Metabolomics Signatures of Exposure to Ambient Air Pollution: A Large-Scale Metabolome-Wide Association Study in the Cancer Prevention Study-II Nutrition Cohort

Abstract

Existing air pollution metabolomics studies showed inconsistent results, often limited by small sample size and individual air pollutants effects. We conducted a metabolome-wide association study among 1096 women (68.2 ± 5.7 years) who provided blood samples (1998-2001) within the Cancer Prevention Study-II Nutrition Cohort. Annual average individual exposures to particulate matter, nitrogen dioxide, ozone, sulfur dioxide, and carbon monoxide in the year of blood draw were used. Metabolomics profiling was conducted on serum samples by Metabolon. We evaluated the individual air pollutants effects using multiple linear regression and the mixture effect using quantile g-computation, adjusting for confounders and false discovery rate (FDR). Ninety-five metabolites were significantly associated with at least one air pollutant or mixture (FDR < 0.05). These metabolites were enriched in pathways related to oxidative stress, systemic inflammation, energy metabolism, signals transduction, nucleic acid damage and repair, and xenobiotics. Sixty metabolites were confirmed with level 1 or 2 evidence, among which 21 have been previously linked to air pollution exposure, including taurine, creatinine, and sebacate. Overall, our results replicate prior findings in a large sample and provide novel insights into biological responses to long-term air pollution exposure using mixture analysis.

Keywords: air pollution; high-resolution metabolomics; inflammation; metabolome-wide association study; mixture; oxidative stress.

Figure 1

Volcano plots of associations between metabolite intensities and individual air pollutants or air pollution mixture. The x-axis denotes the coefficients of metabolite-pollutant associations. For individual air pollutant model, the coefficient is the change in natural log-transformed standardized metabolite intensity per interquartile range increase in air pollutant exposure levels. For the air pollution mixture model, the coefficient is the change in natural log-transformed standardized metabolite intensity per two quartiles (50%) increase in all air pollutant exposure levels. The y-axis denotes the negative natural log of false discovery rate (FDR) in metabolite-pollutant association. Different colors were used to represent different pathways where the metabolites are involved. The black solid line represents FDR = 0.05 and the black dashed line represents FDR = 0.2. For individual air pollutant model, the labeled metabolites were associated with two air pollutants. For the air pollution mixture model, the labeled metabolites were those meeting FDR < 0.05. PM_2.5, fine particulate matter; PM₁₀, coarse particulate matter; NO₂, nitrogen dioxide; O₃, ozone; SO₂, sulfur dioxide; CO, carbon monoxide; Mixture, air pollution mixture. *A compound that has not been confirmed based on a standard, but Metabolon is confident in its identity (not tier 1).

Figure 2

Potential molecular mechanism underlying the ambient air pollution toxicity using high-resolution metabolomics in the Cancer Prevention Study-II Nutrition cohort. Colored molecules are those identified in our study, with different colors corresponding to different biomolecule categories. Green molecules are amino acids/peptides; yellow molecules are cofactors/vitamins; gray molecules are lipids. The red arrow denotes the metabolite in positive association with air pollution exposure level, while the dark green arrow denotes the metabolite in negative association with air pollution exposure level. The solid black arrow indicates a single-step reaction between the molecule at the arrow’s end and the molecule at the arrow’s top, while the dashed black arrow indicates multiple steps between the molecule at the arrow’s end and the molecule at the arrow’s top. The gray arrow indicates the reaction between molecules and their corresponding receptors. TCA: citric acid cycle; GLR: glycine receptor α-1; TRPV1: transient receptor potential cation channel subfamily V member 1. *Three lysophospholipids were found positively associated with ozone, which were 1-linoleoyl-GPG (18:2), 1-oleoyl-GPE (18:1), and 2-stearoyl-GPE (18:0). ^∧Two sphingomyelins were associated with air pollution: sphingomyelin (d18:2/18:1) was in positive association with coarse particulate matter, while sphingomyelin (d18:1/24:1, d18:2/24:0) was in negative association with ozone.