Cancer Susceptibility of Mice with a Homozygous Deletion in the COOH-Terminal Domain of the Brca2 Gene¹ Free

Women who inherit germ-line defects of the BRCA2 breast cancer susceptibility gene have up to an 85% risk of breast cancer development by 70 years of age (1, 2). BRCA2 mutation carriers also have increased risks for a variety of cancers including ovarian, pancreatic, colon, prostate, stomach, laryngeal, thyroid, male breast cancer, and ocular melanoma (3). Substantial evidence from studies of human and mouse cells indicates that the BRCA2 protein is an important component in DNA damage response pathways, and loss of this function is considered a major factor in cancer predisposition (4, 5). Brca2-mutant cells and embryos are hypersensitive to ionizing radiation and other DNA-damaging agents and develop numerous spontaneous chromosomal abnormalities (6, 7, 8, 9, 10). In addition, several domains of the BRCA2 protein appear to interact directly with RAD51, a protein having distinct roles in normal meiotic and mitotic recombination, DNA damage repair, and chromosome segregation. These BRCA2 domains include the BRC repeats in exon 11 and a highly conserved RAD51 binding domain in exon 27 (6, 11, 12, 13). The interaction of Brca2 with Rad51 is pivotal for the role Brca2 plays in the activation of double-strand break repair and/or homologous recombination, and disruption of these processes may predispose individuals to cancer development.

Functional studies of Brca2 in adult tissues have been hindered by the embryonic lethal phenotypes of Brca2 knockout mice reported to date (14). We and others previously described mutant mice with disruptions in the 5′ half of Brca2 that result in embryonic lethality during early to mid-gestation (6, 15, 16, 17). The few homozygous mutant Brca2 mice that have been reported to survive to adulthood display gross developmental abnormalities, are infertile, and develop thymic lymphomas by 5 months of age (7, 8). Here, we describe the creation and initial phenotypic characterization of mutant mice that carry a homozygous germ-line deletion of exon 27 of the Brca2 gene. We chose to target exon 27 because a Rad51 binding domain has been identified between Brca2 amino acids 3196–3232 in yeast two-hybrid studies (6, 11). In addition, this COOH-terminal domain of the Brca2 protein contains nuclear localization signals that have been highly conserved among species (18, 19). Finally, cells lacking this COOH-terminal domain are hypersensitive to γ-radiation (10), and they have recently been shown to be deficient in error-free, homology-directed DNA repair (20, 21). The Brca2-mutant mice described in this report exhibit a low penetrance of perinatal lethality and an increased susceptibility to spontaneous tumorigenesis in a variety of tissues.

A targeting construct was generated that contained exon 27 of the Brca2 gene flanked by loxP sites (Fig. 1). The 5′ targeting fragment consisted of a 4-kb EcoRI fragment containing exons 25, 26, and 27, and this was subcloned from a genomic mouse bacterial artificial chromosome clone (18). Likewise, a 3-kb NsiI fragment distal to the 3′ untranslated region of Brca2 was also subcloned. Double-stranded loxP oligonucleotides flanked by appropriate restriction sites were inserted into a MunI site in intron 26 of the 5′ targeting fragment and a BamHI site of the 3-kb Nsi targeting fragment. Both fragments were then inserted into a previously described pgkNeoTK targeting vector (17). After linearization with SalI, this targeting vector was introduced into 129/Ola-derived BK-4 ES³ cells by electroporation as described previously (17). Electroporated cells were subjected to positive and negative selection with geneticin (250 μg/ml; Life Technologies, Inc., Rockville, MD) and gancyclovir (2 μm; Roche, Hertfordshire, United Kingdom). A properly targeted ES cell clone with the “floxed” Brca2 allele (Brca2^Flox27) was identified by Southern analyses with unique probes outside the Brca2 targeting construct as well as PCR analyses using loxP-specific primers.

A Cre expression plasmid (generously provided by Dr. Robert Sobol, National Institute of Environmental Health Sciences) was transiently electroporated into ES cells carrying a single floxed Brca2 allele. This Brca2 allele was successfully deleted in ∼10% of the floxed ES cells as determined by PCR using primers that flanked the 5′ and 3′ loxP sites (Fig. 1). The Brca2^Wildtype and Brca2^Flox27 alleles were amplified using the following PCR primers: B2F1, 5′-GGAGGAGGAGGAGTTGTTGA-3′ and B2R1, 5′-ATCTCGTTCTCTCCACTCCA-3′, whereas the Brca2^Δ27 allele was detected using primers B2F1 and B2R2, 5′-CAAAAAAGCCCAGATGATGAG-3′. Brca2^Δ27 ES cells were then injected into C57BL/6N blastocysts and transplanted into pseudopregnant CD-1 females. Eight chimeras were generated, and germ-line transmission of their Brca2^Δ27 allele was obtained by standard breeding techniques. Genomic DNA from pups was isolated from tail biopsies at weaning for genotyping by PCR.

Mice carrying the Brca2^Δ27 mutation were backcrossed from a targeted 129-derived chimeric founder onto the C57BL/6N background for one to three generations, followed by intercrosses to generate mice homozygous for the Brca2-exon 27 deletion. Both virgin and multiparious females as well as males carrying this Brca2^Δ27 mutation were group housed in plastic cages on pressed wood-chip bedding. Animals had access to an NIH-31 diet (18% protein, 4% fat, and 5% fiber; Zeigler Bros., Gardeners, PA) and water ad libitum. Mice were monitored for 17 months and were sacrificed when palpable tumors developed or when signs of morbidity became apparent. Terminal sacrifices were performed on all surviving animals between 17 and 19 months of age. All tissue and tumor samples were removed postmortem and fixed in 10% neutral buffered formalin. Specimens were processed for routine histology, embedded in paraffin, sectioned, and stained with H&E. Histological examination was performed by veterinary pathologists (C. H. and B. D.). In all female mice, the right #4 and #5 mammary glands were fixed with neutral buffered formalin on the pelts for 24 h, dissected, fixed, and then stained with carmine as described previously for whole-mount analysis (22). Routine H&E staining was performed on the left #4 and #5 mammary gland of each animal, followed by histopathological analysis.

A χ² goodness-of-fit test was performed for segregation analysis of the three genotypic classes from heterozygous intercrosses. The overall difference in survival between Brca2^Δ27//Δ27 and Brca2^Δ27/+ virgin females was compared with a life table test. Both a Peto analysis (23) and a Fisher’s exact test were used to test for a difference in tumor rates between genotypic classes. Both procedures produced similar results.

ES cells carrying a targeted Brca2 allele with loxP sites flanking exon 27 were generated by homologous recombination (Fig. 1). After transient Cre expression in vitro, ES clones with a Brca2^Δ27 allele were identified. PCR products generated with the B2F1 and B2R2 primers that flank the 5′ and 3′ loxP sites (Fig. 1) were sequenced to confirm the expected exon 27 deletion. Germ-line transmission of the Brca2^Δ27 allele was identified by standard breeding techniques.

As expected, reverse transcriptase-PCR analysis with primers specific for exon 27 does not yield detectable messages in testes RNA from Brca2^Δ27/Δ27 mice, although this reverse transcription-PCR product is easily detectable for Brca2^Δ27/+ mice. Assuming a mutant transcript is expressed, it could give rise to a truncated protein product of 3142 amino acids compared with the wild-type murine Brca2 protein that contains 3329 amino acids. Unfortunately, we are unable to confirm the presence or relative levels of this putative mutant Brca2 protein because of the unavailability of specific antibodies directed against the murine gene product.

Although the Brca2^Δ27/Δ27 animals are viable compared with previous Brca2 germ-line mutants, survival analysis indicates an overall decrease in viability of homozygous mutants compared with heterozygous and wild-type littermates:

First, cumulative genotyping at weaning from Brca2^Δ27/+ intercrosses revealed a significant (P < 0.008) deficit of Brca2^Δ27/gD27 mice from the expected 1:2:1 Mendelian ratio. When both sexes are combined, only 47 of 278 animals analyzed at weaning were of the homozygous mutant genotype. Thus, there are ∼33% fewer Brca2^Δ27/Δ27 mice than expected, and both male and female offspring were under represented. To further characterize the decreased overall viability of Brca2^Δ27/Δ27 mice during development, we examined embryonic fibroblasts from Brca2^Δ27/+ intercrosses at 13.5 days of gestation. Analysis of 39 such embryos shows that the genotypic distribution does not deviate from the expected Mendelian ratio. Thus, a subset of Brca2^Δ27/Δ27 offspring appear to either die during late gestation or shortly after birth, suggesting that loss of Brca2 function impacts overall viability.

Secondly, the overall survival of Brca2^Δ27/Δ27 compared with Brca2^Δ27/+ virgin female mice is significantly decreased (P < 0.05) by a life table test, providing additional support that lifetime survival is affected by this Brca2 mutation. Interestingly, test matings of both adult Brca2^Δ27/Δ27 males and females with wild-type mice do not reveal apparent infertility or difficulty in raising litters.

Alterations in normal growth and differentiation of mammary tissue from Brca2-mutant virgin females were determined using whole-mount analysis. At 6 months of age, mammary tissue from three Brca2^Δ27/Δ27 animals exhibited a dramatic lack of side branching and a much lower density of ductules compared with that observed for three heterozygous and wild-type littermates, which were indistinguishable (Fig. 2). Although the ducts reached the limits of the mammary fat pad in the homozygous mutant animals, a substantial lack of side branching was observed in these Brca2^Δ27/Δ27 females. These observations were confirmed by examining histological sections from these mice. This general trend of inhibited side branching observed in Brca2^Δ27/Δ27 females was maintained in the nine animals of all three genotypic classes that were examined subsequently at 9 months of age.

Brca2^Δ27/Δ27 mice display an increased incidence of a wide variety of solid tumors compared with their Brca2^Δ27/+ and Brca2^+/+ littermates (Tables 1 and 2). We observed untreated animals on a mixed 129 × C57BL/6N genetic background for tumor development or signs of morbidity. Prior to 17 months, 9 spontaneous tumors were detected in Brca2^Δ27/Δ27 mice, whereas only a single tumor was observed in the Brca2^Δ27/+ animals, and no tumors were observed in Brca2^+/+ animals during this time period (Table 2).

Terminal sacrifices of the remaining animals were performed between 17 and 19 months of age. This enabled us to develop a comprehensive tumor spectrum for mice from each genotype (Table 2). With the combined data from all of the moribund and terminal sacrifices, the Brca2^Δ27/Δ27 mice exhibited a >2-fold increase in overall tumor incidence compared with their Brca2^Δ27/+ and Brca2^+/+ littermates. In 41 Brca2^Δ27/Δ27 mice, 30 tumors were observed, compared with only 11 tumors in 43 Brca2^Δ27/+ animals and 4 tumors in 15 Brca2^+/+ animals. Overall, 25 tumor-bearing animals were observed among 41 Brca2^Δ27/Δ27 mice for a tumor incidence of 61% (Table 1). In contrast, only 11 of 43 Brca2^Δ27/+ animals were tumor bearing (26% incidence), and 3 of 15 animals (20% incidence) were tumor-bearing in the Brca2^+/+ littermates (Table 1). These data from all mice combined show a highly significant difference (P < 0.01) in the overall tumor rates between the Brca2^Δ27/Δ27 and Brca2^Δ27/+ animals. These results were also analyzed separately for virgin females alone, virgin plus multiparous females, and males only. The tumor response patterns and incidences for each of these smaller subgroups are similar to the tumor response of all animals combined. Statistically significant differences (P < 0.05) in tumor incidences between the Brca2^Δ27/Δ27 and Brca2^Δ27/+ animals are observed when the data for each subgroup is considered independently. Because a smaller number of Brca2^+/+ mice were examined, there was not enough statistical power to distinguish tumor rates between the wild-type and other genotypic classes. We also did not observe a significant difference in overall tumor incidence between males and females for the Brca2^Δ27/Δ27 genotype with a 67% (6 of 9) incidence observed in males compared with a 59% (19 of 32) incidence observed for virgin and multiparous females combined (Table 1).

The tumor spectrum is diverse and includes carcinomas, adenomas, lymphomas, and sarcomas (Table 2). Of particular interest is the exclusive appearance of various types of carcinomas in the Brca2^Δ27/Δ27 mice. These include one mammary adenosquamous carcinoma (Fig. 3), five squamous cell carcinomas (three of gastric origin), two gastric adenocarcinomas, one endometrial carcinoma, and two A/B lung carcinomas. Fig. 3 illustrates representative histological sections for the two types of gastric tumors that were observed. Interestingly, a single mammary adenomyoepithelioma was detected in one Brca2^Δ27/+ mouse. The sarcoma incidence also appears to be increased in the Brca2^Δ27/Δ27 mice compared with their littermates. Three histiocytic sarcomas, along with a hemangiosarcoma and a leiomyosarcoma, arose in Brca2^Δ27/Δ27 mice, whereas only two histiocytic sarcomas were identified in their Brca2^Δ27/+ and Brca2^+/+ littermates. We found a wide variety of tumor types in the Brca2^Δ27/Δ27 mice that were sacrificed before 17 months of age because of morbidity (Table 2). Therefore, no obvious differences in latency between distinct tumor types were observed.

Here we describe germ-line Brca2 mutant mice with a predisposition for tumor development. In contrast to previous reports of mice with homozygous germ-line mutations in the Brca2 gene (6, 7, 8, 15, 16, 17), the Brca2^Δ27/Δ27 mice exhibit only a 33% decrease in expected viability during perinatal development. Brca2^Δ27/Δ27 mice of both sexes are fertile and lack gross developmental abnormalities. Mice harboring this germ-line mutation in the COOH-terminal domain of Brca2 have an increased susceptibility for a wide spectrum of solid tumors. The fact that 9 of 10 spontaneous tumors that developed prior to 17 months arose in Brca2^Δ27/Δ27 mice indicates that not just the incidence but also the survival, because of tumor development, is significantly affected by the presence of this Brca2 mutation. Overall, the tumor spectrum we observed is similar to that reported for knockout mice eliminating the COOH-terminal region of the Brca1 gene, where an increased incidence of a wide variety of carcinomas, sarcomas, and lymphomas was found (24). One unique finding of our study is the exclusive presence of carcinomas in Brca2^Δ27/Δ27 mice and specifically a substantial number of stomach cancers, both adenocarcinomas and squamous cell carcinomas. We have shown previously that Brca2 expression in adult mouse tissues correlates with cell proliferation and is relatively high in the glandular mucosa of the stomach (25). Interestingly, stomach cancers are among the various tumor types that have been associated with BRCA2 mutations in humans, with a 2.59 relative risk (3).

This report supports a strong correlation between the Brca2 mutation position and the resulting Brca2 mouse knockout phenotype. We and other laboratories have shown previously that mice homozygous for targeted Brca2 mutations 5′ of the BRC repeats in exon 11 exhibit early embryonic lethal phenotypes (6, 7, 8, 15, 16, 17). These BRC repeats interact with Rad51 (12, 13) and have been highly conserved in evolution (18, 26), which suggests that they are important functional domains of the Brca2 protein. A few viable Brca2-mutant mice have been generated that retain at least some of these BRC repeats, and these animals display multiple severe developmental abnormalities, infertility, and early thymic lymphoma development (7, 8). The Brca2^Δ27/Δ27 mice lack only the COOH-terminal domain and could produce a truncated Brca2 gene product that preserves all eight BRC repeats. In contrast to all other Brca2-null mice reported to date, these Brca2^Δ27/Δ27 mice display a modest loss of viability and have no gross developmental abnormalities or obvious infertility. Thus, full retention of the BRC repeats and other functional domains may direct a genotype-phenotype correlation.

This study extends previous findings by several laboratories that have demonstrated the functional significance of the COOH-terminal domain of both human and rodent BRCA2. Morimatsu et al. (10) demonstrated that mouse ES cells and embryonic fibroblasts lacking exon 27 are hypersensitive to γ-radiation and undergo premature senescence. Recent reports indicate that the COOH-terminus of Brca2 is required for error-free, homology-directed repair of DNA double strand breaks and the ability to facilitate the induction of Rad51 nuclear foci after ionizing radiation (20, 21). Thus, disruption of the exon 27 Brca2-Rad51 interaction appears to lead to defective repair of DNA damage and generalized chromosomal instability with a subsequent increased risk of neoplastic progression in Brca2-deficient cells (4, 5). The human COOH-terminal region of BRCA2 also appears to be critical because a truncating mutation (9808delCC) at position 3195, which occurs just upstream of the predicted Rad51-interacting domain in human BRCA2 exon 27, is associated with an elevated breast cancer risk (27).

Despite the increased susceptibility of Brca2^Δ27/Δ27 mice to tumorigenesis, these animals are not highly susceptible to mammary tumorigenesis. Several factors may account for this observation, but the most likely reason is the fact that the Brca2^Δ27 mutation was examined on a mixed C57BL/6N and 129 genetic background. C57BL/6N and 129/SvEvS6 mice are both extremely resistant to spontaneous as well as radiation-induced mammary carcinogenesis.⁴ Thus, we are currently using microsatellite marker-assisted breeding techniques to transfer this Brca2^Δ27 mutation onto inbred strains such as BALB/cJ that are susceptible to mammary carcinogenesis.

The reduced ductal branching phenotype seen in Brca2^Δ27/Δ27 virgin female mice may be associated with increased mammary tumor risk in combination with other environmental or genetic interactions. We cannot exclude the possibility that the subtle mammary gland phenotype we have observed may be attributable to hormonal differences in the Brca2^Δ27/Δ27 mice compared with their littermates rather than a more direct Brca2 effect. However, Deng and Brodie (14) and Xu et al. (28) described a blunted ductal branching phenotype in the mammary glands of mice with a mammary-specific targeted Brca1 mutation that was associated with subsequent tumor development after a long latency. In addition, a high incidence of mammary adenocarcinomas was reported recently in mice carrying a mammary tissue-specific mutation that completely disrupts the Brca2 gene (29). Given their substantial viability and a cancer predisposition, Brca2^Δ27/Δ27 mice and cells derived from them will be useful to further define the role of Brca2 in tumorigenesis.

We thank Drs. Cynthia Afshari and Richard DiAugustine for critical review of the manuscript. The expert technical assistance of William Gersch, Brian Register, Sarah Hagevik, Christine Furmick, Laura Wharey, and David Godley is greatly appreciated. We thank Dr. John Seely for his pathology support. We also thank the staff and managers of Integrated Laboratory Systems, Inc. for their substantial animal husbandry, necropsy, and histology support.