DNA Methylation Study Reveals Inflammation Drives Abdominal Aortic Aneurysm Risk
The largest epigenome-wide study of AAA identifies 1,253 methylation markers and points to inflammatory and lipid pathways as key causal drivers.
Summary
Abdominal aortic aneurysm is a bulge in the body's largest artery that usually causes no symptoms until it ruptures, and rupture is fatal in over 40% of cases. No drug can stop it from growing. Researchers studied blood DNA methylation, chemical tags on DNA that record exposures like smoking, in about 44,000 U.S. veterans. They found 1,253 methylation sites linked to developing an aneurysm, and genetic analyses supported a likely causal role for 143. These sites clustered in gene-control regions tied to inflammatory switches (AP-1 and NF-κB). Blood cholesterol pathways explained over 40% of the risk at two sites, near LDLR and TRIB1. Inflammatory proteins including MMP9, MMP13, and IL1RN were flagged as possible intermediates. A methylation risk score added information beyond standard risk factors in three cohorts, though it needs prospective testing before clinical use.
Detailed Summary
Abdominal aortic aneurysm (AAA), a focal widening of the abdominal aorta beyond 3 cm, stays silent until it ruptures, and rupture carries a mortality rate above 40%. Prevalence reaches up to 8% in older adults, and no drug has been approved to halt its growth. DNA methylation records cumulative exposures such as smoking, diet, and stress, and it can change gene regulation. This study asked which blood methylation marks are linked to AAA, whether they are plausibly causal, and what pathways they reveal.
Researchers used the Veterans Affairs Million Veteran Program. After quality control, 44,045 participants with Illumina EPIC array methylation data were included. The 656 veterans with prevalent AAA were removed to limit reverse causation, leaving 1,324 incident cases, diagnosed a median of 2.91 years after blood draw, and 42,065 controls. Logistic regression tested 757,266 CpG sites, adjusting for age, sex, blood cell proportions, and technical and genetic principal components, with BACON inflation correction (λ=1.16) and Bonferroni thresholds.
The analysis identified 1,253 CpGs associated with incident AAA: 545 hypermethylated and 708 hypomethylated sites linked to higher risk. Results held after excluding cases diagnosed within 12 months and in Cox models. Only 2 CpGs, near MYOM1 and AHRR, differed significantly between European and African genetic-similarity groups. The strongest signals included known smoking markers (AHRR, PRSS23, F2RL3). Adjusting for AHRR methylation or a 1,255-CpG smoking score removed test inflation, whereas self-reported smoking did not.
Mendelian randomization, using cis methylation quantitative trait loci and AAAGen data (37,214 cases, 997,776 controls), supported a putative causal role for 143 CpGs. Annotation showed mostly distal, enhancer-centered regulation enriched for AP-1 and NF-κB inflammatory transcription factors. Expression quantitative trait methylation in 2,115 Framingham participants showed many trans associations. Colocalization found 213 CpG–gene pairs with strong support, including AGER, C2, and DDR1, and 8 multi-trait links sharing one genetic signal.
Network Mendelian randomization identified 135 mediation pathways: 58 through cardiometabolic traits and 77 through inflammation-related traits. Blood lipids mediated over 40% of the association at CpGs near LDLR (cg19305903) and TRIB1 (cg14128479). Circulating DAG1, NEXN, MMP9, IL1RN, and MMP13 emerged as candidate intermediates. A LASSO-derived methylation risk score, built on an 80:10:10 train, recalibrate, and test split, was strongly associated with AAA in the Million Veteran Program and in two external cohorts, the Cardiovascular Health Study and a New Zealand case-control study, even after full clinical adjustment.
Important caveats apply. Methylation was measured in blood rather than aortic tissue. Causal inference relied mainly on European-similarity genetic instruments with partial sample overlap. The external cohorts included only prevalent disease. The authors stress that prospective validation is needed before the score can refine screening or risk stratification.
Key Findings
- 1,253 CpG sites were significantly associated with incident AAA (1,324 cases vs 42,065 controls; Bonferroni-corrected P<0.05), including 545 hypermethylated and 708 hypomethylated sites linked to higher risk
- Mendelian randomization supported a putative causal role for 143 of the 1,253 CpGs
- 213 CpG–gene expression pairs showed strong colocalization (PPH4≥0.7), including AGER, C2, and DDR1, and 8 multi-trait links shared one genetic signal with AAA
- Network MR identified 135 mediation pathways: 58 via cardiometabolic traits and 77 via inflammation-related traits
- Blood lipid traits mediated more than 40% of the CpG–AAA associations near LDLR (cg19305903) and TRIB1 (cg14128479)
- Annotation showed predominantly distal, enhancer-centered regulation with enrichment of inflammatory transcription factors AP-1 and NF-κB; candidate protein mediators were DAG1, NEXN, MMP9, IL1RN, and MMP13
- A methylation risk score was associated with AAA after full clinical adjustment in the discovery cohort and in 2 external cohorts; only 2 of 1,253 CpGs (MYOM1, AHRR) differed between European and African groups
Methodology
This epigenome-wide association study used 44,045 Million Veteran Program participants with EPIC array blood methylation, comparing 1,324 incident AAA cases (identified by at least 2 ICD codes) with 42,065 controls across 757,266 autosomal CpGs. Logistic regression adjusted for age, sex, blood cell proportions, and technical and genetic principal components, with BACON inflation control, Bonferroni correction, and Cox and 12-month-exclusion sensitivity analyses. Two-sample Mendelian randomization used cis-meQTL instruments (F>10) and AAAGen summary statistics, followed by HyPrColoc colocalization, eQTM analysis in 2,115 Framingham participants, and two-stage network MR. A LASSO methylation risk score was trained on 80% of data and tested against a clinical model, with Bayesian equivalence testing (±5%), and externally evaluated in the Cardiovascular Health Study and a New Zealand case-control study.
Study Limitations
Methylation was measured in blood rather than aortic tissue, AAA was ascertained from electronic health record codes, and the cohort is a veteran population. Mendelian randomization relied mainly on European-similarity genetic instruments with partial sample overlap, and the external validation cohorts contained only prevalent AAA, so prospective validation of the risk score is still needed. Conflict-of-interest disclosures were not available in the supplied text, and the results sections on the later analyses were truncated.
Enjoyed this summary?
Get the latest longevity research delivered to your inbox every week.
Enter your email to subscribe:
