Summary: The coral holobiont is the community of metazoans, protists and microbes associated with scleractinian corals. Disruptions in these associations have been correlated with coral disease, but little is known about the series of events involved in the shift from mutualism to pathogenesis. To evaluate structural and functional changes in coral microbial communities, Porites compressa was exposed to four stressors: increased temperature, elevated nutrients, dissolved organic carbon loading and reduced pH. Microbial metagenomic samples were collected and pyrosequenced. Functional gene analysis demonstrated that stressors increased the abundance of microbial genes involved in virulence, stress resistance, sulfur and nitrogen metabolism, motility and chemotaxis, fatty acid and lipid utilization, and secondary metabolism. Relative changes in taxonomy also demonstrated that coral‐associated microbiota ( Archaea , Bacteria , protists) shifted from a healthy‐associated coral community (e.g. Cyanobacteria , Proteobacteria and the zooxanthellae Symbiodinium ) to a community (e.g. Bacteriodetes , Fusobacteria and Fungi ) of microbes often found on diseased corals. Additionally, low‐abundance Vibrio spp. were found to significantly alter microbiome metabolism, suggesting that the contribution of a just a few members of a community can profoundly shift the health status of the coral holobiont.

The species composition and metabolic potential of microbial and viral communities are predictable and stable for most ecosystems. This apparent stability contradicts theoretical models as well as the viral–microbial dynamics observed in simple ecosystems, both of which show Kill-the-Winner behavior causing cycling of the dominant taxa. Microbial and viral metagenomes were obtained from four human-controlled aquatic environments at various time points separated by one day to >1 year. These environments were maintained within narrow geochemical bounds and had characteristic species composition and metabolic potentials at all time points. However, underlying this stability were rapid changes at the fine-grained level of viral genotypes and microbial strains. These results suggest a model wherein functionally redundant microbial and viral taxa are cycling at the level of viral genotypes and virus-sensitive microbial strains. Microbial taxa, viral taxa, and metabolic function persist over time in stable ecosystems and both communities fluctuate in a Kill-the-Winner manner at the level of viral genotypes and microbial strains.