Long-term estimates of ambient air pollution and perturbations in the peripheral blood transcriptome in a large Dutch twin cohort

Abstract

Background/Aim Studying the internal exposome has the potential to provide insight into the biological pathways through which air pollutants affects disease risk. In this study we linked long-term estimates for seven air pollutants and their elemental composition to transcriptomics data collected in the Dutch Twin Register (NTR) cohort. In addition to standard regression techniques we exploited the twin design by comparing gene expression among pairs of monozygotic and dizygotic twins. Replication of findings will be conducted in the Dutch NESDA cohort. Methods In both discovery (NTR, n= 2438) and replication (NESDA, n= 2341) cohorts, long-term land use regression model estimates were generated for nitrogen oxides (NO2, NOx), particulate matter (PM2.5 (including elemental composition), PM2.5abs, PM10, PMcoarse), and ultrafine particles (UFP). Other covariates were assessed using geospatial methods or questionnaires. Gene expression was assessed using Affymatrix U219 arrays (n=44,241 probes). Multi-variable univariate mixed-effect models were used to assess the association between these air pollutants and the transcriptome. Functional analysis was conducted using DAVID (v6.8). Results In the NTR cohort long-term exposure to PM2.5 (and few of the other pollutants) was associated with 374 transcripts (Benjamini- Hochberg p<0.05). Elemental Copper and Sulphur in PM2.5 significantly contributed to this signal. Associations were stronger among residents of less urban areas, with lower exposure levels, and weaker among men. Results were robust in a series of sensitivity analyses. Genes most significantly differentially expressed were ZNF791, BTBD1 and OSBPL8 (upregulated), and WIPF2 (downregulated). We observed strong evidence for enrichment of phosphoprotein and alternative splicing pathways. Twin-only models did not yield strong associations. Replication analyses in the NESDA cohort are ongoing. Conclusions Results from our analyses in the NTR cohort suggest a distinct PM2.5 signal in the peripheral blood transcriptome, primarily occurring in phosphoprotein and alternative splicing pathways.