Genetic Variants in Immune-Related Pathways and Breast Cancer Risk in African American Women in the AMBER Consortium

Endemic infectious diseases fundamentally shape the immune systems of populations at the genetic level, selecting for immunity that preserves well-being through the reproductive years (1). The immune system is organized into two central arms of defense: the nonspecific innate arm that provides protection against a wide variety of pathogens by activating the inflammatory response, and the more specific adaptive arm that targets specific antigens of pathogens (2). Individuals of African ancestry (AA) show inherited differences in constitutional immune function from European Americans (EA), partly due to evolutionary responses to infection and endemic helminth exposure in Sub-Saharan Africa, causing both a stronger inflammatory immune response as well as a type II immunosuppressive phenotype (1, 3).

As reviewed by Palmer and colleagues (4), AA women are more likely than EAs to be diagnosed with breast cancer before age 50, and to have more aggressive disease of higher grade, lacking receptors for estrogen (ER), progesterone (PR), and HER2, which precludes use of targeted therapies, often resulting in poorer survival. Although it is becoming clearer that some reproductive and hormonal factors may, in part, account for some of the disparities in breast cancer aggressiveness (5, 6), it is possible that constitutional immune phenotype could play a role in breast cancer etiology, particularly for ER⁻ disease (7). In the AMBER Consortium, we hypothesized that the higher prevalence of ER⁻ tumors in AAs could arise, in part, from differences in immunologic profiles such as those that typically define inflammation and/or suppressive phenotypes. Thus, we examined genetic variation in immune response pathways involved in host defense and/or self-tolerance in relation to overall breast cancer risk and risk of ER⁻ compared with ER⁺ cancers. Because ER⁻ breast cancers are most prevalent among premenopausal AAs, and immune biomarkers were previously observed to be more strongly associated with being diagnosed with ER⁻ compared with ER⁺ breast cancer among premenopausal women (8), we also examined the role of genetic variation in immune pathways and breast cancer risk by menopausal status.

These analyses were conducted in the context of the African American Breast Cancer Epidemiology and Risk (AMBER) Consortium designed to investigate the etiology of breast cancer subtypes among AA women (4). Studies forming the basis for AMBER were the Women's Circle of Health Study (WCHS), the Carolina Breast Cancer Study (CBCS), the Black Women's Health Study (BWHS), and the Multi-Ethnic Cohort Study (MEC; 9–13), which are described in Supplementary Methods S1. Research protocols were conducted in accordance with the ethical standards of the Declaration of Helsinki, and were approved by the Institutional Review Board at each participating institution. Informed consent was provided by all study participants. Table 1 shows the total number of cases and controls included in the analysis by age, menopausal status, and ER status.

A total of 433 genes in 45 curated pathways involved in immune response were selected from the Molecular Signatures Database (MSigDB) and included BIOCARTA, KEGG, REACTOME, and Gene Ontology (GO) pathways (14). Pathways considered for inclusion were those related to cytokines, Th1- and Th2-related immunity, inflammation, T-cell activation, and immune response regulation. Priority was given to those containing the highest proportion of genes identified in GWAS (full catalog downloaded from the National Genome Research Institute on March 8, 2012) found to be associated with breast cancer, or disease conditions and/or phenotypes associated with dysregulated immune function, hormonal or reproductive factors, obesity-related traits, or circulating vitamin D levels as risk factors for breast cancer. A full list of traits used to generate the list of GWAS genes used to help prioritize immune pathways is provided in Supplementary Table S1. The list of GWAS genes used for prioritization is provided in Supplementary Table S2, with N = 74 found within the 45 immune pathways prioritized. Tag single-nucleotide polymorphisms (SNP) were selected for the 433 unique genes ± 10 kb to capture SNPs with minor allele frequency ≥ 10% (at r² ≥ 0.8) based on the haplotype structure of the Yoruban population (YRI) in 1000 Genomes Phase I reference panel (http://www.1000genomes.org).

Genotyping was performed by the Center for Inherited Disease (CIDR) using the Illumina Human Exome Beadchip v1.1 with additional custom content SNPs, as described previously (15). A total of 13,235 SNPs in the 433 immune-related genes examined passed quality control (QC). Imputation was performed using IMPUTE2 software (16) and the 1000 Genomes Phase I reference panel (May 21, 2011 1000G data, December 2013 haplotype release in IMPUTE2 site). Genetic data from 533 cases and 989 controls were available in the MEC, from a previous GWAS on the Illumina Human 1 M-Duo chip, with SNPs imputed to the same release of 1000 Genomes. Imputed genotypes were combined into a final dataset after omitting variants with mismatching alleles in AMBER and MEC, allele frequencies that were different by more than 0.15, and had imputation INFO scores < 0.5 in either study. A total of 149,514 SNPs were included in these analyses. Genotype principal components were computed using the smartpca program in the EIGENSOFT package to examine population structure.

The Search Tool for the Retrieval of Interacting Genes (STRING, V.10) database was used to construct a protein–protein interaction network to identify potentially relevant immune pathways. The STRING database provides integrated knowledge on the known and predicted functional relationships between proteins (17). Only associations with high confidence (score ≥ 0.7) are presented. Networks were generated for significant genes identified for breast cancer overall, and by ER and menopausal status.

As described previously (15), gene-based pathway analyses were conducted using the adaptive rank truncated product (ARTP) statistic as implemented in the R package PIGE, which allowed us to optimize the number of SNP P values combined in each pathway and in each gene-level test. To avoid capturing highly correlated signals within a gene, a subset of “pruned-in” SNPs were selected to ensure that all SNP pairs had linkage disequilibrium (LD) r² < 0.8, using the R2 filter option in the R package AdaJoint. The ARTP gene-level tests combined the optimal number of the most significant SNP P values from among the top 10 pruned-in SNPs for each gene. Single variant tests were only conducted within genes reaching a nominal significance level of P = 0.05. Unconditional logistic regression analysis was performed using imputed dosage genotype data. A Bonferroni correction was applied for the number of pruned-in SNPs tested within each gene to identify SNPs with gene-wide significance. We performed case–case analyses to compare odds of being diagnosed with ER⁻ to ER⁺ disease to identify pathways, genes, and SNPs that may differentially impact ER status. Although none of the genotype principal components tested were strongly associated with risk, we included three principal components that were associated at P < 0.10 in all regression models to control for any potential confounding by population structure. Models were also adjusted for study, age, geographic location, and DNA source (saliva, blood, mouthwash). Lifestyle factors and comorbid conditions that affect immune function and breast cancer risk were not treated as potential confounders in our study because they could be on the causal pathway between genetic variation and breast cancer risk, and adjustments may inappropriately attenuate risk estimates. Functional follow-up was performed in RegulomeDB, PolyPhen-2, HaploReg v4.1, and GTEx databases (18–21).

Pathway analyses yielded nominally or borderline significant associations for overall breast cancer with cytokine-related pathways (IL receptor activity, P = 0.05; IL binding, P = 0.01; cytokine and chemokine mediated signaling, P = 0.04) as well as the immune effector process pathway (P = 0.04; Supplementary Table S3). When stratifying by ER status, only the Biocarta TNFR2 pathway achieved nominal significance for ER⁺ cancer (P = 0.005), and the positive regulation of cytokine production pathway was associated with ER⁻ disease (P = 0.024).

When considering menopausal status, pathways most significant for premenopausal breast cancer (Supplementary Table S4) were those associated with regulation of immune system processes (positive regulation of immune system process, P = 0.005; regulation of immune system process, P = 0.01; immune effector process, P = 0.02) and lymphocyte and T-cell activation (regulation of lymphocyte activation, P = 0.03; regulation of T-cell activation, P = 0.06). Premenopausal ER⁺ cancers were most strongly related to programmed death-1 (PD-1) signaling (P = 0.04), while ER⁻ cancers were associated with the adaptive immune response pathway (P = 0.03), and cytokine and inflammation-related pathways (Biocarta cytokine pathway, P = 0.004; Biocarta inflammation pathway, P = 0.03). Immune activation pathways were also important among postmenopausal women (Supplementary Table S5; regulation of lymphocyte activation, P = 0.06; T-cell activation, P = 0.06), but cytokine-related pathways appeared to play a more prominent role (IL receptor activity, P = 0.01; IL binding, P = 0.02; cytokine and chemokine-mediated signaling pathway, P = 0.03), involved in both ER⁺ (cytokine binding, P = 0.03; Biocarta IL5 pathway, P = 0.05), and ER⁻ cancers (cytokine and chemokine mediated signaling pathway, P = 0.04).

In gene-based testing, a number of genes were nominally related to overall breast cancer risk, and by ER status at P < 0.01 (Table 2; Supplementary Table S3). Gene functions are summarized in Supplementary Table S6. No genes remained significant after correcting for the overall number of genes tested. For overall breast cancer, the most notable genes were those involved in cytokine-related pathways, which included IL2RB (P = 0.001), CADM1 (P = 0.003), ATP6AP2 (P = 0.006), and TLR6 (P = 0.006). Subsequent SNP testing of these genes showed five variants significant at the gene-wide level (Table 3). The top variant was rs228952 in IL2RB (P_adj = 0.007). In SNP-based analyses, several intronic variants in POU2AF1 were inversely associated with risk, including rs145624147 (P_adj = 0.008).

Among premenopausal women (Table 2; Supplementary Table S4), genes most strongly related to risk were IL21, SLA2, and CFHR1 (P ≤ 0.004), involved in regulation of immune system processes. An intronic SNP in SLA2 (rs221310; P_adj = 0.003) was associated with approximately 30% increased risk for each copy of the G allele. Two intergenic SNPs associated with IL21, rs115698762 (P_adj = 0.008), and rs143266239 (P_adj = 0.01), increased risk by approximately two-fold. Among postmenopausal women, genes associated with overall risk included IL2RB, CYP4F11, and POUF2AF1 (P ≤ 0.007), as well as IL21 (P = 0.004). Two genes, CASP8 (P = 0.006) and IRAK2 (P = 0.008) were specific to overall breast cancer risk in postmenopausal women. When SNPs were examined, several in genes involved in interleukin and inflammatory response pathways, that is, IL2RB, CYP4F11, IRAK2, and IL21 were associated with risk. Of these, the most significant were in IL21 (rs2390350, P_adj = 0.008; rs17005895, P_adj = 0.01).

None of the genes associated with overall breast cancer risk were significantly associated with ER⁺ cancers at P ≤ 0.01. Only BCL3, a putative proto-oncogene important for the development, survival, and activity of adaptive immune cells (22) contained a variant (rs34698726) associated with overall breast cancer risk (P_adj = 0.01) and the risk of ER⁺ cancers (P_adj = 0.01). Except for CADM1 (P = 0.006), none of the genes associated with overall breast cancer risk were associated with ER⁻ cancers. Instead, genes involved in TNFR1 (PRKDC, P = 0.002) and inflammatory response pathways (IL8, P = 0.003; NFATC3, P = 0.008), regulation of cytokine production (CARD11, P = 0.007), and MAPK signaling pathways (MAPK3, P = 0.006) were related to risk of ER⁻ breast cancers. Several SNPs in IL8 (rs188246983, rs113973067), PRKDC (rs148411126, rs8178033, rs56411879), and MAP3K1 (rs863839, rs191188130) were associated with risk of ER⁻ breast cancer.

Among premenopausal women, genes associated with overall breast cancer risk were also associated with ER⁺ cancers, but not ER⁻ disease (Table 2; Supplementary Table S4). None of the genes associated with overall risk of ER⁻ breast cancer were observed when limited to premenopausal women, except for IL8, with two significant variants (rs113976067, rs188246983). The genes IL10 (P = 0.001) and HELLS (P = 0.008) were also associated with risk of premenopausal ER⁻ cancer, with ORs for SNPs associated with IL10 ranging from 1.5 to 1.8 for each additional copy of the variant allele. Several SNPs within the HELLS gene were significant (rs200175744, P_adj = 0.02; rs10882476; P_adj = 0.05; rs11188009, P_adj = 0.03).

Similar to the pattern observed in premenopausal women, genes associated with overall breast cancer risk in postmenopausal women were also associated with ER⁺ disease (IL2RB, CYP4F11, CASP8, POU2AF1, P ≤ 0.01; Table 2). Several SNP variants in POU2AF1, UBE2N, CD274, and XCR1 were associated with postmenopausal ER⁺ risk, including rs75716067 in POU2AF1 (P_adj = 0.003). The genes IL2RB (P = 0.008) and IRAK2 (P = 0.001), associated with overall breast cancer risk in postmenopausal women, were also associated with ER⁻ cancers. Generally, few SNPs were significantly associated with postmenopausal ER⁻ breast cancer. The most significant were intronic SNPs in IRAK2 (rs149858020, P_adj = 0.02) and PRKDC (rs8178153, P_adj = 0.03).

In case–case analyses comparing odds of being diagnosed with ER⁻ versus ER⁺ breast cancers, the only genes showing heterogeneity between ER⁻ and ER⁺ cancers for all women combined were TRAF1 (P = 0.001) and MAP3K1 (P = 0.007). There was also some indication that MAPK3 (P = 0.02) plays a more pronounced role in ER⁻ cancers. No heterogeneity were observed in pre- or postmenopausal women at the P = 0.01 level. Of the above genes, only rs863839 in MAP3K1 was significantly different in ER⁻ versus ER⁺ disease among all women combined (OR = 0.66; 95% CI: 0.52–0.82, P_adj = 0.04), with the study having 80% power to detect this level of difference.

In the PPI network for all significant immunity genes identified at P ≤ 0.01 (N = 38; Supplementary Table S6), there were a total of 22 genes with proteins (nodes) that were connected to at least one other protein (Fig. 1). The greatest number of protein–protein interactions were observed with IL2RB (node degree = 5), MAPK3 (node degree = 4), MAP3K1 (node degree = 6), IL8 (node degree = 5), and IL10 (node degree = 5). For ER⁺ cancers (N = 29 genes), MAP3K1 showing the most protein–protein interactions (node degree = 4), while IL2RB interacted with the most proteins (node degree = 4) for ER⁻ cancers (N = 26 genes).

The immune system plays a critical role in preventing and inhibiting tumor development, but may also act to promote tumor growth and progression (23). In these analyses, we used a comprehensive approach to investigate the role of hereditary immunity on breast cancer risk in AA women by assessing multiple immune response pathways, individual genes within these pathways, and the contribution of specific gene variants. Pathways and genes associated with regulation of immune system processes, immune activation, and inflammation were associated with overall breast cancer, with genes and SNPs in the NFκB pathway associated with innate immunity and activation of the inflammatory response playing a role in ER⁺ breast cancers and pathways associated with MAP3K1 activation playing a role in ER⁻ cancers. Of the genes identified by these analyses, only MAP3K1 was previously identified as a breast cancer GWAS locus, and previously reported to be related to ER⁻ cancer in the AMBER consortium (see Haddad and colleagues for discussion; ref. 15). We also noted that a number of genes related to dysregulation of humoral immunity, and involved in autoimmune and atopic disorders were associated with risk.

Individuals and populations vary in their resistance to infectious disease and pathogens. Much of this variation in phenotype is genetic, with pathogens acting as a selective force on genetic diversity (1, 24). Studies in West Africa show inheritance of immune phenotypes (3). The higher prevalence of bias toward a stronger T-helper type 2 (Th2) immune response among AAs is due in part to evolutionary responses to endemic helminth exposure in Sub-Saharan Africa, and is traditionally viewed as immunosuppressive, favoring tumor growth by inhibiting cell-mediated immunity and promoting angiogenesis. This immune bias contributes to higher incidence of atopic conditions such as asthma, which is characterized by chronic activation of humoral immunity pathways (25) and can lead to allergy-driven inflammatory cytokines (25, 26). In this study, a number of genes identified in relation to breast cancer risk were previously associated with asthma in GWAS studies, including IL2RB, HLA-DQ, and SMAD3, the main mediator of TGFβ signaling, which is important in T-cell activation and immune tolerance (27, 28), and promotes a Th2 immune phenotype (29). The variant rs2289259 in SMAD was strongly related to increased risk of ER⁻ cancer among postmenopausal women and is in high LD with rs2033784 (r² = 0.81), which affects SMAD3 expression in thyroid tissue and blood (18).

IL5 was the Th2-related pathway most strongly related to breast cancer risk in these analyses, and is the most important driver of eosinophil production and host defense against helminth parasite infections. Excessive production of IL5 and eosinophilia would be expected to dampen antitumor immune responses and increase allergy-related inflammation. Among postmenopausal women, the Biocarta IL5 pathway was significantly related to ER⁺ breast cancer, with significant associations observed for IL5, IL5RA, and CCR3 in gene-level analyses (P < 0.05). No significant variants, however, were found in any of these genes. These findings were similar to those in a recent pooled analysis of approximately 40,000 cases and 40,000 controls in the Breast Cancer Association Consortium, which found an association with IL5 in gene-level analyses, but did not identify individual SNPs associated with risk (30). PRG3, coding for proteoglycan 3, an eosinophil protein that is involved in the positive regulation of IL8, histamine, and leukotriene C4 release was also associated with breast cancer risk, but only among premenopausal women. The PRG3 variant rs4411290 and rs1867128 are eQTL sites and are associated with increased PRG2 expression in adipose tissue (31). These findings support the original hypothesis that a type 2, immunosuppressive phenotype plays a role in breast cancer risk among AA women, although it appears that this phenotype may be relevant for both ER⁺ and ER⁻ cancers.

In further support of a potential role for immunosuppression and breast cancer risk was the identification of several genes involved in PD-1 signaling, including among postmenopausal women an association between ER⁺ cancers and CD274, the primary ligand for the PD-1 checkpoint molecule on T cells, as well as an association between ER⁻ cancers and CD247, which is involved in the downregulation of T-lymphocytes due to chronic inflammation, and has been identified as a susceptibility locus for systemic sclerosis and rheumatoid arthritis (32, 33).

Several of the genes associated with breast cancer risk in our study have previously been associated with autoimmune diseases in GWAS studies, pointing to the possibility that dysregulated immune responses associated with autoimmune conditions, including chronic activation and proliferation of B cells, can alter breast cancer risk in AA women. Genes identified include IL21, HLA-DQA1, HLA-DMA, TRAF1, ICOSLG, and SMAD3. The IL21 rs2390350 variant associated with increased risk of breast cancer among postmenopausal women is in high LD with rs907715 (r² = 0.82), an independent susceptibility locus for systemic lupus erythematosus in both AAs and EAs (34, 35), and was found to be borderline significant in our analyses (OR = 0.86; 95% CI: 0.79–0.95; P_adj = 0.072). The gene PRKDC involved in regulation of autoimmune responses (36), as well as T-cell tolerance, inflammatory disease, and DNA repair was the gene most strongly related to risk of ER⁻ breast cancer. The missense variant (rs8178033) identified in this gene, however, was considered to be benign by PolyPhen-2 (Score: 0.29). Given that AA women are more likely to be diagnosed with systemic autoimmune diseases compared with EAs (37), greater understanding is needed of how these pathways might impact breast cancer risk in these women.

Several variants in genes that play a role in B-cell activation and development were also noted in this study. This included SNPs in POU2AF1, a B-cell specific transcriptional coactivator required for B-cell maturation (38), and a polymorphism in SLA2, which encodes for a SRC-like adaptor protein (SLAP) required for maintaining normal levels of B-cell receptor expression and development that is associated with several oncogenic signaling pathways and rheumatoid arthritis (39). The variant rs17848049 in the CXCR4 gene region was found to be associated with premenopausal breast cancer. The CXCR4 receptor binds to stromal cell-derived factor 1 and controls B-cell development, can activate inflammatory signaling pathways, and is dysregulated in autoimmune conditions (40). A recent finding from the Women's Health Initiative study of a specific prediagnostic autoimmune response signature related to humoral immunity in triple-negative breast cancers provide some support for B-cell pathways playing a role in breast cancer etiology (41), although our findings support an association for B-cell pathways in both ER⁺ and ER⁻ cancers among AAs.

In this study, the associations found between MAP3K1, IL1R1, IRAK2, and TRAF1 in the PPI network generated for ER⁺ cancers supports a role for inflammation related to innate immunity and the NFκB pathway as a contributor to disease. MAP3K1 activates CHUK and IKBKB, the central protein kinases of the NFκB pathway, and IL1R1 mediates IL1-dependent activation of NF-κB and MAPK, and strongly induces IL8 expression, a major mediator of the inflammatory response (42). Signaling involves recruitment of adaptor proteins such as IRAK2, and IRAK2, MAPK3, and IL8 were all found to be associated with increased risk of ER⁻ breast cancer. Associations with ER⁻ disease were also noted with other genes in innate immunity pathways including HELLS in premenopausal women. All three variants identified in HELLS (rs200175744, rs10882476, rs11188009) are associated with increased gene expression (31). Our findings are consistent with recent results indicating that inflammation is associated with breast cancer development (43).

The closely related IL2, IL15, and IL21 signaling pathways were found to play a prominent role in breast cancer risk in this study, and are involved in leukocyte development, and immune response activation and cessation. Signal transduction occurs via the Janus Kinase (JAK)–STAT pathway, the P13K–AKT pathway, and the MAPK pathway (reviewed in refs. 44, 45). IL21 triggers rapid activation of ERK1/2 required for IL21-induced cytokine production, which includes IL1, IL8, CCL3, CCL5, and CCL11 (46), and induces differentiation and production of cytotoxic T cells, macrophages, NK, and B cells. This is the first study to observe associations between breast cancer risk and IL21 variants.

IL2RB, coding for one of 3 subunits in the IL2 receptor complex, was the most significant gene associated with overall breast cancer risk, and found to be relevant for both ER⁺ and ER⁻ cancers in postmenopausal women, with rs228952 being the top variant identified. Generally, few studies have examined the role of IL2 receptors in breast cancer. The high affinity IL2 receptor composed of a β, γ, and α subunit binds to IL2 and is highly expressed on Treg cells (47). Increased ratios of Treg cells to total T lymphocytes, an important determinant of immune suppression, was recently shown in a case–cohort study in EPIC-Heidelberg to be associated with increased risk of ER⁻ breast cancers (7). Dysregulation of IL15 expression, potentially mediated by altered IL-2Rβγ function, may alternatively contribute to autoimmune diseases by inhibiting Treg-mediated self-tolerance (44, 48). In this study, there were suggestions that the IL15RA gene was associated with ER⁻ breast cancer, with significant heterogeneity by ER status (P = 0.02), but no SNPs were found to be significant.

Despite this being the first large-scale study of AAs that comprehensively examines the role of immune-related pathways, there was still limited power to detect associations by ER and menopausal status, and to observe significant pathway and gene-level associations after correcting for multiple testing. Limited sample size may have also contributed to the observation that specific genes and variants showing the strongest associations differed by menopausal and ER status, similar to previous findings (49). These differences, however, may have been due in part to age-related changes in immune function. Premenopausal women, for instance, are more prone to autoimmunity, while older postmenopausal women are at higher risk of chronic low-grade inflammation due to increased activation of the innate immune system (50). Most of the genetic variants of interest identified in this study were imputed, which is a potential study limitation, although we focused only on SNPs with high imputation INFO scores that achieved gene-wide significance. Regardless, this is the largest study to date in AAs that has examined the role of specific sub-components of immunity, such as the adaptive and innate immune response, inflammatory response, and cytokine-related pathways.

Results from this study represent an important extension of our understanding of how inherited genetic variation in immune pathways is relevant to breast cancer susceptibility and show the importance of pathways involved in innate immune response, immune activation, and immune suppression for both ER⁺ and ER⁻ cancers among AA women, including a role for Th2, B-cell, and PD-1–related pathways.

No potential conflicts of interest were disclosed.

Conception and design:C.-C. Hong, L.E. Sucheston-Campbell, S. Liu, K.P. Lee, A.F. Olshan, J.R. Palmer, C.B. Ambrosone

Development of methodology:C.-C. Hong, L.E. Sucheston-Campbell, S. Liu

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.):C.-C. Hong, S. Yao, E.V. Bandera, L.A. Rosenberg, C.A. Haiman, J.R. Palmer, C.B. Ambrosone

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis):C.-C. Hong, L.E. Sucheston-Campbell, S. Liu, Q. Hu, S. Yao, K.L. Lunetta, J.T. Bensen, C.A. Haiman, S.I. Abrams

Writing, review, and/or revision of the manuscript:C.-C. Hong, L.E. Sucheston-Campbell, S. Liu, S. Yao, K.L. Lunetta, S.A. Haddad, E.A. Ruiz-Narváez, J.T. Bensen, T.-Y. David Cheng, E.V. Bandera, S.S. Evans, S.I. Abrams, E.A. Repasky, A.F. Olshan, J.R. Palmer, C.B. Ambrosone

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases):C.-C. Hong, A.F. Olshan

Study supervision:S. Liu, K.L. Lunetta, C.B. Ambrosone

The research conducted by the AMBER Consortium is funded by the NIH and Foundation grants P01 CA151135 (to C.B. Ambrosone, J.R. Palmer, A.F. Olshan), R01 CA058420 (to L. Rosenberg), UM1 CA164974 (to J.R. Palmer, L. Rosenberg), R01 CA098663 (to J.R. Palmer), R01 CA100598 (to C.B. Ambrosone), R01 CA185623 (to E.V. Bandera, C.C. Hong, K. Demissie), UM1 CA164973 (to L. Le Marchand, L.N. Kolonel, C.A. Haiman, L.R. Wilkens), R01 CA54281 (to L.N. Kolonel), R01 CA063464 (to B.E. Henderson), P50 CA58223 (to M.A. Troester, A.F. Olshan), U01 CA179715 (to M.A. Troester, A.F. Olshan), Department of Defense Breast Cancer Research Program, Era of Hope Scholar Award ProgramW81XWH-08-1-0383 (to C.A. Haiman), the Susan G. Komen for the Cure Foundation (to M.A. Troester, A.F. Olshan), the Breast Cancer Research Foundation (to C.B. Ambrosone), and the University Cancer Research Fund of North Carolina (to M.A. Troester, A.F. Olshan). This work was also supported by the National Cancer Institute's Cancer Center Support Grant to Roswell Park Cancer Institute (P30CA016056).

Data on breast cancer cases in the Black Women's Health Study were obtained from 24 state cancer registries (Arizona, California, Colorado, Connecticut, Delaware, District of Columbia, Florida, Georgia, Illinois, Indiana, Kentucky, Louisiana, Maryland, Massachusetts, Michigan, New Jersey, New York, North Carolina, Oklahoma, Pennsylvania, South Carolina, Tennessee, Texas, and Virginia), and these results do not necessarily represent their views.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.