The Vaginal Microbiome: Current Understanding and Future Directions

Abstract

This article summarizes the highlights of the expert technical consultation on bacterial vaginosis (BV), sponsored by the National Institute of Allergy and Infectious Disease and held in Washington, DC, on 8–9 April 2015. Many issues touched on in this article are discussed in much greater detail in the 6 preceding articles in this supplement to The Journal of Infectious Diseases. There was a consensus among the meeting attendees concerning the most important research issues in the field: the pathogenesis of the syndrome, way to optimize treatment, and the relative roles of sexual transmission and endogenous infection in BV epidemiology. This article concludes with a listing of BV and genitourinary tract research priorities that were discussed and agreed on by attendees. The most important of these included better characterization of vaginal microbiome community state subtypes, application of advanced “-omic” technologies to improve understanding of BV pathogenesis, further investigation of the relationships between the male and female genitourinary tract microbiomes, and the development of new drugs for BV treatment.

bacterial vaginosis, vaginal, microbiome, genitourinary tract, antibiotics, women's health

Bacterial vaginosis (BV) represents a dysbiosis of the vaginal microbiome that is associated with significant adverse healthcare outcomes, including increased risk of abnormal pregnancy outcomes, pelvic inflammatory disease, and increased susceptibility to sexually transmitted infections, most importantly human immunodeficiency virus type 1 infection. BV is a highly prevalent condition worldwide but disproportionately afflicts women of African descent, in whom it is an important contributor to reproductive healthcare disparities. BV is associated with dramatic shifts in the vaginal microbiota, characterized by a change from Lactobacillis sp. dominance to dominance by a mixture of organisms including Gardnerella vaginalis, various anaerobic species, and the genital mycoplasmas [1–3]. Markers of BV that are used by clinicians to diagnose BV include the following: (1) vaginal secretions with a pH of >4.5, (2) a fishy odor that is best elicited by mixing vaginal secretions with a 10% potassium hydroxide solution, (3) ≥20% of microscopically observed vaginal epithelial cells coated with bacteria (“clue cells”), and (4) a white, skim milk–like vaginal discharge. These are known as the Amsel criteria, and a clinical diagnosis of BV requires ≥3 of the 4 markers to be present. In 1991, Nugent et al [4] reported the use of Gram stain criteria to diagnose BV, using a numerical score based on semiquantization of gram-positive rods, gram-negative coccobacilli forms, and curved gram-negative rods. These morphotypes were thought to represent Lactobacillus spp., G. vaginalis and Mobiluncus spp., respectively. Scores of 0–3 were considered normal (Lactobacillus dominant), scores of 4–6 were labeled as intermediate (mixed morphotypes), and scores of 7–10 were indicative of BV (absence of lactobacilli and predominance of the other 2 morphotypes). The Nugent score has become the standard for diagnosing BV in research studies.

Since the National Institutes of Health (NIH) last sponsored a technical consultation on BV in 2008 [5], there have been significant advances in our understanding of the structure of the vaginal microbiome, and new concepts have emerged concerning the pathogenesis of BV. These advances have led to new ideas for the prevention and treatment of this ubiquitous condition. In this article, we highlight the major issues discussed at the 2015 NIH-sponsored BV technical consultation, many of which are discussed in detail in the other 6 articles included in this supplement. We conclude with a list of the research priorities agreed on by the meeting attendees.

ADVANCES IN THE UNDERSTANDING OF THE VAGINAL MICROBIOME BASED ON 16S RIBOSOMAL RNA BACTERIAL GENE SEQUENCING

Significant time at the meeting was devoted to reviewing the results of vaginal microbiome high-throughput sequencing studies over the last 15 years and how this research has advanced knowledge of the vaginal microbiome in women with or without BV. Early molecular studies of the vaginal microbiota began more than a decade ago, using general 16S ribosomal RNA (rRNA) gene primers and sequencing techniques, such as denaturing gradient gel electrophoresis and amplicon cloning and sequencing [6–9]. These studies led to the discovery of organisms not previously recognized as belonging to the vaginal microbiota, including Lactobacillus iners, Atopobium vaginae, Sneathia, Leptotrichia, Megasphaera, Dialister, and Eggerthela.

Fredricks et al [8] discovered several organisms that were heretofore completely unknown. They named these organisms BV-associated bacteria (BVAB) 1, 2, and 3. BVAB3 subsequently was discovered in a collection of unknown anaerobic bacteria isolated from stored isolates obtained from vaginal microbiology studies conducted by Austin et al and has been formally named Mageeibacillus indolicus [10]. These studies also demonstrated the nearly ubiquitous presence in the vaginal environment of L. iners, whose relatively fastidious growth requirements probably explain why it was not previously recognized in earlier culture-based studies.

During the last 6 years, the advent of inexpensive high-throughput sequencing technologies, such as the 454 pyrosequencing platform [11], in concert with increasing availability of computational analysis programs to analyze the large numbers of sequences (readings) generated, have resulted in the ability to process thousands of 16S rRNA gene sequences in a single specimen at relatively low cost. Further, the creation of vaginal microbiota 16S rRNA gene sequence databases by individual investigators has allowed identification of the representative organisms at the species level in many cases [12–14]. The large number of readings available enables characterization of each individual's vaginal microbiome to the extent that there is reasonable confidence that all but the rarest taxa are being included in the analyses.

The first studies to report high-throughput sequencing results obtained from sexually active premenopausal women provided fine detail that significantly expanded our understanding of the vaginal microbiota [13–16]. These studies identified groups of women with 3 main types of vaginal microbiomes. One is dominated by Lactobacillis crispatus (L. crispatus group), another by L. iners (L. iners group), and a third is more diverse without a single dominant taxon (the “diverse group”). The abundance of taxa in this group is more evenly distributed than in the first 2 groups. L. crispatus is almost never present in the diverse group, whereas L. iners is almost uniformly present in all groups but in much lower abundance than in the L. iners group [17]. In addition, the largest of these 4 studies also found 2 less common types of vaginal microbiome dominated by either Lactobacillis jensenii or Lactobacillis gasseri [16].

L. iners does not produce lactic acid at the same level as L. crispatus, but at very high vaginal concentrations it seems to produce sufficient concentrations to maintain vaginal pH at <4.5 [18]. Unlike those with an L. crispatus–dominant vaginal microbiome, some women with an L. iners–dominant vaginal microbiome have clinical or microscopic evidence of BV or fall into the intermediate Nugent score category [13, 14, 16]. Several studies have documented a vaginal microbiome subgroup in which L. crispatus and L. iners share dominance [13, 14, 19].

As noted above, with a few exceptions, clinical and microscopically defined BV cases are almost always found within the diverse vaginal microbiome group. In some of these studies including the most recent and largest [20], subgroups of cases have been identified in women who have a diverse vaginal microbiome characterized by one highly abundant (though not always dominant) organism. Organisms that are most predominant in these subgroups or community state types (CSTs) include BVAB1 [13, 14, 18, 20], A. vaginae [19], Sneathia/Leptotrichia spp. [14], Prevotella spp. [14, 18], and G. vaginalis [18, 20].

Identification of different stable CSTs within the diverse vaginal microbiome group is potentially important because they may pose differing risks for adverse health outcomes. Candidate CSTs based on a survey of the studies referenced in the paragraph above, titled according to the predominant organism in each, are as follows: Gardnerella, Prevotella, BVAB1, and Sneathia. Differences between studies in identifying these CSTs may be attributable to differences in the women sampled but also to important methodological differences. Methodological components that influence 16S rDNA analyses include the polymerase chain reaction primer sequences used to amplify the 16S rRNA gene, the region of the 16S rRNA gene amplified, specimen collection methods, and the bioinformatics approaches to data analyses. A critical need in this field is standardization of research methods to the extent possible in order to identify women with BV who may be at highest risk of adverse outcomes.

Longitudinal studies have demonstrated the dynamic nature of vaginal microbiome, especially during menstruation, with dramatic shifts in the microbiota occurring in some women [14, 21]. Gajer et al [19] carried out the most comprehensive longitudinal study of the vaginal microbiome in individual women, using 454 pyrosequencing to analyze vaginal specimens collected at weekly intervals over a period of 16 weeks. Overall, women with as L. crispatus–dominant, L. iners–dominant, or diverse type vaginal microbiome, as defined in that study, did not change over time in 85% of the cases. However, the rate of stability was significantly lower (54%) in cases where L. crispatus and L. iners shared dominance. Despite significant shifts in relative abundance of taxa in some of these women, the changes were usually transient, with the vaginal microbiome returning to its prior state in most cases. Shifts to a different vaginal microbiome state were restricted, depending on the type of vaginal microbiome. The L. crispatus–dominant vaginal microbiome shifted either to an L. iners–dominant microbiome or to the mixed Lactobacillus microbiome and only rarely to the diverse, vaginal microbiome, whereas women with an L. iners–dominant vaginal microbiome switched to the mixed Lactobacillus state or the diverse state but only occasionally switched to an L. crispatus–dominant vaginal microbiome. When switches in women with diverse vaginal microbiomes occurred, they were almost always to an L. iners dominated state.

Once CSTs within the diverse group are more clearly defined it will be important to perform similar longitudinal studies of women with these microbiomes to determine their stability. This should be a research priority because the risk of BV-associated adverse health events may differ between BV CSTs. If so, prevention efforts could be more efficiently focused. Such studies will require the vaginal microbiome research community to agree on which primer sequences and 16S rRNA variable region(s) should be used as the basis of the research.

A study in pregnant women demonstrated that the vaginal microbiome differs significantly from that in nonpregnant women [22]. The researchers compared vaginal 16S rRNA gene sequence data obtained from the Human Microbiome Project database for pregnant and nonpregnant women. The results showed that vaginal microbiome richness and diversity are significantly decreased in pregnant women, and there was a shift in the vaginal microbiome overall toward a greatly increased abundance of Lactobacillus spp. This finding supports the concept that hormonal changes have a significant influence on the vaginal microbiome, as did a prospective study conducted in Australia [23].

Further complicating the characterization of the vaginal microbiome is the possibility that subspecies or strain variants of vaginal bacteria could have functionally important roles to play. Documentation of such low level sequence variations (oligotypes) within the highly prevalent vaginal organism, G. vaginalis, has been demonstrated [24]. This study showed that G. vaginalis oligotypes are often congruent between sexual partners and that oligotype distribution in women differs depending on whether or not BV is clinically present, thus supporting the hypothesis of G. vaginalis as a keystone pathogen, as discussed in the first article in this supplement. It is likely that there are similar strain differences among other highly prevalent members of the vaginal microbiome, such as L. iners.

THE ROLE OF “-OMICS” AND SYSTEMS BIOLOGY IN ADVANCING UNDERSTANDING OF THE VAGINAL MICROBIOME

The role of metagenomics, transcriptomics, proteomics, and metabolomics in characterizing the vaginal microbiome was also discussed by the group. 16S rRNA gene sequence data tell us that bacterial communities are present in the genitourinary tract but do not tell us what they are doing there. Metagenomics studies are currently limited but will be more useful in the near future as sequencing costs continue to decline and better bioinformatics tools for “big data” are developed. Metagenomics not only tells us which organisms are present but also informs us of the genetic potential of the entire bacterial community. Metagenomics also gets around the primer bias issues of partial 16S rRNA gene sequence–based studies, as alluded to above, and thus is a potential solution to the need for standardization of methods for such studies. Messenger RNA–based transcriptomics is attractive because the approach informs us which genes are being expressed in vivo, and at what levels, thus providing clues as to BV pathogenesis. Unfortunately, technical challenges with this approach have limited its potential utility.

Proteomics shows which genes are actually being translated, which gets us closer to what the vaginal microbiome is actually doing in vivo, but this approach is also technically challenging in complex human specimens and is also somewhat dependent on the completeness of the metagenome database for vaginal bacteria. At present, this database is insufficient for this task. Proteomics holds out the hope for the discovery of protein biomarkers of the underlying vaginal microbiome which could be used in diagnosing the hypothetical pathogenic vaginal microbiome CSTs discussed above. Metabolomics is a rapidly maturing discipline that may be more immediately applicable to understanding what individual vaginal microbiomes are doing than either transcriptomics or proteomics. Moreover, the development of metabolic biomarkers for vaginal microbiome CST pathogenic potential is also within the near-term realm of possibility. Increasing the vaginal microbiome bacterial metagenome database is important to optimizing all of these newer “-omic” approaches. This supplement includes an article by Srinivasan et al, describing the use of optimized vaginal culture techniques to identify bacteria whose genomes are unknown, for the purpose of building this database.

PATHOGENESIS OF BV

The weight of the evidence supports the concept that in some cases (perhaps most, and especially with incident infections), BV can be sexually transmitted between men and women, and between female sex partners. The epidemiologic data evidence for this is summarized in the article by Muzny and Schwebke in this supplement. This hypothesis is also supported by studies employing cultivation-independent analyses, which show that many BV-associated species can also be detected in male penile skin, semen, urethral, and urine specimens [25, 26]. Liu et al [27] have shown that among uncircumcised men the presence of BV-like flora in the coronal sulcus skin specimens correlates with BV diagnosed by the Nugent score in their sex partners. Finally, a more recent study has shown that the microbiomes of the penile skin and the urethras in male partners of women with BV match the vaginal microbiomes of their partners more closely than they match those of randomly selected women with BV [28].

Taken together, these studies provide powerful additional evidence that sexual exchange of BVAB between male and female partners does occur. However, as highlighted by Muzny and Schwebke, the major controversy is whether new and/or recurrent cases of BV are related primarily to the transmission of the pathogenic consortium of the organisms known to be associated with BV or to the transmission of a single species, which then creates a vaginal environment favorable to proliferation of the organisms commonly associated with BV. The leading candidate species for the latter role is G. vaginalis. However, not all strains of G. vaginalis seem to be associated with BV. Evidence showing that sexually inexperienced adolescents are colonized with this organism early in life was discussed at this technical consultation and has been published [29]. Finally, there is the question of whether endogenous colonization of BVAB at extravaginal sites somehow plays a role in incident BV. One prospective study of women without BV at enrollment demonstrated that detection of G. vaginalis and other BVAB in the oral cavity or in the rectum predicted subsequent BV incidence [30] And, conversely, that rectal carriage of L. crispatus conferred a decreased risk [31].

HOST IMMUNITY

The precise role of the host immune response in BV, and in the risk of associated consequences, is yet to be defined. Data do suggest that genetic variations that decrease the mucosal innate immune response are associated with increased risk for BV. For example, polymorphisms in immune response genes that predict lower production of cytokines and chemokines are associated with a higher risk of BV, polymorphisms in interleukin 6 associated with less of an immune response to interleukin 1 and lipopolysaccharide were associated with increased BV in black women, and an interleukin 8 allele associated with increased interleukin 8 production is less common in women with BV [32].

Single-nucleotide polymorphisms in Toll-like receptor (TLR) genes have been associated with higher vaginal quantities of BVAB and increased prevalence of BV [33–35]. The extent to which these TLR polymorphisms intersect with—and may contribute to—the racial variability seen with BV remains to be studied. As noted by Murphy and Mitchell in their paper which is included in this supplement, the effects of exogenous mediators, such as hormones, smoking, and sexual behaviors (especially exposure to semen), need to be integrated into such analyses. For example, estradiol itself may decrease TLR2 and TLR6 messenger RNA expression, as well as expression of human beta defensin 2 and secretory leukocyte protease inhibitor by vaginal epithelial cells in vitro [36]. Given the dynamic nature of the vaginal microbiome, prospective studies with frequent sampling and collection of relevant metadata—especially hormone use—will be required to disentangle these potential modifiers.

IN VITRO MODELS OF BV

BV research has been inhibited by the lack of adequate in vitro model systems and even more inadequate animal models. There is little experience with modeling human diseases that are caused by complex microbiomes. The multiple issues involved in developing models of the vaginal microbiome in health and disease are well summarized by the article in this supplement by Herbst-Kralovetz et al. A multilayer primary cell culture system grown as Transwell inserts in such a way as to create an air epithelial cell interface, thus mimicking the in vivo environment, has promise as a model for studying the interactions of the vaginal microbiome and its human host in real time. Three-dimensional cell organotypic models of human mucosal sites grown on microcarrier beads in rotating well vessel bioreactors are another promising approach.

One of the major obstacles to the development of animal models for the study of the human vaginal microbiome is the fact that the microbiota of animals is strikingly different from that of humans. Though limited success has been achieved, endogenous microbiomes are often highly resilient and thus resist investigators' efforts to establish human microbiomes in an animal model system. A gnotobiotic mouse has been successfully used in one study to model the vaginal growth and interaction of Lactobacillus and G. vaginalis [37]. Although it will be impossible to overcome all limitations of in vitro human vaginal microbiome modeling, there is hope that some recently developed systems will be of use in addressing carefully defined questions concerning interactions of the host and genitourinary tract microbiota. Investigators in this field should pay careful attention to the results of ongoing in vivo studies designed to more carefully distinguish the roles of genetic variants of organisms such as G. vaginalis in BV pathogenesis and associated biofilms [38].

TREATMENT OF BV

Current recommended treatments for BV include multidose oral metronidazole, topical metronidazole gel, and topical clindamycin cream. Response rates immediately after treatment range from 75% to 85%, and relapse rates among initial responders are high. Changes in the vaginal microbiome after treatment, as assessed by high-throughput sequencing, have not yet been carefully documented, and these data are necessary for understanding why treatment failure and recurrence rates are high. There are no new drugs available for BV treatment but, as detailed by Bradshaw and Sobel in their review paper included in this supplement, promising new approaches are being explored. As described above, formation of biofilm is hypothesized to be an important factor in the pathogenesis of BV. In particular, it is hypothesized that the bacteria critical to BV pathogenesis are protected from the effects of systemic and topical antimicrobial agents by virtue of sequestration within a vaginal biofilm. If this is true, a case can be made for exploring the use of biofilm-disrupting agents for the treatment of BV, either alone or in combination with antibiotics. At least one such compound is currently in early clinical trials (NCT01812889).

Restoration of the normal microbiota after antibiotic treatment with probiotics containing Lactobacillus species is an approach that has been explored, but without great success. However, this may be the result of poorly chosen probiotic candidates. Current probiotic studies will take advantage of better understanding of BV pathogenesis and formulated products, including one with high concentrations of L. crispatus. Finally, recognition of the fact that BV is often sexually transmitted has resulted in renewed interest in partner treatment as a means of preventing relapse. Designing heterosexual partner treatment clinical trials will be complicated by lack of certainty about the male reservoir of infection. If it is primarily the coronal sulcus, topical treatment would likely be required, but systemic drugs would be required if it is within the urethra. Combinations of topical and systemic antimicrobials may be the answer. In any case, it is clear that a great deal of work in this area is needed, but it is encouraging that improved understanding of the pathogenesis and epidemiology of BV is guiding innovative new approaches to treatment.

FUTURE RESEARCH DIRECTIONS

Considerable time during this meeting was devoted to discussion of research approaches to the difficult problems facing BV investigators, as highlighted in the articles presented in this issue. We conclude with a list of future research recommendations and questions agreed on by the majority of meeting attendees.

1. Centers engaged in descriptive studies of the vaginal microbiome should work together to develop a well-documented vaginal specimen bank from around the world that can be shared between laboratories. The goal would be to characterize the differences in vaginal microbiome structure resulting primarily from differing high-throughput sequencing primers but also other more minor differences in specimen preparation and processing methods. Ideally, agreement on standardization of high-throughput sequencing methods could be reached, though this is a difficult goal, given the time and effort experts in the field have put into developing their individual approaches.

2. Different BV CSTs should be documented, and criteria established for assigning individual specimens to them, addressing the following questions: How stable are these BV CSTs over time? If found to be stable, do these BV CSTs have different adverse effects on host women and/or different interactions with the human mucosal immune response? Finally, can biomarker profiles be developed to more easily identify specimen membership in specific BV CSTs?

3. Can bacterial culture methods be advanced to increase the number of cultivatable vaginal microbiome species available for whole-genome sequencing, which is essential for expanding knowledge of the human genitourinary tract metagenome?

4. Proteomic and metabolomic studies in conjunction with metagenomic studies of well-characterized vaginal microbiome specimens are needed to determine the pathogenic mechanisms associated with disease induced by defined BV CSTs. Such studies are also needed to identify non–sequence-based biomarkers for pathogenic CSTs.

5. What is the effect on the vaginal microbiome of host hormonal variation, including the use of hormonal contraception?

6. Do the microbiomes of other body sites, such as the rectum and the mouth, influence the composition of the vaginal microbiome, and vice versa?

7. To what extent do in vitro vaginal microbiome models reliably predict what happens in the human vaginal environment?

8. Longitudinal studies of the genitourinary tract microbiomes in women and their partners over a period of at least several months, with relatively frequent sampling after BV treatment, are critical to understanding the poor long-term outcomes of currently available treatment modalities.

9. A better understanding is needed of the role of reinfection by the male partners of women treated for BV and the mechanisms of reinfection. If reinfection is occurring, is the primary reservoir the penile skin, the male urethra, or both? Once this is understood, the next goals would be to determine the most effective topical agents for eliminating BV-associated organisms from the skin and the most effective systemic antimicrobials for eliminating them from the urethra.

10. Although biofilm disruptive agent early clinical trials are underway, better methods of characterizing vaginal biofilms are sorely needed.

Notes

Supplement sponsorship. This article appears as part of the supplement “Proceedings of the 2015 NIH/NIAID Bacterial Vaginosis Expert Consultation,” sponsored by the Division of Microbiology and Infectious Diseases of the National Institute of Allergy and Infectious Diseases in partnership with the University of Alabama at Birmingham Sexually Transmitted Infections Clinical Trials Group; contract HHSN272201300012I.

Potential conflict of interest. J. M. M. is the protocol chair of an National Institutes of Health-sponsored clinical trial to evaluate a new treatment agent for bacterial vaginosis. D. H. M. report no potential conflicts. Both authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.