Bacteriophages typically have small genomes 1 and depend on their bacterial hosts for replication 2 . Here we sequenced DNA from diverse ecosystems and found hundreds of phage genomes with lengths of more than 200 kilobases (kb), including a genome of 735 kb, which is-to our knowledge-the largest phage genome to be described to date. Thirty-five genomes were manually curated to completion (circular and no gaps). Expanded genetic repertoires include diverse and previously undescribed CRISPR-Cas systems, transfer RNAs (tRNAs), tRNA synthetases, tRNA-modification enzymes, translation-initiation and elongation factors, and ribosomal proteins. The CRISPR-Cas systems of phages have the capacity to silence host transcription factors and translational genes, potentially as part of a larger interaction network that intercepts translation to redirect biosynthesis to phage-encoded functions. In addition, some phages may repurpose bacterial CRISPR-Cas systems to eliminate competing phages. We phylogenetically define the major clades of huge phages from human and other animal microbiomes, as well as from oceans, lakes, sediments, soils and the built environment. We conclude that the large gene inventories of huge phages reflect a conserved biological strategy, and that the phages are distributed across a broad bacterial host range and across Earth's ecosystems.Phages-viruses that infect bacteria-are considered distinct from cellular life owing to their inability to carry out most biological processes required for reproduction. They are agents of ecosystem change because they prey on specific bacterial populations, mediate lateral gene transfer, alter host metabolism and redistribute bacterially derived compounds through cell lysis 2-4 . They spread antibiotic resistance 5 and disperse pathogenicity factors that cause disease in humans and animals 6,7 . Most knowledge about phages is based on laboratorystudied examples, the vast majority of which have genomes that are a few tens of kb in length. Widely used isolation-based methods select against large phage particles, and they can be excluded from phage concentrates obtained by passage through 100-nm or 200-nm filters 1 . In 2017, only 93 isolated phages with genomes that were more than 200 kb in length were published 1 . Sequencing of whole-community DNA can uncover phage-derived fragments; however, large genomes can still escape detection owing to fragmentation 8 . A new clade of human-and animal-associated megaphages was recently described on the basis of genomes that were manually curated to completion from metagenomic datasets 9 . This finding prompted us to carry out a more-comprehensive analysis of microbial communities to evaluate the prevalence, diversity and ecosystem distribution of phages with large genomes. Previously, phages with genomes of more than 200 kb have been referred to as 'jumbophages' 1 or, in the case of phages with genomes of more than 500 kb, as megaphages 9 . As the set reconstructed here span both size ranges we refer to them simply as 'huge phage...

No abstract

Permanent anoxic layers in natural freshwater basins are rare and of considerable interest to microbial ecologists because of their potential undisturbed climax microbial communities and because of their relationship to an earlier biosphere. Lake Pavin in France provides such an environment. It is unusual because the water column has been stratified for a very long period and there has been a lack of mixing (meromixis) and its anoxic zone is in steady state (2). Despite its unique character, information on the distribution of microbial communities in the anoxic water column of Lake Pavin is limited to a terminal restriction fragment length polymorphism study of populations (16). Lehours et al. (16) found that the structures of both the bacterial and archaeal communities changed with depth. The results suggested that communities at interfaces played a predominant role in the water column. To obtain detailed phylogenetic information on the diverse populations in the differing anaerobic communities, 16S rRNA genes in samples collected at three interface layers in the anoxic water column of Lake Pavin were amplified, cloned, sequenced, and analyzed. Bacterial and archaeal clone libraries were also constructed from a sample collected at a depth of 90 m adjacent to the sediment to determine whether sediment fluxes influenced lake bottom community composition.Sample collection and library construction. Samples from depths of 60 m, 70 m, and 90 m in the water column of Lake Pavin were collected in August 2004 using an 8-liter horizontal Van Dorn bottle; samples were also collected from the sediment-water interface (Inter) at a depth of 92 m using a JenkinMortimer multiple corer (21) (see reference 16 for site characteristics). Microbial samples were prepared on site from water samples (500 ml) by filtration through polycarbonate membrane filters (GTTP; Millipore) (47 mm diameter; pore size, 0.2 m) and stored at Ϫ80°C. The sites sampled in the water column are shown in Fig.