A Study of the Vaginal Microbiome in Healthy Canadian Women Utilizing cpn60-Based Molecular Profiling Reveals Distinct Gardnerella Subgroup Community State Types

Abstract

The vaginal microbiota is important in women's reproductive and overall health. However, the relationships between the structure, function and dynamics of this complex microbial community and health outcomes remain elusive. The objective of this study was to determine the phylogenetic range and abundance of prokaryotes in the vaginal microbiota of healthy, non-pregnant, ethnically diverse, reproductive-aged Canadian women. Socio-demographic, behavioural and clinical data were collected and vaginal swabs were analyzed from 310 women. Detailed profiles of their vaginal microbiomes were generated by pyrosequencing of the chaperonin-60 universal target. Six community state types (CST) were delineated by hierarchical clustering, including three Lactobacillus-dominated CST (L. crispatus, L. iners, L. jensenii), two Gardnerella-dominated (subgroups A and C) and an "intermediate" CST which included a small number of women with microbiomes dominated by seven other species or with no dominant species but minority populations of Streptococcus, Staphylococcus, Peptoniphilus, E. coli and various Proteobacteria in co-dominant communities. The striking correspondence between Nugent score and deep sequencing CST continues to reinforce the basic premise provided by the simpler Gram stain method, while additional analyses reveal detailed cpn60-based phylogeny and estimated abundance in microbial communities from vaginal samples. Ethnicity was the only demographic or clinical characteristic predicting CST, with differences in Asian and White women (p = 0.05). In conclusion, this study confirms previous work describing four cpn60-based subgroups of Gardnerella, revealing previously undescribed CST. The data describe the range of bacterial communities seen in Canadian women presenting with no specific vaginal health concerns, and provides an important baseline for future investigations of clinically important cohorts.

Fig 1. Vaginal microbiome profiles of Canadian women.

Heatmap representing hierarchical clustering of Jensen-Shannon distance matrices with Ward linkage on the relative proportions of reads for each OTU within individual vaginal samples collected from healthy, reproductive-aged Canadian women (n = 310). Each column represents a woman’s vaginal microbiome profile, and each row represents an OTU. For clarity, only the top 65 OTU by read abundance are shown. The proportion of the total microbiome comprised is indicated in the yellow to red colour scheme. Community state type (CST) and Nugent category (Nugent) for each woman are indicated by the top bars.

Fig 2. Abundance and phylogenetic relationships of cpn60 defined species detected in samples from each CST.

Neighbour-joining phylogeny of 164 unique cpn60 universal target using MEGA v6 for Mac, that are nearest neighbours for OTU detected at an abundance of at least 1% of at least one woman's sample. Branches proceeding from nodes with less than 50% bootstrap percent (100 replicates) are shown in grey. Circle area represents the proportion of each taxon (branch) in total normalized reads from women in different CST. Therefore, the combined area of all circles of the same colour equals the area of the central grey circle that represents 100% of reads attributed to that CST. Only the most abundant taxa are labelled. L. = Lactobacillus, P. = Prevotella, E. = Escherichia

Fig 3. Correspondence between CST and Nugent score.

(A) Significant (Fisher exact test, Benjamini-Hochberg adjusted p < 0.0001) relationship between CST and Nugent score, including predominance of CST I, III and V in BV- samples, and CST IVC and IVD in BV+ samples. BVI samples were observed most frequently in CST IVA. (B) Principal components of microbiome for all members of study group, with individuals coloured by CST, and C) Nugent score.

Fig 4. OTU associated with BV-, BVI and BV+ categories.

Boxplots showing differences in relative abundance (as center log transformed counts) among Nugent categories across 35 OTU that were significantly different among Nugent categories from the ALDeX analysis. These plots show all reads +1 so that zero values could be included in the log transformation. All p-values shown are corrected by the Benjamini-Hochberg method as indicated in the text.