Patient cohort and characteristics
We recruited 106 women undergoing excisional treatment; three became pregnant after treatment and before the 6 months follow-up visit and were excluded from further analysis. We recruited a further 39 healthy controls with normal cytology.
The patient characteristics across all analyses are detailed in Table 1 & Supplementary Table 1. The average time to follow-up was six months (range: 5–12 months). The control group were found to be younger than the treated women at the time of follow-up (p=0.013). Other characteristics were not significantly different between the compared groups. There was equal distribution of samples collected in the follicular or luteal phase of the cycle and in the rate of women that had intercourse within 48 hours from sample collection between all three groups. Two women were retreated due to the presence high-grade disease at 6 months. The majority were treated for high-grade CIN (CIN1: 15/103, 15%; CIN2: 40/103, 39%); CIN3: 48/103, 46%).
In total, 2977082 reads were obtained from 245 samples with an average number of reads per sample of 11,400 and the mean and median read lengths of 513 and 520 bp respectively. To avoid sequencing bias, operational taxonomic units (OTUs) were sub-sampled to the lowest read count of 296, which retained 99.7% of OTU counts and still provided coverage of >97.9% for all samples. Following removal of singletons and rare OTUs, a total of 77 taxa were identified in the vaginal microbiota of the study cohort.
Three of the five CST’s previously described by Ravel and colleagues [36] were observed in the study cohort, CST I (L. crispatus), III (L. iners) and IV (High-diversity, Lactobacillus-deplete) (Figure 2A).
The vaginal microbiota composition
Pre-treatment versus normal controls – Analysis 1
A significantly greater rate of CST IV VMB composition was observed in our cohort of 103 women undergoing excisional treatment for CIN when compared to 39 healthy untreated controls (Pre-treatment CST IV; 21/103, 20% vs controls; 1/39, 3%; p=0.0081). There was no significant difference between the rates of CST I (Pre-treatment CST I; 45/103, 44% vs controls; 22/39, 56%; p=0.1920, chi-squared test) or CST III (Pre-treatment CST III; 37/103, 36% vs controls; 16/39, 41%; p=0.6978, chi-squared test) (Figure 2B, Table 2). Consistent with increased rates of CST IV in the pre-treatment cohort compare to controls, vaginal microbiota richness as measured by number of species observed, and diversity, represented by the inverse Simpson index were also found to be higher in women pre-treatment (p=0.0003 and p<0.0001 respectively, chi-squared test) (Figure 2C & D). LEfSe analysis demonstrated a significantly increased abundance of Prevotella bivia and Sneathia amnii in women prior to treatment compared to controls, who had significantly greater levels of the Lactobacillus genus (Figure 3A).
Pre- versus post-treatment – Analysis 2
Excisional treatment did not have any significant impact on the VMB composition when comparing paired samples from 103 women on the day of treatment and at 6-month follow-up. Proportions of each of the three CST’s was not significantly different at the two timepoints (CST I; p=0.5715, CST III; p=0.3922, CST IV; p=0.8604, chi-squared test) (Figure 2, Table 2) There was no difference between richness or diversity between pre- and post-treatment samples in the 103 matched women (Figure 2C & D). LEfSe analysis did not detect any differentially abundant taxa between the two groups. The dynamics of the distribution of CST’s pre- and 6-months post-treatment are shown in a Sankey plot (Figure 4). Of the 103 treated patients, 58 (56%) had the same vaginal microbiota CST on the day of treatment and at their 6-month follow-up appointment. Women with CST III were most likely to have the same CST before and after treatment (22/37, 59%), followed by CST I (26/45, 58%) and finally CST IV (10/21, 48%). Of the 45 women who switched to another CST, the majority switched between Lactobacillus spp.-dominant states (25/45, 56%). Nine women switched from a Lactobacillus spp.-dominant CST to CST IV (9/45, 20%) and 11 changed in the opposite direction from CST IV to a Lactobacillus spp.-dominant CST (11/45, 24%).
There was no evidence to suggest that more extensive excisions with larger cone length affected the vaginal microbiota. The CST distribution of rates and dynamics was similar when comparing the upper versus the lower 50th percentile for absolute or proportional length excised and proportional cervical length regeneration (Supplementary Table 2). The mean absolute and proportional cone length and cervix were similar for all CSTs (Supplementary Table 3). We performed the same analysis for absolute and proportional volume excised and found no difference (data not shown).
Post-Treatment versus normal controls – Analysis 3
Compared to controls, CST IV type VMB remained more prevalent in women post-CIN excision at the 6-month follow up (post-treatment CST IV; 19/103, 18% vs controls; 1/39, 3%; p=0.0142, chi-squared test). There was no difference in rates of CST III (44/103, 43% vs controls; 16/39, 41%, p=1.000, chi-squared test), and there was a trend for slightly higher rates of CST I in controls which was approaching significance (Post-treatment CST I; 40/103, 39% vs controls; 22/39, 56%; p=0.0873, chi-squared test) (Figure 2, Table 2). Similarly, richness and diversity remained significantly higher post-treatment compared to controls (p=0.0003 and p<0.0001 respectively, unpaired t-test) (Figure 2C & D). LEfSe analysis showed that BVAB2, Prevotella bivia and Veillonelaceae OTU were significantly overrepresented in the treated cohort compared to controls who had greater levels of Lactobacillus crispatus present (Figure 3B).
To explore whether it is the disease or inherent factors that drive changes in VMB composition, we restricted our subgroup analyses to post-treatment women with negative cytology (Subgroup 3A) and negative HPV and cytology (Subgroup 3B) (Table 2, Supplementary Table 1, Supplementary Figure 2). For sub-analysis 3A we removed 22 treated women with abnormal cytology, leaving 81 treated women versus 39 normal cytology controls. The rate of CST IV after treatment compared to controls remained to be significantly higher (post-treatment CST IV; 14/81, 17% vs controls; 1/39, 3%; p=0.0354, chi-squared test) (Table 2). CST I and III rates were not significantly different between the two groups (p=0.1166 and p=0.8463 respectively, chi-squared test). LEfSe analysis again found Lactobacillus crispatus to be significantly enriched in untreated controls compared to the treated group, and the same three species; BVAB2, Prevotella bivia and Veillonelaceae OTU, all known to be associated with a high-diversity bacterial vaginosis-type VMB composition were significantly enriched in treated women (Supplementary Figure 1A). For sub-analysis 3B, we removed 33 women leaving 70 women with negative HPV and cytology and 16 HPV positive controls leaving 23 HPV and cytology negative controls. In treated women, the rate of CST IV remained almost four times that of healthy controls (post-treatment CST IV; 13/70, 19% vs controls; 1/23, 4%; p=0.1758, chi-squared test), but was no longer significant, likely due to small sample size. Relative abundance of Lactobacillus crispatus remained higher in samples from untreated controls compared to treated women according to LEfSe analysis, with the latter enriched for Atopobium vaginae and a Veillonelaceae OTU (Supplementary Figure 1B).
hBD1 and SLPI levels
The levels of hBD1 and SLPI were normalised to total protein levels in 80 treated women and 34 controls (Supplementary Table 4, Figure 5). Levels for both peptides were higher prior to treatment compared to healthy controls (hBD-1; p=0.0033, SLPI; p=0.0006; unpaired t-test, Analysis 1). The paired samples pre- and post-treatment showed a significant reduction for both (hBD-1 p<0,0001, SLPI p<0.0001; paired t-test, Analysis 2). The levels fell significantly after treatment and were significantly lower than healthy controls (hBD-1 p<0.0029, SLPI p=0.0382; unpaired t-test, Analysis 3). In the comparison of only post-treatment cytology negative women (n=65), AMP levels were significantly less in treated women compared to the healthy controls (n=34) (hBD-1 p=0.0003 and SLPI, p=0.0009; unpaired t-test, Subgroup Analysis 3A), In the comparison of HPV and cytology negative women (n=50) to HPV negative cytology controls (n=20), the same observations were seen (hBD-1 p=0.0006 and SLPI, p=0.0016; unpaired t-test, Subgroup analysis 3B). There was no correlation between levels of hBD1 or SLPI and change in cervical length, volume, or according to proportion of length or volume excision (data not shown). Immunohistochemical staining revealed that expression for both hBD-1 and SLPI was strongest in the glandular epithelium and areas of high-grade CIN when compared to normal tissue within the same sections. For two women that required a second treatment, the scarred epithelium of the repeat cone exhibited weak staining for both peptides compared to the previously untreated epithelium (Supplementary Figure 2).
Proinflammatory cytokine levels
The levels of proinflammatory cytokines were measured in 80 treated women and 34 healthy controls and normalised to levels of total protein in each sample (Supplementary Table 5, Figure 6). Only four out of the 10 cytokines had detectable levels. The proinflammatory cytokines IL-1 and IL-8 were both significantly elevated prior to treatment compared to healthy controls (p<0.0001 and p=0.0014 respectively, unpaired t-test), did not change after excision and maintained significantly higher levels post-treatment than controls (p<0.0001 and p=0.0035 respectively, unpaired t-test). The length of excision and regeneration did not impact on concentration (data not shown). The same patterns were seen for sub-analyses 3A and 3B with IL-1 and IL-8 both being higher pre-treatment compared to controls and remaining elevated after treatment. TNF- was also elevated prior to treatment compared to controls but this was not significant (p=0.62, unpaired t-test), and after treatment levels fell although this was not significant (p=0.10, paired t-test) when compared to pre-treatment levels. The levels of TNF- post-treatment were similar to those seen in untreated controls (p>0.99, unpaired t-test). IFN- was significantly lower prior to treatment compared to controls (p=0.01, unpaired t-test), whilst levels after treatment were significantly higher than before (p=0.002, paired t-test) and similar to levels seen in healthy controls. IL-2, IL-4, IL-6, IL-10, MIP-1 and RANTES were below the lower limit of quantification in the majority of samples.
Correlation between bacterial species, cytokines and antimicrobial peptides.
Correlation analysis was performed to further understand the interplay between the bacterial species present and expression of cytokines and antimicrobial peptides. We found that anaerobic species were positively correlated with expression of antimicrobial peptides and IL-1 (Supplementary Figure 4). In the subgroup analysis comparing only to women that were HPV and cytology negative at follow-up (56 women), we found even stronger correlations. Anaerobes such as Prevotella bivia (p=0.02, Pearson’s correlation coefficient), Atopobium vaginae (p=0.005, Pearson’s correlation coefficient), BVAB2 (p<0.001, Pearson’s correlation coefficient), Streptomyces sp. (p<0.001, Pearson’s correlation coefficient) and Gemella asaccharolytic (p<0.001, Pearson’s correlation coefficient) were most positively correlated with AMP expression prior to treatment, all of which were also positively correlated with expression of IL-1 (p<0.001, Pearson’s correlation coefficient). After treatment there is no significant change in bacterial composition, but AMP expression is no longer correlated with these bacterial species (Figure 7).