Evolutionary age of repetitive element subfamilies and sensitivity of DNA methylation to airborne pollutants

Background Repetitive elements take up >40% of the human genome and can change distribution through transposition, thus generating subfamilies. Repetitive element DNA methylation has associated with several diseases and environmental exposures, including exposure to airborne pollutants. No systematic analysis has yet been conducted to examine the effects of exposures across different repetitive element subfamilies. The purpose of the study is to evaluate sensitivity of DNA methylation in differentially‒evolved LINE, Alu, and HERV subfamilies to different types of airborne pollutants. Methods We sampled a total of 120 male participants from three studies (20 high-, 20 low-exposure in each study) of steel workers exposed to metal-rich particulate matter (measured as PM10) (Study 1); gas-station attendants exposed to air benzene (Study 2); and truck drivers exposed to traffic-derived elemental carbon (Study 3). We measured methylation by bisulfite-PCR-pyrosequencing in 10 differentially‒evolved repetitive element subfamilies. Results High-exposure groups exhibited subfamily-specific methylation differences compared to low-exposure groups: L1PA2 showed lower DNA methylation in steel workers (P=0.04) and gas station attendants (P=0.03); L1Ta showed lower DNA methylation in steel workers (P=0.02); AluYb8 showed higher DNA methylation in truck drivers (P=0.05). Within each study, dose–response analyses showed subfamily-specific correlations of methylation with exposure levels. Interaction models showed that the effects of the exposures on DNA methylation were dependent on the subfamily evolutionary age, with stronger effects on older LINEs from PM10 (p‒interaction=0.003) and benzene (p‒interaction=0.04), and on younger Alus from PM10 (p-interaction=0.02). Conclusions The evolutionary age of repetitive element subfamilies determines differential susceptibility of DNA methylation to airborne pollutants.


Exposure assessment
In Study 1, PM was measured using a GRIMM 1100 light-scattering dust analyzer (Grimm Technologies, Inc. Douglasville, GA, USA). Measures of airborne PM mass and PM metal components were obtained from 11 work areas of the steel production facility in order to estimate individual exposures.
In Study 2, the participants wore a passive sampler (stainless steel tube, internal diameter of 9 mm, length of 90 mm) containing Chromosorb 106, near the breathing zone during the work shift. Air benzene level in the passive sampler was measured by thermal desorption followed by gas chromatography/flame ionization detector analysis.
In Study 3, we measured personal EC levels using gravimetric samplers worn near the breathing zone by the study participants during the eight hours of work. Each sampler setup included an Apex pump (Casella Inc., Bedford, UK), a Triplex Sharp-Cut Cyclone (BGI Inc., Waltham, Massachusetts), and a 37-mm Teflon filter placed on top of a drain disc and inside a metal filter holder. The filters were kept under atmosphere-controlled conditions before and after sampling and were weighed with a microbalance (Mettler-Toledo Inc., Columbus, Ohio, USA).

Statistical models
We used the following model: where !"# represents the methylation level for the i-th subject, the j-th position and the k-th duplicate run (i=1,…,40; j=1,…,m, where m varies depending on the total number of CpG sites measured in the sequence; and k=1,2). !! !!" are the random intercept for subject and S 4 random slope for CpG position, respectively. ! is the overall intercept and ! is the fixed effect which expresses the association between exposure and DNA methylation. ! … ! and ! … ! represent covariates and their regression coefficients; !"# is the residual term error. Age and smoking were considered a priori as possible confounders and therefore included as covariates in all the models of the analysis.
We first fitted a set of models in which DNA methylation was regressed over dichotomous exposure variables (high-exposure vs. low-exposure control groups). In a second set of models, we evaluated dose-response relationships by regressing DNA methylation over We also used mixed-effect regression models to determine whether the correlation between DNA methylation and exposures within each repetitive element family varied as a function of the evolutionary age of the subfamilies. The corresponding model was: respectively. The interaction slope was used to model the existence in the data of positions in common for all the subfamilies. Hence, by using the same labels for the positions in the same common sequences -even if belonging to different subfamilies -and rescaling the others accordingly, the interaction can describe unambiguously to which position and subfamily the measure refers. Finally, ! is the overall intercept; ! represents the fixed effect for the exposure, ! for the age of the subfamily; ! expresses the interaction between exposure and evolutionary age; ! … ! and ! … ! are the covariates and their regression coefficients; and !"#$ is the residual term error.
A two-sided P<0.05 was considered statistically significant. All statistical analyses were performed in SAS (version 9.2; SAS Institute Inc., Cary, NC, USA). We used the PROC MIXED procedure to run the mixed-effect models.
S 6 The participants were divided in high-and low-exposed control group according to their exposure levels.