PCH-2 collaborates with CMT-1 to proofread meiotic homolog interactions

The conserved ATPase, PCH-2/TRIP13, is required during both the spindle checkpoint and meiotic prophase. However, its specific role in regulating meiotic homolog pairing, synapsis and recombination has been enigmatic. Here, we report that this enzyme is required to proofread meiotic homolog interactions. We generated a mutant version of PCH-2 in C. elegans that binds ATP but cannot hydrolyze it: pch-2E253Q. In vitro, this mutant can bind a known substrate but is unable to remodel it. This mutation results in some non-homologous synapsis and impaired crossover assurance. Surprisingly, worms with a null mutation in PCH-2’s adapter protein, CMT-1, the ortholog of p31comet, localize PCH-2 to meiotic chromosomes, exhibit non-homologous synapsis and lose crossover assurance. The similarity in phenotypes between cmt-1 and pch-2E253Q mutants suggest that PCH-2 can bind its meiotic substrates in the absence of CMT-1, in contrast to its role during the spindle checkpoint, but requires its adapter to hydrolyze ATP and remodel them.


Introduction
Sexual reproduction relies on meiosis, the specialized cell division that generates haploid gametes, such as sperm and eggs, from diploid progenitors so that fertilization restores diploidy. To ensure that gametes inherit the correct number of chromosomes, meiotic chromosome segregation is exquisitely choreographed: Homologous chromosomes segregate in meiosis I and We previously showed that in the absence of PCH-2, homolog pairing, synapsis and recombination are accelerated and associated with an increase in meiotic defects, leading us to speculate that PCH-2 disassembles molecular intermediates between homologous chromosomes that underlie these events to ensure their fidelity [50]. Specifically, we hypothesized that these molecular intermediates involve meiotic HORMADs and PCH-2 remodels meiotic HOR-MADs to proofread these homolog interactions [50]. To test this model, we generated a mutant version of PCH-2 that binds ATP, and its substrates, but cannot hydrolyze ATP and therefore cannot remodel and release its substrates. We reasoned that if our hypothesis is correct, this mutant protein will "trap," and allow us to more easily visualize, inappropriate homolog interactions that are incorrectly stabilized. This is accurate: pch-2 E253Q mutants delay homolog pairing and accelerate synapsis, producing non-homologous synapsis. Meiotic DNA repair occurs with normal kinetics but crossover assurance is lost, suggesting that crossoverspecific recombination intermediates between incorrect partners may also be inappropriately stabilized. Surprisingly, loss of CMT-1, an adapter protein that is thought to be essential for PCH-2 to bind its substrates, phenotypically resembles pch-2 E253Q mutants. Since PCH-2 localizes normally to meiotic chromosomes in the absence of CMT-1, these data indicate that CMT-1 is dispensable for PCH-2 to bind its meiotic substrates on chromosomes but is essential to hydrolyze ATP and remodel its substrates.

pch-2 E253Q mutants delay pairing
Next, we analyzed pairing in syp-1 mutants. SYP-1 is a component of the synaptonemal complex (SC), and syp-1 mutants fail to assemble SC between paired homologs, allowing us to more easily visualize pairing intermediates in the absence of synapsis [6]. In C. elegans, cis-acting sites called pairing centers (PCs) are required for pairing and synapsis [54]. In the absence of synapsis, homologous chromosomes exhibit stable pairing of PC, but not non-PC, ends of chromosomes [54]. We monitored synapsis-independent pairing at PC sites by performing immunofluorescence against HIM-8, which binds PCs of X chromosomes [55] (Fig 2B). A single focus of HIM-8 indicates paired X chromosomes while two foci indicate unpaired X chromosomes. When we imaged stained germlines, we divided germlines into six equal sized zones (see cartoon in Fig 2A). Because meiotic nuclei are arranged spatiotemporally in the germline, this allows us to assess pairing as a function of meiotic progression.
syp-1 mutants initiated pairing in zone 2, maintained pairing at the X chromosome PC in zones 3,4 and 5 and destabilized pairing in zone 6 ( Fig 2C). As we previously reported [50], syp-1;pch-2Δ double mutants also initiated PC pairing in zone 2 but had significantly more nuclei with paired HIM-8 signals than syp-1 single mutants (p value < 0.0001, two-tailed Fisher's exact test), indicating that pairing is accelerated in this background. By contrast, syp-1; pch-2 E253Q double mutants had significantly fewer nuclei with paired HIM-8 signals in zone 2 than both syp-1 single (p value < 0.0001, two-tailed Fisher's exact test) and syp-1;pch-2Δ double mutants (p value < 0.0001, two-tailed Fisher's exact test). In later zones, all three genotypes closely resembled each other in that chromosomes are paired. However, in zone 4, significantly fewer X chromosomes were paired in syp-1; pch-2 E253Q double mutants than syp-1 single mutants (p value = 0.0003, two-tailed Fisher's exact test). Thus, syp-1; pch-2 E253Q mutants delay and exhibit defects in pairing at PCs, consistent with our model that PCH-2 proofreads homolog pairing intermediates: If pch-2 E253Q mutants "trap" incorrect pairing intermediates, between non-homologous chromosomes, for example, this would affect pairing between homologous chromosomes.
We reasoned that the delay in pairing, combined with the acceleration in synapsis, raised the possibility that these events, which are typically coupled, had been uncoupled. To test this possibility, we evaluated whether non-homologous synapsis occurs in pch-2 E253Q mutants. We stained wildtype, pch-2Δ and pch-2 E253Q mutant germlines with antibodies against SC components HTP-3 and SYP-1 to evaluate synapsis and HIM-8 to assess pairing. All nuclei in wildtype animals, pch-2Δ, and most nuclei in pch-2 E253Q mutants exhibited homologous synapsis (uncircled nuclei in Fig 3C). However, we identified meiotic nuclei in which all chromosomes were synapsed but contained two HIM-8 foci, indicating these chromosomes had synapsed with non-homologous partners (circled nuclei in Fig 3C). We also observed non-homologous synapsis when we monitored pairing of autosomes using antibodies against ZIM-2, a protein that binds an autosomal PC [56] (S1A Fig). When we quantified this defect, we detected nonhomologous synapsis of X chromosomes in 2.5% of meiotic nuclei ( Fig 3D) and non-homologous synapsis of Chromosome V in 6% of meiotic nuclei (S1B Fig). Unlike HIM-8 [55], ZIM-2 only stains meiotic chromosomes in a subset of meiotic nuclei [56]. Since fewer nuclei are positive for ZIM-2 than HIM-8 in all germlines, this explains the higher frequency of non-homologous synapsis that involves Chromosome V. C. elegans have six pairs of homologous chromosomes. Therefore, it's possible that as many as 18% of meiotic nuclei have at least one pair of chromosomes undergoing non-homologous synapsis in pch-2 E253Q mutants. However,   E253Q and pch-2Δ mutant germlines. C. Images of meiotic nuclei stained with antibodies against HTP-3, SYP-1 and HIM-8 in wildtype animals, pch-2 E253Q and pch-2Δ mutants. Circled nuclei have undergone non-homologous synapsis. D. Quantification of non-homologous synapsis wildtype animals, pch-2 E253Q and pch-2Δ mutants. All error bars indicate 95% confidence intervals. Significance was assessed by performing two-tailed Fisher exact tests. A ��� indicates a p value < 0.0001. it's also formally possible that some nuclei contain multiple pairs of chromosomes that are non-homologously synapsed and that 18% may be an overstatement. Unfortunately, we were unable to directly test this possibility due to technical limitations.

pch-2 E253Q mutants lose crossover assurance
We then monitored meiotic DNA repair and the formation of crossovers. To assess DNA repair, we monitor the appearance and disappearance of RAD-51 from meiotic chromosomes as a function of meiotic progression ( Fig 4A). RAD-51 is required for all meiotic DNA repair in C. elegans [57]. Its appearance on chromosomes signals the formation of double strand breaks (DSBs) that need to be repaired (see zones 3 and 4 in wildtype in Fig 4B) and its disappearance indicates the entry of DSBs into a DNA repair pathway (see zones 5 and 6 in wildtype in Fig 4B) [3]. The appearance and disappearance of RAD-51 on meiotic chromosomes in pch-2 E253Q mutants was indistinguishable from that in wildtype in zones 1-4. In zone 5, we detected fewer RAD-51 foci, suggesting that meiotic DNA repair may occur slightly more rapidly in pch-2 E253Q mutants (p value < 0.0001, two-tailed t-test). However, we observed an acceleration in DNA repair much earlier in pch-2Δ mutants (see zone 4, Fig 4B and [50] (p value < 0.0001, two-tailed t-test).
We assayed crossover formation in pch-2 E253Q mutants by visualizing GFP::COSA-1 foci formation in late meiotic prophase. GFP::COSA-1 cytologically marks putative crossovers and its appearance as robust foci is mechanistically associated with the process of crossover designation [58]. Almost all nuclei in wildtype worms had six GFP::COSA-1 foci per nucleus, one per homolog pair (Fig 4C, top, and 4D) [58]. In contrast, pch-2 E253Q mutants had a greater number of nuclei with five GFP::COSA-1 foci, similar to pch-2Δ mutants, indicating a defect in crossover assurance (Fig 4C, bottom, 4D and [50]). Given the in vitro behavior of the PCH-2 E253Q hexamer [53], this phenotype is likely because inappropriate crossover intermediates are stabilized at the expense of appropriate ones. Alternatively, these meiotic nuclei may result from the non-homologous synapsis we also observe in this mutant background. However, we do not observe a delay in meiotic DNA repair in pch-2 E253Q mutants ( Fig 4B). Further, pch-2 E253Q mutants do not activate feedback mechanisms, as visualized by the extension of DSB-1 on meiotic chromosomes [59,60] (S2A and S2B Fig). Therefore, we favor the first interpretation.

CMT-1 is required for the synapsis checkpoint
During the spindle checkpoint, PCH-2 requires the presence of an adapter protein, CMT-1 (p31 comet in mammalian cells), to bind and remodel a HORMAD protein required for the spindle checkpoint response, Mad2 [53,61,62]. Based on a recent report that the rice ortholog of CMT-1 is an SC component and required for pairing, synapsis and double strand break formation [63], we tested whether cmt-1 had a role in meiotic prophase by assessing whether it was required for meiotic prophase checkpoints in C. elegans. We introduced a null mutation of cmt-1 into syp-1 mutants. The absence of SC in syp-1 mutants results in unsynapsed chromosomes, which activate very high levels of germline apoptosis in response to both the synapsis and the DNA damage checkpoints [64] (Fig 5A). cmt-1 single mutants had wildtype levels of apoptosis (Fig 5B). In contrast to syp-1 single mutants, syp-1;cmt-1 double mutants had intermediate levels of germline apoptosis (Fig 5B), indicating that CMT-1 is required for either the synapsis or DNA damage checkpoint. To distinguish between these two checkpoints, we used syp-1;spo-11 double mutants, which do not generate double strand breaks [18]. As a result, these double mutants do not activate the DNA damage checkpoint ( Fig 5A) and produced an intermediate level of apoptosis ( Fig 5B). [64]. When we generate syp-1;spo-11;cmt-1 triple mutants, we observed wildtype levels of apoptosis, similar to but slightly higher than cmt-1 single mutants (Fig 5B) (p value < 0.05, two tailed t-test). Therefore, CMT-1 is required to activate apoptosis in response to the synapsis checkpoint, similar to pch-2Δ mutants.

cmt-1 mutants phenotypically resemble pch-2 E253Q mutants
Based on the requirement for CMT-1 in the synapsis checkpoint, we assessed pairing, synapsis, meiotic DNA repair and crossover formation in cmt-1 mutants. Similar to our analysis of pch-2 E253Q mutants, pairing was delayed in syp-1;cmt-1 double mutants (see zone 2, Fig 6A) (p value < 0.0001, two-tailed Fisher's exact test) and synapsis was accelerated in cmt-1 single mutants (see zone 3, Fig 6B) (p value < 0.0001, two-tailed Fisher's exact test). We also detected non-homologous synapsis at levels similar to that of pch-2 E253Q mutants (Figs 6C, 6D and S1). Meiotic DNA repair in cmt-1 mutants, as visualized by the appearance and disappearance of RAD-51 (S3A Fig), and meiotic progression, as visualized by the appearance and disappearance of a protein required for DSB formation, DSB-1 (S2A and S2B Fig), closely resembled that of wildtype germlines. However, we did observe a delay in RAD-51 removal in cmt-1 mutants (zones 5 and 6, p value < 0.0001, two-tailed Fisher's exact test). It is unclear whether this delay is biologically relevant, since we only observe a slight delay in DSB-1 removal (S2B Fig). In addition, crossover assurance was reduced (S3B Fig).

Non-homologous synapsis in cmt-1 mutants relies on PCH-2
We previously showed that CMT-1 is required for PCH-2 to localize to mitotic chromosomes during the spindle checkpoint [66]. We tested whether this was also true during meiotic prophase and found that CMT-1 was dispensable for PCH-2's localization to meiotic chromosomes ( Fig 7A, grayscale images in S4 Fig). Consistent with CMT-1 being required for PCH-2's hydrolysis of ATP and release of substrates, we observed that PCH-2 stayed on meiotic chromosomes slightly longer than in wildtype germlines (Fig 7C).
During mitotic divisions in the C. elegans embryo, loss of CMT-1 partially suppresses the spindle checkpoint defect observed in pch-2 mutants [66]. Since PCH-2 ensures availability of the inactive conformer of Mad2 so that it can be converted to the active one during checkpoint activation [67,68], we hypothesize that this genetic interaction results because loss of CMT-1 makes more inactive Mad2 available. In this way, PCH-2 and CMT-1 would control the spindle checkpoint by regulating available, soluble pools of inactive and active Mad2. We  wondered if there was a similar relationship between PCH-2 and CMT-1 in meiotic prophase.
To test this, we generated cmt-1;pch-2Δ double mutants and assayed non-homologous synapsis. Unlike cmt-1 single mutants and similar to pch-2Δ mutants, we did not detect non-homologous synapsis in cmt-1;pch-2Δ double mutants (Fig 6C and 6D). Thus, PCH-2 is epistatic to CMT-1 during meiosis, at least in the context of non-homologous synapsis, supporting our interpretation that PCH-2 can bind its meiotic substrates independent of CMT-1 and suggesting that some of PCH-2's regulation of its meiotic substrates occurs directly on chromosomes.

Discussion
In vitro, PCH-2 E253Q protein binds ATP and its spindle checkpoint substrate, Mad2, but fails to remodel it, due to an inability to hydrolyze ATP [53]. Here, we show that the meiotic phenotypes observed in pch-2 E253Q mutants are consistent with a role for PCH-2 in disassembling inappropriate pairing, synapsis and crossover recombination intermediates in C. elegans, resulting in non-homologous synapsis (Figs 3C, 3D and S1) and a loss of crossover assurance (Fig 4D). The effect we observe on these meiotic prophase events seems limited, affecting a small proportion of nuclei, but this may reflect either the inherent fidelity of these processes or the existence of redundant mechanisms that contribute to fidelity in C. elegans. Further, a role for PCH-2 in proofreading may be more apparent in organisms or situations in which homolog interactions are more challenging. Indeed, PCH-2's involvement in the interchromosomal effect in Drosophila [69], in which chromosome rearrangements affect crossover control on other chromosomes, supports this possibility. Nevertheless, our results demonstrate that PCH-2 proofreads interactions between homologous chromosomes to ensure that they are correct. We previously showed that PCH-2 decelerates homolog pairing, synapsis and recombination, coordinating these meiotic prophase events [50]. Here, we argue that PCH-2 acts directly on chromosomes to proofread meiotic interhomolog interactions and ensure their fidelity. This role is supported by several pieces of data. First, PCH-2 localizes to meiotic chromosomes, both before and after synapsis [50]. This localization is functional, as demonstrated by its persistence on the synaptonemal complex and correlation with an extension of homolog access in mutants defective in recombination [50,70]. Second, despite the defects we report, pch-2Δ, pch-2 E253Q and cmt-1 mutants do not show acceleration of meiotic progression, as assayed by SUN-1 phosphorylation [50], appearance of transition zone nuclei (Figs 1 and 7 and [50]), ZIM-2 (S1 Fig) or DSB-1 localization (S2 Fig), indicating that PCH-2 and CMT-1 do not coordinate these meiotic prophase events indirectly through misregulation of meiotic progression. Finally, meiotic defects associated with loss of CMT-1 rely on PCH-2 function (Fig 6D), unlike what we observe in mitosis [66]. In mitosis, CMT-1 is epistatic to PCH-2, which we attribute to misregulation of soluble pools of Mad2. Soluble pools of at least one meiotic HORMAD, HTP-1, have been implicated in ensuring both accurate homolog synapsis and proper meiotic progression [23]. If PCH-2 and CMT-1 were regulating homolog pairing, synapsis and recombination via control of meiotic progression, we might predict that CMT-1 would be epistatic to PCH-2 in meiosis, as it is in mitosis. Taken together, these data favor a model in which PCH-2 and CMT-1 proofread interactions between homologous chromosomes that underlie pairing, synapsis and recombination in C. elegans. Since pairing, synapsis and recombination depend on meiotic HORMADs adopting their closed conformations [45], we hypothesize that these interactions between homologous chromosomes either involve or are stabilized by closed versions of meiotic HORMADs and PCH-2, in collaboration with CMT-1, remodels them to disassemble or destabilize them. In systems such as plants, yeast and mice, we propose that PCH-2 plays a similar role in addition to removing or relocalizing meiotic HORMADs to signal meiotic progression [28,[30][31][32].
We were surprised to see that PCH-2 localized to meiotic chromosomes in cmt-1 mutants (Fig 7), unlike what we observe during the spindle checkpoint response [66], and that cmt-1 mutants closely resembled pch-2 E253Q mutants in exhibiting non-homologous synapsis and a loss of crossover assurance (Figs 6 and S2). Since PCH-2's localization to meiotic chromosomes depends on synapsis [50], this localization may not accurately represent interaction with its proposed meiotic substrates, meiotic HORMADs. Alternatively, these data may suggest that PCH-2 is competent to bind meiotic HORMADs in the absence of CMT-1 and that CMT-1's meiotic function is to promote PCH-2's ATPase activity and ability to remodel these substrates. This is unlike its role in the spindle checkpoint, in which CMT-1/p31 comet is essential for PCH-2 to bind its substrate, Mad2 [53,61,62,71]. This difference raises the possibility that the interaction between PCH-2 and meiotic HORMADs during their remodeling is substantially different than that with Mad2. This hypothesis is supported by the findings that budding yeast, in which PCH2 was originally identified [72], does not appear to have a CMT-1/p31 comet ortholog [73,74] and budding yeast Pch2 can directly interact with and remodel the budding yeast meiotic HORMAD, Hop1, in vitro [29]. However, another possibility is that PCH-2's ability to remodel meiotic HORMADs exists in two modes: 1) a mode that proofreads meiotic interhomolog interactions during pairing, synapsis and recombination and depends on CMT-1 for ATP hydrolysis, but not substrate binding; and 2) one that removes meiotic HORMADs during meiotic progression which relies on CMT-1 for both substrate recognition and ATP hydrolysis. Such a model raises the immediate question of how these two modes might be regulated in an organism with both.
The PCH-2/HORMAD genetic module is evolutionarily ancient [75,76], having been identified as operons in several bacteria [77], suggesting that this pair of proteins has been coopted to function in multiple molecular contexts. While we can easily detect an interaction between PCH-2 and CMT-1, and CMT-1 and Mad2, by two-hybrid experiments [66], we have not been able to observe a similar interaction between PCH-2 or CMT-1 and any of the four meiotic HORMADs present in C. elegans, HTP-3, HIM-3, HTP-1 and HTP-2, making it difficult to identify what regions of these proteins are required to interact with PCH-2. Genetic mutations, combined with cytological analysis, seems the most straightforward way, at least in C. elegans, to understand whether and how PCH-2 and meiotic HORMADs interact to regulate meiotic pairing, synapsis and recombination.
All images were acquired using a DeltaVision Personal DV system (Applied Precision) equipped with a 100X N.A. 1.40 oil-immersion objective (Olympus), resulting in an effective XY pixel spacing of 0.064 or 0.040 μm. Three-dimensional image stacks were collected at 0.2μm Z-spacing and processed by constrained, iterative deconvolution. Image scaling and analysis were performed using functions in the softWoRx software package. Projections were calculated by a maximum intensity algorithm. Composite images were assembled and some false coloring was performed with Adobe Photoshop.
Scoring of germline apoptosis was performed as previously described in [64] in strains containing bcIs39 [lim-7p::ced-1::GFP + lin-15(+)] with the following exceptions. L4 hermaphrodites were allowed to age for 22 hours. They were then mounted under coverslips on 1.5% agarose pads containing 0.2mM levamisole and scored. A minimum of twenty-five germlines was analyzed for each genotype.
Quantification of pairing, synapsis, RAD-51 foci, GFP::COSA-1 foci, and DSB-1 positive nuclei was performed on animals 24 hours post L4 stage and with a minimum of three germlines per genotype. Relevant statistical analysis, as indicated in the Figure Legends, was used to assess significance. The number of nuclei assayed for each genotype for all figures is shown in S1 Table. Supporting information S1 Fig. Autosomes undergo non-homologous synapsis in pch-2 E253Q and cmt-1 mutants. A. Images of meiotic nuclei stained with antibodies against HTP-3, SYP-1 and ZIM-2 in wildtype animals, pch-2 E253Q and cmt-1 mutants. Circled nuclei have undergone non-homologous synapsis. B. Quantification of non-homologous synapsis wildtype animals, pch-2 E253Q and cmt-1 mutants. Error bars indicate 95% confidence intervals. Significance was assessed by performing two-tailed Fisher exact tests. A ��� indicates a p value < 0.0001.