Clonal hematopoiesis of indeterminate potential, DNA methylation, and risk for coronary artery disease

Abstract

Age-related changes to the genome-wide DNA methylation (DNAm) pattern observed in blood are well-documented. Clonal hematopoiesis of indeterminate potential (CHIP), characterized by the age-related acquisition and expansion of leukemogenic mutations in hematopoietic stem cells (HSCs), is associated with blood cancer and coronary artery disease (CAD). Epigenetic regulators DNMT3A and TET2 are the two most frequently mutated CHIP genes. Here, we present results from an epigenome-wide association study for CHIP in 582 Cardiovascular Health Study (CHS) participants, with replication in 2655 Atherosclerosis Risk in Communities (ARIC) Study participants. We show that DNMT3A and TET2 CHIP have distinct and directionally opposing genome-wide DNAm association patterns consistent with their regulatory roles, albeit both promoting self-renewal of HSCs. Mendelian randomization analyses indicate that a subset of DNAm alterations associated with these two leading CHIP genes may promote the risk for CAD.

Fig. 1. Results from epigenome-wide association studies of four CHIP phenotypes.

a Directional Manhattan plot of discovery multi-ancestry meta-EWAS for any CHIP in CHS cohort, where direction indicates positive vs. negative correlations between CHIP and DNAm. Each dot represents a CpG site, with genomic location on the x-axis and –log₁₀(P)*sign(test statistic) on the y-axis, where P values are based on a two-sided inverse-variance-weighted meta-analysis. Solid horizontal line indicates Bonferroni significance, and dashed line indicates 5% FDR. b–d Volcano plots depicting the effect size and −log₁₀(P) from CHS meta EWAS of b any CHIP, c DNMT3A CHIP, and d) TET2 CHIP. Dashed line indicates FDR < 5%, and colored points highlight CpGs replicated in ARIC cohort. e Overlap of replicated CpGs among the four CHIP EWAS. f Distribution of DNAm at the eight most significant replicated CpGs associated with both DNMT3A and TET2. Colored points show DNAm proportions at each CpG for individuals with DNMT3A (blue) or TET2 (green) CHIP, overlaid by density functions for each group and lines representing medians of each distribution. For comparison, medians for individuals without CHIP are shown as white circles.

Fig. 2. Enrichment patterns among DNMT3A- and TET2-associated CpGs.

a–c Distribution of average methylation levels estimated from external WGBS data for myeloid cells (a), lymphoid cells (b), and HSCs (c) for three sets of CpG sites: all CpGs on Illumina 450 K array (gray), and CpGs showing replicated association with DNMT3A CHIP (blue) or TET2 CHIP (green). Each point represents a CpG, while filled curves show the density function corresponding to all CpGs in each set. Horizontal lines indicate median of distribution. Because of the large number of CpGs considered (N = 478,661), all pairwise comparisons between cell types were significant (Two-sided Wilcoxon P < 2 × 10⁻¹⁶). d Enrichment in cell-specific DHS among the top 1000 DNMT3A- or TET2-associated CpGs, compared to 1000 random genomic-context-matched CpGs (one-sided binomial test; N = 2000). Estimated OR (x-axis, indicated by filled squares) shows extent to which DNMT3A- or TET2-associated CpGs are enriched (or depleted) for DHS regions in six distinct cell types (y-axis), compared to other sites on the array. Horizontal lines indicate 1−α confidence intervals for estimated OR, using a Bonferroni-adjusted α of 0.05/12. Th1/2: Type 1/2 T helper cells. e–f Comparison of DNAm profiles associated with gene-specific mutations in CHIP vs. AML. Test statistics from EWAS of mutations in DNMT3A (e) or TET2 (f) in the context of blood samples from healthy individuals with or without CHIP (x-axis; Z-statistics from discovery sample meta-analysis, N = 582) vs. tumor samples from patients with AML (y-axis; T-statistics from EWAS of mutation type in TCGA data, N = 127 (e) or 108 (f)). Black points: FDR < 0.05 in CHS discovery sample but did not replicate; Blue or green points: FDR < 0.05 and replicated in ARIC.

Fig. 3. Mendelian randomization analysis of CHIP-associated CpGs and CAD risk.

For sets of replicated CpGs associated with CHIP, DNMT3A, and TET2 (y-axis), the odds ratio (x-axis, indicated by filled squares) reflects the change in CAD risk associated with each SD increase in DNAm, with lines representing 95% confidence intervals estimated by GSMR. GSMR analysis was based on published summary statistics (effect estimates) for cis-mQTL (N = 32,851) and CAD GWAS (N = 547,261). Only exposure CpGs showing causal evidence in the MR analysis (FDR < 0.05 based on P-values from two-sided χ² test) are presented here; full summary statistics are available in Supplementary Data 7. *”Association with CHIP”: “+” or “-” signs indicate effect directions for associations with CHIP, DNMT3A or TET2 in the meta-EWAS of CHS AA, CHS EA, ARIC AA, and ARIC EA EWAS. #SNP: number of SNPs included in the MR analysis for each CpG.