School of Public Health

Gloria C. Chi

Methylomic and Transcriptomic Insights Linking Air Pollution and Atherosclerosis


Current literature demonstrates a compelling relationship between long-term air pollution exposure and cardiovascular diseases such as atherosclerosis. DNA methylation is a potential mediator of the air pollution-atherosclerosis association because 1) exposure to particulate matter is associated with alterations in DNA methylation, and 2) alterations in DNA methylation are found in atherosclerotic lesions and are associated with ischemic heart disease and blood pressure. However, findings of past studies may be confounded by cell type composition due to the use of mixed leukocytes that combine distinct methylation profiles.


In this dissertation, we tested the associations between long-term ambient air pollution exposure and DNA methylation (global, epigenome-wide, and candidate site) in purified monocytes. We also tested the associations between air pollution-associated DNA methylation patterns and gene expression and subclinical atherosclerosis.


The Multi-Ethnic Study of Atherosclerosis (MESA) and its ancillary studies provided rich information on air pollution, methylomics, transcriptomics, and coronary artery calcium (CAC) from a diverse cohort of adults. A sample of 1,264 participants were randomly selected for DNA methylation and gene expression assessment using monocytes. Of these, 1,207 participants with air pollution exposure predictions were included in the analytic sample. Ambient fine particulate matter (PM2.5) and oxides of nitrogen (NOX) concentrations were predicted at participants’ residences using sophisticated spatiotemporal models that incorporated both monitoring data and geographical variables. Estimates were averaged over one year prior to blood draw. Epigenome-wide DNA methylation and genome-wide gene expression were measured using the Illumina Infinium HumanMethylation450 BeadChip (450k) and the Illumina HumanHT-12 v4 Expression BeadChip, respectively. Candidate site DNA methylation levels were assessed from the 450k array. Mean methylation of probes mapping to long interspersed element 1 (LINE-1) and Alu repetitive elements on the 450k array were used as surrogates for global DNA methylation. CAC was measured using computed tomography following standardized protocols. Linear models were used to investigate the cross-sectional associations between air pollution and global and site-specific DNA methylation, DNA methylation and cis-gene expression, and DNA methylation and CAC. An innovative “bump hunting” method was adapted to the methylation array to identify regions differentially methylated with respect to long-term PM2.5 and/or NOX exposure. Effect modification by sex and race/ethnicity were also investigated by including interaction terms with air pollution. The association between DNA methylation and CAC progression was analyzed using linear mixed models with random coefficients and slopes for each participant. The false discovery rate (FDR) was controlled at 0.05 following the Benjamini-Hochberg method to account for multiple comparisons.


The study included a sample of 1,207 participants (mean age=69.6) that was 21.2% Black, 31.6% Hispanic, 47.2% White, and 51.6% female. In this study, global DNA methylation (LINE-1 and Alu) was not significantly associated with long-term ambient air pollution exposure in monocytes (p>0.05). We identified nine methylations sites and four methylation regions significantly associated with long-term ambient air pollution exposure in monocytes (FDR<0.05). In general, we did not find strong evidence of effect modification by sex or race/ethnicity. Cg05926640 (near LOC101927851) was the CpG site with the smallest p-value for the association with PM2.5 (β=0.049 M-value units per 2.5 µg/m3; 95% confidence interval [CI]: 0.032, 0.067; p=5.6x10-8). Methylation of cg05926640 was also associated with mRNA expression of genes located within 1 Mb including ARID4B and IRF2BP2. Cg11756214 (within ZNF347) was the only CpG site with methylation significantly associated with NOX (β=0.078 M value units per 30 ppb; 95% CI: 0.050, 0.106; p=5.6x10-8), but it was not associated with cis gene expression. The differentially methylated region (DMR) with the smallest p-value for the association with PM2.5 was located on chromosome 5, overlapping the 5’ untranslated region of SDHAP3 (average methylation difference=14.6% per 2.5 µg/m3; p=1.4x10-6), and methylation of this DMR was also associated with expression of MRPL36. This DMR and the DMR located on chromosome 6 near ZFP57 were significantly associated with both PM2.5 and NOX (FDR<0.05). A PM2.5-associated DMR located on chromosome 7 overlapping the 5’ untranslated region of HOXA5 was associated with expression of HOXA5, HOXA9, and HOXA10. Methylation of this DMR was also negatively associated with CAC progression (β=-4.6 Agatston units/year per 2 fold higher DNA methylation; 95% CI: -7.5, -1.7; p=0.002).


Our results support the possibility that exposure to long-term ambient air pollution may lead to alterations in DNA methylation with potential functional gene expression consequences. The findings identified novel associations between long-term air pollution exposure and genes such as the HOXA cluster genes and IRF2BP2, opening up new avenues of research. Genes with DNA methylation associated with both air pollution and gene expression may be more biologically relevant and are involved in various cellular functions including inflammation and hematopoiesis. Our study found methylation of a DMR located on chromosome 7 to be positively associated with PM2.5 exposure and negatively associated with CAC progression. However, a prior study in MESA found a positive association between PM2.5 exposure and CAC progression; thus the direction of the association between the DMR and CAC does not support this DMR as a potential mediator of the air pollution-CAC association. Future studies are needed to replicate our study results and to explore the biological relevance of these air pollution-associated DNA methylation signals and genes in atherosclerosis.