Asymmetric partitioning of transfected DNA duringmammalian cell divisionXuan Wanga,b, Nhung Lea,b, Annina Denoth-Lippunera,b, Yves Barrala, and Ruth Kroschewskia,1

aInstitute of Biochemistry, Department of Biology, Swiss Federal Institute of Technology Zurich (ETH Zurich), CH-8093 Zurich, Switzerland, and bMolecularLife Science PhD Program, Life Science Zurich Graduate School, CH-8057 Zurich, Switzerland

Edited by Yukiko M Yamashita, University of Michigan, Ann Arbor, MI, and accepted by Editorial Board Member Brigid L. Hogan May 9, 2016 (received forreview April 18, 2016)

Foreign DNA molecules and chromosomal fragments are generallyeliminated from proliferating cells, but we know little about howmammalian cells prevent their propagation. Here, we show thatdividing human and canine cells partition transfected plasmid DNAasymmetrically, preferentially into the daughter cell harboring theyoung centrosome. Independently of how they entered the cell,most plasmids clustered in the cytoplasm. Unlike polystyrene beadsof similar size, these clusters remained relatively immobile andphysically associated to endoplasmic reticulum-derived membranes,as revealed by live cell and electron microscopy imaging. At entry ofmitosis, most clusters localized near the centrosomes. As the twocentrosomes split to assemble the bipolar spindle, predominantlythe old centrosome migrated away, biasing the partition of theplasmid cluster toward the young centrosome. Down-regulationof the centrosomal proteins Ninein and adenomatous polyposiscoli abolished this bias. Thus, we suggest that DNA clustering, clusterimmobilization through association to the endoplasmic reticulummembrane, initial proximity between the cluster and centrosomes,and subsequent differential behavior of the two centrosomes togetherbias the partition of plasmid DNA during mitosis. This process leads totheir progressive elimination from the proliferating population andmight apply to any kind of foreign DNA molecule in mammalian cells.Furthermore, the functional difference of the centrosomes might alsopromote the asymmetric partitioning of other cellular components inother mammalian and possibly stem cells.

foreign DNA | asymmetric cell division | centrosome | endoplasmicreticulum | Ninein

Generally, noncentromeric DNA molecules are mitoticallyinstable in eukaryotes. This results in their apparent disap-

pearance from an ever-increasing proportion of the progeny ofan affected cell (e.g., 1–3). Endogenous sources of such DNA arerecombination byproducts [double minutes, extrachromosomalribosomal (r)DNA circles (ERCs) and other DNA circles (3–6)]or mitotic defects generating noncentromeric chromosomalfragments and cytoplasmic micronuclei (1, 7). Exogenous sour-ces are DNA of pathogens or DNA, typically plasmids, artificiallyintroduced into cells. For the latter, decades of work establishedthat plasmid-born protein expression is transient, persisting onlyfor a few cell cycles (8). This finding is consistent with plasmidDNA being somehow eliminated through divisions. Thus, somemechanisms seem to prevent the propagation of foreign DNAand extrachromosomal DNA in proliferating eukaryotic cells.However, how this is achieved is unclear.In animal cells, DNA sensors mediate the early detection of

exogenous DNA, such as DNA of invading pathogens and arti-ficially introduced DNA (9–11). Both in leukocytes and non-professional immune cells, these can trigger innate immuneresponses, such as cytokine production, autophagy, and apoptosis(9). However, what happens to the DNA molecules themselvesover time is unclear. When microinjected into the nucleus, plas-mid DNA clusters and is expelled into the cytoplasm at mitosis(12). In the cytoplasm, the amount of DNA decreased within a fewhours without completely disappearing, suggesting a rapid degra-dation of a major fraction and the persistence of a minor fraction

of the molecules (13). Within a few hours after introducing DNAinto the cytoplasm, tubular membranes and Emerin, a proteinsynthesized in the endoplasmic reticulum (ER) and subsequentlypredominantly present in the inner nuclear membrane, appear inthe cytoplasm (10, 11). However, the functional relevance of theseobservations is not clear. Furthermore, the destiny of the DNAmolecules, especially during subsequent mitoses, is elusive.To better understand these phenomena and their functional

relevance, we analyzed the fate of transfected plasmid DNA individing mammalian tissue culture cells.

ResultsTransfected Plasmids Localize in a Few Clusters, Which Are Predominantlyin the Cytoplasm. To image the subcellular localization of plasmidDNA upon transfection into cells, we used two fluorescence-basedapproaches. In the first one, rhodamine was covalently coupled to aplasmid (Rho-plasmid) encoding Histone2B fused to enhanced(e)GFP (H2B-eGFP), whereas in the second we introduced a plasmidcontaining Lac operon repeats (pLacO) into cells expressing the Lacinhibitor (LacI) fused to a fluorescent protein [eGFP, monomeric(m)Cherry]. Because of its high affinity to the LacO sequence, theLacI fusion protein efficiently decorated the plasmid.Twenty-four hours after transfection of the Rho-plasmid into

HeLa cells using lipofection, we detected rhodamine fluores-cence either in cells with a few bright foci only in the cytoplasm(61%, filled circle, Fig. 1B) or in cells with foci in the cytoplasmand foci in the nucleus (39%, open circle, Fig. 1B) (Fig. 1 A andB and SI Appendix, Fig. S1A). No cell with only nuclear foci wasfound. Foci intensities varied over four orders-of-magnitude(Fig. 1B), indicating that plasmids aggregated into clusters

Significance

What happens to foreign DNA in mammalian cells during di-visions? We provide evidence that the mitotic machinery helpscope with cytoplasmic clusters of plasmid DNA, a prototype offoreign DNA. The endoplasmic reticulum embraces these clus-ters, probably restricting their mobility and preventing themfrom interfering with chromosomal DNA. Subsequently, theseclusters predominantly partition into the daughter cell with theyoung centrosome and with successive divisions end up in adecreasing fraction of cells within the growing population. Thisprocess is similar to the partition of noncentromeric DNA inbudding yeast and cell fates in stem cells. The behavior dif-ference of the two centrosomes documented here might alsofacilitate asymmetric partition of other cellular components.

Author contributions: X.W., Y.B., and R.K. designed research; X.W. and N.L. performedresearch; A.D.-L. contributed new reagents/analytic tools; X.W. analyzed data; and X.W.,Y.B., and R.K. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission. Y.M.Y. is a guest editor invited by the EditorialBoard.

Freely available online through the PNAS open access option.1To whom correspondence should be addressed: Email: ruth.kroschewski@bc.biol.ethz.ch.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1606091113/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1606091113 PNAS | June 28, 2016 | vol. 113 | no. 26 | 7177–7182

CELL

BIOLO

GY

Dow

nloa

ded

by g

uest

on

July

25,

202

0

containing possibly up to tens of thousands of molecules.Moreover, the cytoplasmic foci were hundreds to thousands foldmore intense than the nuclear ones (Fig.1B). Despite covalentlabeling, the plasmids were successfully transcribed becauseeGFP-H2B was detected in 69% of the analyzed cells (SI Ap-pendix, Fig. S1B). Interestingly, eGFP-H2B was detectable in allcells containing nuclear foci, but only in a fraction of cells con-taining no or only cytoplasmic foci (SI Appendix, Fig. S1C). Werationalized that the cells without nuclear foci might have lostsuch foci after transgene expression. Taken together, our dataare consistent with the notion that nuclear localization is impor-tant for successful transcription.To avoid possible effects of the covalent labeling and the un-

certainty in distinguishing intracellular from extracellular plas-mids, we turned to the LacO/LacI system. The LacI-eGFP andLacI-mCherry proteins contained a nuclear localization signal(NLS) to reduce the cytoplasmic background and a point mutationto avoid tetramerization (14). Twenty-four hours after pLacOtransfection into HeLa stably expressing LacI-eGFP, bright LacIfluorescence was observed in one or a few cytoplasmic foci ofvarious sizes (Fig. 1 C and D). Confirming that these foci con-tained pLacO, a control plasmid without LacO sequence did notinduce such foci (Fig. 1C). Furthermore, the LacI-eGFP focinearly perfectly colocalized with pLacO DNA, as detected byfluorescence in situ hybridization (SI Appendix, Fig. S1 D and E).Similar foci were also observed after pLacO electroporation (SIAppendix, Fig. S1 F and G), indicating that plasmid clustering

was not caused by the lipofection method. Using both methods,the cells containing only one cluster remained predominant evenafter transfection of increasing DNA concentrations (Fig. 1Dand SI Appendix, Fig. S1G). We rarely observed nuclear foci inthese cells, possibly because they were drowned in the nuclearsignals of the LacI fusion proteins.We concluded that independently of transfection and detection

methods the transfected plasmid is essentially deposited in onemain cytoplasmic cluster that contains thousands of molecules.

DuringMitosis, Plasmid Clusters Are Preserved and Partition Predominantlywith the Young Centrosome.Next, we imaged the cytoplasmic pLacOclusters in dividing cells. The clusters did not disassemble duringmitosis and remained in the cytoplasm after mitosis (n >100 clus-ters) (Figs. 2A and 3D). Remarkably, four of eight initial cells withtwo or three pLacO clusters copartitioned all clusters to one of thetwo daughter cells in three consecutive mitoses, resulting in a ma-jority of progeny cells being free of plasmids (Fig. 2A and MovieS1). Analysis of fixed telophase cells showed that in cells with twoor three clusters asymmetric partition patterns were significantlyhigher than random [65% vs. 50% expected for 2:0 (n = 84 cells)and 41% vs. 25% expected for 3:0 (n = 29 cells)] (Fig. 2B).Every mitosis is asymmetric, at least for what concerns cen-

trosomes. Indeed, the centrosomes at the poles of the bipolarspindle and inherited separately by the two resulting daughtercells have different ages (15). Although the relevance of thisasymmetry is unclear, the inheritance of the old and youngcentrosomes correlates with specific cell fates in many stem cells(16–18). We investigated whether the partition of the plasmidclusters correlated with centrosome age. To this end, the cen-trosomes were labeled with stably expressed Centrin1-eGFP,whose abundance correlates with centriole age (19), or with ananti-outer dense fiber protein 2 (ODF2) antibody. ODF2 (Cenexin)is an old centrosome reporter (20). In late mitotic cells, bothproteins decorated most strongly the old centrosome (Fig. 2Cand SI Appendix, Fig. S2). Remarkably, in HeLa and Madin-Darby canine kidney (MDCK) cells the DNA clusters coparti-tioned with the young centrosome in 64.9% and 66.9% of thedivisions, respectively (x: 0 partition patterns, x = 1, 2, or 3clusters) (Fig. 2D). The same effect was observed with a linearfragment (SI Appendix, Fig. S3 A and B). These results weresignificantly different from random, showing that in distinct mam-malian cells the partition of plasmid clusters was biased towardthe young centrosome. To analyze whether this fate is somewhatspecific for DNA or observed for any other particle of similarsize (SI Appendix, Fig. S8D), we investigated how negativelycharged polystyrene beads of 1 μm diameter partition at mitosiswhen introduced into HeLa cells. Unlike pLacO clusters, thebeads did not preferentially copartition with either centrosomeindependently of whether they were transfected alone or to-gether with pLacO (Fig. 2E and SI Appendix, Fig. S3 C–F). Thus,the biased partition of the DNA clusters was driven by neithertheir charge nor their size.Ninein, a component of mature centrioles involved in micro-

tubule anchoring, is thought to mediate the functional differencesbetween old and young centrosomes (18, 21). Furthermore, likeODF2, adenomatous polyposis coli (APC) localizes to maturecentrioles (20, 22). Thus, we tested whether down-regulation ofany of these proteins (validated in SI Appendix, Figs. S5 A and Band S6) interfered with the biased partition of the plasmid inHeLa cells (Fig. 2F). Strikingly, upon Ninein and APC down-regulation, the results were no longer significantly different fromrandom (HeLa siNinein: 44.9%; siAPC 46.2%) (Fig. 2F), sug-gesting that the partition bias was lost in both cases. Ninein down-regulation had a similar effect in canine cells (43.4% MDCK) (SIAppendix, Fig. S5C). As the fraction of cells with only cytoplasmic,nuclear and cytoplasmic, or only nuclear clusters was not changedbetween siControl and siNinein-treated cells, it is unlikely that achange in the localization of the clusters contributed to the lossof the partition bias (SI Appendix, Fig. S9H). Strikingly, neitherNinein overexpression nor ODF2 down-regulation affected the

C D

mCherry

A

Rho

-pla

smid

+Lam

inB

1

B

nr. clusters/cell:

2>2

1

1 m

xz at y*

Focu

s in

tens

ity (A

U)

39%, cytoplasm + nucleus

61%, only cytoplasm

plasmid dose (ng/cm2)

Cel

ls (%

)

0

50

100

5 25 50

LacI-eGFP

10

10²

10

10

10

1 2 >2nr. clusters/cell

(n=69)

no plasmid +pLacO +pControl

y*

Fig. 1. The majority of transfected plasmid DNA localizes in the cytoplasm.(A and B) Cells 24 h after lipofection of rhodamine-labeled plasmid (Rho-plasmid). (A) Cell images with Rho-plasmid foci. LaminB1 immunostaining tovisualize nuclei. (Left, one cytoplasmic focus; Right, two cytoplasmic and onenuclear (arrowhead) foci with z-scan along white dashed line y*; gray line,cell outlines; yellow Inset, enlarged nuclear focus) (B) Fluorescence intensitiesof cytoplasmic (red, blue) and nuclear (black) Rho-plasmid foci were mea-sured in cells with only cytoplasmic (solid circle) and with cytoplasmic andnuclear foci (open circle). Percentage of cells in each pattern (n, pooled cellnumber of three experiments; 17–27 cells per experiment; median valuesindicated). (C and D) Cells stably expressing LacI-eGFP lipofected with pLacO.(C) Images after no plasmid, pControl (as pLacO but LacO repeats were replacedby ORF for mCherry) and pLacO transfection (arrowheads, pLacO clusters).(D) DNA dose dependency of the number of cytoplasmic plasmid clusters percell (mean with SEM of three experiments, n > 50 per experiment). (A and C)Maximal intensity-projected image stacks. [Scale bars (unlabeled), 10 μm.]

7178 | www.pnas.org/cgi/doi/10.1073/pnas.1606091113 Wang et al.

Dow

nloa

ded

by g

uest

on

July

25,

202

0

biased correlation of pLacO clusters with the young centrosome,indicating that the reported link between Ninein and ODF2 isirrelevant for Ninein’s effect on the partition of DNA clustersduring mitosis (Fig. 2F and SI Appendix, Fig. S5 D and E) (23).We conclude that the biased partition of plasmid clusters towardthe young centrosome is the result of a process controlled at thecentrosome level and might be conserved across mammalian cells.

Plasmid Immobility, Initial Vicinity to Centrosomes, and High Mobilityof the Old Centrosome Bias Plasmid Partition. Next we wonderedhow cells biased the inheritance of the plasmid clusters towardthe young centrosome. We first noticed that despite their similarsize the pLacO clusters exhibited a very limited motion in inter-phase cells compared with the cotransfected beads, indicative thatthey interacted differently with cellular structures (Fig. 3 A–C, SIAppendix, Figs. S4 and S8D, and Movie S2). Cluster motion was

also very limited during mitosis in comparison with chromosomeand centrosome dynamics (Fig. 3D, SI Appendix, Figs. S7A andS8B, and Movie S3). Thus, their biased partition was not driven byan active movement of the cluster toward the young centrosome,but rather by a biased movement of the centrosomes at some pointduring mitosis. Because a substantial fraction of HeLa cells elon-gate their spindle asymmetrically (24), we next asked whether thisdrove the segregation of the young centrosome with the clusters.Quantitative analysis indicated that all spindles elongated asym-metrically but to different extents (SI Appendix, Fig. S7 B and C).The plasmid clusters remained in the nonelongating half of thecells, provided that they were at a relative center position inmetaphase (n = 10 cells, SI Appendix, Fig. S7A). However, theelongating half of the cell contained the old centrosome nearly asfrequently as the young centrosome, irrespectively of whetherthe cells had pLacO clusters or not (SI Appendix, Fig. S7C).Therefore, anaphase elongation is not biased to a centrosome typeand thus is not biasing plasmid partition toward any centrosome.Consequently, we analyzed the impact of earlier mitotic events

by live-cell imaging (Fig. 3 D–H). The two centrosomes start toseparate shortly before the nuclear envelope breaks down. Aspreviously reported (25), the two centrosomes reached theirmaximal distance to build the spindle either right before (30%)or shortly after (70%) nuclear envelope breakdown (prophaseand prometaphase pathways) (n = 75). In movies of both cen-trosome-splitting types, we measured the 3D distances of indi-vidual pLacO clusters (C) to the old (O) and the young (Y)centrosome (OC and YC distances, respectively). Remarkably,5 min after they split the old centrosome was on average furtheraway from the plasmid cluster than the young centrosome (SIAppendix, Fig. S8A). To characterize this behavior in individualcells, we calculated the differences between OC and YC (OC-YC)at four stages of mitosis: the last movie-frame before the centro-somes split (before split), when the centrosomes were maximallydistant from each other before metaphase (maximal split), duringmetaphase, and at the second frame of anaphase (Fig. 3E). Thedistance of the cluster to the old centrosome was larger than to theyoung centrosome in 69% of the analyzed cells at the maximalsplit stage; the fraction increased slightly in metaphase (74%) andcame back to 69% in anaphase, consistently with the analysis offixed cells (Fig. 2D). Remarkably, 77% of the clusters that werecloser to the young (red lines) or old centrosome (blue lines) inanaphase were also closer to the same centrosome at the maximalsplit stage (Fig. 3E). Therefore, the spatial association of thecluster with the young centrosome emerged as soon as the twocentrosomes split and was generally maintained afterward.Strikingly, we noticed that before centrosome splitting (−10,

0 min) (Fig. 3D), the majority of plasmid clusters were in thevicinity of the centrosome pair: In 73% of the cases (44 of 60cells), the distance from the cluster to the centrosomes wasshorter than the short diameter of the nucleus (average 17.5 μm),the biggest obstacle in the cell (Fig. 3F and SI Appendix, Fig.S8C). Furthermore, we noticed that in the majority of the ana-lyzed movies the old centrosome moved further than the youngone during splitting, leaving the young centrosome closer to thepLacO cluster [Fig. 3D (0–10 min), Fig. 3 E–G, and Movie S3].Indeed, the displacement of the old centrosome between thetime points “before split” and “maximal split” was in 70% of thecases bigger than that of the young centrosome [displacementratio (old/young centrosome) > 1] (SI Appendix, Fig. S8E, pluscluster). Among the 44 cells in which the cluster was within17.5 μm of the centrosomes before split, 34 old centrosomesmigrated further than their younger counterpart, resulting in82% of these anaphases in a copartition of the young centrosomewith the cluster (28 of 34) (Fig. 3F). In the six remaining cells,partition of the cluster with the old centrosome seemed to becaused by the cells being very small (two cells), by spindle ro-tation (three cells), and by the cluster moving (one cell). In the10 cells where the cluster was close to the centrosome pair at thetime of splitting and the young centrosome moved further thanthe old one, the clusters did not preferentially copartition with

mitosis I

67 h

6 h 7 h

66.5 h

0 hA

40 h

35 h 36 h13.5 h

mitosis II Centrin1-eGFPC

LacI-mCherry Hoechst

ODF2

old young

B

Cel

ls (%

)

**

*

2 (n=84)

rand

om

3(n=29)

0

20

40

60

80

rand

om

2 : 0 3 : 0

clusters

D

HeLa

(n=12

4)

MDCK

(n=57

)

0

50

66.9

64.9

Cel

ls (%

)

100

*** *

Centrosomeyoungold

FE

Bead

pLac

O

46.1

73.2

(n=52)

*

0

50

Cel

ls (%

)

100

n.s.

**

SiN

inei

n(n

=69)

(n=1

53)

SiC

ontro

l62

.844

.9

* *n.s.***

77.2

70.2

46.2

SiO

DF2

(n=5

7)

(n=9

2)S

iCon

trol

SiA

PC

(n=5

2)

0

50

Cel

ls (%

)

100

n.s.

** ***

n.s.

**

mitosis III

LacI-eGFP

Fig. 2. Plasmid clusters are partitioned to the daughter cell harboring theyoung centrosome in a controlled manner. (A) Time-lapse images of the par-tition of three pLacO clusters (arrowheads) in three consecutive mitoses of a cell(dashed line, outline of mitotic cell). (B) Quantification of cells with x:0 partitionpatterns in fixed telophase cells with x clusters (x = 2 or 3 clusters) (dashed bar,theoretical random expectation). (C–F) Correlation analysis between the pLacOclusters and the centrosome type in dividing cells. (C) Representative image of atelophase cell, expressing Centrin1-eGFP and LacI-mCherry, with one pLacOcluster (yellow arrowhead) anti–ODF2-immunostained. (Insets) Enlarged cen-trosome areas. (D–F) Correlations between pLacO clusters and indicated cen-trosome types in late mitotic cells with x: 0 partition patterns (x = 1, 2, and 3clusters or beads). (D and E) Cells without pretreatment. (E) Cells with bothbeads and pLacO clusters. (F) Cells pretreated with siRNA oligos before pLacOtransfection. (A and C) Maximal intensity-projected image stacks; (B and D–F) n,pooled cell number of ≥3 experiments [cell number range per experiment: 14–38 (B, two clusters); 4–19 (B, three clusters); 30–59 (D, HeLa); 17–22 (D, MDCK);16–18 (E), 26–75 (F, siControl, Left graph); 10–34 (F, siNinein), 13–37 (F, siControl,Right graph); 13–24 (F, siODF2); 10–22 (F, siAPC)]; statistics: (B) Binomial test,(D–F) Binomial test to compare experimental to random frequency, χ2 test tocompare pLacO with bead, siRNA with siControl, and siODF2 with siAPC.*P < 0.05; **P < 0.01; ***P < 0.001; n.s., not significant. (Scale bars, 10 μm.)

Wang et al. PNAS | June 28, 2016 | vol. 113 | no. 26 | 7179

CELL

BIOLO

GY

Dow

nloa

ded

by g

uest

on

July

25,

202

0

either centrosome (Fig. 3F). Detailed measurements of displace-ment and speed during splitting confirmed the relative immobilityof the plasmid cluster, and, on average, the bigger displacement ofthe old centrosome compared with the young one (Fig. 3G and SIAppendix, Fig. S8 B and E). The differential displacement of thetwo centrosomes in control cells without pLacO clusters was notsignificantly different compared with cells with clusters (Fig. 3Hand SI Appendix, Fig. S8E). Finally, the LacI-GFP intensities andapproximated cluster sizes were independent of whether they wereclose or far from the centrosome pair (SI Appendix, Fig. S8D).Thus, our data indicate that the two centrosomes behave intrinsi-cally differently during spindle formation and that the plasmidcluster does not induce this difference.We performed this analysis also upon Ninein knockdown

to determine why this affected the partition bias (SI Appendix,Fig. S9). The distance between cluster and centrosomes before

centrosome split and the centrosome displacements early inmitosis revealed no major differences between siControl andsiNinein-treated cells (SI Appendix, Fig. S9 B, C, and E). How-ever, upon Ninein down-regulation, the mitotic spindles con-taining aligned centrosomes moved more frequently comparedwith siControl cells (40% SiNinein, n = 15 cells; 5% siControl,n = 19 cells), randomizing the position of the centrosomes rel-ative to the cluster (SI Appendix, Fig. S9F). Thus, Ninein down-regulation did not abolish the establishment but the maintenanceof the spatial association between the cluster and the youngcentrosome. In these cells, the microtubules emanating from thecentrosomes might no longer establish a stable connection to thecell cortex, as already proposed (26).In summary, the three requirements driving the partition of

plasmid clusters toward the young centrosome are (i) that theplasmid cluster must be relatively immobile and (ii) in proximity

b1

b2

b3b4

cluster

b1

b2b3

b4

cluster

b1

b2b3

b4

cluster

b1

b2

b3

b4

cluster

b1

b2

b3

b4

cluster

far

LacI-mCherry Centrin1-eGFPD

-10 min 15 min

metaphase

25 min 35 min 45 min

before split

0 min

split

5 min

max split

10 min

anaphase

H

Dis

plac

emen

t (m

)

5

10

15

0

centrosome

youngold

(min)100 5 15 20(n)67 3267 53 10

G

(n)

5

10

15

0

Dis

plac

emen

t (m

)

cluster

(min)60 5160 59

100 5 153420

****

****

****

********

centrosome

youngold

44

16

34

10

28 6

4 6

4 4

3 5

8

8

cluster-centrosome: <17.5 m

displacement:old >young

before split during split

close

anaphase

cluster-centrosome: >17.5 m

displacement:old < young

displacementold >young

displacement:old < young

old youngcentrosome

clusternucleus

-20

-10

0

10

20

OC>YC: 69% 69%74%

max sp

lit

metaph

ase

anap

hase

befor

e spli

t

(n=87)

OC

-YC

(m

)

old

young cluster

YC

OC

A

-5 5

-5

5

cluster

beads

x( m)

y( m)

bead (n=22)

cluster (n=17)

0 50 100 150 2000

5

10

15

dT(s)

MS

D (

m²)

C0 s 100 s 200 s 300 s 400 s B

E F

Fig. 3. Predominantly the old centrosome moves away from a stable plasmid cluster and the young centrosome during centrosome splitting. (A) Repre-sentative images of a cell with four beads (blue, b1–b4) and one pLacO cluster (red) at indicated times (gray lines, reference at t = 0 s. (B) Trajectories (0–400 s)of beads and the pLacO cluster of one cell. (C) Mean square displacements (MSD) of beads and pLacO clusters in 16 cells with both beads (one to four beadsper cell) and pLacO clusters (one to two clusters per cell) (mean with SEM; dT, interval time). (D) Representative time-lapse images of a cell with one pLacOcluster during mitosis (white and gray arrowheads, old and young centrosomes, respectively; dashed circle, original location of centrosome pair at 0 min).(E) Difference of the distances of pLacO cluster to the old (OC) and young centrosome (YC), respectively, in individual cells (OC-YC). Results at the times:before split, maximal split, metaphase, and the second frame at anaphase. Percentage of cells with OC > YC indicated above. [3D distances; red, blue lines,clusters associated with young (red) and old centrosomes (blue) at anaphase] (F) Schematic behavior summary of cluster and centrosome dynamics in 60 cellscontaining one plasmid cluster. (G) Displacements of the centrosomes and pLacO clusters during splitting in cells with one plasmid cluster (mean with SD; **P <0.01; ****P < 0.0001). (H) As in G, but for cells without plasmid clusters. (G and H) Displacements are 3D distances to the cluster’s or centrosomes’ start locations5 min before split; (C) n, cluster or bead number from multipositions of one reproduced experiment; (E–H) n, pooled cluster (E) or cell (F–H); number of > 3experiments, number range per experiment: 16–31 (E), 6–19 (F and G), 11–24 (H); statistics, Dunn’s multiple comparisons test (G) or Wilcoxon matched-pairssigned rank test (H) was performed for each time point separately; only significant differences were indicated. (A and D) Maximal intensity-projected imagestacks. (Scale bars, 10 μm.)

7180 | www.pnas.org/cgi/doi/10.1073/pnas.1606091113 Wang et al.

Dow

nloa

ded

by g

uest

on

July

25,

202

0

of the centrosomes before splitting. (iii) During spindle formation,the old centrosome migrates away from the young centrosome andplasmid cluster, leaving the latter two together. The maintenanceof this association toward the end of mitosis requires the orien-tation of the mitotic spindle to remain stable.

Mitotic Plasmid Cluster Appears to Be Physically Linked to the ER.One of the intriguing prerequisites for the biased inheritance ofthe plasmid cluster is its immobility. Because beads of similarelectrostatic nature and size do not show such immobility (Fig. 3A–C and SI Appendix, Fig. S4), we reasoned that DNA clustersmight be anchored to some structure. Supporting this idea, mostof the plasmid clusters were very close to the cell cortex duringmitosis (Fig. 3D). The intermediate filament protein Vimentincan cage damaged proteins and partitions asymmetrically inmammalian cells (27, 28). It is also well established that the ERis close to the cortex, away from the mitotic spindle (29). Wetherefore probed if the DNA cluster contacted cytoskeletal ele-ments or membrane structures during mitosis.None of YFP-αTubulin, Vimentin, and actin showed any evi-

dent association with the plasmid clusters in metaphase cells (SIAppendix, Fig. S10A). However, the ER membrane reporterSec61-mCherry always surrounded the plasmid clusters and wassometimes (10%, n = 21 mitotic cells) even enriched at that po-sition (Fig. 4A). A reporter of ER lumen, eGFP–Lys-Asp-Glu-Leu(KDEL), and the nuclear membrane proteins Emerin and Lap2β,which localize in the ER during mitosis, also closely surroundedthe mitotic plasmid clusters (SI Appendix, Fig. S10B). The tightassociation of the DNA cluster, but not of the beads, with the ERwas also observed in all analyzed interphase cells (Fig. 4A).To visualize the relationships of the pLacO cluster and mem-

branes at high resolution, we performed correlative light with fo-cused ion beam scanning electron microscopy. Tubular membranes,reminiscent of ER tubules, were in close proximity to the plasmidlocation (Fig. 4B and SI Appendix, Fig. S10C). Most remarkably, in

several distinct regions, electron-dense bridges were found betweenthe plasmid cluster and the surrounding tubular structures (Fig.4B, red arrowheads, and SI Appendix, Fig. S10C), suggestingphysical connections. No such structures were observed aroundindividual metaphase chromosomes (Fig. 4B, gray box). Fur-thermore, the plasmid clusters were of equal or higher electrondensity compared with the chromosomal DNA, indicating a tightpacking or high DNA concentration (Fig. 4B and SI Appendix,Fig. S10C). Together, our data suggest that plasmid clusters arephysically linked to ER membranes. This confinement separatesthe plasmid from the mitotic spindle area and probably restrictsthe mobility of the plasmid.

DiscussionHere, we show that 24 or more hours after transfection, themajority of transfected plasmid is mostly in one cluster in thecytoplasm, and only a very small fraction is present in the nu-cleus. Moreover, we show that during mitosis, plasmid clustersare outside of the spindle region and appear to be physicallyassociated to the ER, even in mitosis. Sting (stimulator of IFNgenes), an ER membrane protein involved in the transmission ofsignals from DNA sensors to the native immune system (9),might be implicated therein. Plasmid DNA microinjected intothe nucleus is expelled into the cytoplasm after mitosis (12).These results support the notion that plasmid DNA, as a modelfor foreign DNA, is actively separated from chromosomal DNAin interphase and during the open mitosis of mammalian cells.We also show that the plasmid clusters preferentially partition

to the daughter cell with the young centrosome. This asymmetricpartition is the result of a combination of events. First, clusteringof the DNA molecules promotes the fidelity of partition by re-ducing the number of entities to be partitioned. However, althoughclustering is observed in yeast cells as well (3), its mechanisms arecurrently unknown. Second, a close proximity of the plasmidcluster with the centrosomes at the start of mitosis is required for

xz at y*

y*

embedding: 90 %

metaphase

B

Merge+HoechstA Sec61-mCherryLacI-eGFP

interphase 1 m

LacI-mCherry

DIOC6

1 mz3z1 z2

1 m

Bead Sec61-eGFP Merge+Hoechst

interphase 1 m

Merge+HoechstSec61-mCherryLacI-eGFP

enriched: 10 %

metaphase

Fig. 4. Plasmid clusters connect to ER in mitotic cells. (A) PLacO clusters in mitotic (Left) and interphase (Right, panel 1) cells expressing LacI-eGFP andSec61-mCherry and beads in interphase cells expressing Sec61-eGFP (Right, panel 2). (Insets, areas of clusters and beads; Left, panel 1: arrowhead, pLacOcluster; panels 1 and 2: quantification of 21 metaphase cells (pool of three experiments); Right, panel 1: representative of 70 cells (pool of three ex-periments); Right, panel 2: representative of 27 cells (pool of two experiments, reproduced also with other membrane reporters), (B) Correlative fluo-rescence with electron microscopy (EM) of a metaphase cell containing one pLacO cluster. The EM image corresponds to the confocal image at aperpendicular view (xz). Boxed areas show enlarged plasmid cluster (blue) and chromosomes (gray) (yellow arrowhead, tubular membrane; red arrow-head, connecting junction between the pLacO cluster and membrane; z1, z2, and z3, the blue-boxed area at three different z levels). (Single-z-focusimages) [Scale bars (unlabeled), 10 μm.]

Wang et al. PNAS | June 28, 2016 | vol. 113 | no. 26 | 7181

CELL

BIOLO

GY

Dow

nloa

ded

by g

uest

on

July

25,

202

0

biasing plasmid partition. Future work will determine how thisproximity is established in the first place. Third, the two centro-somes move differentially during early mitosis. During centrosomesplitting, which prepares the formation of the bipolar spindle, theold centrosome moves further away from its original positioncompared with the young centrosome. Importantly, this differen-tial behavior already exists in HeLa cells without plasmid DNA.This might be due to the old centriole nucleating more astralmicrotubules (19) and the amount of a specific subgroup of astralmicrotubules required for asymmetric stem cell divisions (18, 30).Last but most importantly, we found the plasmid cluster to berelatively immobile during mitosis, which is a prerequisite fordifferential centrosome movement to bias the partition of theDNA clusters together with the young centrosome. We suggestthat physical links between the plasmid cluster and the ERmembrane anchor the plasmid DNA and thus restrict its mobility.In budding yeast, noncentromeric plasmid DNA and endoge-

nous circular DNA popping out from the rDNA locus act asaging factors (31). They are inherited by the yeast mother cell,reducing her division potential with each subsequent division,compared with newborn daughter cells (3, 32). The yeast mothercell also inherits the young spindle pole, the functional homologof the young centrosome in mammalian cells (33). Therefore, wespeculate that the asymmetric partition of the plasmid clusterpromotes the maintenance of a high division potential in themajority of the progeny, namely, in those cells that are therebycleared of foreign DNA. In reverse, cells containing foreign DNAmight be more prone to aging. In support of that notion, we foundthat the cell cycle duration of the daughter cell with plasmid clusteris on average 0.7 h longer than that of its sibling (SI Appendix, Fig.S11), similar to the effect of protein aggregates on the proliferationof neural stem cells (28). Germ and renewing stem cells often retainthe old centrosome upon division. Therefore, our data predict thatthese cells are cleared of any entering foreign DNA. Future studieswill determine whether this is the case and how relevant this is forthe biology of these cells.

To conclude, we identified a mechanism facilitating the biasedpartition of plasmid DNA to the daughter cell inheriting theyoung centrosome during mitosis. By this, the fraction of cellsharboring DNA clusters in a population is expected to decreasewith continued divisions, resulting in their apparent elimination.Further, we suggest that this mode of asymmetric partition ofplasmid DNA acts in many mammalian cell types and protectsthe chromosomes from foreign DNA. We postulate that this isespecially important for cell lineages with high division potential,such as germ and stem cells. Last but not least, we expect that thedifferential behavior of the two centrosomes identified will proverelevant for the asymmetric partition of other cellular compo-nents, possibly in many mammalian cell types.

Materials and MethodsDetailed materials and methods are provided in SI Appendix. Presentedexperiments are approved by the Swiss Federal Office of Public Health.

LacO Plasmid. PLacO (13.5 kb) with 256 x LacO spacer (aattgtgagcggataa-caattt-gtggccacatgtgg) is a gift from Susan M. Gasser, Friedrich MiescherInstitute, Basel, Switzerland (34).

Centrosome Classification. The method to distinguish old and young cen-trosomes is described in SI Appendix, Fig. S2A. To validate the classificationof the centrosome age in HeLa cells stably expressing Centrin1-eGFP, thefluorescence intensity of Centrin1-eGFP was compared with that of immu-nostained ODF2 at centrosomes (SI Appendix, Fig. S2 B and C).

ACKNOWLEDGMENTS. We thank the R.K. and Y.B. groups for discussionsand feedback, and the I. Näthke laboratory for generous support with theadenomatous polyposis coli experiments. This study was supported in part bythe ScopeM/Swiss Federal Institute of Technology Zurich, the National Cen-ters of Competence in Research (NCCR) “CO-ME” (to R.K.); European Re-search Council Program ERC2-73915-09 (to Y.B.); Swiss National ScienceFoundation Grant NF2-77714-13 (to Y.B.); Eidgenössiche Technische HochschuleZürich Research Grant TH0-20534-09 (to Y.B.); and by Novartis Foundation Grant270-638-14 (to Y.B.).

1. Crasta K, et al. (2012) DNA breaks and chromosome pulverization from errors inmitosis. Nature 482(7383):53–58.

2. Sinclair DA, Guarente L (1997) Extrachromosomal rDNA circles—A cause of aging inyeast. Cell 91(7):1033–1042.

3. Denoth-Lippuner A, Krzyzanowski MK, Stober C, Barral Y (2014) Role of SAGA in theasymmetric segregation of DNA circles during yeast ageing. eLife 3:e03790.

4. Cowell JK (1982) Double minutes and homogeneously staining regions: Gene ampli-fication in mammalian cells. Annu Rev Genet 16:21–59.

5. Gaubatz JW (1990) Extrachromosomal circular DNAs and genomic sequence plasticityin eukaryotic cells. Mutat Res 237(5-6):271–292.

6. Møller HD, Parsons L, Jørgensen TS, Botstein D, Regenberg B (2015) Extrachromo-somal circular DNA is common in yeast. Proc Natl Acad Sci USA 112(24):E3114–E3122.

7. Hatch EM, Fischer AH, Deerinck TJ, Hetzer MW (2013) Catastrophic nuclear envelopecollapse in cancer cell micronuclei. Cell 154(1):47–60.

8. Hahn P, Scanlan E (2010) Gene delivery into mammalian cells: An overview on existingapproaches employed in vitro and in vivo. Top Curr Chem 296:1–13.

9. Paludan SR (2015) Activation and regulation of DNA-driven immune responses.Microbiol Mol Biol Rev 79(2):225–241.

10. Kobayashi S, et al. (2015) BAF is a cytosolic DNA sensor that leads to exogenous DNAavoiding autophagy. Proc Natl Acad Sci USA 112(22):7027–7032.

11. Ibrahim N, Wicklund A, Wiebe MS (2011) Molecular characterization of the hostdefense activity of the barrier to autointegration factor against vaccinia virus. J Virol85(22):11588–11600.

12. Shimizu N, Kamezaki F, Shigematsu S (2005) Tracking of microinjected DNA in livecells reveals the intracellular behavior and elimination of extrachromosomal geneticmaterial. Nucleic Acids Res 33(19):6296–6307.

13. Lechardeur D, et al. (1999) Metabolic instability of plasmid DNA in the cytosol: Apotential barrier to gene transfer. Gene Ther 6(4):482–497.

14. Kumaran RI, Spector DL (2008) A genetic locus targeted to the nuclear periphery inliving cells maintains its transcriptional competence. J Cell Biol 180(1):51–65.

15. Nigg EA, Stearns T (2011) The centrosome cycle: Centriole biogenesis, duplication andinherent asymmetries. Nat Cell Biol 13(10):1154–1160.

16. Januschke J, Llamazares S, Reina J, Gonzalez C (2011) Drosophila neuroblasts retainthe daughter centrosome. Nat Commun 2:243.

17. Yamashita YM, Mahowald AP, Perlin JR, Fuller MT (2007) Asymmetric inheritance ofmother versus daughter centrosome in stem cell division. Science 315(5811):518–521.

18. Wang X, et al. (2009) Asymmetric centrosome inheritance maintains neural progen-itors in the neocortex. Nature 461(7266):947–955.

19. Piel M, Meyer P, Khodjakov A, Rieder CL, Bornens M (2000) The respective contri-butions of the mother and daughter centrioles to centrosome activity and behavior invertebrate cells. J Cell Biol 149(2):317–330.

20. Anderson CT, Stearns T (2009) Centriole age underlies asynchronous primary ciliumgrowth in mammalian cells. Curr Biol 19(17):1498–1502.

21. Delgehyr N, Sillibourne J, Bornens M (2005) Microtubule nucleation and anchoring atthe centrosome are independent processes linked by ninein function. J Cell Sci 118(Pt8):1565–1575.

22. Louie RK, et al. (2004) Adenomatous polyposis coli and EB1 localize in close proximityof the mother centriole and EB1 is a functional component of centrosomes. J Cell Sci117(Pt 7):1117–1128.

23. Ibi M, et al. (2011) Trichoplein controls microtubule anchoring at the centrosome bybinding to Odf2 and ninein. J Cell Sci 124(Pt 6):857–864.

24. Kiyomitsu T, Cheeseman IM (2013) Cortical dynein and asymmetric membrane elon-gation coordinately position the spindle in anaphase. Cell 154(2):391–402.

25. Tanenbaum ME, Medema RH (2010) Mechanisms of centrosome separation and bi-polar spindle assembly. Dev Cell 19(6):797–806.

26. Srivatsa S, Parthasarathy S, Molnár Z, Tarabykin V (2015) Sip1 downstream Effectorninein controls neocortical axonal growth, ipsilateral branching, and microtubulegrowth and stability. Neuron 85(5):998–1012.

27. Ogrodnik M, et al. (2014) Dynamic JUNQ inclusion bodies are asymmetrically in-herited in mammalian cell lines through the asymmetric partitioning of vimentin.Proc Natl Acad Sci USA 111(22):8049–8054.

28. Moore DL, Pilz GA, Araúzo-Bravo MJ, Barral Y, Jessberger S (2015) A mechanism forthe segregation of age in mammalian neural stem cells. Science 349(6254):1334–1338.

29. English AR, Voeltz GK (2013) Endoplasmic reticulum structure and interconnectionswith other organelles. Cold Spring Harb Perspect Biol 5(4):a013227.

30. Mora-Bermúdez F, Matsuzaki F, Huttner WB (2014) Specific polar subpopulations ofastral microtubules control spindle orientation and symmetric neural stem cell di-vision. eLife 3:e02875.

31. Denoth Lippuner A, Julou T, Barral Y (2014) Budding yeast as a model organism tostudy the effects of age. FEMS Microbiol Rev 38(2):300–325.

32. Shcheprova Z, Baldi S, Frei SB, Gonnet G, Barral Y (2008) A mechanism for asymmetricsegregation of age during yeast budding. Nature 454(7205):728–734.

33. Pereira G, Tanaka TU, Nasmyth K, Schiebel E (2001) Modes of spindle pole body in-heritance and segregation of the Bfa1p-Bub2p checkpoint protein complex. EMBO J20(22):6359–6370.

34. Rohner S, Gasser SM, Meister P (2008) Modules for cloning-free chromatin tagging inSaccharomyces cerevisae. Yeast 25(3):235–239.

7182 | www.pnas.org/cgi/doi/10.1073/pnas.1606091113 Wang et al.

Dow

nloa

ded

by g

uest

on

July

25,

202

0