5rpS16-hGH transcript was present almost exclusively in polysomes (Fig. 5e). After rpS6 depletion, wtrpS16-hGH transcript, as with rpL11, largely shifted to polysomes, behaving as the Cm5rpS16-hGH transcript (Fig. 5f), despite the inhibition of global translation, as reflected by the shift of wtrpS16-hGH and Cm5rpS16-hGH transcripts to polysomes with a smaller mean size. These data show that translational upregulation of rpL11 mRNAs is mediated through derepression of the 5´-TOP.

0

20

40

60

80

100

120

p53

p21

siRNAs:

NS

+ rp

S6

NS

+ rp

L11

rpS6

+ rp

L11

NS

NS

NS

+ rp

S6

NS

+ rp

S23

NS

+ rp

L7a

rpS2

3 +

rpL1

1

rpL7

a +

rpL1

1

rpS6

+ rp

L11

NS

+ rp

L11

p53

p21

rpS6 mRNA

rpS23 mRNA

rpL7amRNA

rpL11 mRNA

siRNAs:

siRNAs:

rpS23mRNA

NSs

iRN

A

NSs

iRN

A

rpS2

3 si

RN

A

rpL7

a si

RN

A

NSsiRNA rpS23 siRNA

a b c

rpL7a siRNA

NS NS + rpS6

rpS6 +rpL11

NS +rpL11

NS + rpS6

rpS6 +rpL11

NS +rpL11

NS + rpS6

rpS6 +rpL11

NS +rpL11

NS NSsiRNAs:

0.0

0.5

1.0

1.5

2.0

2.5

0

1

2

3

rpL7amRNA

d erp

S6

mR

NA

(arb

itrar

y un

its)

rpL1

1 m

RN

A (a

rbitr

ary

units

)

Rel

ativ

e p

erce

ntag

e of

B

rdU

-pos

itive

cel

ls

A26

0

Figure 3 Upregulation of p53 by depletion of ribosomal proteins is mediated by rpL11. (a) Western blots showing the levels of p53 and p21 in A549 cells transfected for 24 h with the indicated siRNAs. (b) Total RNA extracted from the indicated transfections, performed in parallel with those in a, were used to measure the levels of the rpS6 and rpL11 mRNAs by quantitative real-time PCR. Each bar represents the ratio of the indicated mRNA to that of β-actin mRNA, and results are shown as means ± s.e.m. for three independent transfection experiments. (c) BrdU incorporation in A549 cells transfected for 24 h with the indicated siRNAs. For each treatment the values

are normalized to values for cells transfected with NSsiRNA. Results are shown as means ± s.e.m. for three independent experiments. (d) Polysome profiles from extracts of A549 cells treated for 24 h with either NSsiRNA (left panel) or siRNAs specific for rpS23 or rpL7a mRNAs. Insets: northern blots hybridized to probes specific for rpS23 and rpL7a mRNAs. (e) Western blots showing the levels of p53 and p21 proteins, and northern blots showing the levels of rpS6, rpS23, rpL7a and rpL11 mRNAs in A549 cells treated for 24 h with the indicated siRNAs. Uncropped images of blots are shown in Supplementary Information, Fig. S5.



l e t t e r s

During growth, rpL11 is assembled into nascent 60S ribosomes; how-ever, impairment of this process, or of rRNA transcription, allows excess rpL11 to inhibit MDM2 and stabilize p53 (Fig. 5g). In contrast, when 40S ribosome biogenesis is impaired, 60S ribosome biogenesis continues, leading to the translational upregulation of 5´-TOP mRNAs, despite inhibition of global protein synthesis (Fig. 5g). Apparently, the translational upregula-tion of 5´-TOP mRNAs is not highly energy-consuming, because the rate of translation in rpS6-depleted cells is equivalent to that in rpL7a-depleted cells (Fig. 5b), in which 5´-TOP mRNA translation is suppressed (Fig. 4d; Supplementary Information, Fig. S4d). The importance of the cellular response to disrupted ribosome biogenesis is underscored by the finding that hypomorphic mutations in 40S ribosomal proteins rpS19, rpS24, and rpS14 are causal agents in Diamond–Blackfan anaemia6,21 and 5q− syn-drome7, which are disorders characterized by erythroid hypoplasia and increased susceptibility to leukemia. Depletion of ribosomal proteins results in alterations in the ratio of mRNA to ribosomes, potentially changing the pattern of translation and the genetic program5. Such deficiencies lead to p53 induction, which is also associated with other pathological conditions such as Treacher–Collins syndrome (TCS). TCS is a congenital disorder charac-terized by hypoplasia of craniofacial elements caused by mutations in the nucleolar protein Treacle that impair ribosome biogenesis and result in p53

induction22. These effects are reversed in a p53 heterozygous background, suggesting that suppression of p53 could be used in the treatment of TCS22. However, given that ribosomal proteins are haploinsufficient tumour sup-pressors in humans6,7,21, inactivation of p53 in conditions under which ribos-ome biogenesis is impaired could induce leukaemia in Diamond–Blackfan anaemia6,21 and 5q− syndrome7, as recently described in zebrafish23. In contrast, our studies suggest that signalling components involved in rpL11 upregulation could be clinically targeted without interfering with other path-ways involved in p53 induction8. It will therefore be important to elucidate the mechanisms by which translation of 5´-TOP mRNAs is controlled in 40S ribosomal-protein-depleted cells, including those regulated by lupus antigen and mTOR. Lupus antigen has been shown to regulate the translation of 5´-TOP mRNA; however, it is unclear whether it is a positive or negative effector24,25. An inhibitor of mTOR function, rapamycin, has been shown to selectively attenuate the translation of 5´-TOP mRNAs20; however, we found the inhibitory effect of rapamycin on the translation of rpL11 mRNA in either rpS6-depleted or rpS23-depleted cells to be much less pronounced than in NSsiRNA-treated cells (S.F. and G.T., unpublished observations). Unravelling the signalling pathways that regulate 5´-TOP mRNA translation will be a key step in developing specific therapies for diseases associated with hypomorphic lesions in ribosomal protein expression.

siRNA siRNA siRNA

1

Fraction no. Fraction no. Fraction no.

3 5 7 9 112 4 6 8 10 12 1 3 5 7 9 112 4 6 8 10 12 1 3 5 7 9 112 4 6 8 10 12

NS

rpL7a

rpS6

NS

rpL7a

rpS6

NS

rpS7

rpL23

rpS23

0

10

20

30

5

15

25

35

40

0

10

20

30

5

15

25

35

40

0

10

20

30

5

15

25

35

40

MDM2

rpL11

NSs

iRN

Arp

S6 s

iRN

Arp

L7a

siR

NA

ActD

ActD

Anti-MDM2 NRSIP:

ba

c ed

*

NS

rpL11

p53

β-Actin

siRNA: rpS6

rpL7

a

Per

cent

age

of β

-act

in m

RN

A

Per

cent

age

of r

pL1

1 m

RN

A

Per

cent

age

of r

pL1

1 m

RN

A

Figure 4 Depletion of 40S ribosomal proteins results in translational upregulation of the rpL11 mRNA. (a) Western blots showing the levels of rpL11, p53 and β-actin proteins in cells transfected for 48 h with NSsiRNA, rpS6 siRNA2 or rpL7a siRNA. (b) Western blots showing the levels of rpL11 and MDM2 present in immunoprecipitates (IP) by using either an anti-MDM2 antibody or normal rabbit serum (NRS). Extracts used in the immunoprecipitation are from cells transfected for 24 h with NSsiRNA or with rpS6 or rpL7a siRNAs, or treated with actinomycin D (ActD) at

10 ng ml−1 for 10 h. The asterisk indicates the light chain of the anti-MDM2 antibody, which reacts with the secondary antibody. (c, d) Northern blot analysis of β-actin mRNA (c) and rpL11 mRNA (d) distribution on polysomes of A549 cells transfected with the indicated siRNAs for 30 h. (e) Northern blot analysis of rpL11 mRNA distribution on polysomes of A549 cells transfected with the indicated siRNAs. Fractions 4–12 in c–e contained translating polysomes. Uncropped images of blots are shown in Supplementary Information, Fig. S5.



l e t t e r s

MeTHODSCell culture and transfection. A549 and HEK 293 cells were cultured and transfected with siRNAs as described10,19, with the final concentration of each siRNA being 16.7 nM. For co-transfection, cells in 10-cm dishes were transfected with calcium phosphate26 and 2 µg of plasmid DNA, followed 24 h later with 6 nM siRNA. The siRNAs employed targeted mRNAs encod-

ing human rpS6 (5´-AAGAAAGCCCTTAAATAAAGA-3´ (siRNA1) and 5´-TTGTAAGAAAGCCCTTAAATA-3´ (siRNA2)), rpS7 (5´-TAGATGG-CAGCCGGCTCATAA-3´), rpS23 (5´-CAGCCCTAAAGGCCAACCCTT-3´), rpL7a (5´-CACCACCTTGGTGGAGAACAA-3´), rpL11 (5´-AAGGTGC-GGGAGTATGAG TTA-3´), rpL23 (5´-CTGGTGCGAAATTCCGGATTT-3´), p53 (ref. 12) and Nonsilencing siRNA (Qiagen).

L11NO

60S

40S

L11 mRNA L11 mRNA L11 mRNA

L11L1

L11NO

60S

40S

L11NO

60S

40S

Small-subunit proteindepletion

Large-subunit proteindepletion

Unperturbed ribosome biogenesis

0

50

100

150

200

250

p53

β-actin

NS NS rpS6 rpS6 rpL7a rpL7a

– + – + – +siRNA

CHX(0.1 µg ml–1)

CHX:

rpL7a siRNArpS6 siRNA

0 0.1 0.5 1 0 0.1 0.5 1

d

b c

0

0.2

0.4

0.6

0.8

1.0

0 1 2 3 4 5 6

Hours after CHX addition

0

5

10

15

20

25

0

5

10

15

20

25

0

5

10

15

20

25

1 1 12 2 23 3 34 4 45 5 56 6 67 7 78 8 89 9 910 10 1011 11 1112 12 12

e f

g

a

NPLMDM2p53

L11L11

L11NPL

MDM2p53L11L11 L11

L11 L11NPL

MDM2p53L11

µg ml–1

Rel

ativ

e am

ount

of

p53

pro

tein

c.p

.m. o

f 35S

-met

hion

ine

µg o

f pro

tein

A26

0

Per

cent

age

of r

pL1

1 m

RN

A

Per

cent

age

of h

GH

mR

NA

NSsiRNA + WT rpS16-hGH

rpS6 siRNA + WT rpS16-hGH

WT rpS16-hGHmRNA

Cm5rpS16-hGHmRNA

WT rpS16-hGHmRNA

Cm5rpS16-hGHmRNA

Fraction no. Fraction no. Fraction no.

Per

cent

age

of h

GH

mR

NA

NSsiRNA transfection S6 siRNA transfection

Figure 5 Depletion of rpS6 upregulates 5´-TOP mediated translation. (a) Western blots showing levels of p53 and β-actin in A549 cells transfected with either rpS6 or rpL7a siRNA and treated for 6 h with the indicated doses of cycloheximide (CHX). (b) Measurement of [35S]methionine incorporation in cells transfected with the indicated siRNAs and treated, or left untreated, for 4 h with 0.1 µg ml−1 cycloheximide. Results are shown as means ± s.e.m. for three independent experiments. (c) Quantification of the ratio of p53 to β-actin (measured by densitometry on western blots) in cells transfected for 24 h with either rpS6 siRNA (open diamonds) or rpL7a siRNA (filled squares) and treated for the indicated times with 0.1 µg ml−1 cycloheximide. Results are shown as means ± s.e.m. for three independent experiments. A representative western blot used for this analysis is shown in Supplementary Information, Fig. S4c. (d) Polysome profiles (top) and northern blot analysis

of the distribution of rpL11 mRNA on polysomes (bottom) in HEK 293 cells co-transfected with wild-type (WT) rpS16-hGH plasmid DNA in combination with either NSsiRNA (black line) or rpS6 siRNA (grey line). (e) Northern blot analysis of the distribution on polysomes of hGH reporter mRNA in HEK 293 cells co-transfected with NSsiRNA in combination with either wild-type rpS16-hGH or Cm5rpS16-hGH plasmid DNA. (f) Northern blot analysis of the distribution on polysomes of hGH reporter mRNA in HEK 293 cells co-transfected with rpS6 siRNA2 in combination with either wild-type rpS16-hGH or Cm5rpS16-hGH plasmid DNA. In the northern blot analysis of all three panels, fractions 4–12 contained translating polysomes. (g) Model of p53 stabilization in response to impaired 40S ribosome biogenesis. See the text for details. NPL, nucleoplasm; NO, nucleolus. Uncropped images of blots are shown in Supplementary Information, Fig. S5.



l e t t e r s

Mice. The floxed S6 mice and the rpS6 gene deletion have been described (Sigma)5. Three days after rpS6 gene deletion, partial hepatectomy was performed27, after injec-tion of 0.08% Narketan (Vetoquinol AG)/0.24% Rompun (Bayer) at 10 µl g−1 body weight. Livers of mice killed at 20, 30 and 40 h after surgery were quickly frozen and stored at −80 °C. The portion of liver resected was used for the 0 h time point.

Measurement of [35S]methionine incorporation. Cells were labelled for 30 min with a 35SMeth/Cys mix (10 µCi ml−1; NEG072; Perkin Elmer); proteins were extracted and precipitated and the amount of radioactivity incorporated was determined28.

Protein extraction. For western blot analysis of total extracts, cells were washed with ice-cold PBS and lysed in 50 mM Tris-HCl pH 8, 250 mM NaCl, 1% Triton X-100, 0.25% sodium deoxycholate, 0.05% SDS, 1 mM dithiothreitol (DTT). The lysates were cleared by centrifugation. For co-immunoprecipitation of MDM2–rpL11 complexes, cells were lysed in 50 mM Tris-HCl pH 8, 250 mM NaCl, 0.5% Nonidet P40, 1 mM DTT, and the lysates were cleared by centrifugation. For the analysis and distribu-tion of rpL11 protein on polysome profiles, cells were washed with ice-cold buffer A (20 mM Tris-HCl pH 7.5, 250 mM KCl, 10 mM MgCl2) and lysed in buffer A contain-ing 0.5% Triton X-100, 0.5% sodium deoxycholate, 120 U ml−1 RNAsin (Promega), 3 mM DTT. The extracts were incubated on ice for 30 min and cleared by centrifuga-tion. All extraction buffers contained a cocktail of protease inhibitors (Roche).

Immunoprecipitation. Samples (1.6 mg of extract) were precleared for 1 h with protein A-Sepharose and then incubated for 4 h at 4 °C with anti-MDM2 antibody (H-221; Santa Cruz Biotechnology). After 1 h with protein A-Sepharose, immu-noprecipitates were washed three times in extraction buffer and resuspended in protein sample buffer and analysed as described29.

Western blotting and enzyme-linked immunosorbent assay (ELISA). Protein samples were resolved by SDS–PAGE and transferred to poly(vinylidene difluoride) membranes (Millipore), with a semi-dry transfer apparatus (LTF-Labortechnik). The membranes were blocked in Tris-buffered saline (TBS) containing 0.2% Tween 20 and 5% milk. Both primary and horseradish peroxidase-conjugated secondary antibodies (Amersham Biosciences) were added in TBS containing 0.2% Tween 20 and 3% BSA (Sigma). After incubation with an enhanced chemilumines-cence (ECL) reagent (Amersham Biosciences), blots were exposed to X-ray films (Kodak). Primary antibodies used were as follows: mouse anti-rpS6 (Novartis), rabbit anti-rpL11 (from Karen Vousden), anti-rpL11 (Zymed Laboratories) anti-p53 (Santa Cruz Biotechnology), anti-p21 (Pharmingen), anti-MDM2 (Santa Cruz Biotechnology), and anti-β-actin (Cell Signaling). Quantification of western blots by densitometry was performed with ImageQuant software (Molecular Dynamics). For the determination of liver p53 levels, an ELISA kit was used as described by the manufacturer (Roche), with duplicate measurements for each sample.

Analysis of mRNA and protein distribution on polysome profiles. Analysis of mRNA distribution on polysome profiles was performed as described29. To analyse the distribution of rpL11 protein on polysome profiles, cell extracts were centrifuged on a shallow sucrose gradient12 and fractionated. Proteins from each fraction were precipitated with 20% trichloroacetic acid. After centrifugation, the pellets were washed with 80% ethanol/20% diethyl ether, resuspended in protein sample buffer and processed for western blot analysis.

RNA extraction, northern blot analysis, and probes. Isolation of total RNA and northern blot analyses were largely performed as described previously12. The probes used were as follows: rpS6, 5´-GAGATGTTCAGCTTCATC-TTGAAGCAGC TGAACGCC TCCGACGCCACGGAAAAGAGG-3´; rpL7a, 5´-GGCGATACGAGCCACAGAC TTAGGACCCAGGACATTGCCACCCCA-GTGACGGCGGATC-3´; rpL11 (human), 5´-GTTTGGCCCCAATGCAGCC-TGTC C TGC GC T TCT TGTCTGC GATGC TGAAAC C TGGC C-3´ ; rpS23, 5´-CCAC TTC TGGTC TCGTGCGTGAC TACGGAGC TTCC-TAGCAGTACGAAGTCCACGACAC-3´; β-actin, 5´-CAGATTTTC-T C C A T G T C G T C C C A G T T G G T G A C G A T G C C G T G C -T C G AT G G G G TAC T T C AG - 3 ´ ; r p L 1 1 ( m o u s e ) , 5´-GTGTATCTAGCTTTGGAGAACACCGGGGTC TGGCC TGTGAGC TGC-TCCAACACCTTGGC-3´; rRNA, 5´-GGCGAGCGACCGGC AAGGCGGAGGTCGACCCACGCCACACGTCGCACGAACGCC TGTC-3´. The hGH probe has been described20. The blots were developed with a Storm 840 phosphorimager and quantified with ImageQuant software.

Quantitative real-time PCR. Total RNAs were reverse-transcribed with a Superscript III kit (Invitrogen). The complementary DNAs were used in PCR reactions containing Absolute BlueSYBR Green Fluorescein mix (Thermo Scientific). The reactions were analysed in a Realplex S apparatus (Eppendorf). A standard curve was generated for each assay. Reactions were performed as fol-lows: 95^°C for 15^min followed by 40 cycles of 95^°C for 15^s, 60^°C for 20^s, and 72^°C for 20^s. Melting-curve analyses were performed to verify the syn-thesis of single products. The primers used were as follows: rpS6, 5´-TCTTGAC-CCATGGCCGTGTC-3´ (forward) and 5´-GCGGCGAG GCACTGTAGTAT-3´ (reverse); rpL7a, 5´-GCTGAAAGTGCCTCCTGCGA-3´ (forward) and 5´-CACCAAGGTGGTGACGGTGT-3´ (reverse); rpL11, 5´-TCCACTGCACA-GTTCGAGGG-3´ (forward) and 5´-AAACCTGGCCTACCCAGCAC-5´ (reverse). The β-actin primers have been described30.

Immunofluorescence. Cells seeded on coverslips were washed in PBS, fixed for 10 min at room temperature 20–22 °C with 4% paraformaldehyde, permeabilized for 5 min in PBS containing 0.5% Triton X-100 and washed in PBS. After blocking for 30 min in PBS containing 1% BSA at room temperature, the preparations were incubated overnight at 4 °C with the following antibodies diluted in PBS contain-ing 1% BSA: mouse anti-rpS6, rabbit anti-NHP2 (ref. 14), anti-nucleolin (H-250; Santa Cruz) or rabbit anti-rpL7a. The coverslips were then washed in PBS and incubated for 30 min at room temperature with the secondary antibodies Alexa Fluor 488 goat anti-mouse and Alexa Fluor 555 donkey anti-rabbit (Molecular Probes) diluted in PBS containing 1% BSA and then washed in PBS. The cover-slips were mounted on slides and analysed by fluorescence imaging with a Zeiss Axioplan II confocal microscope.

Microarray analysis. RNAs were extracted from liver portions, with the guanidin-ium thiocyanate–phenol–chloroform extraction protocol31, and further purified with an RNeasy Kit (Qiagen). Total RNA (10 µg) was reverse transcribed with the Affymetrix cDNA synthesis kit. The cDNA was then used in an in vitro transcrip-tion reaction using the in vitro transcription kit from Affymetrix. The resulting cRNAs were hybridized to MGU-74A GeneChips (Affymetrix) in accordance with the protocols provided by Affymetrix. Expression values were estimated with the robust multichip analysis summary algorithm (RMA)32 implemented in RMAExpress (http://rmaexpress.bmbolstad.com/). Expression analysis was performed with GeneSpring 6.2 (Silicon Genetics). For each time point we pro-duced a list of genes that were increased and a list of genes that were decreased in S6fl/flMxCre− mice in comparison with S6fl/flMxCre+ mice. Consideration was limited to those genes whose change in expression was at least 1.5-fold relative to the control genotype at one or more time points and that passed a one-way analysis of variance (cutoff P < 0.01). Microarray data can be accessed at http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE13864.

Flow cytometry. Cells pulsed for 10 min with 10 µM BrdU (Sigma) were col-lected and fixed in 70% ethanol. Staining with a fluorescein isothiocyanate-con-jugated anti-BrdU antibody (Becton Dickinson) was performed in accordance with the manufacturer’s instructions. A propidium iodide/RNAse solution (Phoenix Flow Systems) was added to the cell suspension before the analysis. Flow cytometry data were acquired on a Coulter Epics XL (Beckman Coulter) with a 488-nm argon-ion laser. Data acquisition was performed with System II software (Beckman Coulter). Doublet discrimination was performed by gating on peak versus integral signals. Cell-cycle analysis was performed with ModFit LT version 2.0 (Verity Software House, Inc.). Histograms were made with FlowJo software version 6.3.2 (Tree Star, Inc.). In all experiments, 10,000 gated events were collected.


ACkNowleDGemeNTSWe are indebted to P. D. Plas, A. Selvaraj and P. B. Dennis for their critical reading of the manuscript, as well as members of the laboratory for numerous discussions. We are also thankful to K. Vousden for the L11 polyclonal antibody. Finally we thank G. Doerman and M. Daston for computer graphics and the editing of the manuscript, respectively. A.D.C. was supported by fellowship from the Boehringer Ingelheim Fund. These studies were supported by the Mouse Models in Human Cancer Consortium grant U01 CA-84292 to P.P.P. and G.T., and G.T. is also supported by the Gladys and John Strauss Chair in Cancer Research and start-up funds from the University of Cincinnati.



http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE13864

http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE13864

l e t t e r s

ComPeTiNG FiNANCiAl iNTeReSTSThe authors declare that they have no competing financial interests.

Published online at http://www.nature.com/naturecellbiology reprints and permissions information is available online at http://npg.nature.com/reprintsandpermissions/

1. Lohrum, M. A., Ludwig, R. L., Kubbutat, M. H., Hanlon, M. & Vousden, K. H. Regulation of HDM2 activity by the ribosomal protein L11. Cancer Cell 3, 577–587 (2003).

2. Dai, M. S. & Lu, H. Inhibition of MDM2-mediated p53 ubiquitination and degradation by ribosomal protein L5. J. Biol. Chem. 279, 44475–44482 (2004).

3. Dai, M. S. et al. Ribosomal protein L23 activates p53 by inhibiting MDM2 function in response to ribosomal perturbation but not to translation inhibition. Mol. Cell. Biol. 24, 7654–7668 (2004).

4. Yuan, X. et al. Genetic inactivation of the transcription factor TIF-IA leads to nucleo-lar disruption, cell cycle arrest, and p53-mediated apoptosis. Mol. Cell 19, 77–87 (2005).

5. Volarevic, S. et al. Proliferation, but not growth, blocked by conditional deletion of 40S ribosomal protein S6. Science 288, 2045–2047 (2000).

6. Draptchinskaia, N. et al. The gene encoding ribosomal protein S19 is mutated in Diamond–Blackfan anaemia. Nature Genet. 21, 169–175 (1999).

7. Ebert, B. L. et al. Identification of RPS14 as a 5q− syndrome gene by RNA interference screen. Nature 451, 335–339 (2008).

8. Vousden, K. H. & Lane, D. P. p53 in health and disease. Nature Rev. Mol. Cell Biol. 8, 275–283 (2007).

9. Sulic, S. et al. Inactivation of S6 ribosomal protein gene in T lymphocytes activates a p53-dependent checkpoint response. Genes Dev. 19, 3070–3082 (2005).

10. Panic, L. et al. Ribosomal protein S6 gene haploinsufficiency is associated with activation of a p53-dependent checkpoint during gastrulation. Mol. Cell. Biol. 26, 8880–8891 (2006).

11. McGowan, K. A. et al. Ribosomal mutations cause p53-mediated dark skin and pleio-tropic effects. Nature Genet. 40, 963–970 (2008).

12. Beuvink, I. et al. The mTOR inhibitor RAD001 sensitizes tumor cells to DNA-damaged induced apoptosis through inhibition of p21 translation. Cell 120, 747–759 (2005).

13. Rubbi, C. P. & Milner, J. Disruption of the nucleolus mediates stabilization of p53 in response to DNA damage and other stresses. EMBO J. 22, 6068–6077 (2003).

14. Pogacic, V., Dragon, F. & Filipowicz, W. Human H/ACA small nucleolar RNPs and telomerase share evolutionarily conserved proteins NHP2 and NOP10. Mol. Cell. Biol. 20, 9028–9040 (2000).

15. Aloni, R., Peleg, D. & Meyuhas, O. Selective translational control and nonspecific post-transcriptional regulation of ribosomal protein gene expression during development and regeneration of rat liver. Mol. Cell. Biol. 12, 2203–2212 (1992).

16. Meyuhas, O., Avni, D. & Shama, S. in Translational Control (eds Hershey, J. W. B., Mathews, M. B. & Sonenberg, N.) 363–388 (Cold Spring Harbor Laboratory Press, 1996).

17. Meyuhas, O. Synthesis of the translational apparatus is regulated at the translational level. Eur. J. Biochem. 267, 6321–6330 (2000).

18. Fumagalli, S. & Thomas, G. in Translational Control of Gene Expression (eds Sonenberg, N., Hershey, J. W. B. & Mathews, M.) 695–717 (Cold Spring Harbor Laboratory Press, 2000).

19. Levy, S., Avni, D., Hariharan, N., Perry, R. P. & Meyuhas, O. Oligopyrimidine tract at the 5´ end of mammalian ribosomal protein mRNAs is required for their translational control. Proc. Natl Acad. Sci. USA 88, 3319–3323 (1991).

20. Jefferies, H. B. J. et al. Rapamycin suppresses 5´TOP mRNA translation through inhibi-tion of p70s6k. EMBO J. 12, 3693–3704 (1997).

21. Gazda, H. T. et al. Ribosomal protein S24 gene is mutated in Diamond–Blackfan anemia. Am. J. Hum. Genet. 79, 1110–1118 (2006).

22. Jones, N. C. et al. Prevention of the neurocristopathy Treacher Collins syndrome through inhibition of p53 function. Nature Med. 14, 125–133 (2008).

23. MacInnes, A. W., Amsterdam, A., Whittaker, C. A., Hopkins, N. & Lees, J. A. Loss of p53 synthesis in zebrafish tumors with ribosomal protein gene mutations. Proc. Natl Acad. Sci. USA 105, 10408–10413 (2008).

24. Crosio, C., Boyl, P. P., Loreni, F., Pierandrei-Amaldi, P. & Amaldi, F. La protein has a positive effect on the translation of TOP mRNAs in vivo. Nucleic Acids Res. 28, 2927–2934 (2000).

25. Schwartz, E. I., Intine, R. V. & Maraia, R. J. CK2 is responsible for phosphorylation of human La protein serine-366 and can modulate rpL37 5´-terminal oligopyrimidine mRNA metabolism. Mol. Cell. Biol. 24, 9580–9591 (2004).

26. Chen, C. A. & Okayama, H. Calcium phosphate-mediated gene transfer: a highly effi-cient transfection system for stably transforming cells with plasmid DNA. Biotechniques 6, 632–638 (1988).

27. Blindenbacher, A. et al. Interleukin 6 is important for survival after partial hepatectomy in mice. Hepatology 38, 674–682 (2003).

28. Iordanov, M. S. et al. Ribotoxic stress response: activation of the stress-activated pro-tein kinase JNK1 by inhibitors of the peptidyl transferase reaction and by sequence-specific RNA damage to the α-sarcin/ricin loop in the 28S rRNA. Mol. Cell. Biol. 17, 3373–3381 (1997).

29. Pende, M. et al. S6K1−/−/S6K2−/− mice exhibit perinatal lethality and rapamycin-sensi-tive 5´-terminal oligopyrimidine mRNA translation and reveal a mitogen-activated pro-tein kinase-dependent S6 kinase pathway. Mol. Cell. Biol. 24, 3112–3124 (2004).

30. Young, A. R. et al. Starvation and ULK1-dependent cycling of mammalian Atg9 between the TGN and endosomes. J. Cell Sci. 119, 3888–3900 (2006).

31. Chomczynski, P. & Sacchi, N. Single-step method of RNA isolation by acid guanidinium thiocyanate-phenol-chloroform extraction. Anal. Biochem. 162, 156–159 (1987).

32. Irizarry, R. A. et al. Exploration, normalization, and summaries of high density oligo-nucleotide array probe level data. Biostatistics 4, 249–264 (2003).






ERRATUM

Electrochemical cues regulate assembly of the Frizzled/Dishevelled complex at the plasma membrane during planar epithelial polarizationMatias Simons1, William J. Gault1, Daniel Gotthardt2, Rajeev Rohatgi3, Thomas J. Klein1, Youming Shao4, Ho-Jin Lee4, Ai-Luen Wu5, Yimin Fang5, Lisa M. Satlin3, Julian T. Dow6, Jie Chen5, Jie Zheng4, Michael Boutros7 and Marek Mlodzik1

Nature Cell Biol. 11, 286–294 (2009); published online 22 February 2009; corrected after print 24 February 2009

In the version of this article initially published, Fig. 5s had an erroneous 8th transmembrane domain on Frizzled. The corrected panel is shown

below. This error has also been corrected in the HTML and PDF versions of the article.

s

–––

H+

H+ H+H+

DIXPDZDEP

+ ++

Nhe2 Fz

Fz FzDsh

Dsh-K/EDsh

DIX

PDZDEPDIXPDZDEP

Low pH

+ ++

–––

–––



Date post:	13-Apr-2015
Category:	Documents
Upload:	sachin
View:	193 times
Download:	22 times

Nature CellBiology

Documents