Complex species and strain ecology of the vaginal microbiome from pregnancy to postpartum and association with preterm birth

Abstract

Background: Lactobacillus was described as a keystone bacterial taxon in the human vagina over 100 years ago. Using metagenomics, we and others have characterized lactobacilli and other vaginal taxa across health and disease states, including pregnancy. While shifts in community membership have been resolved at the genus/species level, strain dynamics remain poorly characterized.

Methods: We performed a metagenomic analysis of the complex ecology of the vaginal econiche during and after pregnancy in a large U.S. based longitudinal cohort of women who were initially sampled in the third trimester of pregnancy, then validated key findings in a second cohort of women initially sampled in the second trimester of pregnancy.

Findings: First, we resolved microbial species and strains, interrogated their co-occurrence patterns, and probed the relationship between keystone species and preterm birth outcomes. Second, to determine the role of human heredity in shaping vaginal microbial ecology in relation to preterm birth, we performed a mtDNA-bacterial species association analysis. Finally, we explored the clinical utility of metagenomics in detection and co-occurrence patterns for the pathobiont Group B Streptococcus (causative bacterium of invasive neonatal sepsis).

Conclusions: Our highly refined resolutions of the vaginal ecology during and post-pregnancy provide insights into not only structural and functional community dynamics, but highlight the capacity of metagenomics to reveal finer aspects of the vaginal microbial ecologic framework.

Funding: NIH-NINR R01NR014792, NIH-NICHD R01HD091731, NIH National Children's Study Formative Research, Burroughs Wellcome Fund Preterm Birth Initiative, March of Dimes Preterm Birth Research Initiative, NIH-NIGMS (K12GM084897, T32GM007330, T32GM088129).

Keywords: 16S rRNA; GBS; Lactobacillus; bacteria; metagenomics; microbiome; preterm birth; strains; streptococci; vagina.

Figure 1.

Characterization of the vaginal microbiome using WGS metagenomics and targeted 16S rRNA gene amplicon sequencing. (A-B) Relative abundance of the top twenty most prevalent microbial species of the vaginal microbiome based on WGS metagenomics and targeted 16S rRNA gene amplicon analysis. (A) WGS metagenomics derived species relative abundances. The 20 most prevalent taxa contribute 94 percent of the total abundance. (B) 16S rRNA ASV derived species level relative abundances. The 20 most prevalent taxa contribute 96 percent of the total abundance. Samples/columns are grouped according to k-means cluster membership and rows are clustered hierarchically using complete linkage. Column annotations indicate race/ethnicity, time point that sample was acquired, vaginal subsite (for WGS metagenomics), and term/preterm outcome; NA indicates not available. (C-D) Vaginal microbiome community structure. (C) Multidimensional scaling (MDS) ordination of Bray-Curtis distance of WGS vaginal samples supports the k-means clustering of five distinct communities (PERMANOVA, p = 0.001; pairwise PERMANOVA, all p = 0.001). (D) MDS ordination of Bray-Curtis distance of 16S rRNA vaginal samples at the species level supports the k-means clustering of ten clusters (PERMANOVA, p = 0.001; pairwise PERMANOVA, all p < 0.05). Samples are color coded according to cluster membership. Arrows/black dots denote landmark samples – samples with the highest relative abundance (in parentheses) of the indicated species. The percentage of variation explained by each axis is shown in parentheses on the x- and y-axis. Ellipses represent 95% confidence intervals.

Figure 2.

Transitions between and within predominant species clusters during pregnancy and at postpartum. (A) Transitions from third trimester to delivery. Limited transitions occurred from the third trimester to delivery between the predominant species (k-means) clusters. (B) Transitions from delivery to postpartum. A majority of Lactobacillus dominated participants transition to the mixed community from delivery to postpartum. (C) Transitions between and within predominant species clusters during the perinatal period (during pregnancy and at postpartum) within vaginal subsites. Edge widths and colors represent transition probabilities. The mixed L. jensenii/L. iners cluster is denoted here as L. jensenii.

Figure 3.

Probabilistic model of species co-occurrence demonstrates Lactobacillus spp. are relatively exclusionary. (A) Global network of significant positive and negative co-occurrences for species from the vagina. (B) Significant pair-wise co-occurrence networks for select keystone/predominant vaginal bacterial species. Orange dashed lines represent significant negative co-occurrences (p ≤ 0.05), solid blue lines represent significant positive co-occurrences (p ≤ 0.05), line widths represent the strength of the co-occurrence, with node sizes scaled within each subplot according to the average relative abundance across all samples.

Figure 4.

Species co-occurrences during pregnancy (A, left panel) and at postpartum (A, right panel). (B) Taxonomic associations with preterm birth. Subjects delivering preterm were enriched for G. vaginalis, A. vaginae, and other BV-associated taxa, whereas subjects delivering at term were enriched for Lactobacillus ^spp. (C) The effect size for differentially enriched taxa based on preterm birth outcomes was calculated from the average relative abundance derived from WGS metagenomics during pregnancy (3^rd trimester and at delivery) via LEfSe.

Figure 5.

Associations of the vaginal microbiome (WGS metagenomics) with mitochondrial DNA polymorphisms in the context of preterm birth. (A) Manhattan plot demonstrating significant associations between identified taxa and mtSNPs as determined by PLINK associations. (B) Identification of taxa-SNP associations significantly different in the context of preterm birth. Q-values for taxa-SNP associations are plotted on the x-axis, while q-values for quantitative trait interaction (taxa-SNP-preterm birth) are plotted on the y-axis. (C) Relative abundance of vaginal species identified as significantly different between subjects with term and preterm deliveries. Horizontal red bars represent group means.

Figure 6.

Keystone species are present as multiple strains. (A) Principal coordinate analysis (PCoA) of the binary Jaccard distances of pangenome centroids for vaginal samples and reference strains of keystone species. Bacterial reference strains are colored by their respective cluster. Vaginal samples are represented as filled black circles. The percentage of variation explained by each axis is shown in parentheses. Ellipses show the 95% confidence intervals for the reference strains. (B) Species and strain level co-occurrences for pairwise complete observations. (C) Strain-specific functional capacity in vaginal samples during the perinatal interval. Relative proportions of strain-specific metabolism, cellular processes, and environmental information processing KEGG pathways for G. vaginalis, L. crispatus, and L. jensenii.

Figure 7.

Metagenomic identification of GBS. (A) GBS reference genome coverage. Samples from participants with positive GBS clinical cultures are shaded red and negative GBS clinical cultures are shaded green, circle size represents GBS relative abundance. Representative samples annotated i-iii indicate the i) sample with the highest relative abundance of GBS that had a positive clinical culture, ii) sample with a zero-relative abundance of GBS that had a negative clinical culture, iii) sample with the highest relative abundance of GBS that had a negative clinical culture. (B) Genomic coverage of reference genome 2603V/R binned at 1 kb for representative samples i-iii from panel A. Positions of 16S genes (16S) are highlighted by the dashed red lines and the CAMP factor gene (cfb) is highlighted by the dashed purple line. (C) GBS positively co-occurs with bacterial vaginosis associated species and differentially co-occurs with Lactobacillus spp. Positive co-occurrence with L. iners, but a negative co-occurrence (exclusion) with L. crispatus.