BRCA2 deficiency in mice leads to meiotic impairment and infertility

Since the cloning of the human breast cancer susceptibility genes BRCA1 and BRCA2, their roles in various biological processes has been extensively investigated (reviewed by Venkitaraman, 2002). Although it is well known that mutations in these genes predispose individuals to breast and ovarian cancer (Rahman and Stratton, 1998), the roles of the BRCA1 and BRCA2 proteins in other biological processes are not fully understood. Both BRCA1 and BRCA2 proteins have been implicated in DNA repair processes. The BRCA2 protein consists of 3418 amino acids with no clearly defined functional domains(Wooster et al., 1995; Tavtigian et al., 1996). Sequence analysis revealed presence of eight novel internal repetitive units,the `BRC-repeats' (Bork et al.,1996). A role for BRCA2 in DNA repair became evident when its interaction with the DNA repair protein RAD51 was identified(Sharan et al., 1997; Patel et al., 1998). The C terminus of the BRCA2 protein and six out of the eight BRC repeats was shown to directly interact with RAD51 (Sharan et al., 1997; Wong et al.,1997). A recent investigation of the crystal structure of the C terminus of BRCA2 revealed the presence of several single-stranded DNA-binding motifs, suggesting that BRCA2 could bind directly to DNA and stimulate RAD51 mediated recombination (Yang et al.,2002). Studies showing its upregulation in rapidly dividing cells,and its cell-cycle dependent expression peaking at the G₁/S boundary, suggest that BRCA2 may be involved in cell-cycle regulation(Rajan et al., 1996; Vaughn et al., 1996).

The involvement of BRCA1 and BRCA2 mutations in human ovarian cancer as well as the possible roles of the proteins in DNA dynamics has suggested the possibility that breast cancer susceptibility genes may function during normal gametogenesis. Until recently, evidence supporting this idea has derived primarily from expression studies. In mice, both Brca1 and Brca2 are expressed during spermatogenesis,particularly during meiotic prophase(Zabludoff et al., 1996; Blackshear et al., 1998; Chen et al., 1998). Intriguingly, the BRCA2 protein is localized on meiotic chromosomes(Chen et al., 1998), suggesting a role in the dynamics of meiotic recombination and/or chromosome pairing and synapsis.

However, the role of BRCA2 in meiosis and gametogenesis has been difficult to define because loss-of-function mutations in the mouse Brca2 gene result in early embryonic lethality (Hakem et al., 1998). We report a viable but infertile mouse model with impaired BRCA2 expression. The mutant mice lack endogenous Brca2function but carry a human BRCA2 gene present in a bacterial artificial chromosome (BAC). The human transgene rescues the embryonic lethality phenotype of the Brca2-null mice; however, there is poor expression of the human BRCA2 transgene in the testes and ovaries,with ensuing infertility. We use this mouse model to define stages of gametogenesis impaired by deficient BRCA2 expression and find evidence for a role in meiosis and gametogenic success in both males and females.

Human genomic BAC library filters obtained from Roswell Park Cancer Institute (now available at BACPAC Resource Center at the Children's Hospital Oakland Research Institute in Oakland, California) were screened using probes derived from PCR amplified DNA fragments from the 5′ (nucleotides 331-850) and 3′ (nucleotides 9852-10420) ends of the human BRCA2 cDNA (NM 000059). Thirteen clones were obtained from the human BAC library and seven were hybridized to both end probes. To identify the clone with the largest upstream and downstream regions, the T7 and SP6 ends of each BAC clone were sequenced and compared with the published sequence of human BRCA2 region. Human BRCA2 BAC, RP11-777I19 was selected to generate transgenic mice. The insert size (∼165 kb) was confirmed by pulsed field gel electrophoresis of NotI-digested BAC DNA.

BAC DNA was extracted using standard alkaline lysis and purification on a cesium chloride gradient. Supercoiled DNA was dialyzed overnight in 1×TE(10mM Tris pH 7.6, 1mM EDTA) and diluted to 0.5-1.0 ng/μl for pronuclear microinjection. Pronuclear microinjection was performed as described by Hogan et al. (Hogan et al.,1994).

Transgenic mice were identified by Southern analysis of BamHI digested tail DNA using the chloramphenicol resistance gene present in the BAC vector as a probe to detect a 8.0 kb fragment. Brca2 heterozygous and homozygous mutant mice were genotyped by hybridizing EcoRV digested genomic DNA on a Southern blot with a 1.5 kb fragment from exon 11 of murine Brca2 gene nucleotides 5208-6710 of NM 009765. The probe detects a 5 kb wild-type fragment and a 6.5 kb fragment corresponding to the mutant allele. BAC transgenic founders were crossed with Brca2 Ko/+ mice to obtain Brca2 Ko/+;Tg/+ F1 progeny. The Brca2 Ko/+;Tg/+ F1 mice were crossed to Brca2 Ko/+ mice to obtain Brca2 Ko/Ko;Tg/+ mice.

E9.5 and E11.5 embryos, ovaries and testes were fixed in freshly prepared 4% paraformaldehyde and embedded in paraffin wax for sectioning. In situ hybridization was performed as described by Tessarollo and Parada(Tessarollo and Parada,1995).

PCR fragments containing regions of mouse Brca2 (nucleotides 9500-10129, NM 009765) and human BRCA2 (nucleotides 10141-10770, NM 000059) cDNA were cloned into the pGEM-T vector (Promega) and sequenced to determine their orientation. ³⁵S-labeled-antisense human BRCA2 and mouse Brca2 probes were generated by in vitro transcription using T7 RNA polymerase and NotI linearized template. The corresponding sense probes were generated using SP6 RNA polymerase and NcoI (mouse Brca2) or SacII (human BRCA2) linearized templates. In situ hybridization was performed in triplicate for each sample and the experiments were repeated at least once to evaluate the consistency of the results.

E9.5 and E11.5 embryos, testes and ovaries were fixed in Bouin's solution. Samples were dehydrated through an ethanol series, embedded in paraffin wax,serially sectioned and stained with Hematoxylin and Eosin. Histological sections of testes were stained with periodic acid Schiff (PAS)-Hematoxylin and the ovaries from postnatal day 2 pups were stained with Weigert's Hematoxylin-picric acid Methyl Blue. Slides were examined using brightfield microscopy. To examine the consistency of the phenotype, samples from at least two different mice were examined with the exception of the testis from a 7-month-old mouse where a single male was examined.

A TUNEL assay (In Situ Cell Death detection kit, Roche Molecular Biochemicals) was performed according to the manufacturer's instructions on testes of 3-week-old mice. Testes were fixed in 4% paraformaldehyde and processed as described by Tessarollo and Parada(Tessarollo and Parada, 1995). Anti-fluorescein antibody conjugated with horse-radish peroxidase (POD) was used to detect fluorescein labeled dUTP. DAB substrate (Roche Molecular Biochemicals) was used to detect POD and visualized by light microscopy.

Cell preparations enriched in germ cells were prepared as previously described (Cobb et al., 1999). Briefly, testes were detunicated and digested in 0.5 mg/ml collagenase (Sigma)in Krebs-Ringer buffer for 20 minutes at 32°C and then in 0.5 mg/ml trypsin (Sigma) for 13 minutes, followed by filtering through 80 μm Nitex mesh and washing in buffer. Surface-spread preparations were used to visualize nuclei as previously described (Cobb et al., 1997). Cell preparations from at least two rescued mice were used for each antibody staining.

Antisera used were polyclonal human BRCA2 antibody, Ab-2 (Oncogene Research, 1:50), anti-mouse BRCA2 (Pep-3, polyclonal antisera against mouse BRCA2 amino acids 2330-2345, 1:50), anti-SYCP3 (1:500)(Eaker et al., 2001),anti-SPO11 (Trevigen, 1:25), anti-DMC1(Masson et al., 1999) (a generous gift from Madalena Tarsounas and Stephen West, Cancer Research UK),anti-RAD51 (Oncogene 1:100), anti-H1t (polyclonal antisera raised against full-length H1t cDNA, 1:1000), anti-γH2AX (Upstate; Lake Placid, NY,1:500) and anti-RPA (Oncogene 1:25). After overnight incubation in primary antibody, slides were incubated with rhodamine- or fluorescein-conjugated secondary antibodies (Pierce, 1:500), followed by mounting with Prolong Antifade (Molecular Probes) containing DAPI (Molecular Probes) to stain DNA. Control slides were stained with either secondary antibodies only, or pre-immune sera to replace primary antibody. Localization was observed with an Olympus epifluorescence microscope and images were captured and transferred to Adobe PhotoShop with a Hamamatsu color 3CCD camera.

Quantitative analysis of ovarian follicles was completed by marking every tenth section and counting the total number of primordial, primary, preantral and antral follicles present in each of these marked sections. Follicles were counted as primordial if they contained an intact oocyte and were surrounded by a single layer of flattened granulosa cells. Follicles were counted as primary if they contained an intact oocyte and were surrounded by a single layer of cuboidal shaped granulosa cells; as preantral if they contained an intact oocyte, more than one layer of granulose cells and lacked antral spaces; and as antral if they contained an intact oocyte, more than one layer of granulosa cells and antral spaces. To avoid double-counting follicles, only follicles containing an oocyte with a visible nucleus were counted. The number of follicles was multiplied by 10 to account for the fact that every tenth section was used in the analysis (Smith et al., 1991). To avoid bias, all ovaries were analyzed without knowledge of genotype or age.

Females were superovulated and oocytes were collected as described by Hogan et al. (Hogan et al., 1994). Briefly, 0.1 ml of PMS (5 IU) was injected into the intraperitoneal cavity of 6-week-old females. After 46 hours, 0.1 ml of hCG (5 IU) was injected into each female. After 18 hours the females were sacrificed and the oviducts were dissected out. The oviducts were opened with fine forceps and the eggs were collected. Cumulus cells were dissociated in presence of 700 units/ml of hyaluronidase.

Forty-eight hours after PMSG injection, germinal vesicle (GV)-intact oocytes were collected from ovarian follicles by piercing the ovaries of two control and two rescue mice (7 and 9 weeks of age) with 27.5 gauge syringe needles in Whitten's medium containing 0.05% polyvinyl alcohol (Sigma, St. Louis, MO) and 0.25 mM dibutyryl cAMP (dbcAMP, Sigma)(Cho et al., 1974). All oocytes collected, those enclosed in cumulus cells and those not enclosed in cumulus cells, were used in studies of in vitro maturation. For those oocytes that emerged from follicles enclosed in cumulus cells, the cumulus cells were removed by pipetting the oocyte-cumulus cell complexes through a thin-bore pipet. GV-intact oocytes were matured in vitro by washing the oocytes through six 100 μl drops of Whitten's medium lacking dbcAMP, and then cultured overnight at 37°C in 100 μl drops Whitten's medium (10-25 oocytes per drop) in 5% CO₂ in air. At 16 hour post-dbcAMP removal, the cells were viewed by dissecting microscope to assess what percentage of the oocytes had emitted the first polar body (PB1), undergone GV breakdown (GVBD) but not emitted PB1, were still GV intact, or had other phenotypes. Less than 50% of the control oocytes had emitted the first polar body by this time. As oocytes of some strains of mice undergo meiotic maturation more slowly than others(Polanski et al., 1998), we cultured these oocytes for additional 3 hours to 19 hours post-dbcAMP removal. Some oocytes remained GV intact after these 19 hours [27% (22/81), control;14% (4/29)]; this might be attributable to the use of all oocytes collected(i.e. potentially including some meiotically incompetent oocytes) in these maturation studies. At 19 hour post-dbcAMP removal, the cells were prepared for staining as follows. The zonae pellucidae were removed by a brief incubation in acidic medium-compatible buffer (10 mM HEPES, 1 mM NaH₂PO₄, 0.8 mM MgSO₄, 5.4 mM KCl, 116.4 mM NaCl, final pH 1.5), and then the cells were fixed in freshly prepared 3.7%paraformaldehyde in PBS for 1-2 hours. Fixation and all subsequent steps were performed at room temperature in a humidified chamber as described previously(Evans et al., 2000). The fixed cells were washed in PBS, then permeabilized in PBS containing 0.1%Triton X-100 for 15 minutes, and then incubated in blocking solution (PBS containing 0.1% BSA and 0.01% Tween-20) for 60 minutes. Cells were stained with 50 ng/ml of TRITC-conjugated phalloidin (Sigma) for 45 minutes to label F-actin. Cells were then washed three times (in blocking solution, 10-20 minutes each), and then mounted in VectaShield mounting medium (Vector Labs;Burlingame, CA) containing 1.5 μg/ml 4′,6-diamidino-2-phenylindole(DAPI; Sigma). Cells were viewed on a Nikon Eclipse fluorescent microscope,and digital images were captured with a Princeton 5 MHz cooled interlined CCD camera (Princeton Instruments, Trenton, NJ) using IP Labs software(Scanalytics, Fairfax, VA). All images were collected with similar exposure times and were not further manipulated, except for cropping in Photoshop 6.0(Adobe systems Incorporated, San Jose, CA) for figure preparation. Throughout this work, the term `oocyte' refers to GV-intact oocytes at prophase I; the term `egg' refers to a metaphase II egg.

A BAC clone containing full-length human BRCA2 gene was obtained from a genomic library generated in BAC vector pBACe3.6. The BAC clone,RP11-777I19, with a 165 kb insert that included the 80 kb BRCA2 gene and 30 kb of upstream and 55 kb of downstream region, was used to generate transgenic mice (Fig. 1A). Four founder transgenic mice carrying one to three copies of the BAC (data not shown) were then crossed to mice heterozygous for a mutant allele of Brca2. The heterozygous mice (Brca2^Brdm1/+, in brief represented as Brca2 Ko/+) are viable and fertile, whereas the homozygous mutant (designated as Brca2 Ko/Ko) embryos die during development around E7.5 (Sharan et al.,1997). The expression of the human transgene in multiple tissues was examined in 6- to 8-week-old Brca2 heterozygous transgenic F1(Brca2 Ko/+; TgN(RP11-777I19)1sks/+, in brief represented as Brca2 Ko/+; Tg/+) animals by RT-PCR. Of the four lines that were examined, only one showed expression of the human gene (data not shown).

The Brca2 Ko/+; Tg/+ mice were crossed to Brca2 Ko/+mice, producing Brca2 Ko/Ko; Tg/+ mice(Fig. 1B). These rescued homozygous mutant mice were viable and morphologically indistinguishable from their littermates. The mutant mice were obtained in the expected Mendelian ratio (P=0.59) (Fig. 1C), suggesting no embryonic lethality. The three lines that did not express the human BRCA2 gene failed to rescue the lethality of the Brca2 homozygous embryos (data not shown). The observation that the BRCA2 transgene rescued embryonic lethality of the homozygous mutant mice suggests that the human gene is functional in mice. However, the rescued mice failed to produce offspring when intercrossed. When the Brca2 Ko/Ko; Tg/+ mice were crossed to wild-type mice, both wild-type and rescued females showed presence of copulatory plugs. However, neither Brca2 Ko/Ko; Tg/+ females nor wild-type females mated to Brca2 Ko/Ko; Tg/+ males produced any pups, suggesting that both Brca2 Ko/Ko; Tg/+ males and females were infertile. The fact that Brca2 Ko/+; Tg/+ mice are fertile strongly suggests that the infertility of the Brca2 Ko/Ko; Tg/+ mice is not due solely to a deleterious effect of the BAC, but is caused by loss of endogenous BRCA2 protein function.

To investigate the rescue of embryonic lethality by the human BRCA2-containing BAC, the expression of the human transgene was examined by RT-PCR. Although the level of expression of the human gene appeared lower than that of the endogenous Brca2 gene, the two exhibited qualitatively identical expression patterns revealed by RT-PCR. Both were expressed in brain, heart, liver, lung, kidney, spleen, ovary, testis,thymus and the mammary gland of pregnant and lactating females (data not shown). Interestingly, neither the mouse nor the human gene was expressed in the mammary gland of virgin females. These RT-PCR results suggest that the tissue-specific regulation of the human gene under the control of its own promoter is similar to the endogenous mouse Brca2 gene.

Expression was further examined in embryos, where expression of endogenous Brca2 is first detected at E7.5(Sharan et al., 1997). As development progresses, Brca2 expression appears to be upregulated in cells that are rapidly dividing while reduced in cells undergoing differentiation. We examined the expression of the BRCA2 transgene and compared it with the endogenous Brca2 expression pattern in embryos at E9.5 by in situ hybridization. The human BRCA2 gene was expressed in the developing embryo and the pattern of expression was identical to that of the endogenous Brca2 gene(Fig. 2B-D). However,expression of the human BRCA2 gene was reduced compared with the mouse gene. At E11.5 expression of both the human transgene and the mouse gene was found to be restricted to regions of rapid cell division (data not shown). From the expression pattern seen in the brain, both the mouse and the human genes were expressed in the neuroepithelium of the ventricular layer. Similar to E9.5 embryos, expression of the human BRCA2 gene in E11.5 embryos was considerably lower than that of the endogenous gene. The qualitatively identical expression pattern of the mouse and human genes during embryogenesis provides insight into the viability of the Brca2 homozygous mutant mice. Although the level of transcript expression is reduced, it must provide an amount of BRCA2 protein that is above the threshold level required for normal function during embryogenesis.

To assess further the expression of the human BRCA2 transgene in testes and ovaries, we performed in situ hybridization with mouse Brca2- and human BRCA2-specific probes(Fig. 2). Although spermatocytes of 3-week old males showed high expression of the mouse gene(Fig. 2F), the human gene expression was barely above the background signal(Fig. 2G,H). In ovaries from a 3-week-old mouse, the endogenous Brca2 gene was expressed in all the developing follicles as described previously(Fig. 2J), but there was lack of any specific signal of the transgene in these follicles(Blackshear et al., 1998). Similarly, in a 10-week-old ovary, expression of the endogenous mouse Brca2 gene was detected in the oocyte and the surrounding granulosa cells but not in the corpora lutea, while the level of human BRCA2mRNA was not above background levels (Fig. 2N-P). The weak expression of the human BRCA2 gene apparently did not lead to detectable protein expression, as we failed to observe the human BRCA2 protein in the ovaries and testes by immunohistochemistry (data not shown).

We conclude from these expression analyses that the infertility of the rescued animals is due to poor expression or function of the human BRCA2 protein in germ cells. Thus, the Brca2 Ko/Ko; Tg mice, in effect, are equivalent to a conditional knockout with respect to the gonads and can be used to study the function of BRCA2 in gametogenesis.

None of the four Brca2 Ko/Ko; Tg/+ males produced any offspring after repeated mating for 2-6 months with either rescued or wild-type females,while their Brca2 Ko/+; Tg/+ littermates sired multiple litters suggesting that the rescued males were infertile. To define further the infertility phenotype, the histology of the testes from males of different ages was examined. Testes of the rescued mutant males were small compared with controls. Histological examination of 3-week-old Brca2 Ko/+; Tg/+males revealed seminiferous tubules that were filled with spermatocytes(Fig. 3A). The rescued male testes also showed the presence of spermatocytes in seminiferous tubules but their number appeared reduced and some seminiferous tubules were completely devoid of germ cells (Fig. 3B). The seminiferous tubules of 2-month-old Brca2 Ko/Ko; Tg/+ males appeared smaller compared with their normal littermate controls and showed interstitial cell hyperplasia (Fig. 3C,D). The Leydig cells appeared to be functionally normal as the weight of the seminal vesicles was unaffected in the mutant mice (data not shown). No germ cells beyond early meiotic prophase were present in the seminiferous tubules of the rescued males. Several tubules were completely devoid of germ cells and contained only Sertoli cells, suggesting that the germ cells were undergoing rapid degeneration. With germ cell deficiency,there was an increase in the relative number of Sertoli cells per seminiferous tubule in the mutants compared with the control tubules. The degeneration was more severe in 7-month-old rescued males, where very few seminiferous tubules had spermatocytes and majority of the tubules were filled with vacuolated Sertoli cells (Fig. 3E). The nuclei of these remaining spermatocytes appeared condensed and darkly stained.

To determine if the loss of germ cells from the seminiferous tubules was associated with cell death, the TdT-mediated fluorescein-dUTP nick-end-labeling (TUNEL) assay was used to detect the single and double strand DNA breaks that occur during apoptosis. Testes from 3-week-old rescued males were analyzed and compared with Brca2 Ko/+; Tg controls. Although very few TUNEL-positive spermatocytes were seen in the seminiferous tubules of control testis, the seminiferous tubules of the rescued animals had a large number of TUNEL-positive cells(Fig. 3F,G). This suggests that the BRCA2-deficient spermatocytes that are arrested in meiotic prophase undergo apoptosis, causing germ-cell depletion.

As spermatocytes in testes of Brca2 Ko/Ko; Tg/+ rescued animals appeared to be blocked during prophase I of meiosis, spermatocytes were examined by immunofluorescence with various antibodies to determine more precisely the timing of the block in prophase I. First, the localization of the endogenous mouse BRCA2 protein during normal meiosis was confirmed. As seen in Fig. 4A, signal for BRCA2 protein in zygotene spermatocytes is diffuse throughout the nucleus. As synapsis occurs in normal spermatocytes, BRCA2 staining begins to disappear and is no longer detected by the end of pachynema (data not shown) in accordance with previous findings (Chen et al., 1998). Mouse BRCA2 protein was not detected in Brca2 Ko/Ko; Tg/+ spermatocytes (Fig. 4B). Similarly, no human BRCA2 protein was detected in either the heterozygous or homozygous mutant transgenic spermatocytes, confirming the negligible expression of the human protein in spermatocytes of transgenic mice(Fig. 4C,D).

The localization of synaptonemal complex protein SYCP3(Dobson et al., 1994) in BRCA2-deficient spermatocytes revealed partial, but not complete synapsis of chromosomes, suggesting that the spermatocytes are blocked before or at the transition from zygonema to pachynema. To provide further evidence, we examined the expression of male germ-cell specific histone H1t protein, which appears during midpachynema (Cobb et al.,1999). Although the chromatin of control spermatocytes stains positively for histone H1t, the chromatin of spermatocytes in rescued mutants does not, suggesting arrest of both meiotic and spermatogenic developmental programs (Fig. 4E,F).

Events surrounding the initiation of meiotic recombination in the BRCA2-deficient spermatocytes were examined by determining the localization of two proteins known to be associated with the formation of recombination-related DNA double-strand breaks. First, we assessed the presence of SPO11, an evolutionarily conserved protein required for normal formation of meiosis-specific double strand breaks(Keeney et al., 1997). Positive staining for SPO11 protein was found in both Brca2 Ko/Ko;Tg/+ and Brca2 Ko/+; Tg/+ control spermatocytes(Fig. 4G,H). Second, we determined the presence of the phosphorylated form of histone H2AX(γ-H2AX), a marker for chromatin containing DNA double-strand breaks. Brca2 Ko/Ko; Tg/+ spermatocytes show positive γ-H2AX staining,again suggesting that the chromatin in BRCA2-deficient spermatocytes contains DNA with double-strand breaks. Normally, γ-H2AX staining disappears in pachynema, except for its localization with the sex chromosomes(Mahadevaiah et al., 2001). However, mutant spermatocytes do not exhibit either loss or re-localization of staining with anti-γ-H2AX, providing further evidence that they are blocked at a stage prior to pachynema.

As BRCA2 interacts with RAD51 protein(Patel et al., 1998) and co-localizes with RAD51 in human spermatocytes(Chen et al., 1998), we examined localization of RAD51 protein and its meiosis-specific variant, DMC1,in BRCA2-deficient spermatocytes. Western blot analysis of the testes of Brca2 Ko/Ko; Tg/+ mice showed presence of RAD51 and DMC1 proteins at levels similar to the control testes (data not shown). Interestingly, the number of RAD51 foci per nucleus was reduced in Brca2 Ko/Ko; Tg/+spermatocytes compared with heterozygous transgenic control spermatocytes(Fig. 5A-C). Analysis of RAD51 foci in mutant and control nuclei revealed that 98% of mutant spermatocytes in the leptotene and zygotene stages had reduced numbers of RAD51 foci (58% with no foci and majority of nuclei in 1-100 group had less than 10 foci and only a few had 20-30), while greater than 98% of control spermatocytes had 100-250 RAD51 foci per nuclei (Fig. 5C). Additionally, although many discrete DMC1 foci were visible over zygotene-stage spermatocytes in normal animals, few to no foci were detected in BRCA2-deficient spermatocytes, similar to the reduction seen in RAD51 foci (Fig. 5D,E). Finally, we examined the expression of replication protein A (RPA), a single-stranded DNA-binding protein involved in DNA replication and repair,which is associated with synapsed regions of the chromosomes during the zygotene and pachytene stages (Plug et al., 1997; Plug et al.,1998), and has been shown to interact with RAD51 (Golub, 1998). Surprisingly, despite reduced number of RAD51 foci and lack of extensive synapsis, we found abundant RPA foci present in mutant spermatocytes(Fig. 5F,G).

To determine the phenotype of the BRCA2-deficient ovaries (obtained from the rescued females, Brca2 Ko/Ko; Tg/+), we examined their histology at various ages and compared them with the heterozygous transgenic ovaries(Brca2 Ko/+; Tg/+). Ovaries from postnatal day 2 (PND2) rescued females were morphologically similar to those from heterozygous transgenic females (Fig. 6A,B). Rescued female ovaries contained primordial and primary follicles and their number was similar to control ovaries. The ovaries of 3-week-old females also appeared indistinguishable from the littermate control ovaries(Fig. 6C,D). The rescued ovaries contained morphologically normal follicles in all stages of development. However, at this time point the number of primordial and primary follicles were eight- to tenfold reduced in the Brca2 Ko/Ko; Tg/+females (P=0.005 and 0.013, respectively, Fig. 6G). These data suggest massive loss in the oocyte pool between PND2 and 3 weeks. By contrast, there was no significant reduction in number of preantral follicles between the two genotypes (P=0.32). At 6 weeks of age, the rescued ovaries appeared smaller in size than heterozygous transgenic ovaries. The rescued ovaries contained follicles but their number was severely reduced(Fig. 6E,F). The follicles appeared degenerating and the surrounding layer of granulosa cells was poorly organized compared with follicles from ovaries of heterozygous transgenic littermates.

To determine if their infertility could be rescued by hormonal treatment,6-week-old females were treated with pregnant mare serum gonadotropin (PMSG)to stimulate follicle maturation and with human chorionic gonadotropin (hCG)to initiate ovulation. The number of eggs was significantly reduced in rescued females compared with controls (P=0.004). Although each heterozygous transgenic oviduct contained on an average 17.6±3.2 eggs(n=5), oviducts of rescued females contained 7.6±3.6 eggs(n=5). To determine the developmental potential of the oocytes, some of the superovulated females were mated with wild-type males following the hCG injection. The females were sacrificed after 12.5 days and the embryos were dissected from the uterine horns. Morphologically normal appearing embryos were harvested from some of the rescued females. However, the mean number of implanted embryos was 1.33±1.5 (n=3) compared with 14(n=2) in control females. This suggested that the fertilization and/or developmental competence of the oocytes in the rescued mice were in some way compromised.

Because meiotic progress is arrested in mutant males, we hypothesized that there could be some abnormalities in meiosis in females as well. However, the presence of oocytes in young animals suggested that impairment to meiotic progress could be at a stage after the late prophase, or dictyate, arrest that normally occurs during female gametogenesis. Successful oocyte `meiotic maturation' or progress through meiosis from prophase I to metaphase II is crucial for successful fertilization and embryo development.

To address our hypothesis, oocytes from control and rescued females were examined for their abilities to undergo meiotic maturation in vitro. Ninety GV-intact oocytes were collected from two control females (55 from a 9-week-old and 35 from a 7-week-old) and 31 oocytes were collected from two rescued females (21 from a 9-week-old and 10 from a 7-week-old), demonstrating that the rescued females had fewer germinal vesicle (GV)-intact oocytes than did the controls, as was anticipated from the ovarian histology. Additionally,the oocytes from rescued females showed a reduced ability to complete oocyte maturation successfully when compared with controls. GV-intact oocytes from control and rescued females were cultured for 16 hours in medium lacking dbcAMP to allow progression into meiosis from prophase I to metaphase II(Cho et al., 1974). Entry into meiosis is indicated by the breakdown of the nuclear envelope of the GV, known as GV breakdown (GVBD). Completion of the first stage of female meiosis(progression to metaphase II) is indicated by the emission of the first polar body (PB1), the product of asymmetric cytokinesis. As shown in Fig. 7, oocytes from the rescued females appeared capable of initiating meiotic maturation, as only 14%(4/29) remained arrested at GV stage after 16 hours of culture. However,abnormalities of meiotic maturation were observed; most notably, 21% (6/29) of the polar bodies appeared to be abnormally large, and a low number (9/29, 31%)had normal first polar bodies. It should be noted that the PB1 emission in the controls by 16 hours was relatively low (44%). Because oocytes of some strains of mice undergo meiotic maturation more slowly(Polanski, 1986), we cultured the oocytes that had begun maturation (59 controls, 25 rescued) for 3 additional hours, to 19 hours post-dbcAMP removal. In this time, ∼50% of the GVBD oocytes in both the control and rescue groups progressed to emit the first polar body (Fig. 7B). To asses meiotic progress further, cells then were fixed and stained to label the F-actin and DNA (four out of the 59 control oocytes were lost in this process,and therefore data are reported for 55 oocytes.) Normal metaphase II-arrested eggs show first polar body (indicative of completion of meiosis I), chromatin arranged on the metaphase II plate and an actin-rich cap over the meiotic spindle (Fig. 8A). Although nearly 50% (12/25) of the rescued oocytes had this phenotype, abnormalities were observed in the others. The most common anomaly was an oversized polar body-like structure (7/25, 28%; Fig. 8B,C). While eggs with large polar bodies have been observed in some instances of female meiosis abnormalities (e.g. the Mos-deficient and endothelial nitric oxide synthase-deficient mice(Choi et al., 1996; Jablonka-Shariff and Olson,1998), the eggs with oversized polar body-like structures observed here were different. In the majority of these eggs (6/7), all of the DNA was in the PB-like structures and no DNA present in the oocyte(Fig. 8B,C). Other abnormalities in the oocytes from rescued females included apparent impairment of the first meiotic division, including abnormal chromatin condensation(Fig. 8D), chromatin condensation but little organization of the metaphase I spindle(Fig. 8E), formation of a metaphase I spindle but failure to emit PB1(Fig. 8F), or failure to progress to metaphase II and instead of having an abnormal, tightly condensed chromatin `wad' in an egg that had emitted PB1(Fig. 8G). By contrast, 85%(47/55) of the control oocytes emitted the first polar body and 71% (39/55)showed proper metaphase II arrest. The other eight control cells with PB1s had slight metaphase II abnormalities, namely incomplete chromosome congression to the metaphase II spindle (Fig. 8H). Of the eight control oocytes that had undergone GVBD but not emitted PB1, seven of these appeared to be normal metaphase I cells, with the meiotic spindle positioned at the periphery of the cell.

A unique mouse model, deficient for endogenous BRCA2 protein, but carrying a BAC containing the human BRCA2 gene, rescues the embryonic lethality of Brca2-null mice. In spite of poor sequence identity, the genetic complementation of a mutation in the mouse Brca2 gene by the human BRCA2 transgene sheds light on conservation of function of the mouse and human proteins (Sharan and Bradley, 1997). Similar functional conservation was described for the human BRCA1 gene, which can fully complement the Brca1mutant phenotype (Lane et al.,2000; Chandler et al.,2001). For reasons not known, the human BRCA2 transgene is poorly expressed in gonads with ensuing infertility. The infertility of the rescued mice is due to loss of endogenous BRCA2 protein function and not to any detrimental effect of the BAC, as the Brca2 Ko/+; Tg/+ mice are fertile even though they have the same copy number of the BAC transgene as the Brca2 Ko/Ko; Tg/+ mice. This model provides evidence for key roles of BRCA2 protein in both male and female gametogenesis, particularly in the meiotic stages. Interestingly, the meiotic phenotype of BRCA2 protein deficiency is variable: in males there is arrest of meiosis in early prophase,while in females some oocytes are capable of progressing through the meiotic division phase and giving rise to embryos, although with a high frequency of abnormalities.

Previously, the expression of BRCA2 protein in adult testes and ovaries and its colocalization with RAD51 on synapsed chromosomes in spermatocytes had suggested a role for this protein in meiosis(Chen et al., 1998), but this was difficult to test experimentally because of the lethality of Brca2-null embryos and the early germ-cell loss in mice homozygous for either of the two hypomorphic alleles of Brca2(Connor et al., 1998; Friedman et al., 1998). Here,we have circumvented these problems by showing that Brca2-null mice rescued by a human BRCA2 transgene survive to adulthood, but with impaired germ-cell development. As the human BRCA2 transgene is poorly expressed in the gonads and human protein could not be detected in ovaries or testes, we presume that infertility results from BRCA2 deficiency in the gonads, especially in the germ cells.

The phenotype of spermatocytes in Brca2 Ko/Ko; Tg/+ males reveals a requirement for adequate levels of BRCA2 protein for normal meiotic prophase progression, including assembly of protein complexes onto chromosomal axes. Although synapsis is initiated in BRCA2-deficient spermatocytes, it is never completed and mutant spermatocytes do not reach the pachytene stage of meiotic prophase. The presence of both of SPO11 and γ-H2AX proteins on chromatin of mutant spermatocytes suggests that early recombination-related DNA double-strand break formation probably occurs in BRCA2-deficient spermatocytes. However, recombination complexes of RAD51 and DMC1 are not appropriately assembled onto meiotic chromosomal axes in the absence of BRCA2 protein. Thus, the role of BRCA2 in meiotic recombination could be similar to that hypothesized for double-strand break repair in mitotic cells, where it is thought that BRCA2 recruits RAD51 to sites of DNA damage, thus initiating repair(Davies et al., 2001). The BRCA2 protein may also be directly or indirectly responsible for recruiting DMC1 protein to meiotic chromosomal axes at the site of RAD51 foci, as DMC1 localization was dramatically diminished in mutant spermatocytes. In normal spermatocytes, RAD51 and DMC1 colocalize and interact with SYCP3(Tarsounas et al., 1999),suggesting functional interaction of these proteins in recombination events. The presence of RAD51 foci in DMC1-deficient spermatocytes(Pittman et al., 1998; Yoshida et al., 1998) implies that DMC1 is not required for localization of RAD51 onto chromosomal axes. However, the role of RAD51 in DMC1 recruitment to chromatin is not known, as there is early embryonic lethality of Rad51-null mice(Lim and Hasty, 1996; Tsuzuki et al., 1996). Hence,the reduced number of DMC1 foci in BRCA2-deficient spermatocytes could be because of lack of BRCA2 or because of failure of RAD51 to load onto chromatin. In spite of greatly diminished RAD51 and DMC1 localization in BRCA2-deficient spermatocytes, numerous foci of RPA are found on chromosomal axes. RPA and RAD51 do not always colocalize in spermatocytes(Tarsounas and Moens, 2001)and the pattern of RPA localization in BRCA2-deficient spermatocytes suggests that neither BRCA2 nor RAD51 may be required for RPA localization on chromosomal axes, but that BRCA2- or RAD51-mediated processes could be required for removal of RPA from chromosomal axes. Taken together, these observations of BRCA2-deficient spermatocytes suggest a pivotal role for BRCA2 protein in organizing the complex of proteins that assemble onto chromosomal axes during early meiotic prophase. Additionally, the failure of spermatocytes to progress in their developmental program suggests a strong dependency upon completion of early events of recombination.

Although it is not known if absence of BRCA2 protein can cause some prophase arrest of oocytes, it is clear that some oocytes survive and progress to the end of meiotic prophase, where they become normally arrested. Thus,there is sexual dimorphism between spermatocytes and oocytes in their dependency on BRCA2-mediated meiotic events. However, the fact that we do detect the human BRCA2 transcript by RT-PCR in the gonads, the possibility that some human protein may be present cannot be ruled out. This level may be beyond the level of detection by immunohistochemistry but may be sufficient to partially rescue the phenotype. Therefore, the possibility that a low level of BRCA2 protein in the ovary might be sufficient for meiotic maturation in females but not in males cannot be ruled out. Nonetheless,oocyte survival, the progress of oogenesis and acquisition of meiotic competence is indeed compromised by the absence of BRCA2; there is a marked reduction in the number of oocytes in the rescued females. Even though the rescued females appear to make some normal GV-intact, diplotene oocytes,multiple abnormalities are seen in oocytes undergoing maturation in vitro. Additionally, after gonadotropin treatment, there were 2.5-fold fewer metaphase II eggs from the BRCA2-deficient females compared with the controls,and 10-fold fewer E12.5 embryos in rescued females compared with the controls. These analyses of both in vitro and in vivo maturation, fertilization and embryogenesis suggest that the oocytes from rescued BRCA2-deficient females have reduced competence to undergo maturation and fertilization and to support embryonic development. Taken together, these abnormalities reveal a complex role for BRCA2 in oogenesis, and raise the possibility that all oocytes might be abnormal, but some escape arrest of development.

Thus, aspects of the meiotic phenotype of BRCA2 deficiency appear sexually dimorphic: spermatocytes arrest in early meiotic prophase, while at least some oocytes survive to the end of meiotic prophase, but many exhibit abnormalities of subsequent maturation and developmental competence. Similar sexually dimorphic phenotypes have been observed for many meiotic mutations(Hunt and Hassold, 2002). Interestingly, these include many of those mutations where the male phenotype is meiotic arrest at or around the transition from zygonema to pachynema: Brca2 (this report), Msh4(Kneitz et al., 2000), Scp3 (Yuan et al.,2000; Yuan et al.,2002) and Mei1 (Libby et al., 2002); in these mutant female mice, oocytes can progress through meiotic prophase. Taken together, these phenotypes reveal profound differences between male and female meiosis in mice. At the moment, it is not possible to discriminate between two alternative explanations: either the mechanics of meiosis differ between sexes in such a way that the relevant proteins are not required for recombination and meiotic progress in oocytes or meiotic checkpoints present in spermatocytes are lacking or inefficient in oocytes.

The phenotype of the rescued Brca2 mutant mice suggests that human BRCA2-mutation carriers could also be predisposed to fertility problems. In one recent study (Vachon et al., 2002), individuals with a family history of breast and ovarian cancer were found to be at risk for nulliparity. Thus, given the proposed role of BRCA2 in DNA repair in somatic as well as germ cells, cancer predisposition and infertility may represent two different tissue-specific phenotypes of mutation in a single gene, more fully revealing its function.

We thank Drs Allan Bradley, Neal Copeland, John Eppig, Amy Inselman, Bruce McKee and Sundar Venkatachalam for helpful discussions and critical review of the manuscript; Madalena Tarsounas and Stephen West for DMC1 antibody; Jairaj Acharya for help with the TUNEL assay; Keith Rogers and his group in the Histotechnology Laboratory for excellent technical assistance; Dr Dahlem Smith for help with histopathological analysis; and Richard Frederickson of the Publication Department for the illustrations. This research was supported by the National Cancer Institute, the Department of Health and Human Services(S.K.S. and L.T.), National Institutes of Health grants HD 37696 (J.P.E.), HD 38955 (J.A.F.) and HD 33816 (M.A.H.).