The genetics of sex hormones and their effects on mammographic density in women

Thesis (Ph.D.)--University of Washington, 2021In this work we leveraged genomic information from large-scale population-based studies to explore the relationships between three epidemiologic factors associated with breast cancer in women: 1) mammographic density, 2) sex hormone concentrations, and 3) body mass index (BMI). Mammographic density, which describes the proportion of dense (i.e., epithelial and stromal) tissue in the breast, is one of the strongest predictors of breast cancer in women. Women with extremely dense breasts have a 3 to 6-fold increased risk of breast cancer compared to those with primarily fatty breasts. Breast cancer is generally considered to be a primarily hormone-driven cancer, an attribute that has led to the development of effective treatment and prophylactic strategies for hormone receptor positive subtypes and cause for investigating the role of endogenous hormones in breast cancer etiology. Finally, BMI has been consistently observed to have paradoxical associations with breast cancer across menopause, with evidence of preventative effects associated with higher BMI in premenopausal women but increased risk in postmenopausal women. We first built on recent analyses that investigated the genetic architecture of testosterone and sex hormone binding globulin (SHBG) in men and women of European ancestry by conducting genome-wide association studies (GWAS) of estradiol concentrations in women. Additionally, we investigated the generalizability of previous findings in women of African ancestry. We further conducted menopausal status specific GWAS of these sex hormones to identify loci with heterogeneous effects across menopause. We found that the strongest overall genetic predictor of testosterone concentrations, located in the CYP3A7 gene, had an effect nearly twice as large in premenopausal women compared to postmenopausal women. Similarly, genetic variants in the AKR1C4 gene were strongly associated with concentrations of SHBG in premenopausal women, but not in postmenopausal women, with a 5-fold difference in effect estimates between the two. We also estimated the shared heritability across menopausal status specific hormone concentrations, and observed a relatively low genetic correlation between pre- and postmenopausal detectable levels of estradiol, whereas comparisons of pre- and postmenopausal shared heritability for SHBG and testosterone were both close to one, indicating near identical genetic architectures. We performed gene-level tests for enrichment of genetic associations within tissue-specific gene expressions by collapsing multiple SNP-level associations in a gene while accounting for linkage disequilibrium. Using this gene-set analysis for tissue specificity we observed a change from strong adrenal gland tissue specificity of testosterone in premenopausal to adipose tissue specificity in postmenopausal women, suggesting that adiposity may play a more important role in determining circulating concentrations of testosterone after menopause. To understand the directional relationships between overall and menopausal status specific concentrations of sex hormones and BMI on mammographic dense and non-dense area we performed Mendelian Randomization analyses. We created menopausal status specific genetic instruments for SHBG, testosterone, and estradiol based on our previous work. We obtained single nucleotide polymorphisms (SNP)-specific association statistics from a recent GWAS of mammographic density of up to 27,900 women of European ancestry. Effect estimates for BMI were obtained from the largest meta-GWAS of BMI to date, comprising more than 700,000 individuals. We observed an inverse relationship between overall genetically predicted testosterone and dense area. Increasing genetically predicted BMI was strongly associated with an increase in genetically predicted non-dense area, as previously observed. However, we also observed an inverse association between genetically predicted BMI and absolute dense area, which might explain some of the reduced risk of breast cancer associated with an increase in genetically predicted BMI. Higher genetically predicted BMI was also strongly associated with decreasing SHBG concentrations, as well as increasing concentrations of testosterone. Based on the inverse-variance weighted results, we observed increasing genetically predicted BMI to be associated with a decrease in genetically predicted detectable levels of overall and premenopausal specific estradiol concentrations, but not for postmenopausal only. Multivariable MR approaches for the association of BMI and mammographic density adjusting for sex hormones did not substantively change the effect estimates of BMI. Building on the strong association between BMI and mammographic density, we sought to identify genetic loci that interact with BMI to alter mammographic density phenotypes. We conducted genome-wide tests for the interaction between SNPs and BMI on percent mammographic density, absolute dense area, and absolute non-dense area in 14,837 women. Despite having the largest sample size to date with genetic and phenotypic data for mammographic density, we did not find any loci that reached standard Bonferroni correction for statistical significance. This work presents novel findings of the unique genetic architectures of menopausal specific concentrations of sex hormones in women and extends these findings to investigate their associations with mammographic density. We show that BMI plays an important role in determining not only non-dense area, but also dense area and a possibly separate mechanism for breast cancer etiology. Additionally, there is evidence based on our MR approaches of a regulatory role of BMI on endogenous estradiol as yet another possible pathway to tumorigenesis. We did not identify any genetic variant that has a strong modifying effect of BMI on mammographic density phenotypes. It is possible that larger studies are merited to investigate the interactions between germline genetic variants and BMI on mammographic density variation