Description
Complementing 16S rRNA gene amplicon sequencing with estimates of total bacterial load can infer absolute bacterial species concentrations as measured by targeted assays in the vaginal microbiome
Florencia A. T. Boshier1, Sujatha Srinivasan1, David N. Fredricks 1,2,3,4, Joshua T. Schiffer1,2,3
1Vaccine and Infectious Disease Division, Fred Hutchinson Cancer Research Center, Seattle
2Clinical Research Division, Fred Hutchinson Cancer Research Center, Seattle
3Department of Medicine, University of Washington, Seattle
4Department of Microbiology, University of Washington, Seattle
A majority of microbiome studies use 16S rRNA gene amplicon sequencing to quantify bacterial taxa as relative abundances. Though this fractional approach can detect disease-associated bacterial variation, fluctuating total bacterial load between samples restricts its ability to reflect absolute abundance of individual species.
We developed a prospective cohort of 20 women with a history of recurring bacterial vaginosis (BV) who self-collected daily vaginal swabs for 60 days. We performed targeted quantitative PCR (qPCR) assays for seven key bacterial species (Atopobium vaginae, BV-associated bacterium 2 (BVAB2), Gardnerella vaginalis, Lactobacillus crispatus, Lactobacillus jensenii, Lactobacillus iners, and Megasphaera), as well as 16S rRNA gene amplicon sequencing on each sample. We then estimated inferred concentrations of absolute bacterial abundance by taking the product of each species' relative abundance and total bacterial load measured by broad range qPCR. All data was log-converted for comparison.
The inferred concentrations correlated well with measurements from targeted qPCR (r = 0.935, p<2.2e-16). Correlations were strong for each species (median r = 0.916, range 0.872 - 0.978, p<2.2e-16). Median error of the inferred concentrations was 1.97 log (IQR = 1.61 log, range: 0 - 7.84 log). Approximately 95% of errors greater than 0.5 log were accounted for by relative abundances below 10%. Minority species were associated with larger mean errors.
When the relative abundance of a given species is sufficiently high (above 10%), inferred concentrations of absolute abundance are reliable proxies for targeted qPCR assays. Because rapid shifts in the vaginal microbiota associated with incident BV often initiate when key species are at low relative abundance, targeted qPCRs are still required to accurately measure bacterial kinetics.
This work is funded by the NIH NIAID STI CRC (U19 AI 113173)