The human vaginal microbiota: from clinical medicine to models to mechanisms

Abstract

The composition of the vaginal microbiota is linked to numerous reproductive health problems, including increased susceptibility to infection, pregnancy complications, and impaired vaginal tissue repair; however, the mechanisms contributing to these adverse outcomes are not yet fully defined. In this review, we highlight recent clinical advancements associating vaginal microbiome composition and function with health outcomes. Subsequently, we provide a summary of emerging models employed to identify microbe-microbe interactions contributing to vaginal health, including metagenomic sequencing, multi-omics approaches, and advances in vaginal microbiota cultivation. Lastly, we review new in vitro, ex vivo, and in vivo models, such as organoids and humanized microbiota murine models, used to define and mechanistically explore host-microbe interactions at the vaginal mucosa.

Introduction

The vaginal microbiota is the totality of bacteria, archaea, viruses, and fungi at the mucosal vaginal surface. Community composition fluctuates across the human lifespan with marked changes occurring at puberty onset, in pregnancy and the postpartum period, and during menopause[1-5]. With the emergence of modern sequencing methods, the vaginal microbiota is categorized into five primary community state types (CSTs) based on community composition[6,7]. Four CSTs (I, II, III, and V) display lower diversity with dominance of a single Lactobacillus species and are associated with, but not predictive of, health[6]. Metabolites produced by Lactobacillus species, such as lactic acid, keep the vaginal tract weakly acidic and reduce colonization by pathobionts[8-10]. CST IV, defined by higher species diversity and non-Lactobacillus dominance, is associated with increased risk for a variety of health problems, including bacterial vaginosis (BV), urinary tract infections, HIV acquisition, pregnancy complications, and preterm birth[11-13]. Despite associations between the vaginal microbiota and urogenital conditions, many questions remain regarding the functional roles of host-microbe interactions. In this review, we highlight recent clinical findings and therapeutic strategies targeting vaginal microbiota disorders. We also discuss new and emerging models used to study the vaginal microbiota outside of the human host to bridge the gap between correlation and causation.

Recent clinical insights into ecological factors shaping the vaginal microbiota and microbial-based treatment approaches

Factors shaping the vaginal microbiota are unclear but these likely include both genetic and environmental components. Across multiple cohorts, women of African descent have increased prevalence of BV-associated bacteria and risk of BV[17-19], prompting several recent studies to address factors underlying observed differences between racial groups [14-16]. A paired analysis of 16S rRNA V3-V4 amplicon sequencing and genome-wide association study in native Kenyan women positively associated single nucleotide polymorphisms in innate immune pathways, including mannose binding lectin and interleukin (IL)-5, with G. vaginalis colonization and Shannon diversity[17]. In addition to host genetics, the residential neighborhood of an individual, combined with individuals’ socioeconomic status and resources, influenced the likelihood of vaginal Lactobacillus dominance in a pregnant African American cohort[18]. Understanding how reproductive health aspects impact vaginal microbial composition has continued to be a focus in the field. In agreement with prior cross-sectional studies, oral contraceptive use is associated with increased prevalence of Lactobacillus-dominant and more stable communities over a two-year period[19]. Menstrual products may also influence the vaginal microbiota. Several studies evaluating the use of menstrual cups in comparison to other methods found consistently increased proportions of Lactobacillus crispatus-dominant communities in individuals using menstrual cups, suggesting menstrual product use may also influence vaginal microbial composition[20,21].

Numerous clinical studies have associated specific vaginal species with health (i.e., Lactobacillus crispatus) or disease (i.e., Gardnerella vaginalis association with BV), but many individuals have disease-associated vaginal profiles, yet rarely or never develop the associated disease[22-24]. Ultimately, the presence or abundance of specific species does not reliably predict disease risk for everyone, making scientific findings challenging to translate to clinical practices. Nugent and Amsel scores are used to clinically diagnose BV based on bacterial morphology or clinical symptoms, respectively [6,25]. Recent advances in molecular-based diagnostics enable more precise identification of vaginal taxa, discovery of unique taxonomic biomarkers for BV, and predictors of treatment success; however, these techniques have not yet been implemented in clinical diagnosis as reviewed recently[26-28]. Traditional BV treatment is oral or topical antibiotics, but with the growing concern of antibiotic resistance and high BV recurrence rates, alternative treatments are being pursued[29-32]. Clinical trials have evaluated the efficacy of a variety of approaches ranging from commonly used probiotics to vaginal microbiota transplants (VMT) to treat BV with varying degrees of success (Table 1). Of note, some of these trials focused on clinical outcomes and thus the impact of the interventions on the vaginal microbiota was not fully examined. Incorporation of microbe-microbe and host-microbe interactions into vaginal microbiota-targeting therapies is necessary to enhance future success.

Table 1:

Recent clinical studies of microbial-based interventions for bacterial vaginosis

Intervention	Cohort Size	Geographic Region	Participant Demographics	Follow up period	Clinical Outcome	Microbiota changes	Reference
Antibiotics
Topical metronidazole	48	United States	Patients diagnosed with BV	24–48 hours after last treatment	Not assessed	Decreased relative and absolute abundance of BV associated bacteria detected and increased relative, but not absolute, abundance of Lactobacillus spp. detected via qPCR	[72]
Antibiotics with Probiotics
Topical metronidazole followed by vaginally administered LACTIN-V, a live biotherapeutic containing L. crispatus	66	United States	Patients diagnosed with BV	24 weeks after start of treatment	Reduced recurrence of BV that was sustained for 3 months	qPCR showed reduced Prevotella species and increased L. crispatus and L. gasseri	[73]
Topical metronidazole alone or followed by oral probiotic of L. reuteri RC-14 and L. rhamnosus GR-1	126	China	Patients diagnosed with BV	90 days after start of treatment	No difference in cure rate compared to antibiotics alone and no adverse events due to probiotic adjunct	Alpha diversity decreased in both groups. Probiotic species rarely detected in the vaginal microbiota. L. crispatus was significantly increased in the probiotic group.	[74]
Topical metronidazole followed by oral probiotic of L. gasseri TM13 and L. cripsatus LG55 isolated from the intestine	59	China	Symptomatic BV	90 days after start of treatment	No significant difference in cure rate compared antibiotics alone	No detection of probiotic strains in the vaginal tract and no difference in diversity (alpha or beta) compared to antibiotics alone.	[75]
Probiotics
Vaginally administered capsule probiotic of L. gasseri and L. rhamnosus	74	Denmark	Women diagnosed with unfavorable vaginal microbiota while undergoing fertility treatments	2 months after start of intervention	No significant change	No improvement of the microbiota profile in the probiotic group compared to placebo. Spontaneous improvement occurred in 35% of the placebo group	[76]
Oral probiotic of L. reuteri RC-14 and L. rhamnosus GR-1	86	Canada	Asymptomatic pregnant women with BV	24 weeks after start of treatment	No adverse pregnancy outcomes associated with the probiotic, no difference between groups in returning to a low Nugent score	No significant changes in vaginal community composition	[77]
Oral probiotic of L. reuteri RC-14 and L. rhamnosus GR-1	238	United Kingdom	Healthy, asymptomatic pregnant women	27 weeks after start of intervention	No change in adverse pregnancy outcomes, no significant change in Nugent score	No change in diversity or taxa between probiotic and placebo	[78]
Oral MED-01, a probiotic of 5 Lactobacilli isolated from healthy vagina, or placebo	92	Korea	Intermediate Nugent score of 4–6	12 weeks after start of intervention	Significant decrease in Nugent score in MED-01 group, significant decrease in symptoms in MED-01 group from baseline but not compared to placebo.	qPCR showed a significant increase in 3 of the probiotic strains by 12 weeks. A not significant decrease of BV associated (Mobilunucus spp., Gardenerella) species was seen in MED-01 group	[79]
Oral 10-strain probiotic consisting of 8 lactobacilli and 2 bifidobacteria strains for 28 or 42 days	36	North America	Healthy, pre-menopausal women with an intermediate Nugent score of 4–6	42 days after start of intervention	Reduced vaginal pH compared to baseline, no difference in Nugent score	No significant change in alpha diversity	[80]
Postbiotic
Postbiotic gel from Lactobacillus species	42	China	Patients diagnosed with BV	7 days after treatment	Improved symptoms of BV	Decrease in alpha diversity post gel use, increased relative abundance of Lactobacillus spp. in 57%, decrease in pathogens	[8]
Vaginal Microbiota Transplant
Topical clindamycin or metronidazole followed by VMT from healthy donors	5	Israel	Intractable and recurrent BV	1–2 years post first VMT	No adverse events, response varied by recipient, 4 had improved Amsel score	4 had drastic microbiota changes as early as 1 month post VMT. Expansion of Lactobacillus spp. and decrease in Bifidobacterium and Prevotella spp.	[25]
VMT from a healthy donor without the use of antibiotics	1	Denmark	Diagnosed with vaginal dysbiosis	15 months post first VMT	Significant improvement in symptoms, no adverse effect on subsequent pregnancy	Increased L. crispatus of donor origin. Upon pregnancy, Gardnerella increased in prevalence. Second VMT resulted again in stable L. crispatus dominance.	[81]

Recent insights into microbe-microbe interactions

Much of our understanding of microbial interactions within the vaginal microbiota derives from microbiome profiling via 16S rRNA gene amplicon sequencing or shotgun metagenomics sequencing. Though 16S rRNA gene amplicon sequencing is the most common microbiome sequencing method employed to date, these 16S rRNA gene profiles provide only taxonomic identity, not functional information, and closely related species are difficult to discern due to sequence similarities of the marker gene. Recently developed bioinformatics tools, pipelines and databases are summarized in Box 1. Using metagenomics, which can provide strain-level resolution, over 25 vaginal metagenomic CSTs have been described, highlighting the genetic diversity in the vaginal microbiota not seen in species-level analyses[33,34]. Metagenomic profiles also reveal the functional capacity of a vaginal community, but do not confirm whether bacteria within the community actually perform their predicted functions. This methodological gap is bridged using multi-omics approaches (usually an integration of metagenomics with metabolomics, metatranscriptomics, or proteomics) in which an array of bacterial and host transcripts and products can be computationally mapped to source species based on their predicted gene functions. Vaginal metabolomics studies have identified metabolic markers for HPV, cancer, microbial infection, and preterm birth[22,35-38]. In a study that generated predictive models for preterm birth, models that incorporated metabolomics or exclusively used metabolomics out-performed models that used only microbiome or clinical data[38]. Multi-omics methods are impactful given their ability to identify links between the microbiome and its products, but they cannot definitively trace a given metabolite or protein to a particular microbial species/strain and can be challenging to parse from host-derived products. These limitations highlight the need for in vitro models to cultivate vaginal microbes and communities to disentangle complex metabolic and proteomic signals, and to experimentally manipulate communities for mechanistic studies.

Box 1: New bioinformatics tools and databases.

VALENCIA - VAginaL community state typE Nearest CentroId classifier

Nearest centroid classification approach for vaginal CST classification, using 13,160 taxonomic datasets from 1975 women in the United States to define reference centroids. Compared to reference-free hierarchical clustering, allows for between-study comparisons.[6]

VIRGO - Vaginal non-redundant gene database

Reference database of nearly 1 million non-redundant genes from the vaginal microbiome, including functional and taxonomic annotations. Genes are clustered into vaginal orthologous groups (VOGs) to provide a reference of functional gene/protein families relevant to the vaginal microbiome.[33]

Metagenomic subspecies (mgSs) and metagenomic CST (mgCST) classifier

Hierarchical clustering approach to identify metagenomic subspecies based on the presence or absence of species-specific genes, followed by clustering based on the subspecies composition to identify 27 total metagenomic CSTs. Designed to work with the VIRGO gene database. Based on 1,890 vaginal metagenomes of reproductive age women, mostly from North America.[34]

Vaginotypes of the human vaginal microbiome

Meta-analysis of whole metagenome shotgun sequence databases totaling 1312 vaginal samples from healthy non-pregnant and pregnant women. Used hierarchical clustering to confirm (CSTs) seen in Valencia and identify new subgroup CSTs. Also performed functional prediction from 133 samples and identified functional capacities associated with particular CSTs.[82]

VMAP - Vaginal Microbiome Atlas during Pregnancy

Visualization and analysis tool for vaginal microbiome sequencing data. Calculates alpha diversity metrics, assigns VALENCIA CSTs, and analyzes microbiome composition.[83]

Isala Project

Database of 16S rRNA gene V4 amplicon sequences from the vaginal swabs of 3,345 healthy women living in Western Europe (mostly Flanders, Belgium). [84]

In silico predictions for BV

Computational discovery of microbial network structures (species trio motifs) as indicators of bacterial vaginosis.[27]

Cultivation of human-derived bacterial communities, particularly while retaining their original community structure, is a longstanding obstacle for microbiome research. Certain vaginal species are notoriously fastidious, including species from abundant (e.g. Gardnerella and Lachnocurva vaginae [BVAB1]) and rarer (e.g. Megasphera, Prevotella) genera. Hence, vaginal communities are rarely cultivated, and many studies opt to use individually cultivated strains. Even so, experiments with defined consortia can reveal clinically relevant inter-species relationships such as decreased antibiotic sensitivity of Gardnerella vaginalis in a four species biofilm compared to mono-species biofilms[39]. Vaginal communities have been cultivated on immortalized vaginal epithelial cells and maintain community structure similar to what is found in the host, but in such systems, microbe-microbe interactions are never isolated from interactions with host cells, and the low oxygen preferences of some bacteria create technical challenges for growth conditions[40]. Improved genome resolution of fastidious vaginal species may inform design of optimal cultivation conditions[41] Additionally, genomic insights into individual species nutrient requirements, such as predicted L. iners dependency on exogenous cysteine, can be validated in mixed species culture with other vaginal taxa[42]. Development of simulated vaginal fluid for bacteriological studies brings promise for the potential to cultivate intact communities, and continued improvement of cultivation methods will be necessary to reliably model vaginal microbe-microbe interactions in a wet lab setting[43].

Recent insights into host-microbe interactions

Host-microbe interactions in the vaginal tract have been studied through a variety of models, each with advantages and disadvantages. In vitro, immortalized cell lines derived from human vaginal and ectocervical epithelium are widely used to understand how microbes alter the innate immune response and epithelial barrier integrity[44-47]. To replicate vaginal tissue morphology, several groups have employed three-dimensional (3D) cultivation of immortalized cell lines[48,49]. Laniewski and Herbst-Kralovetz demonstrated that specific species of BV-associated bacteria, such as Gardnerella vaginalis, Atopobium vaginae, Prevotella bivia, and Sneathia amnii induce unique metabolic, immune, and epithelial barrier responses in 3D cervical cultures. These findings are critical as they indicate that certain species may differentially contribute to disease states, illustrating the need for tailored therapies. 3D cell culture provides the benefit of a physiologically relevant basement membrane on a permeable surface coupled with multilayer cell structure, allowing researchers to characterize directional immune responses in the apical and basolateral space, niche occupation of vaginal microbes, and complex cellular junctions[50]. These in vitro models are an essential facet of studying host-microbe interactions, facilitating mechanistic studies in a reproducible, cost-effective, and high-throughput manner; however, immortalized cell lines, such as VK2/E6E7 and Ect1/E6E7, have been immortalized through infection with human papillomavirus, which downregulates the epithelial interferon response and toll-like receptors[51,52]. Additionally, immortalized cell lines fail to replicate the cellular complexity and genetic diversity of the human vagina.

Emerging organoid models have addressed some limitations of immortalized cell lines; they can be differentiated from stem cells into a variety of cell types present in the tissue of origin without genetic manipulation[53] and replicate complex tissue architecture[54]. Though human vaginal organoids have yet to be described, mouse vaginal organoids and human endo- and ectocervical organoids have been used to study Wnt signaling, a highly conserved pathway in cell fate determination, in epithelial differentiation and proliferation[55,56]. In the context of host-microbe interactions, human endo- and ectocervical organoids have modeled Chlamydia trachomatis infection [57] but this system has not yet been extended to other vaginal microbes. Organ-on-a-chip models have also been recently applied to the reproductive tract. Vagina-on-a-chip utilizes primary human vaginal epithelial cells seeded on primary uterine fibroblasts in a microfluidic system that provides continuous flow, mimicking the flow of vaginal secretions and nutrient replacement in humans, and enables prolonged microbial colonization [58]. Using this system, Mahajan et al. showed a multi-strain L. crispatus consortium acidified the chip environment and downregulated inflammatory cytokines, while a non-optimal consortium dominated by G. vaginalis induced epithelial damage and increased inflammatory cytokines such as IL-6, IL-8, and IL-1β[58].

While organoids provide significant advances in our understanding of host-pathogen interactions within a given tissue, they cannot reveal systemic impacts across neighboring reproductive organs or peripheral responses by the immune system. Thus, in vivo models are necessary to complement in vitro and ex vivo studies. The conventional mouse vaginal microbiota was first comprehensively and longitudinally described by Vrbanac et al., revealing five murine community state types (mCST), similar in structure, but not taxonomy, to human CSTs [59]. Conventional mice are dominated by Staphylococcus and Enterococcus with only one rare CST (mCST IV) showing Lactobacillus dominance. Despite these differences in microbial composition, conventional mice have been vaginally colonized by disease-associated bacteria, such as group B Streptococcus (GBS), G. vaginalis, and P. bivia[60]. These efforts have revealed meaningful microbe-microbe and host-microbe interactions, such as work by Gilbert et al. identifying the importance of G. vaginalis in ascending GBS infection [61]. Several studies have used rodent models to evaluate the therapeutic potential of vaginal microbes, using Lactobacillus isolates, synthetic Lactobacillus-based bacterial consortia, and vaginal microbiota transplant (VMT)[62-64]. While this research shows promise, it is important to note that rodents possess a unique vaginal microbiota, which may influence the response to pathogens and Lactobacillus-based therapeutics. Several efforts have attempted to produce a humanized vaginal microbiota model in mice; while mice inoculated with human vaginal microbiota samples did not recapitulate their inoculum, mice orally gavaged with human stool exhibited enhanced frequency of Lactobacillus in the vaginal tract and differential response to infection with dysbiosis-associated vaginal microbes compared to conventional mice[65-67]. Non-human primates (NHPs) are also a promising model to study the vaginal microbiota. Anaerobes typically found in human CST IV, such as Sneathia, Prevotella, and Mobiluncus dominate the vaginal microbiota of rhesus macaques, while Langner et al. showed NHPs can be transiently colonized with L. crispatus [68,69]. Like rodent models, it is not yet clear how colonization with exogenous human microbes may impact reproductive tract immune response in NHPs. Further, work with NHPs requires specialized facilities and handling that are not widely accessible.

Outlook

Significant advances have been made in understanding correlations of the vaginal microbiota with female reproductive health; yet, the vast majority of data remains associative with disease, not predictive, and improved model systems remain a critical need. Metagenomic sequencing, complemented with multi-omics approaches, enables more nuanced description of the microbiota in tandem with functional predictions; however, mechanistic validation of these functional predictions for microbe-microbe interactions requires improved cultivation systems (Fig. 1). Vaginal microbiome researchers may benefit from models for gut-derived communities; continuous-flow cultivation in bioreactors facilitate long-term, reproducible cultivation of gut communities to identify taxa conferring benefits such as pathogen resistance[70]. Adaptation of these systems for vaginal communities could similarly identify phenotypes of interest, while allowing experimental manipulation (e.g., taxonomic, genetic, and nutritional) to mechanistically define bacterial factors contributing to these phenotypes. Considerable progress has been made in developing models for studying host-microbe interactions with the vaginal microbiota; however, additional work is necessary to better define the potential and limitations of these models. Use of human ectocervical and murine vaginal organoids suggests that human vaginal organoids are feasible (Fig. 1). Cultivation of vaginal organoids in a two-compartment system with apical vaginal microbiota and basolateral primary immune cells could capture more complex host-microbe dynamics, as seen with endometrial organoids[71]. Such work may be complemented with in vivo models as described above. Advances in humanized microbiota models are a significant step forward; however, these current models have limited applicability to the human vaginal microbiota and more comprehensive characterization of these models is needed (Fig. 1). Both humanized and conventional animal models could benefit from improved understanding of how the endogenous microbiota shapes the vaginal immune response at baseline and in response to infection before parallels can be drawn between animals and humans.

Figure 1. Current and proposed models to study host-microbe and microbe-microbe interactions at the vaginal mucosa.

Upper left) In vitro host-microbe interactions are most frequently studied with well-characterized immortalized vaginal epithelial cell lines, but vagina-on-a-chip and murine vaginal organoid models possess multiple cell types, offering more granular study of host-microbe interactions. We propose human-derived vaginal epithelial organoids as a further advancement. Upper right) Cultivation-independent models for microbe-microbe interactions generally constitute functional predictions from 16S rRNA gene amplicon sequencing to determine bacterial taxonomic presence and abundance within communities, or whole genome sequencing for bacterial, fungal, archaeal, and viral taxonomy and identification of individual microbial genes. Cultivation-dependent models are generally limited to batch cultivation in bacteriological media, but co-culture with vaginal cell lines may improve retention of fastidious but clinically relevant bacteria. We propose continuous flow cultivation of vaginal microbial communities as a further advancement to characterize microbe-microbe interactions. Lower right) The most common in vivo model is the conventional microbiota mouse, which possesses complex tissue/organ systems and a complete immune system. The humanized gut microbiota mouse adds to these strengths by possessing a more human-like vaginal microbiota for improved translational study of human host interactions with vaginal microbes. We propose a humanized vaginal microbiota mammalian model as a further advancement. Figure Created with BioRender.com

Overall, implementation of multiple complementary models to understand microbe-microbe and host-microbe interactions in the vaginal microbiota is key to improving patient outcomes. Thorough mechanistic understanding coupled with microbial-targeted, or host-microbe interface-targeted therapeutics, can facilitate personalized medicine to restore optimal microbial communities for each person, reducing dependence on antibiotics and treatment of asymptomatic individuals.