Presence and Persistence of Putative Lytic and Temperate Bacteriophages in Vaginal Metagenomes from South African Adolescents

Abstract

The interaction between gut bacterial and viral microbiota is thought to be important in human health. While fluctuations in female genital tract (FGT) bacterial microbiota similarly determine sexual health, little is known about the presence, persistence, and function of vaginal bacteriophages. We conducted shotgun metagenome sequencing of cervicovaginal samples from South African adolescents collected longitudinally, who received no antibiotics. We annotated viral reads and circular bacteriophages, identified CRISPR loci and putative prophages, and assessed their diversity, persistence, and associations with bacterial microbiota composition. Siphoviridae was the most prevalent bacteriophage family, followed by Myoviridae, Podoviridae, Herelleviridae, and Inoviridae. Full-length siphoviruses targeting bacterial vaginosis (BV)-associated bacteria were identified, suggesting their presence in vivo. CRISPR loci and prophage-like elements were common, and genomic analysis suggested higher diversity among Gardnerella than Lactobacillus prophages. We found that some prophages were highly persistent within participants, and identical prophages were present in cervicovaginal secretions of multiple participants, suggesting that prophages, and thus bacterial strains, are shared between adolescents. The number of CRISPR loci and prophages were associated with vaginal microbiota stability and absence of BV. Our analysis suggests that (pro)phages are common in the FGT and vaginal bacteria and (pro)phages may interact.

Keywords: CRISPR; Sub-Saharan Africa; bacteriophages; lytic; prophages; stability; vaginal microbiota.

Figure 1

The vaginal microbial community composition of South African adolescents. Bar plot summarising the relative abundance of the 30 most abundant bacteria and viruses on kingdom, family, and species level identified by shotgun metagenomic profiling in vaginal samples collected at baseline and week 16 and from participants who did not have any sexually transmitted infections (STIs), including C. trachomatis, N. gonorrhoea, T. vaginalis, and M. genitalium throughout the study period, did not take any antibiotics or medication that might have influenced their microbiota, and did not practice vaginal insertional practices.

Figure 2

Vaginal microbial community clusters in South African adolescents. (A) Metagenomic data from baseline samples of eligible adolescents (n = 13) grouped into community clusters established using Manhattan clustering. Cluster 1 (C1) is dominated by a diverse range of BV-associated bacteria, C2 by L. iners, and C3 by a range of microbes including Lactobacillus and BV-associated bacterial spp. as well as viruses. BV status was assessed by Nugent Scoring. (B) Box-and-whisker plots depicting alpha diversity for each baseline sample (Shannon Index) by community cluster. (C) Alluvial plot showing the change in in community cluster from baseline to week 16 to week 32, coloured by participant ID (PID).

Figure 3

Prokaryote-infecting viruses in the vaginal metagenomes of South African adolescents. (A) Presence (black square) and absence (white square) of prokaryote-infecting viral families identified by shotgun metagenomic profiling of vaginal samples based on BLASTx analysis using the NCBI viral Refseq database. (B) Examples of circular/full-length Siphoviruses identified in adolescents with bacterial vaginosis (BV). Sequences were annotated using RAST and genes colour-coded based on function. Predicted hosts were determined and participant ID (PID) indicate in which participant they were present.

Figure 4

Presence of prophages in vaginal metagenomes of South African adolescents. The heatmap shows the number of “confident” and “likely” prophage-like elements identified by VirSorter in the metagenomes of the 13 South African adolescents included in this study. Host bacteria were identified by aligning the flanking regions of the predicted prophage-like element with the NCBI bacterial Refseq database using BLASTx. Samples are annotated by BV status (assessed by Nugent Scoring), visit (baseline, week 16, and week 32), participant ID (PID), and hormonal contraceptive use at time of sampling (combined oral contraceptive pills [COCP], the injectable NET-EN, none, or the combined contraceptive vaginal NuvaRing^®).

Figure 5

Persistence of prophages in vaginal metagenomes of South African adolescents and association with microbiota composition and stability. (A) Prophage-like elements identified by VirSorter were mapped against all reads in the dataset to evaluate the persistence of prophage-like elements over time. Samples are annotated by predicted bacterial host, BV status (assessed by Nugent Scoring), visit (baseline, week 16, and week 32), and participant ID (PID). (B) Box-and-whisker plots showing the number of prophages per million reads of participants who did not have Nugent-BV (BV-neg; n = 8) or those who had Nugent-BV (BV-pos; n = 4) at the final visit. (C) Prophage number per million reads of participants who remained Nugent-BV negative and those experiencing a change in Nugent-BV status from week 16 to week 32.

Figure 6

Sharing of putative prophages between participants. (A) Ninety-six putative prophages were identified in cervicovaginal secretions of more than one participant with ≥99% nucleotide identity, and the number of shared and unique (only present in a single participant) prophages are displayed by identified bacterial host. (B) The number of participants sharing prophages of a given bacterial host is shown using box-and-whisker plots.

Figure 7

Diversity of Lactobacillus prophages identified in vaginal metagenomes of South African adolescents. To identify prophage clusters, multiple alignment of all sequences from (A) Lactobacillus crispatus, (B) L. jensenii, and (C) L. iners prophages was performed, a correlation matrix generated using the percentage alignment, and a tree generated using FastTree and visualized using iTOL. Each correlation matrix was annotated by identified prophage cluster, visit (baseline, week 16, and week 32), and participant ID (PID).

Figure 8

Diversity of G. vaginalis prophages identified in vaginal metagenomes of South African adolescents. (A) To identify prophage clusters, multiple alignment of all sequences from G. vaginalis prophages was performed, a correlation matrix generated using the percent alignment, and a tree generated using FastTree and visualized using iTOL. The correlation matrix was annotated by visit (baseline, week 16, and week 32), and participant ID (PID). (B) A proteomic tree including the identified putative G. vaginalis prophages was generated using VipTree. Red branches indicate the putative G. vaginalis prophages from this dataset, while grey branches indicate the viruses included in the reference database. The inner ring annotates the family of a given virus, and the outer ring identifies the host group.

Figure 9

Sequence similarity network of proteins encoded by the identified prophages. Translated prophage protein sequences were used to generate a sequence similarity network (SSN) using EFI-EST. The network was visualized using Cytoscape and protein clusters of bacteriophage hallmark genes (red; including terminase, capsid, integrase, tail, portal and baseplate proteins, head–tail connector, holin, XRE-family HTH domain, DEAD-like helicase, DNA primase/helicase [DnaB], DNA polymerase A and B, DNA gyrase B, DNA topoisomerase IV, DNA ligase, DNA methylase, ribonucleotide reductase, thioredoxin, UvrD-like helicase, AAA family ATPase, holliday junction resolvase, HNH endonuclease, dUTPase, phage antirepressor BRO, peptidoglycan recognitions protein, ssDNA-binding protein, thymidine kinase) and potential antibiotic resistance genes (blue; including ABC transporters, beta-lactamase, and bacitracin resistance protein) highlighted.