Introduction

Women with breast cancer have a twofold to sixfold increased risk of developing a new primary cancer in the contralateral breast (CBC) compared with the risk of developing a first primary breast cancer among the general population [1]. Genetic factors play a critical role in CBC development, including germline pathogenic variants in BRCA1/2, TP53, CHEK2, and PALB2 [2,3,4]. Previous studies have investigated individual common variants in high- or moderate-penetrance breast cancer susceptibility genes [5] or drug metabolizing genes [6] and reported associations of breast cancer susceptibility variants identified from genome-wide association studies (GWAS) [7] with CBC risk. Studies have further demonstrated a positive cumulative effect of genetic variants, i.e., the polygenic risk score (PRS), on CBC risk, using a limited number of SNPs [8], in high-risk populations [9] or with limited adjustment for covariates [10]. However, a comprehensive GWAS assessing the associations between common variants and CBC risk has not been reported.

To advance the understanding of the genetic susceptibility of CBC for a large and growing population of breast cancer survivors, we carried out a GWAS in the Women’s Environmental Cancer and Radiation Epidemiology (WECARE) Study and evaluated the association between the updated breast cancer PRS [11] and CBC risk.

Methods

Study participants

The WECARE Study is a multi-center, population-based case–control study of CBC conducted in two phases: the WECARE I Study (2001–2004) and WECARE II Study (2009–2012) [12, 13]. Due to the word limit, we described the study design and participants in details in Additional file 1. The final analytic data set included 2829 participants (1161 cases and 1668 controls) for the main analysis and 2483 (1017 cases and 1466 controls) for the PRS analysis involving non-Hispanic White women only.

CBC GWAS analysis

The genome-wide association analysis was performed in the combined data of the WECARE I and II Studies. Details about genotyping, quality control and imputation could be found in Additional file 1. Conditional logistic regression models with adjustment for the top five principal components (PCs) and age at first breast cancer diagnosis were performed to test additive effects of genetic variants. Genome-wide statistical significance was determined by the threshold of P < 5 × 10–8 with P < 1 × 10–6 considered as suggestive significance. The functional annotation was performed using Functional Mapping and Annotation (FUMA) [14]. We applied the Sum of Single Effects (SuSiE) method to identify credible sets in each identified locus [15]. Stratified analyses were further performed, and heterogeneity was assessed using the likelihood ratio test for nested models.

PRS analysis

We constructed a weighted PRS, consisting of the 313 known breast cancer risk susceptibility SNPs [11]. Genotyping data were available at 239 of the 313 loci and proxies were determined for 18 of the remaining 74 loci. Detailed information could be found in Additional file 1. The association of the PRS with CBC risk was assessed using the continuous PRS, per standard deviation (SD) of the PRS, and PRS categorized by median, and quartile cut points based on UBC controls. Multivariable adjusted rate ratios (RR) and corresponding 95% confidence interval (CI) were estimated by fitting conditional logistic regression models adjusted for age at first breast cancer diagnosis, the top five genetic PCs, and known or suspected CBC risk factors. Area under the curve (AUC) of receiver operating characteristic (ROC) curves for various nested models were compared using the DeLong test [16].

All analyses were performed using R v4.1.3 or SAS v9.4 (The SAS Institute, Cary, NC).

Results

The quantile–quantile (Q–Q) plot is shown in Additional file 2: Fig. S1. The inflation factor of the genome-wide scan was 1.034, indicating that the population structure was not an issue for the current analysis. Two loci associated with an elevated but not statistically significant CBC risk, 9q32 (rs59657211, P = 2.96 × 10–7, SLC31A2/FAM225A) and 6p22.1 (rs3815096, P = 9.58 × 10–7, TRIM31), were identified (Fig. 1a and Additional file 2: Fig. S2). One credible set, consisting of rs10817445, rs12337704, rs59657211, and rs9632905, was identified for 9q32 using SuSiE. However, SuSiE failed to identify any credible set for 6p22.1. There was no heterogeneity in associations of CBC with rs59657211 and rs3815096 by age at first breast cancer diagnosis, first-degree family history of breast cancer, ER and PR status, and chemotherapy or radiotherapy for the first breast cancer (Fig. 1b).

Fig. 1
figure 1

a Manhattan plot of GWAS for contralateral breast cancer risk in the WECARE STUDY; b stratified analyses for the two loci with suggestive genome-wide significant associations (P < 1 × 10–6) in the WECARE Study. CI, confidence interval; ER, estrogen receptor; No., number; PR, progesterone receptor; RR, rate ratio

Full size image

Among non-Hispanic White women, the weighted PRS without proxies (239 SNPs) was associated with an increased CBC risk of 46% (RR = 1.46, 95% CI 1.25–1.71 per weighted risk allele; RR = 1.25, 95% CI 1.14–1.36 per SD estimate). PRS were evaluated both as below and above the median and by quartiles; the above the median PRS category and the highest PRS quartile were both statistically significantly associated with increased CBC risk (Table 1). ROC curves were generated and AUCs were estimated to compare the discrimination ability of two models in the combined WECARE data (Table 2): CBC risk factors alone and PRS plus risk factors. The AUC of PRS plus risk factors model was 62.4 (95% CI 60.2–64.7), which was significantly higher than the model of the risk factors alone (AUC: 60.73, 95% CI 58.5–63.0, P = 0.01). We repeated the analysis in the WECARE I Study where information regarding BRCA1/2 mutations was known. The AUC of model with PRS, risk factors, and BRCA1/2 mutations was 67.7 (95% CI 64.7–70.7), significantly higher than the model with risk factors alone (AUC: 63.0, 95% CI 60.0–66.1, P < 0.0001) and the model with PRS plus risk factors (AUC: 65.1, 95% CI 62.0–68.1, P = 0.01). The association of PRS with CBC risk was modified by chemotherapy (Pheterogeneity = 0.04) such that the association between the PRS and CBC risk was stronger among women who did not receive chemotherapy for their first primary breast cancer compared to women who had chemotherapy (data not shown in Tables). When focused on the effects of chemotherapy, our data showed a reduced CBC risk among patients with higher PRS (RR = 0.61, 95% CI 0.46–0.81), but no association among patients with lower PRS (RR = 0.90, 95% CI 0.67–1.22) (Table 3). There was no heterogeneity in the association of the PRS with CBC risk by age at first diagnosis, family history of breast cancer, radiation treatment, ER status of first breast cancer, PR status of first breast cancer, hormone replacement therapy at first breast cancer, age at menopause, or parity (Table 3). Similar findings were observed when using the PRS with proxies (257 SNPs) (Table 1).

Table 1 Associations of weighted polygenic risk score comprised of known breast cancer susceptibility SNPs with contralateral breast cancer risk
Full size table
Table 2 Areas under the receiver operating characteristic curve for contralateral breast cancer risk models
Full size table
Table 3 Associations between estrogen receptor, progesterone receptor, family history of breast cancer, age at first breast cancer diagnosis, chemotherapy, radiotherapy, hormone replacement therapy, age at menopause, and parity and risk of CBC, stratified by high/low weighted PRS (239 SNPs) in Non-Hispanic White women
Full size table

Discussion

The present study is the largest population-based GWAS analysis of CBC risk to date and identified two loci with suggestive genome-wide significance. rs59657211 at the FAM225A locus has been reported to be involved in the tumorigenesis and metastasis of several types of cancers, including nasopharyngeal, colorectal, and esophageal squamous cell cancer [17, 18]. rs3815096, an intronic variant of TRIM31, is located within the major histocompatibility complex (MHC) region. It has also been reported to be nominally associated with risk of first primary breast cancer (OR 1.02, 95% CI 1.00–1.03, P = 0.007) in a prior GWAS [19], consistent with our results. TRIM31, a member of the TRIM family and acting as an E3 ubiquitin ligase, may play a promoting or suppressing role in malignant processes of multiple cancers [20, 21]. In breast cancer, TRIM31 was found to suppress the cancer progression through the stabilization and activation of p53 [22]. Further investigation into these loci is needed to determine the underpinning mechanisms involved in CBC development.

We further confirmed an elevated risk of CBC is associated with the established breast cancer susceptibility PRS after the adjustment for other risk factors. Our findings corroborate prior studies that found a PRS consisting of the 313 breast cancer susceptibility SNPs associated with CBC risk [9, 10]. Moreover, the AUC that included the PRS and known breast cancer risk factors with or without BRCA1/2 mutations was significantly higher than that of the risk factors alone, suggesting the PRS may add additional predictive values in identifying breast cancer patients with an elevated risk of CBC. Our study also reported novel findings: i.e., chemotherapy was found to be protective among patients with higher PRS but was not among those with lower PRS. This suggests that breast cancer survivors with an unfavorable genetic background may benefit more from chemotherapy; when chemotherapy was not a viable option or the patients declined to receive the treatment, a more intense surveillance strategy might serve better for these patients for an early detection of CBC and treatment.

Our study has several strengths. Most notably, we included the largest number of CBCs reported in a GWAS study with available detailed risk factors, treatment, and clinical information. One primary limitation pertains to the generalizability across racial and ethnic groups as the WECARE Study included predominantly women of European ancestry and we lacked the statistical power to examine subgroups of interest.

In summary, our findings further the understanding of the genetic risk involved in CBC etiology, conferred by common SNPs. In turn, these results will be useful for the development of prevention strategies for CBC as well as for the long-term management of patients with breast cancer.