CRL4 regulates recombination and synaptonemal complex aggregation in the Caenorhabditis elegans germline

To maintain the integrity of the genome, meiotic DNA double strand breaks (DSBs) need to form by the meiosis-specific nuclease Spo11 and be repaired by homologous recombination. One class of products formed by recombination are crossovers, which are required for proper chromosome segregation in the first meiotic division. The synaptonemal complex (SC) is a protein structure that connects homologous chromosomes during meiotic prophase I. The proper assembly of the SC is important for recombination, crossover formation, and the subsequent chromosome segregation. Here we identify the components of Cullin RING E3 ubiquitin ligase 4 (CRL4) that play a role in SC assembly in Caenorhabditis elegans. Mutants of the CRL4 complex (cul-4, ddb-1, and gad-1) show defects in SC assembly manifested in the formation of polycomplexes (PCs), impaired progression of meiotic recombination, and reduction in crossover numbers. PCs that are formed in cul-4 mutants lack the mobile properties of wild type SC, but are likely not a direct target of ubiquitination. In C. elegans, SC assembly does not require recombination and there is no evidence that PC formation is regulated by recombination as well. However, in one cul-4 mutant PC formation is dependent upon early meiotic recombination, indicating that proper assembly of the SC can be diminished by recombination in some scenarios. Lastly, our studies suggest that CUL-4 deregulation leads to transposition of the Tc3 transposable element, and defects in formation of SPO-11-mediated DSBs. Our studies highlight previously unknown functions of CRL4 in C. elegans meiosis and show that CUL-4 likely plays multiple roles in meiosis that are essential for maintaining genome integrity.

Introduction SC was shown to be regulated by SYP-4 phosphorylation [23]. While the processes controlling SC dynamics may be related to SC assembly, so far recombination appears to play no role in SC assembly; in the absence of SPO-11, or any protein involved in recombination, the SC assembles normally as in wild type. One exception is observed in the cra-1 mutant background in which SC assembly defects become more severe in the absence of early meiotic recombination intermediates [25]. How, and under what conditions, recombination affects SC assembly in C. elegans, is not yet understood.
The post-translational modifier, ubiquitin, is known to play an important part in meiosis in several organisms (reviewed in: [26,27]). Ubiquitination involves a protein cascade; the E1 enzyme activates ubiquitin so the E2 conjugation enzyme can transfer ubiquitin to the E3 ubiquitin ligases. E3 enzymes then attach ubiquitin to target proteins ( Fig 1A). The most wellknown function of ubiquitination is targeted degradation by the proteasome. Loss of some E3 activity or inhibition of the proteasome in meiosis can lead to delayed meiotic entry (C. elegans, and mice; [28,29]), improper meiotic recombination (yeast, C. elegans, and mice; [30][31][32]), and perturbed assembly of SC central elements (C. elegans; [33,34]).
One of the major families of ubiquitin ligases are the Cullin Ring E3 ubiquitin ligases (CRL) which are activated by neddylation. The activity of CRLs are controlled by deneddylation through the CSN/Cop9 Signalosome. One particular CRL, Cullin RING E3 ubiquitin ligase 4 (CRL4), has been shown to play an important role in the processes required for proper DNA function and maintenance (e.g. DNA repair, histone deposition, and replication (reviewed in: [35]). CRL4 is typically composed of 4 subunits; a Cullin scaffold protein, a RING box protein (Rbx) that binds the E2 enzyme, an adaptor protein that binds the ubiquitin-targeted substrate, and Ddb1 which is a mediator between the cullin scaffold and the adaptor protein. While Cul4, Ddb1, and Rbx are found in all CRL4s, the adaptor protein varies and is specific to a few substrates (reviewed in: [36]). In C. elegans, one known role of CRL4 is in the prevention of rereplication during S-phase in the soma [37][38][39], but not in the germline [40]. In meiosis, deletion of Cul4A in mice leads to defects in recombination and increased apoptosis [29,41], and in Arabidopsis Cul4 mutants affect the distribution of crossovers [42]. We previously identified CUL-4 as a member of the SC assembly pathway in C. elegans meiosis, and CRL4 is a plausible target of the CSN/Cop9 signalosome [34]. In wild type C. elegans, recombination depends on SC assembly but not vice versa [6,22]. Therefore, we suggested that the defects in recombination identified in this mutant are due to SC assembly defects [34]. The role of other components of CRL4 on SC assembly were unknown, nor the identity of the adaptor protein.
In this study, we determined novel roles for the CUL-4 E3 ubiquitin ligase complex in meiotic prophase I by examining oogenesis in the model organism C. elegans.
We examined four proteins (CUL-4, DDB-1, GAD-1, and RBXs) that were part of the CRL4 E3 ligase complex to determine if they were involved with SC assembly and/or meiotic recombination. We show that in C. elegans, the CRL4 E3 ligase plays a role in meiotic homologous recombination. Surprisingly, SC proteins are affected by meiotic recombination in one cul-4 mutant, as inactivation of proteins involved in meiotic recombination prevents PC formation in this cul-4 mutant. CUL-4 also appears to have a role in some TE silencing leading to another source of DNA damage. Our studies suggest that CUL-4 performs a conserved role in meiosis in DSB repair and reveal a novel effect of recombination on SC assembly behavior.

SC assembly is perturbed in CRL4 mutants
We have shown previously that the CUL-4 scaffold of the CRL4 ubiquitin ligase is required for proper SC assembly in C. elegans [34]. The CUL-4 RING E3 Ligase (CRL4) complex is ligase complex and its components examined in this study, based on what is known about its biochemical activity and composition from studies in other systems. The complex is presented in two states: active (neddylated/upper) and inactivated (deneddylated/lower), the active form is depicted before and after ubiquitination of the substrate. Gray subunits represent deletion mutations that do not result in a phenotype (RBX-1/2) or show minor phenotype (CAND-composed of 4 subunits, some of which vary between organisms and processes [35]. In C. elegans, the CRL4 complex is made up of four main components: CUL-4, the main scaffolding protein for the complex, a RING box protein that interacts with the E2 conjugating enzyme for targeted ubiquitination, DDB-1 which binds CUL-4 so that an adaptor protein can bind to the complex, and an adaptor protein (DDB1-Cul4-associated factor, DCAF) that recognizes target proteins for ubiquitination ( Fig 1A). We aimed to identify proteins that were part of the CRL4-mediated, SC assembly pathway by focusing on candidates that were part of the CRL4 in other cellular functions, as well as through a targeted RNAi screen (described in Materials and Methods). CUL4, DDB1 and RBX homologs have been previously found in C. elegans. CUL-4 and DDB-1 were shown to have a conserved role in mitotic replication and physically interact with each other [40,43]. RBX-1 acts with other CRL complexes found in C. elegans, but has not been shown to act within CRL4 [44]. To identify the adaptor protein we performed an RNAi screen of genes encoding for proteins containing a WD40-DDB1 domain (see below).
To examine SC assembly dynamics, immunofluorescent staining was performed on dissected C. elegans germlines (Fig 1B and S1 Fig). Nuclei in the germlines are positioned in a sequential manner allowing for a time course analysis of early meiotic events. Based on the timing of HTP-3 and SYP-1 localization we can conclude that meiotic entry occurs in the genotypes tested (S1K-N and [34]). To properly compare dynamics of SC assembly, length measurements of wild type, csn-5(ok1064), and cul-4(ok1891) mutant germlines were taken. Both cul-4(ok1891) and csn-5(ok1064) germlines were significantly shorter than wild type germlines (cul-4(ok1891): p<0.0001, csn-5(ok1064): p = 0.0007, Mann-Whitney, Fig 1C). These analyses were used as a basis to define the progression of meiotic prophase I, in that each zone, or image, represents a different portion of the pachytene time-course (description in Materials and Methods). In wild type germlines, Zones 1 and 2 correspond to the pre-meiotic tip (PMT) of the germline where mitotic proliferation occurs prior to meiotic entrance. In cul-4(ok1891) and csn-5(ok1064) mutants, as well as other CRL4 mutants, Zone 1 and the initial portion (approximately 1/3) of Zone 2 represent the PMT. In wild type, Zones 3-7 progressively represent meiotic prophase I. Zone 3 includes leptotene/zygotene, where SC assembly initiates and SPO-11 dependent DNA double strand breaks (DSBs) occur; Zones 4-7 represent pachytene, where SC assembly is completed and meiotic recombination is finished. Based on the timing of HTP-3 and SYP-1 localization to chromosomes in cul-4(ok1891) and csn-5 1) B) Full germline images of wild type (top) and cul-4(ok1891) mutants (bottom); blue (DAPI) and red (SYP-1). White lines are the border between the image and a black background that was added so a rectangular shape will be created. Scale bars are 20μm. C) Graphical comparison of germline length in wild type, csn-5(ok1064), and cul-4 (ok1891) mutants (Mann-Whitney; p-values, � < 0.001). D-J) Graphical analyses of SYP-1 localization analysis ("No SYP-1" was defined as having no SYP-1 immunofluorescence present along chromosomes (DAPI). "SYP-1 PC" were nuclei with only PC formation present, no elongation of SYP-1 along DAPI. "SYP-1 PC and Some Linear" was the presence of PC(s) in the nucleus but also partial elongation of SYP-1, <50% of DAPI. "SYP-1 PC and Linear" was similar to "SYP-1 PC and Some Linear" with the exception that SYP-1 is elongated along >50% of DAPI. "SYP-1 Partial Linear" nuclei had elongated SYP-1 along up to 50% of DAPI. "SYP-1 Mostly Linear" nuclei had SYP-1 along up to 50% of DAPI but less than 100%. "SYP-1 Linear" nuclei had fully elongated SYP-1 along all DAPI. "Other" was defined as nuclei that had abnormal DAPI appearance.); statistical comparisons of each mutant were made against wild type (Fischer's exact test; p-values, � < 0.001). Scale bars are 2μm. K) Analysis of percentage of germlines with PC formation within each individual genotype. The number of PCs per germline were grouped into three main categories: No PCs = germlines with no SC PC formation, Few PCs = germlines no more than 5 PCs or less than ½ of a zone with PCs, and Abundant PCs = germlines with more than 5 PCs or greater than ½ of a zone with PCs. Statistical comparisons were made to wild type for each mutant individually (Fisher's exact test; p-values, � < 0.0001). L) Analysis of homologous and non-homologous pairing by FISH to the 5S locus coupled with SYP-1 immunostaining in pachytene nuclei. In wild type the majority of nuclei are paired and synapsed (green) while in cul-4(ok1891) mutants less nuclei are paired. Unpaired FISH foci frequently associated with SYP-1 on at least one locus (purple and red) indicating non-homologous synapsis occurs in cul-4(ok1891) mutants regardless if the nuclei contain or do not contain PC. cul-4(ok1891) nuclei with PCs also contain FISH singnals that do not associate with SC (black and gray).
SC assembly can be observed by staining with SYP-1, a component of the central region of the SC. In wild type, SYP-1 is absent in PMT nuclei. Upon meiotic entrance, SYP-1 begins to assemble (in a linear pattern) along homologous chromosomes and becomes fully assembled along homologs in early pachytene (Fig 1D; S1A Fig). In cul-4(ok1891) mutants, large polycomplexes (PCs) are formed, which are defined as at least twice the width of typical SYP-1 structures [34] found in wild type nuclei [ Fig 1D vs. 1E; S1B Fig; PCs (purple, pink and red categories) are absent in wild type but present in~half of meiotic prophase I nuclei of cul-4 (ok1891) mutants]. These PCs may be due to SC assembly defects suggesting a role for CRL4 in SC assembly, consistent with [34]. Although PCs were abundant in cul-4(ok1891) mutants, nuclei frequently showed linear SC even in the presence of PCs. It is possible that this SC would not form between homologs, leading to non-homologous synapsis. Analysis of SYP-1 combined with FISH showed that most SC assembled between homologs ( Fig 1L). However, about 25% of cul-4(ok1891) mutant nuclei showed non-homologous synapsis of at least one chromosome, regardless whether they exhibit PCs or only linear SC ( Fig 1L).
Only 64% (n = 23/36) of cul-4(ok1891) mutant germlines have PCs ( Fig 1K). This incomplete penetrance of cul-4(ok1891) is observed with a 6X outcrossed line and is independent of the balancer used, indicating that the incomplete penetrance phenotype is specific to cul-4 (ok1891). We monitored the incomplete penetrance another way to determine if the phenotype could be observed in a single worm or if the phenotype differed between individuals. Using a gfp::syp-3 construct, SC assembly was monitored in the two gonadal arms of each worm allowing for a thorough analysis of incomplete penetrance. When gfp::syp-3 was present in a cul-4(ok1891) mutant background, 49% of worms (n = 33, S1C Fig) had only one arm with PCs. This supports our view that the phenotype is due to the cul-4(ok1891) allele since the genetic background of the two arms is, of course, identical. cul-4 null mutant strains containing the gk434 or gk511 alleles or cul-4(RNAi) treatment arrest at the L2 larval stage and do not have germlines that can be scored ( [38]and this study). In agreement, gk434 and gk511 alleles are expected to lead to larger deletion of the protein compared to ok1891. RT-PCR of cul-4 (ok1891) revealed that this mutant has an out-of-frame deletion with a premature stop codon in the cul-4 transcript. cul-4(ok1891) is expected to encode for a C' truncated protein that does not contain both the RBX/ROC binding site and the neddylation site of CUL-4. Thus, cul-4 (ok1891) likely exhibits partial penetrance as it is not a full loss-of-function allele.
Next, we examined the phenotype of mutations in the two rbx genes of C. elegans. rbx-1 (ok782) is a deletion of the entire gene, while rbx-2(ok1617) removes the last 48% of the coding sequence. Interestingly, neither rbx-1(ok782) nor rbx-2(ok1617) mutants had defects in SC assembly (S1E & S1F Fig). To test for a possible redundant function, rbx-1(ok782);rbx-2 (ok1617) double mutants were created; these mutants also showed no sign of PC formation ( Fig 1G; S1G Fig). This suggests that a RING box protein may not be required for SC assembly or that a non-canonical RING protein is involved in this function (Fig 1A and Discussion). These data collectively show that the CRL4 (CUL4 and DDB-1), as well as its regulation (CSN-5: Fig 1H; S1H Fig; [34]) are important for proper SC assembly, independent of RBX-1/2.
To identify the adaptor component of the CRL4 that is involved in SC assembly, we conducted a targeted screen using candidate genes that encode for proteins with WD40-DDB1 interaction domains (see Materials and Methods). This screen led to the isolation of gad-1 as a candidate gene. The gad-1(ok573) mutant exhibited PC formation, similar to ddb-1(tm1769) (Fig 1I; S1I Fig). Like ddb-1(tm1769) mutants, gad-1(ok573) showed 100% penetrance, all germlines analyzed had PCs (Fig 1K). We therefore propose that the CRL4 complex that is required for proper SC assembly includes DDB-1 and GAD-1 but not RBX-1 or RBX-2.
We have previously shown that csn-5(ok1064) mutants exhibited PC formation throughout meiotic prophase I [34]. This suggests that overactive (neddylated) CRL4s that cannot be recycled lead to PC formation ( Fig 1H; S1H Fig). CAND-1 is a known regulator of cullin neddylation through its ability to prevent Cullin-RING E3 ligase neddylation [46]. However, unlike other proteins in this pathway, cand-1 is not an essential gene (cand-1(tm1683) homozygote mutants are viable [46]). CAND-1 is proposed to sequester inactive CRL4 following its deneddylation, prolonging its time in the deactivated form. If the balance between neddylated and deneddylated forms is important for proper CRL4 activity in SC assembly, then inactivation of cand-1 should result in defects in SC assembly. We observed that cand-1(tm1683) mutants had a small percentage of nuclei with PCs, observed only in Zones 2 and 3 (Fig 1J and 1K; S1J Fig). This indicates that the sequestering of inactive CRL4s is not important for SC assembly as there are very few PCs formed in cand-1(tm1683) mutants.
To determine where the CRL4 complex is localized, we tagged three of the CRL4 complex proteins by CRISPR/Cas9 insertion of sequence tags encoding for OLLAS and FLAG. CUL-4, GAD-1, and DDB-1 were all enriched in germline nuclei, consistent with their proposed meiotic function (S2 Fig).

Depletion of CUL-4 levels induces PC formation
The cul-4(ok1891) mutant is not a null allele as these worms are able to develop to adulthood, whereas successful RNAi of cul-4 and other mutants of the gene (null alleles) result in larval lethality. Thus, the cul-4(ok1891) mutant phenotype can be attributed either to a reduction in CUL-4 protein levels or to the formation of a truncated CUL-4 protein. To test if cul-4 (ok1891) mutants have reduced levels of CUL-4 protein, we introduced a FLAG tag to the cul-4(ok1891) mutants and quantified the nuclear localization of CUL-4 by FLAG staining intensity (Fig 2A; S2C Fig). As expected, CUL-4 was present in cul-4(ok1891) mutants but at reduced levels in late pachytene nuclei (~15% of wild type) indicating reduced CRL4 levels. Interestingly, the variation in nuclear localization of CUL-4 in cul-4(ok1891) mutants was high and 25% of nuclei had no detectable levels of FLAG::CUL-4 similar to wild type (no FLAG:: CUL-4). This variability of CRL4 nuclear localization could explain the incomplete penetrance of PC formation in the cul-4(ok1891) mutants at 20˚C (Fig 1K). However, these data do not exclude the possibility that the phenotype observed in cul-4(ok1891) mutants may be attributed to the formation of aberrant CRL4 complexes in these mutants (see below).
If the phenotype of cul-4(ok1891) mutants is caused by a reduction of CUL-4 levels, PC formation may be induced by reduction of CUL-4 levels through a different experimental approach. To do so, we mutated cul-4 such that expression of cul-4 was in the PMT (mitotically dividing nuclei) but not pachytene (meiotic nuclei). Expression of genes in the germline are  post-transcriptionally by proteins and small RNAs that bind the 3'UTR of mRNAs [47]. We took advantage of this system of regulation through CRISPR/Cas9 replacement of the 3'UTR of the endogenous cul-4 gene with that of the 3'UTR of the polymerase epsilon gene, a replicative polymerase. This mutant should then have CUL-4 expression in the embryo, during development, and in the mitotic germline, but not in the meiotic germline. Any meiotic CUL-4 proteins present in these nuclei should be protein carried over from mitosis. We also FLAG tagged this mutant protein and observed that CUL-4 levels in the germline are reduced to~70% of that in the wild type flag::cul-4 strain, in both mitotic and meiotic nuclei (Fig 2A; S2C Fig). The fact that FLAG::CUL-4 is observed in late pachytene nuclei of cul-4::pole-1 3'UTR mutants (S2C Fig) indicates that most of the CUL-4 protein that is made in mitotic cells is not degraded upon entrance of, or in, meiotic prophase I.
Since cul-4::pole-13'UTR mutants had reduced levels of nuclear FLAG::CUL-4, we performed SYP-1 staining to test for the formation of PCs in these mutants. Unlike cul-4(ok1891) mutants, CRL4 that is formed in this mutant is expected to be wild type in nature (intact CUL-4 and interacts properly with other CRL4 components). cul-4::pole-13'UTR mutants did not show any PC formation at 20˚C, but when worms were shifted to higher temperatures, PCs were observed (Fig 2B and 2C). As little as a 2˚C temperature shift (22˚C) for 24 hours could induce PC formation. Increased shifts in temperature and length of temperature exposure corresponded to an increase in number of germlines that exhibited PC formation. No PCs were observed in wild type gonads under any of the temperature shifted conditions tested. As the cul-4(ok1891) allele has been described before [32], we continued our studies with the ok1891 allele as opposed to cul-4::pole-13'UTR mutants.
Based on fluorescence analysis, the level of CUL-4 protein in the cul-4(ok1891) mutants was lower than that in cul-4::pole-13'UTR mutants, which may explain why cul-4::pole-13'UTR mutants exhibited a phenotype at 25˚C but not at 20˚C. We would therefore expect that the penetrance of cul-4(ok1891) mutant phenotypes will be enhanced at 25˚C and suppressed at 15˚C. Indeed, cul-4(ok1891) mutants were almost completely penetrant for PC formation at 25˚C and show lower penetrance at 15˚C compared to 20˚C (Fig 2D). If cul-4(ok1891) mutants exhibited defects in SC assembly due to a gain-of-function activity, we expected the CUL-4 protein generated in this mutant to be dominant. However, cul-4(ok1891) mutant heterozygotes [cul-4 (ok1891)/+] show very limited PC formation that is only significant at 25˚C (Fig 2D). Since cul-4 (ok1891)/+ likely produce less CUL-4 protein product than +/+ (wild type), the appearance of PCs at 25˚C can be attributed to the reduction in protein levels. These data indicate that PCs can be formed due to a reduction in CUL-4 protein levels, and not only through perturbations of CUL-4 protein function such as the gain of function activity of the cul-4(ok1891) allele. genotypes had positive average values, indicating nuclear staining. Intensity is in arbitrary units. B) cul-4 mutant with 3'UTR replacement shows temperature dependent PC formation. Categorization of germlines indicated below the graph. The categories are presented next to the graph (Abundant PC in red, Few PC in purple, No PC in black). Temperatures are presented on the X axis. For example, "20C to 25C(12h)" means: worms were grown from egg at 20C, and shifted at the last 12 hours before analysis to 25C. All worms analyzed as 1 day old adults. C) Whole germline analysis of SC morphology in cul-4 mutants with 3'UTR replacement grown at 25C from the L1 stage. Wild-type worms were grown under the same conditions and exhibit no PC formation. Categories on the right. D) CRL4 mutant with 3'UTR replacement shows temperature dependent PC formation. Categorization of germlines indicated below the graph. E) Representative images of untreated and 1,6-Hexanediol treated germlines in wild type, htp-3(tm3655), csn-5 (ok1064), and cul-4(ok1891) strains. Blue (DAPI), green (HTP-3), and red (SYP-1). Images were taken from mid-pachytene (Zone 4/ 5). Scale bars are 2μm. F) Results of 1,6-Hexanediol treated germlines. Categorization of SYP-1 immunofluorescent staining after 1,6-Hexanediol treatment are defined as follows: "No SYP-1 (dispersed)" represents germlines where nuclei no longer have SYP-1 localization to chromosomes, "PCs" represents germlines where nuclei have PCs localized to chromosomes (for non-treated these also contain linear SC alongside PCs, and treated did not contain linear SC), and "Linear SYP-1" represents germlines where nuclei have linearized SYP-1 localized to chromosomes. All germlines had linear HTP-3 staining. Statistical comparisons were performed versus wild type worms (Fisher's exact test dispersed or with SYP-1 vs. PC; p-values, ��� p< 0.0001, �� 0.01> p> 0.001). https://doi.org/10.1371/journal.pgen.1008486.g002 The correlation between temperature and PC formation may be specific to the perturbation of CUL-4 activity, and not the CRL4 complex. To examine if other CRL4 mutants exhibit temperature-dependent PC formation as well, we have grown ddb-1(tm1769) and gad-1(ok573) mutants at 15˚C. These two mutants were completely penetrant at 20˚C, but had decreased penetrance of PC formation when grown at 15˚C from L4 to the adult developmental stage. These data altogether indicate that perturbation of CRL4 activity correlates with defects in PC formation in a temperature-dependent manner.

PC formation in CRL mutants are resistant to 1,6-Hexanediol treatment
PCs can be categorized into two main forms: dissolvable PCs that contain organized SC-like structures with liquid crystal properties and indissoluble PCs that have lost their liquid-like properties (Rog et al, 2017). The fact that the penetrance of PC formation phenotype is enhanced by increased temperature suggests that the PCs found in CRL4 mutants may be due to SYP protein misfolding, which would lead to loss of the liquid-like properties of the PCs in these mutants. There is evidence that misfolded PCs cannot be disassembled with 1,6-Hexanediol, while properly assembled SC can [48]. When treated with 1,6-Hexanediol, normal SC found in wild type germlines become mostly dispersed, no longer forming linear structures along chromosomes (82% of germlines; Fig 2E & 2F). HTP-3 is a component of the SC axial element; htp-3(tm3655) mutants and loss of this protein in meiotic prophase causes the formation of PCs that are known to maintain liquid-like properties (similar to linear SC). In htp-3 (tm3655) mutants, nuclei also had mostly dispersed localization of SYP-1 (76% of germlines; [48]). In both csn-5(ok1064) and cul-4(ok1891) mutants, PCs remained largely intact; only 8% and 15% of germlines had disassembled PCs, respectively (Fig 2E & 2F). Similar resistance to 1,6-Hexanediol treatment was observed in ddb-1(tm1769) and gad-1(ok573) mutants ( Fig 2F) (Fig 2F). This indicates that SC assembly is perturbed in CRL4 mutants in a way that prevents dissolution of PCs, likely due to the loss of the liquid crystal properties in these PCs.

Central region SC proteins are likely not targeted for ubiquitination by the CRL4 complex
PC formation in CRL4 pathway mutants affects all central region proteins of the SC (SYP-1/2/ 3/4), whereas it appears to have no effect on the axis (HTP-3). One possible explanation for PC formation in CRL4 pathway mutants is that ubiquitination of any or all SC central region components is different compared to wild type. In this model, the CRL4 E3 ligase complex would target one or more of the SYPs for ubiquitination. To examine this, we performed western blot analysis of central region SC components in wild type compared to csn-5(ok1064) and cul-4 (ok1891) mutants. Ubiquitination is expected to add~9kDa to the final protein product (or more for poly-ubiquitination). If one of the SYP proteins was ubiquitinated it should show a lower mobility shift in cul-4(ok1891) mutants (which lack CRL4-mediated ubiquitination) as well as a change of protein size in csn-5(ok1064) mutants (depending on whether these mutants contain hyper active or destabilized CRL4). To test this, we epitope tagged all four SYPs using CRISPR/Cas9 (see Materials and Methods). The observed SYP-2 molecular weight from western blot analysis was as expected (~24 kDa), while all other SYPs run at a higher molecular weight than expected (expected: 57, 26, and 67kDa, observed:~75,~30, and 80kDa for SYP-1, SYP-3 and SYP-4 respectively), which may be evidence of post-translational modifications. However, examination of SYP-1 and tagged versions of SYP-2, SYP-3, and SYP-4 in both wild type and, csn-5(ok1064) and cul-4(ok1891) mutants showed no evidence for ubiquitination of any SYP protein, as the molecular weight did not change between mutants and wild-type (V5 antibody: SYP-2::V5 and V5::SYP-4, and FLAG antibody: SYP-1:: FLAG and SYP-3::FLAG; S3 Fig). SYP proteins that are modified by ubiquitination may be subjected to rapid degradation. In this case, ubiquitination by CRL4 may not be detected by a standard western blot. Exposure of worms to MG132, a proteasome inhibitor, should lead to accumulation of rapidly degrading forms of SYP proteins, if they are present. However, we did not observe any change in mobility of SYP proteins under these conditions and the bands corresponding to tagged proteins showed similar mobility between wild type and cul-4(ok1891) mutants. In most cases, SYP levels did go down in the cul-4 mutant, but this can be attributed to the reduction in germline size. These data altogether suggest that the effect on PC formation is likely not due to ubiquitination of the four known SYP SC proteins by the CUL-4 E3 ligase.

CRL4 mutants show reduced pairing of homologous chromosomes
SC defects and PC formation often accompany a defect in homolog pairing [18,49,50] or PC formation may be due to an effect on SC assembly downstream from initiation of pairing interactions (e.g. [25,34]). To determine if homologous chromosome pairing is perturbed in CRL4 mutants, HIM-8 immunofluorescent analysis and fluorescent in situ hybridization (FISH) were utilized (Fig 3). HIM-8 is a pairing center protein that localizes to the X-chromosome pairing center (a small repeat sequence near the telomere) [51]. It is required for X chromosome associations with the nuclear envelope that are important for timely and proper pairing of the X chromosome. In wild type, approximately 53% of nuclei have paired X-chromosomes upon entrance into meiosis (Zone 3), and by pachytene (Zones 4-5) over 90% of nuclei have paired X homologs (Fig 3A & 3B). cul-4(ok1891) and ddb-1(tm1769), follow a similar pattern with approximately 50% paired X-chromosomes in Zone 2, but only reach approximately 70% of paired X-chromosomes in nuclei by pachytene (Zone 3; Fig 3A & 3B). This is similar to what was previously shown for csn-5(ok1064) mutants ( [34] and this study). To examine whether autosomal pairing is affected, FISH was performed at the 5S RNA locus on chromosome V. In wild type, approximately 80% of nuclei had paired 5S loci (Zone 3, Fig 3C  & 3D). In cul-4(ok1891) and ddb-1(tm1769) mutants, there was a delay in 5S loci pairing (53% and 49%, respectively) at the onset of meiosis (Zone 2) but reached higher levels of pairing, 80-90%, in pachytene (Zone 5; Fig 3C & 3D). This was also similar to what was previously shown for csn-5(ok1064) mutants ( [34] and this study). Overall, the CRL4/CSN-Cop9 pathway mutations affect pairing, however the severity of pairing defects observed were milder compared to mutants involved in establishment of pairing (e.g. [21]).
In C. elegans, chromosomes are attached to the nuclear envelope by SUN/KASH domain proteins that associate with cytoplasmic Dynein, forming patches [49,52]. This leads to microtubule dependent chromosome movement at meiotic entry that is required for chromosome pairing and homologous synapsis. Mutants that abrogate this movement also lead to formation of PCs and delayed pairing. To test if the defects in pairing in CRL4 mutants stem from an inability to establish pairing interactions we assayed for formation of SUN-1 patches and chromosome movement. SUN-1 forms patches in early prophase nuclei in all genotypes tested, indicative of proficient clustering of pairing center proteins (S4 Fig,[52]). In agreement with lack of defects in SUN-1 patch formation, we observed no defects in chromosome movement (S4 Fig). This suggests that the impaired pairing outcome is due to an inability to stabilize pairing interactions and not due to defects in chromosome movement important for homolog pairing.

Meiotic recombination is affected in CRL4 mutants
Meiotic recombination in C. elegans requires the formation of functional SC. Previous studies have shown that PC formation involving proteins of the central region of the SC correlated with the accumulation of recombination intermediates in chromosomes, despite assembling intact lateral elements (e.g. [22]). These breaks are eventually repaired as sister chromatids become accessible for repair in late pachytene [53]. Single strand recombination intermediates can be visualized by antibody staining for RAD-51, which forms foci in meiotic nuclei starting in Zones 3-4 ( Fig 4A). As has been shown previously, csn-5(ok1064) mutants have increased numbers of RAD-51 foci as nuclei progress through pachytene compared to wild type levels ( Fig 4B, [34]). Similar to csn-5(ok1064) mutants, the CRL4 mutants ddb-1(tm1769) and gad-1 (ok573) (Fig 4C; S5A Fig) showed an increase in RAD-51 foci in the germline. Interestingly, rbx-1(ok782);rbx-2(ok1617) mutants exhibited no observable defects in SC assembly ( Fig 1G,  S1G Fig), but had an increase in the amount of RAD-51 foci (Fig 4D). This effect is largely due to rbx-1 since rbx-1 (but not rbx-2) single mutants showed an increase in the amount of RAD-51 foci (S5C and S5D Fig). These results suggest that regulation of SC assembly may not be the sole function of the CRL4 complex and suggests that it is also involved in meiotic recombination. cand-1(tm1683) mutants did not show an increase in RAD-51 foci numbers compared to wild type (S5D Fig).
When RAD-51 foci numbers were analyzed, we found that the distribution of RAD-51 foci in cul-4(ok1891) mutants was unlike what was found in csn-5(ok1064), gad-1(ok573), or ddb-1 (tm1769) mutants. We observed variation in the number of RAD-51 foci in Zones 4-6 where about 30% of the nuclei had no RAD-51 foci in cul-4(ok1891) mutants ( Fig 4E). One explanation for this observation is that cul-4(ok1891) mutants have defects not only in DSB repair but also in DSB formation. A small fraction of cul-4(ok1891) mutant nuclei showed an increase in RAD-51 foci numbers, but not to the same extent found in gad-1(ok573), ddb-1(tm1769), or csn-5(ok1064) mutants. There is a correlation between the severity of the aggregation defects and the number of RAD-51 foci per nucleus in cul-4(ok1891) mutants; nuclei with PCs had more RAD-51 foci than nuclei with normal, linear SC ( Fig 4G). Thus, it is possible that the defects in SC formation are responsible, in part, to the accumulation of RAD-51 foci in cul-4 (ok1891) mutants, and this may be the same as in the other CRL4 mutants.
Along with the variation in the number of RAD-51 foci, there was an increase in irregular focus appearance in cul-4(ok1891) mutants (examples in Fig 4E & 4F). These irregular foci appeared as large lobed foci or string-like structures and were found in cul-4(ok1891), ddb-1 (tm1769), and csn-5(ok1064) mutants (Fig 4H). These irregular foci were found throughout the germline, but their numbers increased as meiosis progressed. It is possible that irregular foci were a combined fluorescence of adjacent foci, which would be more common in mutants with high RAD-51 levels. If so, rad-54 mutants, which are defective in RAD-51 unloading and exhibit elevated numbers of RAD-51 foci, should also have increased irregular foci. However, rad-54 mutants have fewer irregular foci compared to cul-4(ok1891) mutants (Fig 4H), indicating that these foci are a product of improper meiotic recombination and its progression. Another possibility for the formation of irregular RAD-51 foci is an increase in nuclear RAD-51 protein. To examine this, fluorescence intensity of RAD-51 in nuclei was measured in cul-4 (ok1891) and csn-5(ok1064) mutants in comparison to wild type nuclei. In cul-4(ok1891) mutants, intensity of RAD-51 fluorescence per nucleus was similar to that of wild type. In csn-5(ok1064) mutants, there was a significantly lower intensity of RAD-51 fluorescence per nucleus compared to wild type ( Fig 4I). Thus, CUL-4 likely has a role in regulation of RAD-51, but not through RAD-51 degradation. Unlike the defects in SC assembly, which was found only in~2/3 of the germlines analyzed, the aberrant accumulation of RAD-51 foci was found in all germlines examined (n = 39).

CUL-4 promotes meiotic recombination and represses SPO-11 independent DSB formation
Because of the wide range of RAD-51 foci numbers in cul-4(ok1891) mutants, we tested whether nuclei with no RAD-51 foci may lack the ability to form SPO-11 generated DSBs. DSB formation and repair is a dynamic process, therefore the steady state number of RAD-51 foci in a given zone is an underestimation of the overall numbers of breaks generated. RAD-54 is required for DSB repair after the RAD-51 coated ssDNA invades the homologous sequence [54]. In the absence of RAD-54, RAD-51 filaments are not processed and the overall number of DSBs can be estimated from the number of RAD-51 foci. rad-54(ok615) mutants exhibit 24X increased levels of RAD-51 foci compared to wild type (Zone 7, Fig 5C). Interestingly, cul-4(ok1891);rad-54(ok615) double mutants had significantly lower numbers of RAD-51 foci in Zones 5 and 6 compared to rad-54 (ok615) single mutants (Zone 5: 0.5X; Zone 6: 0.27X; Fig 5C & 5D). Approximately 13% of nuclei in Zone 5 of cul-4(ok1891);rad-54(ok615) double mutants had no observable RAD-51. On the other hand, rad-54(ok615) single mutants had no nuclei without RAD-51 foci supporting the idea that cul-4(ok1891) mutants have reduced numbers of DSBs (Fig 5B-5D).
One possible explanation for the source of SPO-11-independent DNA damage is the movement of transposable elements. C. elegans transposable element movement is inhibited by the RNA interference pathway and by a set of mutator genes [12,55,56]. Approximately 12% of the C. elegans genome is made up of transposable elements (TEs), but only 6 TEs are found to be active, all of which are part of the Tc1/Mariner DNA TE class (reviewed in: [57]). Expression levels of two TEs, Tc1 and Tc3, were examined by qRT-PCR. In wild type, Tc1 had low levels to no levels of detectable expression. Tc1 activity is known to be specific to somatic tissue [58], and therefore our results are likely basal expression readouts of somatic tissue transposition. In cul-4 (ok1891) mutants, Tc1 had no detectable expression, but Tc3 had an average of 2.5-fold increased expression level compared to wild type (S5G Fig). Thus, in cul-4(ok1891) mutants, certain TE transposases are more highly expressed which could lead to increased mobility. An increase in Tc3 transposition may be the SPO-11-independent source of DNA damage leading to the observed RAD-51 foci in the cul-4(ok1891);spo-11(ok79) double mutants. To test if loss of the CRL4 complex and not just perturbation of CUL-4 leads to increased expression of the Tc3 transposase, we performed qRT-PCR of Tc3 in ddb-1(tm1769) mutants. We observed a similar increase in Tc3 expression in ddb-1(tm1769) compared to cul-4(ok1891) (S5G Fig), suggesting that perturbation of CRL4, and not just CUL-4, may lead to increased transposition.

The CRL4 pathway is required for wild-type numbers of crossovers in meiosis
We have previously shown that the CSN/Cop9 signalosome is required for crossover formation [34]. Crossovers are divided into two subgroups: interfering (non-randomly spaced) and non-interfering crossovers (randomly spaced). In the C. elegans wild type germline all crossovers are interfering and are marked by COSA-1, forming one focus per chromosome [59]. To determine if CRL4 components are also required for crossovers, CRL4 mutants were crossed into a gfp::cosa-1 transgenic background. Mutants of the CRL4 complex (that also exhibit PC formation), on average, had lower levels of GFP::COSA-1 foci as compared to wild type (~61-77% to that of wild type; Fig 6A & 6B). There was also a small population of nuclei with increased levels of GFP::COSA-1 in CRL4 mutants (up to 20%; Fig 6B). This could be due to incomplete stabilization of crossover sites with COSA-1 localizing to multiple sites, indicative of meiotic recombination delay. In meiotic prophase I, progression of nuclei past pachytene are defined as part of the diplotene/diakinesis stage. CRL4 mutants, and csn-5(ok1064) mutants, do not progress to this stage; instead meiotic progression halts at late pachytene (Zone 6). This is true for all CRL4 mutants except for rbx-1 mutants. rbx-1 mutants have nuclei which progress into the final stage of diakinesis called D-1. Nuclei at this stage have 6 DAPI bodies representing the six bivalents (Fig 6C & 6D). rbx-1 mutants had an average of 6.8 DAPI bodies indicating that chiasma in some nuclei had not occurred or had been lost (Fig 6C &  6D), suggesting a minor role for RBX-1 in crossover formation.
To determine the point at which meiotic recombination no longer affects PC formation, we assessed cul-4(ok1891);msh-5(me23) double mutants expecting no reduction in PC formation if only early recombination components had an effect. In these double mutants, we found that PC formation was similar to cul-4(ok1891) single mutants (Fig 7H). Therefore, PC formation in cul-4 (ok1891) mutants is dependent upon DSB formation and early meiotic recombination progression but not later recombination events. cul-4(ok1891);rad-51(ok2218) mutants were also analyzed to identify if the recombinase RAD-51 is required for PC formation in cul-4(ok1891) mutants. Interestingly, PC formation occurs in cul-4(ok1891);rad-51(ok2218) double mutants, but to a lesser degree than in cul-4(ok1891) single mutants (~30% and 95% in Zones 3 and 4, respectively and 23% of germlines with abundant PCs compared to 49% in cul-4 single mutants; Fig 7B & 7J). Thus, the initiation of homologous recombination through SPO-11, and in part, the formation of recombination intermediates is required for PC formation in cul-4(ok1891) mutants ( Fig 7K depicts the percentage of germlines with abundant, few, or no PCs). We did not observe suppression of PC formation by spo-11 in cul-4::pole-1-3'UTR or ddb-1(tm1769) (Fig 7K) indicating that the effect is specific to the cul-4(ok1891) mutants. Regardless, our data demonstrates that PC formation can be dependent on repair of SPO-11 generated DSBs, under some circumstances, and PC formation is not necessarily uncoupled from recombination as previously thought.

Discussion
In this study, we examined the mutant phenotypes of the CRL4 E3 Ligase complex in the context of meiotic prophase I in the C. elegans germline. All mutants tested in genes encoding for the CRL4 E3 ligase complex exhibited PC formation except for rbx mutants (see model in S6A Fig). Interestingly, PC formation in cul-4 C' terminal truncation mutants was dependent on early recombination. Mutants of CRL4 also had increased levels of RAD-51 foci, indicating delays in meiotic recombination. cul-4 mutants had a mixed population of nuclei that either had no RAD-51 foci or increased levels of RAD-51 foci. Due to these recombination defects, crossovers in CRL4 mutants were at lower levels. CRL4 also may play a role in germline Linear" was similar to "SYP-1 PC and Some Linear" with the exception that SYP-1 is elongated along >50% of DAPI. "SYP-1 Partial Linear" nuclei had elongated SYP-1 along up to 50% of DAPI. "SYP-1 Mostly Linear" nuclei had SYP-1 along up to 50% of DAPI but less than 100%. "SYP-1 Linear" nuclei had fully elongated SYP-1 along all DAPI. "Other" was defined as nuclei that had abnormal DAPI appearance). Left: representative images of Zone 4 (EP) in each genotype; right: quantification of SYP-1 morphology throughout the germline of each genotype. Statistical comparisons were performed with double mutants versus single mutants (Fisher's exact test; p-values, � < 0.05). Scale bars are 2μm. K) Analysis of PC presence in all germlines analyzed for each genotype. The number of PCs per germline were grouped into three main categories: No PCs = germlines with no SC PC formation, Few PCs = germlines no more than 5 PCs or less than ½ of a zone with PCs, and Abundant PCs = germlines with more than 5 PCs or greater than ½ of a zone with PCs. Statistical comparisons were performed against cul-4(ok1891) mutants. Statistics: Fisher's exact test; p-values, � < 0.001, except cul-4; rad-51 that was significant only using 2x3 contingency protection from TE movement as mutants have increased expression of the Tc3 TE. Our data indicate that CRL4 plays multiple roles in meiotic prophase I for SC assembly, meiotic recombination, and the preservation of germline integrity.

The CRL4 E3 Ligase complex is required for preventing SC protein aggregation into PC structures
PCs can be categorized based on their reaction to 1,6-Hexanediol, a chemical that dissolves weak hydrophobic interactions. In the context of the PC, the ability to be dissolved by 1,6-Hexanediol, was interpreted as an indication that molecules can move rapidly in and out of the PC structure, similarly to what is found in a wild type PC. The inability to dissolve the PC by 1,6-Hexanediol indicated that the structure is rigid. The nature of this rigid structure is open to debate. In C. elegans PCs can be formed in wild type germlines, at elevated temperatures [60]. These heat induced PCs were shown to be resistant to 1,6-Hexanediol treatment, and therefore likely rigid [48]. In contrast, PCs formed by precocious self-assembly of the SC proteins [61][62][63][64], defects in lateral element formation [18], and perturbation of chromosome movement [49,50] likely still retain the dynamic characteristic of wild type SC, as PCs in some of these mentioned perturbations were shown to be sensitive to 1,6-Hexanediol. CRL4 is likely involved in prevention of formation of the rigid PC structure as 1,6-Hexanediol was unable to disperse the PCs formed in CRL4 mutants (cul-4, gad-1, ddb-1, or csn-5 mutants). PC formation in CRL4 mutants also exhibited temperature-dependence as the number of SC PCs increased at elevated temperatures [cul-4, gad-1, ddb-1, as well as the partial loss-of-function mutant of cul-4 ( See Fig 2 and model in S6B Fig)].
Two possible explanations exist as to why PC structures are rigid/1,6-Hexanediol-resistant in CRL4 mutants. One possibility is that SC proteins are changing their structure in a way that causes decreased dynamics. It cannot yet be determined if this change in structure is minimal (conformational change) or severe (misfolded protein). This model suggests that CRL4 promotes directly, or indirectly, protein structure of one or more of the SYP proteins. These misshaped proteins can be degraded or refolded as nuclei progress through prophase, as most nuclei do not contain PCs in late pachytene. The movement of nuclei in the germline is approximately 1 row per hour which leaves significant amount of time for the process of protein degradation/refolding (estimated~10 hours from mid to late pachytene). Alternatively, in the absence of CRL4, altered protein modifications can lead to a rigid SC structure by adding/ removing binding partners, or generating a conformation different from wild type. Due to the temperature sensitivity of the phenotype and the lack of direct evidence for modifications of SYP proteins in cul-4 mutants we prefer the first model, but both models are plausible. Our ability to detect a small population of modified SYP by this assay may be difficult, but according to our current data, the role of CRL4 appears to be indirect.
A number of post-translational modifications have been shown to target and affect SC proteins. One of the most iconic is sumoylation of Ecm11 in yeast, important for both SC assembly and recombination [65][66][67]. In mouse and yeast, it was shown that both SUMOylation and ubiquitination play an important role in SC assembly and recombination progression, both through proteasome recruitment to chromosome axes [33,68]. SUMOylation plays no role in SC assembly in C. elegans [69,70]; instead, SYP-1 acetylation [71] and phosphorylation [72], are important for assembly and disassembly, respectively. Phosphorylation of SYP-4 was found to be important for changing the structure of the SC from a dynamic nature, where SC proteins are mobile, to one that is less dynamic [23]. We provide evidence that another posttranslational modification, ubiquitination, affects SC assembly in C. elegans ([34]and this manuscript). This is in agreement with studies demonstrating that proteasome inhibition leads to PC formation in mouse and C. elegans [33,68]. Like our study, these studies did not demonstrate a direct effect of ubiquitination on SC proteins.
We have shown that mutants of csn-5, cul-4, ddb-1, and gad-1 all lead to similar phenotypes: defects in SC assembly accompanied by PC formation. We have shown that the PCs formed in CRL4 mutants are specific to central region proteins and do not affect axial element proteins of the SC. However, we found no evidence for ubiquitination of SYP proteins, as no reproducible change in molecular mass was observed in SYP proteins in mutants of either cul-4 (ok1891) or csn-5(ok1064). This suggests that the effect of ubiquitination on SC formation is indirect. It is possible that proteins involved with regulating the rigidity of the SC may be targeted for ubiquitination.
In previous work, we have shown that mutants of the CSN/Cop9 complex have defects in SC assembly [34]. Based on the similarity of phenotypes observed between cul-4 mutants and CSN/Cop9 mutants and studies in other systems (reviewed in: [73]), we propose that CSN/ Cop9 deneddylates CRL4 (Fig 1A). After deneddylation, CRL4 then targets another protein(s) for ubiquitination that indirectly maintains the liquid property of the SC, which is required for SC assembly. The current model of CRLs states that their neddylated form is the active form, and the lack of deneddylation prevents the recycling of the E3 ligase complex and/or its stability [74]. Based on this model, an increase in neddylation (e.g., csn-5 mutants) or loss of CRL4 activity (e.g., cul-4, gad-1, ddb-1 mutants), both lead to the same outcome. The data we present here are consistent with this model.
Another interesting finding is that the rbx genes have no role in SC assembly, which does not fit the canonical CRL4 complex model (Fig 1A). It is possible that RBX-1 acts solely as a member of CRL4 that is involved in DNA repair, as we have no evidence supporting its function in SC assembly, and instead see a difference in recombination intermediate abundance (RAD-51). Both rbx mutants are expected to be null based on the fact that rbx-1(ok782) deletes the whole gene and rbx-2(ok1617) deletes half of the coding frame. RBX-1 and RBX-2 are the only ROC1 homologs in C. elegans. It is possible that other functional homologs exist that are unidentifiable by sequence homology or that CRL4 in C. elegans can interact with other RING domain proteins. The RING domain proteins acting in CRLs are small and part of the H2 subfamily of RING finger proteins [73]. The C. elegans genome encodes for 101 other RING proteins, any of which may act redundantly with rbx-1/2 (Moore & Boyd, 2004). Consistent with these possibilities, rbx-1(ok782);rbx-2(ok1617) mutants progress through larval stages, while cul-4 null mutants cannot. Alternatively, CUL-4 may function in regulation of PC formation independently from its activity as an E3 ligase. This is supported by the fact that cul-4(ok1891) mutants which have a C' truncation that removes the RBX binding site, can support SC formation in~35% of germlines at 20˚C. According to this model, the phenotypes associated with cul-4(ok1891) mutants are attributed to the reduction in CUL-4 levels and not to the removal of the RBX domain. As far as we know, there is no evidence for CRL4 acting outside of its role as an E3 ligase, which may favor our first model (alternative RING protein to RBX-1/2).
The role of CAND-1 is to sequester the Cullin complexes once they are deneddylated to prevent over-activation via rapid neddylation. In terms of PC formation, cand-1 mutants have a relatively mild phenotype in comparison to other mutants of the CRL4 E3 ligase complex, which is consistent with a previous study where defects of cand-1 mutants mildly mimicked Cullin mutant phenotypes [46]. Along with the fact that cand-1 homozygous mutants are fertile; these data suggest that cand-1 activity is only partially impactful where Cullin activity occurs. We conclude that deactivation of Cullin ligases by deneddylation associated with the binding of the CSN/Cop9 complex is more important for controlling CRL4 activity in the germline than sequestering of the deneddylated ligases with CAND-1, indicating that quick recycling of CRL4 is key for SC assembly.

CRL4 is required for proper progression of recombination
In our previous work, we had shown that mutants of the CSN/Cop9 complex have defects in recombination that were attributed to defects in SC assembly [34]. Since CSN/Cop9 acts through CRL4, the same logical argument can be applied here. Lack of assembly of the central region of the SC led to defects in recombination; DSBs are formed but cannot be repaired though homologous recombination with the homologous chromosome [For example: [22,50,75]]. Thus, PC formation in CRL4 mutants may explain why RAD-51 foci accumulate in these mutants and crossover levels are reduced. Unlike what is found for SYP null mutants [20][21][22]75], CRL4 mutants still show~50-70% homologous synapsis which explains why crossover levels are reduced but not eliminated in these mutants. This may be the most parsimonious explanation for the results. However, our data suggests that this may not be a complete explanation. rbx-1;rbx-2 double mutants that have no PC formation still exhibit a small but significant elevation in RAD-51 foci, compared to wild type. This may indicate that the CRL4 complex in which RBX-1/2 are not participating regulates SC assembly, while the CRL4 complex in which RBX-1/2 are present has additional function(s) in DSB repair. Since rbx-1 mutants show only a mild effect on COSA-1 foci numbers, this indicates that the direct effects of the CRL4 complex in DSB repair are minor compared to the indirect effects of SC assembly.
CUL4 has an established function in somatic DSB repair and DNA damage signaling. After UV damage, CUL4 plays a role in recognition and nucleotide excision repair of damage as well as halting of cell cycle progression after this damage [76,77]. CUL4 signals for DNA damage recognition through ubiquitination of histones, H2B, H3, and H4, which has been shown to facilitate Exo1-mediated resection [78,79]. CUL4 also facilitates DNA damage recognition by stabilization of p53 through PCNA interactions [37], and CRL4 inactivation leads to the induction of p53 [80]. Upstream of p53 in DNA damage recognition and signaling is the Chk1 kinase. Chk1 has also been shown to be regulated by CUL4 in both normally cycling cells and cells under replication stress [81]. If CRL4 directly affects repair of DSBs, the appearance of irregular RAD-51 foci in cul-4 and ddb-1 mutants may be a cytological manifestation of a direct effect on recombination.

Truncation of the C' terminal domain of CUL-4
While ddb-1(tm1769) and gad-1(ok573) are likely null alleles, cul-4(ok1891) confers a partial loss-of-function. cul-4(ok1891) mutants produce viable adults in low frequency, while two other mutants of cul-4, gk434 and gk511, arrest in the larval stages. Why ddb-1 and gad-1 null mutants still progress through development unlike the null cul-4 allele may be explained by redundancy: CRLs are known to work with many alternative adaptor/substrate recognition pair(s) [36]. Thus, DDB-1 and GAD-1 may have a more important function in the germline than in larval development.
The ok1891 allele is a deletion located near the C-terminus of the protein, leading to a frameshift mutation; while cul-4(gk434) and cul-4(gk511) are deletions predicted to cause earlier truncations in the coding sequence. The ok1891 deletion removes the neddylation target site and also the binding site for the E2 enzyme. Therefore, the cul-4(ok1891) mutation likely affects CUL-4's ability to form a fully functional CRL4 complex. However, the fact that some worms of cul-4(ok1891) mutants [but not cul-4(gk434) and cul-4(gk511) mutants] do not trigger larval arrest, indicates that this malformed CRL4 complex maintains partial activity circumventing the essential function of CUL-4 during larval development. It is possible that in the absence of the lysine modified by neddylation in wild type CUL-4, another lysine serves as an alternative site for modification in cul-4(ok1891) mutants, rendering it partially active. It is perplexing how CUL-4 protein that lacks the C' domain (RBX/ROC-E2 binding) was able to bypass the need for a functional CRL4 in some worms. It is possible that the E2 enzyme can modify some CRL4 substrates without directly binding to CUL-4; the C' truncated form of CUL-4 that is still able to bind the substrate may provide some assistance to the E2 creating a partial loss of function phenotype.
cul-4::pole-1 3'UTR have similar increased levels of RAD-51 as to these observed in the csn-5, ddb-1, and gad-1 mutants. However, in cul-4(ok1891) mutants, meiotic prophase I nuclei have SPO-11 independent DSBs as well as nuclei that lack SPO-11 DSBs. Moreover, in cul-4::pole-1 3'UTR mutants, PC formation is independent of SPO-11, whereas in cul-4(ok1891) mutants PC formation is dependent on SPO-11. These findings indicate that germline nuclei of these two different cul-4 mutants did not produce the same defects, aside from PC formation, and this can be explained by a gain-of-function in the cul-4(ok1891) allele. The gain-of-function phenotypes in cul-4(ok1891) mutants may be due to the formation of an aberrant CRL4 complex (that binds the substrate through DDB-1/GAD-1, but not the E2) that interrupts SPO-11 induced DSB formation. Alternatively, the unique phenotypes for cul-4(ok1891) mutants compared to cul-4::pole-1 3'UTR mutants may be due to a manifestation of a more severe reduction in CUL-4 levels. In this model, CUL-4 forms a complex with other CRL4 complex proteins, beside GAD-1/DDB-1, and performs germline functions outside SC assembly. In this case, CUL-4 activity in SC assembly (GAD-1/DDB-1 dependent) is perturbed due to the reduction of CUL-4 availability, however other CUL-4 activities such as DSB -dependent PC formation and transposition, are only abrogated when CUL-4 levels drop even further. This may explain why cul-4::pole-1 3'UTR mutants (~70% of wild type protein levels) are similar to ddb-1 and gad-1 mutants, while cul-4(ok1891) mutants (~15% of wild type protein levels) exhibit distinct phenotypes (S6A Fig). In this model, different CRL4 complexes have different required CUL-4 proteins levels for proper function. The caveat with this model vs. the gain-of-function activity model is that it assumes that the C' truncated CUL-4 is at least partially functional, retaining E3 ligase activity.
Another unique feature of cul-4(ok1891) mutants particularly in comparison to csn-5, ddb-1, and gad-1 mutants is the incomplete penetrance of the PC formation phenotype. PCs were present in 2 out of 3 germlines analyzed and two germlines of the same worm may exhibit different phenotypes. The high variation in levels of nuclear localization of CUL-4 in cul-4 (ok1891) mutants may explain the incomplete penetrance observed at 20˚C. Similarly, cul-4:: pole-1 3'UTR mutants also shows partial penetrance, but is observed at higher temperatures (22-24˚C). In cul-4::pole-1 3'UTR mutants, CUL-4 levels are reduced but not to the same extent as cul-4(ok1891) mutants. This may explain why PC formation only appears at higher temperatures in cul-4::pole-1 3'UTR mutants.

Interplay between SC assembly and recombination in C. elegans
In organisms such as mouse or yeast, SC assembly depends on the initiation of meiotic recombination by SPO-11 [7,8]. In the absence of SPO-11 in C. elegans, the SC assembles along homologous chromosomes similar to wild type [6]. Thus, recombination is not typically seen as an interfering process in SC assembly; it either has a positive or neutral effect. In cul-4 (ok1891) mutants, PC formation is reduced when early recombination is inhibited (ex. loss of SPO-11 or RAD-54), but no such effect was observed when CRL4 levels are reduced by 3'UTR replacement or by removing ddb-1, despite high penetrance of the PC-formation phenotype. This suggests that in the presence of functional CRL4, crosstalk between recombination and PC assembly is prevented in C. elegans, but aberrant CRL4 formation or its very low levels (in cul-4(ok1891) mutants) leads to a link, or crosstalk, between PC formation and recombination.
The only other known example of crosstalk between recombination and SC assembly in C. elegans is in the cra-1 mutant background where loss of SPO-11 caused increased assembly defects [25]. CRA-1 is a NatB domain-containing protein shown to have roles in synapsis and crossover formation on autosomes as well as global histone acetylation [25,82]. In cra-1 mutants, SC assembly defects become more severe in the absence of early meiotic recombination intermediates [25]. Thus, CRL4 prevents recombination interfering with SC assembly and CRA-1 inhibits some mechanism that promotes SC assembly through recombination.
Another important connection between SC and recombination was revealed in studies of SC dynamics. Despite the fact that the SC may appear as a rigid structure, SC proteins were shown to be exchanged in a dynamic manner throughout early to mid-prophase [23,24]. This exchange was attenuated in late prophase and is dependent on the formation of crossovers [24]. These findings imply that the recombination process may negatively regulate SC assembly by limiting the influx of new SYP proteins to the SC. Unlike CRL4 mutants which mainly affect early to mid-prophase, the effect on SC dynamics is restricted to late pachytene. However, if PCs can be considered a less dynamic SC structure than assembled SC, then our studies and that of [23] both show similar effect that are executed at different time points in meiotic prophase I. Both studies show that recombination can promote a less dynamic SC structure, but depending on the point in prophase can lead to a different effect: PC formation in early prophase, while promoting SC disassembly in late prophase. This may also relate to the desynapsis pathway that opposes SC stability after the SC is formed. In this pathway, recombination intermediates destabilize the SC for chromosomes in which not all crossovers have been formed, leading to SC disassembly but without PC formation [83]. However, this pathway is not only phenotypically distinct but also known to act later in prophase, whereas CRL4 acts much earlier in prophase during SC assembly.
In cul-4 C-terminal truncation mutants, PC formation is reduced when recombination is prevented (loss of SPO-11) or inhibited at its early steps (loss of RAD-54 and to a lesser degree RAD-51). Loss of the pro-crossover factor MSH-5 had no effect on PC formation. This indicates that PC formation is inhibited when recombination intermediates (RAD-51 covered ssDNA) do not form or when recombination intermediates are locked in position (form joint molecules but are unable to disassemble). However, the effect of removing RAD-51 on PC formation in cul-4(ok1891) mutants is milder than that of removing RAD-54. This suggests that the mechanism involved responds to perturbation of recombination at different levels, when perturbation of early (spo-11) or late (rad-54) events in recombination is more important. In this model, the RAD-54 protein itself may directly participate in the PC formation. Alternatively, RAD-51 may be required for amplification of the signal caused by impaired recombination, proposing that in rad-51 mutants the signal existed but cannot be transmitted. As discussed above, some aspects of DSB repair defects in CRL mutants may be independent of PC formation, thus it is plausible that CRL4 has more than one target for ubiquitination; one target that promotes formation of functional recombination intermediates and another target involved in SC assembly.

CUL-4's multiple roles in meiotic prophase I
In previous studies, CUL4 has been shown to play multiple roles in meiotic prophase I. In mouse and the Chinese mitten crab, loss of CUL4 led to increased levels of apoptosis, decreased and malformed spermatozoa, and a delay of meiotic recombination with accumulation of late recombination intermediates [29,41,84]. This study provides further evidence that CUL-4 ubiquitination activity has an important function in meiosis. Despite the clear importance of CRL4 in meiosis, the question remains as to the identity of CRL4 targets in meiotic prophase I.
CUL-4 plays an important role in cell cycle progression and meiotic entrance. CUL-4 prevents re-replication in S-phase via ubiquitin-signaled degradation of CDT-1, a replication licensing factor. This function is well conserved and is present in humans, mice, frogs, worms, flies, and yeast (reviewed in: [43]). In C. elegans, CRL4 was found be important for export of CDC-6, another replication licensing factor that complexes with CDT-1 [39]. We have shown that CSN/Cop9 mutants have proliferation defects that account for the smaller germlines observed in these mutants [34], and CRL4's function in replication licensing may explain this phenotype.
We have shown here that CRL4 acts in SC assembly and has both direct and indirect roles in recombination. Through the analysis of cul-4(ok1891) mutants, we have shown that dysfunctional CRL4 can interfere with both DSB formation and downregulation of SPO-11 independent DSBs. Loss of CUL-4 could affect localization or function in a number of different proteins important for DSB formation. Another possibility is that DSB formation defects stem from CUL-4's effect on chromosomal structure, since CUL-4 is known to ubiquitinate histones in humans and mice (discussed in: [85]). Our analysis also demonstrated that dysfunctional CRL4 can interfere with the repression of the Tc3 transposase. While we have not directly measured mobility due to the sterility of CRL4 mutants, we suspect that it is increased; if so, transposon movement could contribute to the SPO-11 independent DSBs observed in the cul-4 mutant. [13] have shown that the PIWI pathway downregulates Tc3 expression and has no role in the regulation of Tc1 transposase expression [13]. Therefore, one explanation for CRL4 mutants' effects on Tc3 transposase expression would be the mis-regulation of PIWI.
To conclude, we propose that CRL4 plays a central role in meiotic prophase events; it is required for maintaining the functional and dynamic SC, and its proper activity promotes formation of recombination intermediates that do not interfere with SC assembly. Dysfunctional CRL4 can interfere with DSB formation and activate transposition. Future studies should aim to identify the targets of CRL4 in meiosis.

'UTR
All CRISPR/Cas9 reactions used ssODNs and crRNAs generated by Integrated DNA Technologies (IDT), materials used are listed below in the table. Recombinant Cas9 protein was isolated by QB3 MacroLab (University of California, Berkeley). A dpy-10 co-injection marker was used as a measure of successful Cas9 excision. CRISPR/Cas9 was performed as described in [86](with slight modifications) and the co-CRISPR protocol described in [87]. spo-11 and ddb-1are linked therefore the double mutant was created by CRISPR/Cas9. Deletion of spo-11 was generated as in [88] but by injection into ddb-1(tm1769/nT1 strain [spo-11(iow110) deletion was verified as homozygous the PCR of ddb-1(tm1769) homozygotes, non-balanced worms].

RNAi Screen
To identify possible adapter proteins and E2 conjugating enzyme(s) involved in SC assembly, we performed RNAi knockdown of 41 different possible adapter proteins, containing the DDB1 binding WD40 domain (DWD), and 22 predicted E2 enzymes. The screen was performed in an akir-1(gk528);rrf-1(pk1417) double mutant background. Previous work in our lab has shown that akir-1(gk528) mutants can be used as a sensitized background to find genes that are involved in SC assembly [34,50]. rrf-1 is a gene encoding an RNA dependent RNA polymerase expressed in somatic tissue; rrf-1 mutants were used to direct RNAi knockdown specifically to the germline, as these RNAi targets likely have a function in somatic tissue. A gfp::syp-3 transgenic line was crossed into these double mutants to quantitatively determine if each knockdown condition influenced SC assembly. Germlines were examined for PC formation in pachytene; PCs were defined as SC structures that are at least two times the width of linearized SC. In this screen, we exposed our double mutants with the gfp::syp-3 transgenic line to RNAi by feeding and performed whole worm ethanol fixation for analysis. RNAi clones were inoculated overnight in Lauria broth and ampicillin (50 ng/μl). Bacterial cultures were seeded on plates containing IPTG then allowed to grow for a minimum of 12 hrs. at 37˚C. RNAi clones from the Ahringer C. elegans RNAi library were used [89]. L1 synchronized larvae were placed on the seeded RNAi bacterial plates, pL4440 empty vector was used as a control. Once the next generation reached L4 stage, they were moved to another set of RNAi seeded plates. The next generation at L4 stage were further moved to another set of plates and as day 1 adults were ethanol fixed and examined for PC formation.

Sample size for figures
GraphPad Prism was used for generating of all graphs and statistical analyses. n reflects the number of either germlines or nuclei analyzed.  Fig 6:

Germline length microscopy and analysis
Adult hermaphrodites were fixed with ethanol and sealed with DAPI and Vectashield upon reaching day one adult stage. Full worm images were taken with a Leica DM RXA compound light microscope with QI Click Camera 20X lens. Images were analyzed using softWoRx software by calculating the distance between the distal tip cell to the end of pachytene. Ten worms were analyzed per genotype. Statistical comparison: two-tailed Mann-Whitney.

Germline zonal determination
Germlines were divided into zones based on length: wild type length having seven zones and cul-4(ok1891) mutants having six zones. Each zone represented a time point in meiotic prophase I progression (described in Results; SC assembly is perturbed in CRL4 mutants). Zones were created by imaging from the distal tip cell using a 512x512 pixel dimension frame of view. Images were taken until the end of pachytene.

GFP::SYP-3 analysis
Adult hermaphrodites were fixed with ethanol and sealed with DAPI and Vectashield upon reaching day one adult stage. Whole worms were observed under the Deltavision System described earlier. Germlines were examined for GFP::SYP-3 PC formation (see RNAi Screen for PC definition).

FISH
5S locus FISH and analysis was performed as in [21]. Creation of the 5S rDNA FISH probe was done by amplification of the 1kb 5S rDNA locus with primers: 5 0 -TAC TTG GAT CGG AGA CGG CC-3 0 and 5 0 -CTA ACT GGA CTC AAC GTT GC-3 0 and fluorescent labeling with terminal deoxynucleotidyl transferase. Freeze crack method was used as in Immunofluorescent Staining and Microscopy except with 7.4% PFA applied before freezing. 2X SSC-tween was used as buffer. Hybridization of the 5S rDNA probe to germline nuclei was performed at 94˚C for 90s. Three germlines per genotype were scored.

FISH and SYP-1 immunostaining
Germline squash preparations were prepared as in [85]. Squash preparations were first processed as in "Immunofluorescent Staining and Microscopy" with α-SYP-1 (1:500, VDAPTEA-LIETPVDDQSSGFL) and secondary antibody α-goat Alexa Fluor 549. Nuclei were fixed in 3.7% PFA then hybridized with the 5S rDNA probe as in "FISH". Nuclei were then stained with DAPI. Analysis of homologous synapsis in pachytene nuclei was performed by searching for nuclei with linear SYP-1 and at least one 5S rDNA focus residing on or next to linearized SYP-1. Paired foci were determined by a distance of 0.7μm or less between 5S rDNA foci; as in "FISH and HIM-8 Analysis" below.

Western blot
Whole

SYP-1 localization analysis
Qualitative scoring of SYP-1 localization patterning within individual nuclei was determined by placement into one of eight different categories. "No SYP-1" was defined as having no SYP-1 immunofluorescence present along chromosomes (DAPI). "SYP-1 PC" were nuclei with only PC formation present, no elongation of SYP-1 along DAPI. "SYP-1 PC and Some Linear" was the presence of PC(s) in the nucleus but also partial elongation of SYP-1, <50% of DAPI. "SYP-1 PC and Linear" was similar to "SYP-1 PC and Some Linear" with the exception that SYP-1 is elongated along >50% of DAPI. "SYP-1 Partial Linear" nuclei had elongated SYP-1 along up to 50% of DAPI. "SYP-1 Mostly Linear" nuclei had SYP-1 along up to 50% of DAPI but less than 100%. "SYP-1 Linear" nuclei had fully elongated SYP-1 along all DAPI. "Other" was defined as nuclei that had abnormal DAPI appearance, for example in csn-5(ok1064) mutants this represented micronuclei formation, and in cul-4(ok1891) mutants this represented an EMO-like dispersed DAPI appearance. Statistical comparison: Fisher's exact test.

RAD-51 foci analysis
All zones were quantified for number of RAD-51 foci per nucleus as in [22]. Three germlines were scored per genotype and the mean was determined. Statistical comparison: two-tailed Mann-Whitney.

Fluorescence intensity measurements
16-bit non-deconvolved images acquired with a one second exposure were analyzed by FIJI software. Intensity measurements using anti-FLAG antibody were gathered by drawing a circle around a nucleus at its greatest width from a single Z-stack from PMT or LP. For each genotype and region, 10 nuclei per germline were measured for three or more germlines. Background was recorded at 4 different positions inside the cytoplasmic space and the average of the cytoplasmic intensities was subtracted from each nuclear intensity recording. Mann-Whitney U test (GraphPad Prism 7 software) was used to statistically compare intensities of staining. For RAD-51 intensity analyses, images acquired for RAD-51 Foci Analysis were used. Full nuclear projections were created and analyzed in ImageJ. An ellipse (1 pixel wide) was placed around a targeted nucleus, based on DAPI staining, using the circle tool. Mean intensity of the RAD-51 channel was calculated and divided by the number of RAD-51 foci with that nucleus. These values were then normalized to the mean intensity of the RAD-51 channel in an ellipse placed in the cytoplasm. Statistical comparison: two-tailed Mann-Whitney.

Irregular RAD-51 foci analysis
Images acquired for RAD-51 Foci Analysis were used. Individual nuclei were analyzed for abnormal RAD-51 appearances. Abnormal RAD-51 were defined as anything that did not form wild type puncta. Examples of this include stretches of RAD-51 or a globular "lobed" appearance larger than wild type puncta. Statistical comparison: Fisher's exact test.

1,6-Hexanediol treatment and analysis
Adult worms were dissected in Egg buffer as in timeline for different mutant strains described in Immunofluorescent Staining and Microscopy after L4 stage selection. Dissected germlines were incubated in 15% 1,6-Hexanediol then fixed in 3.7% PFA. Germlines were then immunofluorescent antibody stained for SYP-1 and HTP-3. Using the DeltaVision system and soft-WoRx software, images were analyzed for presence of SYP-1 PC formation and linearization along the axial element HTP-3. Statistical comparison: Fisher's exact test.

FISH and HIM-8 analysis
Zonal division, as in SYP-1 Localization Analysis, was used to divide germlines and were scored for 5S rDNA FISH signal/HIM-8 foci pairing in each nucleus. Paired foci were determined as being 0.7-μm or less apart [21]. Statistical comparison: Fisher's exact test. Simultaneous antibody staining and FISH was performed on squashes as in [17].

COSA-1 foci analysis
GFP::COSA-1 was quantified in the final zone of each germline (either Zone 6 or 7 dependent upon germline length), corresponding to late pachytene [59]. COSA-1 foci were counted per nucleus with a scoring of at least three germlines per genotype. Statistical comparison: twotailed Mann-Whitney.

Diakinesis nuclei analysis
Images were taken of the final diakinesis nucleus (diakinesis -1) prior to spermathecal entrance. Number of DAPI bodies were totaled following counting in 3D. Statistical comparison: two-tailed Mann-Whitney.

DHC-1::GFP movement analysis
Analysis of movement was performed in DHC-1::GFP transgenic strains in the wild type (N2) and cul-4(ok1891) mutant background. Images of single plane at TZ (based on distance from tip and presence of DHC-1::GFP foci) were taken for 60 s. Analysis was performed in ImageJ using the plugin Manual Tracking for tracking of DHC-1::GFP foci. Data points were taken at the onset and conclusion of each directional movement, obtaining the calculation of total distance traveled, mean velocity, number of times foci changed direction, and time between movements of foci. All DHC-1::GFP foci visualized during the whole duration of the movie (60s) were analyzed. Statistical comparison: two-tailed Mann-Whitney.

Statement on data and reagent availability
Strains are available upon request.