Epigenetic predictors of all-cause mortality are associated with objective measures of neighborhood disadvantage in an urban population

Abstract

Background: Neighborhood characteristics are robust predictors of overall health and mortality risk for residents. Though there has been some investigation of the role that molecular indicators may play in mediating neighborhood exposures, there has been little effort to incorporate newly developed epigenetic biomarkers into our understanding of neighborhood characteristics and health outcomes.

Methods: Using 157 participants of the Detroit Neighborhood Health Study with detailed assessments of neighborhood characteristics and genome-wide DNA methylation profiling via the Illumina 450K methylation array, we assessed the relationship between objective neighborhood characteristics and a validated DNA methylation-based epigenetic mortality risk score (eMRS). Associations were adjusted for age, race, sex, ever smoking, ever alcohol usage, education, years spent in neighborhood, and employment. A secondary model additionally adjusted for personal neighborhood perception. We summarized 19 neighborhood quality indicators assessed for participants into 9 principal components which explained over 90% of the variance in the data and served as metrics of objective neighborhood quality exposures.

Results: Of the nine principal components utilized for this study, one was strongly associated with the eMRS (β = 0.15; 95% confidence interval = 0.06-0.24; P = 0.002). This principal component (PC7) was most strongly driven by the presence of abandoned cars, poor streets, and non-art graffiti. Models including both PC7 and individual indicators of neighborhood perception indicated that only PC7 and not neighborhood perception impacted the eMRS. When stratified on neighborhood indicators of greenspace, we observed a potentially protective effect of large mature trees as this feature substantially attenuated the observed association.

Conclusion: Objective measures of neighborhood disadvantage are significantly associated with an epigenetic predictor of mortality risk, presenting a potential novel avenue by which neighborhood-level exposures may impact health. Associations were independent of an individual's perception of their neighborhood and attenuated by neighborhood greenspace features. More work should be done to determine molecular risk factors associated with neighborhoods, and potentially protective neighborhood features against adverse molecular effects.

Keywords: DNA methylation; Mortality predictors; Neighborhood disadvantage; Social determinants of health; Urban populations.

Fig. 1

Association between principal components for neighborhood quality and eMRS. Each of the principal components (PCs) for neighborhood quality was associated with the epigenetic mortality risk score (eMRS [27]) in a model adjusted for age sex, race (White, African-American, and Other), ever smoking, ever alcohol usage, years spent residing in neighborhood, education (binary indicator for some college or more), and employment (binary indicator for employed vs unemployed). Squares and triangles represent models stratified on sex (males and females, respectively), and in such models, the term for sex was removed. Principal components are numbered in order of their ranking of the percent variance explained, and only the top nine were examined as they explained > 90% of the variance as detailed in the “Methods” section

Fig. 2

Association between PC7 and eMRS stratified on indicators of neighborhood greenspace. To examine if associations were potentially modified by exposure to greenspace, we stratified associations between neighborhood principal component 7 (PC7) and the epigenetic mortality risk score (eMRS [27]). Associations appeared to be substantially stronger in neighborhoods with greater than the median presence of large mature trees (Trees (High)) and with community gardens (Gardens) as compared to neighborhoods with below median presence of large mature trees (Trees (Low)) and no observed community gardens (No Gardens). The median presence of large mature trees was 84.2%. Gray bar indicates region as defined by the 95% confidence interval for the overall association