Shallow shotgun metagenomic sequencing of vaginal microbiomes with the Oxford Nanopore technology enables the reliable determination of vaginal community state types and broad community structures

Abstract

Background: The vaginal microbiome plays an important role in female health; it is associated with reproductive success, susceptibility to sexually transmitted infections, and, importantly, the most prevalent vaginal condition in reproduction-age women, bacterial vaginosis (BV). Traditionally, 16S rRNA gene sequencing-based approaches have been used to characterize the composition of vaginal microbiomes, but shallow shotgun metagenomic sequencing (SMS) approaches, in particular when implemented with the Oxford Nanopore Technologies, have important potential advantages with respect to cost effectiveness, speed of data generation, and the availability of flexible multiplexing schemes.

Results: Based on a study cohort of n = 52 women, of which 23 were diagnosed with BV, we evaluated the applicability of Nanopore-based SMS for the characterization of vaginal microbiomes in direct comparison to Illumina 16S-based sequencing. We observed perfect agreement between the two approaches with respect to detecting the dominance of individual samples by either Lactobacilli, vaginosis-associated, or other taxa; very high concordance (92%) with respect to community state type (CST) classification; and a high degree of concordance with respect to the overall clustering structures of the sequenced microbiomes. Comparing the inferred abundances of individual species in individual samples, we observed significant differences (Wilcoxon signed-rank test p < 0.05) between the two approaches for 12 of the 20 species most abundant in our cohort, indicating differences in the fine-scale characterization of vaginal microbiomes. Higher overall abundance of Gardnerella vaginalis, associated with an increased number of CST IV detections, in the Nanopore shallow SMS data indicated potentially increased sensitivity of this approach to dysbiotic states of the vaginal microbiome. Nanopore shallow SMS also enabled the methylation-based quantification of different human cell types in the characterized samples as well as the detection of non-prokaryotic species, including Lactobacillus phage and Candida albicans in study participants with microscopically detected Candida. One important potential limitation of the evaluated Nanopore-based SMS approach was marked variation in sequencing yields.

Conclusion: Our study demonstrated the successful application and potential advantages of Nanopore-based shallow SMS for the characterization of vaginal microbiomes and paves the way for its application in larger-scale research or diagnostic settings.

Keywords: Community state type; Illumina 16S; Oxford Nanopore Technologies; Sequencing; Shallow shotgun metagenomic sequencing; Vaginal microbiome; Vaginosis.

Fig. 1

Illumina 16S-based characterization of the cohort. A Illumina 16S-based relative abundances; shown are relative abundances of the 20 most abundant species (cumulative abundance across all samples) as well as the cumulative abundance of all other species (“Other”). Filled squares above the heatmap indicate the BV status of study participants (dark blue = BV; light blue = non-BV). Columns are clustered based on Euclidean distances. B Shannon diversity for BV (“Vaginosis”) and non-BV (“Control”) groups after rarefaction to 140,000 reads per sample. C Community State Type (CST) distribution for BV (“Vaginosis”) and non-BV (“Control”) groups. D Principal Coordinates Analysis (PCoA) based on Illumina 16S sequencing and Bray-Curtis inter-sample distances. Dots are colored according to BV status (dark blue = BV; light blue = non-BV). Also shown is a contour plot of the probability density function (KDE) in dashed lines, which represent areas of equal probability density

Fig. 2

Comparison of Nanopore shallow SMS and Illumina 16S assessments of vaginal bacterial communities. A Species-level visualization of the per-sample differences between Nanopore- and Illumina-based abundance estimates, shown for the 20 species exhibiting the highest cumulative abundance across all samples in Illumina 16S data; the “Other” category shows the combined abundance of the remaining species. Two samples were excluded because their prokaryotic read count in the Nanopore data was below the utilized threshold of 200. BV status is indicated by filled squares above the heatmap (dark blue = BV; light blue = non-BV). Columns are arranged in the same order as in Fig. 1A. B Statistical comparison of Nanopore- and Illumina-based abundances of individual species for the species from panel A. Shown are, for each species, mean relative abundance in Nanopore and Illumina data, as well as Spearaman's correlation, Wilcoxon signed-rank p-value, and Hodges-Lehmann estimate of location shift computed for the Nanopore- and Illumina-based abundances of the included species in individual samples. C Confusion matrix for detecting the domination (defined as relative abundance ≥ 50%) of individual samples by lactobacilli, vaginosis-associated taxa (VAT), or other taxa (see main text for definitions). Results are based on Nanopore shallow shotgun metagenomic sequencing (SMS) and Illumina 16S data, for the 50 samples with a prokaryotic read count exceeding the threshold of 200. D Nanopore- and Illumina-based cumulative abundances of Lactobacillus species in individual samples; also shown are Pearson’s r, the corresponding p-value, and a least-squares linear regression line with intercept. E Confusion matrix for detecting the Community State Type (CST) of individual samples based on Nanopore shallow SMS and Illumina 16S; included are the 50 samples with a prokaryotic read count exceeding the utilized threshold of 200. F Joint Principal Coordinates Analysis (PCoA) of Nanopore shallow SMS- and Illumina 16S-based microbiome composition vectors; each biological sample is represented by two dots connected with a line. Also shown is a contour plot of the probability density function (KDE) in dashed lines, which represent areas of equal probability density. G Comparison of Nanopore- and Illumina-based Bray-Curtis inter-sample distances for all 1225 unique pairs of biological samples. Also shown are Pearson’s r, the corresponding p-value, and a least-squares linear regression line

Fig. 3

Human read fraction and presence of non-prokaryotic species determined based on Nanopore shallow SMS. A Fraction of human reads for BV (“Vaginosis”) and non-BV (“Control”) groups; significance was determined with the Mann–Whitney U test and a jitter plot of the distribution of the human reads fraction in the two groups is also shown. B Absolute read counts for detected non-human eukaryotic and viral species; only species detections that passed all verification steps (see text) are included. BV status of the included samples is indicated by filled squares above the heatmap (dark blue = BV; light blue = non-BV)