Gardnerella and Prevotella: Co-conspirators in the Pathogenesis of Bacterial Vaginosis

(See the Major Article by Gilbert et al, on pages 1099–108.)

More than 60 years after the first studies of its potential etiology, bacterial vaginosis (BV), formerly called nonspecific vaginitis, remains enigmatic [1]. BV is associated with a loss of “protective” Lactobacillus species (L. crispatus, L. jensenii) from the vaginal microbiota and an increase in facultative and strict anaerobes, including Gardnerella vaginalis and Prevotella species. Association of BV with important health outcomes—including premature birth, endometritis, and enhanced transmission of human immunodeficiency virus and other sexually transmitted infections—has been demonstrated repeatedly in epidemiologic studies [2]. However, poor reliability of clinical criteria for BV diagnosis, coupled with its high prevalence (in excess of 25% in some samples [3, 4]), has led some investigators to question whether BV should be described as a disease state at all [2]. Recent years have brought major advances, particularly in understanding the profound shifts in vaginal microbiota that occur during BV [5–7]. Nonetheless, major challenges persist in the field. These include suboptimal in vitro and in vivo models to facilitate understanding of disease mechanisms, inability to culture many of the organisms present within the BV microbiome, and frequent treatment failures and recurrences following therapy.

The earliest investigations of BV pathogenesis focused on G. vaginalis, then called Haemophilus vaginalis, which was recovered from vaginal samples from women with BV [8]. Gardnerella vaginalis is a facultative gram-positive organism closely related to the Bifidobacterium genus. It is detectable as a component of the vaginal microbiome in essentially all cases of BV, though it may also be found in the absence of BV. Notably, G. vaginalis has several characteristics consistent with its role as a candidate BV pathogen, including the propensity to form biofilms in vitro and in vivo, production of sialidase by a subset of strains, and elaboration of a human-specific cytolysin, vaginolysin [9–11]. However, problems with the concept of Gardnerella as the necessary and sufficient cause of BV were quite apparent even from early human studies. Gardner and Dukes were successful in inducing BV symptoms in 1 of 13 women via intravaginal delivery of cultured G. vaginalis, but their success rates increased dramatically (to 11 of 15) when vaginal lavage samples, which presumably contained communities of organisms in both biofilm and planktonic states, were used [1]. More recently, the presence of specific members of the vaginal microbiome has been linked to particular clinical characteristics of BV [12], and metabolomic studies have revealed signatures of the BV state [13–15], yet no single inciting pathogen has been identified conclusively. For these reasons, BV is now more commonly described as a dysbiosis, with changes in community composition likely mediating clinical outcomes. Such a shift in thinking poses both challenges (eg, a novel microbiome- or metabolome-based definition of BV would require large clinical studies to validate its association with relevant sequelae) and opportunities, such as proposed trials of “vaginal microbiome transplant” as a novel treatment strategy for recurrent BV [16].

Prevotella, a genus of gram-negative anaerobic bacteria, was differentiated formally from Bacteroides in 1990 [17]. Members of both genera are common colonizers of human mucosal surfaces, and with expanding knowledge of the human microbiome has also come a growing appreciation for the diversity and physiologic importance of both groups. As an example, relative abundance of Prevotella or Bacteroides in the human intestinal microbiome correlates with host diet, autoimmunity, and other important phenotypes [18–20]. The situation at the vaginal mucosa is less clear, though Prevotella species are common vaginal commensals, and the association between BV and increased abundance of anaerobes, especially Prevotella bivia, is well established [21, 22]. A recent study of monozygotic and dizygotic twins demonstrated that Prevotella was the most heritable bacterial group in the vaginal microbiome and that its abundance was highly responsive to host states including body mass index and hormonal milieu [23]. It is important to note that Prevotella was grouped at the genus level in that work, and that species- or strain-level effects may be highly relevant [19]. For example, in a study of the vaginal microbiome, our group found multiple Prevotella species that were positively associated with Streptococcus agalactiae carriage and others that were negatively associated [24].

Several lines of evidence suggest that P. bivia may be more than a bystander in BV. The importance of anaerobes, including P. bivia, in BV pathogenesis was postulated by Spiegel et al [22]. Prevotella bivia is an important source of lipopolysaccharide and ammonia in vaginal mucus, is associated with epithelial cytokine production, and enhances the growth of other BV-associated organisms, including G. vaginalis [25–28]. Machado et al similarly noted a symbiotic relationship between G. vaginalis and P. bivia, demonstrating that the presence of a G. vaginalis biofilm stimulates growth of P. bivia in vitro [29]. Notably, using daily swab samples, Muzny et al found that P. bivia was the first BV-associated species to increase above baseline prior to incident BV, consistent with its potential role as a driver of this condition [30].

The complex, dynamic nature of the vaginal microbiota and the likely role of multiple fastidious bacterial species in the pathogenesis of BV have posed major challenges for mechanistic studies. In vivo models have generally focused on G. vaginalis, as it can be cultured in microaerophilic conditions and has been shown to colonize the murine genital tract [31]. Gilbert et al previously demonstrated that G. vaginalis colonization was sufficient to induce several features of bacterial vaginosis in the mouse, including epithelial cell exfoliation, sialidase activity, and ascension to the nongravid uterus [32]. In this issue of The Journal of Infectious Diseases, Gilbert et al report a major extension of those findings [33]. The authors demonstrate that G. vaginalis and P. bivia can colonize the murine vagina for several days both alone and in combination. Somewhat unexpectedly, given the proposed mechanisms of symbiosis between the 2 organisms, co-inoculation did not have major effects on the recoverable quantities of either organism during colonization, with only a slight increase in P. bivia level at day 1 and a decrease at day 2. No effect on G. vaginalis was noted at any time. Consistent with findings in humans with BV, no overt local inflammatory infiltrate was noted during colonization or co-colonization. The previously noted G. vaginalis–mediated epithelial exfoliation and sialidase activity were observed at essentially equivalent levels in the presence or absence of P. bivia. However, co-colonization strikingly enhanced P. bivia ascension to the uterus, whereas G. vaginalis ascension was unaffected. This latter finding is particularly relevant because, alone or in combination with other organisms, P. bivia has the capacity to induce preterm birth, endometritis, and other uterine pathologies [34–36]. The authors conclude that through its production of sialidase, induction of epithelial exfoliation, and promotion of P. bivia ascension to the uterus, G. vaginalis contributes directly to the pathogenesis of BV and its associated symptoms and outcomes.

There are several limitations to the experiments presented, many of which are acknowledged by the authors. Single strains of both G. vaginalis and P. bivia were used in the experiments, and it will be important to study the relevance of their findings to other clinical isolates and to related species. There are strengths of the murine model but also significant limitations, including vaginal pH and background microbiota that differ from those in humans, the presence of human-specific factors produced by bacteria that may alter host responses, and differences in sialic acid substrates for the enzymes detected. Estradiol pretreatment of mice, while necessary for the establishment of G. vaginalis colonization, is associated with a dampened local inflammatory response, including suppression of polymorphonuclear cell influx and diminished cytokine expression [37, 38]. Exploration of the interactions between G. vaginalis and P. bivia in pregnant animals may reveal a distinctly different host immune response. Nonetheless, the use of a co-colonization model with careful assessment of BV-relevant phenotypes is an important step forward and provides a firm foundation for those future studies. Expansion of such animal models, in combination with emerging technologies such as 3-dimensional tissue culture, reconstitution of gnotobiotic mice with human vaginal microbiota, and improved methods for cultivation of relevant organisms will move the BV field forward and closer to a thorough mechanistic understanding of this important condition [39].

Notes

Financial support. The authors receive support from the National Institutes of Health (grant numbers R56 AI136499 and R01 AI143290 to A. J. R. and R21 AI127957 to T. M. R.).

Potential conflicts of interest. A. J. R. has served as a consultant to Pfizer and Toltec Pharmaceuticals and is listed as inventor on 3 patents relevant to bacterial vaginosis. T. M. R. reports no potential conflicts. All 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.

Disclaimer. The content is solely the responsibility of the authors and does not necessarily represent the official views of the funding body.