RAD51C deficiency in mice results in early prophase I arrest in males and sister chromatid separation at metaphase II in females

Bacterial RecA and its yeast orthologue RAD51 are the founding members of the RecA/RAD51 protein family, which plays a crucial role in DNA repair by homologous recombination (for review see Kawabata et al., 2005). After DNA ends are resected at the site of a DNA break, these proteins form a nucleoprotein filament on the single-stranded DNA and catalyze a strand invasion and strand exchange reaction with a homologous region on another DNA molecule to ensure faithful DNA repair. In addition to RAD51, a few other RAD51-like proteins were found in eukaryotes, with their number increasing from three in budding yeast (Rad55, Rad57, and Dmc1) to six in most of the higher eukaryotes (RAD51B, RAD51C, RAD51D, XRCC2, XRCC3, and DMC1; Dosanjh et al., 1998). These proteins have 20–30% amino acid sequence similarity and share common functional domains (Miller et al., 2004). The core C-terminal domain contains two functionally important ATP-binding Walker A and B motifs and is linked to a small globular N-terminal domain via a linker region that is important for protein–protein interactions.

Except for DMC1, a meiosis-specific protein with functions overlapping those of RAD51 (Bishop et al., 1992), the other five paralogues are auxiliary to RAD51. In yeast, the purified Rad55/Rad57 heterodimer functions as a mediator of Rad51, enabling it to nucleate on single-stranded DNA in the presence of replication protein A (RPA; Sung, 1997). Chicken DT40 cells and hamster cell lines deficient in any one of the five RAD51-like proteins are extremely sensitive to DNA damage caused by DNA cross-linking agents and the topoisomerase-I inhibitor camptothecin (Yonetani et al., 2005). In such cells, RAD51 foci formation after DNA damage is attenuated, and the frequency of homologous recombination is reduced (Lio et al., 2004). RAD51C and XRCC3 directly interact with RAD51, and RAD51C has been shown to stabilize RAD51 after DNA damage by protecting it from ubiquitin-mediated degradation (Bennett and Knight, 2005).

These proteins interact with one another and form multiple protein complexes (Masson et al., 2001; Sigurdsson et al., 2001). Two stable complexes have been biochemically purified. One includes RAD51B, RAD51C, RAD51D, and XRCC2 (BCDX2 complex), and the other contains RAD51C and XRCC3 (CX3 complex). The functional relevance of these complexes is supported by studies of double mutants in DT40 cells, in which two mutations in different complexes as opposed to mutations in the same complex had an additive effect on sensitivity to camptothecin (Yonetani et al., 2005).

Despite all of the aforementioned similarities and a general understanding that all RAD51 paralogues are involved in the homologous recombination process, little is known about the specific roles played by any of them. However, each of these proteins is believed to have a nonredundant function, as the loss of any of these genes cannot be compensated for by the overexpression of other family members (Takata et al., 2001). Mutations in these genes have an unequal impact on drug sensitivity in DT40 cells (Yonetani et al., 2005). Although mice that are null for any of the paralogues are embryonic lethal, they vary in the severity of the phenotype (Shu et al., 1999; Deans et al., 2000; Pittman and Schimenti, 2000). In Arabidopsis, Rad51c and Xrcc3 mutants are sterile, whereas other mutants are fertile (Bleuyard et al., 2005). One family member, RAD51D, has been specifically linked to telomere maintenance (Tarsounas et al., 2004).

Although all RAD51 paralogues are implicated in RAD51 foci formation, it is unclear how this function is shared between different paralogues and their complexes (Yonetani et al., 2005). RAD51C is part of both BCDX2 and CX3 complexes and, therefore, is believed to play a central role. Being part of two different protein complexes, RAD51C may also have several distinct functions. On one hand, it is involved in the early steps of recombination associated with RAD51. On the other hand, it may also be involved in late steps of homologous recombination, as RAD51C, together with XRCC3, was purified from HeLa cells as part of a small protein complex possessing a Holliday junction (HJ) resolvase activity (Liu et al., 2004). RAD51C is an essential part of this complex because the resolvase activity was lost upon depletion of RAD51C from the fraction or when protein extracts from hamster cells lacking functional Rad51c were tested. However, the role of RAD51C as a resolvase remains controversial. RAD51C lacks an endonuclease domain, and efforts to demonstrate resolvase activity for the recombinant RAD51C protein have not been successful so far. Also, there is no in vivo evidence supporting this late role of RAD51C in recombination.

In this study, we report the generation of a viable mouse model carrying a hypomorphic allele of Rad51c, which enabled us to overcome the early embryonic lethality of a null allele and demonstrate the role of RAD51C in meiotic recombination. We demonstrate that RAD51C is involved in the recruitment of RAD51 at early stages of homologous recombination in spermatocytes. In oocytes, we describe a defect in sister chromatid cohesion at metaphase II. We speculate that RAD51C may also be required at late steps of homologous recombination.

The mouse Rad51c gene consists of eight exons and encodes a 366–amino acid protein (Leasure et al., 2001). Exons 2 and 3 code for a 142–amino acid region, including a linker and the Walker A ATPase domain. The deletion of these exons resulted in a functionally null allele, Rad51c^ko, which caused an early postimplantation lethality of Rad51c^ko/ko mouse embryos (unpublished data). In addition to a null allele, we also generated a hypomorphic allele, Rad51c^neo, by inserting a neomycin (neo) resistance gene under the control of a phosphoglycerate kinase (PGK) promoter into intron 1 (Fig. 1 A). Disruption of normal splicing by the presence of the PGK-neo cassette has been reported previously for other genes (Meyers et al., 1998). Aberrant splicing was demonstrated for the Rad51c^neo allele by RT-PCR analysis using primers to amplify a region between the first and fourth exons (Fig. S1). The sequence analysis predicted that the misspliced transcript was likely to encode for a 76–amino acid polypeptide that included 39 amino acids from the first exon of Rad51c and an additional 37 amino acids from the 3′ region of the PGK-neo cassette. Aberrant splicing resulted in an ∼60% reduction in RAD51C protein levels (Fig. 1, C and D). The combination of a hypomorphic and a null allele (Rad51c^ko/neo) produced mice expressing only 5–30% of the normal level of RAD51C protein (Fig. 1 D).

The residual amount of RAD51C protein in Rad51c^ko/neo mice is sufficient to allow normal growth and development. However, 36.6% (n = 82) of such males and 11.6% (n= 173) of the females were infertile. In the control group, which included wild type, Rad51c^ko/+, Rad51c^neo/+, and Rad51c^neo/neo (from here on, these genotypes are referred to as controls for simplicity), only one male and one female out of 105 mating pairs were infertile. To determine the fertility status, 6–8-wk-old Rad51c^ko/neo mice were mated with wild-type mice for up to 6 wk. Infertile Rad51c^ko/neo animals were sexually active as indicated by the repeated detection of vaginal plugs, which did not result in pregnancy. In contrast, fertile Rad51c^ko/neo males and females produced normal sized litters compared with the control group (7.1 ± 2.7 pups [n = 68] and 7.3 ± 2.6 pups [n = 21], respectively).

Testes of infertile Rad51c^ko/neo males were significantly reduced in size (P = 0.042) and weighed 18–48 mg at 8–10 wk of age in contrast to 43–90 mg for fertile Rad51c^ko/neo males. Furthermore, testes of infertile males also had reduced levels of RAD51C protein compared with that of fertile males (Fig. 1, B and C). Unlike testes of the control males, histological examination of testes from 4-wk-old Rad51c^ko/neo males revealed seminiferous tubules that were deformed and were often almost devoid of germ cells (Fig. 2, A and B). There was a marked increase in the number of apoptotic spermatocytes in the testes of infertile mice as determined by TUNEL assay (Fig. 2, C and D). Testis histology of the 12-wk-old infertile Rad51c^ko/neo males revealed a few tubules containing some mature spermatozoa, but the majority still showed abnormal structures compared with control testes (Fig. 2, E–H). Seminiferous tubules from control animals were comprised mostly of multiple layers of round spermatids and usually a single layer of spermatocytes undergoing meiotic divisions (Fig. 2 G). In contrast, most tubules from mutant males contained no mature sperm, few, if any, spermatids, and multiple layers of spermatocytes in the zygotene and pachytene stages of meiotic prophase (Fig. 2 H). Although testes morphology and cell composition improved with age, adult males remained functionally infertile for up to 8 mo of age.

To determine the precise nature of spermatogenesis failure in Rad51c^ko/neo mice, we analyzed surface spreads of spermatocytes using molecular markers specific for different stages of meiosis. SCP3 and SCP1, which are components of the lateral and central elements of the synaptonemal complex, respectively, were used to identify cells at different stages of meiotic prophase based on the degree of chromosome condensation and synaptonemal complex formation (Dobson et al., 1994). In control samples, >60% (n = 184) of the spermatocyte population is comprised of cells at late zygotene, when chromosomes are not fully synapsed at their ends (Fig. 3 A), and at pachytene when synapsis is complete (Fig. 3 B). In infertile Rad51c^ko/neo mutants, we found that some spermatocytes progressed to the pachytene stage (Fig. 3 D), but 63% displayed reduced numbers of MLH1 foci (Fig. 3 G and Fig. S2, B–E), which are markers of the crossover sites (Baker et al., 1996). In contrast, only 11% of the control spermatocytes lacked MLH1 foci on one or more chromosomes, whereas most spermatocytes displayed one or two foci on each chromosome (Fig. 3 E and Fig. S2, A and E). Mutant spermatocytes lacking MLH1 are likely to be at the pre-MLH1 early pachytene stage (Fig. S2 D). Overall in mutant Rad51c^ko/neo males, a shift in spermatocyte distribution toward early stages of prophase I was observed (Table I). The number of late zygotene and pachytene spermatocytes was reduced to 40% (n = 208), whereas cells at leptotene (22% in the mutant vs. 7.6% in the control) and at early to midzygotene (30% in the mutant vs. 21% in the control) became the major fractions (Fig. 3 C). The fact that cells at all stages (leptotene-pachytene) of meiosis could be found in these preparations suggests that the phenotype observed in infertile Rad51c^ko/neo males reflects an impairment of spermatogenesis during meiosis I rather than an absolute developmental arrest.

Formation and repair of DNA double strand breaks (DSBs) by homologous recombination ensures a correct pairing and subsequent segregation of homologous chromosomes during the first meiotic division. A successful generation of DSBs and initiation of their repair is indicated by the presence of phosphorylated histone H2AX (γ-H2AX; Fig. 3, I and K) and RPA (Fig. 3, F and H) protein staining in mutant and control spermatocytes at leptotene and zygotene (Plug et al., 1997; Mahadevaiah et al., 2001). At pachytene, normally only the sex chromosomes stain positively for γ-H2AX (Fig. 3 J; Mahadevaiah et al., 2001). However, in mutants, more than half of all spermatocytes displayed multiple γ-H2AX foci at pachytene, indicating the persistence of DSBs at this stage (Fig. 3 L). We found no marked difference in the number of RPA foci in mutant and control spermatocytes at various stages of prophase I (Fig. 3, F and H; and Fig. S3). As RAD51C has been implicated in RAD51 foci formation in mitotic cells, we tested whether RAD51 foci formation was defective in spermatocytes of infertile males. RAD51 plays an important role in the initial steps of homologous recombination by mediating homologous pairing and strand exchange and is normally observed in multiple foci first appearing at leptotene and sharply decreasing at pachytene (Fig. S4, A–C; Ashley et al., 1995). We found that the number of RAD51 foci was reduced about threefold at leptotene and zygotene stages in spermatocytes of Rad51c^ko/neo 3-mo-old mice (46–57 foci per cell in the mutant compared with 140–169 foci in the Rad51c^neo/+ littermate control; Fig. 3, M and O; and Fig. S4, D–G), indicating an early defect in the homologous recombination process.

γ-H2AX staining marks not only unrepaired DNA but also unsynapsed regions along homologous chromosomes. Although some mutant spermatocytes showed only a few γ-H2AX foci on autosomal chromosomes, we found others that had entire chromosomes staining positively for γ-H2AX at pachytene (Fig. 4 A, arrowheads). Such chromosomes also appeared thinner than others when stained for SCP3 (Fig. 4 B, arrowheads) and lacked SCP1 staining (Fig. 3 D, arrow), suggesting that these were completely unpaired autosomal univalents. Abnormal synapsis between homologous chromosomes was further confirmed by karyotyping spermatocytes at metaphase I. We examined metaphase I spermatocytes from 3-wk-old Rad51c^ko/neo males, which were potentially infertile as identified by testes weight, and a Rad51c^neo/+ littermate control. We found that although a few metaphase I spermatocytes (16%; n = 114) from mutant males appeared normal, similar to the control spermatocytes (Fig. 4, C and D), the majority (84%) were morphologically abnormal, possibly undergoing apoptotic fragmentation of the chromatin (Fig. 4, E and F). Unpaired autosomal univalents were visible in each of these spreads (Fig. 4, E and F; arrowheads). Such univalents are known to induce apoptosis at metaphase I (Eaker et al., 2001). Although metaphase I and II spermatocytes were almost equally represented in control animals (57% and 43%, respectively; n = 90), metaphase II spermatocytes were rarely found in mutant males (9%; n = 125). This suggests that mutant spermatocytes rarely progress beyond metaphase I, as they undergo apoptosis at this stage.

Ovaries from infertile females were similar in size compared with fertile littermates, and histological examination showed the presence of morphologically normal follicles at all stages of development (Fig. 5, A–F). However, ovaries from 6- and 12-wk-old animals revealed the absence of corpora lutea in infertile Rac51c^ko/neo females, suggesting that failure to ovulate may be the cause of infertility (Fig. 5, D and E). This ovulation block could be overcome by hormonal (intraperitoneal injection of 5 IU [0.1 cc] of pregnant mare serum [PMS] followed by a second injection of 5 IU [0.1 cc] of human chorionic gonadotropin [hCG] 48 h later) treatment (see ovary histology after superovulation; Fig. 5 F), after which infertile Rad51c^ko/neo females could become pregnant when mated with wild-type males. When embryos from the uterine horns of these mice were dissected at embryonic day 8.5, we found that the number of embryos obtained from infertile Rad51c^neo/ko females was greatly reduced (5 ± 0; n = 2) compared with the number of embryos from heterozygous Rad51c^neo/+ superovulated littermates (15.5 ± 4.94; n = 2; P = 0.05). In addition, 7/10 embryos (70%) from the mutant females displayed a wide range of developmental abnormalities compared with only two abnormal embryos out of 31 (6.5%) from superovulated control mice (Fig. 5 H). All of the embryos from the control females appeared to develop normally (Fig. 5 G). We hypothesized that the developmental defects observed in these embryos resulted from gross chromosomal defects in the oocytes caused by abnormal meiosis (Yuan et al., 2002).

To determine the cause of such abnormalities in oocytes, we examined the meiotic progression of germinal vesicle (GV)–intact oocytes by in vitro maturation. Chromosomes of the control Rad51c^neo/+ oocytes were aligned at the metaphase plate after 8 h of maturation (Fig. 5 I). Mutant oocytes progressed to metaphase I, and some aligned normally along the metaphase plate (Fig. 5 K). However, 60–75% (n = 50) of them displayed poorly structured metaphase plates (Fig. 5 J) in contrast to only 30% (n = 33) of abnormal oocytes from control Rad51c^neo/+ littermates. The reason for the relatively high number of abnormal oocytes observed in control samples is unclear. One possibility is that this might be an effect of haploinsufficiency, as these animals were heterozygous for the hypomorphic allele.

We examined the karyotypes of metaphase I oocytes, which revealed no apparent defects in pairing and cohesion in infertile Rad51c^ko/neo females (n = 8) compared with the oocytes from Rad51c^neo/+ females (n = 6; Fig. 6, A and B). Evaluation of the centromeric cohesion between sister chromatids was facilitated by using a centromere-specific probe ( Fig. S5, A and B. We then karyotyped the oocytes that were allowed to progress to metaphase II in vivo after hormonal treatment. At metaphase II, oocytes from fertile females show the presence of 20 pairs of chromatids, each consisting of two sister chromatids that are attached together at their centromeres (Fig. 6 C). In oocytes from infertile Rad51c^ko/neo mice, we found a variety of chromosomal abnormalities. A majority of the mutant oocytes (85%; n = 34) showed precocious separation of sister chromatids (PSSC), indicating a problem with chromatid cohesion (Fig. 6 D). The degree of the cohesion defect varied from cases in which all chromosomes were affected (approximately half of all cases) to those in which only a few chromosomes were affected (for more examples, see Fig. S5, D–G). Only one oocyte from the control littermate was found to have a PSSC phenotype (5%; n = 20). Mutant oocytes with a PSSC phenotype often had a few acentric chromatids (40%; n = 10; Fig. 6 D, open arrowhead; and Fig. S5 F, arrowheads). On the other hand, a few other chromatids in the same spreads had two centromeres (Fig. 6 D, double arrows; and Fig. S5, E–G; double arrows). No such phenotype was observed in control oocytes. In addition, 20% of mutant oocytes (n = 10) in which chromosomes could be counted revealed >20 chromosomes or chromatid pairs, whereas others had 20 or fewer pairs of centromeres, indicating the abnormal segregation of homologous chromosomes during anaphase I of meiosis (Fig. S5 G).

The chromosomal defects observed in oocytes after metaphase I suggest a late role for RAD51C in recombinational repair. We did not find any evidence to support a direct role of RAD51C in sister chromatid cohesion (see Discussion; unpublished data). However, based on a biochemical study (Liu et al., 2004), the late function of RAD51C is likely to be associated with the resolution of HJs. To test whether HJ resolvase activity is indeed affected in Rad51c-deficient mouse cells, we examined protein extracts from mouse embryonic fibroblasts (MEFs) in an in vitro HJ resolution assay. Two Rad51c^ko/ko MEF lines were established on a p53-null background, which partially rescued the early embryonic lethality of Rad51c-null embryos (unpublished data). We used MEFs derived from p53-null and wild-type embryos as controls. HJ resolution activity of protein extracts from the two Rad51c mutant MEFs was at the background levels between 1.9 and 3.8% and did not correlate with the increasing amounts of protein extract (Fig. 6, E and F). The HJ resolution activity of the p53-null control cells increased from 5.4 to 11.3% as the amount of protein extract increased from 0.9 to 3.5 μg. This was slightly higher than the activity of the wild-type primary MEFs (between 1.0 and 7.5%; Fig. 6, E and F). These activity differences correlated with the amount of RAD51C protein detected in these extracts by Western blotting (Fig. 6 G). Thus, these data further support a RAD51C role in the resolution of HJs.

To test whether the premature separation of chromatids occurred even in males, we examined the spermatocytes that progressed to metaphase II (n = 39 in control and n = 11 in mutant samples). In control spermatocyte samples, we found 20 pairs of chromatids attached at the centromeres as observed in oocytes (Fig. 7 A and Fig. 6 C, respectively). Although the chromosomes in some mutant spermatocytes appeared to be normal (Fig. 7 B), others revealed fragmented chromosomes with broken centromeres or had aberrant chromosome numbers (Fig. 7, C and D). However, none of the mutant spermatocytes displayed the sister chromatid cohesion defect. Thus, the PSSC defect is a sexually dimorphic feature restricted only to females. It is unclear whether this phenotype in males is caused by a defect in DSB repair or is caused by the defect in HJ resolution. We present a model to explain how a defect in HJ resolution may result in PSSC in oocytes (see Discussion and Fig. 8).

To date, the only well established function of RAD51C is its role in the process of homologous recombination by facilitating RAD51 foci formation after DNA damage. The precise role of RAD51C in the recruitment of RAD51 and how it differs from other RAD51 paralogues is still unclear. More importantly, recent studies of AtRad51c- and AtXrcc3-deficient Arabidopsis plants and fruit flies deficient in a RAD51C-like gene, spn-D, point to their unique requirement in meiosis unlike other RAD51 paralogues (Abdu et al., 2003; Bleuyard et al., 2005). Characterization of the AtRad51c mutant plants showed that this defect is associated with the repair of Spo11-induced DSBs, leading to abnormal synapsis and severe chromosomal fragmentation at the pachytene stage of prophase I (Abdu et al., 2003; Li et al., 2005).

In this study, we describe a meiotic defect in Rad51c mutant male mice that is associated with abnormal synapsis between homologous chromosomes at pachytene. At this stage, abnormal synapsis is indicated by the presence of γ-H2AX foci on some of the autosomal chromosomes. Occasionally, entire chromosomes appeared positive for γ-H2AX, indicating completely unsynapsed chromosomes (Fig. 4, A and B). Unlike Arabidopsis, mouse spermatocytes with unsynapsed chromosomes undergo apoptosis either at pachytene or metaphase I (Mahadevaiah et al., 2000; Eaker et al., 2001). Consistent with this, we observed a massive increase in abnormal metaphase I spermatocyte spreads.

Chromosome synapsis is dependent on RAD51-mediated recombination machinery, which helps bring homologous chromosomes together to repair DNA breaks introduced at leptotene using homologous regions as a template. We found that in Rad51c^ko/neo spermatocytes, the number of RAD51 foci was reduced about threefold as early as at leptotene. This is consistent with observations of attenuated RAD51 foci formation in RAD51C-deficient cell lines and with a recent finding that RAD51C may interact with RAD51 directly or as a complex with XRCC3 (Eaker et al., 2001; Rodrigue et al., 2006). A similar role for Rad55/Rad57, the budding yeast paralogues of RAD51, has been described previously (Gasior et al., 2001). The lack of RAD51 foci revealed in spermatocytes of Brca2-deficient mice results in developmental arrest at late zygotene (Sharan et al., 2004). However, Rad51c^ko/neo spermatocytes do not completely arrest at this stage but clearly accumulate at leptotene and at early to midzygotene. This suggests that the RAD51C defect in these mice causes impairment of the RAD51 function; however, because of the hypomorphic nature of the mutation, some spermatocytes escape an early arrest at zygotene, progress further with unrepaired DSBs, and exhibit partially or completely unsynapsed chromosomes so that most of them are eventually blocked at metaphase I.

Like the males, a subset of the Rad51c^ko/neo females failed to produce any litters. Female infertility was associated with an ovulation failure, as no corpora lutea were found in ovaries of such mice. In ovaries of adult mice, oocytes are normally arrested in meiotic prophase I (Borum, 1961). As ovaries of the infertile Rad51c^ko/neo mice were of the same size as the wild-type animals and contained follicles at all stages of maturation, we conclude that Rad51c-deficient oocytes are able to progress to pachytene normally without an early arrest like in males.

However, oocytes from infertile Rad51c^ko/neo mice do suffer a maturation defect preventing ovulation, which could be overcome only by external hormonal stimulation. Although the number of GV-intact oocytes isolated from ovaries of such females was not considerably reduced compared with littermate controls, they often appeared more fragile to handle during the in vitro maturation experiments (unpublished data). Mutant oocytes display other early meiotic defects like an increased incidence of dysregulated chromosome alignment at the metaphase plate during metaphase I. Such defects may result from the impaired DSB repair as observed in mutant spermatocytes, but, unlike spermatocytes, oocytes could progress to metaphase I even with unrepaired DSBs as a result of sexually dimorphic checkpoint mechanisms (Eaker et al., 2001; Hunt and Hassold, 2002). Sexually dimorphic phenotypes in mice have been reported for several other meiotic mutations. Mutations in genes like Scp3, Mei1, and Brca2 result in male meiosis arrested during a zygotene to pachytene transition, whereas mutant oocytes can progress through pachytene all the way to metaphase I (Libby et al., 2002; Yuan et al., 2002; Sharan et al., 2004). Similarly, sexual dimorphism in Rad51c-deficient mice is reflected by the fact that 37% of Rad51c^ko/neo males were infertile, whereas only 12% of females were infertile.

Despite a slightly dysregulated chromosome alignment at the metaphase plate, mutant oocytes were karyotypically normal and did not display unsynapsed chromosomes, which is characteristic of metaphase I oocytes with early meiotic dysfunctions (such as MLH1 and Mei1-deficient mice; Woods et al., 1999; Libby et al., 2002). Thus, abnormal embryos produced by infertile females after superovulation must be caused by defects occurring after the metaphase I stage. Indeed, major chromosomal abnormalities were found in mutant oocytes later at metaphase II such as PSSC, aneuploidy, and broken chromosomes.

PSSC is the most prominent defect at metaphase II, affecting almost all oocytes of the Rad51c mutant infertile females. The PSSC phenotype has been previously demonstrated for several genes regulating meiotic cohesion such as SMC1β in mice and Sgo1, Bub1, and PP2A in yeast (Bernard et al., 2001; Kitajima et al., 2004; Revenkova et al., 2004; Riedel et al., 2006). Sister chromatid cohesion is mediated by the multiprotein complex cohesin, comprising REC8, STAG3, SMC1β, and SMC3 proteins (Prieto et al., 2004). During prophase I of meiosis, cohesin is required for synaptonemal complex formation and recombination to occur between homologous chromosomes rather than between sister chromatids (Schwacha and Kleckner, 1997). Although cohesin is cleaved and released from chromosome arms by separase at metaphase I, Shugoshin protects it at centromeres until anaphase II (Fig. 8; Wang and Dai, 2005). This ensures a correct segregation of homologues at anaphase I and sister chromatids at anaphase II into separate daughter cells. Mice deficient for meiotic cohesins are infertile and display a wide range of defects, from the failure of interhomologue synapsis and recombination for Rec8^−/− mice (Xu et al., 2005) to the occasional breakup of bivalents at metaphase I and premature sister chromatid separation that is detectable at metaphase I and is fully exposed at metaphase II for SMC1β^−/− mice (Revenkova et al., 2004). This sister chromatid cohesion defect at metaphase II seen in infertile Rad51c^ko/neo females is remarkably similar to defects found in SMC1β knockout mice (Revenkova et al., 2004). Interestingly, a hamster cell line lacking functional RAD51C has also been reported to have defects in sister chromatid cohesion (Godthelp et al., 2002). Altogether, these findings suggested a possible role for RAD51C in sister chromatid cohesion.

We tested whether RAD51C protein can directly interact with proteins involved in sister chromatid cohesion like RAD21, REC8, SMC1β, and Shugoshin (Sgo1) by coimmunoprecipitation but found no evidence to support this hypothesis (unpublished data). This does not eliminate the possibility that RAD51C affects cohesins indirectly via other binding partners or signaling pathways. However, other evidence suggests that RAD51C may not be directly involved in sister chromatid cohesion. First, immunostaining of spermatocyte spreads for SMC1β, RAD21, and REC8 did not reveal any obvious abnormalities during prophase I in male meiosis (Fig. 3, N and P; and not depicted). Second, neither PSSC nor univalents were observed in Rad51c mutant oocytes at metaphase I unlike SMC1β^−/− mice (Revenkova et al., 2004). Third, cohesin failure alone does not explain the presence of acentric chromatids and chromatids with two centromeres seen in the Rad51c-deficient oocytes. Fourth, the few mutant spermatocytes that reach metaphase II do not show any defect in sister chromatid cohesion (Fig. 7). Finally, no sister chromatid cohesion defects were found in Rad51c-null MEFs generated from Rad51c^−/− embryos (unpublished data). On the other hand, in addition to biochemical activity, a role for RAD51C in HJ resolution is supported by its colocalization with MLH1 to the site of an obligate crossover at the pseudoautosomal region of the sex chromosomes and by the observation that the number of RAD51C foci is substantially reduced in spermatocyte spreads of Mlh1^−/− mice (Liu et al., 2007). Also, we have shown that protein extracts from Rad51c-null MEFs lack HJ resolution activity, which further supports this finding (Fig. 6, E and F).

If RAD51C indeed plays a role in the resolution of HJ, why do RAD51C mutant oocytes exhibit a sister chromatid cohesion defect? A precise phenotype of an HJ resolvase deficiency in the mouse is difficult to predict because so far no other mouse protein has been implicated in this process. Mus81-Eme1 was identified as an HJ resolvase in fission yeast (Boddy et al., 2001). Yeast cells deficient for any of these genes are sterile as a result of a defect in chromosome segregation. However, in higher eukaryotes, HJ resolution is likely to be performed by other molecules because Mus81-deficient mice are fertile and do not show any meiotic defect (Dendouga et al., 2005). We propose that the PSSC phenotype reflects the response of unresolved chromosomes to the increased tension exerted by the spindle at the kinetochores when the chromosomes are pulled to the opposite poles. This model is supported by the behavior of dicentric chromatids in mouse oocytes, which revealed a range of meiotic defects that were surprisingly similar to those of Rad51c-deficient oocytes, including PSSC as well as broken or acentric chromosomes and aneuploidy at metaphase II (Koehler et al., 2002). Such dicentric chromosomes were generated in strains of mice that were heterozygous for an inversion in a large region of chromosomes X or 19. Meiotic recombination in the inverted region led to the generation of dicentric chromatids. When pulled in opposite directions during anaphase I, such chromatids resulted in PSSC in 90–97% of cases and rarely in broken chromosomes (3–10%). Interestingly, the meiotic behavior of such chromosomes in female mice was different from that in male mice, flies, and maize, which were more prone to breakage or selective loss (Koehler et al., 2002). Consistent with this observation, metaphase II spermatocytes from Rad51c^ko/neo mutant mice do not show any sister chromatid cohesion defect but do show chromosomes with broken centromeres (Fig. 7, C and D).

We speculate that a similar situation would be created if homologous chromosomes failed to dissolve chiasmata. Chiasmata are the cytological manifestation of crossovers, which are believed to arise from double HJs (Szostak et al., 1983). Therefore, we conclude that the late meiotic defects found in oocytes of infertile Rad51c^ko/neo mice can be associated with the HJ resolution failure and propose a model to explain its mechanism (Fig. 8).

In normal meiocytes, HJs are established between homologous chromosomes by the pachytene stage of prophase I using the homologous recombination machinery. At metaphase I, bivalents align in the middle of the spindle so that homologous chromosomes can be pulled in opposite directions by microtubules attached to kinetochores of sister chromatids that are oriented toward the same pole. Correct chromosomal alignment triggers anaphase-promoting complex activation, and cohesion is released along the chromosome arms, whereas Sgo1 protects it at the centromere to ensure that sister chromatids stay together during the reductional division. In fission yeast, a spindle checkpoint regulator, Bub1, is essential for the recruitment of Sgo1 to centromeres and, together with Sgo2, promotes sister kinetochore coorientation during metaphase I (Vaur et al., 2005). After chiasmata are dissolved, homologous chromosomes can segregate in separate cells. When chiasmata are not formed, homologous chromosomes cannot align properly at the metaphase plate. This activates a spindle checkpoint leading to metaphase I arrest (Fig. 8).

We propose that in Rad51c-deficient oocytes, meiosis proceeds normally until anaphase I. However, there is an accumulation of recombination intermediates such as double HJs that hold the homologous chromosomes together even though the chiasmata fail to fully mature. At the onset of anaphase, there is an increase in tension at the centromere as a result of the persistence of unresolved double HJs. The increased tension may disrupt sister chromatid cohesion at the centromere, as has been demonstrated for dicentric chromosomes and the unpaired X chromosome of XO mice (Hodges et al., 2001; Koehler et al., 2002). The exact mechanism of this process is unclear. However, centromeric sister chromatid cohesion is sensitive to chemicals interfering with the metaphase I to anaphase I transition (Yin et al., 1998; Mailhes et al., 1999). In addition, Sgo1 has been shown to be a sensor of kinetochore tension in mitotic cells (Indjeian et al., 2005). Although it remains to be shown in meiosis, it is likely that Shugoshin degradation/disruption caused by increased tension plays a critical role in the PSSC phenotype that is so prominent in Rad51c-deficient oocytes.

In addition to the disruption of centromeric cohesion, chromosome breakage is another consequence of increased tension. Indeed, half of all Rad51c-deficient oocytes displayed broken chromosomes (Fig. 6 and Fig. S5, E–G). We predicted that in such cases, the centromeric cohesion would remain unaffected. As shown in Fig. 6 D (inset), in several oocytes, we did observe single chromatids with an extra centromere that is likely to have originated from a sister chromatid. It is also possible that both homologous chromosomes may segregate into one daughter cell, which would lead to aneuploidy detectable at metaphase II. We did see aneuploidy in 2/10 oocytes in which individual chromosomes could be counted (Fig. S5 G).

Based on the meiotic defects observed in infertile Rad51c^ko/neo females, we speculate that RAD51c functions in late stages of meiotic recombination, possibly participating in the resolution of HJs. We cannot rule out the possibility that RAD51C may play a role in sister chromatid cohesion, which may explain the PSSC defect in oocytes. Similarly, it is possible that the primary defect may be the same (i.e., a defect in RAD51-mediated DSB repair) in males and females, but the phenotypes are different because of sexually dimorphic checkpoints. However, based on the evidence presented here, we have proposed a model connecting the impairment of HJ resolution function in Rad51c^ko/neo infertile females to the observed phenotype in oocytes.

The Rad51c^neo allele was generated in embryonic stem cells in which a neo resistance gene flanked by two loxP sites was inserted into intron 1 and a single loxP was inserted into intron 3. Heterozygous offspring in the C57BL/6J × 129/Sv mixed genetic background were crossed with β-actin–cre transgenic mice (Lewandoski et al., 1997) to obtain the Rad51c^ko allele. Mutant Rad51c^ko/neo mice were generated by crossing Rad51c^ko/+ or fertile Rad51c^ko/neo mice to Rad51c^neo/neo mice.

To determine the fertility status of Rad51c^ko/neo mice, 6-wk-old males and females were mated with fertile animals for 4–10 wk. To monitor the mating behavior, mice were checked for vaginal plugs daily in a course of at least 6 wk. Mice that plugged several times but failed to produce any offspring were identified as infertile. Infertile Rad51c^ko/neo mice are referred to as mutant mice in the text for simplicity.

Protein lysates from control and mutant testes were prepared in cold radioimmunoprecipitation assay buffer. Samples containing 80 μg of protein were separated in NuPAGE 4–12% Bis-Tris polyacrylamide gels (Invitrogen) and transferred onto a nylon membrane. The membrane was probed with mouse or rabbit α-RAD51C antibody (Novus Biologicals or Chemicon International, respectively) at a 1:500 dilution or were probed with α-panactin antibody (NeoMarkers) diluted 1:400 according to standard procedures. Secondary α-mouse IgG-HRP antibody (1:2,000; Santa Cruz Biotechnology, Inc.) and an ECL chemiluminescence system (GE Healthcare) were used for signal visualization.

Testes and ovaries were fixed in Bouin's solution. Samples were dehydrated through an ethanol series, embedded in paraffin, serially sectioned, and stained with hematoxylin and eosin. Slides were examined using brightfield microscopy. For TUNEL staining, testes were fixed in 10% neutral buffered formalin and were stained using the ApopTag kit (Chemicon International) according to the manufacturer's instructions.

Surface spreads of spermatocytes from the testes of mutant and control animals were prepared and stained as described previously (Romanienko and Camerini-Otero, 2000). Another method was used to prepare spermatocytes for the RAD51 and RPA staining and was described previously (Counce and Meyer, 1973). The difference between the two protocols is that no enzymatic treatment is involved in the second method, and cells are separated by pipetting and spread in a hypotonic solution directly on a slide. The following primary antibodies were used for immunofluorescence: rabbit anti–γ-H2AX (1:1,000; obtained from W. Bonner, National Cancer Institute, Bethesda, MD), mouse anti-MLH1 (1:10; BD Biosciences), rabbit polyclonal α-SCP3 (1:500), α-SCP1 (1:1,000), α-RPA (1:100; all provided by P. Moens, York University, Toronto, Canada), mouse monoclonal α-SMC1β (1:10; provided by E. Revenkova, Mount Sinai School of Medicine, New York, NY), and rabbit α-Rad51 (1:500; obtained from S. West, Cancer Research UK, South Mimms, UK). Secondary antibodies used were goat anti–rabbit AlexaFluor488, goat anti–rabbit AlexaFluor568, goat anti–mouse AlexaFluor488, and goat anti–mouse AlexaFluor568 (Invitrogen). Secondary antibodies were used at a 1:250 dilution.

Images were acquired with a microscope (Axioplan 2; Carl Zeiss MicroImaging, Inc.) using an oil plan Neofluar 100× 1.3 NA objective (Carl Zeiss MicroImaging, Inc.). Images were taken with a CCD camera (Quantix; Photometrix) and processed using SmartCapture software (Desksoft). Images were further processed with Photoshop software (Adobe) to adjust for size and contrast.

Females were superovulated, and oocytes and embryos were collected as described previously (Hogan et al., 1994). In brief, 0.1 ml PMS (5 IU; Sigma-Aldrich) was injected intraperitoneally into female mice. GV-intact stage oocytes were collected 44–48 h later in flushing and holding medium with 3 mg/ml BSA (Specialty Media) by puncturing ovaries with a 27-gauge needle. Oocytes were incubated briefly in 0.1 mg/ml hyaluronidase type IV-S (Sigma-Aldrich), and cumulus cells were removed by pipetting before transfer into Chatot, Ziomek, and Bavister media containing l-glutamate and 3 mg/ml BSA (Specialty Media) for 8 h at 37°C and 5% CO₂ to obtain metaphase I stage oocytes for immunocytochemistry and karyotyping. Alternatively, 48 h after PMS treatment, 6-wk-old female mice were given an intraperitoneal injection of 0.1 ml hCG (5 IU; Sigma-Aldrich), and, 14 h later, metaphase II–stage oocytes were collected from ampullae. To obtain embryos, 6-wk-old female mice given both PMS and hCG injections were placed with stud males.

Metaphase I and II stage oocytes were washed in Dulbecco's PBS without CaCl₂ or MgCl₂ (Invitrogen) and were fixed for 10 min in cold methanol. Oocytes were incubated in blocking buffer (4% BSA; Sigma-Aldrich), 10% normal goat serum (Vector Laboratories) overnight at 4°C, monoclonal anti–β-tubulin clone TUB2.1 (Sigma-Aldrich) for 1 h at RT, AlexaFluor488 goat anti–mouse IgG (Invitrogen) for 1 h at RT, and in a 1:5,000 dilution of 1 mg/ml DAPI (Roche) for 20 min at RT before mounting in Vectashield Mounting Medium (Vector Laboratories) on Shandon Multi-Spot microscope slides (Thermo Savant).

Metaphase spreads of spermatocytes were prepared by the Evans method from the testes of 3-wk-old mutant and control males (Evans et al., 1964). Oocytes were karyotyped by an air-drying method as described previously (Tarkowski, 1966). In brief, mutant and control females were injected with PMS followed by hCG as described above (see Superovulation and collection of oocytes and embryos), and, 14 h later, oocytes arrested at metaphase II were collected from the ampullae in Weimouth media containing penicillin-streptomycin, 10% FBS, and 2.5 mg/ml sodium pyruvate. Chromosome spreads were prepared from the oocytes after a 2-min incubation in 0.9% sodium citrate solution.

Protein extraction was performed as follows: exponentially growing MEFs were collected from ∼20 15-cm tissue culture dishes after trypsinization. Approximately 1 g of a cell pellet was resuspended in prechilled lysis buffer (10 mM Tris, pH 8.0, 1 M KCl, 1 mM EDTA, and 1 mM DTT) in the presence of complete EDTA-free protease inhibitor cocktail (Roche). Cells were homogenized by sonication followed by incubation on ice for 1 h. The insoluble pellet was removed by centrifugation at 13,000 rpm for 1 h. Solid (NH₄)₂SO₄ (25%; 134 g/liter) was added to the supernatant and dissolved by gentle stirring on ice for 30 min. Insoluble materials were removed by low speed centrifugation (30 min at 9,000 rpm). The (NH₄)₂SO₄ concentration was then raised to 55% (an additional 179 g/liter), and the proteins were precipitated during 30 min of stirring on ice. The precipitate was recovered by centrifugation and resuspended in buffer A1 (50 mM K₂HPO₄/KH₂PO₄, pH 6.8, 10% glycerol, 1 mM EDTA, 1 mM DTT, and 0.01% NP-40) containing 250 mM KCl and was dialyzed against the same buffer for 2 h before storage at −80C.

The resolvase assay was performed as follows: for each 10-μl reaction, 1 μl of extract containing 0.9, 1.8, and 3.5 μg of protein was assayed on 1 ng 5′-[³²P]end-labeled synthetic HJ DNA (X0) in phosphate buffer (60 mM Na₂HPO₄/NaH₂PO₄, pH 7.4, 5 mM MgCl₂, 1 mM DTT, and 100 μg/ml BSA) in the presence of 200 ng of competitor poly[dI:dC] DNA. Incubation was performed for 30 min at 37°C. Products were deproteinized and analyzed by native 10% PAGE. The percentage of resolution was quantified by a phosphoimager (Typhoon 9400 Scanner; Molecular Dynamics).

Fig. S1 shows aberrant splicing of the Rad51c transcript as a result of the presence of the PGK-neo cassette. Fig. S2 shows quantitative analysis of MLH1 foci formation in spermatocytes from control and infertile Rad51c^ko/neo mice. Figs. S3 and S4 show quantitative analysis of RPA foci (Fig. S3) and RAD51 foci (Fig. S4) in spermatocytes from control and infertile Rad51c^ko/neo mice. Fig. S5 shows chromosomes from control and Rad51c^ko/neo oocytes at metaphase I hybridized with a pancentromeric probe as well as sister chromatid cohesion and other defects at metaphase II.

We thank Drs. Jairaj Acharya, Patricia Hunt, and Philipp Kaldis for helpful discussions and critical review of the manuscript; Dr. Mary Ann Handel for a protocol for metaphase I chromosome preparation from spermatocytes; Martha Susiarjo and Aaron Batterbee for providing valuable technical guidance; Amy Gueye and Chistopher Leasure for technical help; Drs. Peter Moens and Ekaterina Revenkova for antibodies; Keith Rogers and his group in the Histotechnology Laboratory for excellent technical assistance; and Allen Kane of the Publication Department for illustrations.

This research was supported by the National Cancer Institute and the Department of Health and Human Services (grant to A. Nussenzweig, L. Tessarollo, and S.K. Sharan).