Experimental evolution at ecological scales allows linking of viral genotypes to specific host strains.

Viruses shape microbial community structure and activity through the control of population diversity and cell abundances. Identifying and monitoring the dynamics of specific virus-host pairs in nature is hampered by the limitations of culture-independent approaches such as metagenomics, which do not always provide strain-level resolution, and culture-based analyses, which eliminate the ecological background and in-situ interactions. Here, we have explored the interaction of a specific "autochthonous" host strain and its viruses within a natural community. Bacterium Salinibacter ruber strain M8 was spiked into its environment of isolation, a crystallizer pond from a coastal saltern, and the viral and cellular communities were monitored for one month using culture, metagenomics, and microscopy. Metagenome sequencing indicated that the M8 abundance decreased sharply after being added to the pond, likely due to forces other than viral predation. However, the presence of M8 selected for two species of a new viral genus, Phoenicisalinivirus, for which 120 strains were isolated. During this experiment, an assemblage of closely related viral genomic variants was replaced by a single population with the ability to infect M8, a scenario which was compatible with the selection of a genomic variant from the rare biosphere. Further analysis implicated a viral genomic region putatively coding for a tail fiber protein to be responsible for M8 specificity. Our results indicate that low abundance viral genotypes provide a viral seed bank that allows for a highly specialized virus-host response within a complex ecological background.