Cervicovaginal Human Papillomavirus Genomes, Microbiota Composition and Cytokine Concentrations in South African Adolescents

Abstract

The interaction between cervicovaginal virome, bacteriome and genital inflammation has not been extensively investigated. We assessed the vaginal DNA virome from 33 South African adolescents (15-19 years old) using shotgun DNA sequencing of purified virions. We present analyses of eukaryote-infecting DNA viruses, with a focus on human papillomavirus (HPV) genomes and relate these to the vaginal bacterial microbiota (assessed by 16S rRNA gene sequencing) and cytokines (assessed by Luminex). The DNA virome included single-stranded (Anelloviridae, Genomoviridae) and double-stranded DNA viruses (Adenoviridae, Alloherpesviridae, Herpesviridae, Marseilleviridae, Mimiviridae, Polyomaviridae, Poxviridae). We identified 110 unique, complete HPV genomes within two genera (Alphapapillomavirus and Gammapapillomavirus) representing 40 HPV types and 12 species. Of the 40 HPV types identified, 35 showed positive co-infection patterns with at least one other type, mainly HPV-16. HPV-35, a high-risk genotype currently not targeted by available vaccines, was the most prevalent HPV type identified in this cohort. Bacterial taxa commonly associated with bacterial vaginosis also correlated with the presence of HPV. Bacterial vaginosis, rather than HPV, was associated with increased genital inflammation. This study lays the foundation for future work characterizing the vaginal virome and its role in women's health.

Keywords: Alphapapillomavirus; DNA viruses; Gammapapillomavirus; HPV; Sub-Saharan Africa; genital inflammation; microbiome; vaginal virome; whole genome; women’s health.

Figure 1

Eukaryotic DNA viruses in South African adolescents. Presence (black square) and absence (white square) of eukaryote-infecting families of DNA viruses identified by viral metagenome sequencing after rolling-circle-amplification of viral-nucleic acids extracted from vaginal samples from 33 South African women based on BLASTx analysis of contigs >500 bp against the NCBI viral Refseq database. PID and study visit at which samples were collected (V1 = baseline, V2 = 4 months, V3 = 8 months) are shown. This analysis includes samples from the 33 women included in the viral metagenome sequencing project.

Figure 2

Maximum likelihood phylogeny of L1 amino acid sequences identified in this study. HPV species and types identified in 19 individuals are shown in bold font, and the colored boxes represent the presence of HPVs in that individual. Branches with >0.6 approximate likelihood-ratio test (aLRT) branch support are shown. This analysis includes all complete HPV genomes identified in the viral and whole community metagenome sequencing projects.

Figure 3

Occurrence of HPV genotypes. (A) Heat map showing the positive HPV genotype associations determined by a probabilistic co-occurrence model [36]. Genotype names are positioned to indicate the columns and rows that represent their pairwise relationships with other species. Blue indicates that the co-occurrence of HPV genotypes is significantly large and greater than expected (statistically significant positive association). This analysis included the HPV genotypes identified in the viral metagenome and whole community sequencing projects. The plot only includes the 35 HPV genotypes that had significant associations with one or more other HPV genotypes. (B) The occurrence of the 39 HPV genotypes identified in the 33 participants collected when women had BV (light blue, n = 19) and when BV was absent (dark blue, n = 22) is shown in a cross-sectional comparison. HPV genotypes are annotated by risk classification and species. This analysis includes samples from the 33 women in the viral metagenome sequencing project.

Figure 4

Distinguishing women with HPV versus without HPV based on vaginal microbiota composition. (A) Loadings plot showing the contribution of the different bacterial taxa selected by sparse PLS-DA (sPLSDA) that explain the highest variance in the comparison of samples from women with any HPV (orange) and from women without any HPV (blue). (B) A network analysis of the bacterial taxa selected by sPLSDA. Positive correlations are shown in red, while negative correlations are shown in blue. (C) Unsupervised clustering of the selected bacterial taxa, annotated by HPV status with orange indicating the presence of any HPV and blue indicating the absence of HPV in a sample. This analysis includes samples from the 33 women included in the viral metagenome sequencing project.

Figure 5

Evaluating presence of HPV variants in longitudinal samples from participants who experienced a change in BV status. From eight participants included in the viral metagenome sequencing project longitudinal samples were collected 4 months apart, and HPV genotype presence was evaluated in the absence (purple) and presence (blue) of BV. A variant was defined based on a 5% difference (virus operational taxonomic unit, vOTU), and the analysis was limited to vOTUs that had >75% genome coverage.

Figure 6

Associations between cervicovaginal cytokines and presence of HPV genotypes. (A) In a univariate analysis, the measured concentrations of cytokines were compared in samples of women with a given HPV genotype (labelled 1) to that of samples from women without the given HPV genotype (labelled 0). Significant differences are highlighted using a black box. (B) The same comparisons were made for samples from women with any HPV genotype present versus those without, and those with multiple HPV genotypes, any high- (hr) or any low-risk (lr) HPV genotypes. This analysis includes samples from the 33 women included in the viral metagenome sequencing project.