Nontarget mass spectrometry and in silico molecular characterization of air pollution from the Indian subcontinent
Particulate matter (PM) is a major component of air pollution that impacts health and global climate. The fine fraction (<2.5 μm, PM2.5) is responsible for millions of premature deaths annually1,2and is a risk factor for chronic illness and cancer3,4. Ambient levels in low- and middle-income countries of Asia have been declared public health emergencies1,5,6. PM2.5has intercontinental spatial impacts and transports persistent organic contaminants from populated regions to remote global regions7,8,9. PM furthermore affects aerosol-sunlight and aerosol-cloud interactions10and is generally believed to contribute to climate cooling through light-scattering and cloud condensation, although black carbon and brown carbon (BrC; light-absorbing organic matter) components may also lead to warming through absorption of solar radiation11,12. These opposing factors remain poorly constrained in climate models11,12,13. Organic molecules can be a major mass fraction of total PM14, thus a comprehensive molecular characterization of PM2.5could contribute to improved understanding of global air pollution sources, climate impacts, and health effects.
High-resolution mass spectrometry (HRMS) is an established instrumental technique that can reveal the molecular complexity of PM2.5organic compounds, however, most substances remain uncharacterized beyond assignment of molecular formula or the presence of certain functional groups15,16,17,18,19,20,21,22. Characterization of PM2.5samples by HRMS also creates high demands on data processing which has limited previous detailed studies to only a few atmospheric samples. In metabolomics and proteomics, high-throughput workflows for batch-processing of full-scan HRMS chromatographic data (i.e. MS1) and the associated fragmentation spectra (MS2) are now applied to explore the chemical structures of ‘molecular dark matter’ in biological systems23,24,25,26. Such methods have yet to be applied in atmospheric research, but could open new molecular-level windows for studies of air pollution. We hypothesized that new insights into the molecular composition and effects of PM2.5organic compounds could be achieved by combining comprehensive nontarget HRMS analysis with computational workflows23,24,25,26.
Here, PM2.5was continuously collected throughout January-April by high-volume sampling (n = 40 samples) at the Maldives Climate Observatory at Hanimaadhoo (MCOH) as part of the South Asian Pollution Experiment, 2018 (SAPOEX-18)11. During these months, the MCOH enables sampling of highly polluted plumes originating from the Indian subcontinent1,11, and occasional pristine air from the Southern Indian Ocean27,28. Polluted air in this geographical hotspot leads to millions of premature deaths1,2,4and the associated ‘Atmospheric Brown Cloud’ extends south of the equator, influencing atmospheric energy balances over a vast region29. Previous studies of PM in this region have highlighted the major fraction of sunlight-absorbing BrC30.
To achieve a broad molecular characterization of organic compounds, each PM2.5sample was extracted with a range of solvents and analyzed by gas-chromatography (GC)-HRMS electron ionization (EI) and negative chemical ionization (NCI), or by high-performance liquid chromatography (LC)-HRMS electrospray ionization (ESI) in positive and negative mode. This approach resulted in six unique molecular profiles per sample (Fig. 1a), and revealed high molecular complexity (Fig. 1b–d). Known anthropogenic contaminants were confirmed (Level 1), including legacy persistent organic pollutants, polycyclic aromatic hydrocarbons (PAHs), plasticizers, pesticides, and associated transformation intermediates. However, these identifications represented only a minor portion of all detected molecules. Hence, by integration of open-source cheminformatics and computational workflows, including molecular networking25, MS2-guided in silico structural predictions31, and physicochemical property estimation of optical and toxicological relevance32,33,34, we proceeded to characterize thousands of structures among the remaining unknown molecules. The molecular properties of these structures were evaluated with consideration of potential health and climate impacts.
