Exposome profiling reveals widespread environmental pollutant exposure in seminal plasma and previously unknown associations with male fertility.

There is evidence that indicators of male fertility are in decline globally, but the underlying causes have yet to be fully elucidated. While the environment is likely a major contributor, our current knowledge of environmental determinants of male fertility does not explain this phenomenon. This study aimed to examine and profile organic pollutants in seminal plasma, including both known priority environmental chemicals and previously uncharacterized chemicals, and to discover previously uncharacterized male reproductive environmental toxicants. Semen samples were collected from male partners from 100 couples undergoing assisted reproductive treatment at the Sheba Medical Center (Israel) after 2-7 day abstinence. Semen parameters were assessed for sperm concentration, percent motility, and total motile sperm. Using a QuEChERS (quick, easy, cheap, efficient, rugged, and safe) extraction method, targeted and non-targeted organic pollutant exposures were measured from seminal plasma using gas chromatography. We used linear regression for individual exposure modeling, principal component pursuit (PCP) to remove noise and identify latent patterns in exposure data, and Bayesian Kernel Machine Regression (BKMR) to model multiple pollutants simultaneously. In individual exposure models, we corrected for multiple testing via false discovery rate (FDR). We detected 118 of 119 organic pollutants in our targeted panel in ≥1 sample and 814 non-targeted spectral peaks in all samples, showing widespread detection of organic pollutants in seminal plasma. We used PCP, a machine learning pattern recognition approach, on our targeted panel and derived a low-rank matrix in which one component (explaining 17.4% of the variance in the data) was both driven by etriadizole, a common pesticide, and associated with total motile sperm (p<0.001) and concentration (p=0.03). This was confirmed by the exposome-wide association modeling approach using individual chemicals, where we found that etriadizole was negatively associated with total motile sperm (FDR q=0.01) and concentration (q=0.07). Using PCP on 814 non-targeted spectral peaks identified a component that was associated with total motile sperm (p=0.001). BKMR identified one principal driver of this association, which was analytically confirmed to be N-Nitrosodiethylamine (level-1 confidence) and consistent with linear models (p=0.01). Among the many detectable chemicals in seminal plasma, we identified etridiazole and N-nitrosodiethylamine as previously uncharacterized potential reproductive toxicants. We show that our approach, combining novel machine learning methods and modern statistical models, is consistent with results obtained from traditional statistical approaches, but is far more efficient in identifying associations using realistic sample sizes.