Abstract
Introduction: The methylomic mechanisms underlying abdominal aortic aneurysm (AAA) and the elevated risk observed in smokers remain poorly understood.
Methods: The analytic cohort comprised individuals from the VA Million Veteran Program who did not have a clinically diagnosed AAA at the time of enrollment and had available DNA methylation data. DNA methylation was profiled using the Illumina 850K EPIC array, with quality control performed via SeSAMe. We conducted logistic regression to assess associations between CpGs and incident AAA and corrected for bias using a Bayesian method. A replication analysis was performed in an independent cohort of 473 individuals with AAA and 488 controls with DNA methylation profiled by the Illumina 450K array. Pathway enrichment was conducted using the "enrichR" package. Smoking-methylation associations were estimated using linear regression with adjustment for covariates. A network Mendelian randomization (MR) was performed. Bayesian colocalization was used to determine whether CpG and AAA associations shared causal variants. We estimated the proportion of smoking-AAA association mediated by CpGs using a two-stage MR framework. We then compared those estimates against cohort data.
Results: The analysis included 40,492 individuals; 1,276 developed incident AAA during follow-up. Among 757,266 autosomal CpG sites, 816 CpGs were identified as being significantly associated with AAA after Bonferroni correction. In the replication cohort, 80.7% of the available 374 associations demonstrated consistent associations (P < 0.05). Gene-set enrichment of 449 unique genes mapped from these AAA-associated CpGs identified pathways including JAK2/STAT3 signaling, osteoblast differentiation, TGF-β and NF-κB intracellular cascades, etc. All 816 CpGs were significantly differentially methylated between never and ever smokers (P < 1.97×10-4). In MR analyses, genetically predicted methylation at 125 CpGs was associated with AAA; 49 of these associations had concordant directions in cohort and MR analyses. Colocalization analysis provided strong-to-moderate evidence at two CpG sites: cg26242531 (ZFYVE21) and cg15474579 (CDKN1A). Genetically proxied smoking initiation was significantly associated with 23 CpGs. MR-based and cohort-based mediation analyses uncovered multiple smoking-CpG-AAA pathways.
Conclusions: Our integrative epigenetic and genetic analyses establish a causal role for DNA methylation in AAA development.