Cis-meQTL for cocaine use-associated DNA methylation in an HIV-positive cohort show pleiotropic effects on multiple traits

Abstract

Background: Cocaine use (CU) is associated with psychiatric and medical diseases. Little is known about the mechanisms of CU-related comorbidities. Findings from preclinical and clinical studies have suggested that CU is associated with aberrant DNA methylation (DNAm) that may be influenced by genetic variants [i.e., methylation quantitative trait loci (meQTLs)]. In this study, we mapped cis-meQTLs for CU-associated DNAm sites (CpGs) in an HIV-positive cohort (N_total = 811) and extended the meQTLs to multiple traits.

Results: We conducted cis-meQTL analysis for 224 candidate CpGs selected for their association with CU in blood. We identified 7,101 significant meQTLs [false discovery rate (FDR) < 0.05], which mostly mapped to genes involved in immunological functions and were enriched in immune pathways. We followed up the meQTLs using phenome-wide association study and trait enrichment analyses, which revealed 9 significant traits. We tested for causal effects of CU on these 9 traits using Mendelian Randomization and found evidence that CU plays a causal role in increasing hypertension (p-value = 2.35E-08) and decreasing heel bone mineral density (p-value = 1.92E-19).

Conclusions: These findings suggest that genetic variants for CU-associated DNAm have pleiotropic effects on other relevant traits and provide new insights into the causal relationships between cocaine use and these complex traits.

Keywords: Cis-methylation quantitative trait loci (cis-meQTL); Cocaine use; Complex trait; Epigenome-wide association study (EWAS); Mendelian randomization.

Fig. 1

Overall study design. A Flowchart for the analyses performed. Selection of CU-related CpGs and meQTL identification were performed in 423 samples with DNAm profiled by 450K arrays (blue) and 388 samples with DNAm profiled by EPIC arrays (green), separately. Meta-analyses were followed to combine results (grey). Downstream analyses were conducted based on the identified meQTLs. B Paths involved in the analyses. After selection of candidate CpGs associated with CU (Step 1) and meQTL identification (Step 2), downstream analyses including PheWAS and meQTL trait enrichment were performed to investigate the pleiotropic effects of the meQTLs on other traits. Mendelian randomization was then performed to build causal paths. CU: cocaine use; PheWAS: phenome-wide association study; meQTL: methylation quantitative trait loci

Fig. 2

Cis-meQTLs identified for cocaine use-associated CpG sites. A Manhattan plots of meQTL results following meta-analysis. The red line indicated the FDR-corrected 0.05 level (p-value = 8.9e-4). The top SNP (index meQTL) in the most significant meQTL clumps was marked with the corresponding CpG site and gene. B The top 11 index meQTLs marked in (A). meQTL: methylation quantitative trait loci

Fig. 3

Manhattan plot of the traits associated with the top index meQTLs in PheWAS analysis. The associations and classification of traits (domains) were obtained from 4,756 GWAS studies available on GWAS Atlas. The red dash line indicated the Bonferroni-corrected 0.05 level (p-value = 1.70E-07). PheWAS: phenome-wide association study; meQTL: methylation quantitative trait loci

Fig. 4

Results of meQTL trait enrichment analysis. A Dot plot for the 16 complex human traits which the meQTLs were enriched in (p-value < 0.05). B Venn plot to compare the 36 phenotypes identified by PheWAS and 16 human traits identified by meQTL trait enrichment. The 9 overlapped traits were marked in bold in (A). PheWAS: phenome-wide association study; meQTL: methylation quantitative trait loci