Distinct Brca1 Mutations Differentially Reduce Hematopoietic Stem Cell Function.

BRCA1 is a well-known DNA repair pathway component and a tissue-specific tumor suppressor. However, its role in hematopoiesis is uncertain. Here, we report that a cohort of patients heterozygous for BRCA1 mutations experienced more hematopoietic toxicity from chemotherapy than those with BRCA2 mutations. To test whether this reflects a requirement for BRCA1 in hematopoiesis, we generated mice with Brca1 mutations in hematopoietic cells. Mice homozygous for a null Brca1 mutation in the embryonic hematopoietic system (Vav1-iCre;Brca1F22-24/F22-24) developed hematopoietic defects in early adulthood that included reduced hematopoietic stem cells (HSCs). Although mice homozygous for a huBRCA1 knockin allele (Brca1BRCA1/BRCA1) were normal, mice with a mutant huBRCA1/5382insC allele and a null allele (Mx1-Cre;Brca1F22-24/5382insC) had severe hematopoietic defects marked by a complete loss of hematopoietic stem and progenitor cells. Our data show that Brca1 is necessary for HSC maintenance and normal hematopoiesis and that distinct mutations lead to different degrees of hematopoietic dysfunction.


In Brief
Mgbemena et al. report that hematopoietic stem cells have an absolute requirement for Brca1 to survive. They also show that humanization of the mouse Brca1 gene with a knocked-in human BRCA1 cDNA, but not a mutant BRCA1/5382insC cDNA, fully substitutes for mouse Brca1 during both embryonic development and hematopoiesis.

INTRODUCTION
Hematopoietic stem cells (HSCs) depend on DNA repair mechanisms to maintain their genomic integrity (Beerman et al., 2014;Mohrin et al., 2010;Rossi et al., 2007). Deficiencies in DNA repair proteins impair HSC function, although the nature and severity of the defects vary among DNA repair pathways. Deficiency for proteins involved in non-homologous end joining does not affect HSC frequency or hematopoiesis in normal young adult mice, but it does reduce HSC function in response to stress (Rossi et al., 2007) and can lead to HSC depletion during aging (Nijnik et al., 2007). Deficiency for proteins involved in DNA mismatch repair does not appear to have major effects on hematopoiesis under normal conditions, but it impairs the capacity of HSCs to reconstitute irradiated mice (Reese et al., 2003). Deficiency for homologous recombination-mediated double-strand-break repair proteins, however, can lead to hematopoietic failure in patients (Kottemann and Smogorzewska, 2013), and it can impair hematopoiesis in mice as well as HSC function upon transplantation into irradiated mice (Bender et al., 2002;Carreau et al., 1999;Haneline et al., 1999;Ito et al., 2004;Navarro et al., 2006). Fanconi anemia is caused by at least 18 different autosomal recessive mutants in the FA-BRCA repair pathway, including BRCA2 (Howlett et al., 2002;Xia et al., 2007), PALB2 (Reid et al., 2007), and BRIP1 (Seal et al., 2006). All three of these proteins physically interact with BRCA1 during DNA repair (Baer and Ludwig, 2002;Prakash et al., 2015;Xia et al., 2006;Zhang et al., 2009), raising the question of whether mutations in BRCA1 also could influence HSC function or hematopoiesis. Two individuals with developmental defects consistent with Fanconi anemia were identified with genetic variants in both BRCA1 alleles (Domchek et al., 2013;Sawyer et al., 2015); however, it is not clear that these were all deleterious mutations and neither individual was reported to have hematopoietic defects. If loss-of-function mutations in BRCA1 impair DNA repair in hematopoietic cells, this would have broad implications for patients with BRCA1 mutations, as these patients are at increased risk of certain cancers that are commonly treated with DNA-damaging chemotherapies.
Homozygosity for germline loss of function in Brca1 is embryonic lethal in mice (Drost and Jonkers, 2009). Conditional deletion of Brca1 from breast epithelium in mice leads to the development of breast cancer, but only when combined with p53 deficiency (Drost and Jonkers, 2009;McCarthy et al., 2007). Two recent studies conditionally deleted Brca1 from hematopoietic cells (Santos et al., 2014;Vasanthakumar et al., 2016). One showed that leukemia cells transformed by MLL-AF9 exhibited reduced proliferation and increased differentiation in the absence of Brca1 (Santos et al., 2014). The second study showed that conditional Brca1 deletion reduced blood cell counts and colony-forming progenitors. Transplantation of Brca1-deficient bone marrow cells into irradiated mice was associated with lower blood cell counts in recipient mice and a trend toward lower levels of donor cell reconstitution 10-15 days after transplantation. However, this study did not detect a significant reduction in HSC frequency, and the consequences for the long-term reconstituting capacity of bone marrow cells was not assessed (Vasanthakumar et al., 2016). Therefore, it has not yet been tested whether Brca1 is required for HSC function or whether heterozygosity for BRCA1 mutations affects recovery after chemotherapy in humans or in mice.
We evaluated the hematologic effects of chemotherapy on cancer patients with germline BRCA1 or BRCA2 mutations, and we found that, in our small cohort, BRCA1 mutations were associated with an increased risk of hematopoietic toxicity. Based on these clinical observations, we tested the effects of BRCA1 mutations on hematopoiesis in mice. To do this, we characterized the effects of two different mutant Brca1 alleles on mouse hematopoiesis. We show that Brca1 is necessary for HSC maintenance and normal hematopoiesis but that different alleles exhibit differences in the severity of the HSC phenotype that do not correlate with differences in the severity of their effects on embryonic development.

RESULTS
Association of BRCA1 Mutations with Hematopoietic Toxicity from Chemotherapy Patients with germline BRCA1 mutations commonly develop cancers that are treated with DNA-damaging chemotherapies. To test whether those patients are at increased risk for hematopoietic complications, we analyzed hematopoietic parameters in patients heterozygous for deleterious mutations in BRCA1 at baseline (healthy patients) and after chemotherapy. We compared our patients to those who carry deleterious mutations in BRCA2 for two reasons. First, it allowed for a comparison cohort that was similar in gender and age (Table 1), and second because prior data suggested that BRCA2 mutation carriers may experience fewer episodes of neutropenia compared to BRCA1 mutation carriers (Shanley et al., 2006). In the latter report, it was not possible to match for type of chemotherapy regimen and most of the patients did not receive doxorubicin, a standard component of current breast cancer treatment. In our cohorts, most of the patients with BRCA1 or BRCA2 mutations that received chemotherapy had breast cancer, and they were treated with four cycles of dose-dense doxorubicin plus cyclophosphamide (both drugs are DNA damaging) followed by four cycles of paclitaxel.
BRCA1 and BRCA2 mutation carriers at our institution had normal blood cell counts at steady state ( Figures 1A-1F). Both sets of patients experienced hematopoietic toxicity after chemotherapy, though BRCA1 mutation carriers tended to experience more frequent and severe hematopoietic toxicity from chemotherapy than BRCA2 mutation carriers ( Figures 1G-1K). This was evident only in a subset of the patients, as shown by the wide variations in the post-chemotherapy blood counts (Figures 1H-1K; blood counts from each patient at the time of maximum neutrophil toxicity). Severity of maximal toxicity for blood parameters also was quantified as recommended by the National Cancer Institute (NCI)'s criteria for adverse events (Figures 1L-1O; Table 2). Since prophylactic granulocyte-colony stimulating factor (G-CSF) is standard for patients receiving dose-dense doxorubicin plus cyclophosphamide and paclitaxel therapy, the overall rate of febrile neutropenia (neutropenia with fever) in breast cancer patients at our institution during the last 3 years has averaged $5%. This is consistent with a recently reported 3.4% overall febrile neutropenia rate for this regimen in the United States (Caggiano et al., 2005). In contrast, 34% of BRCA1 mutation carriers experienced febrile neutropenia, a significantly higher frequency than observed among BRCA2 mutation carriers (0%; p < 0.0001; Figure 1G). The BRCA1 and BRCA2 mutation carriers were relatively young patients (median age 41.5 and 46, respectively; Table 1), without co-morbidities and in whom febrile neutropenia would not be expected (Kouroukis et al., 2008). Consistent with the increased incidence of febrile neutropenia, we observed a trend toward an increased incidence of neutropenia among BRCA1 mutation carriers as compared to BRCA2 mutation carriers ( Figure 1I; p < 0.1). BRCA1 mutation carriers also had a significantly higher frequency of grade 3/4 leukopenia (28% for BRCA1 versus 8% for BRCA2; Figure 1L; p < 0.05) and grade 3/4 lymphopenia (31% for BRCA1 versus 0% for BRCA2; Figure 1N; p < 0.05)  compared to BRCA2 mutation carriers. There was also a trend toward an increased incidence of grade 3/4 anemia (31% for BRCA1 versus 0% for BRCA2; Figure 1O) after chemotherapy. There were no differences in platelet counts between the two cohorts (data not shown).
There were no correlations of specific BRCA1 mutations with febrile neutropenia. The small numbers of individuals with each mutation precluded our ability to determine whether or not specific mutations in the BRCA1 gene were associated with toxicity (Table S1). Nevertheless, these data suggest that, collectively, germline BRCA1 mutations are associated with a higher-than-expected risk for chemotherapy-associated hematopoietic complications.
Vav1-iCre;Brca1 F22-24/F22-24 mice developed severe pancytopenia. The 3-to 6-week-old Vav1-iCre;Brca1 F22-24/F22-24 mice had significantly decreased absolute numbers of white blood cells, including neutrophils and lymphocytes, as well as significantly reduced numbers of platelets compared to controls . Deletion of a single allele of Brca1 was sufficient to slightly but significantly reduce white blood cell (WBC) and lymphocyte levels ( Figures 2B and 2D).
Vav1-iCre;Brca1 F22-24/F22-24 mice had a shortened lifespan and most died spontaneously without appearing ill, likely as a result of hematopoietic failure and its consequences (acute infection, bleeding, etc.). Half of the Vav1-iCre;Brca1 F22-24/F22-24 mice died by 75 days ( Figure 2H). A fraction (27%, 7/26) of the Vav1-iCre;Brca1 F22-24/F22-24 mice that survived beyond 3 months of age did become moribund (hunched, immobile, and cold) prior to death, and they developed lymphocyte-infiltrated splenomegaly ( Figure S2A). Unlike the prior report that found p53 mutations in some spleens that were Brca1 null (Vasanthakumar et al., 2016), when we used RNA sequencing (RNA-seq) to analyze for mutations in expressed genes, only wild-type p53 was found in these enlarged spleens. However, consistent with the same report, T cell infiltration was present based on a decreased B cell-specific gene expression and an increased T cell gene expression pattern in the enlarged spleens ( Figure S2B). These data suggest that deletion of Brca1 from mouse hematopoietic cells has the potential to promote the development of hematopoietic malignancies from surviving progenitors (Vasanthakumar et al., 2016).

Heterozygosity for Brca1 Reduces HSC Reconstituting Capacity
To begin to determine if the bone marrow of Vav1-iCre; Brca1 F22-24/+ mice has increased sensitivity to DNA stress, we treated a cohort of Vav1-iCre;Brca1 F22-24/+ mice with two cycles of cyclophosphamide, and we monitored their blood counts for recovery abnormalities. Under these specific conditions, we did not observe consistent differences between heterozygous and wild-type mice ( Figures S3R-S3W). To observe more subtle differences, it may be necessary to either give different doses of cyclophosphamide, treat with several more drug cycles, or treat with other drugs used in our patients, such as cisplatin, doxorubicin, and paclitaxel.
To further test if heterozygosity impaired HSC self-renewal potential, we serially transplanted bone marrow cells from the primary recipient mice into secondary recipient mice; 16 weeks after primary transplantation, we transplanted bone marrow cells from primary recipient mice with levels of donor cell reconstitution (CD45.2+) nearest the median values in each treatment. Vav1-iCre;Brca1 F22-24/+ cells gave significantly lower levels of donor cell reconstitution in all lineages compared to Brca1 +/+ control cells in secondary recipients ( Figures S3N-S3Q). These results suggest that Vav1-iCre;Brca1 F22-24/+ HSCs do in fact exhibit a reduced self-renewal capacity as compared to wildtype HSCs. Although single Vav1-iCre transgenic mice were not used as controls in this cohort, complete blood count (CBC) abnormalities or defects in primary or secondary reconstitution by bone marrow cells from Vav1-iCre mice in the same pure C57BL/6 background have not been observed previously (Foley et al., 2013). These observations in mice are consistent with greater chemotoxicity in humans with BRCA1 mutations, and they suggest that heterozygosity for a loss-of-function mutation in Brca1 can impair the ability to regenerate hematopoiesis after one or more cycles of chemotherapy ( Figure 1G).
Because serial bone marrow transplantation was required to observe a deleterious effect of proliferative stress on Brca1 haploinsufficient HSCs, it remains possible that the heterozygous genotype does not impose as much chemotherapeutic toxicity as our human data would suggest. Since our patient data are from a retrospective analysis of a small cohort ( Figure 1G), further work to prospectively observe more humans with inherited cancer predisposition mutations who are treated with chemotherapy is necessary. Also, generation and serial treatment of more Brca1 haploinsufficient mice with various bone marrow stresses that include repeated cycles of cyclophosphamide, doxorubicin, cisplatin, and paclitaxel will need to be completed.

Generation of Wild-Type and Mutant Knockin Alleles of Human BRCA1
We and others have found that frameshift or stop-gain mutations in the last few exons of the BRCA1 gene encode non-functional mutant proteins (Scully et al., 1999) expressed from messages that do not experience RNA decay (Perrin-Vidoz et al., 2002;Soyombo et al., 2013). These C-terminal mutations frequently lead to cancer phenotypes distinct from those caused by mutations elsewhere in the gene (Rebbeck et al., 2015). We have found that the common Ashkenazi Jewish BRCA1 5382insC founder mutation results in expression of a mutant transcript in the same amounts as BRCA1 wild-type message in fibroblasts, induced pluripotent stem cells, and teratomas . The 5382insC mutation leads to a C-terminal frameshift mutation. We wondered whether this hypomorphic allele would have a less severe hematopoietic phenotype as compared to the null mutation.
To assess the 5382insC mutation's effects on embryogenesis and hematopoiesis and to ensure that any abnormalities observed with humanization of the Brca1 locus with the BRCA1 5382insC allele were due to the abnormal BRCA1, we generated mice that were humanized with wild-type human BRCA1 in place of mouse Brca1. To do this, we designed a targeting vector that allowed for expression of both a wildtype BRCA1 allele and, upon Cre-mediated recombination, the BRCA1 5382insC mutation (Figures 4A and S4A). C57BL/6 embryonic stem cells (ESCs) were electroporated with the targeting construct and screened for correctly targeted knockin alleles ( Figure S4B). Two lines that were correctly targeted and expressed human BRCA1 ( Figure S4C, lanes 3-6) also were electroporated with CMV-Cre to generate ESC lines that expressed the recombined Brca1 5382insC mutant allele ( Figure S4C, lanes  7 and 8). The Brca1 BRCA1 allele substituted for wild-type mouse Brca1 function, as evidenced by the fact that fully humanized homozygotes (Brca1 BRCA1/BRCA1 ) were born at Mendelian frequencies ( Figure 4B, left), and they remained alive and well with no hematopoietic abnormalities ( Figures 4D-4P) for up to 1.5 years of age ( Figure S4D).
To generate mice with the BRCA1 5382inC allele (Brca1 5382insC ) in the germline, Brca1 BRCA1/BRCA1 mice were mated with CMV-Cre deletor mice (Dupé et al., 1997). The progeny with recombination in the germline were then used for further analysis of mice who carried this human mutation. In contrast to the Brca1 BRCA1 allele, homozygosity for the Brca1 5382insC allele was embryonic lethal, as 208 progeny from the heterozygous Brca1 5382insC/+ parents included no Brca1 5382insC/5382insC homozygotes and an expected frequency of heterozygotes and wildtype mice ( Figure 4B, right; p < 0.0001). This lethality confirms that the 5382insC allele encodes a severe loss-of-function mutation. These data thus indicate that the wild-type human BRCA1 cDNA rescues mouse embryonic lethality (suggesting that alternative splicing is not necessary for this gene to function properly in mice) and that the BRCA1 5382insC mutant does not.

Hematopoiesis in Mice with the Brca1 5382insC Mutation
To assess the effects of the germline Brca1 5382insC mutant allele on hematopoiesis, we crossed the Brca1 5382insC/+ mice with the Vav1-iCre;Brca1 F22-24/+ mice to generate Vav1-iCre; Brca1 F22-24/5382insC biallelic mutant mice. This is a similar genetic configuration predicted to occur in many human cancers-a germline mutation (Brca1 5382insC ) in one allele followed by somatic loss of heterozygosity as a result of deletion of the second allele. The biallelic Vav1-iCre;Brca1 F22-24/5382insC mice were healthy at weaning. In contrast to the severe hematopoietic defects in Vav1-iCre;Brca1 F22-24/F22-24 mice, average peripheral blood counts (Figures S5A-S5F) and bone marrow stem and progenitor cell frequencies were normal in adult Vav1-iCre; Brca1 F22-24/5382insC mice ( Figures S5H-S5P). However, the normal blood counts may be attributed to the main presence of cells that lacked recombination of the floxed null allele ( Figure S5G). In contrast to the presence of only non-recombined hematopoietic cells in Vav1-iCre;Brca1 F22-24/5382insC mice (same amplification curve as obtained from DNA derived from Crenegative Brca1 F22-24/+ control bone marrow), hematopoietic cells from Vav1-iCre;Brca1 F22-24/F22-24 mice exhibited significant recombination ( Figure S5G). A possible explanation for the lack of somatic recombination of the Brca1 F22-24 allele when the germline allele was Brca1 5382insC is that the BRCA1 5382insC protein is more deleterious to hematopoietic cells than the simple null allele, and, thus, the only cells that survived into adulthood were those that were not somatically recombined.
The severe anemia in the Mx1-Cre;Brca1 F22-24/5382insC mice was the largest and most significant difference from the Mx1-Cre;Brca1 F22-24/D mice ( Figure 5G). Since the half-life of mouse red blood cells in wild-type mice has been measured at more than 20 days (Van Putten, 1958), it is possible there was bleeding secondary to the severe thrombocytopenia. In fact, in two mice that were necropsied immediately after death, we observed large pools of blood in the abdominal and thoracic cavities. Thrombocytopenia can occur quickly with the loss of progenitors, as mouse platelet half-life has been estimated to be 3.4 days (Jayachandran et al., 2010).

DISCUSSION
In this article, we show a cohort of patients with BRCA1 mutations that experienced increased hematopoietic toxicity and complications after cancer chemotherapy. We also observed that Brca1 is required for HSC function and normal hematopoiesis in mice. When Brca1 was conditionally deleted from embryonic hematopoietic cells, young adult mice developed pancytopenia ( Figure 2) and a loss of nearly all HSCs (Figure 3). Moreover, heterozygosity for a loss-of-function allele of Brca1 in mouse hematopoietic cells led to a slight but significant decrease in white blood cells and lymphocytes (Figure 2), as well as deficits in HSC reconstituting potential upon serial transplantation ( Figure S3).
These results are consistent with a reduced hematopoietic regenerative capacity in BRCA1 heterozygous humans after chemotherapy, suggesting that even a partial loss of BRCA1 function reduces the capacity for hematopoietic recovery due to direct DNA damage or replication stress after myeloablation. The concept that replication stress leads to more chemotherapeutic toxicity for BRCA1 mutation carriers is also consistent with prior work that has suggested there is enhanced replication stress (due to decreased stalled fork repair) in BRCA1 heterozygous epithelial cells. This abnormality in heterozygous cells was hypothesized to enhance the formation of tumors in epithelial cells (Pathania et al., 2014).
These data suggest a cell death or transformation tissue specificity hypothesis, and they could explain why patients with germline BRCA1 mutations have a predisposition to epithelial cancers but do not have a predisposition to hematological malignancies. The ultimate loss of BRCA1 heterozygosity, which is promoted by diminished DNA repair in the heterozygous state and is thought to be required for the transformation of epithelial cells to cancer (Pathania et al., 2014), is not tolerated by hematopoietic stem cells. This is the first report of generation and characterization of a humanized Brca1 allele. The human BRCA1 cDNA was knocked into the mouse Brca1 locus to study its function. Humanization of mouse genes has proven useful for in vivo functional evaluation of human p53 mutations (Song et al., 2007). Like p53, the introduction of human mutations into the mouse Brca1 allele is advantageous, as there are significant differences in amino acid sequence between mouse Brca1 and human BRCA1. The mouse protein is only 60% identical to the human BRCA1 protein (Sharan et al., 1995). Our finding of embryonic lethality for the Brca1 5382insC/5382insC genotype, but not in un-recombined Brca1 BRCA1/BRCA1 mice (Figure 4), confirms that human BRCA1 can perform many of the necessary functions of mouse Brca1 after being knocked into the mouse Brca1 locus.
Conditional deficiency for Brca1 using Mx1-Cre in adult mice previously has been reported to increase differentiation in MLL-AF9-induced leukemia (Santos et al., 2014), diminish hematopoietic cell proliferation in vitro, and lead to mild leukopenia and anemia (Vasanthakumar et al., 2016). However, neither of these studies reported a deleterious effect of Brca1 deficiency on HSC frequency or function. Use of the Vav1-iCre allele to delete Brca1 in embryonic and adult HSCs and use of the new human Brca1 5382insC allele were not part of the prior studies. The use of different Cre alleles suggests that deletion of Brca1 in the embryonic HSCs (Vav1-iCre) may be more deleterious than deletion in adult HSCs (Mx1-Cre).
A trivial explanation for why an HSC defect was only observed in mice using the Vav1-iCre allele for conditional deletion of Brca1 is that there was more recombination in HSCs with the Vav1-iCre allele compared to those with the Mx1-Cre allele. Although this explanation is not possible to confirm or refute without analysis of the original mice, because Vasanthakumar et al. (2016) did provide evidence for full recombination in hematopoietic cells from their Mx1-Cre transgenic mice, it is an unlikely explanation. Different BRCA1 mutations have been shown to have distinct effects on cancer phenotypes. Humans with mutations at the extreme C and N termini of BRCA1 experience more breast cancer and less ovarian cancer compared to humans with mutations in the middle of the BRCA1 gene (Rebbeck et al., 2015). Our surprising observation that the Brca1 5382insC mutation led to a more severe adult hematopoietic phenotype than the Brca1null mutation ( Figure 5) suggests that distinct germline BRCA1 mutations may result in different degrees of chemotherapeutic toxicity as well. More patients with these mutations are needed to make associations, and more investigation into the mechanism of this increased toxicity of the Brca1 5382insC allele in the mice will be important.
Because the Brca1 5382insC allele, in contrast to the null allele, expresses a mutant protein ( Figure S4C versus Figure 2A), it could indeed have hypo-or hypermorphic effects on cells. Expression of the BRCA1 5382insC mutant mRNA from humans heterozygous for the BRCA1 5382insC mutation is equivalent to the expression of the wild-type mRNA from the other allele in primary human fibroblasts, induced pluripotent stem cells, and teratomas. However, the expression of BRCA1 5382insC in these cells does not promote excessive cell death, differentiation, survival, or growth . Further, heterologous expression of this mutant BRCA1 in cell lines does not . Statistical significance was assessed using a two-tailed Student's t test except in (A) where a log-rank test was used (*p < 0.05, **p < 0.01, and ***p < 0.001). lead to altered growth or survival (data not shown). These observations may be due to the expression from the normal BRCA1 allele. The mutant protein may only be detrimental in a completely deficient BRCA1 background. Further studies to understand why the BRCA1 5382insC allele leads to an in vivo phenotype distinct from the Brca1-null allele are necessary.
Stem cells are susceptible to DNA damage due to their longevity and self-renewal potential. HSCs from mice with mutations in DNA damage repair proteins that also lead to cancer susceptibility syndromes, such as Brca2 (Navarro et al., 2006) and Msh2 (Reese et al., 2003), have defects in their ability to reconstitute bone marrow in irradiated mice, and mice with mutant Rad50 exhibit hematopoietic failure (Bender et al., 2002). However, the hematopoietic phenotype we observed after Brca1 deletion is much more severe than the phenotypes reported in these studies.
Several mouse models have been generated to study BRCA1mutant breast cancer (Dine and Deng, 2013;Drost et al., 2011;Drost and Jonkers, 2009;Evers and Jonkers, 2006;Shakya et al., 2011). These models confirm that Brca1 maintains genome stability in vivo and that, without normal Brca1 in breast epithelial tissues, breast tumorigenesis occurs. However, breast cancer develops in Brca1-knockout mice only after a long latency (even if p53 is also deficient). This is consistent with the fact that human BRCA1 mutation carriers are only diagnosed with cancer as adults, if ever.
Here we describe mice with different Brca1 alleles mutated specifically in the hematopoietic system that have distinct phenotypes, which, in contrast to the breast cancer phenotype, occur rapidly and are fully penetrant (for allele/phenotype summary, see Table S2). In addition to the new information about the role of Brca1 in hematopoiesis, these allele combinations provide the field with powerful tools for rapid investigation of the pathogenicity of BRCA1 variants of unknown significance.
Finally, given the potent requirement for Brca1 in HSCs, an inherited BRCA1 mutation may be a marker to add to the list of patient risk factors, such as age and co-morbidities (Caggiano et al., 2005), that support the prophylactic use of growth factors and antibiotics and close monitoring for chemotherapy-related hematopoietic complications. Preventative use of myeloid growth factor support may, however, be counterproductive if unrepaired replication-induced mutations are increased in BRCA1 heterozygotes. Prophylactic growth factor support and antibiotics should, therefore, be evaluated prospectively in BRCA1 mutation carriers who are receiving chemotherapy.

EXPERIMENTAL PROCEDURES Patients
A list of patients with BRCA1 or BRCA2 mutations, treated between January 1, 2011, and October 31, 2014, was identified from the University of Texas Southwestern Medical Center's Cancer Genetics database. Patients were categorized based on cancer type and chemotherapy treatments. A retrospective chart review was then conducted on these patients to collect information on patient characteristics (Table 1), as well as co-morbidities and past medical/ surgical histories, type of cancer, age of diagnosis, treatment, treatment complications (if applicable), and complete blood cell counts. For any patient who had at least one complete blood cell count recorded in their medical record, baseline complete blood cell count values were selected for each patient based on the following criteria: pre-treatment (but as close to beginning of therapy as possible within 5 years of cancer diagnosis), no active infection, no procedural context (e.g., post-biopsy or post-operative), and did not appear to be an outlier if other complete blood cell counts were available for comparison.
For the patients who underwent chemotherapy for their cancer, the most severe adverse hematopoietic event during chemotherapy and its associated toxicity score were recorded. Grades of blood cell count toxicity were assigned based on the National Cancer Institute Common Terminology Criteria for Adverse Events v3.0 guidelines (Trotti et al., 2003) (Table 2). Neutropenic fever was defined as an absolute neutrophil count <500 cells/mm 3 and fever. Fever was defined as a single oral temperature of >38.3 C (101 F) or a temperature of >38.0 C (100.4 F) sustained for more than 1 hr.
Following collection of these data, statistical analyses were conducted on de-identified data. Range, mean, and SDs of complete blood cell components (neutrophils, platelets, hemoglobin, etc.) in BRCA1/2 mutation carriers were compared to the normal ranges. Analyses of variations in complete blood cell components in response to different chemotherapy regimens also were evaluated for differences between the BRCA1 and BRCA2 mutant patients. This study (STU 072014-043; Analysis of Complete Blood Counts in BRCA Mutation Carriers) was approved by the University of Texas Southwestern Medical Center Institutional Review Board.
The BRCA1 knockin mice were generated as described in detail in the Supplemental Experimental Procedures and Figure S4. Briefly, the targeting vector was constructed to generate a knockin allele that conditionally generated the BRCA1 5382insC mutation as well as constitutively humanized BRCA1 mice ( Figures 4A  and S4). These strains are available through the Jackson Laboratory Repository (JAX Stock No. 030081, Humanized BRCA1 KI -or-[Brca1 huBKI/huBKI ] and JAX Stock No. 030082, Humanized BRCA1 5382insC KI -or-[Brca1 5382insC ]).
All Brca1-mutant mice were genotyped from tail snips using real-time PCR assays designed by and available from Transnetyx. The assays were designed to detect the wild-type and mutant alleles in the presence or absence of recombination. Mice were housed in the Unit for Laboratory Animal Medicine at the University of Texas Southwestern Medical Center under specific pathogen-free conditions, and they were monitored regularly for evidence of disease and abnormal peripheral blood cell counts. The animal use protocol was approved by the University of Texas Southwestern Institutional Animal Care and Use Committee (APN 2011-0143).

Bone Marrow Transplantation
Adult recipient mice (CD45.1) were administered a minimum lethal dose of radiation using an XRAD 320 X-ray irradiator (Precision X-Ray) to deliver two doses of $540 rad (1,080 rad in total) at least 3 hr apart. Cells were injected into the retro-orbital venous sinus of anesthetized recipients. For competitive bone marrow transplants, 5 3 10 5 donor and 5 3 10 5 recipient cells were transplanted. Blood was obtained from the submandibular plexus of recipient mice at the indicated time points after transplantation. Red blood cells were lysed with ammonium chloride potassium buffer. The remaining cells were stained with antibodies (Tonbo Biosciences) against CD45.2, CD45.1, CD45R (B220), CD11b, CD3, and Gr-1 to assess donor cell engraftment. Mice that died were omitted from the analyses.

Hematopoietic Analysis
Bone marrow cells were isolated by flushing the long bones (femurs and tibias) in Ca 2+ -and Mg 2+ -free Hank's buffered salt solution (Corning Life Sciences) supplemented with 3% heat-inactivated bovine serum (Gibco). Spleens were prepared by crushing tissues between frosted slides. Cell number and viability were assessed by a Vi-CELL cell viability analyzer (Beckman Coulter) or by counting on a hemocytometer.
Flow cytometric analysis of specific hematopoietic progenitors was performed as previously described (Foley et al., 2013;Signer et al., 2014). Complete blood cell count analysis was performed on peripheral blood using the Hemavet 950 with MULTI-TROL Mouse as an equilibration control (Drew Scientific).

Statistical Analysis
Statistical significance was assessed using a two-tailed Student's t test with p values (*p < 0.05, **p < 0.01, and ***p < 0.001). A Fisher's exact test (*p < 0.05) was used to asses statistical significance in Figures 1G and 1L-1O. For Kaplan-Meier curves depicting survival analyses, a log-rank test was used. All statistical analyses were performed using GraphPad Prism version 7.00 for Windows. All RNA-seq expression data and accession codes can be found at GEO: GSE91390.

ACCESSION NUMBERS
The accession number for the RNA-sequencing data reported in this paper is GEO: GSE91390.

SUPPLEMENTAL INFORMATION
Supplemental Information includes Supplemental Experimental Procedures, five figures, and two tables and can be found with this article online at http:// dx.doi.org/10.1016/j.celrep.2016.12.075.

AUTHOR CONTRIBUTIONS
V.E.M. and R.A.J.S. designed experiments, collected data, interpreted results, and edited the manuscript. R.W. and T.L. collected data, interpreted results, and edited the manuscript. S.J.M. interpreted data and edited the manuscript. T.S.R. designed experiments, interpreted data, and wrote the manuscript.