The synaptonemal complex has liquid crystalline properties and spatially regulates meiotic recombination factors

1In most eukaryotes, chromosome pairing during meiosis culminates with synapsis, defined as the assembly of synaptonemal complexes (SCs) between homologous chromosomes (homologs). Upon meiotic entry, prior to pairing and synapsis, replicated chromosomes reorganize around a central ‘axis,’ a linear structure comprised of cohesin complexes and associated meiosis-specific proteins. Once chromosomes establish local interactions with their homologs through recombination or other mechanisms, the central region of the SC nucleates and assembles progressively between paired axes, resulting in close side-by-side alignment of homologous chromosomes along their entire lengths. SC assembly is required for stable interhomolog pairing, normal levels of crossover (CO) recombination, cell cycle progression, and faithful chromosome segregation.

2While disruption of the SC dramatically alters the CO distribution in a variety of organisms, the functional contribution of the SC to CO regulation remains hotly debated. Evidence from budding yeast has suggested that the SC may be a passive ‘glue’ that merely stabilizes the physical pairing of homologs to permit efficient CO formation. However, evidence from C. elegans has indicated that the SC plays a direct role in the conserved phenomenon of crossover interference, which distributes COs in a non-random, widely spaced pattern along each chromosome.

3The SC was initially observed sixty years ago by thin-section transmission electron microscopy (TEM), which revealed electron-dense linear structures at the interface between meiotic chromosomes. Subsequent observations of meiocytes from diverse eukaryotes led to a consensus view that the SC is a symmetrical, tripartite structure, with two parallel, darkly-staining lateral bands flanking an electron-lucent, transversely striated, central region (and Figure 1B). Electron microscopy, superresolution fluorescence imaging, and protein-protein interaction analysis have clearly indicated that the proteins that make up this complex form a highly ordered, periodic structure with bilateral symmetry. Serial-section TEM analysis also led to the discovery of ‘recombination nodules’ associated with SCs at sites of genetic exchange, and specific recombination factors have been localized to these sites by a variety of cytological methods. However, how the SC assembles between chromosomes, and how this polymer might govern or respond to meiotic recombination are largely mysterious.

4Synapsis of homologous chromosomes typically requires both nucleation factors and structural proteins that localize throughout the SC, which spreads along the interface between aligned chromosomes. Structural components, defined here as proteins required for SC assembly, have been identified through genetic screens and/or immunocytochemistry in a variety of organisms. Three to five such proteins have been characterized in budding yeast, Drosophila, C. elegans, and mammals, and at least one SC protein has been identified in plants. However, it is not yet clear whether the full cohort of SC structural elements has been defined in any organism. While the dimensions and appearance of the SC are conserved among diverse phyla, these structural proteins show remarkable divergence in their lengths and primary sequences. All known SC structural proteins contain extensive regions that score highly on coiled-coiled prediction scales. Together with the transverse striations observed by TEM, this has led to a prevalent view of the SC as a stable, ‘crosslinking’ polymer assembled through coiled-coil interactions, perhaps resembling intermediate filaments. However, other observations have hinted at a more dynamic structure. In particular, evidence from budding yeast has revealed that SCs can incorporate newly translated proteins after assembly, consistent with dynamic exchange of subunits. Additionally, in many organisms SCs are known to undergo ‘synaptic adjustment,’ post-assembly progressive reorganization that tends to simplify their topology (e.g., the extent of loop structures and unsynapsed axes is minimized) – a behavior that implies extensive rearrangement of subunits after initial assembly of the complex.

5Here we probe the biophysical properties of the SC in vivo, and show that it exhibits liquid-like behaviors and likely assembles through a regulated coacervation process. We further report that SCs in C. elegans, budding yeast, and Drosophila are rapidly and reversibly dissolved by aliphatic alcohols, indicating that their integrity relies on weak hydrophobic interactions. Finally, we demonstrate that the dynamic localization and interdependence of two essential CO factors is recapitulated by chromosome-free assemblies of SC proteins, demonstrating how an SC compartment may dynamically partition enzymatic activities to regulate meiotic recombination.

6The SC normally assembles between the paired axes of meiotic chromosomes. Note that throughout this work we use terminology that differentiates between the SC (sometimes called the 'central region') and the chromosome axis. In the nematode C. elegans, the SC comprised of at least four structural proteins (SYP-1–4), which are mutually dependent for their chromosome association. When the chromosome axis is disrupted — for example, in C. elegans mutants lacking either the essential meiotic axis protein HTP-3, or all three meiotic kleisins (REC-8, COH-3, and COH-4) — SC proteins fail to assemble between chromosomes, and instead form large bodies in the nucleoplasm of meiotic cells (Figure 1A and Figure 1—figure supplement 1A). Using transmission electron microscopy (TEM), we found that these nuclear bodies have a multilaminar internal structure, in which each layer recapitulates the appearance and dimensions of SCs (Figure 1B and Figure 1—figure supplement 2A and C). Similar periodic, multilaminar structures formed by SC proteins have been observed in wild-type nuclei or cytoplasm of meiocytes or nurse cells from a wide variety of organisms, as well as under perturbed conditions where SCs cannot assemble between chromosomes, and are known as ‘polycomplexes’. In C. elegans these structured bodies have also been observed in wild-type meiocytes prior to SC assembly. Importantly, polycomplexes in htp-3 mutants do not contain any of the known chromosome axis components (cohesins, HIM-3, or HTP-1–3) (, and our observations), indicating that these proteins are not essential for self-assembly of SC proteins into an ordered structure, but only for their assembly as a unilamellar structure between chromosomes. Additionally, this reveals that the electron-dark bands with 100 nm periodicity within polycomplexes, which correspond to the cytologically-defined ‘lateral elements’ of the SC, are not equivalent to chromosome axes.

7Polycomplexes contain all known SC structural proteins and depend on the full cohort of these proteins for their formation (, and our observations). Coupled with their structural similarity to SCs, we reasoned that it might be possible to learn more about the SC, independent of its interactions with chromosome axes, by analyzing polycomplex dynamics in living worms. Time-lapse imaging of htp-3 mutant worms expressing GFP-SYP-3 revealed that prophase nuclei frequently contained more than one fluorescent polycomplex. Unexpectedly, we observed that when these came into contact, they readily fused to form a single larger body (Videos 1 and 2, Figure 1C and Figure 1—figure supplement 2B). We also observed polycomplexes to undergo large-scale structural deformations, particularly towards the end of prophase as the nuclear volume increased (Video 1 and Figure 1C). The ability of such bodies to merge together and to readily change in shape are defining characteristics of liquids, since they imply that molecules can rapidly rearrange within a material. These observations were thus initially very surprising, given the highly ordered appearance of polycomplexes. However, materials with liquid-like properties and ordered arrangements of molecules, known as structured fluids or liquid crystals, are a well-known class of soft condensed matter, and a variety of structures within cells have been postulated to have such properties.

8In vivo imaging also revealed that polycomplexes likely assemble by regulated phase separation. Our early efforts to carry out long-term recordings in living worms consistently led to meiotic arrest, which we deduced to be a physiological response to food withdrawal. Arrest was evident as abatement of the dramatic chromosome movements that normally occur throughout early prophase, which typically ceased within 15–20 min of immobilization (Figure 2A and Figure 2—figure supplement 1 and Videos 3 and 4; see Figure 2B for a schematic of the C. elegans gonad). In nuclei that had initiated synapsis before arresting, we observed that diffuse GFP-SYP-3 gradually coalesced to join existing segments of SC between chromosomes. These partial SCs became brighter but not longer, likely reflecting addition of subunits to form a multilayered structure, much as they must stack within 3D polycomplexes. Consistent with this idea, EM analysis of lateral and cross-sectional views of SCs has indicated that they contain varying numbers of layers. The observation that these pre-existing SC segments did not spread along partially synapsed chromosomes following arrest suggests that longitudinal extension is an active process. In earlier, presynaptic nuclei, GFP-SYP-3 coalesced within each nucleus over a period of tens of minutes to form several compact aggregates. Similar coalescence of SC proteins was observed in animals treated with sodium azide, which reversibly arrests meiosis and other physiological processes by depleting ATP (Figure 2C), suggesting that phosphorylation and/or other ATP-dependent mechanisms normally maintain the solubility of SC subunits.

9The liquid-like behaviors we observe for polycomplexes, as well as the spontaneous coalescence of SC components from the nucleoplasm upon arrest, indicate that these nuclear bodies arise through coacervation, or phase separation, as been inferred for other self-assembling subcellular compartments. SC proteins have also been shown to form nuclear aggregates in C. elegans following extended incubation at elevated temperatures, which also renders the animals sterile. We found that these heat-induced bodies lack periodic striations in electron micrographs and do not undergo fusions or shape deformations (Figure 1—figure supplement 1B–D and Video 5), indicating that they lack both the liquid-like properties and internal order of polycomplexes. This irreversible aggregation likely represents conversion to a denatured state, similar to the way that other coacervates can transition to aberrant solid or amyloid-like states.

10We next investigated whether SCs assembled between chromosomes also exhibit liquid-like behaviors by assessing the mobility of SC proteins within these structures. We engineered worm strains expressing SYP-3 fused to the photoconvertible fluorescent protein mMaple3. We photoconverted subnuclear regions in live worms and imaged these nuclei over time. The photoconverted proteins rapidly redistributed throughout the SCs within the same nucleus, becoming homogeneously dispersed within 22 min (Figure 3A,C and Figure 3—figure supplement 1A). By contrast, when the axis protein HIM-3 was tagged with mMaple3, we observed no significant redistribution of converted fluorophores over more than 30 min of observation (Figure 3B–C and Figure 3—figure supplement 1B). This implies that other axis components, particularly HTP-3, which scaffolds the recruitment of HIM-3, are also stably associated with the axis. We conclude that SC central region proteins are far more mobile than axis components. The relatively slow rate of subunit turnover within the SC compared to that observed for other liquid-like compartments might reflect a high viscosity of the liquid-like phase, and/or other constraints such as confinement within the very thin laminar space, or a slow rate of exchange of subunits between the different SC compartments in the nucleus. Future studies using high-speed imaging and other tools should provide a more complete understanding of the movement of proteins within the SC. Taken together, our observation of coacervation following meiotic arrest, of liquid-like behaviors of polycomplexes, and of the dynamic exchange of subunits all indicate that the SC is a phase-separated compartment with liquid crystalline properties.

11Recent work has illuminated several mechanisms that can promote intracellular phase transitions, including multivalent interactions, low-complexity protein domains with β-strand propensity, and hydrophobic interactions. The latter are thought to be essential for the integrity of the nuclear pore, within which a network of unstructured proteins forms a diffusion barrier, and for the formation of P-granules in the C. elegans germline, among other liquid-like cellular compartments. In support of this idea, these compartments are reversibly disrupted by moderate concentrations (~5%) of 1,6-hexanediol or similar aliphatic alcohols, which lower aqueous surface tension and reduce the hydrophobic effect.

12To test the idea that hydrophobic interactions promote coacervation of SC proteins, we extruded worm gonads in physiological buffer and added varying concentrations of 1,6-hexanediol. When gonads expressing GFP-SYP-3 were exposed to 5–10% 1,6-hexanediol, we observed immediate disruption of SCs and dispersion of the fluorescence throughout the nucleoplasm. By contrast, the distributions of many other chromosome-associated proteins were unaffected by this treatment, including axis components (HTP-3, HIM-3, HTP-1/2, LAB-1, and cohesins); the pairing center protein HIM-8; and proteins associated with CO sites (COSA-1 and ZHP-3) (Figure 4A–B, Figure 4—figure supplement 3A, Videos 6 and 7, and data not shown; see Figure 4—figure supplement 1C for an explanatory diagram; in all figures, panels showing samples treated with 1,6-hexanediol have pink borders). In some nuclei a short stretch of SC associated with the X chromosomes persisted after treatment with 1,6-hexanediol (Figure 4—figure supplement 3), perhaps reflecting enhanced recruitment of SC proteins to the paired sex chromosomes. SC dissolution following 1,6-hexanediol exposure was also observed in wild-type gonads, indicating that it was not a consequence of the fluorescent tag (Figure 4—figure supplement 3A). Polycomplexes showed a very similar response to 1,6-hexanediol: they rapidly dissolved upon exposure to the solvent, while heat-induced SC aggregates were resistant to dissolution (Figure 5B, Figure 4—figure supplement 4 and Video 8 and Video 9).

13Despite the widespread conservation of SC structure among eukaryotes, central region proteins are notorious for their divergence in primary sequence among different lineages. We therefore wondered whether the biophysical properties we observed in C. elegans are conserved. We found that SCs in both Drosophila and budding yeast were rapidly dispersed by 1,6-hexanediol, while axis markers (Rec8p in budding yeast; C(2)M in Drosophila) remained associated with chromosomes, as in C. elegans (Figure 4C–E; see Figure 4—figure supplement 1C for an explanatory diagram). This indicates that weak hydrophobic interactions are important for SC integrity in diverse lineages.

14While SCs in budding yeast, Drosophila, and C. elegans meiocytes were readily disrupted by 1,6-hexanediol, we found that similar concentrations of the more polar solvents ethanol or dimethyl sulfoxide failed to dissolve these structures (data not shown). Because the effects of 1,6-hexanediol on cellular structures are not well understood, we also probed the sensitivity of SCs to disruption by a number of other water-miscible alcohols. These di-alcohols vary in the length of the hydrocarbon chain and the positions of their hydroxyl groups. Most experiments were performed with yeast cells arrested in mid-pachytene, since it is simple to manipulate the media in which these cells are immersed and no dissection or cell permeabilization was required to see the effects of 1,6-hexanediol. For each di-alcohol we determined the minimal concentration required for complete SC disruption, based on loss of any discernible structure by fluorescence microscopy of unfixed yeast cells (Figure 4—figure supplement 1B). We observed a clear correlation between the extent of the hydrophobic domains of different solvents and their effects on the SC. For example, 1,2-heptanediol disrupted the structure at lower concentrations than 1,2-hexanediol, which was more potent than 1,2-pentanediol. Similar results were obtained for the series 1,7-heptanediol, 1,6-hexanediol, and 1,5-pentanediol. Together, these results reinforce the idea that solvation by aliphatic alcohols is due to suppression of the hydrophobic effect, and support a major contribution of hydrophobic interactions to SC stability.

15We also probed the effect of electrostatic interactions on SC stability. We noted that SC central region proteins in various organisms are predominantly polyampholytic, i.e., they contain an unusually large fraction of charged residues, both acidic and basic, which are well-mixed within the primary sequences. This suggests that electrostatic interactions likely contribute to SC integrity, and that higher salt concentrations, which shield electrostatic interactions, might disfavor assembly. Consistent with this, we found that increasing the concentration of KCl potentiated the ability of 1,6-hexanediol to dissolve the SC in budding yeast (Figure 4—figure supplement 1A; other salts showed similar effects [data not shown]), highlighting a role for electrostatic interactions, in addition to hydrophobic interactions, in SC assembly and stability. The effects of increasing salt concentration were fairly modest in light of the highly charged nature of SC proteins; however, only a limited range of ionic strength could be investigated in living cells, and the effects of these buffer manipulations on intracellular ionic strength could not be directly monitored.

16Both of these trends observed in budding yeast — disruption of SCs by lower concentrations of more hydrophobic diols, and synergy between electrostatic shielding and amphipathic solvents — also held true for C. elegans SCs (data not shown). Experiments conducted using this system were less comprehensive since they required extensive hand dissection and, probably as a consequence of this manipulation, were somewhat less consistent and thus harder to score. Nevertheless, we tested whether various mutations in C. elegans that perturb meiotic recombination or chromosome structure might affect 1,6-hexanediol sensitivity, and observed no major differences, except that the aberrant SCs in syp-3(me42) mutants, which carry a small C-terminal truncation in an SC protein, were hypersensitive to 1,6-hexanediol (Figure 4—figure supplement 2A–B). We further probed the effects of this mutation, and found that the syp-3(me42) truncation also prevents polycomplex formation in an htp-3 background (Figure 4—figure supplement 2C), as well as heat-induced aggregation (Figure 4—figure supplement 2D). The correlation between 1,6-hexanediol hyper-sensitivity, defects in SC assembly and lack of SC aggregation further underscore the contribution of hydrophobic interactions to the functional integrity of the SC.

17When SCs in C. elegans were disrupted by 1,6-hexanediol, followed by rapid dilution of the solvent with buffer, GFP-SYP-3 immediately relocalized along the chromosomes (Video 10 and Figure 5A; throughout the figures, panels showing samples subsequent to the dilution are shown with a yellow border). Some fluorescent material that had exited the nucleus, perhaps due to transient disruption of the nuclear pores, formed small puncta in the cytoplasm. Notably, reassembly of SCs following 1,6-hexanediol dilution occurred much more quickly than physiological SC assembly between axes, which requires 20–30 min per chromosome pair. Polycomplex dissolution was also reversible: upon dilution of 1,6-hexanediol, multiple small puncta formed in each nucleus, and rapidly fused with each other to form larger bodies (Video 9 and Figure 5B), qualitatively similar to the behavior of polycomplexes in live animals (Figure 1C, Figure 1—figure supplement 2B and Videos 1 and 2). This reassembly following hexanediol dilution occurred much more quickly than under metabolic arrest (Figure 2A and Figure 2—figure supplement 1). Dissolution of the SC by 1,6-hexanediol in budding yeast cells was also reversible: we observed some re-association of SC components with chromosomes after both short (2 min) and longer (>30 min) incubations of yeast cells arrested at mid-pachytene (Figure 5C). These results provide additional evidence that neither the chromosome axes, which are essential for SC assembly between chromosomes, nor the SC proteins themselves, are irreversibly perturbed by exposure to 1,6-hexanediol.

18Subcellular liquid-like compartments are selectively permeable to macromolecules, and can regulate biochemical reactions by concentrating enzymes and substrates and restricting their diffusion (reviewed by ). We therefore wondered whether the SC might regulate and surveil recombination by partitioning CO factors into a distinct compartment between homologous chromosomes. At least one essential CO factor that is not required for SC assembly, the conserved RING-finger protein ZHP-3, is observed throughout the SC during early prophase, before becoming restricted to designated CO sites (Figure 6—figure supplement 1A), whereupon it colocalizes with COSA-1, another conserved protein required for COs. We found that CO formation altered not only the localization of ZHP-3 but also its sensitivity to 1,6-hexanediol: Prior to CO designation, the ZHP-3 distributed throughout SCs was dispersed by the solvent, while CO-associated ZHP-3 foci remained intact even after disruption of the SC by 1,6-hexanediol (Figure 6—figure supplement 1A–B).

19If ZHP-3 concentrates between chromosomes by partitioning into the SC, we expected that it might also localize within polycomplexes. Indeed, we found that in htp-3 mutants, ZHP-3 was concentrated throughout polycomplexes during most of meiotic prophase. However, near the ‘loop’ region of the gonad, where exit from pachytene normally occurs (see Figure 2B for reference), we observed an abrupt change in ZHP-3 localization: The protein largely disappeared from the interior of the polycomplexes and became restricted to a single focus conspicuously abutting the surface of each polycomplex. Intriguingly, at the same stage, COSA-1, which was not detected earlier, became localized to these same small, peripheral puncta (Figure 6A–C). This dynamic relocalization of ZHP-3 and COSA-1 is highly analogous to what is observed along bona fide SCs upon CO designation. We further determined that ZHP-3 is required for the appearance of COSA-1 foci at polycomplexes, while COSA-1 is required for the relocalization of ZHP-3 to the edge of these bodies and its disappearance from the interior (Figure 6D–E and Figure 6—figure supplement 1C–D), mirroring the interdependence of these factors for their localization to CO sites. We also found that the earlier ZHP-3 localization throughout polycomplexes is readily disrupted by 1,6-hexanediol, whereas the foci of COSA-1 and ZHP-3 foci at the edges of polycomplexes are 1,6-hexanediol-resistant (Figure 6—figure supplement 1E), again mirroring the properties of ZHP-3 and COSA-1 associated with SCs and designated CO sites, respectively. Polycomplexes do not appear to be associated with chromosomes and moreover, meiotic recombination does not initiate in htp-3 mutants. Thus, localization of COSA-1 and ZHP-3 to foci on polycomplexes does not require their interaction with CO intermediates such as Holliday junctions, or indeed with chromosomes. These results imply that in response to exogenous cell cycle signals, polycomplexes can recapitulate a signaling network that normally designates COs, and hence suggest how a liquid-like SC compartment may promote and regulate CO recombination along meiotic chromosomes.

20Sixty years of observation by EM has revealed that SC structure, as well as the tendency of its components to assemble into ordered polycomplexes, are widely conserved. Here we report another conserved property of SCs and polycomplexes: they assemble through coacervation of polypeptides into compartments with liquid crystalline properties. While the symmetric, transversely striated appearance of the SC, combined with known properties of coiled-coil proteins, has suggested a filamentous and stable internal structure, our observations indicate that SCs instead resemble smectic (planar) liquid crystal bilayers, structurally analogous to lipid bilayers. Moreover, the SC can also assemble into its canonical smectic structure when its normal association with chromosome axes is prevented. The preferential assembly of SC proteins into a thin lamellar structure between two axes is further evidenced by our observation that upon meiotic arrest, nuclear aggregates appear only if no pre-assembled SC bilayers are present (Figure 2). While biological membranes are well-known examples of smectic liquid crystals, to our knowledge, the SC is the first known example of such a highly ordered, liquid-like structure arising from protein-protein interactions. However, a variety of biological assemblies of less regular structure have been described as liquid crystals, and many others have been created in vitro by mixing synthetic peptides in aqueous solution.

21Liquid crystals are considered to be a distinct phase of matter from liquids, which are disordered and anisotropic. Thus, it is perhaps unsurprising that the proteins that assemble to form SCs are markedly different from those that have been implicated in the assembly of other liquid-like compartments within cells. Many such proteins have low-complexity, intrinsically disordered domains that are essential for their phase transition behavior, and some of these have been shown to adopt cross-beta strand structures in vitro. By contrast, SC proteins are predicted to be largely alpha-helical in structure, and have extensive domains predicted to form coiled-coils. Smectic liquid crystals assemble from elongated, amphipathic rod-like molecules. It is known that some of the longer SC proteins, particularly SYCP1 in mammals and Zip1p in budding yeast, have a head-to-head orientation within the SC, with their N-termini near the center and their C-termini near the chromosome axes. Interaction of their C-termini with axis-associated proteins seems to promote their coacervation with other SC proteins, possibly by promoting high local concentrations. Axis association also favors the formation of single bilayers (lamellae), rather than the stacks of bilayers that make up polycomplexes. By specifying the orientation of these ‘transverse filament’ proteins, the chromosome axes may also act as a ‘director’ for the liquid crystalline array.

22Many of the best-studied coiled-coil proteins form highly stable interactions through regions of superhelical coiling. Some, such as muscle myosin, keratins, and lamins, form long-lived filamentous networks, which has suggested that the SC might have similar properties. However, it has long been appreciated that interruptions in the phasing of the canonical heptad repeat sequence disrupt supercoiling and reduce the stability of interactions. We note that even the ‘transverse filament’ proteins SYCP1 (vertebrates and other metazoans), Zip1p (budding yeast), C(3)G (Drosophila), and SYP-1 (Caenorhabditis) contain very few perfect heptads arranged in tandem, which may underlie the lability of the SC.

23Meiotic synapsis can be disrupted by small increases in temperature, as seen in C. elegans and many other organisms, and we propose that this stems from the liquid crystalline properties of the SC. While the C. elegans Bristol N2 strain remains fertile at temperatures up to 25°C, at 26.5°C hermaphrodites are sterile, and in such animals we observed SC protein assemblies that were neither crystalline nor liquid-like. Because liquid crystals depend on ordered but weak interactions, they are considered to be a ‘mesophase’ between liquids and crystalline solids. They are often exquisitely sensitive to variations in temperature and/or other perturbations such as electric fields, making synthetic materials useful as thermometers (on fish tanks and mood rings) and in video displays. Thus strong selective forces to maintain the responsive, liquid-like properties of the SC, which plays a central role in fertility and fitness, may underlie the sequence diversity observed among its constituent proteins, especially among ectothermic organisms. In support of this idea, direct evidence has recently indicated that SC proteins, among other chromosome structural components, are subject to rapid positive selection in plant populations growing at different temperatures.

24Our observation that meiotic arrest or azide treatment promotes the coalescence of SC components is consistent with the idea that phosphorylation and/or other modifications regulate the coacervation of these proteins. These modifications presumably tune the strength of interactions among SC proteins to control the timing and location of assembly and disassembly, as has been described for other cellular coacervation processes, such as the formation of P-granules. Consistent with this, PLK-2, a kinase that promotes SC disassembly at the end of meiotic prophase, localizes to SCs, and we have found that several of the SYP proteins can be phosphorylated by PLK-2 in vitro (Y. Kim and AFD, unpublished). Because liquid crystals are ultrasensitive to perturbations of their intermolecular interactions, we envision that posttranslational modifications may also modulate the dynamic properties of assembled SCs during meiotic progression.

25Phase-separated liquid-like compartments in vitro and in vivo are selectively and tunably permeable to macromolecules. This property can concentrate specific cellular activities while excluding others. While the liquid crystalline properties of the SC set it apart from other compartments that resemble amorphous liquids, we provide evidence that it may have some analogous properties. Specifically, this assembly dynamically regulates the localization of at least two conserved recombination factors, ZHP-3 and COSA-1 (Figure 6 and Figure 6—figure supplement 1), which are not required for SC assembly, in response to meiotic progression. Previous work has shown that Zip3p, the budding yeast homolog of ZHP-3, also localizes throughout polycomplexes, while the CO factors Zip2p, Zip4p, Msh5p and the 9-1-1 clamp proteins localize to one or two foci abutting polycomplexes, sometimes referred to as a ‘capping complex’, very similar to the physical distribution we observe for ZHP-3 and COSA-1 during late meiotic prophase. Homologs of ZHP-3 in plants, mammals and fungi also localize to SCs before concentrating at sites of COs, suggesting that spatial regulation of COs by the SC is widely conserved. We envision that the SC acts as an active ‘scaffold’ to localize and concentrate ‘client’ proteins, defined here as SC-resident proteins that are not required for SC assembly. These factors, including ZHP-3, are recruited by the SC to control the processing of double-strand breaks that are induced during meiosis. We also speculate that the SC may share an evolutionary relationship with other liquid-like nuclear compartments, such as repair foci, which concentrate repair factors at sites of DNA damage.

26Our observations support a central role for the SC in CO interference, which acts over many microns and even over entire chromosomes. Evidence that SCs share physical properties with liquid crystals, a particularly responsive class of active materials, illuminates how they might contribute to the transduction of a long-range, cis-acting interference signal (Figure 6C). In particular, the knowledge that macromolecules can diffuse laterally within the plane of smectic liquid crystals suggests that the dynamics of diffusion within the SC likely determine the rate at which biochemical signals are propagated along the chromosomes. (By analogy, 2D diffusion within lipid membranes profoundly impacts the spatial organization and kinetics of signaling by membrane-associated proteins.) Thus, our discovery that the SC acts as a non-membrane-bound compartment suggests how COs could be regulated in cis through a reaction-diffusion mechanism, which is consistent with quantitative analysis of CO interference in several organisms. This idea is also consistent with evidence that reduction in the expression levels of SC proteins, which likely results in discontinuities of the SC along individual chromosomes, can permit additional COs.

27We further report that both SCs and polycomplexes undergo an abrupt biochemical change in response to cell cycle signals, which results in the appearance of a single CO-like focus and removal of ZHP-3 from the interior, indicating that this material can mediate switch-like regulation that spans an entire compartment. We have found that depletion of the ERK kinase MPK-1, which has been implicated in regulating late meiotic prophase events in C. elegans, does not affect the relocalization of COSA-1 or ZHP-3 on SCs or polycomplexes (Liangyu Zhang, OR and AFD, unpublished), so the nature of the ‘exit pachytene’ signal that triggers this switch is unknown.

28Finally, we note that long-range order within liquid crystals allows propagation of structural transitions, known as ‘disclinations,’ over long distances, similar to a ‘domino effect.’ Energy can also be stored within liquid crystals as elastic strain, and released through abrupt topological changes. This property could provide a rapid response to crossover formation. Specifically, we imagine that structural changes could be induced within the SC at sites of CO formation and transmitted laterally, in which case interference might be mediated by a mechanochemical rather than a purely biochemical signal. This idea is analogous to a ‘beam-film’ model that has been proposed to explain interference, but posits that the medium that stores strain is the SC, or more precisely, the SC as confined by its interaction with chromosome axes. Ultrastructural analysis of SC organization before and after CO formation may eventually allow this idea to be tested. Additionally, future studies of the role of SC organization and dynamics in CO interference may enable development of new synthetic materials with long-range signaling capabilities.

29All strains were cultured using standard methods. Worms were maintained at 20°C, except that animals expressing mMaple-HIM-3 and mMaple-SYP-3 were cultured at 25°C to enhance transgene expression. Where indicated, worm plates were incubated overnight at 26.5°C to induce SC aggregation. See Table 1 for a full list of the strains used.

30HTP-3-mRuby (ieSi17) and mRuby-SYP-3 (ieSi19) were constructed and inserted using MosSCI, essentially as previously described for GFP-SYP-3 and HTP-3-GFP. A transgene including a 3x(Gly-Gly-Ser-Gly) linker and mRuby immediately downstream of the coding sequnce of htp-3 was constructed using the MosSCI repair template pCFJ178. This construct was integrated at the cxTi10882 IV site by MosSCI, and later homozygosed. mMaple3-SYP-3 was constructed using a codon-optimized synthetic DNA sequence, with modifications to improve germline expression, and inserted by MosSCI. All SYP-3 and HTP-3 transgenes were crossed into mutants lacking the corresponding wild-type proteins (syp-3(ok758) or htp-3(tm3655), respectively). mMaple3-HIM-3 was constructed by inserting a codon-optimized mMaple3 sequence at the N-terminus of the endogenous him-3 gene using CRISPR-Cas9. Analysis of htp-3 mutant animals was performed on homozygous progeny of balanced heterozygotes, and their heterozygous siblings served as wild-type controls. A cosa-1(ie98) htp-3(tm3655) strain was constructed by inserting a premature stop codon to disrupt the cosa-1 gene using CRISPR-Cas9 in htp-3(tm3655)/hT2 animals. A zhp-3(ie97) htp-3(tm3655); COSA-1-GFP strain was constructed by inserting a premature stop codon disrupting the zhp-3 gene using CRISPR-Cas9 in COSA-1-GFP; htp-3(tm3655)/hT2 animals.

31Adult hermaphrodites were immobilized for imaging as previously described. Briefly, worms were placed on freshly prepared agarose pads (7.5% in water) overlaid with 100 nm polystyrene beads (Polysciences, cat#00876). The pad and beads included freshly-dissolved serotonin creatinine sulfate (Sigma-Aldrich, St. Louis, MO) at a final concentration of 25 mM. Worms were overlaid with high-performance coverslips (0.17 ± 0.005 mm; Schott) that were sealed to the slide with VALAP (1:1:1 vaseline:lanolin:paraffin), and imaged immediately afterwards. Time-lapse images were recorded from the posterior gonad arm, the motion of which was less perturbed by pharyngeal pumping than the anterior arm. Only gonads for which chromosome motion was detected during the entire recording were analyzed, except where otherwise indicated.

32Hexanediol treatment during time-lapse imaging was performed by dissecting adult hermaphrodites in 30 μL 1X egg buffer (EB) on poly-L-lysine covered coverslips (Neuvitro, Vancouver, WA). Extruded gonads were gently pressed against the coverslip to promote adhesion. At the indicated times, an equal volume of hexanediol in EB was added to the drop to attain the indicated final concentrations of hexanediol. Where indicated, the buffer was subsequently diluted with additional EB lacking hexanediol.

33Images were acquired using a Marianas spinning-disc confocal microscope (Intelligent Imaging Innovations, Inc. [3i], Denver, CO) at ambient temperature (19°C–24°C), using a 100 × 1.46 NA oil immersion objective, yielding a pixel spacing of 133 × 133 nm. 3D image stacks of 11–17 sections at 0.5 µm z-spacing were acquired over 1–3 s. Raw image stacks were analyzed using Imaris 7.3 or 8.0 (Bitplane AG, Zurich, Switzerland). Time-lapse series were segmented, tracked and aligned based on overall fluorescence using the Spots tool. Compactness was measured using the Surface tool, analyzing at least three different gonads for each condition.

34Photoconversion of mMaple3 fusion proteins was performed using the Marianas spinning-disc confocal system described above, equipped with a Vector FRAP module. Two-color 3D image stacks of 11 sections at 0.8 µm z-spacing were acquired over 1–3 s. The first time point included activation of 10 × 10 pixel regions (1.33 × 1.33 µm) using a 405 nm laser at each focal position. Following photoconversion, 3D stacks were acquired every 1–5 min, as indicated.

35Our recordings indicate that photoconverted molecules can spread both along SCs and between SCs on different chromosomes. We thus assume that this movement has multiple rate components. At the temporal and spatial resolution we could achieve in living animals with spinning disk confocal microscopy, it was not possible to directly measure diffusion rates or the surface tension of individual SC compartments. Instead, we compared the spreading rates of SYP-3 (SC) and HIM-3 (axis) proteins by calculating the rate of change of the volume of the nucleus containing photoconverted (red) signal, based on 3D images from time-lapse recordings. The volume of the nucleus occupied by photoconverted molecules at each time point was placed in one of four bins (25%, 50%, 75% or 100%), which were then plotted as a function of time elapsed since photoconversion. At least 15 nuclei were scored for each fluorescent marker. The expansion rate was defined as the slope of a line fitted to the data. These values were compared to simple simulations in which a point source of fluorescence became homogeneously distributed throughout the nucleus in 1 hr.

36Immunofluorescence was performed as previously described, except that where indicated, dissected hermaphrodites were briefly incubated with 1,6-hexanediol in egg buffer immediately before adding fixative. The following antibodies, all of which have been previously described, were used: anti-SYP-2 (rabbit, affinity purified, 1:1,000), anti-SYP-1 (goat, affinity purified, 1:1,000), anti-GFP (mouse monoclonal, Roche, Indianapolis, IN, 1:400), anti-SUN-1 (rabbit, affinity-purified, Novus Biologicals, Littleton, CO, 1:10,000), anti-HTP-3 (guinea pig, 1:500), anti-HIM-8 (rat, 1:500), anti-ZHP-3 (rabbit, affinity-purified, Novus Biologicals, 1:10,000) and secondary antibodies conjugated to Alexa 488, Cy3, or Cy5 (Jackson ImmunoResearch, West Grove, PA or Molecular Probes, Thermo-Fisher, Waltham, MA; 1:300). Imaging was performed with a Marianas spinning-disc confocal microscope system (Intelligent Imaging Innovations, Inc. [3i]) using a 100 × 1.46 NA oil objective, or with a DeltaVision Elite system (GE Healthcare, Pittsburgh, PA) equipped with an Olympus 100 × 1.4 NA oil-immersion objective. Wide-field images were deconvolved using the SoftWoRx Suite (GE Healthcare). 3D stacks were visualized in Imaris 7.3 or 8.0 (Bitplane).

37To evaluate the sensitivity of the SC to hexanediol in various mutants (Figure 4—figure supplement 2A), at least 3 slides with >10 gonads per slide were scored. The data are presented as the average value for a set of slides. Gonads were scored as ‘dissolved’ if no filaments were visible, other than short SC stretches on the X chromosome and/or within apoptotic nuclei.

38The following yeast strains were used:

39Brün292: MATa/alpha; ndt80::LEU2:ndt80::LEU2 flo8::KanMX6/flo8::KanMX6 ZIP1::GFP(700)/ ZIP1::GFP(700) (SK1 strain background)

40Brün793: MATa/alpha; GAL-NDT80::TRP1/GAL-NDT80::TRP1 ura3::pGPD1-GAL4(848).ER::URA3/ura3::pGPD1-GAL4(848).ER::URA3 leu2::URA3p-tetR-tdTomato::LEU2 CENV::tetOx224::HIS3 REC8-GFP-URA3/REC8-GFP-URA3 (SK1 strain background)

41Sporulation and pachytene arrest of ndt80 strains was was carried out as described. Where indicated, 1,6-hexanediol diluted in sporulation medium was added to cells immediately prior to imaging. Where indicated, cells were treated with various concentration of 1,6-hexanediol and KCl by mixing a 2x concentrated solution with cells washed 3 times with 20 mM phosphate buffer (pH 7). ‘Washing’ of hexanediol was performed by mixing the treated cells with 10x excess of 20 mM phosphate buffer (pH 7). Unfixed cells were imaged using a Marianas spinning-disc confocal microscope, as described above.

42To analyze the effects of hexanediol on Drosophila melanogaster SCs, ovaries from young, mated c(3)G-GFP; c(3)G/+ females fed on yeast paste were dissected in PBS, separated into ovarioles, and transferred to a drop of PBS in a glass-bottomed petri dish. 3D wide-field images were acquired using a DeltaVision microscope equipped with a 100 × 1.4 NA objective. An equal volume of 2x 1,6 hexanediol in PBS was added during imaging. For immunofluorescence, ovarioles from c(3)G-GFP; c(3)G/+ or from GAL4-nos/+; UASp-c(2)M-3xHA/+ females were dissected in Modified Robb’s Saline (MRS), incubated briefly with hexanediol in MRS (or just MRS) and fixed with 4% formaldehyde in MRS, following by immunostaining as described. The following antibodies were used: anti-GFP (1:400, mouse, Roche), anti-CID (1:500, chicken), anti-C(2)M (1:500, rabbit), anti-HA (mouse, 1:500), anti-C(3)G (Rabbit, 1:3000), and appropriate secondary antibodies conjugated to dyes (1:300). Samples were imaged on a Marianas spinning-disc confocal microscope, as described above.

43C.elegans adult hermaphrodites were high-pressure frozen and freeze-substituted according to previously described methods. Briefly, worms were frozen in 50 µm deep specimen carriers in a HPM-010 high pressure freezer (Bal-Tec, Balzers, Liechstenstein), freeze-substituted in 1% osmium tetroxide plus 0.1% uranyl acetate in acetone over 2.5 hr, and embedded in Epon-Araldite resin over a period of 2.5 hr. Serial sections (70 nm thick) were picked up on slot grids, post-stained with uranyl acetate and lead citrate, and imaged in Tecnai 12 transmission electron microscope operating at 120kV (FEI, Hillsboro, OR), using an Ultrascan 1000 charge coupled device camera (Gatan, Pleasanton, CA).

44All movies show maximum-intensity projection images from 3D time-lapse recordings of C. elegans gonads acquired with a spinning disk confocal microscope. In most cases the worms were intact and alive during recording; in experiments involving 1,6-hexanediol treatment, gonads were extruded by nicking the cuticle with a scalpel blade shortly before imaging began. Elapsed times are indicated at the bottom right corner of each frame as hours:minutes:seconds:(milliseconds).