In silico Prediction of Virus-Host Interactions for Marine Bacteroidetes With the Use of Metagenome-Assembled Genomes

Bacteroidetes is one of the most abundant heterotrophic bacterial taxa in the ocean and play crucial roles in recycling phytoplankton-derived organic matter. Viruses of Bacteroidetes are also expected to have an important role in the regulation of host communities. However, knowledge on marine Bacteroidetes viruses is biased toward cultured viruses from a few species, mainly fish pathogens or Bacteroidetes not abundant in marine environments. In this study, we investigated the recently reported 1,811 marine viral genomes to identify putative Bacteroidetes viruses using various in silico host prediction techniques. Notably, we used microbial metagenome-assembled genomes (MAGs) to augment the marine Bacteroidetes reference genomic data. The examined viral genomes and MAGs were derived from simultaneously collected samples. Using nucleotide sequence similarity-based host prediction methods, we detected 31 putative Bacteroidetes viral genomes. The MAG-based method substantially enhanced the predictions (26 viruses) when compared with the method that is solely based on the reference genomes from NCBI RefSeq (7 viruses). Previously unrecognized genus-level groups of Bacteroidetes viruses were detected only by the MAG-based method. We also developed a host prediction method based on the proportion of Bacteroidetes homologs in viral genomes, which detected 321 putative Bacteroidetes virus genomes including 81 that were newly recognized as Bacteroidetes virus genomes. The majority of putative Bacteroidetes viruses were detected based on the proportion of Bacteroidetes homologs in both RefSeq and MAGs; however, some were detected in only one of the two datasets. Putative Bacteroidetes virus lineages included not only relatives of known viruses but also those phylogenetically distant from the cultured viruses, such as marine Far-T4 like viruses known to be widespread in aquatic environments. Our MAG and protein homology-based host prediction approaches enhanced the existing knowledge on the diversity of Bacteroidetes viruses and their potential interaction with their hosts in marine environments.