The conserved LEM-3/Ankle1 nuclease is involved in the combinatorial regulation of meiotic recombination repair and chromosome segregation in Caenorhabditis elegans

Homologous recombination is essential for crossover (CO) formation and accurate chromosome segregation during meiosis. It is of considerable importance to work out how recombination intermediates are processed, leading to CO and non-crossover (NCO) outcome. Genetic analysis in budding yeast and Caenorhabditis elegans indicates that the processing of meiotic recombination intermediates involves a combination of nucleases and DNA repair enzymes. We previously reported that in C. elegans meiotic joint molecule resolution is mediated by two redundant pathways, conferred by the SLX-1 and MUS-81 nucleases, and by the HIM-6 Bloom helicase in conjunction with the XPF-1 endonuclease, respectively. Both pathways require the scaffold protein SLX-4. However, in the absence of all these enzymes, residual processing of meiotic recombination intermediates still occurs and CO formation is reduced but not abolished. Here we show that the LEM-3 nuclease, mutation of which by itself does not have an overt meiotic phenotype, genetically interacts with slx-1 and mus-81 mutants, the respective double mutants displaying 100% embryonic lethality. The combined loss of LEM-3 and MUS-81 leads to altered processing of recombination intermediates, a delayed disassembly of foci associated with CO designated sites, and the formation of univalents linked by SPO-11 dependent chromatin bridges (dissociated bivalents). However, LEM-3 foci do not colocalize with ZHP-3, a marker that congresses into CO designated sites. In addition, neither CO frequency nor distribution is altered in lem-3 single mutants or in combination with mus-81 or slx-4 mutations. Finally, we found persistent chromatin bridges during meiotic divisions in lem-3; slx-4 double mutants. Supported by the localization of LEM-3 between dividing meiotic nuclei, this data suggest that LEM-3 is able to process erroneous recombination intermediates that persist into the second meiotic division.


Introduction
Meiosis is comprised of two specialized cell divisions that elicit the reduction of the diploid genome to haploid gametes. Homologous recombination occurs in the first meiotic division and is required for meiotic crossover (CO) formation [1]. COs are needed to shuffle genetic information between maternal and paternal chromosomes and are thus required to ensure genetic diversity. COs become cytologically visible as chiasmata and also provide stable connections between maternal and paternal homologous chromosomes (homologues). Chiasmata counter the spindle force and thereby facilitate the accurate segregation of homologues in the first meiotic division.
Meiotic recombination is initiated by DNA double-strand breaks (DSBs) generated by the conserved meiosis-specific Spo11 protein [2]. The number of DSBs generated by Spo11 exceeds the number of COs, ranging from~2:1 in Saccharomyces cerevisiae to~20:1 in maize [3][4][5][6]. In Caenorhabditis elegans, each chromosome pair receives 4-7 DSBs over the course of prophase I and typically only one DSB per homologous pair will mature into a CO event [7,8]. It is thought that the excessive number of DSBs is required to ensure that at least one CO occurs on each homologue, a notion supported by checkpoint mechanisms that delay meiotic prophase progression when the number of DSBs is reduced [9][10][11]. It is unclear how the obligate CO is selected from the pool of DSBs. The CO selection (or designation) correlates with the congression of several pro-CO factors into six distinct foci, one on each paired chromosome starting from the mid-pachytene stage. These include the cyclin-related protein COSA-1/CNTD1, MSH-4/MSH-5 components of the MutSγ complex, the predicted ubiquitin ligases ZHP-3/RNF212 and HEI10, the Bloom (BLM) helicase HIM-6 as well as its regulatory subunit RMH-1 [12][13][14][15][16][17].
When the CO designation sites are associated with those pro-CO factors, processing of meiotic recombination intermediates is biased towards the CO outcome. One of the meiotic recombination intermediates is called a Holliday junction (HJ), a cruciform DNA structure formed as a result of a reciprocal exchange of DNA strands between homologous chromosomes [18]. While in fission yeast single HJs appear to predominate, direct evidence for the occurrence of double HJs (dHJs) was obtained in budding yeast [19]. dHJs can be processed to result in CO or a non-crossover (NCO) outcome, depending on the directionality of the cut made by structure-specific endonucleases [20]. There is emerging evidence that a combination of nucleases is required for the processing of meiotic HJs to promote CO formation and to resolve joint DNA structures that might impede proper chromosome segregation [21]. Only in fission yeast, deletion of a single nuclease MUS81, leads to a defect in meiotic CO formation [22]. In budding yeast, absence of the MUS81-MMS4, SLX1-SLX4 or YEN1 nucleases exhibits a modest reduction of meiotic COs [23]. The Exo1 nuclease and the mismatch-repair MutLγ complex Mlh1-Mlh3 have also been shown to contribute to HJ resolution [23,24]. Mouse gen1 mutants have no meiotic phenotype, while mus81 animals only show minor defects [25]. In C. elegans HJ resolution and CO formation appear to be conferred by at least two redundant pathways [26][27][28]. One pathway is defined by the MUS-81 and SLX-1 nucleases. Consistent with in vitro nuclease assays, it appears that SLX-1 might confer a first nick on a HJ, the nicked HJ being the preferred substrate for MUS-81 [29]. The second pathway comprises the XPF-1 nuclease and the BLM helicase HIM-6. It is possible that HIM-6 might be able to unwind a HJ, which would generate a substrate cleaved by XPF-1. Both pathways require SLX-4 as a scaffold protein. When both pathways are compromised, the CO frequency is reduced by roughly one third [26].
Since only a small subset of DSBs are designated as CO sites, the majority of DSBs have to be processed to favour inter-homolog NCO and/ or inter-sister recombination [30]. In budding yeast recombination events leading to the majority of NCO events mature early, whereas the CO events mature later [31,32]. Several helicases are proposed to mediate the disassembly of early recombination intermediates such as D-loop structures in a pathway called synthesisdependent strand annealing (SDSA). In budding yeast, this is driven by the Srs2 helicase [33]. In animals, this activity is ascribed to the BLM and RTEL helicases. In C. elegans deletion of the RTEL-1 helicase leads to an elevated number of meiotic COs [34]. In contrast, deletion of him-6, the C. elegans BLM homologue, leads to reduced meiotic CO formation [34]. The presence of HIM-6 at CO designation sites infers a late pro-CO function [35,36]. Once dHJs are formed, they can either be dissolved by the BLM helicase and Top3 topoisomerase in a NCO manner or resolved by nucleases to form CO or NCO [33].
In this study, we report on roles of the LEM-3/Ankle1 nuclease in processing meiotic recombination intermediates. LEM-3 is only conserved in animals and the mammalian ortholog is referred to as Ankle1 [37,38]. C. elegans lem-3 mutants are hypersensitive to ionizing irradiation, UV treatment and DNA cross-linking agents [37]. LEM-3/Ankle1 contains an Nterminal LEM domain, Ankyrin repeats and a GIY-YIG nuclease motif. The same nuclease motif can also be found in bacterial UvrC nucleotide excision repair proteins and in the distantly related SLX1 nuclease [39]. Our data show that LEM-3 and MUS-81 act in conjunction to process early recombination intermediates in meiosis. Loss of LEM-3 and MUS-81 leads to aberrant profiles of recombination markers, delayed processing of markers for CO designation, increased apoptotic cell death and the formation of dissociated bivalents. In addition, we found that a considerable pool of LEM-3 localizes between dividing meiotic nuclei and chromosome segregation is compromised due to persistent chromatin linkages in the absence of both LEM-3 and SLX-4, indicating that LEM-3 is able to process erroneous recombination intermediates that persist into meiotic divisions.

LEM-3 acts in a genetic pathway parallel to SLX-4 and MUS-81/SLX-1
We and other groups previously showed that there are at least two pathways needed for the resolution of meiotic recombination intermediates: one dependent on SLX-1-MUS-81 and the other relying on XPF-1. SLX-4 acts as a scaffold component in both pathways [26][27][28]. Given that viability and CO recombination are reduced but not eliminated when both the MUS-81 and XPF-1 pathways are compromised, we considered that there might be at least one additional nuclease that had not been identified. We therefore searched for nucleases which are synthetic lethal with SLX-4 and focused on LEM-3 in this study. Out of the three previously reported LEM-3 alleles, we used the lem-3 (mn155) and the lem-3 (tm3468), the former leads to a premature stop codon at amino acid 190 leaving the N-terminal Ankyrin Repeat domain intact, but eliminating the nuclease domain, thus representing a null allele [37]. lem-3 (tm3468) bears an in-frame deletion of 110 amino acids between the Ankyrin Repeat and the LEM domain [37]. We found that the lethality of lem-3 (tm3468); slx-4 was increased to 90% while lem-3 (mn155); slx-4 worms were 100% embryonic lethal (Fig 1). We employed lem-3 (mn155) for our genetic analysis since it is a null allele for lem-3, and generated double mutants with slx-1, mus-81 and xpf-1: 100% of eggs laid by the slx-1 lem-3 double mutants failed to develop. 100% embryonic lethality was also observed in broods laid by mus-81 lem-3 double mutants (Fig 1). In contrast, the lethality of lem-3; xpf-1 double mutants was not different from LEM-3/Ankle1 dependent processing of meiotic recombination intermediates xpf-1 mutants (Fig 1). In summary, these genetic data support two potential roles for LEM-3 to maintain embryonic viability: one parallel to SLX-4 and another parallel to MUS-81 and SLX-1.

Meiotic chromosome axis formation is normal in lem-3 mutants
We next wanted to test if the synthetic phenotypes we observed were linked to defects in meiotic chromosome axis formation, which plays a central role in organization and dynamics of meiotic chromosomes. In C. elegans, meiotic prophase progression, which occurs in a gradient of differentiation, can be visualized using dissected germlines. At the distal end of the gonad germ cell mitotically divide, before entering the transition zone where meiotic chromosomes reorganise into arrays of chromatin loops anchored to the chromosome axis [40]. The chromosome axis establishes a platform for homologous chromosome paring, DSB induction, synaptonemal complex assembly, and CO formation [41]. Highlighting the axial element HTP-3 indicated that chromosome axis formation occurred normally in lem-3 and slx-4 single mutants and lem-3; slx-4 double mutants (S1 Fig).

LEM-3 is required for processing of meiotic recombination intermediates
Since there are no overt defects in chromosome axis formation, we assessed whether the synthetic lethality we observed was due to a defect in meiotic recombination. In C. elegans, unrepaired DSBs activate checkpoints that induce apoptosis of late pachytene stage germ cells [42]. Directly scoring for the number of apoptotic corpses by DIC (differential interference contrast) microscopy revealed that apoptosis was increased in mus-81 worms, further increased in slx-4 worms, and that the highest level of apoptosis occurred in both mus-81 lem-3 and lem-3; slx-4 double mutant worms (Fig 2A). Apoptosis was reduced in mus-81 lem-3; spo-11 triple mutants (Fig 2A). Careful examination of DAPI stained germ cell nuclei by fluorescence microscopy revealed that pyknotic cells, which have abnormally condensed nuclei, became apparent in mid/late pachytene in the mus-81 single mutant and to a larger extent in the mus-81 lem-3 and lem-3; slx-4 double as well as the mus-81 lem-3; spo-11 triple mutant ( Fig 2B). In mus-81 lem-3; spo-11 triple mutant, some pyknotic cells were already evident in the transition zone and early pachytene stages, where apoptotic cells are not apparent based on morphology under DIC microscopy [42,43]. Given that mutation of spo-11 does not fully bypass excessive apoptosis, some DNA lesions in mus-81 lem-3 double mutants might be independent of the meiotic DSBs.
We next wished to analyse key recombination intermediates in lem-3 single and mus-81 lem-3 double mutant worms compared to wild type. During meiosis, an excess of meiotic DSBs are generated but most breaks are repaired without leading to a CO outcome and generally only one break for each chromosome pair matures into a CO-designated site in C. elegans [44]. RAD-51 foci mark early recombination intermediates engaged in strand invasion. RAD-51 foci accumulate in the transition zone where meiotic DSBs are initiated ( Fig 3A, zone 3) and peak in early pachytene ( Fig 3A, zone 4) [45]. We found that the number of RAD-51 foci was comparable to wild type in mus-81 lem-3 double mutant worms, despite an increase of RAD-51 foci in mus-81 and lem-3 single mutants ( Fig 3A). RMH-1 foci label both CO and NCO recombination intermediates and appear and disappear later than RAD-51, consistent with RMH-1 marking recombination intermediates after strand invasion [17]. We found that both lem-3 and mus-81 single mutants showed a wild type level of RMH-1 foci in early pachytene (with an average of 11 foci) while the number of foci seen in mid-pachytene in lem-3 (on average 14.8 foci) and mus-81 (12.9 foci) was higher than wild type (10.8 foci) (Fig 3B-3D), indicative of delayed recombination intermediate processing. Increased RMH-1 foci numbers were also observed in xpf-1 single mutants and mus-81; xpf-1 double mutants, in which, respectively, one or two redundant pathways needed to resolve meiotic HJ are compromised (Fig 3B-3D) [26][27][28], In contrast, the number of RHM-1 foci in mus-81 lem-3 double mutants was significantly lower than wild type throughout the pachytene stage (Fig 3B and 3C). The number of recombination foci reflects the number of meiotic DSBs and the kinetics of their processing. Thus, our data suggest that LEM-3 and MUS-81 act in a similar step of DSB repair to ensure the proper maturation and/or turnover of meiotic recombination intermediates.
We also tested whether CO designation occurs normally using a strain expressing a functional GFP::COSA-1 fusion in mus-81 lem-3 and lem-3; slx-4 double mutant worms. COSA-1 foci mark CO designated sites in late pachytene [12]. As previously reported for wild type, slx-4 and mus-81 single mutants, also only 6 CO designated sites are apparent in the lem-3 single and lem-3; slx-4 double mutants in late pachytene ( Fig 4A and 4B), indicating that CO designation is not perturbed. We note that while the majority of nuclei display 6 COSA-1 foci, a small number of nuclei with 7 COSA-1 foci (5/131 nuclei, p = 0.503 compared with 3/127 in wild type, not significant) and 8 COSA-1 foci (4/131 nuclei, p = 0.0477 compared with 0/127 in wild type) can be observed in mus-81 lem-3 nuclei. In the wild type, COSA-1 protein develops as prominent foci at late pachytene and gradually dissociates from chromosome pairs in diplotene [12]. While COSA-1 foci fully disassemble in wild type and lem-3 single mutant before the -3 oocyte (Fig 4C and 4E), COSA-1 foci could still be detected in 50% of -3 oocytes in mus-81 single mutants (n = 12). Strikingly, 45% of the -2 oocytes (n = 22) in mus-81 lem-3 double mutants displayed COSA-1 foci, which eventually disappeared in -1 oocytes (Fig 4C and 4E). Persistent COSA-1 foci were also observed up to the -1 oocytes in slx-4 single and lem-3; slx-4 double mutants (Fig 4C and 4E). The COSA-1 foci localized to the chromosome linkages between dissociated bivalents, which could have been destined to become a CO (Fig 4D). Taken together, these data suggest that compromised recombination intermediate processing in mus-81 lem-3 double mutant leads to a delay in the dismantling of COSA-1 foci.
The morphology and number of diakinesis chromosomes can serve as readout for meiotic recombination defects. In diakinesis, homologous chromosome pairs restructure to form bivalents, which can be observed as 6 DAPI-stained bodies in wild type maturing oocytes [7]. Defects in meiotic recombination can result in a failure to stably connect homologous chromosomes, which becomes apparent as univalents at diakinesis (12 DAPI-stained bodies when physical linkages between all six homologue pairs fail to form). Our prior analysis of slx-4, as well as mus-81; xpf-1 double mutants, revealed a distinct phenotype [26]. In contrast to spo-11, pairs of 'univalents' were found associated with each other, being linked by SPO-11 dependent chromatin bridges. We termed these structures as´dissociated bivalents´. We interpreted these structures as chromosome pairs that engage in meiotic recombination but do not resolve recombination intermediates thus leading to the linkage of maternal and paternal chromosomes. As expected, wild type, lem-3 and mus-81 single mutants predominately had 6 bivalents ( Fig 5). In contrast, mus-81 lem-3 double mutant showed elevated numbers of dissociated bivalents (Fig 5, red arrow) and chromosome fragments (Fig 5, red arrowhead), which can be interpreted as unrepaired meiotic DSBs. The dissociated bivalents can also be detected in slx-1 lem-3 mutants (S2 Fig). Importantly, the analysis of slx-1 lem-3; spo-11 and mus-81 lem-3; spo-

Fig 3. Comparison of RAD-51 and RMH-1 foci localization in wild type, lem-3, mus-81 single mutants and mus-81 lem-3 double mutant. (A)
Quantification of RAD-51 profiles over the course of meiotic prophase. C. elegans gonads were divided into seven equal zones. We determined the number of RAD-51 foci in each zone. Quantifications were done based on three representative gonads for each genotype. Error bars are standard error of the mean. (B) Quantification of RMH-1::GFP foci in wild type, lem-3, mus-81, xpf-1,mus-81 lem-3 and mus-81; xpf-1 mutants in early Pachytene, mid Pachytene and late Pachytene stages. Quantifications were done for three gonads per genotype. Asterisks indicate statistical significance as determined by student T test. P Values below 0.05 were considered as significant, p < 0.05 is indicated with Ã , p < 0.01 with ÃÃ , p < 0.005 with ÃÃÃ and p < 0.0001 with ÃÃÃÃ . (C) Representative images of gonads stained with DAPI. Yellow numbers represent those nuclei with RMH-1::GFP foci between 1 and 10, pink numbers represent nuclei with RMH-1::GFP counts between 11 and 25, and white numbers represent nuclei with RMH-1::GFP foci above 25

Elimination of lem-3 does not lead to reduced CO recombination
The rate of CO recombination is reduced in slx-1; xpf-1 and mus-81; xpf-1 double mutants by approximately one third [26]. Given that slx-1 lem-3, mus-81 lem-3 and lem-3; slx-4 double mutants only sire dead embryos, we wanted to examine whether CO recombination is abolished in those double mutants. CO frequency and distribution can be investigated by meiotic recombination mapping [46]. We generated the lem-3 single and double mutants with chromosome V being heterozygous for the Hawaiian and Bristol backgrounds. To determine the recombination frequency and distribution we employed five single nucleotide polymorphisms (snip-SNPs), which together cover 92% of chromosome V [26]. Given the lethality of various lem-3 double mutants, embryos were used for recombination mapping to avoid biasing the analysis on the basis of viability. We found that CO recombination rates were comparable to wild type in lem-3 single mutants (S3 Fig). Furthermore, when analysing the respective compound mutants, we found that lem-3 did not lead to a decreased CO rate in conjunction with mus-81 or slx-4 single mutants (S3 Fig). Thus, despite the chromatin linkages that occur in various lem-3 double mutants, CO recombination is not reduced.

LEM-3 acts in conjunction with MUS-81 to prevent illegitimate recombination
We next investigated if LEM-3 was involved in inter-sister recombination repair pathway. Depletion of SYP-2, a component of the synaptonemal complex, abolishes the inter-homolog recombination and leads to the formation of 12 univalents, since breaks are likely repaired by using the sister chromatids as repair template (S4 Fig) [47]. If LEM-3 was able to promote inter-sister repair, lem-3; syp-2 double mutants would be expected to have an increased number of DAPI-stained bodies at diakinesis; a phenotype indicative of chromosome fragmentation. However, we observed an average of 11.4 DAPI-stained bodies in lem-3; syp-2 double mutants (S4 Fig). Furthermore, when we blocked the formation of the synaptonemal complex in mus-81 lem-3 double mutants by syp-2 RNAi, the average number of DAPI staining bodies did not increase (12.1, n = 19, p = 0.102 >0.05 compared to syp-2 mutant). This result indicates that LEM-3 and MUS-81 are not involved in inter-sister recombination, or that a role in the resolution of inter-sister recombination intermediates is masked by a redundant pathway.
In C. elegans, COs trigger the restructuring of bivalents in CO distal and CO proximal domains revealed by the differential location of phospho-histone H3 (pH3) and the synapsis protein SYP-1 to the short arm, and the axial proteins LAB-1 (Long Arm of the Bivalent) and HTP-1/2 to the long arm of the 6 bivalents [48]. The lack of CO recombination as is the case in spo-11 mutants, leads to the formation of 12 univalents without such bifurcation. In contrast, "peculiar univalents" display the same reciprocal localization as bivalents, suggesting that a CO site has been designated leading to enrichment of the above-mentioned markers to apposed chromosomal domains [17,26,36]. 'Peculiar univalents' were previously observed in rmh-1, him-6 and xpf-1 single mutants [17,26,36]. We found an increased prevalence of univalents in rmh-1 mus-81 lem-3 triple mutants compared to rmh-1 single mutants, as revealed by an increased number of DAPI-stained bodies that display the features of "peculiar univalents" (Fig 6A and 6B, Please note, most quantification of univalents was done by DAPI staining, due to limited amounts of reagents to determine domain organization). Univalents were not detected in the lem-3 or mus-81 single mutants as well as in mus-81 lem-3 double mutants ( Fig  6C) as evidenced by the 6 DAPI stained bodies we observed. In contrast, rmh-1 mus-81 and rmh-1 lem-3 double mutants both showed an average of 8 DAPI stained bodies compared to the average of 10 observed in the rmh-1 mus-81 lem-3 triple mutant (Fig 6C and S5 Fig, clearly discernible univalents are highlighted). Thus, the increased prevalence of univalents in rmh-1 mus-81 lem-3 triple mutant might be the result of compromising several parallel recombination pathways.
We next set out to test if these univalents result from the mis-direction of recombination intermediates towards a NCO pathway such as inter-sister repair or SDSA. In C. elegans, the BRC-1 homologue of the mammalian BRCA1 recombination protein, which forms a heterodimer with BRD-1, has been proposed to be important for inter-sister repair in the germline [49]. Indeed, blocking inter-sister repair by introducing a brd-1 mutation into rmh-1 mus-81 lem-3 triple mutants resulted in a reduced number of univalents as revealed by an average of 8.5 DAPI stained bodies (Fig 6C and S5 Fig). The rmh-1 mus-81 lem-3; brd-1 quadruple mutants also displayed a reduction of RAD-51 compared to rmh-1 mus-81 lem-3 triple mutants, indicative of altered processing of meiotic recombination intermediates or reduced HR repair (Fig 6D). Altogether, these results indicate that LEM-3 might act in conjunction with MUS-81 and RMH-1 to process early recombination intermediate and the simultaneous absence of LEM-3, RMH-1 and MUS-81 could lead to illegitimate recombination intermediates impeding CO formation.

LEM-3 promotes proper chromosome segregation during meiotic cell division
We next investigated the localization of LEM-3 using a strain expressing a GFP::LEM-3 fusion. Consistent with previous reports [37,50], we found that LEM-3 localized as dots outside of the nucleus in the mitotic germ cells of wild type worms (Fig 7A). LEM-3 foci (typically no more than one per cell) were occasionally observed in pachytene, both in and outside of the nucleus (Fig 7B). These LEM-3 foci did not co-localize with the ZHP-3 marker that congresses into CO designated sites (Fig 7C) [51]. Interestingly, careful examination of cells undergoing meiotic divisions revealed that LEM-3 localized between dividing nuclei in meiosis II (Fig 7D and  7E). To analyse whether LEM-3 has a role in meiotic chromosome segregation, we performed live cell imaging of the first and second meiotic cell divisions by using an integrated Histone mCherry::H2B fusion. We reasoned that the chromosome segregation in meiosis I and II might be affected if a chromosome linkage remains present between two homologues, or sister chromatids, respectively. Chromosome segregation in the lem-3 single mutant was similar to wild type (Fig 7F, S1 and S2 Movies). As we had previously reported for double mutants affecting both the SLX-1/MUS-81 and the HIM-6/XPF-1 pathway, chromosome linkages appeared during the first meiotic division in slx-4 mutants (9/14 embryos, Fig 7F and 7G), consistent with an important role for SLX-4 in resolving inter-homolog recombination intermediates (Fig 7E and 7G, S3 Movie) [26] [48]. While the chromosome linkage could only be detected in 28.6% of slx-4 mutant embryos (4/14) during the second meiotic division, all lem-3; slx-4 double mutant embryos (19/19) showed extensive chromosome linkage formation in meiosis II (Fig 7F and 7G, S4 Movie). These data indicate that LEM-3 might have a role in processing recombination intermediates that persist into the second meiotic division.

Discussion
In this study, we investigated the interplay between the C. elegans LEM-3/Ankle1 nuclease and nucleases previously implicated in meiotic CO resolution. We provide evidence for two roles of LEM-3 during meiosis. First, LEM-3 acts in conjunction with the MUS-81 and SLX-1-SLX-4 nucleases to process various recombination intermediates during meiotic prophase. Second, LEM-3 functions as a backup nuclease to deal with persistent DNA linkages during meiotic divisions.
https://doi.org/10.1371/journal.pgen.1007453.g006 CO intermediates in the absence of CO resolvases such as MUS-81. In contrast, the lem-3; xpf-1 double mutant showed no synthetic lethality, indicating that LEM-3 is able to process specific aberrant recombination intermediates that arise in mus-81 mutants. However, the lem-3 mutation in combination with slx-4 or mus-81 did not lead to a further reduction of CO frequency, or to an altered CO distribution (S3 Fig). Thus, synthetic lethality of the mus-81 lem-3 mutant is not a result of decreased CO formation.
RMH-1 foci label both CO and NCO recombination intermediates. A previous study showed that RMH-1 does not colocalize with RAD-51 and appears and disappears later than RAD-51, suggesting that RMH-1 might act after RAD-51 removal and mark late recombination intermediates [17]. We found that the number of RMH-1 and RAD-51 foci was increased in lem-3 and mus-81 single mutants compared to wild type (Fig 3A and 3B), suggesting that both LEM-3 and MUS-81 play a role in the proper maturation/turnover of recombination intermediates. The increased number of RMH-1 and RAD-51 foci could be due to an increased number of processed DSBs, or due to a delay in DSB processing. Interestingly, the number of RAD-51 foci was comparable to wild type in the absence of both MUS-81 and LEM-3 (Fig 3A), indicating that the increased RAD-51 foci in mus-81 single mutants is depended on the activity of LEM-3 and vice versa. In contrast, the number of RMH-1 foci was decreased in the mus-81 lem-3 double mutant in early and mid pachytene compared to wild type ( Fig 3B). These data suggest that MUS-81 and LEM-3 individually have a function in processing recombination intermediates and appear to act in conjunction after strand invasion but before the formation of RMH-1 foci. Irrespective, these alterations in the kinetics of RAD-51 or RMH-1 foci reflect the production and/or processing of early meiotic intermediates but do not overtly affect the number of CO designated sites. The number of COSA-1 foci that mark CO designated sites is roughly normal in the mus-81 lem-3 double mutant (Fig 4A and  4B). Only in 3% of mus-81 lem-3 mutant worms (4/131 nuclei) the number of COSA-1 foci is increased to 8, a finding which may indicate that the mechanisms leading the restriction to one CO per chromosome might be occasionally overwhelmed by unusual recombination intermediates occurring in this double mutant. Alternatively, this slightly increased number of COSA-1 foci could be linked to cells that become pyknotic and have abnormally condensed chromosomes and a higher number of smaller COSA-1 foci. Overall, our data suggest that LEM-3 and MUS-81 can process early recombination intermediates and contribute to the formation of recombination intermediates that recruit RMH-1.
How can we explain that mus-81 lem-3 stains have dissociated bivalents but do not have a reduced rate of CO recombination? We interpreted these structures as chromosome pairs that engage in meiotic recombination but do not resolve recombination intermediates thus leading to the linkage of maternal and paternal chromosomes [26]. These structures were previously observed in slx-4 single mutant worms, as well as in mus-81; xpf-1 double mutants, and the CO rates were reduced as we expected [26,48]. We favour two possible explanations for this discrepancy. DNA linkages might define CO intermediates that can still result in a NCO outcome. For instance, in the classical HJ resolution model, 4-way joint molecules that link maternal and paternal chromosomes can be resolved depending on the symmetry of the cleavage leading to CO or NCO outcome. Alternatively, it might as well be that chiasmata, the LEM-3/Ankle1 dependent processing of meiotic recombination intermediates structures that hold maternal and paternal chromosomes together at the CO site, could be weakened in mus-81 lem-3 double mutants, resulting in their dissociation. This dissociation could equally lead to a 'dissociated bivalent' phenotype. If such dissociation occurs, one does not expect reduced CO recombination. Both hypotheses are consistent with the delayed dissolution of COSA-1 foci.
How might the activity of LEM-3 be involved in processing meiotic recombination intermediates? It has been reported that LEM-3 and its human homologue ANKLE-1 are able to cleave supercoiled plasmid into relaxed circular (nicked) and linear DNA [37,38]. In addition, LEM-3 can cleave a DNA substrate that is rich in secondary structures [37], indicating that LEM-3 might be a structure-specific endonuclease. Therefore, it is possible that LEM-3 can either process early recombination intermediates, such as D-loops, or else act at a late stage of meiotic recombination for HJ resolution to generate NCO products. It will be interesting to investigate the DNA substrate preference of LEM-3 in the future.
Chromosome segregation can be affected if unresolved recombination intermediates remain present during meiotic divisions. We previously observed that the extensive chromatin bridges generated by depleting both XPF-1 and MUS-81 pathways during the first meiotic division are eventually resolved in meiosis II, suggesting that backup activities function at or after anaphase I [26]. Here we found that LEM-3 localises between dividing nuclei during meiotic division (Fig 7D). In addition, depletion of LEM-3 leads to accumulation of chromosome linkages in the slx-4 mutant, especially during meiosis II (Fig 7F), indicating that LEM-3 might have a role in proper chromosome segregation, by directly processing DNA linkages caused by unresolved recombination intermediates. A similar function has also been reported for the HJ resolvase GEN1/YEN1 nuclease in S. cerevisiae [52]. The budding yeast yen1 mutant does not have obvious meiotic defects, whereas the mus81 yen1 double mutant fails to segregate its chromosomes due to unresolved DNA joint-molecules during meiotic anaphase I and II [53]. Furthermore, the enzymatic activity of YEN1 is kept at a very low level in prophase but is highly induced at the onset of meiosis II, suggesting that it provides a safeguard activity that processes DNA linkages that escape the attention of MUS81 during meiotic divisions [53]. Mutation of the YEN1/GEN1 nuclease shows phenotypic variation in different organisms [21]. In C. elegans, no meiotic phenotypes were observed in the gen-1 single mutant on its own or in combination with him-6 or various nuclease mutants [54]. Our data suggest that LEM-3 might provide a failsafe system in C. elegans instead of GEN-1, to ensure that all recombination intermediates are resolved at the final stage of gamete formation. Consistent with this idea we recently provided evidence that LEM-3 might bind to chromatin bridges in the contractile ring in mitotic cells to process a large variety of DNA intermediates linked to recombination failure, DNA catenation, DNA decondensation failure, and to DNA underreplication [50] In summary, we provide evidence for a role of LEM-3 in meiotic recombination intermediate processing in prophase I and in resolving persistent chromatin bridges during meiotic divisions. It will be interesting to see whether the mammalian LEM-3 orthologue Ankle1 has a role in meiosis.

C. elegans strains and maintenance
Strains were grown at 20˚C followed standard protocols [55]. N2 Bristol was used as the wild type. CB4856 Hawaii was used to generate strains for CO recombination frequency analysis. Strains used in this study are listed in S1 Table. The cop859 [Plem-3::eGFP::STag::lem-3::3 0 UTRlem-3] eGFP insertion was generated by Knudra (http://www.knudra.com/) following the procedures described by Dickinson et al [56]. Exact details are available upon request.

Recordings of meiotic divisions
Meiotic divisions were recorded by in utero embryo live imaging [57]. Worms were picked into a solution containing 1 mM levamisole to paralyze worms. Worms were mounted on 2% agar pads covered with a coverslip. Images were captured every 10 seconds using spinningdisk confocal microscopy.

Image acquisition and analysis
Microscopy images acquired with a Delta Vision microscopy were deconvolved and analysed using softWoRx Suite and softWoRx Explorer software (AppliedPrecision, Issaquah, WA, USA). Images acquired with a spinning-disk confocal microscopy were analyzed by ImageJ software.

RNA interference
RNAi was performed by feeding worms with bacteria containing plasmid that express doublestranded RNA for syp-2 [58]. Worms were fed on NGM plates supplied with 100 mg/L ampicillin and 1mM IPTG. An empty L4440 plasmid was used as a control for RNAi experiment.

Determining meiotic crossover frequency and distribution
Meiotic CO frequency and distribution were assayed essentially as described [26] with slight modifications. Five snip-SNPs on Chr. V that differ between N2 Bristol and CB4856 Hawaii were used to determine the crossover landscape in embryos. Single embryo was transferred into lysis buffer by mouth pipetting using a capillary and incubated at -80˚C for at least 5 min to help crack the embryo before lysis.