Study population – inclusion and exclusion criteria
We prospectively recruited non-pregnant women attending outpatient gynaecology and bariatric surgery clinics at Imperial College NHS Healthcare Trust between 2013-2016. A subset of this population was scheduled for bariatric surgery. Women were recruited irrespective of age, menopause status, ethnicity, parity, smoking, phase in menstrual cycle and contraception use. Women who were HIV, hepatitis B or C positive, had autoimmune disorders, or had a previous hysterectomy were excluded. Ethical approval was obtained from the National Research Ethics Service Committee London – Fulham (Approval number 13/LO/0126) and the NHS West of Scotland Research Ethics Service Committee (WoSRES) (REC 14/WS/1098). All patients gave informed consent.
Sample collection and processing
Cervicovaginal secretions were collected during the clinic visit from the posterior vaginal fornix with a BBLTM CultureSwabTM containing liquid Amies (Becton Dickinson, Oxford, UK) using a sterile, disposable speculum, without lubricant and immediately stored at -80°C. A second transport microbiology swab (Transwab®) containing Amies gel medium was simultaneously collected for cytokine analysis. In the subset population of women planned for bariatric surgery (Gastric band, Roux-en-Y Gastric Bypass or Vertical Sleeve Gastrectomy), we collected serial vaginal swab samples on the day of surgery and at months 3 and 6 post-surgery (Supplementary Figure 4). Serial fasting blood samples were also collected on the day of surgery and at 6-month follow-up with the aim to correlate changes in vaginal microbiota to four serum markers known to be affected by surgery-induced weight loss and hyperinsulinaemia correction. Blood samples were centrifuged at 4,472 g for 10 minutes and serum collected for freezing and storage in -80°C.
A comprehensive interview and questionnaire were used to obtain all relevant gynaecological, medical and surgical history. Menopause status, type of contraception or HRT use and menstrual cycle phase (follicular or luteal) were documented. Ethnicity was self-reported as Caucasian, Asian, Black or Other.
Whole-Genomic bacterial DNA was extracted from the CultureSwabTM using a QiAmp Mini DNA kit (Qiagen, Venlo, Netherlands) as described previously . The second swab for cytokine analysis was thawed on ice and re-suspended in 350ml phosphate-buffered saline solution with protease inhibitor (5ml/ml; Sigma Aldrich). The suspension was centrifuged at 402g for 2 minutes, and supernatant collected into a new 1.5ml microcentrifuge tube. This centrifugation step was repeated to remove any remaining cellular debris. The cell-free supernatant was stored in -80°C. Negative control swabs (blank, devoid of patient sample) were processed alongside each DNA extraction set. No amplicons were observed following PCR and gel electrophoresis of the negative controls, and these were not subsequently sequenced.
Illumina MiSeq sequencing of 16S rRNA gene amplicons and data processing
The V1-V2 hypervariable regions of 16S rRNA genes were amplified by PCR using a forward and reverse fusion primer, described in detail in Supplementary Methods S1. Bacterial profiling using a MiSeq platform (Illumina, San Diego, CA, USA) was conducted at Research and Testing Laboratory (Lubbock, TX, USA). The 16S rRNA gene sequence data was analysed with bioinformatic software package Mothur  using the MiSeq SOP Pipeline. Sequence reads were quality checked and normalised to the lowest number of reads (n=1,855) and singleton operational taxonomic units (OTUs); samples containing fewer than 10 reads were excluded. OTU taxonomies (from Phylum to Genus) were then determined using the RDP MultiClassifier script to generate the RDP taxonomy. Taxonomy level for species of the OTUs was determined using the USEARCH algorithm with 16S rRNA gene sequences from the cultured representatives from the RDP database . Rare OTUs were defined as those present at less than 10 counts within the entire cohort. Alpha and beta diversity indices were calculated from these datasets with Mothur and R statistical package using the Vegan package.
Serum biomarker and local vaginal cytokine analyses
Four serum markers including oestradiol (pmol/L), insulin (mIU/L), glucose (mmol/L) and sex hormone binding globulin (SHBG, nmol/L) were quantified using ELISA at the Imperial College Healthcare NHS Trust North West London Pathology laboratory (Supplementary Methods S2).
Levels of interleukin (IL)-4, IL-6, IL-8, IL-1β, tumour necrosis factor-alpha (TNFα), interferon-gamma (IFN-γ), and macrophage inflammatory protein-1 alpha (MIP1α) in cell-free cervicovaginal secretion supernatants were determined using the Magnetic Luminex Screening Assay multiplex kit (R&D Systems, Minneapolis, MN, USA) on a MAGPIX Analyzer (Luminex® Corporation, s-Hertogenbosch, Netherlands), as per manufacturer’s instructions. Analytes were chosen based on evidence of inflammatory markers specific to adiposity [88-90].
The population was categorized into two groups of interest for the main analysis, non-obese (BMI <30kg/m2) versus obese (BMI ³30kg/m2). We performed further supplementary analyses to assess results for different obesity status subcategories and according to insulin resistance and diabetic status. The homeostatic model assessment for insulin resistance (HOMA-IR) was calculated by the following formula: the product of fasting insulin (mU/L) multiplied by fasting glucose (nmol/L) divided by 22.5 . We used the 2nd tertile value of HOMA-IR as the cut-off to determine insulin resistance status (at 2.98). Differences in categorical clinical parameters between the two main groups of interest (non-obese versus obese) were assessed using Fisher’s exact test for each of the listed characteristics; age, ethnicity, parity, smoking, menopause, menstrual cycle, use of contraception, HRT use, diabetes status and treatment, and abnormal high vaginal swab (HVS) results.
Significant differences between vaginal microbiota at genera taxonomic level were assessed using the Statistical Analysis of Metagenomic Profiles (STAMP) software package . Dependent on genera hierarchical clustering analysis, Lactobacillus spp. or Gardnerella spp. abundance among selected phenotypic categories was investigated by assigning each patient sample into one of three groups (Lactobacillus-dominant, Gardnerella-dominant, or high diversity VMB). Linear discriminant analysis (LDA) effect size (LEfSe) modelling was used to identify biomarkers based on obesity status, according to relative taxonomic abundance .
At genus taxonomic level, prevalence of each of the three categories relating to Lactobacillus or Gardnerella presence were compared between the two phenotypic categories (obese or non-obese) using Fisher’s exact test. We performed further sub-analyses for different weight categories, as well as by the presence of diabetes and/or insulin resistance status. A sensitivity analysis assessed whether the exclusion of women that had antibiotics less than 2 weeks before sample collection or those disclosing sexual intercourse within 48 hours from sampling would affect the results. We further analysed the results for pre- and post-menopausal women separately and after exclusion of those taking oral hormonal contraception or hormone replacement therapy (HRT).
Expression levels of assayed cytokines were compared according to obesity status, prevalence of each genus group and diversity (non-parametric Shannon Index) at baseline sampling using two-way ANOVA. Where data fell outside the range of the standard curve for each analyte, either the minimal or maximal extrapolated or minimal or maximal value of the standard curve was used, where appropriate. Analyses were performed using Prism 8, p values <0.05 considered significant.
In the subset of women undergoing bariatric surgery, changes in serum markers introduced by weight loss were analysed. We further assessed the impact of surgically-induced weight loss on the prevalence of each of the three genus groups at baseline, month 3 and 6 using McNemar’s Chi square test. We analysed the results for the full cohort and separately for pre- and post-menopausal women, and according to diabetic and insulin resistance status. Transition in vaginal microbiota across genus groups correlating with weight loss from baseline sampling to 6 months post-surgery was analysed for the total bariatric cohort and pre- and postmenopausal women separately. Cytokine and serum marker expression levels were compared between baseline sampling and 6 months after surgically-induced weight loss.