Next generation sequencing to examine associations between vaginal washing and vaginal microbiota: A cohort study

Study population and procedures

We performed a secondary analysis of data collected from 111/116 females participating in the placebo arm of the Preventing Vaginal Infections (PVI) trial, which was a randomized placebo-controlled trial to assess monthly intravaginal metronidazole plus miconazole for reducing BV and vulvovaginal candidiasis (VVC). ⁶ In brief, a total of 234 participants aged 18–45 were enrolled in the US and Kenya from May 2011 to August 2012. Based on the interventions for the trial, only persons assigned female sex at birth and presenting with a vaginal infection at screening (BV, VVC, or Trichomonas vaginalis [TV]) were eligible. Women with symptomatic vaginal infections or TV at the screening visit were treated and instructed to return in 7–28 days if they wished to enroll in the study. At enrollment, participants were randomized to receive either monthly intravaginal metronidazole plus miconazole or placebo for a total of 12 months. Participants were excluded from the current analyses if they did not consent to storage and future use of specimens (N = 1), exited the study early (N = 1), or were lost to follow-up (N = 3). Thus, final study size (N = 111) was based on the number of women enrolled in the placebo arm of the PVI trial who completed the study and consented to future use of specimens.

Vaginal samples were collected at enrollment and every 2 months thereafter for diagnosis of vaginal conditions (e.g., VVC, BV) and storage for future analyses. Data on vaginal washing practices, contraceptive use, date of the last menstrual period (LMP), and sexual practices were collected monthly during face-to-face interviews. Participants were asked two sequential questions to determine vaginal washing status. First, participants were asked “In the past month, did you douche or wash inside your vagina?” Those that answered “yes,” were asked “How far inside your vagina did you wash?” Participants could indicate that washing was external (e.g., performed at the “introitus only,” with substances used for washing passing “no deeper than a fingertip” beyond the introitus) or internal, “beyond the introitus.” Vaginal washing was defined as answering “yes” to the first question and “beyond the introitus” to the second question. Women who met the definition for vaginal washing were asked to indicate all products used for vaginal washing in the past month (choices included water only, soap and water, vinegar and water, antiseptic, detergent, or the option to specify another substance), what they used to wash inside the vagina (finger, bathing flannel/cloth, douche applicator or other) and how frequently they performed vaginal washing per week (<1, 1–7, 8–14, >14).

Written informed consent for participation and storage and future use of specimens was obtained at enrollment. The trial was approved by the human subject’s research committees at the University of Washington (approval number 39507), the University of Alabama at Birmingham (approval number 100827002), and Kenyatta National Hospital (approval number P378/10/2010), and all procedures were in accordance with the revised Helsinki Declaration. The PVI trial was registered at ClinicalTrials.gov (NCT01230814; http://clinicaltrials.gov).

Laboratory procedures

Vaginal samples were stored and transported as previously described. ⁴ Extraction of DNA was performed using the BiOstic Bacteremia DNA Isolation Kit (MoBio, now Qiagen, Carlsbad, CA, USA). Sham swabs were used as controls to monitor for contamination of PCR reagents, and assessment of PCR inhibitors was performed as previously described. ⁹ Relative abundance data were generated by broad-range PCR targeting the V3-V4 region of the 16S rRNA gene. Next generation amplicon sequencing was performed using the Illumina MiSeq platform (San Diego, CA, USA). ¹⁰ Raw data were analyzed using the DADA2 pipeline (Version 2.1.6.0), ¹¹ and resultant sequences were classified using a set of reference vaginal bacteria ¹² and the pplacer tool (v.1.1alpha19). ¹³ Sequences are available in the NCBI Short Read Archive (PRJNA638104). ¹⁴

Detection of Neisseria gonorrhoeae and Chlamydia trachomatis was performed using the Hologic Aptima Combo-2 assay (Hologic, San Diego, CA, USA). Herpes simplex virus type 2 (HSV-2) serostatus was measured by enzyme linked immunosorbent assay (ELISA) (HerpeSelect 2, Focus Diagnostics, Cypress, CA, USA), with positivity defined as optical density >2.1 in Kenyan participants and positive versus negative in US participants. ¹⁵ BV was diagnosed by Gram stain using the method of Nugent and Hillier. ¹⁶ Wet preparations with microscopy were performed to evaluate for the presence of vaginal yeast and TV.

Statistical analysis

Baseline data on age, contraceptive use, exchange of sex for cash or in-kind payments, vaginal washing practices, and prevalence of N. gonorrhoeae, C. trachomatis, and HSV-2 infection are reported from the enrollment visit. Antibiotic treatment at the screening visit could have altered vaginal microbiota at the subsequent enrollment visit. ⁸ Thus, clinical data collected at the baseline visit that may have been influenced by antibiotic treatment at the screening visit (e.g., Nugent score) were excluded, and data collected at the first follow-up visit are reported as the baseline for the present analyses. As with our earlier vaginal washing study in this dataset, analyses were performed for the full cohort and after stratification by country. ⁴

For these analyses, the primary predictor was vaginal washing (defined under “study populations and procedures”) and the primary outcome was the Shannon Diversity Index (SDI) and the Chao 1 index. The SDI is a summary statistic generated based on the number of different bacterial taxa detected by high throughput sequencing (species richness) and the proportion of the entire population comprised by each species (species evenness). ¹⁷ The Chao 1 index estimates the number of taxa (community richness) in the bacterial population. ¹⁸ The SDI and Chao 1 index were calculated using sequence counts of individual bacterial taxa using the R microbiome package (version 3.3.2) at visits where vaginal washing was reported versus not reported. Results were compared using generalized estimating equations (GEE) with a Gaussian link, independent correlation structure, and robust standard errors. Adjusted analyses were performed for known or potential confounders of the association between vaginal washing and vaginal microbiota including country, age, condomless sex, HSV-2 serostatus, and menstrual cycle phase.^4,19–22 Consistent with our prior analyses, menstrual cycle phase was modeled as a categorical variable defined as 0–14 days since the start of the LMP (follicular phase), 15–28 days since the start of the LMP (luteal phase), >28 days since the start of the LMP, or no menstrual period for >3 months (amenorrheic).^4,19,20 The β coefficients generated from unadjusted and adjusted analyses can be interpreted as the mean difference in the SDI or Chao 1 Index between visits with and without vaginal washing.

The association between vaginal washing (primary predictor) and bacterial relative abundances (secondary outcome) was analyzed. Taxa were selected for analyses using two methods. First, a Wald score was calculated using linear regression to screen for associations between bacterial relative abundances and vaginal washing. Taxa with a p-value cut-off of <0.1 by the Wald test and >1% prevalence were carried forward for further analyses. Second, a subset of bacterial taxa that were associated with a higher likelihood of detection by taxon directed qPCR at vaginal washing visits in a prior study were selected for analysis a priori: Fannyhessea (Atopobium) vaginae; Bacterial Vaginosis Associated Bacteria 1 (BVAB1); BVAB2; Gardnerella species; Mageeibacillus indolicus; Megasphaera lornae (formerly Megasphaera species type 1); Megasphaera hutchinsoni (formerly Megasphaera species type 2); Sneathia vaginalis (formerly Leptotrichia amnii); Sneathia sanguinegens; and Sneathia species. ⁴ Because participants contributed data from multiple visits, associations between vaginal washing and bacterial relative abundances were assessed using GEE with a Gaussian link, independent correlation structure, and robust standard errors. A Benjamini-Hochberg false discovery rate (FDR) of 0.1 was applied to all bacterial taxa analyzed. Adjustment for potential confounding factors was performed using the variables described above. The β coefficients derived from these analyses represent the mean difference in bacterial relative abundance between visits exposed versus not exposed to vaginal washing.

A total of 20/111 (18.0%) participants missed 36/666 (5.4%) follow-up visits. Given the low frequency of missing visits, all participants in the placebo arm who completed the study and consented to storage and future use of samples were included in these analyses (N = 111). All statistical analyses were performed using IBM SPSS Statistics (Version 26.0) and R (Version 3.3.2).