The ambition of chemical exposomics to measure all environmental exposures throughout the lifecourse brings analytical challenges because of the large number of environmental chemicals, their diverse physicochemical properties, their dynamic occurrence, and presence in blood at 1000-fold lower concentrations than endogenous metabolites. Gas chromatography high resolution mass spectrometry (GC-HRMS) complements liquid chromatography (LC) by expanding chemical-space coverage into neutral, hydrophobic and semi-volatile substances. However, development of GC-HRMS chemical exposomics has lagged behind LC-, and workflows are still required for sensitive and quantitative detection for multiple priority chemical classes while simultaneously enabling the discovery of novel exposures by nontarget acquisition, robust data processing and appropriate structural annotation frameworks. In Paper I, an actionable annotation scoring framework was developed to incorporate unique GC-HRMS information into annotation confidence level assignments, these GC-specific criteria were applied in Papers II, III and IV.

In Paper II, I developed and validated a chemical exposomics method using isohexane (H) to quantitatively extract prioritized analytes from protein-free acetonitrile-plasma (A-P), which significantly reduced coextracted lipid interference and enabled large-volume injections (25 µL) to GC. The resulting HA-P method enabled highly sensitive and quantitative detection, achieving a mean method limit of quantification (MLOQ) of 0.09 ng/mL from only 200 µL of human plasma. Application to 32 individual samples (100 µL) allowed quantification of 51 targets and a nontarget molecular discovery of 112 additional substances (Level 2, 12.8% high annotation rate). In Paper III, the HA-P method was then applied in a longitudinal study of 46 healthy individuals in a multiomics cohort, whereby each participant donated 6 plasma samples over 2 years. Overall, the GC chemical-exposome was longitudinally unstable, with mean intraclass correlation coefficients (ICCs) of 0.24, significantly lower than for other omics profiles measured (i.e., proteomics, metabolomics, lipidomics and microbiota). Molecular networks and hierarchical clustering analysis revealed structural similarities and correlated co-exposures for numerous chemical classes that may share common exposure sources.

To enable comprehensive chemical exposomics, in Paper IV I developed and validated an integrated sample preparation method that produces two extracts from a single plasma sample, enabling high sensitivity target/nontarget analysis of separate polar and nonpolar analyte fractions. Application to 32 plasma samples allowed overall identification of 204 chemicals at Level 1 covering a wide chemical space, e.g., 11 orders of magnitude in water solubility. Hierarchical clustering of 348 total annotated features revealed a broad range of common or rare co-exposures, many of which were unique to specific individuals or correlated with endogenous metabolites, thereby revealing a high-relevance to precision public health. Overall, the combined methods and frameworks provide new tools to study the human chemical exposome, and the unprecedented datasets from their first applications will guide sampling design (based on low ICCs), comprehensive analysis and data exploration in future studies.