Cadmium, Smoking, and Human Blood DNA Methylation Profiles in Adults from the Strong Heart Study

Abstract

Background: The epigenetic effects of individual environmental toxicants in tobacco remain largely unexplored. Cadmium (Cd) has been associated with smoking-related health effects, and its concentration in tobacco smoke is higher in comparison with other metals.

Objectives: We studied the association of Cd and smoking exposures with human blood DNA methylation (DNAm) profiles. We also evaluated the implication of findings to relevant methylation pathways and the potential contribution of Cd exposure from smoking to explain the association between smoking and site-specific DNAm.

Methods: We conducted an epigenome-wide association study of urine Cd and self-reported smoking (current and former vs. never, and cumulative smoking dose) with blood DNAm in 790,026 CpGs (methylation sites) measured with the Illumina Infinium Human MethylationEPIC (Illumina Inc.) platform in 2,325 adults 45-74 years of age who participated in the Strong Heart Study in 1989-1991. In a mediation analysis, we estimated the amount of change in DNAm associated with smoking that can be independently attributed to increases in urine Cd concentrations from smoking. We also conducted enrichment analyses and in silico protein-protein interaction networks to explore the biological relevance of the findings.

Results: At a false discovery rate (FDR)-corrected level of 0.05, we found 6 differentially methylated positions (DMPs) for Cd; 288 and 17, respectively, for current and former smoking status; and 77 for cigarette pack-years. Enrichment analyses of these DMPs displayed enrichment of 58 and 6 Gene Ontology and Kyoto Encyclopedia of Genes and Genomes gene sets, respectively, including biological pathways for cancer and cardiovascular disease. In in silico protein-to-protein networks, we observed key proteins in DNAm pathways directly and indirectly connected to Cd- and smoking-DMPs. Among DMPs that were significant for both Cd and current smoking (annotated to PRSS23, AHRR, F2RL3, RARA, and 2q37.1), we found statistically significant contributions of Cd to smoking-related DNAm.

Conclusions: Beyond replicating well-known smoking epigenetic signatures, we found novel DMPs related to smoking. Moreover, increases in smoking-related Cd exposure were associated with differential DNAm. Our integrative analysis supports a biological link for Cd and smoking-associated health effects, including the possibility that Cd is partly responsible for smoking toxicity through epigenetic changes. https://doi.org/10.1289/EHP6345.

Figure 1.

Summary of results from the epigenome-wide association study of cadmium (Cd) and smoking with DNA methylation (DNAm) and subsequent bioinformatics analysis. After considering a false discovery rate (FDR)-corrected p-value < 0.05, we obtained 6 Cd-differentially methylated positions (DMPs) and 312 smoking DMPs in a genomic exploration of $\sim \mathrm{790,000}$ CpG sites with the Infinium MethylationEPIC BeadChip. We excluded nonprotein-coding genomic regions and duplicate genes and conducted gene set enrichment analysis. We subsequently evaluated protein interaction networks among protein-encoding genes from the STRING database, resulting in a protein interaction network of 271 nodes and 1,802 interactions. We observed highly connected nodes, which were directly or indirectly related to DNAm metabolism key proteins.

Figure 2.

Enrichment analysis of top significant cadmium (Cd) and smoking-related differentially methylated positions (DMPs) ($p\text{-value}<{10}^{-5}$) for 15-chromatine states from the ROADMAP project. The area of the solid dots is directly proportional to the strength of the statistical evidence in favor of being annotated to a given chromatin state category comparing statistically significant vs. nonstatistically significant DMPs in our data.

Figure 3.

Protein-interaction network between proteins attributed to cadmium (Cd)- and smoking-differentially methylated positions (DMPs) and key proteins in DNA methylation (DNAm) metabolism pathways from KEGG. The nodes correspond to proteins involved in DNAm-related pathways, and proteins encoded by genes associated to Cd-DMPs and smoking-DMPs. The size of the nodes is proportional to the number of connections. The STRING database provides a confidence score (from 0 to 1) to indicate the estimated likelihood that the annotated interaction between a given pair of proteins is biologically meaningful, specific, and reproducible, according to the evidence derived from in-house predictions, homology transfers, and the externally maintained databases (Szklarczyk et al. 2019). Increasingly darker solid edge lines indicate a protein interaction with increasingly higher confidence scores. The nodes included among the top 25 statistically significant Cd- and/or smoking-DMPs in our EWAS show thicker node lines.