Caenorhabditis elegans RMI2 functional homolog-2 (RMIF-2) and RMI1 (RMH-1) have both overlapping and distinct meiotic functions within the BTR complex

Homologous recombination is a high-fidelity repair pathway for DNA double-strand breaks employed during both mitotic and meiotic cell divisions. Such repair can lead to genetic exchange, originating from crossover (CO) generation. In mitosis, COs are suppressed to prevent sister chromatid exchange. Here, the BTR complex, consisting of the Bloom helicase (HIM-6 in worms), topoisomerase 3 (TOP-3), and the RMI1 (RMH-1 and RMH-2) and RMI2 scaffolding proteins, is essential for dismantling joint DNA molecules to form non-crossovers (NCOs) via decatenation. In contrast, in meiosis COs are essential for accurate chromosome segregation and the BTR complex plays distinct roles in CO and NCO generation at different steps in meiotic recombination. RMI2 stabilizes the RMI1 scaffolding protein, and lack of RMI2 in mitosis leads to elevated sister chromatid exchange, as observed upon RMI1 knockdown. However, much less is known about the involvement of RMI2 in meiotic recombination. So far, RMI2 homologs have been found in vertebrates and plants, but not in lower organisms such as Drosophila, yeast, or worms. We report the identification of the Caenorhabditis elegans functional homolog of RMI2, which we named RMIF-2. The protein shows a dynamic localization pattern to recombination foci during meiotic prophase I and concentration into recombination foci is mutually dependent on other BTR complex proteins. Comparative analysis of the rmif-2 and rmh-1 phenotypes revealed numerous commonalities, including in regulating CO formation and directing COs toward chromosome arms. Surprisingly, the prevalence of heterologous recombination was several fold lower in the rmif-2 mutant, suggesting that RMIF-2 may be dispensable or less strictly required for some BTR complex-mediated activities during meiosis.


Introduction
Damage-induced DNA double-strand breaks (DSBs) pose a threat to genome integrity. Highfidelity repair via homologous recombination (HR) is employed during both mitotic and meiotic cell cycles. It involves the generation of 3 0 overhang ends by DNA resection and their stabilization by the single-stranded DNA-binding protein RPA (replication protein A) (RPA-1 in worms). RPA-1 is subsequently exchanged with the RAD-51 recombinase to allow invasion of a homologous DNA strand, giving rise to a D-loop intermediate structure. After DNA synthesis and second-end capture, DNA joint molecules are generated. These can be processed to produce crossovers (COs), which result in the reciprocal exchange of large regions of chromosomes [1]. In meiosis, where Spo11-mediated DSBs are induced via a highly regulated program, crossing-over and cohesion establish a physical tether between homologous chromosomes, which greatly aids their correct segregation in meiotic anaphase I and drives genetic variability. In meiosis, one chromatid of the homologous chromosome is preferentially used as a repair template for HR. Joint DNA molecules must disengage in order to segregate, and this is achieved by redundant endonucleases (called resolvases) and the BTR complex [2]. Depending on the orientation of the resolvase-induced cut, the outcome is a CO or non-crossover (NCO) product. To ensure at least one CO per chromosome pair, excess DSBs are introduced and those that do not form the CO are repaired to form NCOs [2].
The BTR complex dismantles joint DNA molecules via its decatenation activity, which has been reconstituted in vitro [3]. In decatenation, strand passage is achieved via cutting one DNA strand and then resealing the DNA break. Following HR in mitotically dividing cells, the BTR complex mostly mediates the NCO outcome since COs can have detrimental effects such as loss of heterozygosity [3]. For example, loss of heterozygosity of a tumor suppressor gene can lead to cancer development. Patients with mutations in components of the BTR complex show elevated rates of sister chromatid exchange and aberrant chromosomes [4,5].
In mammals, the BTR complex consists of Bloom helicase, topoisomerase, and the RMI1 and RMI2 scaffolding proteins. In worms, the respective homologs are HIM-6, TOP-3, and RMH-1 and RMH-2 (both RMH proteins are RMI1 homologs-an RMI2 homolog has not been identified). Structural and biochemical analysis of RMI1 suggest its involvement in the strand passage-tyrosine transesterification reaction mediated by the topoisomerase. Based on these activities combined with the DNA unwinding activity of the helicase, the BTR complex has important roles in DNA metabolism [6,7].
Detailed analyses of the homologous yeast STR complex (Sgs1 helicase-topoisomerase-Rmi1) have revealed that its meiotic functions include an important role in D-loop reversion in vivo [8,9], which has also been shown in vitro [10]. D-loop reversion prevents the generation of complex multi-joint molecules; in the absence of STR activity, multi-joint molecules can only be resolved by non-canonical resolvases, which generate a mix of COs and NCOs (i.e. additional COs are formed). It was also observed that Top3-Rmi1 form a sub complex that limits the accumulation of toxic recombination intermediates. Loss of function of both Rmi1 and topoisomerase 3 leads to meiotic catastrophe, due to persistent joint molecules that are resistant to cleavage by resolvases. It is conceivable that these DNA structures represent extended D-loops involving homologous and/or heterologous chromosomes or other branched structures (for a review, see [11]). The Caenorhabditis elegans Bloom ortholog HIM-6, similar to BRC-1/BRCA1, suppresses heterologous recombination in the germline, which could lead to translocations and genome rearrangements [12].
In mammalian cells, RMI2 was identified as an RMI1-or Bloom-interacting protein [27,28]. Like RMI1, it contains a characteristic OB-fold domain (OB, oligonucleotide/oligosaccharide binding), and RMI1 and RMI2 interact via their OB domains (OB2 in RMI1 and OB3 in RMI2). RMI2 is required to stabilize the other members of the complex within recombination foci, and is also suggested to function in governing post-translation modifications of other complex members [27]. Upon RMI2 depletion, elevated sister chromatid exchange and chromosome aberrations have been observed [28].
So far, RMI2 orthologs have not been identified in yeast, Drosophila, or C. elegans. Here we report the identification of a novel C. elegans protein encoded by the open reading frame Y104H12D.4, which we found in RMH-1-containing protein complexes in the germline. Similar to RMI2 proteins, Y104H12D.4 contains an OB-fold domain. Based on its ability to stabilize RMH-1 (worm RMI1) and concentrate HIM-6 (Bloom helicase) and topoisomerase 3 into recombination foci, it qualifies as a functional homolog of RMI2. Thus, we named the protein RMI2 functional homolog-2 (RMIF-2). Similar to the other BTR complex proteins, RMIF-2 displays a dynamic localization pattern in recombination foci. Nevertheless, our detailed analysis of germline recombination revealed marked differences between rmh-1 and rmif-2 mutants, indicating that rmif-2 functions not just as RMH-1 stabilizer for all its activities in the germline as it is observed in mammalian cells. Our data suggest, that without RMIF-2 RMH-1 can function in some of the BTR related meiotic activities.

Identification of Y104H12D.4 as an interaction partner of RMH-1
To purify RMH-1 meiotic interaction partners, we tagged the endogenous rmh-1 locus with a 5 0 HA-degron-tag using CRISPR/Cas9. The tagged line has normal hatching rates (comparable to those of the GFP-tagged line, which we published previously [19]), indicating its full functionality. Biochemical fractionation [29] and western blot analysis showed that the RMH-1 protein is enriched in both nuclear soluble and insoluble fractions. To identify RMH-1 interactors, we used pooled nuclear fractions in immunoprecipitation experiments followed by mass spectrometry analysis. In triplicate experiments, several interaction candidates (Table 1) were found to be enriched over control levels, including RPA-1 and other members of the BTR complex: HIM-6 and TOP-3. The novel open reading frame Y104H12D.4 (RMIF-2) was also identified as an RMH-1 interactor.
Despite the presence of RMI2 family members in other nematodes, the C. elegans proteome does not contain a Pfam RMI2 hit (see Materials and Methods for details of the bioinformatics analysis). Of the co-purified proteins, RMIF-2 was the only candidate with a predicted OBfold domain to be identified with methods to find remote homologs (Fig 1A). A ribbon diagram of the human RMI core complex and a model of the putative C. elegans RMIF-2 OB-fold can be found in S1 Fig. In this report, we will present evidence that RMIF-2 is a true functional homolog of RMI2.
To confirm the interaction between RMH-1 and RMIF-2, we generated a functional rmif-2::3xflag tagged line (Table 2), which we combined with ha::rmh-1 [30]. We used this strain for reciprocal co-immunoprecipitation experiments. Western blot analysis of FLAG pull downs revealed robust co-immunoprecipitation of HA::RMH-1, confirming that these two proteins form a complex in vivo ( Fig 1B).
To examine the role of rmif-2 in meiotic DNA repair, we analyzed the formation and processing of meiotic recombination intermediates by monitoring the dynamic appearance and disappearance of the RAD-51 recombinase [33,34]. For this, we divided the gonads into seven equal zones and quantified the RAD-51 foci per nucleus in each zone (Fig 2C). In the WT, RAD-51 foci begin to accumulate in the transition zone and peak in mid pachynema; in late pachynema, the foci disappear as a consequence of successful repair. In rmif-2, RAD-51 foci appeared with similar dynamics as in the WT. However, greater numbers of foci accumulated and persisted throughout early/mid pachynema ( Fig 2C) with a significantly different increase in zones 3-5; by late pachynema most of the foci had disappeared, and at diplonema no RAD-51 signal remained. Overall, DSB repair in rmif-2 was delayed but accomplished eventually. We also found that aberrant RAD-51 accumulation was SPO-11 dependent (as shown in the rmif-2 spo-11 double mutant; Fig 2C). There was no significant difference between the RAD- Table 2. Offspring analysis including embryonic lethality, brood size and segregation of male progeny in the genotypes used in this study.

PLOS GENETICS
The C. elegans RMI2 functional homolog-2 (RMIF-2) in meiosis 51 foci counts in the spo-11 and rmif-2 spo-11 mutants. This indicates a specific defect in the processing of meiotic recombination intermediates in rmif-2. Concomitant with recombination, chromosome pairing is initiated soon after meiotic entry [32]. We used the X chromosome-binding protein HIM-8 as a specific tool to analyze X chromosome pairing [35]. In the rmif-2 mutant, X chromosome pairing reached WT levels in pachynema (S2A Fig), although with slower kinetics up to early pachynema. We assume that this is a consequence of the slightly extended mitotic zone in the rmif-2 mutant (mean 23.4 (± 3.0 SD) cell rows versus 20.4 (± 2.6 SD) in the WT; S2B Fig). Progressive co-localization of the HTP-3 chromosome axis marker and SYP-1 synapsis marker serves as a read-out for synaptonemal complex formation [36]. Using this read-out, no major defects in synapsis were detected in the rmif-2 mutant (S2C Fig).
Taken together, the phenotypic data reveal that rmif-2 mutants display aberrant recombination and univalent diakinesis chromosomes, which goes in hand with an increased embryonic lethality and X chromosome non-disjunction. All of these phenotypes have also been reported for rmh-1 mutants, although to differing degrees. In rmh-1 mutants, both the number of univalents and the degree of embryonic lethality are higher (Fig 2A and 2B, and Table 2) and [19].

RMIF-2 is required to concentrate the other BTR complex proteins into recombination foci
In tissue-cultured cells, RMI2 is required to localize the Bloom helicase into recombination foci [27], and RMI2 interaction with the other BTR complex proteins is mediated by the OBfold domain [28]. We therefore wanted to examine whether the correct localization of all BTR complex proteins depends on RMI2. We first analyzed GFP::RMH-1 localization in the rmif-2 mutant. In the gfp::rmh-1, faint RMH-1 foci are seen throughout early pachynema, and become brighter from mid to late pachynema ( Fig 3A) [19]. GFP::RMH-1 was not detectable in the rmif-2 mutant: only faint cytoplasmic foci were occasionally seen (Fig 3A), suggesting that RMIF-2 is essential for RMH-1 localization into discrete chromatin-associated foci. To address whether RMIF-2 stabilizes the RMH-1 protein, we isolated germline-enriched nuclei, as described in [29], followed by subcellular fractionation and western blot analysis (Fig 3B  and 3C). RMH-1 was enriched in both the soluble and insoluble nuclear fractions in the ha:: rmh-1 (Fig 3B and 3C). In the absence of RMIF-2, the soluble fraction contained less RMH-1 and the insoluble fraction was below levels of detection (Fig 3B and 3C). Next, we addressed the reciprocal question of whether RMH-1 stabilizes the RMIF-2 protein. In rmh-1 mutants, RMIF-2 foci were undetectable ( Fig 3D). Western blot analysis of whole-cell extracts showed that RMH-1 stabilizes RMIF-2 ( Fig 3E): the quantity of RMIF-2 protein was three times lower in the rmh-1; rmif-2::ha mutant than in the rmif-2::ha ( Fig 3F). These data indicate a reciprocal requirement for RMH-1 and RMIF-2 for their recruitment into recombination foci and for protein stability.

PLOS GENETICS
The C. elegans RMI2 functional homolog-2 (RMIF-2) in meiosis rmif-2(jf113) pachytene nuclei stained with DAPI (magenta) and GFP (green). GFP::RMH-1 localization to nuclear foci starts in early pachynema, peaks in mid pachynema, and becomes concentrated in six foci in late pachynema. In the rmif-2 mutant background, RMH-1 fails to localize into foci throughout pachynema, except for a very few cytoplasmic foci. Scale bar: 10μm. (B) A protein fractionation shows specific HA::RMH-1 enrichment in the nucleus, which is reduced in the rmif-2 mutant. Equal amounts of protein were loaded for each fraction. C = cytosolic fraction, NS = soluble nuclear fraction, IN = insoluble nuclear fraction. LMN-1 was the loading control for nuclear fractions; GAPDH was the loading control for the cytosolic fraction. (C) Western blot normalization and quantification of foci throughout pachynema, but these foci appeared smaller and fainter in the absence of RMIF-2 ( Fig 4A) [19]. Previously it was shown that in the rmh-1 mutant, HIM-6 foci were present throughout pachynema but seemed smaller and fainter, suggesting that RMH-1 is required to stabilize and enrich HIM-6 into foci [19]. Taken together these data suggest that the initial recruitment of HIM-6 is independent of both RMIF-2 and RMH-1 [19], but both of these factors seem to be necessary for HIM-6 accumulation in recombination foci.
To visualize TOP-3, we inserted an internal OLLAS tag into the protein coding sequence (see Materials and Methods). The tagged line was functional since it had a hatching rate of 99.5% ± 0.4 SD ( Table 2) compared with 0% (i.e. 100% embryonic lethality) in the null mutant [37]. TOP-3 also localized to foci throughout pachynema, with similar dynamics and foci numbers to RMH-1 and RMIF-2 ( Fig 4B-4D). The number of TOP-3 foci peaked in mid pachynema, while in late pachynema TOP-3 appeared to be present on the putative six CO sites. TOP-3 failed to localize properly in rmif-2, with the TOP-3 signal detectable as greatly reduced, sporadic foci ( Fig 4B-4D), consistent with a requirement for RMIF-2 in TOP-3 complex stabilization. Taken together, these data suggest that constituents of the BTR complex require RMIF-2 for their proper localization into foci throughout pachynema. Although the initial recruitment of HIM-6 appears to be independent of RMIF-2, RMIF-2 seems to enhance the enrichment of HIM-6 into foci. These findings are consistent with a model in which (1) RMIF-2 stabilizes the BTR complex and (2) RMH-1 and RMIF-2 recruitment into discrete chromatin-associated foci is essential for the mutual stabilization of both proteins.
Interestingly, rmh-1 was the only mutant in the BTR complex that failed to localize MSH-5 into clearly visible foci. In both the him-6 and top-3 mutants, GFP::MSH-5 could be detected in foci, albeit with a delayed appearance and in significantly reduced numbers. In him-6 mutants, the MSH-5 signal can be seen throughout zones 4-7 (range 0-15 foci/nucleus), with an average of 5.7 (± 1.8 SD) foci per nucleus in zone 7 (Fig 6A and 6B). In the top-3 mutant, the number of MSH-5 foci was much lower (range 0-6 foci/nucleus in zones 4-7) (Fig 6B). Although their appearance was delayed, foci were visible as distinct signals in zones 4-7 (mean 2.9 (± 2.0 SD) foci/nucleus in zone 7). Taken together, our data show that in all mutants of BTR complex proteins, CO designation is less efficient as depicted by the significantly delayed and reduced appearance of COSA-1 and MSH-5 foci, with the exception of the rmh-1 mutant, where GFP::MSH-5 foci are completely absent.
Therefore, we next asked whether formation of the putative bivalent chromosomes observed in the diakinesis nuclei of rmh-1 and rmif-2 mutants (Fig 2A and 2B) depended on the canonical CO pathway [39,41]. Indeed, these chromosome structures were cosa-1 dependent in the rmif-2 mutant (S3A and S3B Fig). In contrast to the rmif-2 single mutant, in the cosa-1; rmif-2 double mutant diakinesis nuclei contained mostly univalents, as in cosa-1 single mutants. The data indicate that the diakinesis joint chromosome structures observed in rmif-2 are mostly formed via the class I CO pathway. As expected, RMIF-2 did not localize into recombination foci in the cosa-1 mutant (S3C Fig). Instead, in mid pachynema RMIF-2 was observed in very few, faint foci, with most of the signal located outside the nucleus, and in late pachynema no nuclear foci were detectable.
Although MSH-5 foci were absent in rmh-1 (where the MSH-5 signal forms a nuclear haze (Fig 6A)), we examined whether the putative bivalent chromosomes in rmh-1 were still dependent on msh-5. Analysis of DAPI bodies in the rmh-1; msh-5 double mutant (S3A and S3B Fig) revealed that many of the joint structures seen in diakinesis were indeed dependent on MSH-5. On average, 8.5 (±1.7 SD) DAPI-stained bodies were formed in rmh-1, 12 (± 0 SD) in msh-5 and 10.8 (±1.4 SD) in rmh-1; msh-5. To further analyze whether the remaining physical attachments between univalent chromosomes in rmh-1 mutants were dependent on non-homologous end joining (NHEJ) activity, we constructed the triple mutant rmh-1; cku-70; msh-5, (S3A and S3B Fig). CKU-70 is a key protein component of the NHEJ pathway [47]. Analysis of diakinesis chromosomes in the triple mutant revealed a significant increase in fragmentation, leading to on average 12 (± 0.5 SD) DAPI-stained bodies compared to the average of 6 (± 0 SD) in the cku-70 single mutant and 10.8 (± 1.4 SD) in the double mutant rmh-1; msh-5, suggesting that some of the rmh-1; msh-5 undefined structures (which seem larger than univalents) are dependent on the activity of the NHEJ pathway.

RMIF-2 and RMH-1 may not always act interdependently in meiosis
To investigate whether RMIF-2 acts as a BTR complex stabilizer by binding to and or stabilizing RMH-1, we generated the rmh-1; rmif-2 double mutant and compared its phenotype with those of the single mutants. First, we examined embryonic lethality and the number of males in the viable progeny. We observed that in the rmh-1; rmif-2 double mutant, levels of embryonic lethality were similar to those in rmh-1 mutants, 56% (± 37 SD) although with much higher variability between individual worms, with 5% (± 3 SD) of males in the progeny (Table 2; in comparison, embryonic lethality is 40% (± 4 SD) in rmif-2 and 68% (± 9 SD) in rmh-1 mutants). In the double mutant, the brood size was much smaller (mean 72 (± 83 SD)) than in the single mutants ( Table 2). We next examined DAPI-stained diakinesis chromosomes in rmh-1; rmif-2 worms. There were an average of 6.7 (± 1.5 SD) DAPI bodies per nucleus, and most nuclei contained six aberrant bodies that differed markedly from the well-shaped bivalents in the WT (Fig 2A and 2B). We saw chromatin clumps together with DAPI bodies of differing sizes that resembled DNA fragments and/or univalents (Fig 2B). We wanted to examine whether these DAPI-stained bodies were the results of NHEJ activity and thus, constructed the triple mutant rmh-1; cku-70; rmif-2 (Fig 2A and 2B). Analysis of diakinesis nuclei in this mutant background revealed no difference in the number of DAPI bodies (7.45 (± 1.6 SD)) when compared to rmh-1; rmif-2 (6.7 (± 1.5 SD)). Therefore NHEJ does not cause the chromosome abnormalities observed at diakinesis stage. We also analyzed DSB induction and repair in the double mutant rmh-1; rmif-2. Assessment of RAD-51 loading/unloading on chromosomes revealed a striking accumulation of RAD-51 foci, which persisted throughout pachynema (statistically different from the single mutants; Fig 2C). Exacerbation of the meiotic phenotypes in the rmh-1; rmif-2 double mutant suggests that either RMIF-2 or RMH-1 may have an additional function outside of the mutually dependent formation of recombination foci and/or that RMIF-2 might not merely function as a RMH-1 stabilizer within the BTR complex. It could be that RMIF-2 has a less prominent role during meiotic DSB repair.
We next generated the rmif-2 him-6 double mutant to test whether RMIF-2 was needed to support the activity of the HIM-6 helicase. In these worms, DAPI-body counts showed an increase to an average of 9.4 (± 2.0 SD) in comparison to the single mutants rmif-2, 6.9 (± 1.0 SD) and him-6, 7.5 (± 0.9 SD) (Fig 2A and 2B). In contrast, in the rmh-1; him-6 diakinesis DAPI-bodies were not statistically different in comparison to the him-6 single mutant (7.8 (± 1 SD) vs 7.3 (± 1.1 SD)) [19]. Moreover, in rmif-2 him-6 embryonic lethality strikingly increased to 99.2% (± 1.6 SD) from 41% (± 5.6 SD) in the him-6 single mutant (Table 2), while in the rmh-1; him-6 it was 68.22% [19]. The quantification of RAD-51 in the double mutant rmif-2 him-6 revealed accumulation of meiotic and mitotic recombination intermediates. RAD-51 was loaded earlier than in the WT and single mutants, from the mitotic zone (zone 1) onwards, with high and persisting numbers of unresolved breaks, which were not repaired by late pachynema (zone 7) (Fig 2C). Taken together, RAD-51 quantification revealed high levels of genomic instability in the rmif-2 him-6 animals arising from mitotic and meiotic defects. The severe phenotype of rmif-2 him-6 double mutants suggests that RMIF-2 and HIM-6 might act in independent parallel pathways. Analysis of the accumulation of recombination intermediates in the rmh-1; him-6 revealed differences in comparison to the rmif-2 him-6 double mutant. Here, RAD-51 was also observed in significantly higher numbers throughout pachynema, with accumulation of unrepaired breaks until late pachynema (Fig 2C), however no significant amount of RAD-51 signals was quantified in the mitotic zone (zones 1-2).

RMIF-2 and RMH-1 influence the recombination landscape
Based on observed differences in the localization dynamics of key recombination factors in rmif-2 and rmh-1 mutants, we performed a recombination analysis to compare the recombination rate and CO number and position in the mutants with those of the WT. For this, we introduced homozygous mutations into a C. elegans hybrid strain derived from the Bristol and Hawaii isolates. Single nucleotide polymorphisms (SNPs) on chromosomes IV and V were used for assessing COs [19,48] (Fig 7A). PCR-based analysis of SNPs on chromosomes IV and V was conducted on F2 worms (described in Materials and Methods). As previously observed for rmh-1, we did not find significant differences between the total frequencies of COs in rmif-2 compared to WT (on both chromosomes IV and V), and in rmh-1 on chromosome V (Fig 7). We observed a significant shift of COs towards the central regions of both chromosome IV and V, where CO are usually not favored in the WT (Fig 7B-7D). In addition, an increase in double and triple COs was observed in the rmif-2 in contrast to rmh-1 that displayed only few extra COs. For chromosome IV two double COs were recorded for rmif-2 (n = 364) and one for the WT (n = 281). On chromosome V nine double COs and two triple COs were observed in rmif-2 (n = 362); in contrast, only one double CO was seen in the WT (n = 269) (Fig 7D). In summary, fewer extra COs were present in the rmh-1 mutant than in the rmif-2 mutant. Taken together, the recombination assays revealed that both RMIF-2 and RMH-1 play a role in correctly positioning COs to chromosome arms (away from the chromosome center), with RMIF-2 having a more pronounced role in suppressing the formation of double and triple COs [49].
In C. elegans, HIM-6 (BLM) helicase is involved in rejecting strand invasion into heterologous sequences [12], and lack of this activity leads to genome rearrangements. We used visible phenotypic markers to assess the extent of heterologous recombination within the mIn1 inversion on chromosome II [12]: one copy of chromosome II is marked with the semi-dominant dpy-25 mutation and the second copy contains the mIn1 inversion, which is marked with the recessive rol-1 mutation and has a semi-dominant insertion of a GFP-expressing transgene (Fig 8A). In WT worms, heterologous recombination is rare (Fig 8B), as previously reported [12], and we did not detect a single event (n = 2029). In contrast, in absence of RMIF-2, there was a significant increase in progeny displaying heterologous recombination events (2.3%, n = 2018) (Fig 8B and  8C). Surprisingly, the level was roughly three times higher in the rmh-1 mutants, at 7.24% (n = 1090), consistent with the level seen in him-6 mutants (6.6%), as reported in [12]. Taken together, these data show that both rmif-2 and rmh-1 are required to prevent heterologous recombination; however, the more pronounced defect in rmh-1 mutants compared with rmif-2 worms suggests that RMIF-2 might have a less prominent role in rejecting heterologous strand invasion.

Discussion
In this study, we present a C. elegans functional homolog of the BTR complex stabilizer RMI2. The newly identified RMIF-2 protein contains a characteristic OB-fold domain and is also found in RMH-1-containing protein complexes. Our analysis in meiosis prophase I showed that the recombination foci that mark both CO and NCO recombination sites during pachynema contain HIM-6, TOP-3, RMH-1, and RMIF-2. In late pachynema, these foci decorate the obligate COs on the six C. elegans bivalents. The concentration of most BTR complex proteins into these foci depended on the presence of RMIF-2. The exception was the HIM-6 helicase, which was recruited independently, although its enrichment at recombination sites was RMIF-2 dependent. Furthermore, similar to reported observations in mitotic cells [27,28], we showed that RMIF-2-dependent concentration of RMH-1 into recombination foci strongly influenced the RMH-1 protein stability.
The rmh-1 and rmif-2 single mutants shared phenotypical features, indicating a role in both CO and NCO formation and in the suppression of heterologous recombination. Phenotypes of both mutants included embryonic lethality, segregation of males (through non-disjunction of the second X chromosome), and the presence of univalents in diakinesis nuclei. The rate of embryonic death was higher in the rmh-1 mutant, possibly caused by more frequent random segregation of the univalents. In contrast, the reduced number of univalents in rmif-2 mutants might be due to the higher prevalence of (usually rare) double and triple CO events on

PLOS GENETICS
The C. elegans RMI2 functional homolog-2 (RMIF-2) in meiosis chromosomes. On the other hand, the average of 7 DAPI bodies in rmif-2 could be the result of a premature dissociation of a bivalent after the CO designation, a phenotype also observed in the rmh-1 mutant. The extra COs in rmif-2 mutants might counteract the univalent formation, which likely arise through the absence of a pro-CO activity that is shared by RMH-1 and also seems to be lacking in him-6 mutants [19,22,24]. Thus, the extra COs in rmif-2 mutants might be linked to their lower rate of embryonic death compared with rmh-1 mutants. The C. elegans genome encodes for two RMI1 homologs, namely RMH-1 and RMH-2 [19]. In fact, a degree of redundancy between RMH-1 and RMH-2 is indicated by the embryonic lethal phenotype of the double mutant, where no eggs hatch [19]. Nevertheless, RMH-1 seems to have evolved specialized non-shared meiotic activities, since rmh-2 mutants did not display univalents in diakinesis nuclei and had a rate of larval lethality of only 7% [19].
In both rmif-2 and rmh-1 mutants, there was no difference in the total frequency of crossovers in comparison to the WT, however we detected a significant shift of COs away from chromosome arms toward the center.
Our data clearly indicate that the BTR complex must have activities that are not executed in the quantifiable recombination foci, where RMH-1 and RMIF-2 co-localization is mutually dependent. A striking difference between rmh-1 and rmif-2 mutants was their profiles of MSH-5 recombination foci. These foci were completely absent in the rmh-1 mutant, where the MSH-5 signal was only detectable as a nuclear haze. In the rmif-2 mutant, MSH-5 foci were significantly delayed and reduced in number; however, they co-localized with the ZHP-3 protein, which marks CO sites [50], and showed stringent co-localization with the COSA-1 pro-CO marker almost to WT levels. Therefore, a unique function of RMH-1 is to enrich MSH-5 in recombination foci. Moreover, the formation of joint chromosome structures in rmh-1 partially depends on MSH-5 and to some extent on the unscheduled activity of NHEJ, so a future challenge will be to elucidate this RMH-1-specific activity. It seems as if MSH-5 is operating at recombination intermediates in the rmh-1 mutant, but the levels are too low for detection or MSH-5 activity is not always found in foci.

PLOS GENETICS
The C. elegans RMI2 functional homolog-2 (RMIF-2) in meiosis The rmif-2 and rmh-1 mutants also differed in the extent of heterologous recombination in the mIn1 inversion segment: the number of heterologous recombination events was three-fold lower in the rmif-2 mutant compared with both rmh-1 and him-6 (this study and [12]). This strongly suggests the existence of meiotic BTR activity(ies) that do not strictly depend on the RMIF-2 protein. Alternatively, this might indicate that the BTR complex suppresses heterologous recombination via two independent activities, with only one of them depending on RMIF-2. These activities might include inhibiting the establishment of D-loops (which might be the source of the joint DNA structures) or decatenation by dissolution. It is not yet possible to determine which of these two possibilities is the more likely. Nevertheless, the strict interdependence of RMH-1 and RMIF-2 for localization into foci suggests that the rmif-2-independent activity of RMH-1 might relate to inhibiting D-loop formation, which might not involve the formation of distinct strong foci. The delayed formation and reduced numbers of MSH-5 and COSA-1 foci that we detected in rmif-2 mutants argue that extra COs arising through the lack of such activities are not marked by MSH-5 and are likely class II COs, which are usually resolved by non-canonical resolvases.
Interestingly, the somatic cells of individuals with mutations in RMI2 display a "weak Bloom-like phenotype" [5]. For instance, elevated sister chromatid exchange is less pronounced in the chromosomes of RMI2 patients. This could be explained by residual dissolution activity mediated by Bloom in the absence of RMI2 and without a strict requirement for RMI2 to stabilize the BTR complex. The rate of heterologous recombination is lower in rmif-2 mutants than in him-6 or rmh-1 mutants. Therefore, in meiosis RMIF-2 might not be involved in all HIM-6-mediated activities. It will be interesting to determine whether individuals carrying rmif-2/RMI2 mutations accumulate fewer genome rearrangements or translocations and whether this is associated with a lower risk of developing cancer compared with Bloom patients.
In summary, we have shown that the C. elegans RMI2 homolog RMIF-2 contributes to successful chromosome segregation in meiosis and shares numerous activities with RMH-1. However, specific RMIF-2-independent BTR activities were also identified in the germline, and the reduced susceptibility of rmif-2 mutants to heterologous recombination might lead to fewer genome rearrangements and translocations than in mutants of the other BTR complex proteins. It will be interesting to investigate whether those observations made in the germline also hold true in the soma.

Biochemical studies
Fractionated protein extracts for western blotting and immunoprecipitation assays were prepared as described previously [29]. A total of 50 μg protein from each fraction was used for western blotting, and at least 1 mg pooled soluble and insoluble nuclear fractions for immunoprecipitation assays. HA-tagged protein was immunoprecipitated using HA magnetic beads (Pierce #88836). For all immunoprecipitation experiments, beads were pre-equilibrated in buffer D (20% glycerol, 0.2 mM EDTA pH 8, 150 mM KCl, 20 mM Hepes-KOH (pH 7.9), and 0.2% Triton X-100, supplemented with protease inhibitor cocktail (Roche)) and incubated with the proteins overnight at 4˚C with mild agitation. Beads were then washed three times in buffer D for 10 minutes, followed by acidic elution. For this, 25 μl glycine (100 mM, pH 2) was added to the beads and rotated for 5 mins. After magnetic separation of the beads, the supernatant (containing the target antigen) was neutralized with 5μl 1M Tris pH 9.2. Eluated proteins were separated on pre-cast 4-20% TGX gels (BioRad) in 1× SDS-Tris-glycine buffer and transferred onto nitrocellulose membrane for 1 hour at 4˚C at 100V in 1× Tris-glycine buffer containing 20% methanol. Membranes were blocked for 1 hour in 1× TBS containing 0.1% Tween (TBS-T) and 5% milk; primary antibodies were added in the same buffer and incubated overnight at 4˚C. Membranes were then washed three times in 1× TBS-T for 10 minutes and incubated with appropriate secondary antibodies in TBS-T containing 5% milk for 2 hours at room temperature. After another three washes, membranes were incubated with ECL (Euro-Clone) and developed using a ChemiDoc system (BioRad).

Mass spectrometry analysis
Following co-immunoprecipitation, beads were washed with 50 mM ammonium bicarbonate and incubated with 200ng Lys-C protease for 4 h at 37˚C. The supernatant was collected, and bead-bound proteins and polypeptides were eluted twice with 100mM glycine pH 2. The pH of eluates was adjusted to alkaline by adding 1M TRIS pH 8 and disulfide bridges were reduced by adding dithiothreitol to a final concentration of 10mM and incubated at 45˚C for 30 min. Free thiols were alkylated with iodoacetamide to a concentration of 20mM and incubated at room temperature for 30min in the dark. Proteins were digested with 200ng trypsin overnight at 37˚C, and then acidified by adding TFA to a final concentration of 1%. Peptides were desalted on StageTips [51].
Peptide samples were separated on an Ultimate 3000 RSLC nano-flow chromatography system (Thermo Scientific, Dionex) using a pre-column for sample loading (Acclaim PepMap C18, 2 cm × 0.1 mm, 5 μm) and a C18 analytical column (Acclaim PepMap C18, 50 cm × 0.75 mm, 2 μm; both Thermo Scientific Dionex) with a linear 2-35% gradient of solvent B (80% acetonitrile, 0.1% formic acid; solvent A 0.1% formic acid) for 2 hours at a flow rate of 230 nl/min. Eluting peptides were analyzed on a Q Exactive HF Orbitrap mass spectrometer (Thermo Scientific). In data-dependent mode, survey scans were acquired in a mass range of 380-1650 m/z with lock mass on at a resolution of 120,000 at 200 m/z. The AGC target value was set to 3E6 with a maximal injection time of 60 ms. The 10 most intense ions were selected with an isolation width of 2 m/z, and fragmented in the HCD cell with a normalized collision energy of 28%. Spectra were recorded at a target value of 1E5 with a maximal injection time of 250 ms and a resolution of 30,000. Peptides with an unassigned charge state or a charge of +1 or greater than +6 were excluded from fragmentation. The peptide match feature was set to preferred and the exclude isotope feature was enabled. Selected precursors were dynamically excluded from repeated sampling for 30 s. Raw data were processed using the MaxQuant software package 1.6.17.0 (http://www.maxquant.org/) [52] and searched against a C. elegans reference database (Wormbase, version WS269) and a custom database of common contaminants. The search was performed with full tryptic specificity and a maximum of two missed cleavages. Carbamidomethylation of cysteine residues was set to fixed and oxidation of methionine, and N-terminal protein acetylation as variable modifications-all other parameters were set to default. The "match between runs" feature was enabled, label-free quantification (LFQ) activated but without normalization. Results were filtered at protein and peptide level for a false discovery rate of 1%. The protein-group table was further processed in R as follows (R Core Team (2020). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. https://www.R-project.org/.: data were filtered for reverse hits, contaminants and proteins "only identified by site". Raw protein group intensity values were median normalized per sample. After filtering for at least two valid quantification values in at least one group missing values were replaced by drawing random values from a normal distribution at the detection limit. The LIMMA package was used for statistical comparison, using a paired design to account for batch effects between the three independent replicates [53]. Peptide spectrum matches (PSM) in Table 1 correspond to "MS/MS counts" as defined in the MaxQuant output, representing PSM that were considered by the MaxLFQ algorithm. For the complete list of peptides see, S1 File.

Bioinformatic analysis of Y104H12D.4
The RMI2 protein family is highly conserved in the plant and animal kingdoms. In some nematodes, such as Trichinella pseudospiralis, RMI2 family members can be identified as significant hits using hmmsearch with the hidden Markov model (HMM) of the Pfam RMI2 domain (UniProt|A0A0V0YD80, E-value 6.2e-05), [54,55]. However, no Pfam RMI2 hit was identified in the C. elegans proteome. Likewise, one significant hit to the RMI2 domain was found in the termite Zootermopsis nevadensis (UniProt|A0A067R4M9, E-value 4.2e-21) but none in Drosophila melanogaster. The Pfam RMI2 domain belongs to the OB-fold clan, which comprises 107 domains with a wide range of molecular functions, including oligonucleotide or oligosaccharide binding and protein-protein interactions. The common structural feature of the OBfold is a five-stranded beta-sheet forming a closed beta-barrel [56].
As none of the RMH-1 co-purifying proteins had a significant hit to OB-fold superfamily domains in a Pfam hmmsearch, we applied the HMM-HMM comparison with HHPRED to search for remote homologs [57]. RMIF-2 (Y104H12D.4) was the only candidate with a predicted OB-fold, where the best hits in the PDB structural database were to the yeast RFA2 (Replication factor A protein 2; 6I52_B, probability 74.5%) and human RMI2 (3MXN_B, probability 72.5%). In the Pfam database, the best hit was to the RMI2 domain (probability 66.24%). The hits covered nearly the complete length of RMIF-2 (residues 5-101), including the five-stranded beta-sheet. RMIF-2 orthologs can be identified in other nematodes using NCBI blastp searches of the NCBI non-redundant protein database [58], but not in nematodes with RMI2 orthologs. Since nematode RMI2 orthologs are highly related to RMI2 protein family members, it is uncertain whether RMIF-2 is an RMI2 ortholog. However, the mutually exclusive distribution of RMI2 and RMIF-2 orthologs in nematode taxa and the structural similarity to the OB-fold superfamily, in addition to biochemical and genetic data, are strong indications that RMIF-2 is indeed a functional RMI2 homolog.
In the case of RMIF-2 (UniProt accession Q8MXU4), the conservation histogram and the consensus sequence are based on an alignment of nematode orthologs, and for RMI2 (Q96E14) we used a wide selection of eukaryotic orthologs, including animal and plant sequences. Secondary structure elements were predicted by Jpred, where the helices are marked as red tubes, and sheets as green arrows, [59]. The visualization was performed as, [60].
For the ribbon diagram of the human RMI core complex and a model of the putative C.elegans RMIF-2 OB-fold the 3D coordinates of the crystal structure of RMI1 and RMI2 were retrieved from the RCSB PDB protein database, [61,62]. The model of the RMIF-2 OB-fold was created with MODELLER, [63] based on an alignment performed by the remote homology detection and 3D structure prediction server HHpred, [64]. In this search, a multiple alignment of RMIF-2 orthologs (including region 18-89 of C.elegans RMIF-2) was compared with profile hidden Markov models (HMMs) of sequences from the PDB structure database and the best, but not significant, hit was to RMI2 (region 58-138). The model was aligned to the RMI2 coordinates and visualized with pymol (https://pymol.org/2/).

Worm strains
All worms strains were grown at 20˚C using standard techniques [65] on Nematode Growth Medium seeded with Escherichia coli OP50. The N2 Bristol strain was used as the WT reference. Unless otherwise stated, prepicked L4 hermaphrodite worms grown at 20˚C for 16-24 hours were used for all experiments.
The following mutant alleles and tagged lines were created for this study: rmif-2(jf113)/ tmC25 [unc-5 (tmIs1241) CRISPR-Cas9. All strains generated by CRISPR-Cas9 were confirmed by sequencing and backcrossed to WT worms twice prior to use. Strains were generated using a published protocol [67]. Tagged lines had WT levels of viability, hatch rates and lack of males in their progeny ( Table 2).
Generation of rmif-2::ha. We tagged the endogenous rmif-2 locus at the protein C-terminus with an HA-tag and a 5x-Gly linker sequence. The repair template (from Integrated DNA Technologies, 4 nmole Ultramer DNA Oligo) was composed of 35-bp homology to the rmif-2 sequence, into which the HA sequence was inserted.
The Generation of rmif-2::3xflag. We tagged the endogenous rmif-2 locus at the protein Cterminus with a 3×FLAG-tag and a 5x-Gly linker sequence. The 3×FLAG sequence was inserted into a repair template (from Integrated DNA Technologies, 4 nmole Ultramer DNA Oligo) with 35-bp homology to the rmif-2 sequence. The guide RNA was: crRNA (from Dharmacon, Edit-R CRISPR-Cas9 Synthetic crRNA 20 nmol), 5 0 AGAGATGATCAGTTGGCTGT 3 0 . The sequence of the repair template was:

Viability analysis
Single L4-stage worms were transferred to plates. The worms were picked individually and moved onto new plates every 12 hours for 4 days. Dead eggs and viable larvae were scored 24 hours after the mother was removed, and male progeny were counted 3 days later. The viability of embryos was calculated as the number of hatched eggs divided by the total number of eggs laid, and percentage of males was calculated as the total number of male progeny divided by the number of hatched eggs.

Immunofluorescence analysis
Immunocytological analysis was performed as previously described [68]. L4 hermaphrodite worms were incubated at 20˚C for 20-24 hours. Their gonads were then dissected in 1× PBS on Superfrost slides, fixed in final 1% paraformaldehyde for 5 min at room temperature, and frozen in liquid nitrogen. After freeze-cracking and fixation in ice-cold methanol at −20˚C for 10 minutes, the slides were washed three times in PBS-T (1× PBS, 0.1% Tween) at room temperature for 10 minutes. Non-specific binding sites were blocked by incubation in PBS-T containing 1% BSA for 0.5-1 hour. Primary antibody diluted in PBS-T was applied to slides and incubated overnight at 4˚C in a dark, humid chamber. Slides were then washed three times in PBS-T at room temperature for 10 minutes and incubated with secondary antibody diluted in PBS-T for 2 hours at room temperature in a dark, humid chamber. Slides were again washed three times for 10 minutes in PBS-T. Slides were then incubated with DAPI (60μl of a 2 μg/ml stock solution diluted 1:1000 in water) for 1 minute at room temperature, washed with PBS-T for 30 minutes at room temperature, and mounted with Vectashield Mounting Medium (Vector Labs #H-1000).
For detection of GFP::MSH-5, gonads were dissected and fixed in 1× EGG buffer containing 0.1% Tween instead of PBS-T.
The Quantification of nuclear foci. For quantification of COSA-1, MSH-5, and RAD-51 foci, the hermaphrodite gonad was divided into seven equal zones (in Adobe Photoshop) from the mitotic tip to late pachynema. The number of foci per nucleus was counted in each zone, in at least three gonads per genotype. RAD-51 quantification graphs show the percentage of nuclei corresponding to each of the following categories: 0 foci, 1 focus, 2-3 foci, 4-6 foci, 7-12 foci, and >12 foci per nucleus. For the complete statistics of RAD-51 foci quantification, see S2 File.
For COSA-1/ZHP-3 co-localization in foci, a similar system was used but OLLAS::COSA-1 foci were scored in each nucleus in zone 7 only (late pachynema) for co-localization with ZHP-3 signals.
For quantification of TOP-3 foci, the hermaphrodite gonad was divided into four equal zones from the onset of meiosis in the transition zone to late pachynema (Adobe Photoshop) and the number of foci per nucleus was counted in each zone, in three gonads per genotype.

Length of the mitotic zone
The length of the mitotic zone in WT and mutant strains was measured by counting the number of cell rows in the hermaphrodite gonad from the mitotic tip to the transition zone, where chromatin adopts a half-moon shape. Cell rows were counted in seven WT worms and ten rmif-2 worms.

X chromosome pairing
To assess X chromosome pairing, the number of nuclei containing one (i.e. paired) or two (i.e. unpaired) HIM-8 foci was recorded in each of seven zones (starting from the mitotic tip to late pachynema) of the hermaphrodite gonad. Graphs show the percentage of paired HIM-8 nuclear signals in each zone.

Quantification of Synaptonemal complex assembly
To assess synaptonemal complex assembly, the number of nuclei with full co-localization between HTP-3 (chromosome axis marker) and SYP-1 (marker of the central element of the synaptonemal complex) were quantified. The hermaphrodite gonad was divided in seven equal zones from the mitotic tip to late pachynema. Graphs show the percentage of nuclei with complete synapsis in each zone.

Recombination assays
Recombination frequencies and CO localization were assessed in the WT, rmif-2(jf113), and rmh-1(jf54) strains by determining the differences in unique SNPs for chromosomes IV and V between the N2 Bristol and Hawaiian (Hw) strains. For this, mutants were crossed to the Hw strain to generate a mutant with introgressed Hw chromosomes IV and V. Subsequently, rmif-2 mutant males in Hw and rmif-2 mutant hermaphrodites in WT were crossed to generate F1 rmif-2 mutants heterozygous for Hw on chromosomes IV and V and recombination events took place in F1 worms. F1 worms were mated with WT worms expressing a tomato transgene to introduce a WT paternal chromosome and enable recombination events in oogenesis to be monitored. After laying eggs, F1 parent worms were lysed and genotyped to ensure that the first mating worked, using the presence of Hawaiian SNPs as a read-out for chromosomes IV and V. Single F2 hermaphrodites were lysed and analyzed by PCR (restriction by DraI). SNP positions and primers are shown in Table 3.

Heterologous recombination assay
To examine illegitimate recombination events between heterologous sequences, we used the heterologous recombination assay, as previously described, with the dpy-25/mIn1 [rol-1 GFP]

Image acquisition
All microscopy experiments were done using a DeltaVision Epifluorescence Microscope system with 1.3 Megapixel CCD camera, 7-color LED for fluorescence, white LED for transmitted light (UPlanSApo 100×/1.40 oil immersion objective lens) with softWoRx suite R6.1.1 image analysis deconvolution software (Applied Precision); ImageJ (National Institutes of Health), and Adobe Photoshop software. Unless otherwise stated, images are maximum projections of entire nuclei. Images acquired with the DeltaVision were deconvolved using the softWoRx deconvolution algorithm. Maximum intensity projections of deconvolved images were generated using Fiji/ImageJ after background subtraction using a rolling ball radius of 50 pixels. Images of gonads consist of multiple stitched pictures that were processed in the same manner. This is necessary, due to the size limitation of the field of view at high magnification. Stitching of images to build up entire gonad was performed manually in Adobe Photoshop.

Statistical analysis
Statistical analyses were performed using GraphPad Prism 6 (GraphPad) software and Microsoft Excel. Fisher's exact tests, Student T-tests, χ 2 tests, Mann-Whitney tests mean and standard deviation, and statistically significant differences are shown in figures and reported in figure legends. p values of below 0.05 were considered statistically significant: � p < 0.05, �� p < 0.01, ��� p < 0.005, and ���� p < 0.0001.
Supporting information S1 Fig. A Ribbon diagram of the human RMI core complex and a model of the putative C. elegans RMIF-2 OB-fold. (A) 3D coordinates of the crystal structure of RMI1 (in green) and RMI2 (in grey) were retrieved from the RCSB PDB protein database [61,62]. A model of the RMIF-2 OB-fold is shown in cyan and was created with MODELLER, [63], based on an alignment performed by the remote homology detection and 3D structure prediction server HHpred, [64]. In this search, a multiple alignment of RMIF-2 orthologs (including region 18-89 of C.elegans RMIF-2) was compared with profile hidden Markov models (HMMs) of sequences from the PDB structure database and the best, but not significant, hit was to RMI2 (region 58-138). The model was aligned to the RMI2 coordinates and visualized with pymol (http://www.pymol.org/). Two views on the ensemble are given, rotated by 180 degrees.