Untargeted NMR-based metabolomics for field-scale monitoring: Temporal reproducibility and biomarker discovery in mosquitofish (Gambusia holbrooki) from a metal(loid)-contaminated wetland.

There is considerable interest in applying omics techniques, which have proven extremely valuable for laboratory-based toxicology studies, towards field-scale ecotoxicology and environmental monitoring. Concerns that confounding factors in natural ecosystems may exacerbate variability in omics datasets must be addressed to validate the transition from laboratory to field. This study explores how temporal variability related to seasonal and climatic trends influence qualitative and quantitative metabolomics outcomes, in fish from reference and metal(loid)-polluted wetlands in Australia. Female mosquitofish (Gambusia holbrooki) were sampled on two separate occasions, from a rehabilitated tailings wetland at the site of historic antimony (Sb) processing and a reference wetland with comparable water quality. The first sampling coincided with greater monthly rainfall and colder water temperature, whereas the second sampling was drier and water was warmer. Despite temporal changes and associated differences in metal(loid) concentrations, site differences in metabolite profiles were qualitatively very similar between sampling events. However, quantitative differences were observed, with a greater number of significantly altered metabolites identified during the second sampling event, which coincided with greater metal(loid) concentrations in both water and fish. The majority of identified metabolites were elevated in fish from the contaminated wetland, but with notable decreases in several metabolites that are known to play a role in various aspects of metal(loid) binding, detoxification and excretion. Specifically, decreased aspartate, histidine, myo-inositol, taurine and choline were observed in fish from the contaminated wetland, and may therefore represent a metabolite suite that is broadly indicative of metal toxicity. Quantitative differences between sampling events are suggestive of a dose-response relationship observable at the cellular level which, if harnessed, may be useful for assigning levels of concern based on the degree of change in a multi-parameter set of metabolite biomarkers.