Linking cancer genomics to metabolites: A Mendelian randomization study

Introduction & Objectives: Breast cancer (BC) is the most common cancer in women and the 2nd most common cancer overall (Bray et al., 2018). Several studies have attempted to unravel the underlying physiology; however, a complete knowledge of the exact etiology is still lacking due to the complexity of the disease. Genome-wide association studies (GWAS) have revealed thousands of association signals influencing complex traits or diseases including BC, providing a great opportunity to unravel causal mechanisms. Here we employ BC GWAS results as genetic instruments (or instrumental variables; IV) and try to investigate the causal associations of circulating metabolites with the risk of BC.Materials & Methods: We worked over the largest GWAS study of breast cancer comprising a total of 122,977 cases and 105,974 controls from European ancestry (Michailidou et al., 2017). We first identified the independent associated loci to BC by clumping GWAS results based on GWAS p-value<5.00E-08, r2>0.05 and physical distance of 1Mb. Next we used GWAS results of 123 metabolites, measured by magnetic NMR, reported by studying 24,925 individuals of European descent (Kettunen et al., 2016). Finally, we integrated BC with 123 metabolites using Mendelian randomization (MR) approach (Zhu et al., 2018). Since MR needs strong genetic instruments to prevent inflation of test statistics under the null hypothesis that bxy = 0, we set the significance threshold of GWAS p-value for genetic instruments to 5.00E-08. After exclusion of two metabolites lacking strongly significant SNPs (p-value≥5.00E-08), 121 metabolites remained which were then clumped and used for MR analysis against BC. We controlled inflation of test statistics due to LD using HEIDI-outlier test. We also checked pleiotropic effects of the used IVs by LD Hub (Zheng et al., 2017) to adjust for potential genetic confounders using mtCOJO approach (Zhu et al., 2018).Results: Bi-directional MR analyses of 121 metabolites against BC, controlling for linkage and pleiotropic effects, revealed significant causal associations for 28 metabolites (P-value < 4.13E-04; which is the Bonferroni corrected threshold based on all metabolite-disease tests i.e. 0.05/121) in forward analysis i.e. metabolite → disease. Tyrosine and total cholesterol in HDL with the largest ORs (1.16 and 1.11 respectively) were at the top of risk factors and total lipids in chylomicrons and extremely large VLDL showed the strongest protective effects (OR = 0.87, P-value = 3.41E-06). Reverse MR analyses i.e. disease → metabolite resulted in no significant associations (P-value > 4.13E-04).Conclusion: Our results suggest that high levels of circulating Tyrosine increase the risk for BC which may be explained via Tyrosine kinase pathways (Hsu and Hung, 2016). It also suggest a protective role for chylomicron and VLDL ingredients against the disease free of non-genetic confounders. Our pipeline can be expanded to include whole metabolome and also can be applied by other researchers for any other traits or disease.

کلیدواژه ها:

Association ، breast cancer ، lipids ، Mendelian randomization ، causality