Extensive Unexplored Human Microbiome Diversity Revealed by Over 150000 Genomes from Metagenomes Spanning Age Geography and Lifestyle

摘要

class="head no_bottom_margin" id="sec1title">IntroductionDespite extensive recent studies of the human microbiome using a variety of culture-independent molecular technologies (, , , ), most characterization of these ecosystems is still focused on microbes that are easily cultivable, particularly when those with sequenced isolate genomes are considered. Since physiological characterization of diverse, uncharacterized human-associated microbes by cultivation can be difficult in high throughput (), additional approaches are needed that scale with the extent of populations that can now be surveyed using metagenomic sequencing. Culture-independent genomic approaches that are scalable to large cohorts (, , ) have facilitated access to an expanded set of isolation-recalcitrant members of the microbiome, but they also suggested the presence of a large fraction of still unexplored diversity (, ).Here, we present a set of 154,723 microbial genomes that are often prevalent, population specific, and/or geographically specific that we reconstructed via single-sample assembly from a total of 9,428 global, body-wide metagenomes. Other studies have also succeeded in reconstructing microbial genomes by metagenomic assembly on single human cohorts (, , , , , href="#bib72" rid="bib72" class=" bibr popnode">Sharon et al., 2013), but systematic cross-study cataloging of metagenomically assembled genomes focused so far on non-human environments (href="#bib55" rid="bib55" class=" bibr popnode">Oyama et al., 2017, href="#bib58" rid="bib58" class=" bibr popnode">Parks et al., 2017). Complementary techniques, such as co-abundance of gene groups (href="#bib48" rid="bib48" class=" bibr popnode">Nielsen et al., 2014), can identify genomic bins without reference, but these techniques do not account for sample-specific strains and strain-level differences in the sequence reconstruction and thus require downstream single-nucleotide variation analysis on specific genomic regions to uncover strain variability (href="#bib64" rid="bib64" class=" bibr popnode">Quince et al., 2017b, href="#bib80" rid="bib80" class=" bibr popnode">Truong et al., 2017).Using large-scale single-sample metagenomic assembly supported by strict quality control (including filtering based on nucleotide polymorphisms), we identified 3,796 species-level clades (comprising 34,205 genomes) without previous whole-genome information. This identified several taxa prevalent but previously unobserved even in well-profiled populations (e.g., a genus-level Ruminococcaceae clade phylogenetically close to Faecalibacterium), extensive taxonomically uncharacterized species associated with non-Western populations, and the presence of several taxa from undersampled phyla (e.g., Saccharibacteria and Elusimicrobia) in oral and gut microbiomes. The resulting genome set can thus serve as the basis for future strain-specific comparative genomics to associate variants in the human microbiome with environmental exposures and health outcomes across the globe.