A compartmentalized, self-extinguishing signaling network mediates crossover control in meiosis

Meiotic recombination between homologous chromosomes is tightly regulated to ensure proper chromosome segregation. Each chromosome pair typically undergoes at least one crossover event (crossover assurance) but these exchanges are also strictly limited in number and widely spaced along chromosomes (crossover interference). This has implied the existence of chromosome-wide signals that regulate crossovers, but their molecular basis remains mysterious. Here we characterize a family of four related RING finger proteins in C. elegans. These proteins are recruited to the synaptonemal complex between paired homologs, where they act as two heterodimeric complexes, likely as E3 ubiquitin ligases. Genetic and cytological analysis reveals that they act with additional components to create a self-extinguishing circuit that controls crossover designation and maturation. These proteins also act at the top of a hierarchical chromosome remodeling process that enables crossovers to direct stepwise segregation. Work in diverse phyla indicates that related mechanisms mediate crossover control across eukaryotes.


34
Meiosis gives rise to haploid gametes through two sequential rounds of nuclear 35 division. To ensure faithful meiotic chromosome segregation in most organisms, every pair of 36 homologous chromosomes must attain at least one crossover (CO) recombination product, 37 which creates a stable interhomolog connection known as a chiasma (Page and Hawley, 2003). 38 However, the total number of COs per cell is typically far too low to assure CO formation on  (Dernburg et al., 1998;Keeney, 2008;Keeney et al., 1997). A subset of DSBs are processed to 48 become COs, while the rest are repaired through alternate pathways. In several model 49 organisms, two CO pathways have been elucidated: The "class I" pathway, which is subject to 50 CO assurance and interference, requires meiosis-specific homologs of the bacterial mismatch 51 repair protein MutS (Msh4 and Msh5). Components of the synaptonemal complex (SC), an 52 ordered, periodic proteinaceous structure that assembles between homologous chromosomes 53 during meiotic prophase, are also required for class I COs. An alternate enzymatic pathway can 54 give rise to both noncrossovers and "class II" COs, which do not show interference. To investigate the localization of the ZHP proteins, we inserted epitope tags at the C-129 terminus of each coding sequence; these tagged alleles supported normal meiosis 130 (Supplemental Table S1). All four of the proteins localized to the synaptonemal complex (SC; prophase to mid-pachynema, they became brighter and more contiguous along the length of 137 SCs, but did not appear completely uniform throughout this structure. Upon the appearance of 138 GFP-COSA-1 foci, which mark designated CO sites from mid-pachynema through diplonema 139 (Yokoo et al., 2012), ZHP-1/2 became confined to the SC on one side of each CO. Intriguingly, 140 this restriction was observed in pachytene nuclei that retained SC and the HORMA domain 141 proteins HTP-1/2 on both sides of the CO, and represents the earliest known molecular 142 differentiation of the two chromosome domains that will become the long and short arms of 143 the bivalent. ZHP-1 and -2 remained associated with SC proteins along the short arm of each 144 bivalent as the SC disappeared from the long arm, and persisted as long as SC proteins were 145 present along the short arms, through late diakinesis. 146 ZHP-3 and -4 exhibited a similar but distinct distribution ( Figure 1C-D). These proteins 147 also localized to polycomplexes prior to synapsis, then became discontinuously distributed 148 along the length of SCs in early to mid-pachynema. Upon the appearance of GFP-COSA-1 foci, 149 ZHP-3 and -4 both concentrated at these designated CO sites and gradually disappeared from  Association of ZHPs with meiotic chromosomes depends on synaptonemal complexes 155 The SC in C. elegans comprises at least 4 proteins, known as SYP-1-4, which are 156 mutually dependent for its self-assembly. We found that none of the four ZHP proteins  Because the ZHP proteins relocalized upon CO designation (Figure 1), we examined their 166 distribution under conditions where COs fail to occur. In mutants lacking SPO-11, MSH-5, or 167 COSA-1, ZHPs were detected along the length of SCs through diplonema (data not shown). 168 Thus, relocalization of the ZHPs in late prophase depends on CO designation.

170
ZHPs act as two heterodimeric complexes 171 To investigate the meiotic functions of the ZHP proteins, we first exploited the auxin-172 inducible degradation system. Each of the ZHP genes was tagged with a 3xFLAG epitope and a 173 44-aa degron sequence (hereafter "AID") in a strain that expresses the F-box protein AtTIR1  To determine the dependence of the ZHP proteins on each other for their localization 185 and stability, we exposed worms to auxin for 24 hours, so that a pool of nuclei had entered and  and ZHP-3 levels were somewhat lower when ZHP-4 was depleted ( Figure 2F).

192
Other RING finger proteins such as BRCA1/BARD1 are known to act as obligate 193 heterodimers (Brzovic et al., 2001;Metzger et al., 2014;Wu et al., 1996). The similarity and 194 interdependence of localization of ZHP-1 and -2, and of ZHP-3 and -4, suggested that these 195 four proteins might function as two pairs. Using yeast two-hybrid analysis, we found that ZHP- normally maintained by chiasmata, and six condensed bivalents can be detected in each 206 oocyte nucleus. We exposed zhp-AID strains to auxin for 24 hours and then analyzed chiasma

212
Because a few chiasmata were observed when ZHP-1-AID or ZHP-2-AID were depleted, 213 we were concerned that these proteins might not be efficiently degraded. We therefore 214 engineered null mutations in zhp-1 and zhp-2 (see Materials and Methods). Homozygous 215 mutants produced few viable progeny and a high incidence of male self-progeny ( Figure 3E), as 216 expected for important meiotic factors (Hodgkin et al., 1979). They also showed identical 217 distributions of DAPI-staining bodies at diakinesis to that observed in AID-depleted animals 218 (Figure 3-figure supplement 1A-C and data not shown). Additionally, both zhp-1 and -2 null 219 mutants produced more viable progeny than zhp-3 null mutants ( Figure 3E), consistent with a 220 less absolute dependence of chiasma formation on ZHP-1/2 than on ZHP-3/4. 221 We wondered whether the few chiasmata in hermaphrodites lacking ZHP-1/2 might 222 preferentially occur on the X chromosome, which shows some differences from autosomes in 223 its genetic requirements for DSB induction and CO formation (Yu et al., 2016). To investigate 224 this, we used fluorescence in situ hybridization (FISH) to mark Chromosome V and the X 225 chromosome in zhp-1 mutants. By examining nuclei at diakinesis, we found that chiasma   Thus, the crossover assurance checkpoint is triggered, presumably by CO defects, indicating 245 that the ZHPs are dispensable for this feedback mechanism. 246 We noted differences in RAD-51 dynamics when ZHP-1/2 were depleted as compared 247 to ZHP-3/4 (Figure 3-figure supplement 1G-J). In the absence of ZHP-3/4, RAD-51 foci reached 248 higher peak numbers than in wild-type animals or in the absence of ZHP-1/2, but depletion of 249 ZHP-1/2 caused a greater delay in the appearance and disappearance of RAD-51 foci than 250 depletion of ZHP-3/4. Thus, the two ZHP complexes likely play distinct roles in the processing 251 of recombination intermediates.   To test whether dim GFP-COSA-1 foci in ZHP-1/2-depleted worms mark recombination 273 intermediates, we inhibited DSB formation. We AID-tagged SPO-11 and either ZHP-1 or ZHP-2 274 in the same strain, and found that both proteins could be effectively co-depleted by auxin detected when ZHP-1 and -4 were co-depleted ( Figure 4G). 281 We next sought to understand why oocytes lacking ZHP-1/2 have abundant, dim 282 COSA-1 foci but few bivalent chromosomes. This discrepancy could indicate that the dim foci 283 do not mature as COs, or that they do become COs but fail to give rise to stable chiasmata. To 284 address this, we mapped COs in zhp-1 null mutants by whole genome sequencing (see 285 Materials and Methods). This analysis indicated that COs are strongly reduced in the absence 286 of ZHP-1. We detected an average of 2.8 ± 1.8 (SD) per oocyte in zhp-1 mutants, compared to 287 7.3 ± 3.0 (SD) per oocyte (slightly more than the expected number of 6) in wild-type animals 288 ( Figure 4H). This is consistent with the number of bivalents seen in zhp-1/2 mutants (Figure 3  As another approach to examine how the number of intermediates impacts CO 311 designation in the absence of ZHP-1/2, we co-depleted ZHP-2 and SPO-11 to abolish 312 programmed meiotic DSBs, and exposed these animals to ionizing radiation to induce varying 313 numbers of DSBs. When SPO-11 alone was depleted by the AID system, GFP-COSA-1 foci and 314 bivalents were eliminated. Following 10 Gy of radiation, a dose sufficient to ensure CO COSA-1 foci were observed following low doses of radiation, but these never exceeded 2 per 319 nucleus ( Figure 5D). At 10 Gy, only extremely dim GFP-COSA-1 foci were detected ( Figure 5D).

320
Taken together, these observations indicate that ZHP-1/2 are dispensable for a single 321 break to become a functional interhomolog CO, but act to limit or focus the activity of ZHP-3/4 322 and/or other CO factors to ensure robust CO designation in the context of excess DSBs.

323
However, ZHP-1/2 also contribute to CO formation even when very few breaks are made. A 324 possible interpretation is that ZHP-1/2 stabilize a factor that becomes limiting for CO 325 formation in their absence, particularly when there are abundant recombination 326 intermediates. 327 We next tested whether the function of ZHP-1/2 in limiting designated COs could be   To further probe the contribution of ZHP-1/2 to remodeling, we examined dsb-2 353 mutants, since some COs were robustly designated in this background in the absence of ZHP-  We also observed that PLK-2, which plays a poorly characterized role in remodeling (Harper et

380
Because ZHP-3/4 are strictly required for CO designation, we could not analyze their 381 roles in chromosome remodeling. However, prior work has found that a GFP-tagged ZHP-3 382 transgene acts as a separation-of-function allele that is proficient for CO designation but 383 defective for remodeling (Bhalla et al., 2008).  ZHP-1/2 promote the removal of ZHP-3/4 from these compartments. We also found that MSH-423 5 colocalized with ZHP-3/4 and COSA-1 foci ( Figure 7D). Additionally, PLK-2 localized 424 throughout polycomplexes even prior to the appearance of COSA-1 foci, but became 425 noticeably brighter afterwards, in the presence or absence of ZHP-1/2 ( Figure 7E-F).

426
Together with the observations of dsb-2 mutants described above, these findings 427 demonstrate that key aspects of the regulatory circuit that designates and constrains COs act 428 autonomously within each contiguous SC or polycomplex ( Figure 7G). Remarkably, this circuit 429 is triggered within polycomplexes even in the absence of CO intermediates. These findings 430 support the idea that CO regulation is mediated through this liquid crystalline medium, which   Our analysis also reveals that the C. elegans ZHP proteins function as heterodimers, or 494 perhaps as higher order oligomers. It will be important to determine whether ZHP homologs in 495 other organisms also have obligate partners, or perhaps homooligomerize, since this 496 knowledge may help to illuminate their in vivo activities and substrates. RNF212B, which 497 encodes a mammalian paralog of RNF212, was recently identified as a candidate modifier of 498 meiotic recombination in cattle (Kadri et al., 2016), suggesting that it may act as a partner for 499 RNF212. Multiple protein isoforms are also predicted from the RNF212 locus, which could act 500 together as dimers or higher-order oligomers.  heterochromatin and sometimes at centromere cores, and Shugoshin/MEI-S332, a Rec8-513 specific protective factor, is also recruited to these regions. In principle, programmed release of 514 cohesion along arms and its retention near centromeres could suffice to allow homologs to 515 segregate in MI while maintaining a link between sisters until MII. However, experimental 516 evidence from several organisms indicates that CO formation leads to a local disruption of 517 cohesion, which can predispose chromosomes to missegregate when COs occur close to 518 centromeres (Brar and Amon, 2008). In light of other similarities between ZHP-1/2 and HEI10, 519 this raises the possibility that HEI10 may also influence meiotic cohesion, perhaps via spatial  Table S2 and Table S3. Unless  Injected animals were transferred to individual plates and incubated at 20˚C. After 4 576 days, F1 Rol progeny (dpy-10 heterozygotes) or Dumpy animals (dpy-10 homozygotes) from 577 "jackpot" broods containing multiple Rol/Dpy animals were picked to individual plates, allowed 578 to produce self-progeny, then lysed and screened by PCR.

579
Disruption of the zhp-1 gene was accomplished by inserting the sequence TAAGCTCGAG 580 after the 6 th codon, introducing a stop codon, a frameshift, and an Xho I restriction site. Three 581 independent alleles of the same sequence were generated: ie43 and ie44 in the N2 strain 582 background, and ie46 in the CB4856 strain. The zhp-2 gene was similarly disrupted by insertion 583 of TAATAATTAATTAG after the 7 th codon, resulting in multiple stop codons and a frameshift.

584
All C. elegans strains were maintained on nematode growth medium (NGM) plates 585 seeded with OP50 bacteria at 20°C. See Table S4 for a full list of strains used. Unless otherwise 586 indicated, young adults were used for both immunofluorescence and western blotting assays.

587
Synchronized young adults were obtained by picking L4 larvae and culturing them for 20-24 h 588 at 20°C.  Briefly, NGM agar was supplemented with 1mM IAA just before pouring plates. E. coli OP50 594 bacteria cultures were concentrated and spread on plates, which were allowed to dry before 595 transferring C. elegans onto the plates. 596 We note that two major advantages of the AID system have facilitated our study of these 597 proteins in meiosis: i) this approach enables temporal control and germline-specific degradation 598 of proteins; ii) complex strains that combine various mutations and epitope-tagged alleles can 599 be constructed far more easily, since balancer chromosomes are not required to maintain 600 conditional alleles of genes essential for reproduction. Additionally, control experiments can be 601 performed using the same strains without auxin treatment.       To quantify western blots, TIF images were collected for each blot using a Chemidoc 693 system (Bio-Rad). Integrated intensities for relevant bands were then calculated using ImageJ.

694
To normalize for sample loading, the indicated band intensity was divided by the 695 corresponding α-tubulin or HTP-3 band intensity. Each normalized band intensity was 696 expressed as the percentage of the intensity at t=0 or in a control sample. at UC Berkeley, supported by NIH S10 Instrumentation Grants S10RR029668 and 709 S10RR027303. We are grateful to Ed Ralston for assistance with the irradiator and to Anne