Advertisement
Short Communication

A Genome-wide Association Study of Early-Onset Breast Cancer Identifies PFKM as a Novel Breast Cancer Gene and Supports a Common Genetic Spectrum for Breast Cancer at Any Age

Habibul Ahsan, Jerry Halpern, Muhammad G. Kibriya, Brandon L. Pierce, Lin Tong, Eric Gamazon, Valerie McGuire, Anna Felberg, Jianxin Shi, Farzana Jasmine, Shantanu Roy, Rachelle Brutus, Maria Argos, Stephanie Melkonian, Jenny Chang-Claude, Irene Andrulis, John L. Hopper, Esther M. John, Kathi Malone, Giske Ursin, Marilie D. Gammon, Duncan C. Thomas, Daniela Seminara, Graham Casey, Julia A. Knight, Melissa C. Southey, Graham G. Giles, Regina M. Santella, Eunjung Lee, David Conti, David Duggan, Steve Gallinger, Robert Haile, Mark Jenkins, Noralane M. Lindor, Polly Newcomb, Kyriaki Michailidou, Carmel Apicella, Daniel J. Park, Julian Peto, Olivia Fletcher, Isabel dos Santos Silva, Mark Lathrop, David J. Hunter, Stephen J. Chanock, Alfons Meindl, Rita K. Schmutzler, Bertram Müller-Myhsok, Magdalena Lochmann, Lars Beckmann, Rebecca Hein, Enes Makalic, Daniel F. Schmidt, Quang Minh Bui, Jennifer Stone, Dieter Flesch-Janys, Norbert Dahmen, Heli Nevanlinna, Kristiina Aittomäki, Carl Blomqvist, Per Hall, Kamila Czene, Astrid Irwanto, Jianjun Liu, Nazneen Rahman, Clare Turnbull, for the Familial Breast Cancer Study, Alison M. Dunning, Paul Pharoah, Quinten Waisfisz, Hanne Meijers-Heijboer, Andre G. Uitterlinden, Fernando Rivadeneira, Dan Nicolae, Douglas F. Easton, Nancy J. Cox and Alice S. Whittemore
DOI: 10.1158/1055-9965.EPI-13-0340 Published April 2014

Abstract

Early-onset breast cancer (EOBC) causes substantial loss of life and productivity, creating a major burden among women worldwide. We analyzed 1,265,548 Hapmap3 single-nucleotide polymorphisms (SNP) among a discovery set of 3,523 EOBC incident cases and 2,702 population control women ages ≤ 51 years. The SNPs with smallest P values were examined in a replication set of 3,470 EOBC cases and 5,475 control women. We also tested EOBC association with 19,684 genes by annotating each gene with putative functional SNPs, and then combining their P values to obtain a gene-based P value. We examined the gene with smallest P value for replication in 1,145 breast cancer cases and 1,142 control women. The combined discovery and replication sets identified 72 new SNPs associated with EOBC (P < 4 × 10−8) located in six genomic regions previously reported to contain SNPs associated largely with later-onset breast cancer (LOBC). SNP rs2229882 and 10 other SNPs on chromosome 5q11.2 remained associated (P < 6 × 10−4) after adjustment for the strongest published SNPs in the region. Thirty-two of the 82 currently known LOBC SNPs were associated with EOBC (P < 0.05). Low power is likely responsible for the remaining 50 unassociated known LOBC SNPs. The gene-based analysis identified an association between breast cancer and the phosphofructokinase-muscle (PFKM) gene on chromosome 12q13.11 that met the genome-wide gene-based threshold of 2.5 × 10−6. In conclusion, EOBC and LOBC seem to have similar genetic etiologies; the 5q11.2 region may contain multiple distinct breast cancer loci; and the PFKM gene region is worthy of further investigation. These findings should enhance our understanding of the etiology of breast cancer. Cancer Epidemiol Biomarkers Prev; 23(4); 658–69. ©2014 AACR.

Introduction

Early-onset breast cancer (EOBC) leads to substantial loss of life and productivity, creating a major public health and economic burden in both developed and developing countries. Many patterns of breast cancer incidence, histopathologic characteristics, clinical behavior, and risk factors, including the increase in risk associated with a family history, differ between cases diagnosed during premenopausal and postmenopausal periods; a difference that has prompted speculation that there might be some genetic etiologies that are different for EOBC and later-onset breast cancer (LOBC; refs. 1–4). For example, a study of Utah families estimated that the risk of developing breast cancer for sisters of EOBC cases was 3.70 [95% confidence interval (CI), 2.5–5.2] times that for the general Caucasian population, nearly double the 1.83-fold relative risk (95% CI, 1.65–2.01) among sisters of cases of all ages (5). About 25% of the aggregation is explained by the high risks specific to carriers of deleterious mutations in the major susceptibility genes BRCA1 and BRCA2, but even after excluding carrier families, risks are higher for relatives of EOBC cases than among relatives of LOBC cases (3, 6, 7). Recently, genome-wide association studies (GWAS) have reported many single-nucleotide polymorphisms (SNP) as associated with breast cancer risk (8). To date, however, no published GWASs have focused on EOBC. Here, we report findings from the first large-scale GWAS of EOBC involving a discovery set of 6,225 young Caucasian women from eight sites in the United States, Canada, Australia, and Germany and two replication sets of Caucasian women from Australia, the United States, the United Kingdom, and other European countries.

Materials and Methods

We used a case–control design to investigate EOBC risk among Caucasian women in relation to 1,265,546 SNPs included in the HapMap3 project (http://hapmap.ncbi.nlm.nih.gov/downloads/phasing/2009-02_phaseIII/HapMap3_r2/). Specifically, we used Illumina SNP arrays to genotype 3,523 EOBC cases and 2,702 control women and to impute their genotypes for the HaMap3 SNPs (hereafter called the discovery set). We then conducted two SNP-based analyses and a gene-based analysis. The results of these analyses were examined in two sets of independent data (called replication sets). We begin with a description of subject recruitment, genotyping, and quality control for the discovery set. We then describe the SNP-based analysis and replication, followed by the gene-based analysis and replication.

Discovery set

Subject recruitment.

Population-based subjects were recruited from the eight sites described in Supplementary Table S1, some of which oversampled cases with a personal or family history, suggesting a heritable basis for their disease (9–14). Eligible cases were non-Hispanic White (NHW) women diagnosed with invasive breast cancer when 51 years or younger and not known to carry pathogenic mutations in BRCA1 or BRCA2. Eligible controls were NHW women ages 20 to 51 years without a history of breast cancer, who were identified largely by random-digit dialing. Supplementary Table S2 shows the number of eligible subjects from each of the eight contributing sites after quality control.

Genotyping and quality control.

DNA samples for subjects from all but one of the sites were genotyped at the University of Chicago on Illumina 610-Quad and Cyto12 v2 BeadChips (Illumina Inc.), using the protocol described in the Supplementary Methods. Two hundred and twenty-seven population control subjects from the Colon Cancer Family Registry (CCFR) were genotyped at TGEN (Translational Genomics Research Institute; http://www.tgen.org) using the Illumina Human1M and HumanOmni1-Quad BeadChips. In addition, 27 blinded and 22 unblinded quality control replicates from the study sample were genotyped on the Human1M. Replicates showed concordance of called genotypes >99.94% (for samples with call rates >90%). Standard laboratory quality control procedures were applied and have been described previously (15). Quality control was implemented using a combination of PLINK (16) and custom programs written in C, R, Perl, and the Unix bash shell. Data quality control procedures are described in more detail in Supplementary Methods and summarized in Supplementary Table S3. This table shows that 555,254 of the 1,298,078 SNPs remained after quality control, and that most SNPs were deleted because they appeared only on the 1M and 1M Omni chips that were used to type only 227 controls.

Analysis.

We first identified principal components representing axes of ancestral variation to adjust for population stratification (17) and imputed untyped SNPs using the HapMap3 data. We then conducted two SNP-based analyses and a gene-based analysis. The first SNP-based analysis consisted of SNP-specific logistic regressions for each of the 1,265,548 typed or imputed HapMap3 SNPs using BEAGLE (18). We checked for population stratification using graphical plots of test statistics and the λ measure of overdispersion (19). We used an additive regression model in which the logit of EOBC risk was linearly related to the number of SNP minor alleles, and noted the SNPs with nominal P values less than 4 × 10−8. These SNPs and their minor allele frequencies (MAFs) for cases and controls as well as the discovery set P values are shown in Supplementary Table S4. The second SNP-based analysis was conducted to examine association between EOBC and each of the 82 breast cancer–associated SNPs currently reported and validated in the literature. (We were unable to impute one SNP that was not polymorphic in the HapMap3 data.) Here, we used SNP-specific logistic regressions for each of the 82 SNPs in which the logit of EOBC risk was linearly related to the number of SNP risk alleles, as reported in the literature.

The gene-based analysis was conducted in two steps. First, we attempted to annotate each known human gene with one or more of the SNPs in the discovery set that could affect its expression and/or function. Then, we combined the EOBC discovery set P values of these expression-related SNPs into summary gene-based P values. For step 1, we used expression-quantitative trait locus (eQTL) mapping of SNPs to genes, as implemented in the online database SCAN (20, 21) and used the eQTL significance levels to quantify the likelihood that a SNP [or one in strong linkage disequilibrium (LD) with it] regulates gene transcript levels (22). That is, we assigned an SNP to a gene if the SNP encoded a missense, nonsense, or frameshift (MNF) variant in the gene, or if it met our criteria for an eQTL SNP for the gene. Although not all the SNPs annotated to a gene are likely to be functional, they are clearly enriched for those with functional consequence. We were able to annotate 19,684 genes with one or more putative functional SNPs, 11,040 of which were annotated with at least one e-QTL SNP. In step 2, we calculated a gene-based P value for each of the 19,684 genes by combining the EOBC-association P values for all its putative functional SNPs using methods described elsewhere (23).

The Supplementary Methods contains additional details about both SNP-based and gene-based analyses.

Replication sets

Replication of SNP-based results.

Primary genotype data were obtained from three early-onset breast cancer GWAS in populations of European ancestry (3, 24–28) as described in Supplementary Tables S5 and S6. For each typed or imputed SNP using ProbABEL (29), we combined the SNP-specific regression coefficients obtained for the discovery and replication sets using the commonly-deemed inverse-weighted summary statistic proposed by Cochran (30).

Replication of gene-based results.

To replicate the gene-based association analyses, we used available GWAS data from the CGEMS breast cancer study of 1,145 Caucasian case women and 1,142 Caucasian control women ages 55 to 74 years. Details of subject selection, genotyping methods, and quality control analyses for CGEMS breast cancer project have been published (31, 32). The identical gene-based analytic method, described above and in Supplementary Methods, was applied to the CGEMS data obtained from dbGaP. The gene-based P values from both discovery and replication datasets were combined for the gene with smallest gene-based P value in discovery data using Fisher's method for meta-analysis (33).

Further details of both the SNP-based and gene-based replication sets can be found in Supplementary Method.

Results

SNP-based analysis

Analysis of combined discovery and replication sets identified 96 SNPs from six chromosomal regions as associated with EOBC risk with P < 4 × 10−8 (the threshold for genome-wide significance at level 0.05 with 1.2 million independent tests). These results were not driven by data from a single site. The six regions lie on chromosomes 3p24.1, 5q11.2, 8q24, 10q26.13, 11q13.2, and 16q12.1. Previous GWASs have associated SNPs in these regions with (largely later-onset) breast cancer; however, they have reported only 24 of these 96 SNPs (Table 1; refs. 28, 31, 32, 34–60). To investigate how many of the remaining 72 unpublished SNPs are independently associated with EOBC, we evaluated each of them using a regression model that also contained the published SNP in the region having the smallest P value in the combined discovery and replication data (called the index SNP). These regressions identified 12 of the 72 SNPs as independently associated with EOBC at significance level P < 0.001 (listed in bold in Table 1). Eleven of these 12 SNPs are in the 5q11.2 region and almost all are within or near the MAP3K1 gene; eight are downstream of the published index SNP (Fig. 1). The strongest of these SNPs is rs2229882 with unadjusted P value 1.02 × 10−14 and squared correlation r2 = 0.10 with the published SNP rs889312 (Fig. 1 and Table 1).

Figure 1.
Figure 1.

Manhattan plot of significance levels from combined discovery and replication data for SNPs in the 5q11.2 region. Y-axis, minus log P value for association with EOBC; X-axis, chromosomal position; and SNP color reflects its correlation with SNP rs889312 (SNP with smallest P value in discovery set, marked by arrow). SNPs in red boxes are associated with P value < 0.001 from regression analyses adjusting for rs889312. Horizontal bar, genome-wide significance threshold P = 4 × 10−8; blue curve, recombination rate.

View this table:
Table 1.

Newly identified SNPs associated with EOBC at combined significance level P < 4 × 10−8

To further explore the 5q11.2 association, we examined 2,889 SNPs (278 typed and 2,611 imputed using 1 KG data) within a 2 Mb region centered at rs889312, the strongest published SNP in the region. We found rs7709971 to have the smallest P value (1.01 × 10−9). Adjusting for this SNP in bivariate regressions did not produce strong new associations for any of the other SNPs in the region (results not shown).

Association with known breast cancer SNPs

Table 2 shows 83 SNPs reported in the GWAS catalog http://www.genome.gov/26525384#1 as associated with breast cancer at P < 4 × 10−8 in studies of predominantly LOBC (28, 31, 32, 34–60). We used the discovery set to examine association between EOBC and the 82 SNPs that we could impute using HapMap3 and/or 1 KG data. Table 2 shows that 32 SNPs were associated at P < 0.05 (listed in bold in the table). We also computed the probability that a test of size 0.05 using 3,523 cases and 2,702 controls would detect association with each of the 82 SNPs, given its published effect size as shown in the table. We found that the mean power to detect the 50 missed SNPs was 44%, appreciably lower than the mean power of 77% for the 32 we detected. Thus our failure to confirm the remaining 50 SNPs seems due to insufficient power to detect their small effect sizes. These results suggest that the genetic etiology of EOBC is not different than that of LOBC.

View this table:
Table 2.

EOBC discovery set risk allele frequencies (RAFs) and per-allele odds-ratios (ORs) for known breast cancer hits (GWAS Catalogue) in subjects of European ancestry (P ≤ 5 × 10−8)

Gene-based analysis

Analysis of the discovery set identified the phosphofructokinase muscle-type (PFKM) gene on chromosome 12q13.11 region as associated with EOBC with P value of 9 × 10−7, which meets the genome-wide threshold of P < 2.5 × 10−6 for the 19,684 statistical tests performed. This region is distinct from the regions 12q22 and 12q24 containing SNPs known to be associated with breast cancer (Table 2). When we repeated the same analysis using the predominantly LOBC breast cancer replication data from the CGEMS study, the PFKM gene also was associated with breast cancer (P = 3 × 10−2). Combined analysis of the two data sets yielded an overall gene-based Fisher's meta P value of 5 × 10−7 for the PFKM gene. No other genes met the genome-wide significance threshold.

The association between PFKM and breast cancer risk was based on its annotation with the 35 putative functional SNPs shown in Table 3. This set consists largely of trans e-QTL SNPs rather than MNF SNPs in the coding region of the gene. Nevertheless, we also found evidence implicating SNPs in the 1 M region centered at the PFKM gene. We found that 27 of the 966 SNPs in this region that were included in the EOBC GWAS discovery set were associated with EOBC at P < 0.01. These SNPs are listed in Fig. 2. Also shown in the figure are the genes in this region (Fig. 2A), a Manhattan plot of the 966 P values (Fig. 2B), and the D′ measure of linkage disequilibrium between pairs of SNPs (Fig. 2C). To evaluate the statistical significance of this finding, we permuted subjects' case–control statuses 1,000-times, and in each permutation, we evaluated how many of the 966 SNPs were associated with EOBC at P < 0.01. We found that none of the 1,000 permutations yielded 27 or more such SNPs, giving a significance level of P < 0.001. Most of the 27 EOBC-associated SNPs were located within other nearby genes, suggesting that EOBC risk could be due to some complex gene expression pattern in this gene-rich region (see A and B of Fig. 2). Fig. 2C of the figure shows the correlations among the SNPs in the region.

Figure 2.
Figure 2.

A, chromosomal positions of genes on chr12q13.11 in the 2 Mb region surrounding the PFKM gene. B, Manhattan plot of 27 SNPs in the region associated with EOBC with combined discovery and replication P values of 0.01 or less. These 27 SNPs and their P values are listed on the left in their order of appearance from left to right. SNP colors reflect magnitudes of their squared correlation coefficients with SNP rs7296288 (marked by arrow), which had the smallest discovery set P value. C, linkage disequilibrium measures D′ for all 966 HapMap3 imputed SNPs in the chr12q13.11 region. Dark red squares represent D′ values near 1 and white squares represent D′ values near zero.

View this table:
Table 3.

Significance levels from discovery and replication sets for association of EOBC with 35 putatively functional PFKM SNPs

Discussion

This study identified and replicated EOBC associations with 72 previously unpublished SNPs in six regions known to harbor variants affecting breast cancer risk. Twelve of the 72 SNPs remained associated with EOBC after adjusting for the SNP with smallest published P value in the same region. Eleven of these 12 SNPs lie on chromosome 5q11.2 near the MAP3K1 gene. Their lack of strong correlation with the strongest published SNP rs889312 suggests the presence of multiple causal variants in this region. Future sequence-based studies, coupled with functional experiments, can exploit these associations to identify the causal variant(s) in the region.

We examined association between EOBC and 82 of the 83 common SNPs currently known to be associated with largely LOBC. We found evidence for association with only 32 (39%) of these SNPs. However, comparison of detected and missed SNPs with respect to effect size and power suggests that this low confirmation rate reflects the inadequate power to detect the missed SNPs rather than systematic etiologic differences between EOBC and LOBC. These findings suggest that the genetic factors responsible for breast cancer affect risk at all ages.

The gene-based GWAS analyses identified the PFKM gene region 12q13.11 to be associated with breast cancer risk, independently of the 12q22 and 12q24 regions previously associated with breast cancer. PFKM, one of the three PFK isoenzymes, is the key regulator of cellular glycolysis catalyzing the phosphorylation of fructose-6-phosphate to fructose-1,6-bisphosphate. Disabling PFKM mutations leads to glycogen storage diseases (especially type VII—Tarui's disease) as well as cardiac and hematologic disorders (61–63). The association of PFKM expression with breast cancer risk is plausible for several reasons. First, this gene is expressed in breast cancer cell lines (64). Second, variants in the gene have been related to posttranslational modifications, which have been shown to alter the metabolism and promote the growth of cancer cells (65). Third, an association between breast cancer risk and this gene is consistent with observations that tumor cells can consume large amounts of glucose due to aberrant glucose metabolism, especially through a glycolytic pathway that produces lactate (65). Finally, tumor-suppressor protein p53 has been shown to suppress PFKM expression in model system (66). Because the biology of the PFKM gene and its modulators and inhibitors are well characterized (67, 68), identification of PFKM gene region as a breast cancer susceptibility locus has potential translational implications for breast cancer prevention and treatment.

The present study has several strengths, including its large sample size, its focus on EOBC, its homogenous Caucasian study population, and its novel gene-based analysis involving the functional characteristics of gene-related SNPs. Study limitations include use of somewhat different types of study populations between the discovery (population-based) and replication (both population- and clinic-based) phases, and our inability to replicate the gene-based analysis in an EOBC replication set due to lack of access to necessary relevant data from replication cohorts.

In conclusion, the study identified EOBC risks to be associated with 72 new SNPs in six chromosomal regions that were previously associated with LOBC risks. Eleven of the 72 SNPs, all on chromosome 5q11.2, were associated with EOBC independently of previously reported SNPs. These EOBC-associated SNPs may help in the search for causal variants in the 5q11.2 region. In addition, we found little evidence to support genetic heterogeneity between EOBC and LOBC. Finally, the gene-based analysis identified a region containing the key glycolysis regulation gene PFKM that is worthy of further investigation as a susceptibility locus for breast cancer in Caucasian women of all ages. Future studies need to determine whether the current findings apply to non-Caucasian women.

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.

Disclaimer

The content of this article does not necessarily reflect the views or policies of the NCI or any of the collaborating centers in the BCFR, nor does mention of trade names, commercial products, or organizations imply endorsement by the U.S. Government or the BCFR.

Authors' Contributions

Conception and design: H. Ahsan, M.G. Kibriya, D.C. Thomas, D. Seminara, G.G. Giles, M. Lathrop, R.K. Schmutzler, N. Dahmen, C. Blomqvist, N. Rahman, A.M. Dunning, N.J. Cox, A.S. Whittemore

Development of methodology: J. Halpern, B.L. Pierce, D.C. Thomas, M. Lochmann, N. Rahman, N.J. Cox, A.S. Whittemore

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): H. Ahsan, M.G. Kibriya, F. Jasmine, J. Chang-Claude, I. Andrulis, J.L. Hopper, E.M. John, K. Malone, G. Ursin, M.D. Gammon, D. Seminara, G. Casey, J.A. Knight, M.C. Southey, G.G. Giles, R.M. Santella, E. Lee, D. Duggan, R. Haile, M. Jenkins, N.M. Lindor, P. Newcomb, C. Apicella, J. Peto, O. Fletcher, I. dos Santos Silva, M. Lathrop, D.J. Hunter, S.J. Chanock, A. Meindl, R.K. Schmutzler, M. Lochmann, J. Stone, D. Flesch-Janys, N. Dahmen, H. Nevanlinna, K. Aittomäki, P. Hall, K. Czene, J. Liu, C. Turnbull, A.M. Dunning, P. Pharoah, Q. Waisfisz, F. Rivadeneira, D.F. Easton

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): H. Ahsan, J. Halpern, M.G. Kibriya, B.L. Pierce, L. Tong, E. Gamazon, J. Shi, F. Jasmine, S. Melkonian, J. Chang-Claude, D.C. Thomas, M.C. Southey, D. Conti, S. Gallinger, K. Michailidou, M. Lathrop, D.J. Hunter, B. Müller-Myhsok, M. Lochmann, L. Beckmann, E. Makalic, D.F. Schmidt, Q.M. Bui, P. Hall, A. Irwanto, Q. Waisfisz, D. Nicolae, D.F. Easton, N.J. Cox, A.S. Whittemore

Writing, review, and/or revision of the manuscript: H. Ahsan, J. Halpern, M.G. Kibriya, B.L. Pierce, E. Gamazon, V. McGuire, M. Argos, J. Chang-Claude, I. Andrulis, J.L. Hopper, E.M. John, K. Malone, G. Ursin, M.D. Gammon, D.C. Thomas, D. Seminara, J.A. Knight, M.C. Southey, G.G. Giles, E. Lee, D. Conti, R. Haile, M. Jenkins, J. Peto, I. dos Santos Silva, R. Hein, H. Nevanlinna, C. Blomqvist, P. Hall, K. Czene, A.M. Dunning, P. Pharoah, H. Meijers-Heijboer, A.G. Uitterlinden, F. Rivadeneira, D.F. Easton, N.J. Cox, A.S. Whittemore

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): H. Ahsan, J. Halpern, V. McGuire, A. Felberg, M. Argos, S. Melkonian, E.M. John, D. Seminara, D.J Park, A. Meindl, R.K. Schmutzler, J. Stone, H. Meijers-Heijboer, F. Rivadeneira, A.S. Whittemore

Study supervision: H. Ahsan, M.G. Kibriya, M. Argos, J.L. Hopper, D.J Park, A.S. Whittemore

Other: Running laboratory assay and data generation, S. Roy; laboratory work, R. Brutus

Grant Support

This study was supported by NIH Grants U01CA122171, RC1CA145506, R01CA094069, and U19CA148065. The Breast Cancer Family Registry (BCFR) is supported by the NCI, NIH under RFA-CA-06-503 and through cooperative agreements with members of the BCFR and principal investigators, including Cancer Care Ontario (U01 CA69467), Cancer Prevention Institute of California (CPIC; U01 CA69417), and University of Melbourne (U01 CA69638). Samples from the CPIC were processed and distributed by the Coriell Institute for Medical Research. The Colon Cancer Family Registry (CCFR) is supported by the NCI, NIH under RFA-CA-95-011.

Acknowledgments

The authors are thankful for the use of data from the Database of Genotypes and Phenotypes (dbGaP), and thank all investigators who contributed the phenotype data and DNA samples to the CGEMS project, and the NCI, the primary funder of the CGEMS genome-wide association study.

Footnotes

  • Note: Supplementary data for this article are available at Cancer Epidemiology, Biomarkers & Prevention Online (http://cebp.aacrjournals.org/).

  • Received April 2, 2013.
  • Revision received January 29, 2014.
  • Accepted January 29, 2014.

References

  1. 1.↵
    1. Narod SA
    . Early-onset breast cancer: what do we know about the risk factors?: a countercurrents series. Curr Oncol 2011;18:204–5.
    OpenUrlPubMed
  2. 2.↵
    1. Renehan AG,
    2. Tyson M,
    3. Egger M,
    4. Heller RF,
    5. Zwahlen M
    . Body-mass index and incidence of cancer: a systematic review and meta-analysis of prospective observational studies. Lancet 2008;371:569–78.
    OpenUrlCrossRefPubMed
  3. 3.↵
    1. Dite GS,
    2. Jenkins MA,
    3. Southey MC,
    4. Hocking JS,
    5. Giles GG,
    6. McCredie MR,
    7. et al.
    Familial risks, early-onset breast cancer, and BRCA1 and BRCA2 germline mutations. J Natl Cancer Inst 2003;95:448–57.
    OpenUrlAbstract/FREE Full Text
  4. 4.↵
    1. Whittemore AS,
    2. Gong G,
    3. Itnyre J
    . Prevalence and contribution of BRCA1 mutations in breast cancer and ovarian cancer: results from three U.S. population-based case-control studies of ovarian cancer. Am J Hum Genet 1997;60:496–504.
    OpenUrlPubMed
  5. 5.↵
    1. Goldgar DE,
    2. Easton DF,
    3. Cannon-Albright LA,
    4. Skolnick MH
    . Systematic population-based assessment of cancer risk in first-degree relatives of cancer probands. J Natl Cancer Inst 1994;86:1600–8.
    OpenUrlAbstract/FREE Full Text
  6. 6.↵
    Collaborative Group on Hormonal Factors in Breast Cancer. Familial breast cancer: collaborative reanalysis of individual data from 52 epidemiological studies including 58,209 women with breast cancer and 101,986 women without the disease. Lancet 2001;358:1389–99.
    OpenUrlCrossRefPubMed
  7. 7.↵
    1. Lee JS,
    2. John EM,
    3. McGuire V,
    4. Felberg A,
    5. Ostrow KL,
    6. DiCioccio RA,
    7. et al.
    Breast and ovarian cancer in relatives of cancer patients, with and without BRCA mutations. Cancer Epidemiol Biomarkers Prev 2006;15:359–63.
    OpenUrlAbstract/FREE Full Text
  8. 8.↵
    1. Hindorff LA,
    2. MacArthur J,
    3. Morales J,
    4. Junkins HA,
    5. Hall PN,
    6. Klemm AK,
    7. et al.
    A catalog of published genome-wide association studies. [accessed Sept 2012]. Available from: www.genome.gov/gwastudies.
  9. 9.↵
    1. John EM,
    2. Hopper JL,
    3. Beck JC,
    4. Knight JA,
    5. Neuhausen SL,
    6. Senie RT,
    7. et al.
    The Breast Cancer Family Registry: an infrastructure for cooperative multinational, interdisciplinary and translational studies of the genetic epidemiology of breast cancer. Breast Cancer Res 2004;6:R375–89.
    OpenUrlCrossRefPubMed
  10. 10.↵
    1. Chang-Claude J,
    2. Eby N,
    3. Kiechle M,
    4. Bastert G,
    5. Becher H
    . Breastfeeding and breast cancer risk by age 50 among women in Germany. Cancer Causes Control 2000;11:687–95.
    OpenUrlCrossRefPubMed
  11. 11.↵
    1. Gammon MD,
    2. Neugut AI,
    3. Santella RM,
    4. Teitelbaum SL,
    5. Britton JA,
    6. Terry MB,
    7. et al.
    The Long Island Breast Cancer Study Project: description of a multi-institutional collaboration to identify environmental risk factors for breast cancer. Breast Cancer Res Treat 2002;74:235–54.
    OpenUrlCrossRefPubMed
  12. 12.↵
    1. Friedrichsen DM,
    2. Malone KE,
    3. Doody DR,
    4. Daling JR,
    5. Ostrander EA
    . Frequency of CHEK2 mutations in a population-based case-control study of breast cancer in young women. Breast Cancer Res 2004;6:R629–35.
    OpenUrlCrossRefPubMed
  13. 13.↵
    1. Lee E,
    2. Ma H,
    3. McKean-Cowdin R,
    4. Van Den Berg D,
    5. Bernstein L,
    6. Henderson BE,
    7. et al.
    Effect of reproductive factors and oral contraceptives on breast cancer risk in BRCA1/2 mutation carriers and noncarriers: results from a population-based study. Cancer Epidemiol Biomarkers Prev 2008;17:3170–8.
    OpenUrlAbstract/FREE Full Text
  14. 14.↵
    1. Newcomb PA,
    2. Baron J,
    3. Cotterchio M,
    4. Gallinger S,
    5. Grove J,
    6. Haile R,
    7. et al.
    Colon Cancer Family Registry: an international resource for studies of the genetic epidemiology of colon cancer. Cancer Epidemiol Biomarkers Prev 2007;16:2331–43.
    OpenUrlAbstract/FREE Full Text
  15. 15.↵
    1. Figueiredo JC,
    2. Lewinger JP,
    3. Song C,
    4. Campbell PT,
    5. Conti DV,
    6. Edlund CK,
    7. et al.
    Genotype-environment interactions in microsatellite stable/microsatellite instability-low colorectal cancer: results from a genome-wide association study. Cancer Epidemiol Biomarkers Prev 2011;20:758–66.
    OpenUrlAbstract/FREE Full Text
  16. 16.↵
    1. Purcell S,
    2. Neale B,
    3. Todd-Brown K,
    4. Thomas L,
    5. Ferreira MA,
    6. Bender D,
    7. et al.
    PLINK: a toolset for whole-genome association and population-based linkage analysis. Am J Hum Genet 2007;81:559–75.
    OpenUrlCrossRefPubMed
  17. 17.↵
    1. Price AL,
    2. Patterson NJ,
    3. Plenge RM,
    4. Weinblatt ME,
    5. Shadick NA,
    6. Reich D
    . Principal components analysis corrects for stratification in genome-wide association studies. Nat Genet 2006;38:904–9.
    OpenUrlCrossRefPubMed
  18. 18.↵
    1. Browning BL,
    2. Browning SR
    . Efficient multilocus association testing for whole genome association studies using localized haplotype clustering. Genet Epidemiol 2007;31:365–75.
    OpenUrlCrossRefPubMed
  19. 19.↵
    1. Devlin B,
    2. Roeder K
    . Genomic control for association studies. Biometrics 1999;55:997–1004.
    OpenUrlCrossRefPubMed
  20. 20.↵
    1. Nicolae DL,
    2. Gamazon E,
    3. Zhang W,
    4. Duan S,
    5. Dolan ME,
    6. Cox NJ
    . Trait-associated SNPs are more likely to be eQTLs: annotation to enhance discovery from GWAS. PLoS Genet 2010;6:e1000888.
    OpenUrlCrossRefPubMed
  21. 21.↵
    1. Gamazon ER,
    2. Zhang W,
    3. Konkashbaev A,
    4. Duan S,
    5. Kistner EO,
    6. Nicolae DL,
    7. et al.
    SCAN: SNP and copy number annotation. Bioinformatics 2010;26:259–62.
    OpenUrlAbstract/FREE Full Text
  22. 22.↵
    1. Duan S,
    2. Huang RS,
    3. Zhang W,
    4. Bleibel WK,
    5. Roe CA,
    6. Clark TA,
    7. et al.
    Genetic architecture of transcript-level variation in humans. Am J Hum Genet 2008;82:1101–13.
    OpenUrlCrossRefPubMed
  23. 23.↵
    1. De la Cruz O,
    2. Wen X,
    3. Ke B,
    4. Song M,
    5. Nicolae DL
    . Gene, region and pathway level analyses in whole-genome studies. Genet Epidemiol 2010;34:222–31.
    OpenUrlPubMed
  24. 24.↵
    1. Osborne RH,
    2. Hopper JL,
    3. Kirk JA,
    4. Chenevix-Trench G,
    5. Thorne HJ,
    6. Sambrook JF
    . kConFab: a research resource of Australasian breast cancer families. Kathleen Cuningham Foundation Consortium for Research into Familial Breast Cancer. Med J Aust 2000;172:463–4.
    OpenUrlPubMed
  25. 25.↵
    1. Fletcher O,
    2. Johnson