Gill microbiomes accurately predict the geographic origin of bonefish ( Albula vulpes ) when combined with machine learning models

Understanding movement and dispersal is critical for fisheries management and conservation, but many studies rely on tagging and telemetry, which can be costly and labor-intensive. Recent research on Atlantic bonefish ( Albula vulpes ) has shown that gill microbiomes vary among populations, raising the possibility that microbial communities might prove useful for inferring movement and dispersal. Here, we used bacterial 16S metabarcoding to characterize the structure and composition of the gill microbiomes of 253 A. vulpes from five areas spanning nearly 2000 km throughout their western Atlantic range. We compared gill microbiomes among areas and among nearby locations within areas, and we also compared gill microbiomes to seawater samples collected from the same sampling sites at the same times. We then used machine learning (ML) models to determine whether gill microbiomes accurately predict/recover the location from which a fish was sampled. Gill microbiomes varied among areas and locations, consistent with previous findings, and differed markedly from microbiomes of seawater samples. ML models predicted the geographic origin of a bonefish with 95 and 84% accuracy on the larger scale of area and the smaller scale of location, respectively, demonstrating the utility of the method. Additionally, gill microbiomes from bonefish in Florida were enriched for potentially disease-causing bacterial genera compared to microbiomes from bonefish sampled elsewhere. This result concords with previous findings, suggesting that Florida’s historical bonefish population declines may be linked to health impairments resulting from pollution, runoff, and elevated bacterial inputs into coastal waters from anthropogenic factors such as urbanization and agricultural intensification.