Superresolution microscopy reveals the three-dimensional organization of meiotic chromosomeaxes in intact Caenorhabditis elegans tissueSimone Köhlera,b,1, Michal Wojcikc,d,1, Ke Xuc,d,e,2, and Abby F. Dernburga,b,d,f,2

aDepartment of Molecular and Cell Biology, University of California, Berkeley, CA 94720-3220; bHoward Hughes Medical Institute, Chevy Chase, MD 20815;cDepartment of Chemistry, University of California, Berkeley, CA 94720-3220; dCalifornia Institute for Quantitative Biosciences, Berkeley, CA 94720;eDivision of Molecular Biophysics and Integrated Bioimaging, Lawrence Berkeley National Laboratory, Berkeley, CA 94720; and fDivision of BiologicalSystems and Engineering, Lawrence Berkeley National Laboratory, Berkeley, CA 94720

Edited by Nancy E. Kleckner, Harvard University, Cambridge, MA, and approved April 27, 2017 (received for review February 10, 2017)

When cells enter meiosis, their chromosomes reorganize as lineararrays of chromatin loops anchored to a central axis. Meioticchromosome axes form a platform for the assembly of the synapto-nemal complex (SC) and play central roles in other meiotic processes,including homologous pairing, recombination, and chromosomesegregation. However, little is known about the 3D organization ofcomponents within the axes, which include cohesin complexes andadditional meiosis-specific proteins. Here, we investigate the molec-ular organization of meiotic chromosome axes in Caenorhabditiselegans through STORM (stochastic optical reconstruction microscopy)and PALM (photo-activated localization microscopy) superresolutionimaging of intact germ-line tissue. By tagging one axis protein(HIM-3) with a photoconvertible fluorescent protein, we establisheda spatial reference for other components, which were localized us-ing antibodies against epitope tags inserted by CRISPR/Cas9 genomeediting. Using 3D averaging, we determined the position of allknown components within synapsed chromosome axes to highspatial precision in three dimensions. We find that meiosis-specificHORMA domain proteins span a gap between cohesin complexesand the central region of the SC, consistent with their essential rolesin SC assembly. Our data further suggest that the two differentmeiotic cohesin complexes are distinctly arranged within the axes:Although cohesin complexes containing the kleisin REC-8 protrudeabove and below the plane defined by the SC, complexes containingCOH-3 or -4 kleisins form a central core, which may physically sep-arate sister chromatids. This organization may help to explain therole of the chromosome axes in promoting interhomolog repair ofmeiotic double-strand breaks by inhibiting intersister repair.

meiosis | chromosome axis | superresolution microscopy | C. elegans |cohesin

During meiosis, chromosomes undergo dramatic remodeling toenable homolog pairing, recombination, and segregation. A

hallmark of meiotic entry is the reorganization of meiotic chro-mosomes into linear arrays of chromatin loops anchored to acentral axis. The mechanism of this remodeling is not understood,but it involves replacement of canonical cohesin complexes withvariant complexes containing meiosis-specific subunits. In additionto cohesins, other meiosis-specific proteins are recruited to chro-mosome axes and are required for their roles in synapsis andmeiotic regulation (1, 2). Although axis components have beenidentified and their interactions analyzed in various model or-ganisms, their physical organization is poorly understood.Chromosome axes form an essential substrate for the assembly of

the synaptonemal complex (SC), which bridges the axes of pairedhomologs. In addition to their structural roles in reorganizing mei-otic chromosomes and templating SC formation, axis proteins play acentral role in meiotic chromosome dynamics: Axis assembly is re-quired for homolog recognition as well as homolog-specific synapsis(3, 4). Furthermore, chromosome axes are required for double-strand break (DSB) formation (5, 6) and are thought to regulate

the processing of DSBs as they undergo recombinational repair.Specifically, axis structure and/or activities recruited by axis proteinsinhibit the use of the sister chromatid as a template for homologousrecombination, thereby promoting interhomolog repair, which isessential for chiasma formation and proper homolog segregation (3,7–10). The axis also recruits components of the DNA damage re-sponse pathway, which likely regulate both the abundance of breaksand the choice of recombination pathways (11, 12).In Caenorhabditis elegans, four meiosis-specific HORMA domain

proteins (HTP-1, HTP-2, HTP-3, and HIM-3) localize to the axis,where they play distinct roles (3, 5, 13–15). Biochemical, structural,and genetic evidence has revealed that these HORMA domainproteins form a hierarchical complex. HTP-3 recruits HTP-1, HTP-2,and HIM-3 through interactions of their respective HORMA do-mains with cognate closure motifs in the C-terminal tail of HTP-3(16). However, how HTP-3 is recruited to the axis and how thesemeiotic HORMA domain proteins interact with cohesins is stillunclear. Axis association of some meiotic cohesins and HTP-3 arepartially interdependent, indicating that these components mightinteract (5, 14, 17). Studies in other organisms have shown that aninterdependence between HORMA domain proteins and cohesincomplexes is conserved among metazoans (18, 19).Axial elements were described in early electron micrographs

as electron-dense regions flanking the ladder-like central region

Significance

Meiosis is the essential cell division process that generates haploidgametes from diploid precursor cells. It relies on a dramatic re-organization of chromosomes around a central axis, which es-tablishes a platform for homologous pairing and its subsequentstabilization by the formation of a proteinaceous structure, thesynaptonemal complex, and by physical linkages resulting fromcrossover recombination events. Despite their central role in reg-ulating key meiotic events, little is known about the organizationof the chromosome axes. Here, we use superresolution micros-copy, combined with CRISPR/Cas9 genome editing, to build athree-dimensional model of the synapsed chromosome axis. Ourdata link axis structure to its functions in synapsis and regulationof partner choice during meiotic double-strand break repair.

Author contributions: S.K., M.W., K.X., and A.F.D. designed research; S.K. and M.W. per-formed research; S.K. and M.W. analyzed data; and S.K., M.W., K.X., and A.F.D. wrotethe paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Freely available online through the PNAS open access option.1S.K. and M.W. contributed equally to this work.2To whom correspondence may be addressed. Email: afdernburg@berkeley.edu or xuk@berkeley.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1702312114/-/DCSupplemental.

E4734–E4743 | PNAS | Published online May 30, 2017 www.pnas.org/cgi/doi/10.1073/pnas.1702312114

Dow

nloa

ded

by g

uest

on

Mar

ch 1

1, 2

020

of the SC, which is transversely striated and spans ∼100 nm (Fig.1A, Top) (20). However, recent evidence has indicated that thecentral region proteins alone can self-assemble into structuresknown as polycomplexes, which display both transversely striatedand longitudinal electron-dense components (21). This observa-tion indicates that the electron-dense elements that flank the SCmay not correspond to assemblies of axis proteins, although someaxis proteins appear to localize at or near these structures (22).Although recent work has illuminated some structural details ofaxis proteins and their interactions (16, 22–24), the overall orga-nization of the chromosome axis remains largely undetermined.Superresolution microscopy has emerged as a powerful tool to

bridge the resolving capabilities of conventional fluorescence andelectron microscopy (25, 26). Although the molecular organizationof the chromosome axes cannot be resolved by diffraction-limitedfluorescence methods, we have found that ultrastructural featuresof the axes can be probed by STORM (stochastic optical re-construction microscopy) and PALM (photo-activated localizationmicroscopy) superresolution methods (27–31). Because the archi-tecture of meiotic chromosomes is highly regular and reproducible,we could combine these localization techniques with averagingmethods (32–34) to attain a high-resolution molecular map of thechromosome axes to few-nanometer precision. Taking advantage ofthe well-characterized progression of meiosis within the germ line ofC. elegans, we determined the 3D organization of synapsed axes innuclei in intact germ-line tissue using a combination of STORM andPALM techniques. We use both specific antibodies and epitope tagsinserted by CRISPR/Cas9-based genome editing to determine po-sitions for all known proteins of the chromosome axis, in some cases

tagged at multiple sites. We map the position of these epitopes withrespect to an internal reference to reconstruct a detailed 3D modelof the synapsed chromosome axis. The structure that emerges fromour data is fully consistent with known interactions among theconstituent proteins. In particular, we find that meiosis-specificHORMA domain proteins bridge the distance between cohesincomplexes and the central region of the SC. Additionally, we detectintriguing differences between the conformation of distinct meioticcohesin complexes and provide evidence that complexes containingREC-8 protrude above and below the plane of each chromosomeaxis, suggesting how the axis may contribute to preventing re-combination between sister chromatids.

Results3D-STORM in Intact Tissue. During the pachytene stage of meioticprophase, homologous chromosome axes are held in parallel bythe SC, which assembles between them. The SC has a width ofabout 100 nm in most species, and thus, paired axes cannotusually be resolved by conventional fluorescence microscopy(Fig. 1B). Although optical techniques such as structured illu-mination microscopy can resolve paired axes (35–38), a higherresolution is required to determine the positions of individualproteins within the axes. We thus turned to superresolutionmicroscopy techniques based on the localization of single mol-ecules (27–31, 39–41), which have resolved the positions of in-dividual fluorophores at a resolution approaching 10 nm in plane(xy) and 20 nm in the axial (z) direction. This class of techniqueswas introduced as STORM (27) for dye-labeled samples andPALM (28, 29) for samples tagged by fluorescent proteins.

0.000

0.005

0.010

0.015

-100 -50 0 50 100z [nm]

dens

ityA

E individual stretches F averaged STORM G STORM/PALM

I

centralregionsi

ster

chro

ma t

ids

axis axis~ 100 nm

xz

cross-sectional view

chro

mos

ome

axis axis

frontal view

chromosom

e

central region

xy

coverslip

late pachytene

illumination

DAPIC

B

conv

entio

nal

HIM-3 D

Hcross-sectional view

xy

STO

RM

J

HIM-3(1) HIM-3(2)mEos-HIM-3 mMaple- HIM-3

HIM-3(1) HIM-3(2)

xz

0.000

0.005

0.010

-100 0 100x [nm]

dens

ity

sisterchrom

atids

z-400 400 nm

Fig. 1. PALM and STORM localization microscopy of intact C. elegans gonads yields highly reproducible distance measurements. (A) Chromosome axes (magenta)linearize homologous chromosomes (gray) and provide the platform for SC (black ladder) assembly. The cartoon depicts synapsed chromosomes in frontal xy view(Top) and in a cross-sectional xz view (Bottom). (B) Conventional immunofluorescence microscopy cannot resolve paired chromosome axes stained for HIM-3.(C) Schematic illustrating how data were collected from C. elegans gonads. Tissue was pressed against a coverslip, and the nuclei closest to the coverslip were illuminatedfor STORM/PALM imaging. (D) Paired axes are readily resolved by 3D-STORM of HIM-3. Colors in 3D-STORM images indicate localization along the optical axis, with redbeing closest to the coverslip and violet at the most distant position, as represented by the colored scale bar. (E and F) To determine the localization of HIM-3, multipleindividual stretches in frontal view were aligned to generate an averaged 3D-STORM image. Color denotes the position in z ranging from –100 to 100 nm as shown bythe colored scale bar (or color key, see below) in G, which applies to E–H. (G) Using optical astigmatism, we determined the localization of HIM-3 in three dimensions,which can be visualized in a cross-sectional view of the chromosome axes. (H) The localization of HIM-3 determined by immunofluorescence with STORM (green) isindistinguishable from its localization based on PALM images of endogenously taggedmEos2–HIM-3 (magenta). (I) The distance between HIM-3 and the center of the SCwas determined by fitting two normal Gaussian distributions to histograms of localization events in frontal view. HIM-3 localization is highly reproducible in independentSTORM experiments from two different days (green spheres and cyan squares) and PALM experiments (mEos2, dark purple diamonds; mMaple3–HIM-3, HIM-1–intHA,magenta asterisks). Bars at the bottom of the graph represent the localizations of HIM-3 in different experiments with their respective SDs determined by bootstrapping.(J) The histogram of localization events in z reveals a well-defined localization pattern of HIM-3 across the chromosome axis, which is highly reproducible. Bars at thebottom of the graph represent widths at half maximum. (Scale bars: 5 μm in B; 10 μm in C; 1 μm in D; 100 nm in F, which applies to E–H; magnification: 0.34× in E.)

Köhler et al. PNAS | Published online May 30, 2017 | E4735

CELL

BIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Mar

ch 1

1, 2

020

Although STORM/PALM is more typically applied to thin sam-ples, in this study we worked with intact C. elegans gonads ex-truded from adult animals (42). By adjusting the illuminationangle to be slightly smaller than the critical angle of the glass–water interface (43, 44), we imaged several micrometers into thewhole-mount samples. This approach enabled us to take advan-tage of the information about the meiotic stages of individual cellsbased on their position within the tissue and to minimize potentialartifacts arising from spreading or other sample preparationmethods (45). Because paired axes are held at a precise distancefrom each other by the SC, we were able to localize proteins withinthese structures at significantly higher precision by analyzing thedistributions of positions measured by STORM/PALM.We first performed 3D-STORM on the axial HORMA domain

protein HIM-3, detected with polyclonal antibodies raised againstits C terminus (46) and secondary antibodies conjugated toAlexaFluor 647. Intact gonads were extruded from adult her-maphrodites and sandwiched between a coverslip and a micro-scope slide in a small volume of imaging buffer, so that the tissuesurface was apposed to the coverslip (Methods). In most C. elegansmeiocytes, the chromosomes are confined to a shell near the nu-clear envelope, as the center of each nucleus is occupied by a largenucleolus. Thus, we were able to visualize chromosomes lying nearthe basal surface of these nuclei (adjacent to the sheath thatsurrounds the germ-line tissue) (Fig. 1C). We could readily resolvetwo parallel bands of HIM-3 staining separated by ∼100 nm (Fig.1D and Fig. S1). We found that in these images, synapsed chro-mosomes were predominantly oriented in frontal view, with theparallel axes lying at the same depth within the sample. Thispreferential orientation of paired chromosomes is likely due totheir spatial confinement between the nucleolus and nuclear en-velope. Nevertheless, some chromosome regions were orientedsuch that the plane of the SC was not parallel to the xy imagingplane. We excluded such regions from our analysis of proteinorganization within the axes and used only regions observed infrontal view (xy view, Fig. 1A, Top). Using optical astigmatism(30), individual proteins within these axial structures could belocalized to a resolution of ∼20 nm in xy and ∼50 nm in z, asestimated from the width of Gaussian profiles of single emitters.

Averaging over Many Samples Increases the Precision of SuperresolutionData. To determine the position of HIM-3 within paired, synapsedaxes, we generated averaged images of localization events frommultiple chromosome regions in frontal view (Fig. 1 E and F). Wefit the histograms of these localization events to two normalized,symmetrical Gaussian distributions (Fig. 1I). Using this approach(32–34), we determined the average position of components in thesynapsed axes with a precision of a few nanometers, which we de-termine by calculating the SD using a bootstrapping approach(Methods). Based on this analysis, the center of each band of HIM-3staining was localized to be 48.4 ± 1.6 nm from the midline of theSC, a value that showed remarkably little variance between samplesprepared independently on different days (green and cyan lines inFig. 1I). In electron micrographs, the central region of the SC inC. elegans spans about 96–97 nm (21, 47), Thus, the distance be-tween the edge of this structure and its midline is 48 nm, and HIM-3therefore localizes at or very close to this edge. This finding isconsistent with evidence that HIM-3 is essential for SC assemblybetween chromosomes (9, 15) and suggests that it may directly in-teract with central-region proteins.Using optical astigmatism (30) (Methods), we also determined

the distribution of HIM-3 along the optical axis (z)—that is,perpendicular to the plane of the SC (Fig. 1A, Bottom). In thesecross-sections (xz view), the distribution of HIM-3 in z withineach axis was normal (Fig. 1G), with a half width at half maxi-mum (HWHM) of 32 ± 3 nm (Fig. 1J), which is comparable to itsHWHM in xy of 29 ± 1 nm. These measurements indicate that

HIM-3 is confined to a narrow plane in the center of the synapsedchromosome axes.

Internal Quality Control by Sequential STORM and PALM Imaging.Because the orientation of parallel HIM-3 bands provides a ro-bust, highly reproducible reference for the orientation of the SC,we used it as an internal standard in double-labeling experimentsto define the organization of other axis proteins. MulticolorSTORM in thick samples is challenging due to the high back-ground and poor photoswitching performance of dyes outside ofa narrow spectral regime (41). We therefore engineered photo-switchable fusion proteins to localize HIM-3 using mEos2 (48)and mMaple3 (49). The fluorescent tags were inserted at the 5′end of the endogenous him-3 coding sequence using CRISPR/Cas9 genome editing methods. The resulting N-terminal fusionproteins were fully functional, based on an absence of meioticdefects in worms expressing the fusion proteins in lieu of un-tagged HIM-3 (Table S1).Using 3D-PALM, we determined the position of mEos2–HIM-3

to a resolution of 40 nm in xy, which is lower than the resolution inSTORM due to the limited photon yield from photoswitchablefluorescent proteins (41) (Fig. 1H and I). The separation of mEos2–HIM-3 and mMaple3–HIM-3 from the SC midline was 49.4 ±2.3 nm and 50.2 ± 1.8 nm (mMaple3–HIM-3; HIM-1–intHA, seeOrganization of Cohesins Within the Chromosome Axis), respec-tively (Fig. 1I), virtually identical to the distance determined by3D-STORM using immunofluorescence. Although most of our ex-periments were performed with mEos2–HIM-3, we note thatmMaple3–HIM-3 has slightly superior photochemical properties.Moreover, its faster maturation time of about 30 min allowed im-aging in earlier stages in meiotic prophase compared with mEos2,which restricted our observations to mid-/late pachytene due to itsmaturation time of several hours (49). Nevertheless, the highly re-producible localization of HIM-3 across many samples and with bothimaging methods provides evidence that the structure of the axes isnot easily distorted during dissection, fixation, or mounting of intactgonad tissue for single-molecule imaging. As the spatial resolution ofPALMwas lower than that of STORM, we used STORM to analyzethe localization of other axial components in whole gonads fromworms expressing photoconvertible mEos2–HIM-3 or mMaple3–HIM-3 and used PALM images from the same field of view forstructural alignment and quality control to assure constant localiza-tion of the reference protein HIM-3 in all experiments.

Organization of Meiosis-Specific HORMA Domain Proteins Within theChromosome Axis. HTP-3 is the largest of four meiotic HORMAdomain proteins expressed in C. elegans. It is recruited to the axisindependently of HTP-1/2 and HIM-3 (5, 14, 16), which arerecruited to the axis through binding of their HORMA domainsto “closure motifs,” short peptide sequences within the long tailof HTP-3 (Fig. 2A). This direct physical interaction indicates thatall four HORMA domain proteins should be in close proximitywithin the chromosome axis. To test this hypothesis, we usedSTORM to map the positions of HTP-3 and HTP-1. Using aGFP antibody against a fully functional, C-terminally taggedHTP-3-GFP fusion protein [htp-3(tm3655) I; ieSi6 II] (16), wedetected two strands of HTP-3 at 54.7 ± 2.1 nm off-center infrontal view, placing this epitope 6 nm farther from the center ofthe SC than HIM-3 (Fig. 2 B, Top and E). Additionally, HTP-3-GFP was more broadly distributed along the z axis, with aHWHM of 58 ± 9 nm (Fig. 2 B, Bottom and F). This broaddistribution in z may reflect flexibility of the C terminus of HTP-3,which spans almost 500 amino acids and has no predicted structureddomains. Interestingly, HIM-3, which binds to motifs in themiddle of the long tail of HTP-3 (16), is constrained to thecentral plane of the axes. Given the known interaction of HIM-3and the C-terminal tail of HTP-3, we wondered whether thedistinct localization patterns of HIM-3 and HTP-3 might be

E4736 | www.pnas.org/cgi/doi/10.1073/pnas.1702312114 Köhler et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 1

1, 2

020

caused by the insertion of epitope tags and/or use of antibodies.Alternatively, the long and presumably unstructured C-terminaltail of HTP-3 might be specifically oriented within the chromo-some axes. If so, we expected that the extreme N terminus mightbe resolved from the C terminus of HTP-3 and the position of theHIM-3-binding closure motifs in the center of its C-terminal tail.To test this hypothesis, we measured the position of the N ter-minus of HTP-3 using a worm strain expressing 3×Flag–HTP-3–GFP inserted via MosSCI (50) [htp-3(tm3655) I; ieSi62 II, TableS1]. This N-terminal epitope was located at 60.4 ± 1.8 nm fromthe SC midline in x, and in the center of chromosome axes, with aHWHM of 40 ± 2 nm in z (Fig. 2 C, E, and F). Thus, HTP-3 isoriented diagonally within the chromosome axes relative to theplane of the SC, such that its C terminus, which lies close to thecentral region, is widely distributed in z, whereas the HORMAdomain at its N terminus is pointing away from the SC and isconfined to a narrow plane near the center of the SC (Fig. 2A).The HORMA domain proteins HTP-1 and -2 are closely re-

lated paralogs arising from a recent gene duplication event. Theyshare 83% sequence identity and are both recognized by apolyclonal antibody raised against HTP-1 (3). HTP-2 is not es-sential for proper execution of meiosis but appears to overlap infunction with HTP-1 (9). We therefore refer to this pair ofparalogs as HTP-1/2. In vitro and in vivo experiments havedemonstrated that HTP-1/2 are recruited to the chromosomeaxes through their interaction with closure motifs within the longtail of HTP-3 and at the extreme C terminus of HIM-3 (16). Ouraligned and averaged data confirm that HTP-1/2 localize to thechromosome axes at 57.3 ± 1.6 nm off-center (Fig. 2 D, Top andE). In frontal view, HTP-1/2 are thus between the N and Ctermini of HTP-3, which is consistent with the interaction ofHTP-1/2 HORMA domain with closure motifs in the center ofthe tail of HTP-3. HTP-1/2 are narrowly distributed across the

axes in cross-sectional view, with a HWHM of 33 ± 2 nm, similarto both HIM-3 and the N terminus of HTP-3 (Fig. 2 D, Bottomand F). Thus, we observe distinct but similar localizations foreach of the HORMA domain proteins within synapsed chro-mosome axes, in both frontal and cross-sectional views, whichare fully consistent with their direct physical interactions in vitroand in vivo (Fig. 2A).

Organization of Cohesins Within the Chromosome Axis. AlthoughHORMA domain proteins play essential roles at chromosomeaxes in diverse phyla, they are not found in all clades. However,meiotic cohesins are required for axis function in all known speciesand therefore likely govern their essential structure and functions,including the recruitment of HORMA domain proteins to the axis(5, 14, 51, 52). Therefore, we wondered whether STORM mightilluminate how cohesin complexes are organized within the axes.Electron microscopy and crystallographic analysis have indicatedthat cohesins form a tripartite ring structure through the interac-tions of two large coiled-coil Structural Maintenance of Chro-mosome (SMC) proteins, known as SMC-1/HIM-1 and SMC-3 inC. elegans (53–56). Each of these proteins folds back on itself at its“hinge” domain to form an intramolecular superhelical coiled-coil,bringing together its N and C termini to form an ATPase domain,often called the “head.” The two SMC subunits form hetero-dimers through interactions at both their hinge and head domains.In addition, an essential “kleisin” subunit is thought to connect thehead of one of the SMC proteins with the coiled-coil domain ofthe other (Fig. 3A). Proteolytic cleavage of the kleisin proteinmediates release of cohesion during both mitosis and meiosis (57–59). In C. elegans, there are three meiosis-specific kleisins: REC-8,COH-3, and COH-4 (14, 17). COH-3 and COH-4 are highlysimilar and functionally interchangeable, but at least one of theseproteins is required for proper axis function, as is REC-8.

HTP-1/2D

HTP-3-GFPB

xy

xz

E F

HIM-3HTP-1/2

HTP-3

GFP

Flag

HWHM in z

position in x

cent

ral r

egio

nS

C

A

Flag-HTP-3HTP-3-GFPHTP-1 HIM-3

HTP-3-GFPHTP-1 HIM-3

Flag-HTP-3C

0.000

0.004

0.008

-100 0 100x [nm]

dens

ity

0.000

0.005

0.010

0.015

-100 0 100z [nm]

dens

ity

Flag-HTP-3

Fig. 2. Localization of meiosis-specific HORMA domain proteins in the chromosome axes. (A) The positions of individual HORMA proteins within the axis, asdetermined by 3D-STORM (colored lines in coordinate system), are consistent with the binding interactions among these proteins, as previously determined by invitro and in vivo analysis (16). (B–F) The C terminus of HTP-3–GFP (B, cyan asterisks) localizes more proximally than the N terminus (C, green spheres) and shows abroader distribution in z, based on STORM images obtained in frontal view. HTP-1 (D, magenta squares) localizes between the C terminus and the N terminus ofHTP-3, consistent with its binding to motifs in the C-terminal tail, but it shows a much narrower spatial distribution along the optical axis than a GFP tag at the Cterminus of HTP-3 (Top). (E and F) Distances were determined by fitting the distribution of localization events to double Gaussian distributions in x (E) and singleGaussians in z (F). HIM-3 (purple dashed line) is shown as a reference. The distributions are also represented as bars below the density profiles. Results for positions ± SDin x andHWHMs in z are summarized at the bottom of the graphs (see also Fig. 5A). (Scale bars: white, 100 nm; colored,−100 to 100 nm in z; information applies to B–D.)

Köhler et al. PNAS | Published online May 30, 2017 | E4737

CELL

BIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Mar

ch 1

1, 2

020

Biochemical and structural analyses of the mitotic cohesincomplex have shown that the SMC-1 and SMC-3 head domainsand the C terminus of the kleisin subunit form a tight complex(55, 60). Consistent with this, epitope tags that we introduced atthe C termini of REC-8 (REC-8–3×Flag), SMC-1 (SMC-1–3×Flag), and SMC-3 (SMC-3×HA) (Table S1) were localized inclose proximity to each other at 68.5 ± 1.5 nm, 69.6 ± 3.2 nm,and 70.1 ± 2.6 nm off-center, respectively, in frontal view. Theyalso showed similar distributions along the optical axis, withHWHM values of 59 ± 4 nm, 65 ± 4 nm, and 49 ± 7 nm, re-spectively (Fig. 3 B–D, G, and H). We therefore conclude thatthe head domains are in close proximity, as in the closed ringconformation observed in crystal structures (55, 56).We next asked whether cohesin ring complexes, with an esti-

mated diameter of 65 nm based on EM (53), have a specific ori-entation within synapsed chromosome axes. To this end, weinserted an HA epitope tag into a poorly conserved loop in thehinge domain of SMC-1. This tagged protein supported the de-velopment and fertility of C. elegans, although we did observe aslight decrease in egg viability and an increase in male self-progenyfrom homozygous mutant animals, indicative of compromised

function in mitosis and/or meiosis (Table S1). Nevertheless, thelocalization of HIM-3 was unaffected, suggesting a largely normalaxis structure (Fig. 1I, mMaple3–HIM-3). In frontal view, the hingedomain is localized 82.5 ± 3.2 nm off-center (Fig. 3 E, Top and G)and is thus separated by only 13 nm from the head domains. Theseobservations can accommodate a 65-nm distance between head andhinge domains if cohesin rings are tilted with respect to the plane ofthe SC, but our data are also consistent with a much more compactconformation of the complex, in which the head and hinge are inproximity. We note that a recent study has suggested that the DNA-bound form of mitotic cohesin complexes may differ from the largering structure observed for isolated complexes by EM (61).In cross-sectional view, the cohesin hinge domain is localized

to a narrow plane in the center of the chromosome axes with aHWHM of only 37 ± 3 nm (Fig. 3 E, Bottom and H). In a crystalstructure of the yeast cohesin complex (56), the N-terminal do-main of the kleisin subunit REC-8 interacts with the coiled-coildomain of SMC-3. Thus, we expected that the N terminus ofREC-8 should be closer to the hinge domain than other com-ponents of the cohesin head (Fig. 3A). Indeed, the HWHM oflocalization events for an antibody recognizing amino acids

0.000

0.004

0.008

-100 0 100x [nm]

dens

ity

REC-8(N)REC-8-Flag(C)SMC-1-intHA

(hinge) SMC-3-HA(C)HIM-3SMC-1-Flag(C)

G H

A

SMC-1

head

hingeSMC-3

N-REC-8-C

SMC-3-HA(C)C

SMC-1-intHA(mid)E F

REC-8(N)B

REC-8-Flag(C) SMC-1-Flag(C)D

REC-8(N)

REC-8-Flag(C)

SMC-1-intHA(hinge)SMC-3-HA(C)

HIM-3

SMC-1-Flag(C))

-0.005

0.000

0.005

0.010

0.015

-100 0 100z [nm]

dens

ity

xy

xz

Fig. 3. REC-8 cohesins have a defined orientation in the chromosome axes. (A) A schematic of the putative cohesin ring complex structure, as determined bysingle-particle EM imaging and crystallographic analysis, with the positions of the epitopes engineered in this study indicated. (B–F) STORM images of cohesinin the synapsed axes are used to measure the position of components in the axes in frontal (Top) and cross-sectional (Bottom) views. (Scale bar: white, 100 nm;colored, –100 to 100 nm in z.) (G) Histograms of localization events in x show that all epitopes associated with the head domain localize to ∼70 nm from the SCmidline, including the C-terminal REC-8–3×Flag tag (B, red open squares), C-terminal SMC-3–HA tag (C, teal closed spheres), and C-terminal SMC-1–3×Flag tag(D, light blue closed squares, dashed line), whereas the hinge domain of SMC-1–intHA (E, light blue asterisks) and the N terminus of the kleisin REC-8 (F,orange open spheres) are further from the center of the SC. For reference, the localization of HIM-3 is shown (purple dashed line). (H) Epitopes associatedwith the cohesin head domain are broadly distributed along the optical axis, whereas SMC-1–intHA in the hinge domain and the N terminus of REC-8 exhibitnarrower HWHMs in z. The results of fits are summarized at the bottom of the graphs (positions ± SD in x and HWHMs in z) and in Fig. 5A.

E4738 | www.pnas.org/cgi/doi/10.1073/pnas.1702312114 Köhler et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 1

1, 2

020

171–270 at the N terminus of REC-8 (46) in z is 37 ± 5 nm. Thiswidth is narrower than for other components in the cohesin headdomain, and in frontal view, it is located between head and hingedomain at 74.9 ± 2.4 nm from the midline of the SC (Fig. 3 F–H).These data suggest that REC-8 cohesins are oriented diagonallywithin the chromosome axis, with their hinge domains more distalrelative to the SC and the head domains extending toward theHORMA domain proteins and the central region of the SC.

Localization of COH-3/4 Cohesin Complexes. Meiotic kleisins playmultiple roles, both during and before chromosome segregation.Orthologs of REC-8 were first identified in fungi, where they werefound to hold sister chromatids together near their centromeresduring the first meiotic division. Similarly, shortly before the firstmeiotic division in C. elegans oocytes, cohesin complexes containingREC-8 become asymmetrically enriched along the “long arm” ofbivalents, where they hold sisters together during the first division(62). Conversely, complexes containing COH-3/4 become enrichedalong the “short arm” and are targeted for cleavage to enable re-ductional segregation (disjunction of homologs) during the firstdivision. Thus, specialized kleisins can mediate the unusual, step-wise release of cohesion required for proper partitioning of ho-mologous chromosomes and sister chromatids to form haploidgametes. However, during early meiotic prophase, both types ofcohesin complexes are distributed along the length of chromo-somes, as revealed by immunolocalization of REC-8 and COH-3/4(62). Deletion of rec-8 has distinct effects on chromosome interac-tions from deletion of coh-3/4, indicating that even in early pro-phase, these cohesin complexes play different roles (14, 62).We therefore investigated whether REC-8 and COH-3/4

cohesin complexes are distinctly organized within the axes. Infrontal view, the localization of a C-terminal HA tag on COH-3(Table S1) was located indistinguishably from a C-terminal3×FLAG epitope on REC-8, at 71.7 ± 3.1 nm. However, theCOH-3 C terminus showed a significantly narrower distributionalong the optical axis, with a HWHM of 46 ± 4 nm comparedwith that of REC-8–3×Flag (59 ± 4 nm) (Fig. 4 A, D, and E).It was not possible to analyze cohesin organization within

chromosome axes in the absence of COH-3/4, as the chro-mosomes do not synapse when both genes are deleted, and themethods used here require a regular, symmetrical structure.

However, chromosomes in rec-8 mutants still undergo extensivesynapsis, although it does not appear to be fully normal and mayinvolve some intersister or nonhomologous synapsis (14, 17, 62, 63)(see below in this paragraph). Thus, we analyzed the organization ofCOH-3/4 cohesin complexes in the absence of REC-8. To this end,we used CRISPR/Cas9 mutagenesis to delete rec-8 in both SMC-3–HA; mEos2–HIM-3 and SMC-1–intHA; mEos2–HIM-3 strainbackgrounds. The mutation rec-8(ie41) introduces a stop codon anda frameshift mutation after the fourth amino acid in the REC-8coding sequence and should therefore recapitulate the pheno-types observed in other rec-8 loss-of-function strains. Consistentwith prior studies of rec-8 mutants, we observed extensive synapsisin rec-8(ie41) (Fig. S2 A and B). Synapsis was predominantly be-tween paired homologs, at least for the X chromosome, as we de-tect a single, paired focus of HIM-8, which recognizes the pairingcenter of chromosome X (Fig. S2 C and D). However, chromo-somes in rec-8(ie41) mutant animals failed to form functional chi-asmata, as indicated by the presence of univalents and chromosomefragments at diakinesis (Fig. S2 E and F). Additionally, univalents atdiakinesis showed a “butterfly”-like morphology, indicative of lessextensive cohesion than in most other crossover-defective mutantsbut identical to other rec-8 mutant alleles (14, 17, 62). As suggestedby conventional fluorescence images, superresolution microscopyrevealed a fairly normal organization of chromosome axes inrec-8(ie41) mutants (Fig. 4 B and C). The localization of the headdomain of SMC-3–HA (Fig. 4B) at 71.2 ± 2.6 nm in frontal view(wild type, 70.1 ± 2.6 nm) and a HWHM of 48 ± 1 nm in zremained largely unchanged in the absence of REC-8. By contrast,the hinge domain showed a significantly wider distribution in z, witha HWHM of 45 ± 3 nm (wild type, 37 ± 3 nm) (Fig. 4C), and wasshifted closer to the head domain, at 75.7 ± 3.4 nm off-center infrontal view (wild type, 82.5 ± 3.2 nm). Moreover, the distributionof the hinge domain in frontal view showed a broader distribution inwild-type axes (Fig. 3G, HWHM in x of 38 ± 3 nm) that was re-duced in rec-8(ie41) (HWHM of 32 ± 3 nm). A possible explanationfor this difference is that the hinge domains of COH-3/4 and REC-8cohesin complexes occupy different positions and that the widedistribution seen in wild-type animals reflects a mixture of COH-3/4and REC-8 cohesin complexes. In the absence of REC-8, both thehinge and head domains were more tightly distributed in x and z.These results could indicate that COH-3 cohesin complexes form a

SMC-3-HA(C) SMC-1-intHA(mid) D EBA

COH-3-HA(C)C

rec-8 -/- rec-8 -/-wild-type

SMC-3-HA*SMC-1-intHA*

COH-3-HA

HIM-3*SMC-3-HA*SMC-1

-intHA*

COH-3-HA

*rec-8-/-

0.000

0.004

0.008

-100 0 100x [nm]

dens

ity

0.000

0.005

0.010

0.015

-100 0 100z [nm]

dens

ity

xy

xz

Fig. 4. COH-3/4 cohesin complexes are localized distinctly from REC-8 cohesin complexes within chromosome axes. (A) In frontal view (Top), the C terminus of thekleisin COH-3–HA colocalizes with the C terminus of REC-8, whereas its vertical distribution (Bottom) is significantly narrower. (B) Similarly, the HWHM of SMC-3–HA isdecreased in absence of REC-8, whereas the hinge domain of SMC-1–intHA (C) is shifted closer to the center in frontal view and slightly wider in z in rec-8(ie41). Thepositions for COH-3–HA (pink solid triangles), rec-8(ie41);SMC-3–HA (tan crossed squares), and rec-8(ie41);SMC-1–intHA (yellow crosses) in x and their HWHMs in z arequantified in D and E, respectively. Asterisks denote rec-8(ie41)mutants. For reference, the positions and HWHMs of REC-8–3×Flag (red), SMC-3–HA (teal), and SMC-1–intHA (light blue) in wild type are shown in the summaries (positions ± SD in x and HWHMs in z) at the bottom of the graphs next to their COH-3 cohesin complexcounterparts. HIM-3 in wild type is shown in purple. Notably, mEos–HIM-3 is more distal in rec-8(ie41) in frontal view (navy open triangles). (Scale bars: white, 100 nm;colored, −100 to 100 nm in z; information applies to A–C.)

Köhler et al. PNAS | Published online May 30, 2017 | E4739

CELL

BIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Mar

ch 1

1, 2

020

central core along the length of the chromosome axis, whereasREC-8 mediates a splayed orientation of cohesin complexes withineach axis, giving rise to a broader distribution along the optical axis.Previous work has shown that the HORMA domain protein

HTP-3 is required for robust loading and/or stability of REC-8cohesin complexes along the axis (14, 16). In our experiments,the position of the HORMA domain protein mEos2–HIM-3 wasshifted significantly farther from the SC in the absence of REC-8,from 49.4 ± 2.3 nm to 55.8 ± 1.2 nm off-center (Fig. 4D). Ourresults thus suggest an interplay between REC-8 cohesins andHORMA domain proteins such that each affects the others’ orga-nization and/or stability within the axis.

DiscussionUsing STORM/PALM, we have mapped the localization of allknown protein components of meiotic chromosome axes inC. elegans, in some cases with multiple epitope tags inserted at dis-tinct positions (Fig. 5A). All epitopes on cohesin complexes localizedmore distally, relative to the central region of the SC, than epitopeson meiosis-specific HORMA domain proteins (Fig. 5 and MovieS1). This key result indicates that the HORMA domain proteinsbridge the distance between cohesins and the SC central region.Overall, we find that the width of axial elements in frontal view

spans at least 34.1 nm within the plane of the SC, the distancefrom our most proximal marker, HIM-3, to the most distal epitopeon the cohesin complex. Epitopes that mark the four meiosis-specific HORMA domain proteins, which form a flexible complex(16), span at least 12 nm in x, and cohesin complexes, which span14 nm, are more distal from the central region and are clearlyseparated from the HORMA domain proteins. The localizationpatterns we observe in synapsed chromosome axes in vivo are fullyconsistent with all genetic and biochemical evidence for in-teraction of components in the axis: The N terminus of HTP-3,and thus its HORMA domain, is in proximity to cohesin proteins,suggestive of a direct interaction, which is also supported by im-munoprecipitation experiments (16) (Fig. 5 B and C and MovieS1). The C terminus of HTP-3 is located closer to the centralregion of the SC, as are HIM-3 and HTP-1/2, which are recruitedto the axis by binding to this C-terminal domain.HIM-3 and HTP-3 are strictly required for the assembly of the

central region of the SC between chromosomes (3, 5, 13–15, 64).Our localization data demonstrate that the C terminus of HTP-3

as well as the HORMA domain of HIM-3 are near this interfaceand might thus form a molecular platform for the assembly of thecentral region (Fig. 5 B and C). Although HORMA domainproteins remain strongly associated with synapsed chromosomeaxes in C. elegans, they become depleted from axes upon synapsisin budding yeast and mice (65, 66). However, these proteins donot completely disappear from synapsed axes and may thereforeprovide a persistent bridge between cohesins and the SC centralregion even when their abundance is reduced. Consistent with thisidea, cohesin complexes and proteins of the central region werefound to be separated by about 20 nm in mouse spermatocytes(24), suggesting the presence of other components between thecentral region proteins and cohesin complexes.Through analysis by 3D-STORM/PALM, we find that the axes

show evidence of bilateral symmetry along the optical axis. Ourdata suggest that REC-8–containing cohesin rings are diagonallyoriented in the axes, extending from a distal hinge domain in thecentral plane of the axis to bilaterally protruding head domainsmore proximal to the HORMA domain protein complexes (Fig.5 B and C and Movie S1). Based on a variety of in vitro evidenceand mutational analysis, cohesion has been explained by a simplemodel in which two DNA molecules are entrapped by individualcohesin rings (60, 67, 68). Other models posit a requirement forinteractions among cohesin complexes (reviewed in refs. 69–71).Supporting the latter possibility, a recent study demonstratedinterallelic complementation of mutations in cohesin subunitsfor mitotic function in budding yeast (72). Interactions betweencohesin complexes have not been reported in vitro, perhapsbecause they depend on chromatin association. Intriguingly, thebilateral localization pattern we observe for REC-8 cohesincomplexes with central hinge domains and protruding head do-mains is consistent with such a higher order organization ofcohesins in vivo (Fig. 5C and Movie S1).By contrast to the bilateral symmetry and specific orientation of

REC-8 cohesin complexes, hinge and head domains of COH-3/4cohesin complexes are indistinguishable in x and z cross-sections ofthe axis in absence of REC-8. These data suggest that COH-3/4cohesin complexes are confined to lie close to the plane of the SC,whereas REC-8 cohesin complexes may be in a splayed orientation,protruding above and below this plane (Fig. 5C and Movie S1).Chromosome axes play a central role in homologous recombination:Although DSBs are formed preferentially in chromatin loops,

protein

HIM-3 48.4±1.6 nm 32±3 nm

HTP-1 57.3±1.6 nm 33±2 nm

Flag-HTP-3 60.4±1.8 nm 40±2 nm

HTP-3-GFP 54.7±2.1 nm 58±9 nm

SMC-3-HA 70.1±2.6 nm 49±7 nm

REC-8(N) 74.9±2.4 nm 37±5 nm

REC-8-Flag 68.5±1.5 nm 59±4 nm

SMC-1-Flag 69.6±3.2 nm 65±5 nm

SMC-1-intHA* 75.7±3.4 nm 45±3 nm

SMC-1-intHA 82.5±3.2 nm 37±3 nm

SMC-3-HA* 71.2±2.6 nm 48±1 nm

COH-3-HA 71.7±3.1 nm 46±4 nm

HIM-3* 55.8±1.2 nm n.d.

HWHM in z

*rec-8-/-

A B C

head

HTP-3

HIM-3cent

ral r

egio

n

sisterchromatid

HORMA-domainproteins

HTP-1/2

SMC-1

N-REC-8-C

COH-3 cohesin

SMC-3

sisterchromatid

hinge

position in x [nm]

HWHM in z [nm]

5

5

Flag-HTP-3(N)

HIM-3

HTP-1/2

SMC-1-Flag(C)

SMC-3-HA(C)

REC-8(N)

COH-3-HA(C)

REC-8-Flag(C)

HTP-3-GFP(C)

(hinge)SMC-1-intHA

rec-8-/-SMC-1-intHA

rec-8-/-SMC-3-HA

position in x

30

40

50

60

70

50 60 70 80

position in x [nm]

HW

HM

in z

[nm

]

Fig. 5. Model of synapsed chromosome axes in cross-sectional view. The positions from the midline of the SC in x and the HWHM in z of components within thechromosome axes measured by STORM are summarized in a table (A) and graphically (B). Mean values are shown by symbols, and shaded areas are their corre-sponding SDs. These positions were used to construct a model of the synapsed chromosome axis (C); see Movie S1 for additional information about the model.Asterisks in A denote rec-8(ie41) mutants.

E4740 | www.pnas.org/cgi/doi/10.1073/pnas.1702312114 Köhler et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 1

1, 2

020

recombination intermediates are tethered to the axes (1, 2, 73).Moreover, multiple studies have demonstrated that components ofthe chromosome axes, such as meiosis-specific HORMA domainproteins, are required to establish an interhomolog bias for DSBrepair in meiosis (6–8, 64, 74–76), and the cohesin Rec8 is requiredfor maintenance of this bias (77). Our analysis of axis organizationmay illuminate the mechanism of this interhomolog bias: Cohesincomplexes and associated HORMA domain proteins in the plane ofthe SC may form a physical barrier between sister chromatids. Wespeculate that interposition of a central axis core between sisters(illustrated in Fig. 5C and Movie S1) inhibits untimely sister-chromatid recombination, whereas the establishment of a substratefor SC assembly by HORMA domain proteins may simultaneouslypromote interhomolog recombination. Notably, early electron mi-crographs have suggested that sister chromosome axes and pro-truding sister chromatid loops are stacked perpendicular to the planeof the SC in pigeon spermatocytes (78). Similarly, EM images of ratspermatocytes suggest that axes have three distinct layers, with acentral layer perhaps acting as an interface between axes of sisterchromatids (79). These earlier descriptions are consistent with ourfindings and suggest that a 3D organization of chromosome axeswith central and bilaterally protruding components is conservedacross species. This specialized 3D architecture may explain whymultiple meiosis-specific kleisins are expressed during meiotic pro-phase in diverse eukaryotes.

MethodsWorm Strains and Transgenes. All C. elegans strains were cultured using standardmethods (80). Wormsweremaintained at 20 °C. Strains used in this study are listedin Table S1. Transgenes encoding HTP-3-GFP (16) and 3×Flag–HTP-3 were insertedby MosSCI (50) and crossed into mutants lacking an intact htp-3 gene. For genomeediting, we initially performed CRISPR-Cas9 as described (81), which resulted inediting efficiencies of about 1% or less of F1 progeny positive for coinjectionmarkers. Editing efficiencies were dramatically improved, to about 50% ofmarker-positive F1s, by injecting gRNA-Cas9 ribonucleoprotein (RNP) complexes pre-assembled in vitro. Cas9 protein was purchased from the UC Berkeley Macrolab,and Alt-R tracrRNA and crRNA were purchased from IDT. Equimolar solutions oftracrRNA and crRNA (100 μM each) were heated (95 °C, 5 min) and annealed(room temperature, 5 min). Cas9-RNP was then formed by addition of purifiedCas9 protein to a final concentration of 30 μM gRNA and 28 μM Cas9 protein for5 min at room temperature. Repair templates were provided as plasmid (mEos2–HIM-3), long ssDNA [mMaple3–HIM-3 (21)], or ssODNs (synthesized as “Ultramers”by IDT) for epitope tagging and rec-8(ie41), respectively. All coding sequences forepitope tags and fluorescent proteins were codon-optimized for C. elegans, andthe protein coding sequences also contained short introns (82).

Immunofluorescence. Immunostaining of extruded gonads from adult her-maphrodites at 24 h post-L4 was performed as previously described (42) withminor modifications. Worms were immobilized using 0.02% tetramisole (insteadof azide) during dissection. Fixed tissue was transferred to small polypropylenetubes, and all staining steps were carried out in suspension. Stained tissue wasthen mounted in ∼7 μL imaging buffer (Tris·HCl, pH 7.5, containing 100 mMcysteamine, 5% glucose, 0.8 mg·mL−1 glucose oxidase, and 40 μg·mL−1 catalase).The following primary antibodies were used: anti–HIM-3 (amino acids 154–253)[rabbit SDQ4713, 1:500, ModENCODE project (46)], anti–HTP-1/2 (rabbit, 1:500),anti–REC-8 (amino acids 171–270) [rabbit SDQ3914, 1:10,000, ModENCODEproject (46)], anti-GFP (mouse monoclonal, 1:500, Roche cat. no. 11814460001),anti-FLAG (mouse monoclonal M2, 1:500, Sigma), and anti-HA (mouse mono-clonal 2–2.2.14, 1:500, ThermoFisher Scientific). Secondary antibodies for STORMare Alexa Fluor 647–anti-Rabbit (goat or donkey, 1:500, Invitrogen) or AlexaFluor 647–anti-Mouse (donkey, 1:500, Jackson ImmunoResearch).

Correlative STORM and PALM Imaging. The 3D STORM imaging was performedon a custom microscope based on a Nikon Eclipse Ti-E inverted optical mi-croscope, using an oil immersion objective (Nikon CFI Plan Apochromat λ 100×,N.A. 1.45). Briefly, lasers at 647 nm (MPB Communications), 560 nm (MPBCommunications), 488 nm (Coherent), and 405 nm (Coherent) were coupledinto an optical fiber after an acousto-optic tunable filter and then introducedinto the sample through the back focal plane of the objective. Using atranslation stage, the laser beams were shifted toward the edge of the ob-jective so that emerging light reached the sample at incidence angles slightlysmaller than the critical angle of the glass–water interface to illuminate several

micrometers into the sample. For STORM imaging, continuous illumination ofthe 647-nm laser (∼2 kW∙cm−2; for AF647) was used to excite fluorescencefrom labeled dye molecules and to switch them into the dark state. For PALMimaging, continuous illumination of the 560-nm laser (∼2 kW·cm−2; formEos2) was used to excite fluorescence from photoconvertible mEos2 ormMaple3 proteins. Emitted photons were collected from single mEos2 ormMaple3 proteins until they photobleached. Concurrent illumination of the405-nm laser was used to convert native-state mEos2 or mMaple3 proteins intotheir photoconverted state. This process results in a cycle of photoconversionand photobleaching for each single protein. The power of the 405-nm laserwas adjusted during image acquisition so that at any given instant, only asmall, optically resolvable fraction of the fluorophores in the sample were inthe emitting state, thus avoiding the likelihood of multiple fluorophoresemitting at the same time within a diffraction-limited area. For 3D-STORM and3D-PALM imaging, a cylindrical lens was inserted into the imaging path tointroduce astigmatism so that images of single molecules were elongated in xand y for molecules on the proximal and distal sides of the focal plane (relativeto the objective), respectively (30). Single-molecule images were recorded at110 frames per second on an Andor iXon Ultra 897 EMCCD camera with a fieldof view of 256 × 256 pixels, corresponding to an area ∼40 μm by 40 μm in size.STORM and PALM images were taken consecutively, with STORM precedingPALM imaging due to the longer wavelength absorbed by the STORM dye.

Image Processing. The 3D STORM and 3D PALM datasets were processedaccording to previously described methods (27, 30), using the Insight3 softwarepackage developed by Bo Huang, University of California, San Francisco.Briefly, the optical astigmatism introduced into the optical path distortedsingle-molecule images into ellipses along the vertical or horizontal directionsfor molecules above or below the focal plane, respectively. The centroid po-sitions and ellipticities of the single-molecule fluorescent spots obtained fromraw STORM and PALM data were used to deduce lateral and axial positions ofsingle fluorescent emitters within each sample. The exact z position was cal-culated by mapping the degree of ellipticity of each single molecule to acalibration curve created by moving single dye molecules or fluorescent beadsadhered to a coverslip smoothly through the focal plane (30). Final super-resolution images used in figures result from blurring each single moleculepoint to a 2D Gaussian profile. For both STORM and PALM measurements,resolution was experimentally determined by repeatedly detecting the posi-tions of isolated single fluorophores in the sample over different frames andexamining the distribution of the thus detected positions (27, 30, 41). Thedistributions in x, y, and z directions were respectively fit to Gaussian functions,and the resultant full width at half maximum values from the fits were used torepresent the image resolution achieved (41).

Averaging Individual Synapsed Chromosomes. Once two-color 3D STORM andPALM images of late pachytene nuclei in whole gonads were processed andaligned, individual stretches of synapsed chromosomes from each image werealigned to a consistent orientation for purposes of overlaying and averaging. Thenumber and total lengths of stretches are summarized in Table S1. Individualsynapsed axes are then smoothed in x and z dimensions using a locally weightedscatterplot smoothing algorithm to ultimately produce axes running verticallywith no angular deviation in the xy plane or yz plane. Data from individualsynapsed chromosomes are combined and overlaid. By subsequently blurringthe data points to a 2D Gaussian profile, images that are effectively averagesover several stretches of synapsed chromosome axes are obtained. To determinethe localizations of components in the synapsed chromosome axes, histogramsin x (frontal view) and z (vertical view) were fitted to two and one normalizedGaussian distributions, respectively. To compare localizations, we use the posi-tions from the center of the SC in frontal view (x) and HWHMs in vertical view(z). SDs of the positions and HWHMs, corresponding to the precision of ourmeasurements, were estimated by a bootstrapping approach using a customscript in R. To this end, all possible combinations of subsets containing half thenumber of individual stretches were analyzed and SDs were calculated from theresults. This procedure has enabled a more accurate evaluation of the precisionof measurements than could be obtained from fitting individual stretches byeliminating potential bias existing in any one individual stretch due to experi-mental limitations such as inhomogeneous labeling.

ACKNOWLEDGMENTS. Some nematode strains used in this work were providedby the Caenorhabditis Genetics Center, which is funded by the NIH Office ofResearch Infrastructure Programs (P40 OD010440). This work was supported by apostdoctoral fellowship of the Human Frontier Science Program (LT000903/2013-C)(to S.K.), an NSF Graduate Research Fellowship (DGE 1106400) (to M.W.), thePew Biomedical Scholars Award and the Packard Fellowship for Science andEngineering (to K.X.), and National Institutes of Health Grant R01 GM065591(to A.F.D.) and the Howard Hughes Medical Institute.

Köhler et al. PNAS | Published online May 30, 2017 | E4741

CELL

BIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Mar

ch 1

1, 2

020

1. Zickler D, Kleckner N (1999) Meiotic chromosomes: Integrating structure and function.Annu Rev Genet 33:603–754.

2. Blat Y, Protacio RU, Hunter N, Kleckner N (2002) Physical and functional interactionsamong basic chromosome organizational features govern early steps of meiotic chi-asma formation. Cell 111:791–802.

3. Martinez-Perez E, Villeneuve AM (2005) HTP-1-dependent constraints coordinatehomolog pairing and synapsis and promote chiasma formation during C. elegansmeiosis. Genes Dev 19:2727–2743.

4. Ishiguro K, et al. (2014) Meiosis-specific cohesin mediates homolog recognition inmouse spermatocytes. Genes Dev 28:594–607.

5. Goodyer W, et al. (2008) HTP-3 links DSB formation with homolog pairing andcrossing over during C. elegans meiosis. Dev Cell 14:263–274.

6. Panizza S, et al. (2011) Spo11-accessory proteins link double-strand break sites to thechromosome axis in early meiotic recombination. Cell 146:372–383.

7. Niu H, et al. (2005) Partner choice during meiosis is regulated by Hop1-promoteddimerization of Mek1. Mol Biol Cell 16:5804–5818.

8. Schwacha A, Kleckner N (1997) Interhomolog bias during meiotic recombination: Meioticfunctions promote a highly differentiated interhomolog-only pathway. Cell 90:1123–1135.

9. Kim Y, Kostow N, Dernburg AF (2015) The chromosome axis mediates feedbackcontrol of CHK-2 to ensure crossover formation in C. elegans. Dev Cell 35:247–261.

10. FerdousM, et al. (2012) Inter-homolog crossing-over and synapsis in Arabidopsis meiosis aredependent on the chromosome axis protein AtASY3. PLoS Genet 8:e1002507.

11. Couteau F, Zetka M (2011) DNA damage during meiosis induces chromatin remod-eling and synaptonemal complex disassembly. Dev Cell 20:353–363.

12. Carballo JA, et al. (2013) Budding yeast ATM/ATR control meiotic double-strand break(DSB) levels by down-regulating Rec114, an essential component of the DSB-machinery.PLoS Genet 9:e1003545.

13. Couteau F, Zetka M (2005) HTP-1 coordinates synaptonemal complex assembly withhomolog alignment during meiosis in C. elegans. Genes Dev 19:2744–2756.

14. Severson AF, Ling L, van Zuylen V, Meyer BJ (2009) The axial element protein HTP-3promotes cohesin loading and meiotic axis assembly in C. elegans to implement themeiotic program of chromosome segregation. Genes Dev 23:1763–1778.

15. Zetka MC, Kawasaki I, Strome S, Müller F (1999) Synapsis and chiasma formation inCaenorhabditis elegans require HIM-3, a meiotic chromosome core component thatfunctions in chromosome segregation. Genes Dev 13:2258–2270.

16. Kim Y, et al. (2014) The chromosome axis controls meiotic events through a hierar-chical assembly of HORMA domain proteins. Dev Cell 31:487–502.

17. Pasierbek P, et al. (2001) A Caenorhabditis elegans cohesion protein with functions inmeiotic chromosome pairing and disjunction. Genes Dev 15:1349–1360.

18. Hopkins J, et al. (2014) Meiosis-specific cohesin component, Stag3 is essential formaintaining centromere chromatid cohesion, and required for DNA repair and syn-apsis between homologous chromosomes. PLoS Genet 10:e1004413.

19. Klein F, et al. (1999) A central role for cohesins in sister chromatid cohesion, formationof axial elements, and recombination during yeast meiosis. Cell 98:91–103.

20. Westergaard M, von Wettstein D (1970) Studies on the mechanism of crossing over.IV. The molecular organization of the synaptinemal complex in Neottiella (Cooke)saccardo (Ascomycetes). C R Trav Lab Carlsberg 37:239–268.

21. Rog O, Köhler S, Dernburg AF (2017) The synaptonemal complex has liquid crystallineproperties and spatially regulates meiotic recombination factors. eLife 6:e21455.

22. Ortiz R, Kouznetsova A, Echeverría-Martínez OM, Vázquez-Nin GH, Hernández-Hernández A (2016) The width of the lateral element of the synaptonemal complex isdetermined by a multilayered organization of its components. Exp Cell Res 344:22–29.

23. Syrjänen JL, Pellegrini L, Davies OR (2014) A molecular model for the role of SYCP3 inmeiotic chromosome organisation. eLife 3:e02963.

24. RongM,Matsuda A, Hiraoka Y, Lee J (2016) Meiotic cohesin subunits RAD21L and REC8 arepositioned at distinct regions between lateral elements and transverse filaments in thesynaptonemal complex of mouse spermatocytes. J Reprod Dev 62:623–630.

25. Schücker K, Holm T, Franke C, Sauer M, Benavente R (2015) Elucidation of synapto-nemal complex organization by super-resolution imaging with isotropic resolution.Proc Natl Acad Sci USA 112:2029–2033.

26. Prakash K, et al. (2015) Superresolution imaging reveals structurally distinct periodic patternsof chromatin along pachytene chromosomes. Proc Natl Acad Sci USA 112:14635–14640.

27. Rust MJ, Bates M, Zhuang X (2006) Sub-diffraction-limit imaging by stochastic opticalreconstruction microscopy (STORM). Nat Methods 3:793–795.

28. Betzig E, et al. (2006) Imaging intracellular fluorescent proteins at nanometer reso-lution. Science 313:1642–1645.

29. Hess ST, Girirajan TPK, Mason MD (2006) Ultra-high resolution imaging by fluores-cence photoactivation localization microscopy. Biophys J 91:4258–4272.

30. Huang B, Wang W, Bates M, Zhuang X (2008) Three-dimensional super-resolutionimaging by stochastic optical reconstruction microscopy. Science 319:810–813.

31. Heilemann M, et al. (2008) Subdiffraction-resolution fluorescence imaging withconventional fluorescent probes. Angew Chem Int Ed Engl 47:6172–6176.

32. Dani A, Huang B, Bergan J, Dulac C, Zhuang X (2010) Superresolution imaging ofchemical synapses in the brain. Neuron 68:843–856.

33. Kanchanawong P, et al. (2010) Nanoscale architecture of integrin-based cell adhe-sions. Nature 468:580–584.

34. Suleiman H, et al. (2013) Nanoscale protein architecture of the kidney glomerularbasement membrane. eLife 2:e01149.

35. Wang C-JR, Carlton PM, Golubovskaya IN, Cande WZ (2009) Interlock formation andcoiling of meiotic chromosome axes during synapsis. Genetics 183:905–915.

36. Phillips D, Nibau C, Wnetrzak J, Jenkins G (2012) High resolution analysis of meiotic chro-mosome structure and behaviour in barley (Hordeum vulgare L.). PLoS One 7:e39539.

37. Qiao H, et al. (2012) Interplay between synaptonemal complex, homologous re-combination, and centromeres during mammalian meiosis. PLoS Genet 8:e1002790.

38. Horn HF, et al. (2013) A mammalian KASH domain protein coupling meiotic chro-mosomes to the cytoskeleton. J Cell Biol 202:1023–1039.

39. Hell SW (2007) Far-field optical nanoscopy. Science 316:1153–1158.40. Huang B, Babcock H, Zhuang X (2010) Breaking the diffraction barrier: Super-

resolution imaging of cells. Cell 143:1047–1058.41. Xu K, Shim S-H, Zhuang X (2015) Super-resolution imaging through stochastic switching and

localization of single molecules: An overview. Far-Field Optical Nanoscopy, Springer Series onFluorescence, eds Tinnefeld P, Eggeling C, Hell SW (Springer, Berlin, Heidelberg), pp 27–64.

42. Phillips CM, McDonald KL, Dernburg AF (2009) Cytological analysis of meiosis inCaenorhabditis elegans. Methods Mol Biol 558:171–195.

43. Cui B, et al. (2007) One at a time, live tracking of NGF axonal transport using quantumdots. Proc Natl Acad Sci USA 104:13666–13671.

44. Tokunaga M, Imamoto N, Sakata-Sogawa K (2008) Highly inclined thin illuminationenables clear single-molecule imaging in cells. Nat Methods 5:159–161.

45. Schmekel K, et al. (1996) Organization of SCP1 protein molecules within synaptone-mal complexes of the rat. Exp Cell Res 226:20–30.

46. Gerstein MB, et al.; modENCODE Consortium (2010) Integrative analysis of the Cae-norhabditis elegans genome by the modENCODE project. Science 330:1775–1787.

47. Smolikov S, Schild-Prüfert K, Colaiácovo MP (2008) CRA-1 uncovers a double-strandbreak-dependent pathway promoting the assembly of central region proteins onchromosome axes during C. elegans meiosis. PLoS Genet 4:e1000088.

48. McKinney SA, Murphy CS, Hazelwood KL, Davidson MW, Looger LL (2009) A brightand photostable photoconvertible fluorescent protein. Nat Methods 6:131–133.

49. Wang S, Moffitt JR, Dempsey GT, Xie XS, Zhuang X (2014) Characterization and de-velopment of photoactivatable fluorescent proteins for single-molecule-based su-perresolution imaging. Proc Natl Acad Sci USA 111:8452–8457.

50. Frøkjaer-Jensen C, et al. (2008) Single-copy insertion of transgenes in Caenorhabditiselegans. Nat Genet 40:1375–1383.

51. Brar GA, Hochwagen A, Ee L-SS, Amon A (2009) The multiple roles of cohesin inmeiotic chromosome morphogenesis and pairing. Mol Biol Cell 20:1030–1047.

52. Winters T, McNicoll F, Jessberger R (2014) Meiotic cohesin STAG3 is required forchromosome axis formation and sister chromatid cohesion. EMBO J 33:1256–1270.

53. Anderson DE, Losada A, Erickson HP, Hirano T (2002) Condensin and cohesin displaydifferent arm conformations with characteristic hinge angles. J Cell Biol 156:419–424.

54. Haering CH, Löwe J, Hochwagen A, Nasmyth K (2002) Molecular architecture of SMCproteins and the yeast cohesin complex. Mol Cell 9:773–788.

55. Haering CH, et al. (2004) Structure and stability of cohesin’s Smc1-kleisin interaction.Mol Cell 15:951–964.

56. Gligoris TG, et al. (2014) Closing the cohesin ring: Structure and function of its Smc3-kleisin interface. Science 346:963–967.

57. Uhlmann F, Lottspeich F, Nasmyth K (1999) Sister-chromatid separation at anaphaseonset is promoted by cleavage of the cohesin subunit Scc1. Nature 400:37–42.

58. Siomos MF, et al. (2001) Separase is required for chromosome segregation duringmeiosis I in Caenorhabditis elegans. Curr Biol 11:1825–1835.

59. Kitajima TS, Miyazaki Y, Yamamoto M, Watanabe Y (2003) Rec8 cleavage by separaseis required for meiotic nuclear divisions in fission yeast. EMBO J 22:5643–5653.

60. Gruber S, Haering CH, Nasmyth K (2003) Chromosomal cohesin forms a ring. Cell 112:765–777.61. Stigler J, Çamdere GÖ, Koshland DE, Greene EC (2016) Single-molecule imaging reveals a

collapsed conformational state for DNA-bound cohesin. Cell Reports 15:988–998.62. Severson AF, Meyer BJ (2014) Divergent kleisin subunits of cohesin specify mecha-

nisms to tether and release meiotic chromosomes. eLife 3:e03467.63. Tzur YB, et al. (2012) LAB-1 targets PP1 and restricts Aurora B kinase upon entrance

into meiosis to promote sister chromatid cohesion. PLoS Biol 10:e1001378.64. Couteau F, Nabeshima K, Villeneuve A, Zetka M (2004) A component of C. elegans

meiotic chromosome axes at the interface of homolog alignment, synapsis, nuclearreorganization, and recombination. Curr Biol 14:585–592.

65. Börner GV, Barot A, Kleckner N (2008) Yeast Pch2 promotes domainal axis organi-zation, timely recombination progression, and arrest of defective recombinosomesduring meiosis. Proc Natl Acad Sci USA 105:3327–3332.

66. Wojtasz L, et al. (2009) Mouse HORMAD1 and HORMAD2, two conserved meioticchromosomal proteins, are depleted from synapsed chromosome axes with the helpof TRIP13 AAA-ATPase. PLoS Genet 5:e1000702.

67. Haering CH, Farcas A-M, Arumugam P, Metson J, Nasmyth K (2008) The cohesin ringconcatenates sister DNA molecules. Nature 454:297–301.

68. Huis in ’t Veld PJ, et al. (2014) Characterization of a DNA exit gate in the humancohesin ring. Science 346:968–972.

69. Onn I, Heidinger-Pauli JM, Guacci V, Ünal E, Koshland DE (2008) Sister chromatidcohesion: A simple concept with a complex reality. Annu Rev Cell Dev Biol 24:105–129.

70. Nasmyth K, Haering CH (2009) Cohesin: Its roles andmechanisms.AnnuRevGenet 43:525–558.71. Zhang N, Pati D (2009) Handcuff for sisters: A new model for sister chromatid co-

hesion. Cell Cycle 8:399–402.72. Eng T, Guacci V, Koshland D (2015) Interallelic complementation provides functional

evidence for cohesin-cohesin interactions on DNA. Mol Biol Cell 26:4224–4235.73. van Heemst D, Heyting C (2000) Sister chromatid cohesion and recombination in

meiosis. Chromosoma 109:10–26.74. Hunter N, Kleckner N (2001) The single-end invasion: An asymmetric intermediate at

the double-strand break to double-holliday junction transition of meiotic re-combination. Cell 106:59–70.

75. Niu H, et al. (2007) Mek1 kinase is regulated to suppress double-strand break repairbetween sister chromatids during budding yeast meiosis. Mol Cell Biol 27:5456–5467.

76. Li XC, Bolcun-Filas E, Schimenti JC (2011) Genetic evidence that synaptonemal complex axialelements govern recombination pathway choice in mice. Genetics 189:71–82.

77. Hong S, et al. (2013) The logic and mechanism of homologous recombination partnerchoice. Mol Cell 51:440–453.

E4742 | www.pnas.org/cgi/doi/10.1073/pnas.1702312114 Köhler et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 1

1, 2

020

78. Nebel BR, Coulon EM (1962) The fine structure of chromosomes in pigeon sper-matocytes. Chromosoma 13:272–291.

79. Dietrich AJ, van Marle J, Heyting C, Vink AC (1992) Ultrastructural evidence for a triplestructure of the lateral element of the synaptonemal complex. J Struct Biol 109:196–200.

80. Brenner S (1974) The genetics of Caenorhabditis elegans. Genetics 77:71–94.

81. Dickinson DJ, Ward JD, Reiner DJ, Goldstein B (2013) Engineering the Caenorhabditiselegans genome using Cas9-triggered homologous recombination. Nat Methods 10:1028–1034.

82. Redemann S, et al. (2011) Codon adaptation-based control of protein expression inC. elegans. Nat Methods 8:250–252.

Köhler et al. PNAS | Published online May 30, 2017 | E4743

CELL

BIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Mar

ch 1

1, 2

020