Cervicovaginal microbiome and natural history of HPV in a longitudinal study

Abstract

Background: Human papillomavirus (HPV) infection is one of the most common sexually transmitted infections. However, only a small percentage of high-risk (HR) HPV infections progress to cervical precancer and cancer. In this study, we investigated the role of the cervicovaginal microbiome (CVM) in the natural history of HR-HPV.

Methods: This study was nested within the placebo arm of the Costa Rica HPV Vaccine Trial that included women aged 18-25 years of age. Cervical samples from two visits of women with an incident HR-HPV infection (n = 273 women) were used to evaluate the prospective role of the CVM on the natural history of HR-HPV. We focus specifically on infection clearance, persistence, and progression to cervical intraepithelial neoplasia grade 2 and 3 (CIN2+). The CVM was characterized by amplification and sequencing the bacterial 16S V4 rRNA gene region and the fungal ITS1 region using an Illumina MiSeq platform. OTU clustering was performed using QIIME2. Functional groups were imputed using PICRUSt and statistical analyses were performed using R.

Results: At Visit 1 (V1) abundance of Lactobacillus iners was associated with clearance of incident HR-HPV infections (Linear Discriminant Analysis (LDA)>4.0), whereas V1 Gardnerella was the dominant biomarker for HR-HPV progression (LDA>4.0). At visit 2 (V2), increased microbial Shannon diversity was significantly associated with progression to CIN2+ (p = 0.027). Multivariate mediation analysis revealed that the positive association of V1 Gardnerella with CIN2+ progression was due to the increased cervicovaginal diversity at V2 (p = 0.040). A full multivariate model of key components of the CVM showed significant protective effects via V1 genus Lactobacillus, OR = 0.41 (0.22-0.79), V1 fungal diversity, OR = 0.90 (0.82-1.00) and V1 functional Cell Motility pathway, OR = 0.75 (0.62-0.92), whereas V2 bacterial diversity, OR = 1.19 (1.03-1.38) was shown to be predictive of progression to CIN2+.

Conclusion: This study demonstrates that features of the cervicovaginal microbiome are associated with HR-HPV progression in a prospective longitudinal cohort. The analyses indicated that the association of Gardnerella and progression to CIN2+ may actually be mediated by subsequent elevation of microbial diversity. Identified features of the microbiome associated with HR-HPV progression may be targets for therapeutic manipulation to prevent CIN2+.

Trial registration: ClinicalTrials.gov NCT00128661.

Fig 1. HPV natural history.

The natural history of HR-HPV is depicted. Briefly, an incident HR-HPV infection can occur by entering the basal layer through an epithelial abrasion. Most incidence infections are cleared, however some remain persistent for years and decades. Persistence of a HR-HPV infection combined with known risk factors (e.g., smoking) may allow the persistent HR-HPV infection to progress to precancer (cervical intraepithelial neoplasia, CIN). If the lesion does not regress and the HR-HPV is able to successfully integrate into the host-cell genome, clonal expansion may occur and result in an invasive cancer.

Fig 2. Bacterial Shannon diversity by visit.

HR-HPV category specific microbial Shannon diversity is shown for V1 and V2. Horizontal strip labels at the top of the figure indicate visit number. V1 has an elevated diversity in the progression group, but the overall trend did not achieve statistical significance (p = 0.52). At V2 the observed trend of a rising Shannon alpha diversity from clearance to persistence to progression was statistically significant, p = 0.024.

Fig 3. Bacterial and fungal communities within the study cohort.

A. The abundance plot represents the bacterial community structure of the study subjects. The operational taxonomic units (OTUs) were collapsed at the species level and the top 10 species are presented. Figure boxes labeled: L. iners, L. crispatus, Gardnerella and Diversity represent the vaginal community state types (CSTs) identified using hierarchical clustering. B. Heatmap showing the top 20 fungal species identified within the study subjects. C. albicans has the highest mean abundance. There were three vaginal fungal community states identified using hierarchical clustering as indicated by the separate boxes.

Fig 4. Bacteria associated with progression to CIN2+ identified using LEfSe.

HR-HPV bacterial biomarkers for visit 1 (panel A) and visit 2 (panel B), comparing clearance vs. progression, were identified using the LEfSe tool. Only the significant bacterial taxa (LDA>2.0) are shown for both visits.

Fig 5. Generalized Linear Model (GLM) results showing the odds ratios of key microbial components in association with progression to CIN2+.

The forest plot shows the results of variables evaluated in the univariate analysis that were then entered into a GLM. The model shows ORs (small circle) and 95% confidence interval (line extending on either side of the circle) of the microbial features associated with clearance/progression at either Visit 1 (V1) and/or Visit 2 (V2). The main variables included V1 Gardnerella, V1 Lactobacillus, V1 Fungal Observed OTUs and V1 Cell Motility and V2 Shannon diversity. The model was adjusted for age, bacterial CSTs, smoking and HPV16 infection status. 95% CIs that did not cross the Odds Ratio of 1.0 (dotted vertical line) are considered statistically significant.

Fig 6. Diversity model for HPV progression with mediation analysis.

Panel A shows the results of the mediation analysis that focus on V1 Gardnerella and V2 Shannon diversity. Top row shows Average Causal Mediation Effect (ACME) which is the full mediation effect of V2 Shannon diversity after adjusting for the direct effect of V1 Gardnerella on case status. The second row shows Average Direct Effect (ADE) which is the direct effect of Gardnerella on the clearance/progression outcome after accounting for the mediation effect of V2 Shannon diversity. The third row shows the Total Effect which is the direct, unadjusted effect, of Gardnerella on case outcome. The last row shows the Proportion (Prop.) Mediated, which is the proportion of the model that is mediated by V2 Shannon diversity. Based on GLM modeling, we propose the above model (Panel B) in which V1 Gardnerella causes an expansion of bacterial diversity at V2, which acts as a risk factor for the progression of a HR_HPV infection into a CIN2+ lesion.