RESEARCH ARTICLE SUMMARY◥

CELLULAR STRUCTURE

Increased spatiotemporal resolutionreveals highly dynamic dense tubularmatrices in the peripheral ERJonathon Nixon-Abell,* Christopher J. Obara,* Aubrey V. Weigel,* Dong Li,Wesley R. Legant, C. Shan Xu, H. Amalia Pasolli, Kirsten Harvey, Harald F. Hess,Eric Betzig, Craig Blackstone,† Jennifer Lippincott-Schwartz†

INTRODUCTION: The endoplasmic reticulum(ER) is a continuous, membrane-bound organ-elle, spanning from the nuclear envelope to theouter cell periphery, that contacts and influencesnearly every other cellular organelle. In the peri-pheral ER, prevailing models propose a systemof interconnected tubules and flattened sheetsmaintainedbydistinctproteins.Becausemutationsin these proteins and resultant ER irregularitiescoincidewith various neurologic disorders, char-acterizing ER morphology is critical in under-standing its roles in the basic biology of cells inboth health and disease. Given limitations inimaging technologies, determining the dyn-

amic rearrangements and fine ultrastructureof the peripheral ER has proven challenging.

RATIONALE: Previous work characterizingperipheral ER structure has relied extensivelyon diffraction-limited optical microscopy todescribe gross morphology and dynamics, andelectron microscopy (EM) for ultrastructuraldetails. Regrettably, the respective spatial andtemporal limitations of these techniques canobscure underlying cell processes where intri-cate morphology and/or rapid dynamism areimportant. Additionally, efforts to characterizeprotein distribution in the peripheral ER have

presented confounding evidence regarding thelocalization of tubular junction-forming atlas-tin guanosine triphosphatases to sheets, andconcerning the induction of sheet proliferationafter atlastin overexpression. We exploited avariety of emerging superresolution (SR) mi-croscopy techniques to collectively provide un-precedented spatiotemporal resolution thatchallenges prevailing models regarding peri-pheral ER morphology, dynamics, and pro-tein distribution.

RESULTS: We used a combination of five SRtechnologies, with complementary strengths andweaknesses in the spatial and temporal domains,to image the peripheral ER in live and fixed cells.Using novel analytical approaches to study bothprotein and lipid components, we found that

many structures previous-ly proposed to be flatmem-brane sheets are insteaddensely packed tubulararrays—a previously un-described structurewe terman ER matrix. These ma-

trices can become astoundingly compact, withspaces between the tubules far beneath theresolvable power of even most SR technolo-gies. We observed dynamic oscillations of ERtubules and junctions, with matrices rapidly in-terconverting from tight to loose arrays, givingrise to different apparent morphologies de-pendent upon how closely their three-wayjunctions are clustered. We demonstrate howthese ERmatrices have beenmisinterpreted asa result of the spatiotemporal limitations ofearlier imaging technologies. Finally,we accountfor the distribution of atlastin and other ER-shaping proteins within these structures.

CONCLUSION: Theapplicationof cutting-edgeSR technologies to the peripheral ER has estab-lished a precedent for studying its dynamics andstructural properties in living cells. The specificfinding of dense tubular matrices in areas pre-viously thought of as flat sheets provides a newmodel formaintaining and generating ER struc-ture. Reorganization from tight to loose tubularnetwork arrays may allow the ER to rapidlyreach outward to the cell periphery duringmigration or other cell shape changes. More-over, tight clusters of junctions may functionas sites for sequestering excess membrane pro-teins and lipids or for contacting other organ-elles. Improved spatiotemporal resolution ofER structure and dynamics, as shown here,should help to address these and other keyissues regarding ER function in healthy cellsand during disease pathogenesis.▪

RESEARCH

SCIENCE sciencemag.org 28 OCTOBER 2016 • VOL 354 ISSUE 6311 433

The list of author affiliations is available in the full article online.*These authors contributed equally to this work.†Corresponding author. Email: lippincottschwartzj@janelia.hhmi.org (J.L.-S.); blackstc@ninds.nih.gov (C.B.)Cite this article as J. Nixon-Abell et al., Science 354, aaf3928(2016). DOI: 10.1126/science.aaf3928

ON OUR WEBSITE◥

Read the full articleat http://dx.doi.org/10.1126/science.aaf3928..................................................

Dense tubular matrices in the peripheral ER. New superresolution imaging modalities reveal thatperipheral ER sheets are actually densely clustered tubules and interconnecting junctions. Shown isthe distribution of an ER protein marker (3D-SIM) (upper right), internal cellular lipids (LLS-PAINT)(lower left), and an EM reconstruction (FIB-SEM) (upper left) demonstrating tubular matrices in theperipheral ER at high resolution.

on October 10, 2020

http://science.sciencem

ag.org/D

ownloaded from

RESEARCH ARTICLE◥

CELLULAR STRUCTURE

Increased spatiotemporal resolutionreveals highly dynamic dense tubularmatrices in the peripheral ERJonathon Nixon-Abell,1,2* Christopher J. Obara,3,4* Aubrey V. Weigel,3,4* Dong Li,4,5

Wesley R. Legant,4 C. Shan Xu,4 H. Amalia Pasolli,4 Kirsten Harvey,2 Harald F. Hess,4

Eric Betzig,4 Craig Blackstone,1†‡ Jennifer Lippincott-Schwartz3,4†‡

The endoplasmic reticulum (ER) is an expansive, membrane-enclosed organelle that playscrucial roles in numerous cellular functions.We used emerging superresolution imagingtechnologies to clarify the morphology and dynamics of the peripheral ER, which contactsand modulates most other intracellular organelles. Peripheral components of the ER haveclassically been described as comprising both tubules and flat sheets.We show that thissystem consists almost exclusively of tubules at varying densities, including structures thatwe term ER matrices. Conventional optical imaging technologies had led to misidentificationof these structures as sheets because of the dense clustering of tubular junctions and apreviously uncharacterized rapid form of ER motion.The existence of ER matrices explainsprevious confounding evidence that had indicated the occurrence of ER “sheet” proliferationafter overexpression of tubular junction–forming proteins.

The ER is a continuous, membranous net-work extending from the nuclear envelopeto the outer periphery of cells; it plays vitalroles in processes such as protein synthe-sis and folding, mitochondrial division, cal-

cium storage and signaling, and lipid synthesisand transfer. In the cell periphery, theER is thoughtto exist as an elaborate membrane system thatmakes contact with nearly every other cellularorganelle. Prevailing models of its structure pro-pose a complex arrangement of interconnectedtubules and sheets, each of which is maintainedby distinct mechanisms (1, 2). Numerous proteinsare involved in maintaining this complex struc-tural organization.Membrane curvature-stabilizingproteins, includingmembers of the reticulon (RTN)and REEP families, contain hydrophobic hairpindomains that are thought to be responsible forpromoting curvature in ER tubules via scaffold-ing and hydrophobic wedging. Members of theatlastin (ATL) family of dynamin-related guano-sine triphosphatases (GTPases) are thought tomediate the formation of tubular three-way junc-tions, giving rise to the characteristic polygonal

tubular network (3). Meanwhile, an alternativecomplement of proteins is proposed to regulatethe structure of ER sheets, with p180, kinectin,and CLIMP63 all thought to play a role in shaping,helicoidal stacking, and luminal spacing (3). Mu-tations inmany of these ER-shaping proteins areconnected to a variety of human disease condi-tions, most notably the hereditary spastic pa-raplegias (4). Thus, characterizing ER morphologyis critical to understanding the basic biology ofcells in both health and disease.Determining the structure of the ER is challeng-

ing because of limitations in our ability to visualizethe intricate nature of its morphology. The pe-ripheral ER is particularly susceptible to thisconstraint, given its well-documented dynamicrearrangements and fine ultrastructure (5, 6). Thesecharacteristics impede attempts to derive functionalinformation based on changes to ER structure. Therecent development of various superresolution(SR) imaging approaches, however, offers an op-portunity to examine ER structure and dynamismwith substantially improved spatiotemporal resolu-tion.Here, we used five different SRmodalities, eachhaving complementary strengths and weaknessesin the spatial and temporal domains, to examineER structure and dynamics. A high-speed variationof structured illumination microscopy (SIM) al-lowed ER dynamics to be visualized at unprece-dented speeds and resolution. Three-dimensionalSIM (3D-SIM) and Airyscan imaging allowedcomparison of the fine distributions of differentER-shaping proteins. Finally, lattice light sheet-point accumulation for imaging innanoscale topog-raphy (LLS-PAINT) and focused ionbeamscanningelectron microscopy (FIB-SEM) permitted 3D char-

acterization of different ER structures. Thoroughlyprobing the ER in this manner provides unprece-dented information about the morphology anddynamics of this organelle, including the charac-terization of a previously underappreciated struc-ture within the peripheral ER.

ER tubules and junctions undergo rapidmotion in living cells

ER tubules are known to undergo rapid struc-tural rearrangements, occurring over seconds orminutes, yet examination of these processes hastypically been confined to the extension and re-traction of tubules and the formation of tubularthree-way junctions (5, 6). To obtain a more com-prehensive picture of tubular motion, we usedhigh-speed SIMwith grazing incidence illumina-tion (GI-SIM; seematerials andmethods) (7). Thislive SR imaging modality (resolution ~100 nm)uses light beams counterpropagating just abovethe sample substrate to image cellular featuresnear the basal plasma membrane at frequenciesup to 40 Hz. This translates to a factor of 4 to 10increase in acquisition speed, relative to whatcan be practically achieved with spinning-diskconfocal microscopy for imaging the ER, and afactor of ~2 improvement in resolution.With GI-SIM, we imaged COS-7 cells express-

ing an ER membrane marker (mEmerald-Sec61b,henceforth Sec61b) to track ER tubules. Increasedspatiotemporal resolution revealed a novel formof ER motion consisting of remarkably rapid tu-bular fluctuations (Fig. 1 and movie S1). Using amodified skeletonization algorithm (8) to trackthe movement of ER tubules (Fig. 1B), we iden-tified oscillations with a mean peak-to-peak am-plitude of 70 ± 50 nm, occurring an average of 4 ±1 times per second (means ± SD; n = 1755 tubulesfrom 8 cells) (Fig. 1, C and D). Traditional imag-ing modalities have the ability to localize pre-cisely this tubular motion only if the tubules andjunctions are sufficiently sparse. Also, a largeproportion of this motion often occurs on tooshort a time scale to be effectively tracked usingspinning-diskconfocalmicroscopyat imagingspeedscommonly reported in the literature (6, 9). Thissuggests that in dense regions, tubular ERmotionand morphology are likely to be obscured whenusing traditional imaging modalities. Theserapid fluctuations we observed in COS-7 cellswere also found in an unrelated cell type (U-2OS) as well as in COS-7 cells expressing a luminalER marker (mEmerald-ER3, henceforth ER3;see materials and methods) instead of Sec61b(Fig. 1, C to E, and tables S1 and S2).In addition to the tubules themselves, three-

way junctions also exhibited appreciable motionover very short time scales (Fig. 1, F and G, andmovie S1). Three-way junctions were identifiedfrom the skeletonization of fluorescent ER images.Skeletonized pixels with exactly two neighborswere considered to be part of a branch (Fig. 1F,white), and pixels with more than two neighborswere considered junctions (Fig. 1F, overlaid withcyan dots). Three-way junctions were then treatedas single particles and tracked (Fig. 1F, green).The time-averaged mean square displacement

RESEARCH

SCIENCE sciencemag.org 28 OCTOBER 2016 • VOL 354 ISSUE 6311 aaf3928-1

1Cell Biology Section, Neurogenetics Branch, National Instituteof Neurological Disorders and Stroke (NINDS), Bethesda, MD,USA. 2Department of Pharmacology, UCL School of Pharmacy,University College London, London, UK. 3Cell Biology andMetabolism Program, Eunice Kennedy Shriver National Instituteof Child Health and Human Development (NICHD), Bethesda,MD, USA. 4Janelia Research Campus, Howard Hughes MedicalInstitute (HHMI), Ashburn, VA, USA. 5National Laboratory ofBiomacromolecules, Institute of Biophysics, Chinese Academyof Sciences, Beijing, China.*These authors contributed equally to this work. †These authorscontributed equally to this work. ‡Corresponding author. Email:blackstc@ninds.nih.gov (C.B.); lippincottschwartzj@janelia.hhmi.org (J.L.-S.)

on October 10, 2020

http://science.sciencem

ag.org/D

ownloaded from

(MSD) of three-way junctions can be described asMSD ~ ta, where a describes a particle’s motion asBrownian (a = 1), subdiffusive (a < 1), or super-diffusive (a > 1) (10). Junctiondynamics exhibited abroad distribution of a (Fig. 1G and table S1).Again, these results were consistent among cellsexpressing membrane or luminal markers aswell as in other cell types such as U-2 OS (Fig. 1Gand tables S1 and S2).Such rapid dynamics can often represent ther-

mally derived motion in systems, but in this con-text thermal energy alone does not appear to beresponsible for driving ER dynamics. Indeed, themotion of both tubules and three-way junctionscharacterized above was substantially altered bya variety of biological perturbations. Diverse treat-ments affecting access to cellular energy sources[deoxyglucose (DOG) + sodium azide (NaN3);aluminum fluoride (AlF)], cytoskeletal dynamics(blebbistatin), or protein translation (puromycin;cycloheximide) each reduced the amplitude andincreased the frequency of oscillations to levelsconsistent with thermally derived Brownian mo-tion (11) (Fig. 1, C to E). Additionally, the motionof three-way junctions was dampened (Fig. 1G).Although the broad susceptibility of rapid ERdynamics to pharmacological perturbation doesnot elucidate the direct source of the motion, itsuggests a broader role for cellular dynamics indriving ER motion, as it is affected by a range ofdisparate processes. Of note, treatment with the

microtubule-depolymerizing agent nocodazole in-creased the frequency of motion without anynoticeable effect on amplitude (Fig. 1E and tableS2), so it appears that at least in some situations,the amplitude and frequency of tubule oscilla-tions can be uncoupled.

Peripheral “sheets” appear highlydynamic and are riddled with spaces

GI-SIM also permitted rapid imaging of themorphology and dynamics of structures that ap-peared to be flat peripheral sheets by diffraction-limited epifluorescence, leading to several highlyunexpected observations. At the improved spa-tiotemporal resolution afforded by GI-SIM, mostperipheral “sheets” do not appear continuous,but rather are riddled throughout with spacesdevoid of Sec61b fluorescence (Fig. 2, A and B).These spaces are highly dynamic and denselydistributed across the structure (Fig. 2B, kymo-graphs). To analyze these dynamics, we used afluorescence inversion and image preparationprotocol (see materials and methods) (12), trans-forming the dark areas into particle-like entitiesthat are trackable using single-particle tracking(SPT) algorithms (13) (Fig. 2C andmovie S2). Thespaces were tracked and their lifetimes were ex-tracted from the trajectories. For distances be-tween tubules (spaces) larger than our ~100-nmlimit of resolution, we quantified the average life-span (250 ± 250 ms, n = 4292 tracks from 4 cells)

and detectable separation between tubules (260 ±350 nm, n = 1273 spaces from 4 cells) (Fig. 2D andtables S3 and S4). There was no significant differ-ence in the lifespan or apparent size of voids inintensity in either U-2 OS or COS-7 cells express-ing Sec61b (table S4). Measurable gaps betweentubules did, however, appear significantly smallerin cells expressing Sec61b than in cells express-ing ER3, presumably because of the relocation ofthe fluorescence tag from the lumen to the cyto-plasmic surface of the ERmembrane (Fig. 2Dandtables S3 and S4; see also supplementary text). Thisapparent dilatation of the structure would be pre-dicted to lead to an appearance of constriction inremaining nonfluorescent spaces. Certainly, thesize and lifetime of the spaces observed with GI-SIMwithin theseperipheral structureswould rendera substantial proportion of them undetectable bymore traditional imaging modalities.To accompany the SPT of these transient spaces,

we used an analysis that does not require a best-fit process, termed the temporal intensity deriv-ative (detailed inmaterials andmethods; see alsofig. S1). This techniquemaps locations in the struc-ture where substantial changes in fluorescence in-tensity occur over defined time windows.We foundthat even across a very short time frame (250ms),we were able to detect substantial motion in ERtubules, consistent with the tubular oscillationsdescribed in Fig. 1. Intriguingly, nearly the entirearea of structures that appeared as peripheral

aaf3928-2 28 OCTOBER 2016 • VOL 354 ISSUE 6311 sciencemag.org SCIENCE

Fig. 1. The peripheral ER moves at high speeds. (A) Tubular ER in the periph-ery of a COS-7 cell expressing Sec61b imaged live at 40 Hz using GI-SIM micros-copy. Scale bar, 2 mm. (B) ER tubules within the boxed region in (A), identifiedusing a skeletonization algorithm. Left: The midline of each tubule (green) ismapped onto the fluorescence (magenta). Right: Positions of the midlines areplotted as kymographs against time for each of the three locations shown in cyanat left; scale bars, 200 nm and 0.5 s. (C and D) Amplitudes (C) and frequencies(D) of tubular ER oscillations in COS-7 cells expressing Sec61b treated with de-oxyglucose plus sodium azide (DOG + NaN3), AlF, nocodazole (NZ), blebbistatin

(Bleb), puromycin (Puro), and cycloheximide (CHX). Untreated controls using aluminal ER marker (ER3) and results for a different cell line (U-2 OS–Sec61b) arealso shown. (E) Plot of frequency versus amplitude for tubular oscillations in treatedand untreated cells. Error bars represent SEM. (F) Locations of three-way junctionsderived from skeletonized data (white). Original fluorescence is shown in magenta;example tracks of junctions (cyan) over 2.5 s are indicated in green. Scale bar,2 mm. (G) MSD scaling exponent (a values) for treated and control cells. Boxplots indicate the mean and SD in (C), (D), and (G); range is indicated by outertick marks. See tables S1 and S2 for a detailed list of means and test statistics.

RESEARCH | RESEARCH ARTICLEon O

ctober 10, 2020

http://science.sciencemag.org/

Dow

nloaded from

“sheets” by conventional imaging also underwentfluctuations in fluorescence intensity similar to thatof isolated tubules, suggesting major structuralrearrangements within these structures over thistime frame (Fig. 2, E andF). This phenomenonwasconsistently observed in COS-7 cells expressingeither Sec61b or ER3, as well as in U-2 OS cells ex-pressing Sec61b. Collectively, the rapid rearrange-ment of spaces and themagnitude of fluorescencechanges across very short time scales within pe-ripheral sheet-like structures imply that theseregions are not likely to be continuous in nature.

Rapid assembly and disassembly ofsheet-like structures into isolated tubules

In further support of this idea, longer time-lapseimaging with GI-SIM revealed rapid assembly anddisassembly of sheet-like structures from clearlyisolated tubules (movie S3). Given that this proc-ess occurs over relatively short time frames, itseems unlikely that energetically costly fusion orfission of the ER membrane would be required.Instead, our data suggest a possible mechanismwhereby tubules could coalesce until the spacesbetween them become too small to observe, lead-ing to the discontinuities described above. Conse-quently, dense networks could expand outward toisolated tubules by the reverse mechanism, with-out requiring membrane fusion or fission. Thiscould be achieved by well-characterized motion

through molecular motors or by the sliding ofthree-way junctions along tubules (3).

SR imaging reveals the existence of densetubular matrices in the peripheral ER

To gain a more comprehensive understandingof structures classically defined as peripheralsheets, we performed parallel experiments usingSR imaging to reconstruct the protein and lipiddistribution across the entire ER. COS-7 cells ex-pressing Sec61b as an ER marker (thought to beexpressed uniformly across the ER) were fixedand imaged by wide-field 3D-SIM, providing arepresentative map of ER transmembrane pro-tein distribution throughout the cell (Fig. 3A).Remarkably, many structures that appeared tobe intact sheets by diffraction-limited, wide-fieldimaging instead comprised a dense, cross-linkednetwork of tubules enriched in three-way junc-tions (Fig. 3A). Notably, these structures were flatrelative to their height but showed substantial var-iation in local topology (Fig. 3A, color coding).To verify that the ER membrane itself shared

this structure, we turned to an even higher-resolution imaging technique that directly probesthe locations of the membranes themselves. LLS-PAINT microscopy uses single-molecule localiza-tion of individual fluorescent lipid moleculesas they stabilize on cellular membranes (14).COS-7 cells expressing Sec61b were fixed and the

structure of the internal membranes was ascer-tained at the single-molecule level by LLS-PAINTmicroscopy (movie S4). In addition, for each cell,a single diffraction-limited 3D LLS image (15) wastaken of the Sec61b signal to allowERmembranesto be distinguished from those of other organelles(Fig. 3B and movie S4). The resulting data setconfirms that many ER structures that appear ascontinuous sheets with diffraction-limited imag-ing are shown to be dense tubular matrices whenviewed using LLS-PAINT (Fig. 3B, insets). ManyER matrices had substantial topological variationacross their structures, again supporting thenotion that they are not strictly 2D (Fig. 3B, iii,and fig. S2). Thus, with improved spatial resolu-tion in three dimensions, both protein and lipidcomponents of the peripheral ER appear to com-prise densematrices of highly convoluted tubules.

Limitations in spatiotemporal resolutionobscure dense tubular matrices

Given our observations that ER tubules undergovery rapid motion (Fig. 1), that many spaces instructures previously described as sheets are nearor beneath the diffraction limit (Fig. 2), and that SRimaging of both ER membrane protein and lipidreveals most of these structures to be dense tubularnetworks (Fig. 3), we hypothesized that limita-tions in spatiotemporal resolution might obscuredense tubularmatrices and lead to their frequent

SCIENCE sciencemag.org 28 OCTOBER 2016 • VOL 354 ISSUE 6311 aaf3928-3

Fig. 2. Peripheral ER “sheets” are highly dynamic and riddled withspaces. (A) COS-7 cell expressing Sec61b imaged live by GI at 40 Hz exhibitsmany peripheral sheet-like structures. (B) GI-SIM of the boxed region in (A)shows many discrete spaces throughout the structure. Colored lines at leftcorrespond to the locations of the kymographs shown at right. Voids inintensity within the structure can be seen appearing and disappearing overtime. (C) Single-particle tracking (SPT) of dark spaces within the structure.The fluorescence image (i) was inverted and spaces were tracked using SPTalgorithms. Tracks overlaid onto the inverted image are shown in (ii), withtrajectories shown in different colors. (D) Each track length corresponds to

the lifetime of the space; distance across the space (i.e., distance betweentubules) is also quantified.The box plot indicates the SD and mean; range isindicated by outer tick marks.The asterisks denote significant difference be-tween means, detailed in table S4. (E and F) Temporal intensity derivativeanalysis (see materials and methods) of representative peripheral sheet-likestructures in a COS-7 cell expressing Sec61b, with a luminal ERmarker (ER3)and another cell line (U-2 OS–Sec61b) as controls. (E) Original fluorescenceimages. (F) Each consecutive frame over a 250-ms time period is color-coded,with intensitycorresponding to themagnitudeof fluorescencechange.Scale bars,2 mm. See tables S3 and S4 for a detailed list of means and test statistics.

RESEARCH | RESEARCH ARTICLEon O

ctober 10, 2020

http://science.sciencemag.org/

Dow

nloaded from

misinterpretation as sheets. To test this directly,we compared images of peripheral ER matricescollected by either diffraction-limited GI or GI-SIM using two different simulated exposure times(see materials and methods for details) (Fig. 4A).We found that the loss of either spatial or temporalresolution was sufficient to obscure the majority ofgaps between tubules within the matrix.

In the case of live-cell imaging, tubular motioncan occur faster than the acquisition time of asingle frame, creating a blurring artifact and thusincreasing the apparent diameter of tubules asthe temporal resolution decreases (Fig. 4B andtables S5 and S6). Likewise, the resolvable sepa-ration between tubules in amatrix is affected notonly by the true distance across the space be-

tween tubules, but also by temporal blurring dueto oscillations of the surrounding tubules andmotion of their three-way junctions. We there-fore expected gaps in tight tubular matrices toappear smaller in active, living cells than in fixedones. Indeed, the average apparent distance be-tween tubules in fixed 3D-SIM (220 nm) wasgreater than that observed using live GI-SIM

aaf3928-4 28 OCTOBER 2016 • VOL 354 ISSUE 6311 sciencemag.org SCIENCE

Fig. 3. Many peripheral structures classically identified as sheets areinstead dense matrices of tubules. (A) Left: Fixed COS-7 cell expressingSec61b, imaged by 3D-SIM and color-coded by z position. Scale bar, 10 mm.Right: Magnified regions (i to iii) show that 3D-SIM reveals dense tubular mat-rices, which appear as sheets by diffraction-limited (DL) epifluorescence. Scalebars, 2 mm. (B) Left: Deconvolved, diffraction-limited LLS imaging of a fixedCOS-7 cell overexpressing Sec61b (gray). All internal lipid membranes were

reconstructed using LLS-PAINTmicroscopy. Data from three regions containingER matrices are shown in colored insets. Scale bar, 10 mm. Right: The threeboxed regions are enlarged, showing (i) 3D orientation of LLS-PAINT volumerendering, (ii) overlay of LLS-PAINTand diffraction-limited LLS imaging volumerendering (gray), and (iii) LLS-PAINT color-coded by z position. White arrow-heads mark areas that appear as sheets by diffraction-limited imaging; the redarrowhead [top of (ii)] denotes a mitochondrion. Scale bars, 2 mm.

RESEARCH | RESEARCH ARTICLEon O

ctober 10, 2020

http://science.sciencemag.org/

Dow

nloaded from

(150 nm). Consequently, we predicted that de-creasing temporal resolution would preferentiallyblur smaller spaces, as tubular oscillations andjunctionmovement wouldmask themmore read-ily. To quantify this, we summed GI-SIM imagesof tubularmatrices over an increasing number of25-ms frames to simulate longer exposure times.As predicted, longer simulated exposure timesresulted in blurring of matrices until they ulti-mately resembled continuous sheets (Fig. 4D). Onthe other hand, improving temporal resolutiondecreased the minimum distance required be-tween tubules before a space became detectable(Fig. 4E and tables S5 and S6), thereby increasingthe density of measurable spaces within a dy-namic matrix (Fig. 4F and tables S5 and S6).To quantify the contribution of insufficient

spatial resolution in obscuring the structure ofER tubular matrices, we compared cells in theabsence of motion (i.e., in fixed cells) using threedifferent imaging modalities of varying spatialresolution (diffraction-limited GI, 3D-SIM, or LLS-PAINT) (Fig. 4G). Predictably, improvements inspatial resolution decreased the measurable di-ameter of ER tubules (Fig. 4C and tables S5 and

S6). Moreover, imaging with diminishing spatialresolution limited the detectable degree of sepa-ration of tubules in amatrix (Fig. 4H and tables S5and S6), decreasing their detectable density (Fig. 4Iand tables S5 and S6). Thus, sufficient spatial andtemporal resolution are both required to resolvethe fine structure of tubularmatrices in living cells,and with any imagingmodality an apparently con-tinuous structure may conceal spaces if they arebeneath the resolvable power of the technique.

FIB-SEM reveals tubular matrices withheterogeneous topology

To overcome the limitations of opticalmicroscopyin resolving the very fine structure of tubularmatrices, we studied their morphology by elec-tron microscopy (EM). FIB-SEM was performedin native, untransfected cells to provide an addi-tional control against the possibility that over-expression of both Sec61b and ER3 coincidentallyinduces matrices. The conditions used for FIB-SEM were selected to result in 8-nm steps in thez position (see materials and methods), whichprovides extremely high resolution in the z di-mension and ensures that even thin structures

such as tubularmatrices can be captured (Fig. 5A).This fine z resolution allowed the reconstructionof remarkably intricate 3D tubular structuresevenwithin very thin sections of the cell (Fig. 5BandMovie 1). In contrast, we found that the topo-logical complexity of matrices in the z dimensionwould be lost with diffraction-limited confocalimaging, as nearly the entire structure shownin Fig. 5B falls within the focal plane of a singleconfocal slice. Indeed, a projection of a theoreticalconfocal image derived from the EM data in Fig.5B (see materials and methods) results in an im-age indistinguishable from that of an intact ERsheet (Fig. 5B, green footprint). In agreement withthe LLS-PAINT data described above, we foundthat these structures can contain substantial ver-tical topology evenwithin a thin space, and a singleslice through the structure often revealed only afew isolated tubules (Fig. 5A). Inspection of theFIB-SEM data also revealed heterogeneity inmatrix structures, from highly convoluted 3Dstructures to nearly planar arrays of tubes (e.g., Fig.5B versus Fig. 5C). Additionally, some matriceswere incredibly tightly clustered, with spaces lessthan 50 nm between tubules (Fig. 5C). As such, it

SCIENCE sciencemag.org 28 OCTOBER 2016 • VOL 354 ISSUE 6311 aaf3928-5

Fig. 4. Effects of spatiotemporal blurring on imaging ER tubular matrices.(A) A COS-7 cell expressing Sec61b, imaged live using GI-SIM, showing bothsingle 25-ms frames (i, iii) and 40 frames averaged over 1 s (ii, iv). The top rowshows the GI-SIM images; the bottom row shows the corresponding diffraction-limited images with GI illumination, demonstrating the combined effects ofspatial and temporal limitations in resolution. (B and C) The measurable diam-eter of isolated ER tubules found outside of matrices also decreases with in-creasing temporal (B) and spatial (C) resolution. (D) Structure of a representativetubular ER matrix imaged in a live cell by GI-SIM when integrating imageframes for (i) 25 ms (1 frame), (ii) 250 ms (10 frames), or (iii) 1 s (40 frames)

as in (A). (E and F) Quantification of the measurable distance between tu-bules (E) and density of these spaces (F) within tubular matrices for eachfunctional exposure time. (G) Representative images of tubular matricesimaged in fixed cells by diffraction-limited GI (i), 3D-SIM (ii), and LLS-PAINT(iii). (H and I) Quantification of the measurable distance between tubules (H)and density of these spaces (I) in tubular matrices, as identified by imagingmodalities of increasing spatial resolution. Scale bars, 2 mm. Box plots indi-cate the mean and SD in all panels; range is indicated by outer tick marks.The asterisks denote statistical significance between the means, as detailedin tables S5 and S6.

RESEARCH | RESEARCH ARTICLEon O

ctober 10, 2020

http://science.sciencemag.org/

Dow

nloaded from

aaf3928-6 28 OCTOBER 2016 • VOL 354 ISSUE 6311 sciencemag.org SCIENCE

Fig. 5. ER tubules form highly convoluted and intricate structures withinthe resolvable volume limits of any optical imaging technique. (A) Threeconsecutive FIB-SEM slices through an ER matrix spaced by 8 nm. Scale bar,1 mm. (B) Three-dimensional reconstruction of a tubularmatrix in a thin section(~600 to 1200 nm between the plasma membranes) of the cell. (i) The foot-print shows the theoretically highest resolution that could be achieved with asingle confocal slice through the structure directly shown above. Scale bar,1 mm in each direction. (ii) Close-up of the reconstruction of the boxed region in

(i). (C) 3D rendering of raw EM data showing an example of an approximatelyplanar ER matrix with subresolution spaces (large scale bar, 500 nm; smallscale bar, 50 nm). (D) 3D rendering of the ER at the border of perinuclear andperipheral regions of the cell, showing stacked helicoidal sheets (cyan box) andERmatrices (yellow box). (E) A theoretical confocal image of the structure, show-ing the difficulty in distinguishing these structures by diffraction-limited imaging.(F) Views from the side of a stacked helicoidal sheet in the perinuclear region ofthe ER [from cyan box in (D)], showing the pitch of the intact membranes.

RESEARCH | RESEARCH ARTICLEon O

ctober 10, 2020

http://science.sciencemag.org/

Dow

nloaded from

would be impossible to resolve these structuresas distinct from sheets by SIM, and it would bedifficult to do so even by LLS-PAINT.Generally, areas of flat continuous membrane

are described as having net zero curvature, where-as highly curved regions of membrane have eithernet positive or negative curvature (1). In additionto the aforementioned tubular matrices, we alsoobserved a number of structures in the ER thatshow local regions of near-zero curvature in the ERmembrane. The largest of these regions are in truestacked ER sheets near the nucleus (Fig. 5D, cyanbox), which have been described ex-tensively (16). When imaged withdiffraction-limited techniques, thesestructures are nearly impossible todifferentiate from tubular ERmatrices(Fig. 5,DandE).Wealso observed thatwhenever stacked sheets are seen, theyappear to be connected bymembrane“ramps” [as described in (16)] andgenerally are close to thenucleus,wherethe cell height ismuch greater than inthe periphery (Fig. 5F and movie S5).

Localizing ER-shapingproteins to tubular matrices

Given the highly variable and com-plex nature of ERmatrices identifiedin the FIB-SEM data, we speculatedthat examining the distribution ofwell-characterized ER-shaping pro-teins may provide insights into thediversity of these structures. ATLGTPases have been well character-ized as three-way junction-formingproteins localized within ER tubules,where they also bind ER-shapingRTNs (17, 18). Thus, although theywould not be expected to be en-riched in sheets, they would be ex-pected in matrices composed ofdensely packed tubules and three-way junctions. We transfected COS-7 cellswithHaloTag-ATL1 (one of threehuman ATL paralogs) and assessed the localiza-tion of ATL1 in live cells by means of diffraction-limited confocal microscopy (Fig. 6A). ATL1 waspresent in all structures classically considered asperipheral sheets. This was verified by immunos-taining of endogenousATL3 inHeLa cells (Fig. 6B).Overexpression of ATLs was previously shown

to induce massive proliferation of “aberrant sheet-like structures” in aGTPase-dependentmanner (17).This finding has been difficult to explain in light ofthe known role of ATL GTPases as mediators ofthree-way junction formation between tubules.Wehypothesized that ATL overexpression might drivethe formationof increasinglydense tubularmatrices,which would appear as peripheral sheets understandard confocal imaging because of insufficientspatiotemporal resolution. COS-7 cells coexpressingSec61b andHaloTag-ATL1were fixed and imagedusing both diffraction-limited epifluorescenceand 3D-SIM (Fig. 6C). Epifluorescence revealedthe presence of peripheral sheet-like structures(Fig. 6C, i), as previously reported (17). However,

the improved spatial resolution offered by 3D-SIM showed these “aberrant sheet-like structures”to be dense tubular matrices (Fig. 6C, ii). Thesematrices contained ATL1 throughout (Fig. 6C,iii). Collectively, these data suggest that over-expression of ATLs does not drive formation ofaberrant sheets, but rather induces the formationof dense tubular matrices, consistent with theknown cellular functions of ATLs.We also looked at the distributions of other pro-

teins associated with driving or stabilizing par-ticular ER shapes. RTNs are a highly conserved

family of ER proteins sharing substantial se-quence homology; multiple isoforms are presentin most cell types and are typically associatedwith inducing and maintaining the curvature ofthe ER membrane in tubules (3). Classic work inthe field has demonstrated localization of RTNsto tubular structures and their exclusion fromsheet-like structures (19, 20). Conversely, CLIMP63is a traditional marker used to identify ER sheets,as it is believed to be involved in stabilizing thediameter of the ER lumen through interactionsbetween long, dimeric coiled-coil domains (2). U-2OS cells expressing Sec61b were stained withantibodies to endogenous CLIMP63 and RTN4A/Band imaged using Airyscan, a technique capableof achieving subdiffraction-limited imaging inde-pendent of SIM. In concordance with their tubularnature, many peripheral ERmatrices were foundto be positive for RTN4A/B, and some of thesewere also enriched in CLIMP63 (Fig. 6D). There arealso a smaller number of ER matrices that excludeRTN4A/B staining, as has been previously demon-

strated in some peripheral sheet-like structures (2)(fig. S3). This complex heterogeneity in proteinlocalization across matrices is unlikely to be astaining artifact, as overexpression of CLIMP63or a variety of RTN isoforms also resulted in thepresence of these proteins in some, but not all, ERmatrices (fig. S3 and supplementary text). It is pos-sible that heterogeneity in the distribution of theseproteins is linked to the highly variable topologiesshown in Fig. 5, which may correspond to func-tionally distinct classes of structures that appearidentical when imagedwith insufficient resolution.

To explore whether our observationshold true for multiple cell types, weexamined the structure of the ER inthree dimensions using either 3D-SIMor Airyscan in a variety of cell lineswith highly variant morphology anddiverse organisms and tissues of origin.In all 10 cell lines examined, peripheralER matrices were visible (Fig. 7). Al-though across cell types there seems tobe substantial variation in the topology,density, and cellular location of thesestructures, they are uniformly presentand clearly visible using either of thetwo independent imaging techniques.

Discussion

Taken together, our data indicate thatmost previously described “sheet-like”ER structures within the thin periph-ery of cells are actually dense tubularmatrices. Limitations in spatiotemporalresolution using conventional micro-scopy result in their appearance ascontinuous or fenestrated sheets. Ad-ditionally, we show a previously un-characterized, rapidly dynamic statein the peripheral ER that is broadlydependent on cellular energy sourcesand that contributes to the misiden-tification of ERmatrices in living cells.Why would the peripheral ER be

organized in this way? The structuralconformation of tubular matrices is likely to beimperative to multiple features of ER biology,such as the ability of the ER to rapidly alter itsconformation in response to changing cellularneeds. Interconversions among loose polygonalnetworks and dense matrices could be accom-plishedby simply sliding tubules alongoneanother,rather than requiring energetically costly fusionor fission of ERmembranes. This rapid intercon-version between loose and tight polygonal arraysof tubules (e.g., movie S3) is likely to be importantin enabling the ER to rapidly reconfigure its spatialfootprint in response to intracellular structural re-arrangements or cell shape changes, or during cellmigration. Indeed, theERand cytoskeleton coexerta driving force for cytoplasmic streaming duringcell expansion inArabidopsis, and this is altered inmutants in which ER morphology is affected (21).Clustering of tubules into tight arrays of three-

way junctions might also function to decreasecurvature stress across the ER, because the nega-tive curvature of three-way junctions could help

SCIENCE sciencemag.org 28 OCTOBER 2016 • VOL 354 ISSUE 6311 aaf3928-7

Movie 1. Three-dimensional reconstruction of FIB-SEM data reveals aconvoluted ER matrix. Raw 2D FIB-SEM data of ER tubules are shown ina series of sequential planes. The ER is segmented in green; the 3D recon-struction is shown, revealing an ER matrix. When confocal resolution issimulated, the convoluted nature of the structure is masked.

RESEARCH | RESEARCH ARTICLEon O

ctober 10, 2020

http://science.sciencemag.org/

Dow

nloaded from

aaf3928-8 28 OCTOBER 2016 • VOL 354 ISSUE 6311 sciencemag.org SCIENCE

Fig. 6. Localization of ER-shaping proteins within dense tubular matrices.(A) COS-7 cell expressing low levels of HaloTag-ATL1 imaged live with con-ventional spinning disk microscopy. Boxed regions (i) and (ii) demonstrateATL1 localization throughout structures that appear to be peripheral sheets byspatiotemporally limited imaging techniques. (B) Left: A fixed HeLa cell ac-quired with conventional scanning point confocal microscopy, stained for theendogenous ER marker calnexin and endogenous ATL3 (merged image);boxed regions (i) and (ii) showendogenousATL3 localization to structures thatappear as sheets. Right: Regions shown in (i) and (ii) were stained for the

endogenous ERmarker calnexin and endogenous ATL3. (C) A fixed COS-7 cellexpressing Sec61b and HaloTag-ATL1 imaged by wide-field SIM. Sec61bfluorescence is color-coded by z position. Boxed region is enlarged in panelsat right. Structures that appear as sheets by diffraction-limited imaging (i)are revealed to be dense tubular matrices (ii) that are positive for ATL1 (iii);the merged image of (ii) and (iii) is shown in (iv). (D) U-2 OS cell expressingSec61b stained for endogenous RTN4A/B and CLIMP63 with the boxedregion magnified (bottom row), illustrating localizations of both proteinswithin a tubular matrix. Merged images are at the far right. Scale bars, 2 mm.

RESEARCH | RESEARCH ARTICLEon O

ctober 10, 2020

http://science.sciencemag.org/

Dow

nloaded from

SCIENCE sciencemag.org 28 OCTOBER 2016 • VOL 354 ISSUE 6311 aaf3928-9

Fig. 7. Tubular ER matrices are present in different cell types. (A and B) Various cell lines expressing Sec61b were imaged using 3D-SIM (A) or, wherefluorescence intensity was insufficient, Airyscan (B).The first four cell lines were imaged using both modalities, demonstrating that the dense matrix structures arenot artifacts of any given imaging modality. Boxed regions highlighting representative tubular ER matrices in each cell type are magnified at the right. Whitearrowheads indicate subdiffraction-limited spaces between tubules. Scale bars, 2 mm.The signal of Sec61b fluorescence is color-coded by z position in the left andcenter panels corresponding to each cell line and technique.

RESEARCH | RESEARCH ARTICLEon O

ctober 10, 2020

http://science.sciencemag.org/

Dow

nloaded from

to neutralize the positive curvature associatedwith tubules (1). Thus, a relatively planar networkcould be formed (e.g., Fig. 5C) in the thin peripheryof cells, where a lack of vertical space might pre-vent the formation of stacked helicoidal sheets(16). Dense ERmatrices are also predicted to havegreater membrane surface area than a flat sheetof similar dimensions, so they may allow storageof excess membrane proteins and lipids or provideincreased surface for modulating lipid synthesisor protein folding. Such a tubular membrane res-ervoir may also be needed to facilitate the avail-ability of ERmembrane formodulating interactionswith other organelles, such as mitochondria, lipiddroplets, or endocytic compartments.Our data do not conflict with the impressive lit-

erature describing the structures of flattenedregions of the ER associated with specializedfunctions, such as the nuclear envelope (22), heli-coidal stacks in the perinuclear region (16), orflattened cisternal structures close to the plasmamembrane (23). Rather, these structures representone end of a spectrum of curvatures across the ERmembrane, with the other end of the spectrumdominated by more prevalent ER tubules andthe dense tubular matrices we have described.The heterogeneity observed in fine ultrastructure

and ER protein content also suggests that theremay be several distinct types of ERmatrices. It istherefore possible that these different structuresmay carry out distinct functions. For instance, ourFIB-SEMdata support previous 3DEMreconstruc-tions in suggesting that tubular morphology is farmoreheterogeneous than the cylindrical structuresoften depicted in models based on optical micros-copy studies (24, 25). Tubules can also take onflattened or highly irregular structures along theirlengths, as has been described in ER contact siteswith the plasma membrane (23). There is no rea-son to think that tubules within ERmatrices couldnot also undergo these sorts of deformations;hence, ERmatrices close to the plasmamembraneor other organelles (e.g., Fig. 3B) could potentiallyplay important roles in rapidly facilitating calciumsignaling or lipid transfer. These altered morphol-ogies may also explain some of the variability indistribution of ER-shaping proteins acrossmatrices,as structuresmaybecome toodenseor toodeformedto stably hold certain classes of ER proteins. Thesedatamay thus suggest anothermechanism formod-ulating peripheral ER function. It seems clear, in anyevent, that given the pathogenic role of impairedER shaping and protein distribution in disorderssuch as the hereditary spastic paraplegias (4), theER structure and dynamics described here willhave important implications for understandingboth basic cell biology and disease pathogenesis.

Materials and methods

Plasmids and antibodiesConstructs expressing Myc- and HA-tagged at-lastins have been previously described (26), withHaloTag-ATL1 purchased from Promega. Con-structs expressing mEmerald-Sec61b and mApple-Sec61b were generated by replacing the GFPcassette of pAcGFP1-Sec61b (2) with correspond-ing fluorescent proteins, using flanking AgeI and

BsrG1 restriction sites. Reticulon constructs werecloned fromwhole brain cDNA and inserted intopmCherry-C1 using the BglII and SalI restrictionsites. Lifeact-mApplehasbeenpreviouslydescribed(27), and mEmerald-ER-3 (ER3) was a gift fromMichael Davidson (Addgene plasmid # 54082).ER3 consists of an ER lumen-targeting motif fusedto mEmerald with a C-terminal KDEL tag.Commercially available mouse monoclonal

anti-Myc epitope (1:200; IgG1, clone 9E10; SantaCruz Biotechnology) and rabbit polyclonal anti-HA epitope (1:200; IgG, clone Y-11; Santa CruzBiotechnology) antibodies were used for Immu-nocytochemistry. In addition, a custom affinity-purified mouse monoclonal anti-ATL3 (No. 6115,IgG, clone 9H2B12; residues 561–578; acetyl-CATVRDAVVGRPSMDKKAQ-OH) antibody wasused at 1:100. The anti-RTN4A/B antibody was akind gift from Riqiang Yan, and was used asdescribed previously (28). CLIMP63 was stainedusing a commercial antibody purchased fromEnzo Lifesciences (1:250; IgG2a, clone G1/296).

Cell culture, transfection, and plating

COS-7, U-2 OS andHeLa cells (ATCC) were grownin phenol red-free Dulbecco's Modified EagleMedium (DMEM) supplemented with 10% (v/v)FBS (Corning), 2 mM L-glutamine, 100 U/ml pen-icillin and 100 mg/ml streptomycin at 37°C and5%CO2.MDCK cells were a line stably expressingER-RFP (29,30) generouslyprovidedbyErik Snapp.All other cells (SK-BR-3, BeWo, HeLa-S3, HEK293T,HT1080, NRK and NIH3T3) were grown accord-ing to manufacturer’s specifications (ATCC).Coverslips and chambers were pre-coatedwith

400-600 mg/ml Matrigel (Corning), and cells wereseeded to achieve ~60% confluency at the time ofimaging. Transfections were executed using Lip-ofectamine 3000 (Thermo Fisher Scientific) ac-cording to the manufacturer’s specifications.Fluorescently-tagged Sec61b alone was transfectedat 1 mg/ 35 mm chamber, or else cotransfected ata ~3-4:1 ratio with the additional plasmid. Imag-ing was performed between 14-22 hours post-transfection. Where indicated, HaloTag-ATL1 waslabeled with JF549 as previously described (31),and cells were imaged immediately post-labeling.

Drug treatments

All drugs used in the paper were purchased fromSigma Aldrich and used as has been described (32).Drugs were diluted from high concentration stocksreconstituted in DMSO, and diluted to the appro-priate concentration in complete medium unlessotherwise indicated. Cycloheximide (CHX) andPuromycin (Puro) were used at a final concentra-tion of 100 mg/ml, Nocodazole (NZ) was used at30 mM, and Blebbistatin (Bleb) was used at a finalconcentration of 50 mM. ATP depletion was accom-plished by incubating the cells for one hour in10 mM 2-deoxyglucose (DOG) and 2 mM sodiumazide (NaN3) in PBS at 37°C, and AlF treatmentwas performed as previously described (33) inHBSS. HBSS-only controls were also run for allexperiments using AlF, and no difference in anyphenotype studied was seen compared to un-treated controls (data not shown).

ImmunocytochemistryFor immunocytochemistry staining, cells wereseeded into No. 1.5 imaging chambers (Lab-Tek)coatedwithMatrigel, and then fixedwith 4% (w/v)paraformaldehyde (PFA) for 20 min at room tem-perature (RT). Cells were then permeabilized with0.2% (w/v) saponin (Sigma-Aldrich) for 30 min,and blocked in 5% (v/v) donkey serum (Sigma-Aldrich) dopedwith 0.05% (w/v) saponin for 1 hourat RT. Primary antibodies were diluted in block tothe aforementionedconcentrations, added to sam-ples, and incubated overnight at 4°C. Secondaryanti-mouse and anti-rabbit antibodies conjugatedto Alexa 488, Alexa 555 (1:1000; Life Technologies)weremade in block and incubatedwith samples for30min at RT. Imagingwas performed in fresh PBS.

Confocal microscopy

Live-cell confocal imaging was performed usinga customized Nikon TiE inverted scope outfittedwith a Yokogawa spinning-disk scan head (#CSU-X1, Yokogawa) and a Photometrics EM-CCD cam-era (Evolve 512) with 500 ms exposure time.Fluorescence was collected using a 60× Plan-Apochromat 1.40 NA oil objective (Nikon) withthe additional use of a 1.5× optovar to create afinal pixel size of 130 nm. Cells were imaged inDMEM and incubated with a LiveCell ImagingChamber (Nikon) at 37°C and 5% CO2.Fixed-cell confocal microscopy was performed

using a Zeiss 780 laser scanning confocal micro-scope equipped with a 32-channel multi-anodespectral detector. Excitations were performed se-quentially using 405, 488, 561, or 633 nm lines asneeded, and imaging conditionswere experimentallyselected to minimize crosstalk. The resultingfluorescence was collected using a 100× Plan-Apochromat 1.4 NA oil objective (Carl Zeiss)and images were prepared using the commer-cial Zen software package (Carl Zeiss).Airyscan imaging was performed in fixed cells

using a Zeiss 880 outfitted with an Airyscan mod-ule. Cells were seeded onto matrigel-coated cover-slips, fixed in 4% (w/v) PFA supplemented with0.2% (w/v) glutaraldehyde at RT for 20 min. Afterfixation, cells were washed in RT PBS and imagedin clean PBS. Datawas collected using a 63x 1.4NAobjective and immersion oil optimized for 30°C(Carl Zeiss). Colors were collected sequentially tominimize crosstalk, and Airyscan processing wasperformed using the Airyscan module in the com-mercial ZEN software package (Carl Zeiss).

Structured illumination microscopy (SIM)

Fixed cell, 3D-SIM was performed using a com-mercial Zeiss microscope (ELYRA SR-SIM, CarlZeiss Microimaging) outfitted with a Plan-Aprochromat 63× 1.4 NA objective lens. Sampleswere fixed in 4% PFA supplemented with 0.1%glutaraldehyde at RT for 20 min. Cells wereimaged in PBS atRTwith a final pixel size of 40nmand 110 nm z plane spacing using three rotationsof the SIM grating. SIM processing was per-formed using the SIM module in the Zen soft-ware package (Carl Zeiss Microimaging), andmulticolor images were channel aligned usinga matrix generated with Tetraspeck beads (Life

aaf3928-10 28 OCTOBER 2016 • VOL 354 ISSUE 6311 sciencemag.org SCIENCE

RESEARCH | RESEARCH ARTICLEon O

ctober 10, 2020

http://science.sciencemag.org/

Dow

nloaded from

Technologies) imaged on the same day asthe cells.

Grazing incidence-structuredillumination microscopy (GI-SIM)

GI-SIM was performed in a high speed SIM mi-croscope previously adapted for total internalreflection fluorescence (TIRF)-SIM (34). GI-SIMshared the same beam path configuration as thehigh-speed TIRF-SIM system. In TIRF illumina-tion mode, because the intensity of evanescentwave of excitation light exponentially decayed fromthe interface between cover slip and cell sample,the characteristic penetration depth of the evanes-cent wave limited the TIRF-SIM imaging depthto around 100 nm (34), in which much of the ERnetwork lies outside. In order to increase the im-aging depth, it is straightforward to tune down theincident angle that is inversely proportional to thepenetration depth of the evanescent wave (34). Werealized that when the incident angle was tuned toslightly smaller than the critical angle, where therefraction angle at the interface was near 90 de-grees, and the refracted light grazed the surface ofthe cover slip, the grazing incident light actuallyformed a thin light sheet parallel to the cover slipsurface. The thin light sheet intensity remainedconstant in both lateral and axial directions, andits thickness could be adjusted by tuning the in-cident angle of excitation light. The optimumthickness of grazing incident light sheet shouldmatch the depth-of-focus of the high NA objective(Zeiss alpha Plan-Apochromat 100x/1.57 Oil-HI),which is around 700 nm. To implement GI-SIM,we used the grating pattern generation algo-rithm previously developed for patterned activa-tion nonlinear SIM. It permitted us to finely tunethe incident angle of excitation light, i.e., theperiod of illumination pattern (34).We identifiedthe optimum incident angle by keeping the out-of-focus background of the TIRF image as littleas possible, meanwhile observing as much ERstructure as possible. After we identified the crit-ical angle for grazing incidence, the raw imageacquisition and SIM image reconstruction proce-dure is the same as TIRF-SIM (34). Time-lapseimages were also subject to a traditional bleachcorrection algorithm by histogram matching inImageJ (NIH). Cellswere plated and transfected onMatrigel-coated high-NA coverslips (Zeiss) andimaging was performed the following day.

Lattice light sheet-pointaccumulation for imaging innanoscale topography (LLS-PAINT)

LLS-PAINTwas performed as described elsewhere(14) using a custom built Lattice Light Sheet mi-croscope (15).Membrane labelingwas performedsequentially with BODIPY-TR (LifeTechnologies)followed by AZEP-Rh (31) to label intracellularmembranes and carried out over 14 days total.The final imagewas reconstructed from548,792,627individual molecular localizations with a medianprecision of 7.2 nm laterally and 41.0 nm axially.Immediately prior to PAINT imaging, a diffractionlimited dithered LLS image of mEmerald-Sec61bwas taken for comparison in the same cell.

Electron microscopy (EM)In preparation forEM, cellswere grown in 100mmculture dishes (Corning) in standard cell cul-ture conditions. Cells were fixed in 2% (w/v)glutaraldehyde in 0.08M cacodylate buffer for onehour. Cells were then post-fixed with osmium ac-cording to a modified ROTO (reduced osmiumthiocarbohydrazide-osmium) protocol (35). Brief-ly, fixation was performed in 1% (w/v) OsO4 in0.1 M cacodylate buffer for 30min on ice, followedby a wash in cacodylate buffer. The cells were thenincubated with 1% (w/v) thiocarbohydrazide inwater for 10min at room temperature, followed byimmersion in 1% (w/v) OsO4 in 0.1 M cacodylatefor 30 min at 4°C. Cells were contrasted en blocwith 1% (v/v) uranyl acetate, dehydrated inethanol,andembedded inDurcupanACM(Fluka). FIB-SEMwas performed using a Zeiss NVision40 FocusedIon Beam Scanning Electron Microscope. SEMand FIBmilling steps were optimized to produceisotropic 8 nm voxels. The SEM image stack wasacquired at 300 kHz/voxel using a 3-nA electronbeam at 1.5 kV landing energy for imaging and a30-kV gallium ion beam for FIB milling.

Structured illumination microscopyreconstruction and Fourier filtering

SIM reconstruction was performed utilizing amodified reconstruction algorithm based on thepreviously described Gustafsson algorithm (36).During reconstruction, data was filtered in Four-ier space using a variety of filters tominimize theappearance of reconstruction artifacts. This in-cluded at least a one log scan of the Wiener filterand a variety of suppression radii around thepeaks at Abbe’s limit (fig. S4), in addition to avariety of apodization functions designed toroll off the noise at the limit of resolution. Whilethe first two filters were selected individually foreach image, the apodization function was decidedcollectively for the data and applied to every imagein the paper. The apodization functionwas a singleGaussian blur using a radius that is smaller thanthe resolution limit of the technique, s = 45 nm inreal space. This allows most SIM reconstructionartifacts tobe filteredout,maximizing thepotentialsignal to noise with only a small price in functionalresolution (fig. S4).

Reconstruction of three-dimensionalEM data

Three dimensional FIB-SEM data was recon-structed and the ER segmented using a pseudo-automated approach. First, images were prepped,cropped, and invertedusing ImageJ, so thatosmiumsignal appeared as fluorescence for subsequentanalysis. Images were then loaded into Ilastik (37)for pixel classification. The pixel classification al-gorithm was used to generate a probability mapfor cellular membranes, based upon the osmiumsignal. A carving algorithmwithin Ilastik was thenutilized. The resulting segmentation was overlaidonto the raw EMdata using Amira (FEI) and qual-ity checked by eye throughout each slice of thereconstruction. Simulation of serial sectiondata was performed by simply summing therequisite number of FIB-SEM slices that would

have been present in a single slice acquired byserial section.

Data visualization

Two dimensional image preparation and analy-sis was generally performed using ImageJ (NIH),and three dimensional image preparation wasperformed using Amira (FEI).

Skeletonization

Skeletonization of images was performed usingImageJ (NIH). First, images were pre-processedusing enhanced local contrast (CLAHE) to helpflatten the intensity of the ER. The images werethen manually thresholded, made binary, andskeletonized. Using the AnalyzeSkeleton (8) plug-in in ImageJ, branches and junctions were de-termined from the skeletonized images. In short,skeletonized pixels with exactly two neighborsare considered branches and pixels with morethan two neighbors, junctions.

Analysis of tubule motion

After obtaining the skeletonized image, lines weredrawn perpendicular to the skeletonized structureover a number of tubules that were to be analyzed.In order to avoid the confounding effects of junc-tions crossing the line, lines were placed on sec-tions of tubule that were spatially separated fromany junctions. Kymographs of the skeletonizeddata were then generated along the lines over atotal time lapse of 100 frames (2.5 s) (see Fig. 1B,for example). Amplitude was extracted from thekymograph by using a customwritten peak find-ing algorithm in Labview, then measuring thedistance between the maximum and minimumof peaks of the skeleton during the kymograph’stime window. The frequency was defined as theinverse of the period, which was measured bydividing the length of the time course by the num-ber of pairedmaxima andminimawithin the data.

Junction tracking

Junctions were determined directly from theAnalyzeSkeleton plugin in ImageJ. Junctions fromthe tagged skeleton output of the AnalyzeSkeletonplugin typically occupy 1-5 pixels. The binaryjunction images were smoothed with a Gaussiankernel having a standard deviation of 1.0 pixel,resulting in single-particle-like images. The re-sulting images were then fed directly into theu-track SPT software (13).

Mean square displacement analysis

Trajectories, with a lifetime of at least 10 frames(0.25 s) were obtained. The trajectories were char-acterized through theirmean square displacement(MSD).MSD ¼ ð1=T Þ

XTt¼1ðrðtÞ− r0Þ2, where T

is the total movie time and r the displacement.The MSD can be described as MSD ~ ta wherea can be used to describe a particle’s motion asBrownian (a = 1), subdiffusive (a < 1), or super-diffusive (a > 1) (10).

Kymographs

Kymographs were prepared from time lapseimages by manually drawing lines across the

SCIENCE sciencemag.org 28 OCTOBER 2016 • VOL 354 ISSUE 6311 aaf3928-11

RESEARCH | RESEARCH ARTICLEon O

ctober 10, 2020

http://science.sciencemag.org/

Dow

nloaded from

image and using the standard reslice packagein ImageJ. The resulting figure represents theintensity by pixel along the line graphed againsttime. Axes are labeled to indicate the respectivex (mm) and t (s).

Tracking spaces in sheets

GI-SIM images of the ERwere cropped such thatthe ROI was an individual sheet. The area outsideof the sheet was then subtracted and the intensityof the images inverted. The transformation resultedin local intensity minima (spaces) becoming localintensity maxima. These maxima were then di-rectly entered into the u-track SPT algorithm.

Space lifetime and density

Track length corresponds to the lifetime of thespaces. To account for clipping of lifetimes ateither the start or end of the movie, the distrib-ution of lifetimes was corrected following Loerkeand colleagues (38, 39). To measure the density ofspaces, the area of each sheet was measured bydrawing a freehand ROI around the sheet andthen measuring the area of the ROI in ImageJ.The density of spaces was then calculated as thenumber of spaces within the sheet divided bythe area of that particular sheet.

Temporal intensity derivative

The derivative was calculated by choosing a definedregion and time series of interest and processingthe data as described for GI-SIM (fig. S1A). Eachframe of the time-lapse image was then subtractedpixel by pixel from the following frame using afloating 32-bit depth pixel to ensure negative sig-nalswere not lost (fig. S1B). The resulting imagewassquared on a per-pixel basis, to make all changespositive integers (fig. S1C). The upper limit of thedynamic range was reset to the theoretical maxi-mum in order to normalize the derivative betweensamples. The resulting time-lapse image was tem-porally color coded, yielding a spatial map of thechange in fluorescence intensity over time (fig. S1D).

Measuring the diameter of tubules andspaces in matrices

The size of apparent spaces within tubular ma-triceswasmeasured by fitting the intensity cross-section profile of eachminima to aGaussian curve.The full width at half-maximum of the Gaussiancurves provides a good estimate of the distanceacross the space. The diameter of tubules wasmea-sured in a similar fashion: intensity cross-sectionsalong several locations of tubules were fit to aGaussiancurveand the fullwidthathalf-maximumwas reported as the diameter of the tubule.

Temporal blurring

To simulate the effects of longer exposure timesin GI-SIM, the appropriate number of SIMframes collected at 40 Hz were merged using asimple sum projection in ImageJ. Thus, 250 msimages are the sum of 10 individual 25 msframes, and 1 s images are the sum of 40 separate25 ms frames. When color-coding by frame isshown, the temporal color code projection toolwas used in place of the simple sum projection.

REFERENCES AND NOTES

1. T. Shemesh et al., A model for the generation and interconversionof ER morphologies. Proc. Natl. Acad. Sci. U.S.A. 111, E5243–E5251(2014). doi: 10.1073/pnas.1419997111; pmid: 25404289

2. Y. Shibata et al., Mechanisms determining the morphology ofthe peripheral ER. Cell 143, 774–788 (2010). doi: 10.1016/j.cell.2010.11.007; pmid: 21111237

3. U. Goyal, C. Blackstone, Untangling the web: Mechanismsunderlying ER network formation. Biochim. Biophys. Acta 1833,2492–2498 (2013). doi: 10.1016/j.bbamcr.2013.04.009;pmid: 23602970

4. C. Blackstone, Cellular pathways of hereditary spasticparaplegia. Annu. Rev. Neurosci. 35, 25–47 (2012).doi: 10.1146/annurev-neuro-062111-150400; pmid: 22540978

5. C. Lee, L. B. Chen, Dynamic behavior of endoplasmic reticulumin living cells. Cell 54, 37–46 (1988). doi: 10.1016/0092-8674(88)90177-8; pmid: 3383243

6. A. R. English, G. K. Voeltz, Rab10 GTPase regulates ERdynamics and morphology. Nat. Cell Biol. 15, 169–178 (2013).doi: 10.1038/ncb2647; pmid: 23263280

7. D. Li et al., Extended-resolution structured illumination imagingof endocytic and cytoskeletal dynamics. Science 349, aab3500(2015). doi: 10.1126/science.aab3500; pmid: 26315442

8. I. Arganda-Carreras, R. Fernández-González, A. Muñoz-Barrutia,C. Ortiz-De-Solorzano, 3D reconstruction of histological sections:Application to mammary gland tissue. Microsc. Res. Tech. 73,1019–1029 (2010). doi: 10.1002/jemt.20829; pmid: 20232465

9. S. Chen et al., Lunapark stabilizes nascent three-way junctions inthe endoplasmic reticulum. Proc. Natl. Acad. Sci. U.S.A. 112,418–423 (2015). doi: 10.1073/pnas.1423026112; pmid: 25548161

10. R. Metzler, J. Klafter, The random walk's guide to anomalousdiffusion: A fractional dynamics approach. Phys. Rep. 339,1–77 (2000). doi: 10.1016/S0370-1573(00)00070-3

11. M. Guo et al., Probing the stochastic, motor-driven properties ofthe cytoplasm using force spectrum microscopy. Cell 158,822–832 (2014). doi: 10.1016/j.cell.2014.06.051; pmid: 25126787

12. C. Kural et al., Tracking melanosomes inside a cell to studymolecular motors and their interaction. Proc. Natl. Acad. Sci. U.S.A.104, 5378–5382 (2007). doi: 10.1073/pnas.0700145104;pmid: 17369356

13. K. Jaqaman et al., Robust single-particle tracking in live-celltime-lapse sequences. Nat. Methods 5, 695–702 (2008).doi: 10.1038/nmeth.1237; pmid: 18641657

14. W. R. Legant et al., High-density three-dimensional localizationmicroscopy across large volumes. Nat. Methods 13, 359–365(2016). doi: 10.1038/nmeth.3797; pmid: 26950745

15. B.-C. Chen et al., Lattice light-sheet microscopy: Imaging moleculesto embryos at high spatiotemporal resolution. Science 346,1257998 (2014). doi: 10.1126/science.1257998; pmid: 25342811

16. M. Terasaki et al., Stacked endoplasmic reticulum sheets areconnected by helicoidal membrane motifs. Cell 154, 285–296(2013). doi: 10.1016/j.cell.2013.06.031; pmid: 23870120

17. J. Hu et al., A class of dynamin-like GTPases involved in thegeneration of the tubular ER network. Cell 138, 549–561(2009). doi: 10.1016/j.cell.2009.05.025; pmid: 19665976

18. G. Orso et al., Homotypic fusion of ER membranes requires thedynamin-like GTPase atlastin. Nature 460, 978–983 (2009).doi: 10.1038/nature08280; pmid: 19633650

19. G. K. Voeltz, W. A. Prinz, Y. Shibata, J. M. Rist, T. A. Rapoport,A class of membrane proteins shaping the tubular endoplasmicreticulum. Cell 124, 573–586 (2006). doi: 10.1016/j.cell.2005.11.047; pmid: 16469703

20. Y. Shibata et al., The reticulon and DP1/Yop1p proteins formimmobile oligomers in the tubular endoplasmic reticulum.J. Biol. Chem. 283, 18892–18904 (2008). doi: 10.1074/jbc.M800986200; pmid: 18442980

21. G. Stefano, L. Renna, F. Brandizzi, The endoplasmic reticulumexerts control over organelle streaming during cell expansion.J. Cell Sci. 127, 947–953 (2014). doi: 10.1242/jcs.139907;pmid: 24424025

22. M. W. Hetzer, The nuclear envelope. Cold Spring Harb.Perspect. Biol. 2, a000539 (2010). doi: 10.1101/cshperspect.a000539; pmid: 20300205

23. R. Fernández-Busnadiego, Y. Saheki, P. De Camilli, Three-dimensional architecture of extended synaptotagmin-mediatedendoplasmic reticulum-plasma membrane contact sites. Proc.Natl. Acad. Sci. U.S.A. 112, E2004–E2013 (2015). doi: 10.1073/pnas.1503191112; pmid: 25787254

24. M. Puhka, M. Joensuu, H. Vihinen, I. Belevich, E. Jokitalo,Progressive sheet-to-tubule transformation is a generalmechanism for endoplasmic reticulum partitioning in dividingmammalian cells. Mol. Biol. Cell 23, 2424–2432 (2012).doi: 10.1091/mbc.E10-12-0950; pmid: 22573885

25. M. Puhka, H. Vihinen, M. Joensuu, E. Jokitalo, Endoplasmicreticulum remains continuous and undergoes sheet-to-tubuletransformation during cell division in mammalian cells. J. CellBiol. 179, 895–909 (2007). doi: 10.1083/jcb.200705112;pmid: 18056408

26. P.-P. Zhu, C. Soderblom, J.-H. Tao-Cheng, J. Stadler,C. Blackstone, SPG3A protein atlastin-1 is enriched in growthcones and promotes axon elongation during neuronaldevelopment. Hum. Mol. Genet. 15, 1343–1353 (2006).doi: 10.1093/hmg/ddl054; pmid: 16537571

27. J. Riedl et al., Lifeact: A versatile marker to visualize F-actin.Nat. Methods 5, 605–607 (2008). doi: 10.1038/nmeth.1220;pmid: 18536722

28. W. He et al., Reticulon family members modulate BACE1activity and amyloid-b peptide generation. Nat. Med. 10,959–965 (2004). doi: 10.1038/nm1088; pmid: 15286784

29. C. W. Lai, D. E. Aronson, E. L. Snapp, BiP availability distinguishesstates of homeostasis and stress in the endoplasmic reticulum ofliving cells. Mol. Biol. Cell 21, 1909–1921 (2010). doi: 10.1091/mbc.E09-12-1066; pmid: 20410136

30. E. L. Snapp, A. Sharma, J. Lippincott-Schwartz, R. S. Hegde, Monitoringchaperone engagement of substrates in the endoplasmicreticulum of live cells. Proc. Natl. Acad. Sci. U.S.A. 103,6536–6541 (2006). doi: 10.1073/pnas.0510657103; pmid: 16617114

31. J. B. Grimm et al., A general method to improve fluorophoresfor live-cell and single-molecule microscopy. Nat. Methods 12,244–250 (2015). doi: 10.1038/nmeth.3256; pmid: 25599551

32. N. B. Cole, Compendium of drugs commonly used in cell biologyresearch. Curr. Protoc. Cell Biol. 00:1B:1.B.1–1.B.26 (1998).

33. J. G. Donaldson, R. A. Kahn, J. Lippincott-Schwartz, R. D. Klausner, Binding of ARF and beta-COP to Golgi membranes: Possibleregulation by a trimeric G protein. Science 254, 1197–1199 (1991).doi: 10.1126/science.1957170; pmid: 1957170

34. J. A. Steyer, W. Almers, A real-time view of life within 100 nmof the plasma membrane. Nat. Rev. Mol. Cell Biol. 2, 268–275(2001). doi: 10.1038/35067069; pmid: 11283724

35. T. J. Deerinck, E. A. Bushong, A. Thor, M. H. Ellisman, NCMIRmethods for 3D EM: A new protocol for preparation ofbiological specimens for serial block face scanning electronmicroscopy (National Center for Microscopy and ImagingResearch, 2010); www.ncmir.ucsd.edu/sbem-protocol.

36. P. Kner, B. B. Chhun, E. R. Griffis, L. Winoto,M. G. L. Gustafsson, Super-resolution video microscopy of livecells by structured illumination. Nat. Methods 6, 339–342(2009). doi: 10.1038/nmeth.1324; pmid: 19404253

37. C. Sommer, C. Straehle, U. Kothe, F. A. Hamprecht, Ilastik:Interactive learning and segmentation toolkit. In 2011 8th IEEEInternational Symposium on Biomedical Imaging (ISBI 2011),pp. 230–233 (2011). doi: 10.1109/ISBI.2011.5872394

38. D. Loerke et al., Cargo and dynamin regulate clathrin-coatedpit maturation. PLOS Biol. 7, e1000057 (2009). doi: 10.1371/journal.pbio.1000057; pmid: 19296720

39. A. V. Weigel, M. M. Tamkun, D. Krapf, Quantifying the dynamicinteractions between a clathrin-coated pit and cargomolecules. Proc. Natl. Acad. Sci. U.S.A. 110, E4591–E4600(2013). doi: 10.1073/pnas.1315202110; pmid: 24218552

ACKNOWLEDGMENTS

Supported by a NIGMS Postdoctoral Research Associate Programfellowship (A.V.W.), an intramural AIDS research fellowship from theNIH Office of AIDS Research (C.J.O.), a grant from the NIH IntramuralAIDS Targeted Antiviral Program (J.L.-S. and C.J.O.), National KeyResearch Program of China grant 2016YFA0500200 (D.L.), andHHMI and the Intramural Research Programs of NINDS and NICHD.J.N.-A. is supported by a UCL School of Pharmacy PhD studentship(to K.H.). We thank H. White for assistance with cell culture, L. Shaofor assistance with imaging and data processing, P.-P. Zhu for advice onimmunocytochemistry, R. Yan for a generous gift of an RTN4A/Bantibody, and E. Snapp for provision of an MDCK ER-RFP stable cell lineand critical reading of the manuscript. The authors declare that thereare no conflicts of interest. All data to support the conclusions areprovided either in the manuscript or in the supplementary materials.

SUPPLEMENTARY MATERIALS

www.sciencemag.org/content/354/6311/aaf3928/suppl/DC1Supplementary TextFigs. S1 to S8Tables S1 to S6Movies S1 to S5References (40–48)

3 February 2016; accepted 16 September 201610.1126/science.aaf3928

aaf3928-12 28 OCTOBER 2016 • VOL 354 ISSUE 6311 sciencemag.org SCIENCE

RESEARCH | RESEARCH ARTICLEon O

ctober 10, 2020

http://science.sciencemag.org/

Dow

nloaded from

peripheral ERIncreased spatiotemporal resolution reveals highly dynamic dense tubular matrices in the

Kirsten Harvey, Harald F. Hess, Eric Betzig, Craig Blackstone and Jennifer Lippincott-SchwartzJonathon Nixon-Abell, Christopher J. Obara, Aubrey V. Weigel, Dong Li, Wesley R. Legant, C. Shan Xu, H. Amalia Pasolli,

DOI: 10.1126/science.aaf3928 (6311), aaf3928.354Science 

, this issue p. 433; see also p. 415Scienceneeds.clustering of tubules. This dynamic meshwork may allow the ER to change its conformation rapidly in response to cellular comprise tubules and sheets; however, the higher-resolution view revealed that most of the ''sheets'' consist of a densecontacts many other cellular organelles (see the Perspective by Terasaki). This peripheral ER has been thought to

used superresolution approaches to look at the ER at the periphery of the cell, where the ERet al.structure. Nixon-Abell itsthe cell periphery. It has important roles in many cellular processes, and numerous proteins are involved in maintaining

The endoplasmic reticulum (ER) is a complex membranous structure that extends from the nuclear envelope toA dynamic view of the endoplasmic reticulum

ARTICLE TOOLS http://science.sciencemag.org/content/354/6311/aaf3928

MATERIALSSUPPLEMENTARY http://science.sciencemag.org/content/suppl/2016/10/27/354.6311.aaf3928.DC1

CONTENTRELATED

http://stke.sciencemag.org/content/sigtrans/9/434/ra65.fullhttp://science.sciencemag.org/content/sci/354/6311/415.full

REFERENCES

http://science.sciencemag.org/content/354/6311/aaf3928#BIBLThis article cites 36 articles, 15 of which you can access for free

PERMISSIONS http://www.sciencemag.org/help/reprints-and-permissions

Terms of ServiceUse of this article is subject to the

is a registered trademark of AAAS.ScienceScience, 1200 New York Avenue NW, Washington, DC 20005. The title (print ISSN 0036-8075; online ISSN 1095-9203) is published by the American Association for the Advancement ofScience

Copyright © 2016, American Association for the Advancement of Science

on October 10, 2020

http://science.sciencem

ag.org/D

ownloaded from