Metagenomic analysis with strain-level resolution reveals fine-scale variation in the human pregnancy microbiome

Abstract

Recent studies suggest that the microbiome has an impact on gestational health and outcome. However, characterization of the pregnancy-associated microbiome has largely relied on 16S rRNA gene amplicon-based surveys. Here, we describe an assembly-driven, metagenomics-based, longitudinal study of the vaginal, gut, and oral microbiomes in 292 samples from ten subjects sampled every three weeks throughout pregnancy. 1.53 Gb of non-human sequence was assembled into scaffolds, and functional genes were predicted for gene-and pathway-based analyses. Vaginal assemblies were binned into 97 draft quality genomes. Redundancy analysis (RDA) of microbial community composition at all three body sites revealed gestational age to be a significant source of variation in patterns of gene abundance. In addition, health complications were associated with variation in community functional gene composition in the mouth and gut. The diversity of Lactobacillus iners-dominated communities in the vagina, unlike most other vaginal community types, significantly increased with gestational age. The genomes of co-occurring Gardnerella vaginalis strains with predicted distinct functions were recovered in samples from two subjects. In seven subjects, gut samples contained strains of the same Lactobacillus species that dominated the vaginal community of that same subject, and not other Lactobacillus species; however, these within-host strains were divergent. CRISPR spacer analysis suggested shared phage and plasmid populations across body sites and individuals. This work underscores the dynamic behavior of the microbiome during pregnancy and suggests the potential importance of understanding the sources of this behavior for fetal development and gestational outcome.

Molecular Heterogeneity in Large-Scale Biological Data: Techniques and Applications

“…For de novo approaches, reads are aligned against each other and grouped into OTUs. This strategy does not require references but is sensitive to predefined thresholds and the clustering algorithm, and it ignores information on a finer scale such as strain-level differences (132,133). Although it is popular in 16S sequencing, OTUs are rarely used in whole-metagenome shotgun sequencing due to sequencing errors.…”