Figures
Abstract
While interplay between BRCA1 and AURKA-RHAMM-TPX2-TUBG1 regulates mammary epithelial polarization, common genetic variation in HMMR (gene product RHAMM) may be associated with risk of breast cancer in BRCA1 mutation carriers. Following on these observations, we further assessed the link between the AURKA-HMMR-TPX2-TUBG1 functional module and risk of breast cancer in BRCA1 or BRCA2 mutation carriers. Forty-one single nucleotide polymorphisms (SNPs) were genotyped in 15,252 BRCA1 and 8,211 BRCA2 mutation carriers and subsequently analyzed using a retrospective likelihood approach. The association of HMMR rs299290 with breast cancer risk in BRCA1 mutation carriers was confirmed: per-allele hazard ratio (HR) = 1.10, 95% confidence interval (CI) 1.04 – 1.15, p = 1.9 x 10−4 (false discovery rate (FDR)-adjusted p = 0.043). Variation in CSTF1, located next to AURKA, was also found to be associated with breast cancer risk in BRCA2 mutation carriers: rs2426618 per-allele HR = 1.10, 95% CI 1.03 – 1.16, p = 0.005 (FDR-adjusted p = 0.045). Assessment of pairwise interactions provided suggestions (FDR-adjusted pinteraction values > 0.05) for deviations from the multiplicative model for rs299290 and CSTF1 rs6064391, and rs299290 and TUBG1 rs11649877 in both BRCA1 and BRCA2 mutation carriers. Following these suggestions, the expression of HMMR and AURKA or TUBG1 in sporadic breast tumors was found to potentially interact, influencing patients’ survival. Together, the results of this study support the hypothesis of a causative link between altered function of AURKA-HMMR-TPX2-TUBG1 and breast carcinogenesis in BRCA1/2 mutation carriers.
Citation: Blanco I, Kuchenbaecker K, Cuadras D, Wang X, Barrowdale D, de Garibay GR, et al. (2015) Assessing Associations between the AURKA-HMMR-TPX2-TUBG1 Functional Module and Breast Cancer Risk in BRCA1/2 Mutation Carriers. PLoS ONE 10(4): e0120020. https://doi.org/10.1371/journal.pone.0120020
Academic Editor: Hiromu Suzuki, Sapporo Medical University, JAPAN
Received: June 25, 2014; Accepted: January 22, 2015; Published: April 1, 2015
This is an open access article, free of all copyright, and may be freely reproduced, distributed, transmitted, modified, built upon, or otherwise used by anyone for any lawful purpose. The work is made available under the Creative Commons CC0 public domain dedication
Data Availability: All relevant data are within the paper and its Supporting Information files.
Funding: BCFR—all. This work was supported by grant UM1 CA164920 from the National Cancer Institute. The content of this manuscript does not necessarily reflect the views or policies of the National Cancer Institute or any of the collaborating centers in the Breast Cancer Family Registry (BCFR), nor does mention of trade names, commercial products, or organizations imply endorsement by the United States Government or the BCFR. BFBOCC: BFBOCC is supported by: Lithuania (BFBOCC-LT): Research Council of Lithuania grant LIG-07/2012 and Hereditary Cancer Association (Paveldimo vėžio asociacija); Latvia (BFBOCC-LV) is partly supported by LSC grant 10.0010.08 and in part by a grant from the ESF Nr.2009/0220/1DP/1.1.1.2.0/09/APIA/VIAA/016 and Liepaja's municipal council. BMBSA: BRCA-gene mutations and breast cancer in South African women (BMBSA) was supported by grants from the Cancer Association of South Africa (CANSA) to EJR. BRICOH: SLN was partially supported by the Morris and Horowitz Familes Endowed Professorship. CBCS: This work was supported by the NEYE Foundation. CNIO: This work was partially supported by Spanish Association against Cancer (AECC08), RTICC 06/0020/1060, FISPI08/1120, Mutua Madrileña Foundation (FMMA) and SAF2010-20493. COH-CCGCRN: City of Hope Clinical Cancer Genetics Community Network and the Hereditary Cancer Research Registry, supported in part by Award Number RC4CA153828 (PI: JW) from the National Cancer Institute and the Office of the Director, National Institutes of Health. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. CONSIT TEAM: Funds from Italian citizens who allocated the 5x1000 share of their tax payment in support of the Fondazione IRCCS Istituto Nazionale Tumori, according to Italian laws (INT-Institutional strategic projects ‘5x1000’) to SM. Italian Association for Cancer Research (AIRC) to LO. CORE: The CIMBA data management and data analysis were supported by Cancer Research—United Kingdom grants C12292/A11174 and C1287/A10118. SH is supported by an NHMRC Program Grant to GCT. ACA is a Cancer Research -United Kingdom Senior Cancer Research Fellow. GCT is an NHMRC Senior Principal Research Fellow. DKFZ: The DKFZ study was supported by the DKFZ. DEMOKRITOS: This research has been co-financed by the European Union (European Social Fund—ESF) and Greek national funds through the Operational Program "Education and Lifelong Learning" of the National Strategic Reference Framework (NSRF)—Research Funding Program of the General Secretariat for Research & Technology: ARISTEIA. Investing in knowledge society through the European Social Fund. EMBRACE: EMBRACE is supported by Cancer Research United Kingdom Grants C1287/A10118 and C1287/A11990. DGE and FL are supported by an National Institute of Health Research (NIHR) grant to the Biomedical Research Centre, Manchester. The Investigators at The Institute of Cancer Research and The Royal Marsden NHS Foundation Trust are supported by an NIHR grant to the Biomedical Research Centre at The Institute of Cancer Research and The Royal Marsden NHS Foundation Trust. RE and EB are supported by Cancer Research United Kingdom Grant C5047/A8385. FCCC: The authors acknowledge support from The University of Kansas Cancer Center (P30 CA168524) and the Kansas Bioscience Authority Eminent Scholar Program. AKG was funded by 5U01CA113916, R01CA140323, and by the Chancellors Distinguished Chair in Biomedical Sciences Professorship. GC-HBOC: The German Consortium of Hereditary Breast and Ovarian Cancer (GC-HBOC) is supported by the German Cancer Aid (grant no 109076, RKS) and by the Center for Molecular Medicine Cologne (CMMC). GEMO: The study was supported by the Ligue National Contre le Cancer; the Association “Le cancer du sein, parlons-en!” Award; and the Canadian Institutes of Health Research for the "CIHR Team in Familial Risks of Breast Cancer" program. G-FAST: TVM is a postdoctoral researcher funded by the Fund for Scientific Research Flanders (FWO). GOG: This study was supported by National Cancer Institute grants to the Gynecologic Oncology Group (GOG) Administrative Office and Tissue Bank (CA 27469), the GOG Statistical and Data Center (CA 37517), and GOG's Cancer Prevention and Control Committee (CA 101165). MHG, PLM and Dr. Savage were supported by funding from the Intramural Research Program, NCI. HCSC: Was supported by a grant RD12/00369/0006 and 12/00539 from ISCIII (Spain), partially supported by European Regional Development FEDER funds. HEBCS: The HEBCS was financially supported by the Helsinki University Central Hospital Research Fund, Academy of Finland (132473), the Finnish Cancer Society and the Sigrid Juselius Foundation. HEBON: The HEBON study is supported by the Dutch Cancer Society grants NKI1998-1854, NKI2004-3088, NKI2007-3756, the Netherlands Organization of Scientific Research grant NWO 91109024, the Pink Ribbon grant 110005 and the BBMRI grant NWO 184.021.007/CP46. HUNBOCS: Hungarian Breast and Ovarian Cancer Study was supported by Hungarian Research Grant KTIA-OTKA CK-80745 and the Norwegian EEA Financial Mechanism HU0115/NA/2008-3/ÖP-9. ICO: The ICO-IDIBELL study was supported by grants from the Spanish Ministry of Health ISCIII FIS (PI10/01422, PI12/01528, and PI13/00285) and RTICC (RD12/0036/0008); the Ramón Areces (XV), Eugenio Rodríguez Pascual (2012), and Roses Contra el Cancer (2012) Foundations; the Spanish Association Against Cancer (AECC 2010); and the AGAUR Generalitat de Catalunya (2009-SGR290 and 2009-SGR293). IHCC: The IHCC was supported by Grant PBZ_KBN_122/P05/2004; Katarzyna Jaworska is a fellow of International PhD program, Postgraduate School of Molecular Medicine, Warsaw Medical University, supported by the Polish Foundation of Science. ILUH: The ILUH group was supported by the Icelandic Association “Walking for Breast Cancer Research”, by the Nordic Cancer Union and by the Landspitali University Hospital Research Fund. INHERIT: This work was supported by the Canadian Institutes of Health Research for the “CIHR Team in Familial Risks of Breast Cancer” program, the Canadian Breast Cancer Research Alliance-grant #019511 and the Ministry of Economic Development, Innovation and Export Trade—grant # PSR-SIIRI-701. IOVHBOCS: The study was supported by Ministero dell'Istruzione, dell'Università e della Ricerca and Ministero della Salute. IPOBCS: This study was in part supported by Liga Portuguesa Contra o Cancro. KCONFAB: kConFab is supported by grants from the National Breast Cancer Foundation, the National Health and Medical Research Council (NHMRC) and by the Queensland Cancer Fund, the Cancer Councils of New South Wales, Victoria, Tasmania and South Australia, and the Cancer Foundation of Western Australia. GCT and ABS are NHMRC Senior Research Fellows. MAYO: MAYO is supported by National Institutes of Health grant CA128978, an NCI Specialized Program of Research Excellence (SPORE) in Breast Cancer (CA116201), a United States Department of Defence Ovarian Cancer Idea award (W81XWH-10-1-0341) and grants from the Breast Cancer Research Foundation. MCGILL: Jewish General Hospital Weekend to End Breast Cancer, Quebec Ministry of Economic Development, Innovation and Export Trade. MODSQUAD: The work was supported by the European Regional Development Fund and the State Budget of the Czech Republic (RECAMO, CZ.1.05/2.1.00/03.0101) and MH CZ—DRO (MMCI, 00209805). MSKCC: MSKCC is supported by Breast Cancer Research Foundation, the Niehaus Family Genetics Research Fund, and the STARR Cancer Consortium Grants. NAROD: 1R01 CA149429-01. NCI: The research of MHG and PLM was supported by the NCI Intramural Research Program, National Institutes of Health, and by support services contracts NO2-CP-11019-50 and N02-CP-65504 with Westat, Inc, Rockville, Maryland. This study was supported by National Cancer Institute grants to the Gynecologic Oncology Group (GOG) Administrative Office and Tissue Bank (CA 27469), the GOG Statistical and Data Center (CA 37517), and GOG's Cancer Prevention and Control Committee (CA 101165). NICCC: NICCC is supported by Clalit Health Services in Israel. Some of it's activities are supported by the Israel Cancer Association and the Breast Cancer Research Foundation (BCRF), New York. NNPIO: This work has been supported by the Russian Federation for Basic Research (grants 11-04-00227, 12-04-00928 and 12-04-01490) and the Federal Agency for Science and Innovations, Russia (contract 02.740.11.0780). OCGN: This work was supported by the Canadian Institutes of Health Research for the “CIHR Team in Familial Risks of Breast Cancer” program and grant UM1 CA164920 from the National Cancer Institute. The content of this manuscript does not necessarily reflect the views or policies of the National Cancer Institute or any of the collaborating centers in the Breast Cancer Family Registry (BCFR), nor does mention of trade names, commercial products, or organizations imply endorsement by the United States Government or the BCFR. OSU CCG: OSUCCG is supported by the Ohio State University Comprehensive Cancer Center. SMC: This project was partially funded through a grant by the Isreal cancer association and the funding for the Israeli Inherited breast cancer consortium. SWE-BRCA: SWE-BRCA collaborators are supported by the Swedish Cancer Society. UCHICAGO: UCHICAGO is supported by NCI Specialized Program of Research Excellence (SPORE) in Breast Cancer (CA125183), R01 CA142996, 1U01CA161032 and by the Ralph and Marion Falk Medical Research Trust, the Entertainment Industry Fund National Women's Cancer Research Alliance and the Breast Cancer research Foundation. OIO is an ACS Clinical Research Professor. UCSF: UCSF Cancer Risk Program and Helen Diller Family Comprehensive Cancer Center. UPENN: National Institutes of Health (NIH) (R01-CA102776 and R01-CA083855; Breast Cancer Research Foundation; Rooney Family Foundation; Susan G. Komen Foundation for the cure, Basser Research Center for BRCA. WCP: The Women's Cancer Program (WCP) at the Samuel Oschin Comprehensive Cancer Institute is funded by the American Cancer Society Early Detection Professorship (SIOP-06-258-01-COUN). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: PP is an author of this study and is a member of the PLOS ONE Editorial Board; this does not alter the authors' adherence to PLOS ONE Editorial policies and criteria.
Introduction
An integrative genomics study generated a breast cancer network model that predicted novel genetic and molecular relationships for breast cancer tumor suppressors [1]. Among the predictions, the product of the hyaluronan-mediated motility receptor (HMMR) gene, RHAMM, was found to be biochemically and functionally linked to the breast cancer gene, early onset 1 gene product (BRCA1) [1]. Analysis of common genetic variation in HMMR suggested an association with breast cancer risk in Ashkenazi Jewish women, with a greater increased risk in younger individuals [1]. However, this association was not observed either in a European case-control study [2] or in a genome-wide association study in postmenopausal women of European ancestry [3].
Following the initial functional evidence, a molecular mechanism involving RHAMM and BRCA1 was found to regulate mammary epithelial apicobasal polarization and, possibly, differentiation [4]. The results from this study indicated that RHAMM and BRCA1 play a central role in the cytoskeletal reorganization necessary for epithelial polarization. This functional interplay included interactions with the product of the proto-oncogene aurora kinase A (AURKA) and its major regulator, targeting protein for Xklp2 (TPX2), in addition to g-tubulin (TUBG1) [4]. In this scenario, cell proliferation is endorsed by activated AURKA while polarization and differentiation are mediated by activation of BRCA1 and degradation of RHAMM. Intriguingly, the same HMMR variation as originally detected in the Ashkenazi Jewish population was suggested to be associated with breast cancer risk in BRCA1, but not in BRCA2 mutation carriers [4]. This observation was endorsed by complementary analyses in breast cancer tissue; specifically, loss of cell polarity was revealed in in situ breast tissue lesions of BRCA1 mutation carriers and, accordingly, increased staining of phospho-T703-RHAMM (target of AURKA) was preferentially detected in estrogen receptor α (ERα)-negative and BRCA1-mutated tumors [4].
While the HMMR association study in BRCA1/2 mutation carriers drew on a partial dataset from the Consortium of Investigators of Modifiers of BRCA1/2 (CIMBA), the depicted mechanistic model highlighted additional gene candidates for breast cancer risk; i.e., AURKA, TPX2, and TUBG1 [4]. In a previous CIMBA study, no evidence of association was found between functional variation in AURKA and breast cancer risk among BRCA1/2 mutation carriers [5]. However, these results were based on a more limited CIMBA dataset (4,935 BRCA1 and 2,241 BRCA2 mutation carriers) and did not comprehensively assess variation in the AURKA genomic region. Furthermore, variation in TUBG1 was found to be associated with breast cancer risk in a hospital-based case-control study [6], but has not been assessed in BRCA1/2 mutation carriers.
In addition to multiplicative allele effects, systematic analyses in model organisms have shown that a given phenotype may be substantially determined by genetic interactions (GxG); that is, “epistasis” in statistical terms, defined as deviation from additivity for a quantitative phenotype arising from the effect of genetic variants or mutations in another locus [7]. Importantly, GxG significantly overlap with other types of gene and/or protein relationships [8–10]. Therefore, the functional interplay between the aforementioned genes/proteins in a key mammary epithelial cell process could support the existence of genetic interactions that influence cancer risk.
In the present study, given previous evidence of 1) the functional interplay between BRCA1 and AURKA-RHAMM-TPX2-TUBG1 in mammary epithelial polarization [4], and 2) the potential modification of breast cancer risk in BRCA1 mutation carriers by common genetic variation in HMMR [4], we further assessed the association between variants in AURKA-HMMR-TPX2-TUBG1 and breast cancer risk in BRCA1 and BRCA2 mutation carriers. Genotyped variants from the custom Illumina iSelect array of the Collaborative Oncological Gene-environment Study (iCOGS) were analyzed in a large series of BRCA1/2 mutation carriers [11,12].
Materials and Methods
Study Subjects and Ethics Statement
BRCA1/2 mutation carriers were recruited under the CIMBA initiative following approval of the corresponding protocol by the institutional review board or ethics committee at each participating center, and written informed consent was obtained from the patients when required [11,12]. Sixty CIMBA study centers recruited 15,252 BRCA1 and 8,211 BRCA2 mutation carriers that passed quality control assessment in this study. Most of these individuals were recruited through cancer genetics clinics and enrolled into national or regional studies. The remaining carriers were identified by population-based sampling or community recruitment. Eligibility in CIMBA was restricted to female carriers of pathogenic BRCA1 and BRCA2 mutations who were ≥ 18 years old at recruitment. Information collected included year of birth, mutation description, self-reported ethnicity, age at last follow-up, ages at breast or ovarian cancer diagnosis, and age at bilateral prophylactic mastectomy or oophorectomy. Information regarding tumor characteristics, including ERα status, was collected for 3,458 BRCA1 and 1,924 BRCA2 mutation carriers. Related individuals were identified by a unique family identifier.
iCOGS Design
The iCOGS array, genotyping and quality controls for the CIMBA BRCA1/2 mutation carrier samples have been described recently [11,12]. The final array design included 211,155 manufactured SNPs that were selected on the basis of primary evidence from genome-wide association studies (GWASs) of breast, ovarian and prostate cancer, for fine mapping of known cancer susceptibility loci, and included functional candidate variants of interest [11–15] (also see http://www.nature.com/icogs/primer/cogs-project-and-design-of-the-icogs-array/ and http://ccge.medschl.cam.ac.uk/research/consortia/icogs/). Details of the iCOGS array design have been described elsewhere [11–15]. The genotype data used in this study are available upon request from the CIMBA Data Access Coordinating Committee (contact A.C.A.).
Association Study
Based on previous GWAS results for BRCA1 [16] and BRCA2 [17] mutation carriers, and on the selection of gene candidates, the iCOGS array included SNPs in the AURKA (n = 15), HMMR (n = 14), TPX2 (n = 3) and TUBG1 (n = 4) loci (defined as ± 20 kilobases (kb) from the genomic structure of each gene), and these were analyzed in the present study (S1 Table). In addition, we analyzed five SNPs proximal to the HMMR locus that provided some suggestion of association with breast cancer risk in Ashkenazi Jewish women [1] (S1 Table). In total, these SNPs represented 32 partially independent variants (pairwise r2 < 0.85). To account for multiple testing, we used a FDR approach for the 41 genotyped SNPs that were evaluated for their associations with breast cancer risk in BRCA1 and BRCA2 mutation carriers; significant results are reported for FDR < 5%. The main analyses focused on evaluating associations between each genotype and breast cancer or ovarian cancer risk separately, in a survival analysis framework. In the breast cancer analysis, the phenotype of each individual was defined by age at breast cancer diagnosis or age at last follow-up. Individuals were followed until the age of the first breast or ovarian cancer diagnosis or bilateral prophylactic mastectomy, whichever occurred first, or until age at last observation. Mutation carriers censored at ovarian cancer diagnosis were considered to be unaffected. For the ovarian cancer analysis, the primary endpoint was the age at ovarian cancer diagnosis, and mutation carriers were followed until the age of ovarian cancer diagnosis or risk-reducing salpingo-oophorectomy, or until age at last observation. In order to maximize the number of ovarian cancer cases, breast cancer was not considered to be a censoring event in this analysis, and mutation carriers who developed ovarian cancer after breast cancer diagnosis were considered affected in the ovarian cancer analysis. To adjust for the non-random sampling of mutation carriers with respect to their disease status, data were analyzed by modeling the retrospective likelihood of the observed genotypes conditional on the disease phenotypes [18]. The associations were assessed using the 1-degree of freedom score test statistic based on this retrospective likelihood. To allow for the non-independence among related individuals, the correlation between the genotypes was taken into account using a kinship-adjusted version of the score test statistic [16]. The p values presented were based on the adjusted score test. To estimate the HRs, the effect of each SNP was modeled as either a per-allele or genotype on the log-scale by maximizing the retrospective likelihood. The evidence of heterogeneity in the associations between countries/study-centers was also evaluated. Associations with breast and ovarian cancer risks were assessed simultaneously within a competing risk analysis framework [11,18]. The significant FDR-adjusted associations (for rs299290 in HMMR and for rs2426618 in AURKA/CSTF1) were subsequently explored using imputed genotypes based on data from the 1,000 Genomes project (March 2012 version [19]). The IMPUTE2 software [20] was used for imputation of non-genotyped SNPs. Associations of each marker with cancer risk were assessed using a similar score test to that used for the genotyped SNPs, but based on the posterior genotype probabilities at each imputed marker for each individual. In all analyses, only those SNPs with an imputation information/accuracy of r2 > 0.30 and a minor allele frequency (MAF) > 0.3 were considered. The haplotypes and their posterior probabilities were estimated using the expectation-maximization algorithm [21]. Only the four HMMR haplotypes with the highest probabilities were considered; the rest were grouped into a single rare haplotype. Each carrier was assigned the most likely haplotypes and the association between haplotypes and age at breast cancer diagnosis was evaluated using a standard Cox proportional hazards model. All possible pairwise gene interactions including rs299290 or rs2426618 were evaluated using a standard Cox proportional hazards model that considered the main effects and the interaction term.
Expression Analysis
The association between gene expression and survival after breast cancer diagnosis was assessed using the NKI-295 dataset of sporadic primary breast tumors [22,23] and a standard Cox proportional hazards model. All possible pairwise gene interactions including HMMR or AURKA microarray probes (2 and 1 probes, respectively) were evaluated using this model. The quantitative analysis of HMMR expression isoforms was carried out using mRNA extracted from lymphoblastoid cell lines of nine rs299290-TT and six rs299290-CC BRCA2 mutation carriers, and the following TaqMan (Applied Biosystems) probes in real-time PCR assays: hs01063269 for total HMMR expression; hs0106328 for the inclusion of exon 4; and hs00234864 for the inclusion of exon 11.
Genome Analyses
Data for formaldehyde-assisted isolation of regulatory elements (FAIREs) that marked transcriptionally active regions in normal human mammary epithelial cells (HMECs) were downloaded from the Gene Expression Omnibus (GEO) reference GSE46074 [24]. Sequence reads were trimmed for the adaptor, masked for low-complexity and low-quality sequences/reads and subsequently aligned to the genome version hg19 using TopHat [25] with default parameters. Peaks were called using HOMER [26], applying a triangle-based distribution, a median length of 150 base pairs, and an α value of 0.01 (99.0% CI). Replicates were analyzed individually and uniquely merged using BEDTools [27]. Chromatin immunoprecipitation data of ERα were downloaded from the GEO reference GSE32222 and analyzed with MACS (version 2.0.9; macs2diff function) [28]. Significance was defined as a false discovery rate < 1%, using default values for all other parameters. Differentially bound genomic regions were assigned to the closest ENSEMBL (version 62) annotated gene using the R-Bioconductor package ChIPpeakAnno [29]. Histone modification and chromatin segmentation data in HMECs were obtained from the UCSC Genome Browser (hg19) and correspond to the GEO references GSE29611 and GSE38163, respectively, deposited by the ENCODE project [30].
Evolutionary analysis of BRCA1 and RHAMM
The full-length nucleotide and protein sequences from 20 (for BRCA1/BRCA1) and 26 (for HMMR/RHAMM) mammalian species, which included human and naked mole rat, were downloaded from the OrthoMaM 2.0 database [31] (S2 Table). For evolutionary analysis, a multiple sequence alignment (MSA) of the corresponding amino acid sequences was generated using the algorithm implemented in PRANK v.140110 [32]. To prevent the inclusion of incorrectly aligned positions, all MSA positions with low statistical support (posterior probabilities < 0.99) in the PRANK alignment were excluded. Next, the high-quality protein MSAs were used as guide in the alignment of the corresponding coding sequences (CDS MSAs). The level of functional constraints acting on the coding regions of both genes was analyzed using the maximum likelihood method implemented in the codeml program of PAML v.4 [33]; this approach allows to estimate the non-synonymous (dN) to synonymous (dS) ratio (ω) in a particular coding region by using a codon-based evolutionary model under a phylogenetic framework, allowing comparison of their fit to the data by the likelihood ratio test (LRT). In particular, the goodness-of-fit of two nested evolutionary models was compared: the M7 model, which assumes a β distribution of ω across sites between 0 and 1 (0 ≤ ω ≤ 1); and the M8 model, which adds to M7 an extra category of positively selected sites (ω > 1). To reduce the probability of false positive results from the M7-M8 comparison, we also estimated the likelihood of the data under the model M8a [34], in which ω was set to 1. The posterior probabilities for each site of belonging to the positively selected class were computed using the Bayes empirical Bayes approach in codeml [35]. In all models, the topology of the mammalian phylogenetic tree assumed in OrthoMaM database was used (S2 Table).
Results
HMMR rs299290 Association
The product of the HMMR gene, RHAMM, interacts with BRCA1 in the control of mammary epithelial polarization and this function may be at the basis of a modification of breast cancer risk in BRCA1 mutation carriers [4]. In this iCOGS BRCA1/2 study, 14 SNPs (11 with pairwise r2 < 0.85) at the HMMR locus were genotyped in 15,252 BRCA1 and 8,211 BRCA2 mutation carriers from 60 participating centers. Among these variants, the strongest evidence of association with breast cancer risk in BRCA1 mutation carriers was observed for the originally reported SNP rs299290 [4] (MAF = 0.25): BRCA1 per-allele HR = 1.10, 95% CI 1.04–1.15, p = 1.9 x 10−4 (FDR-adjusted p = 0.043, accounting for 41 genotyped SNPs used in the association analyses in BRCA1 and BRCA2 mutation carriers). In contrast, no evidence of association was obtained between HMMR variation and breast cancer risk in BRCA2 mutation carriers; specifically, rs299290 per-allele HR = 0.98, 95% CI 0.92–1.05, p = 0.57. The effect among BRCA1 mutation carriers was consistent across most participating countries (Fig. 1A) and no heterogeneity was detected (pheterogeneity ≥ 0.30). Importantly, the BRCA1 association remained after excluding the centers participating in the original study [4]: n = 5,039, rs299290 per-allele HR = 1.13, 95% CI 1.04–1.22, p = 0.005.
(A) Forest plots showing rs299290 HRs and 95% CIs (retrospective likelihood trend estimation) for participating countries (relatively small sample sets are not shown) ordered by sample size. Left and right panels show results for BRCA1 and BRCA2 mutation carriers, respectively. The sizes of the rectangles are proportional to the corresponding country/study precision. (B) The rs299290-containing region, including the genes, variation and regulatory evidence mentioned in HMECs. Exons are marked by black-filled rectangles and the direction of transcription is marked by arrows in the genomic structure. The chromosome 5 positions (base pairs (bp)) and linkage disequilibrium structure from Caucasian HapMap individuals are also shown.
Although the original Ashkenazi Jewish population study suggested associations involving SNPs proximal to HMMR (∼450 kb proximal) [1], no evidence of association in BRCA1/2 mutation carriers was obtained for five correlated variants in this region (S1 Table). With respect to rs299290 and ovarian cancer risk, no evidence of association was found under the single disease risk model or the competing risks model (p > 0.65; only breast cancer risk in BRCA1 mutation carriers was significant in this model: p = 2.5 x 10−4). Together, these results corroborate the association between variation at the HMMR locus and breast cancer risk in BRCA1 mutation carriers.
Mapping the HMMR locus association
Allelic imputation within ∼60 kb centered on HMMR (fully including the proximal genes CCNG1 and NUDCD2) did not detect substantially stronger associations than those identified for rs299290: a variant located in HMMR intron 7 (rs61292050; Fig. 1B) was found to be similarly associated (p = 2.7 x 10−4), but this was correlated with rs299290 (r2 = 0.95). Haplotype analyses were then carried out to demarcate the HMMR genomic region potentially harboring a causative variant or mutation. Using the 14 SNPs genotyped in iCOGS, two haplotypes, both characterized by the minor allele of rs299290, were found to be associated with breast cancer risk in BRCA1 mutation carriers (S3 Table). Based on these haplotypes, the minimal region harboring a mutation could be delimited to ∼28 kb between rs299284 and rs10038157 (S3 Table).
Analysis of the potential causative variant
The rs299290 variant represents a missense amino-acid change in HMMR exon 11 that is predicted to be benign/neutral/tolerated by several algorithms: Valine 369 to Alanine in accession number NP_001136028.1; MutationAssessor score = 0; Polyphen score = 0.005; and SIFT score = 0.73. Subsequent examination of the splicing of exon 11 and of exon 4, the latter of which is known to be differentially spliced in different conditions and cell types [36], did not reveal alterations or differences between mRNA samples with different rs299290 genotypes (S1 Fig.). Nonetheless, rs299290 is located ∼14 kb from the HMMR promoter region that is active in mammary epithelial cells, as detected by the analysis of data from genome occupancy profiling [24] (Fig. 1B). In addition, analysis of data for ERα binding plasticity [37] revealed significant binding of this factor at the HMMR promoter in poor-prognosis breast tumors (Fig. 1B).
Causal alleles for different common diseases have shown evidence of positive selection [38]. Notably, a recent report suggested the action of positive selection on the evolution of BRCA1 and RHAMM orthologs in the naked mole rat, which is an exceptionally cancer-resistant species [39]. Following on from this suggestion, we identified footprints of positive selection in the evolution of some amino acids of both proteins. In both cases, model M8 (selection model) better fits the protein alignment data than model M7 (null model): p values = 6.69 x 10−14 and 1.55 x 10−5, for BRCA1 and RHAMM, respectively. Moreover, the likelihood of the data is significantly higher under model M8 than under the nested model M8a: p values = 6.22 x 10−10 and 0.014 for BRCA1 and RHAMM, respectively, confirming the presence of positively selected sites in these alignments. However, Valine 369 RHAMM was not identified in these analyses; the predicted amino acid sites under selection were only linked to rare variants (MAFs < 0.01) (S4 Table). Nonetheless, as suggested in the analysis of the naked mole rat sequence, Valine 369 is within a region with a potential excess of selected positions (Fig. 2A). Analysis of BRCA1 also showed multiple potential selection sites (Fig. 2B), but the specific region or domain mediating the interaction with RHAMM remains unknown [1,4,40].
(A) Plot showing the position of potentially selected sites (p (w > 1)) in the amino acid sequence of RHAMM. The relative position of known protein domains is shown. (B) Plot showing the position of potentially selected sites (p (w > 1)) in the amino acid sequence of BRCA1. The relative position of known protein domains is shown.
Evaluation of HMMR association by ERα tumor status and BRCA1 mutation class
The original study suggested an association between the rs299290 risk allele and ERα-negative breast cancer for BRCA1 mutation carriers [4]. In the present study, no difference was found in the rs299290 effect between ERα-negative and ERα-positive cases: ERα-negative, per-allele HR = 1.09, 95% CI 1.03–1.15, p = 2.3 x 10−3; ERα-positive, per-allele HR = 1.09, 95% CI 0.98–1.21, p = 0.13; pdifference = 0.96. Interestingly, there was a suggestion of an rs299290 association with ERα-negative breast cancer for BRCA2 mutation carriers, but in the opposite direction to that observed for BRCA1 mutation carriers: ERα-negative BRCA2 mutation carriers n = 434, per-allele HR = 0.83, 95% CI 0.70–0.97, p = 0.022. In addition, there was no evidence of an rs299290 association with ERα-positive breast cancer in BRCA2 mutation carriers (n = 1,490, p = 0.40, ERα-negative effect pdifference = 0.019).
Regarding BRCA1 mutation classes, the original study [4] suggested a rs299290 association in carriers of mutations expected to result in a reduced transcript or protein level due to nonsense-mediated RNA decay (Class 1), but not in carriers of mutations likely to generate stable proteins with a potential residual or dominant-negative function (Class 2). The current study indicated a similar association, although the estimations were not significantly different: Class 1, rs299290 per-allele HR = 1.11, 95% CI 1.05–1.18, p = 4.6 x 10−4; Class 2, rs299290 per-allele HR = 1.03, 95% CI 0.94–1.14, p = 0.51. Regarding Ashkenazi Jewish BRCA1 mutation carriers (n = 1,231), there were no significant associations in this population or with founder mutations (185delAG HR = 0.94, p = 0.17; and 5382insC HR = 0.85, p = 0.10). Larger sample series may be required to assess associations in these settings and their consistency with previous observations in the Ashkenazi Jewish population [1].
AURKA/CSTF1 association with breast cancer risk in BRCA2 mutation carriers
As the HMMR and AURKA-TPX2-TUBG1 gene products are functionally related in the regulation of mammary epithelial polarization [4], the associations between variants at these loci that were included on the iCOGS array and cancer risk in BRCA1/2 mutation carriers were assessed (S1 Table). No associations were observed for TPX2 and TUBG1, but there was an indication of an association for a variant relatively close to AURKA; rs2426618 and breast cancer risk in BRCA2 mutation carriers, per-allele HR = 1.10, 95% CI 1.03–1.16, p = 0.005 (FDR-adjusted p = 0.045; Fig. 3A). There was no evidence of association between this variant and breast cancer risk in BRCA1 mutation carriers or between the same variant and ovarian cancer risk in either BRCA1 or BRCA2 mutation carriers (p > 0.30). Consistent with the main analysis, the competing risk model showed an rs2426618 association with breast cancer risk only in BRCA2 mutation carriers (p = 0.002).
(A) Forest plots showing rs2426618 HRs and 95% CIs (retrospective likelihood trend estimation) for participating countries (relatively small sample sets are not shown) ordered by sample size. Left and right panels show results for BRCA1 and BRCA2 mutation carriers, respectively. The sizes of the rectangles are proportional to the corresponding study precision. (B) The rs2426618-containing region, including the genes, variation and regulatory evidence in HMECs. Exons are marked by black-filled rectangles and the direction of transcription is marked by arrows in the genomic structure. The chromosome 20 positions (bp) and linkage disequilibrium structure from Caucasian HapMap individuals are also shown.
The rs2426618 variant is located in intron 5 of a neighboring (distal) gene, CSTF1 (Fig. 3B). Analogous to the HMMR setting, rs2426618 is relatively close (∼30 kb) to the promoter region of AURKA/CSTF1, which is active in mammary epithelial cells and differentially ERα-regulated in poor-prognosis breast tumors (Fig. 3B). Association analysis with imputed SNPs within ∼50 kb centered on rs2426618 (fully including AURKA and CSTF1) did not reveal stronger evidence than that of rs2426618: a variant located in AURKA intron 8 (rs187966577, Fig. 3B) was found to be similarly associated (p = 0.002). However, this variant was rare (MAF = 0.009) and poorly imputed (r2 = 0.37). Finally, the analysis of rs2426618 by ERα tumor status in BRCA2 mutation carriers did not reveal specific associations: ERα-positive, per-allele HR = 1.10, 95% CI 1.02–1.17, p = 0.010; ERα-negative, per-allele HR = 1.11,95% CI 0.97–1.26, p = 0.15, pdifference = 0.89.
Potential genetic interactions in the AURKA-HMMR-TPX2-TUBG1 module
Given the above associations at HMMR and AURKA/CSTF1, the influence of GxG on breast cancer risk in BRCA1/2 mutation carriers was assessed between rs299290 and the genotyped variants in AURKA/TPX2/TUBG1 (n = 22), and between rs2426618 and the genotyped variants in TPX2/TUBG1 (n = 7). No GxG was detected for rs2426618-TPX2/TUBG1, but potential interactions (unadjusted pinteraction values < 0.05) between rs299290 and AURKA/CSTF1 or TUBG1 variants were suggested (Table 1). An interaction between rs299290 and rs6064391 (in intron 2 of CSTF1, Fig. 3B) could reduce breast cancer risk in both BRCA1 and BRCA2 mutation carriers (interaction HRs = 0.87 and 0.73, respectively; Table 1). In the main effect analysis, rs6064391 showed a suggestion of association with reduced breast cancer risk in BRCA2 mutation carriers: HR = 0.87, unadjusted p = 0.036. Conversely, an interaction between rs299290 and rs11649877 (3’ region of TUBG1) could increase breast cancer risk in both BRCA1 and BRCA2 mutation carriers (HRs = 1.33 and 1.21, respectively; Table 1). In the main effect analysis, there was no evidence of rs11649877 association with breast cancer risk in BRCA1 or BRCA2 mutation carriers (p = 0.49). Genetic interactions frequently represent complex molecular relationships and, thus, can be explained by multiple models of phenotypic differences across genetic backgrounds [41], including those identified here. However, additional studies are required to corroborate the findings.
Integrative studies in model organisms have shown that experimentally identified GxG overlap with other types of gene and/or protein interactions (e.g., gene co-expression) to a degree that is significantly higher than expected by chance [8]. Thus, we assessed the GxGs suggested above for their equivalence with interactions between gene expression profiles across breast tumors, considering breast cancer survival as the outcome. There are no large tumor series from BRCA1/2 mutation carriers that allow gene expression interactions in these mutation backgrounds to be assessed; therefore our analysis was restricted to sporadic cases and aimed to explore complex relationships between HMMR and AURKA or TUBG1. Using the NKI-295 dataset [22], HMMR expression was found to be significantly associated with overall survival: two microarray probes gave identical results: HR = 1.78 and p = 0.001. AURKA expression was also found to be associated with overall survival (single probe HR = 2.07, p = 3.3 x 10−9) but, notably, the combined model with HMMR indicated a protective interaction: the HRs for both HMMR and AURKA probe combinations were 0.57 and 0.55, pinteraction values = 0.026, respectively (Fig. 4A). The combined HMMR and TUBG1 expressions also suggested an interaction, but in this case “aggravating”: the HRs for both HMMR probe combinations with a TUBG1 probe were 1.84 and 2.16, pinteraction = 0.041 and 0.009, respectively (Fig. 4B). The expression of TUBG1 alone was not found to be associated with overall survival (p = 0.12). Together, these results could indicate a functional correspondence for interactions influencing both breast cancer development and progression. However, the precise effects of the corresponding risk alleles remain to be elucidated.
(A) Kaplan–Meier survival curves based on categorization of HMMR (probe NM_012484) and AURKA (NM_003600) expression in tertiles (low, medium or high expression). For simplicity, only the tertiles for “high” AURKA are shown. The tumours with high expression levels for both genes were not those with the poorest prognosis. (B) Kaplan–Meier survival curves based on categorization of HMMR (NM_012484) and TUBG1 (NM_016437) expression in tertiles (low, medium or high expression). For simplicity, only the tertiles for “high” HMMR are shown. The cases with high expression levels for both genes were those with the poorest prognosis.
Discussion
The results of this study expand on the previous suggestion that variation in HMMR is specifically associated with breast cancer risk in BRCA1 mutation carriers [4]. By analyzing a much larger number of carriers—genotyped using the high-quality iCOGS approach [11,12]—we were able to confirm the association of this locus, and to demonstrate that the strongest signal corresponded to rs299290. No specific effect was revealed when analyzing this association by ERα tumor status. Larger carrier series may be required to comprehensively evaluate associations by tumor subtype or, conversely, to establish that variation in HMMR interacts with Class I BRCA1 mutations to give rise to any type of tumor. In this context, mapping of the protein and/or functional domains interacting between BRCA1 and RHAMM would be necessary to decipher the role of rs299290, if any, and of the observed potential positive selection. With regard to the results obtained for BRCA2 mutation carriers, the opposite direction of the potential association with ERα-negative breast cancer compared with the association observed for BRCA1 mutation carriers is intriguing. Opposite effects between cancer subtypes have been observed for other modifier loci [16] and, thus, potentially inform on opposing functional roles in biological processes influencing carcinogenesis. In other words, alteration of RHAMM function in mammary epithelial differentiation may have a differential effect on breast cancer risk depending on whether it occurs in a BRCA1- or BRCA2-mutated background.
The study of gene loci functionally related to HMMR suggests an association between variation in AURKA/CSTF1 and risk of breast cancer among BRCA2 mutation carriers. While there were initial conflicting results for AURKA associations with breast cancer risk [5,42], it is interesting to note that a population case-control study [43] identified variants in the AURKA promoter region associated with breast cancer risk in the same direction as detected in our analysis. The minor allele of rs6064389 was shown to be protective in the general population [43], and a similar association was observed in the analysis of BRCA2 mutation carriers: HR = 0.93, p = 0.021; not significant in BRCA1 mutation carriers, p = 0.25. This variant is partially correlated with rs2426618 (r2 = 0.63), and more strongly correlated with rs6064391 (r2 = 0.86), which might interact with rs299290.
In addition to the main effects, interaction between HMMR and AURKA, and between HMMR and TUBG1 could influence breast cancer risk in BRCA1 and BRCA2 mutation carriers. The interactions did not remain statistically significant after corrections for multiple testing and, therefore, additional studies are warranted to investigate these findings further. Combined expression of HMMR and AURKA, and HMMR and TUBG1 in sporadic breast tumors appeared to influence patients’ survival differentially. These results could be analogous to those observed for breast cancer risk, but it remains unknown how the corresponding risk alleles may alter gene expression and/or protein function. In this context, over-expression of HMMR, AURKA or TUBG1 impairs mammary epithelial polarization [4] and the rs299290 risk allele could be associated with relatively higher levels of HMMR expression [1,4], which would be consistent with the observed interactions. However, other studies on the potential regulatory impact of this variant and/or the link to selective constraints are needed. It is also important to note that the altered gene products may not be those included in the depicted functional module, but rather other products from the corresponding chromosomal regions.
In summary, centered on the AURKA-HMMR-TPX2-TUBG1 functional module that regulates mammary epithelial polarization, this study confirms previous association results for HMMR rs299290 and suggests novel associations (for AURKA/CSTF1 and HMMR-rs299290 interactions) with breast cancer risk in BRCA1 and/or BRCA2 mutation carriers.
Supporting Information
S1 Fig. Results of the quantification of the expression of HMMR exons 4 and 11 according to the rs299290 major and minor genotypes.
https://doi.org/10.1371/journal.pone.0120020.s001
(TIF)
S1 Table. Genotyped variants and breast cancer association results in BRCA1 and BRCA2 mutation carriers.
https://doi.org/10.1371/journal.pone.0120020.s002
(XLSX)
S2 Table. Species and genes used in functional constraints analyses.
https://doi.org/10.1371/journal.pone.0120020.s003
(XLSX)
S3 Table. Variation in HMMR and haplotype association study for risk of breast cancer in BRCA1 mutation carriers.
https://doi.org/10.1371/journal.pone.0120020.s004
(PDF)
S4 Table. Evolutionary selection tests and results.
https://doi.org/10.1371/journal.pone.0120020.s005
(XLS)
S5 Table. Details of acknowledgments and funding support.
https://doi.org/10.1371/journal.pone.0120020.s006
(XLSX)
Acknowledgments
This work is dedicated to the memory of Olga M. Sinilnikova, who inspired the CIMBA group and led studies on BRCA1 modifiers. We wish to thank all study participants and clinicians for their valuable contributions to the iCOGS study. We also wish to thank the individual contributors from the following groups: BCFR, EMBRACE, GEMO, HEBON, kConFab, and SWE-BRCA (listed in S5 Table). Additional acknowledgments are also detailed in S5 Table.
Author Contributions
Conceived and designed the experiments: IB K. Kuchenbaecker KO GC-T FJC ACA MAP. Performed the experiments: K. Kuchenbaecker DC GRG NB ACA MAP FM. Analyzed the data: K. Kuchenbaecker DC XW DB GRG NB LM AI A. Berenguer A. Petit IC A. Gómez SS HH EV CG KO GC-T FJC ACA MAP. Contributed reagents/materials/analysis tools: IB K. Kuchenbaecker DC XW DB GRG PL AS-G JR NB LM VSP AI A. Bojesen A. Petit IC ME J. Brunet LF ET J. Benítez AO TRC HN KA BKA AET BYK CW J. Lester MHG PLM RLN ILA SMD KLN TRR RBB AJ J. Lubinski KD KJ-B KC TVM OD TVH LJ A-MG BE MH TC EMBRACE AMD CO EF M. Cook SP EM A. Murray CJ GP FL CC HD JP K-RO MRT HEBON FBLH AHH C. Sutter MGEMA MJLL PD JTW MAR HEJM-H MJBl AMWO CMA GCR K-AAP MP SRN VLB-J DMO ESR RKS BW KR CE A. Meindl ND NA HJP DN C. Singer SW-G DS SP-A K. Kast RV-M A. Gehrig A. Berger ISP LS UBJ MT TAK LF PP LB BP GS S. Mazoyer PR LO MM SA CM JS PS AB AF-R CFS CR DG-K M-KT GP BCFR EMJ A. Miron SLN MBT WKC MBD DEG RJ CMD EJR FF IK JG AKG SWE-BRCA AL JR MS-A NL ENI UH A. Gómez SS HH EV CG kConFab ABS SH JNW JH DM GEMO OMS DS-L ML S. Manoukian FD LB MF-C A. de Pauw CL SFF LC PB FC Y-JB EO SAN GR SSP YL EF CAM JV MR NK M. Corines DV JC AL NL CL DFE KO GC-T FJC ACA MAP MAD RBL. Wrote the paper: IB KK ACA MAP.
References
- 1. Pujana MA, Han JD, Starita LM, Stevens KN, Tewari M, Ahn JS, et al. Network modeling links breast cancer susceptibility and centrosome dysfunction. Nat Genet. 2007;39: 1338–1349. pmid:17922014
- 2. Kalmyrzaev B, Pharoah PD, Easton DF, Ponder BA, Dunning AM Hyaluronan-mediated motility receptor gene single nucleotide polymorphisms and risk of breast cancer. Cancer Epidemiol Biomarkers Prev. 2008;17: 3618–3620. pmid:19064580
- 3. Hunter DJ, Kraft P, Jacobs KB, Cox DG, Yeager M, Hankinson SE, et al. A genome-wide association study identifies alleles in FGFR2 associated with risk of sporadic postmenopausal breast cancer. Nat Genet. 2007;39: 870–874. pmid:17529973
- 4. Maxwell CA, Benitez J, Gomez-Baldo L, Osorio A, Bonifaci N, Fernandez-Ramires R, et al. Interplay between BRCA1 and RHAMM regulates epithelial apicobasal polarization and may influence risk of breast cancer. PLoS Biol. 2011;9: e1001199. pmid:22110403
- 5. Couch FJ, Sinilnikova O, Vierkant RA, Pankratz VS, Fredericksen ZS, Stoppa-Lyonnet D, et al. AURKA F31I polymorphism and breast cancer risk in BRCA1 and BRCA2 mutation carriers: a consortium of investigators of modifiers of BRCA1/2 study. Cancer Epidemiol Biomarkers Prev. 2007;16: 1416–1421. pmid:17627006
- 6. Olson JE, Wang X, Pankratz VS, Fredericksen ZS, Vachon CM, Vierkant RA, et al. Centrosome-related genes, genetic variation, and risk of breast cancer. Breast Cancer Res Treat. 2011;125: 221–228. pmid:20508983
- 7. Hartman JLt, Garvik B, Hartwell L Principles for the buffering of genetic variation. Science. 2001;291: 1001–1004. pmid:11232561
- 8. Dixon SJ, Costanzo M, Baryshnikova A, Andrews B, Boone C Systematic mapping of genetic interaction networks. Annu Rev Genet. 2009;43: 601–625. pmid:19712041
- 9. Costanzo M, Baryshnikova A, Bellay J, Kim Y, Spear ED, Sevier CS, et al. The genetic landscape of a cell. Science. 2010;327: 425–431. pmid:20093466
- 10. Bonifaci N, Colas E, Serra-Musach J, Karbalai N, Brunet J, Gomez A, et al. Integrating gene expression and epidemiological data for the discovery of genetic interactions associated with cancer risk. Carcinogenesis. 2014;35: 578–585. pmid:24296589
- 11. Couch FJ, Wang X, McGuffog L, Lee A, Olswold C, Kuchenbaecker KB, et al. Genome-wide association study in BRCA1 mutation carriers identifies novel loci associated with breast and ovarian cancer risk. PLoS Genet. 2013;9: e1003212. pmid:23544013
- 12. Gaudet MM, Kuchenbaecker KB, Vijai J, Klein RJ, Kirchhoff T, McGuffog L, et al. Identification of a BRCA2-specific modifier locus at 6p24 related to breast cancer risk. PLoS Genet. 2013;9: e1003173. pmid:23544012
- 13. Michailidou K, Hall P, Gonzalez-Neira A, Ghoussaini M, Dennis J, Milne RL, et al. Large-scale genotyping identifies 41 new loci associated with breast cancer risk. Nat Genet. 2013;45: 353–361, 361e351–352. pmid:23535729
- 14. Eeles RA, Olama AA, Benlloch S, Saunders EJ, Leongamornlert DA, Tymrakiewicz M, et al. Identification of 23 new prostate cancer susceptibility loci using the iCOGS custom genotyping array. Nat Genet. 2013;45: 385–391, 391e381–382. pmid:23535732
- 15. Pharoah PD, Tsai YY, Ramus SJ, Phelan CM, Goode EL, Lawrenson K, et al. GWAS meta-analysis and replication identifies three new susceptibility loci for ovarian cancer. Nat Genet. 2013;45: 362–370, 370e361–362. pmid:23535730
- 16. Antoniou AC, Wang X, Fredericksen ZS, McGuffog L, Tarrell R, Sinilnikova OM, et al. A locus on 19p13 modifies risk of breast cancer in BRCA1 mutation carriers and is associated with hormone receptor-negative breast cancer in the general population. Nat Genet. 2010;42: 885–892. pmid:20852631
- 17. Gold B, Kirchhoff T, Stefanov S, Lautenberger J, Viale A, Garber J, et al. Genome-wide association study provides evidence for a breast cancer risk locus at 6q22.33. Proc Natl Acad Sci U S A. 2008;105: 4340–4345. pmid:18326623
- 18. Barnes DR, Lee A, Easton DF, Antoniou AC Evaluation of association methods for analysing modifiers of disease risk in carriers of high-risk mutations. Genet Epidemiol. 2012;36: 274–291. pmid:22714938
- 19. Abecasis GR, Auton A, Brooks LD, DePristo MA, Durbin RM, Handsaker RE, et al. An integrated map of genetic variation from 1,092 human genomes. Nature. 2012;491: 56–65. pmid:23128226
- 20. Howie B, Marchini J, Stephens M Genotype imputation with thousands of genomes. G3 (Bethesda). 2011;1: 457–470. pmid:22384356
- 21. Long JC, Williams RC, Urbanek M An E-M algorithm and testing strategy for multiple-locus haplotypes. Am J Hum Genet. 1995;56: 799–810. pmid:7887436
- 22. van de Vijver MJ, He YD, van't Veer LJ, Dai H, Hart AA, Voskuil DW, et al. A gene-expression signature as a predictor of survival in breast cancer. N Engl J Med. 2002;347: 1999–2009. pmid:12490681
- 23. Chang HY, Nuyten DS, Sneddon JB, Hastie T, Tibshirani R, Sorlie T, et al. Robustness, scalability, and integration of a wound-response gene expression signature in predicting breast cancer survival. Proc Natl Acad Sci U S A. 2005;102: 3738–3743. pmid:15701700
- 24. Rhie SK, Coetzee SG, Noushmehr H, Yan C, Kim JM, Haiman CA, et al. Comprehensive functional annotation of seventy-one breast cancer risk Loci. PLoS One. 2013;8: e63925. pmid:23717510
- 25. Trapnell C, Pachter L, Salzberg SL TopHat: discovering splice junctions with RNA-Seq. Bioinformatics. 2009;25: 1105–1111. pmid:19289445
- 26. Heinz S, Benner C, Spann N, Bertolino E, Lin YC, Laslo P, et al. Simple combinations of lineage-determining transcription factors prime cis-regulatory elements required for macrophage and B cell identities. Mol Cell. 2010;38: 576–589. pmid:20513432
- 27. Quinlan AR, Hall IM BEDTools: a flexible suite of utilities for comparing genomic features. Bioinformatics. 2010;26: 841–842. pmid:20110278
- 28. Zhang Y, Liu T, Meyer CA, Eeckhoute J, Johnson DS, Bernstein BE, et al. Model-based analysis of ChIP-Seq (MACS). Genome Biol. 2008;9: R137. pmid:18798982
- 29. Zhu LJ, Gazin C, Lawson ND, Pages H, Lin SM, Lapointe DS, et al. ChIPpeakAnno: a Bioconductor package to annotate ChIP-seq and ChIP-chip data. BMC Bioinformatics. 2010;11: 237. pmid:20459804
- 30. ENCODE An integrated encyclopedia of DNA elements in the human genome. Nature. 2012;489: 57–74. pmid:22955616
- 31. Douzery EJ, Scornavacca C, Romiguier J, Belkhir K, Galtier N, Delsuc F, et al. OrthoMaM v8: a database of orthologous exons and coding sequences for comparative genomics in mammals. Mol Biol Evol. 2014;31: 1923–1928. pmid:24723423
- 32. Loytynoja A, Goldman N Phylogeny-aware gap placement prevents errors in sequence alignment and evolutionary analysis. Science. 2008;320: 1632–1635. pmid:18566285
- 33. Yang Z PAML 4: phylogenetic analysis by maximum likelihood. Mol Biol Evol. 2007;24: 1586–1591. pmid:17483113
- 34. Swanson WJ, Nielsen R, Yang Q Pervasive adaptive evolution in mammalian fertilization proteins. Mol Biol Evol. 2003;20: 18–20. pmid:12519901
- 35. Yang Z, Wong WS, Nielsen R Bayes empirical bayes inference of amino acid sites under positive selection. Mol Biol Evol. 2005;22: 1107–1118. pmid:15689528
- 36. Sohr S, Engeland K RHAMM is differentially expressed in the cell cycle and downregulated by the tumor suppressor p53. Cell Cycle. 2008;7: 3448–3460. pmid:18971636
- 37. Ross-Innes CS, Stark R, Teschendorff AE, Holmes KA, Ali HR, Dunning MJ, et al. Differential oestrogen receptor binding is associated with clinical outcome in breast cancer. Nature. 2012;481: 389–393. pmid:22217937
- 38. Corona E, Dudley JT, Butte AJ Extreme evolutionary disparities seen in positive selection across seven complex diseases. PLoS One. 2010;5: e12236. pmid:20808933
- 39. Keane M, Craig T, Alfoldi J, Berlin AM, Johnson J, Seluanov A, et al. The Naked Mole Rat Genome Resource: facilitating analyses of cancer and longevity-related adaptations. Bioinformatics. 2014;30: 3558–3560. pmid:25172923
- 40. Hill SJ, Rolland T, Adelmant G, Xia X, Owen MS, Dricot A, et al. Systematic screening reveals a role for BRCA1 in the response to transcription-associated DNA damage. Genes Dev. 2014;28: 1957–1975. pmid:25184681
- 41. Drees BL, Thorsson V, Carter GW, Rives AW, Raymond MZ, Avila-Campillo I, et al. Derivation of genetic interaction networks from quantitative phenotype data. Genome Biol. 2005;6: R38. pmid:15833125
- 42. Fletcher O, Johnson N, Palles C, dos Santos Silva I, McCormack V, Whittaker J, et al. Inconsistent association between the STK15 F31I genetic polymorphism and breast cancer risk. J Natl Cancer Inst. 2006;98: 1014–1018. pmid:16849685
- 43. Shi H, Bevier M, Johansson R, Grzybowska E, Chen B, Eyfjord JE, et al. Single nucleotide polymorphisms in the 20q13 amplicon genes in relation to breast cancer risk and clinical outcome. Breast Cancer Res Treat. 2011;130: 905–916. pmid:21630024