Human-generated negative impacts on aquatic environments are rising. Despite wild fish playing a key role in aquatic ecologies and comprising a major global food source, physiological consequences of these impacts on them are poorly understood. Here we address the issue through the lens of interrelationship between wild fish and their gut microbiota, hypothesizing that fish microbiota are reporters of the aquatic environs. Two geographically separate teleost wild-fish species were studied (Lake Erie, Ohio, and Caribbean Sea, US Virgin Islands). At each geolocation, fresh fecal samples were collected from fish in areas of presence or absence of known aquatic compromise. Gut microbiota was assessed via microbial 16S-rRNA gene sequencing and represents the first complete report for both fish species. Despite marked differences in geography, climate, water type, fish species, habitat, diet, and gut microbial compositions, the pattern of shifts in microbiota shared by both fish species was nearly identical due to aquatic compromise. Next, these data were subjected to machine learning (ML) to examine reliability of using the fish-gut microbiota as an ecomarker for anthropogenic aquatic impacts. Independent of geolocation, ML predicted aquatic compromise with remarkable accuracy (>90%). Overall, this study represents the first multispecies stress-related comparison of its kind and demonstrates the potential of artificial intelligence via ML as a tool for biomonitoring and detecting compromised aquatic conditions.
Keywords: environment; fish; machine learning; microbiota; stress.