Gardnerella Species and Their Association With Bacterial Vaginosis

Abstract

Background

Bacterial vaginosis (BV) is a condition marked by high vaginal bacterial diversity. Gardnerella vaginalis has been implicated in BV but is also detected in healthy women. The Gardnerella genus has been expanded to encompass 6 validly named species and several genomospecies. We hypothesized that particular Gardnerella species may be more associated with BV.

Methods

Quantitative polymerase chain reaction (PCR) assays were developed targeting the cpn60 gene of species groups including G. vaginalis, G. piotii/pickettii, G. swidsinskii/greenwoodii, and G. leopoldii. These assays were applied to vaginal swabs from individuals with (n = 101) and without BV (n = 150) attending a sexual health clinic in Seattle, Washington. Weekly swabs were collected from 42 participants for up to 12 weeks.

Results

Concentrations and prevalence of each Gardnerella species group were significantly higher in participants with BV; 91.1% of BV-positive participants had 3 or more Gardnerella species groups detected compared to 32.0% of BV-negative participants (P < .0001). BV-negative participants with 3 or more species groups detected were more likely to develop BV within 100 days versus those with fewer (60.5% vs 3.7%, P < .0001).

Conclusions

These results suggest that BV reflects a state of high Gardnerella species diversity. No Gardnerella species group was a specific marker for BV.

Bacterial vaginosis (BV) is the most common cause of vaginal discharge in reproductive age women worldwide and is marked by high bacterial diversity, increased abundance of BV-associated anaerobic bacteria, and decreased abundance of lactobacilli in the vagina [1–3]. BV has been associated with a wide range of negative health outcomes such as increased risk of preterm birth [4], acquisition of sexually transmitted infections [5, 6] and pelvic inflammatory disease [7]. BV etiology remains incompletely understood and the heterogeneous, polymicrobial nature of this condition has made efforts to study BV challenging. Hypothetical models have been advanced to help explain the pathogenesis of BV [8].

Gardnerella has been implicated in the development of BV since the 1950s and was the first bacterium linked to nonspecific vaginitis, now known as BV [9]. Previous studies have shown increased abundance of vaginal Gardnerella in women with BV [10, 11] and Gardnerella has been associated with adherent biofilms in BV [12]. There is evidence that Gardnerella can be sexually transmitted [13, 14], and it may be a keystone species in the development of BV [8, 14]. However, Gardnerella has also been detected by polymerase chain reaction (PCR) in more than 50% of women without BV, calling into question its role in BV etiology [15, 16]. Formerly classified as Gardnerella vaginalis, the Gardnerella genus has been shown to consist of at least 13 unique genomospecies, 6 of which have been validly named: Gardnerella vaginalis, Gardnerella piotii, Gardnerella swidsinskii, Gardnerella leopoldii, Gardnerella pickettii, and Gardnerella greenwoodii [17, 18]. Vaginal colonization with multiple Gardnerella species has been demonstrated in prior studies [10, 13, 19–24] and highlights the need to further understand the prevalence and role of each species within the Gardnerella genus. We hypothesized that a subset of Gardnerella species is more prevalent and abundant in women with BV, thereby contributing to the etiology of BV due to pathogenic characteristics of these species. In contrast, we expected women without BV to be colonized by other Gardnerella species, potentially explaining the high prevalence of Gardnerella commonly noted in women without BV.

Previous studies have often utilized the 16S ribosomal RNA (rRNA) gene for Gardnerella identification; however, all Gardnerella species are more than 98.5% identical in this region [17], necessitating the use of other phylogenetic marker genes to differentiate Gardnerella species. We developed 4 quantitative PCR (qPCR) assays targeting the cpn60 gene of species groups including G. vaginalis, G. piotii/pickettii, G. swidsinskii/greenwoodii, and G. leopoldii. These assays were applied to vaginal swabs collected from 101 study participants with BV and 150 participants without BV. Longitudinal weekly swabs from a subset of 42 participants (16 BV positive and 26 BV negative) were used to monitor vaginal Gardnerella population dynamics over time and response to metronidazole treatment.

METHODS

Study Population and Sample Collection

From October 2012 to August 2015, vaginal swabs were collected from 251 nonpregnant study participants with (n = 101) and without (n = 150) BV attending the Public Health Seattle and King County Sexual Health Clinic. The overall study population included individuals assigned female sex at birth in Seattle, Washington, and individuals with and without risk factors for BV to include previous history of BV, identifying as black and/or Hispanic, or high number of sexual partners [25]. The study was approved by the Fred Hutchinson Cancer Center institutional review board (protocol No. 7683) and all study participants provided informed consent. Vaginal swabs were collected by clinicians at baseline and subsequent follow-up clinic visits. Participants were considered BV positive when at least 3 of 4 Amsel clinical criteria were met (symptomatic BV) [26]. Vaginal smears of all clinic samples were prepared on glass slides for Gram stain and assessment of BV by Nugent scoring [27]. Participants with symptomatic BV were treated with metronidazole at each clinic visit following Centers for Disease Control and Prevention guidelines. Most study participants did not remain enrolled beyond 100 days, and this was used as a cutoff for analysis of BV development. A subset of 42 participants provided at least 10 additional weekly self-collected vaginal swabs for up to 12 weeks from baseline and these samples were used for longitudinal analysis. Behavioral data were collected via questionnaires at clinic visits and via daily diaries. A more detailed description of the study population and sample collection can be found in the Supplementary Methods.

DNA Extraction and qPCR

DNA was extracted from vaginal swabs as described in Supplementary Methods. Total Gardnerella concentrations were quantified using a previously developed qPCR assay targeting the Gardnerella 16S rRNA gene [28]. Using the species designations proposed by Vaneechoutte et al [17] and Sousa et al [18], 4 qPCR assays were developed targeting the cpn60 gene of the following species groups: G. vaginalis and Gardnerella genomospecies 2, G. piotii and G. pickettii, G. leopoldii, G. swidsinskii, and G. greenwoodii (Gardnerella genomospecies sp 9 and 10 also detected with reduced sensitivity, see Supplementary Table 1). We report the species groups detected by these assays here as G. vaginalis, G. piotii/pickettii, G. swidsinskii/greenwoodii, and G. leopoldii. Assay conditions and oligonucleotide sequences are shown in Supplementary Table 1 and further described in the Supplementary Methods.

Statistical Analysis

Figures were created using matplotlib, seaborn, and statannotations Python packages [29–31]. Statistical analysis was performed using GraphPad Prism 9.5.1 for Windows and the psych R Statistical software package [32, 33]. Significance of qPCR data was calculated using the Mann-Whitney test. Relative risk confidence intervals (CIs) were calculated using Koopman asymptotic score and significance was calculated using Fisher exact test. Heatmap correlations were calculated using Spearman correlation.

RESULTS

Study Population

The 251 participants enrolled in the study ranged from 19 to 49 years of age; 51.8% identified as white, 35.4% as black, and 10.4% as other; 7.2% of participants identified as Hispanic. Of the participants, 8.8% did not identify as black or Hispanic and reported both no history of BV and 10 or fewer lifetime sex partners (no BV risk factors). In total, 101 participants were BV positive at baseline by Amsel criteria and 150 were BV negative; 115 participants had Nugent scores indicative of no BV (0–3), 32 had Nugent scores indicative of intermediate microbiota (4–6), and 103 had Nugent scores indicative of BV (7–10). Thirty-nine participants were Amsel negative but had Nugent scores >3 (asymptomatic BV/intermediate microbiota); 80 of 101 (79.2%) participants with symptomatic BV reported a history of BV compared to 86 of 150 (57.3%) of BV-negative participants (Table 1). Additional participant characteristics are shown in Supplementary Table 2.

Prevalence and Concentrations of Gardnerella Species

Gardnerella was detected via 16S rRNA gene qPCR in 100 of 101 (99.0%) Amsel BV-positive participants and 112 of 150 (74.7%) Amsel-negative participants. Using the qPCR assays developed in this study targeting the cpn60 gene, G. vaginalis was most prevalent (69.7%), followed by G. piotii/pickettii (62.9%), G. swidsinskii/greenwoodii (58.6%), and G. leopoldii (41.0%). All 4 species groups were more frequently detected in BV-positive participants compared to BV-negative participants at baseline as measured by both Amsel criteria and Nugent score (Table 2 and Supplementary Table 3). Median concentrations of each Gardnerella species group were significantly higher in Amsel BV-positive participants compared to Amsel-negative participants (P < .0001; Figure 1), and significantly higher in participants with Nugent scores of 4–6 (intermediate microbiota) or 7–10 (BV) compared to those with scores of 0–3 (no BV) (Supplementary Figure 1). Study participants with concordant BV (Amsel positive, Nugent 7–10, n = 96), as well as participants with asymptomatic BV/intermediate microbiota (Amsel negative, Nugent 4–10, n = 39), had significantly higher concentrations of all 4 Gardnerella species groups compared to concordant BV-negative participants (Amsel negative, Nugent 0–3, n = 110; P < .0001; Figure 2). Median total bacterial concentration was significantly higher in participants with symptomatic BV as measured by a broad-range 16S rRNA gene qPCR assay (8.65 vs 8.02 log₁₀ gene copies/swab, P < .0001).

Gardnerella Species: Sensitivity and Specificity for BV

Gardnerella 16S rRNA gene detection at baseline had 99.0% sensitivity, 25.3% specificity, and a risk ratio (RR) of 18.4 (95% CI, 3.6–104.3; P < .001) for symptomatic BV. Gardnerella species group detection using cpn60 assays at baseline had the following sensitivities, specificities, and RR for symptomatic BV, respectively: G. vaginalis (98.0%, 49.3%, 21.5, P < .001), G. piotii/pickettii (91.1%, 56.0%, 6.0, P < .001), G. swidsinskii/greenwoodii (82.2%, 57.3%, 3.3, P < .001), and G. leopoldii (64.4%, 74.7%, 2.6, P < .001). Detection of 3 or more Gardnerella species groups at baseline had 91.1% sensitivity, 68.0% specificity, and a RR of 8.1 (95% CI, 4.4–15.4; P < .001) for symptomatic BV. Detection above the median concentration of positive samples for each Gardnerella species group improved specificity for symptomatic BV but decreased sensitivity (Table 2). Similar results were found when using Nugent score for BV diagnosis (Supplementary Table 4).

Cooccurrence of Gardnerella species

Gardnerella species groups were more frequently detected together in Amsel BV-positive participants than Amsel-negative participants (Figure 3A); 91.1% of participants with symptomatic BV had 3 or more Gardnerella species groups detected at baseline compared to 32.0% of BV-negative participants (P < .0001; Table 2). Most Amsel-positive participants had either all 4 Gardnerella species groups detected (46.5%) or G. vaginalis, G. piotii/pickettii, and G. swidsinskii/greenwoodii (29.7%). Of the 32.7% of Amsel-negative participants with 3 or more species groups detected, 45.8% had Nugent scores of 7–10 (BV) and 18.8% had Nugent scores of 4–6 (intermediate microbiota) (Supplementary Figure 2). Concentrations of G. vaginalis, G. piotii/pickettii, and G. swidsinskii/greenwoodii were all positively correlated with each other at baseline in BV-negative participants, but not in BV-positive participants (P < .0001). In those with BV, concentrations of G. swidsinskii/greenwoodii and G. leopoldii were negatively correlated (P < .001; Figure 3B). Study participants reporting a new sex partner within 60 days prior to baseline had significantly more species groups detected at baseline (2.6 vs 2.2, P = .03).

Gardnerella Species and Subsequent Development of BV by Amsel Criteria

Only 3 of 81 (3.7%) Amsel-negative participants with 1 or more clinic follow-up visits and 2 or fewer Gardnerella species groups detected at baseline developed symptomatic BV at follow-up within 100 days (mean 55.3 [SD 30.9] days) compared to 26 of 43 (60.5%) Amsel-negative participants with 3 or more groups detected (53.7 [SD 25.1] days) (P < .0001). Participants with 3 or more Gardnerella species groups detected at baseline were also more likely to have reported a history of BV compared to participants with 2 or fewer groups detected (72.8% vs 48.6%). Median concentrations of G. vaginalis (7.24 vs 5.14 log₁₀ copies/swab, P < .001), G. piotii/pickettii (6.88 vs 4.81 log₁₀ copies/swab, P < .001), and G. swidsinskii/greenwoodii (8.40 vs 5.60 log₁₀ copies/swab, P < .001) were significantly higher at baseline in Amsel-negative participants with follow-up who subsequently developed BV within 100 days compared to those who did not (Figure 4).

Gardnerella Species Acquisition and Loss

In the 42 participants with weekly sampling data (16 Amsel positive, 26 Amsel negative), G. vaginalis and G. swidsinskii/greenwoodii were the most prevalent species groups at baseline (73.8% each), followed by G. piotii/pickettii (64.3%) and G. leopoldii (28.6%). G. leopoldii was the most frequently lost species with 75.0% of participants with detection at baseline having at least 1 weekly time point with concentrations below assay threshold, followed by G. piotii/pickettii (44.4%), G. vaginalis (35.5%), and G. swidsinskii/greenwoodii (29.0%). In participants without detection of a Gardnerella species group at baseline, G. swidsinskii/greenwoodii was the most acquired species (any weekly time point) (45.5%), followed by G. leopoldii (26.7%), G. piotii/pickettii (20.0%), and G. vaginalis (18.2%) (Supplementary Table 5). Instances of G. swidsinskii/greenwoodii acquisition or redetection were significantly higher in participants reporting vaginal intercourse on the same day or week prior (12 of 232 [5.2%] versus 4 of 256 [1.6%]; P = .04). We did not see statistically significant associations with acquisition or loss of other Gardnerella species groups and menses or sexual activity.

Metronidazole Treatment Response

There were 44 episodes of symptomatic BV with subsequent metronidazole treatment (24 episodes treated with metrogel once a day at bedtime for 5 days, 19 with metronidazole 500 mg orally twice a day for 7 days, and 1 with both) in 26 of 42 weekly longitudinal participants. Sixteen participants experienced additional or continuing episodes of symptomatic BV during the 12-week follow-up period and had more than 1 BV episode represented. G. vaginalis, G. piotii/pickettii, G. swidsinskii/greenwoodii, and G. leopoldii were detected in 44, 38, 42, and 20 BV episodes, respectively. Concentrations of each Gardnerella species group were significantly lower in the week after metronidazole treatment compared to prescription date (Figure 5). Of BV episodes, 8.9%, 23.7%, 4.8%, and 35.0% with G. vaginalis, G. piotii/pickettii, G. swidsinskii/greenwoodii, and G. leopoldii detection, respectively, resulted in no detection by week 2 postprescription, and redetection by week 3 occurred in 75.0%, 44.4%, 0.0%, and 28.6% of those episodes, respectively.

DISCUSSION

The mechanisms underlying the onset of BV remain unclear and despite antibiotic treatment more than 50% of women may experience recurrent BV within a year of initial diagnosis [34]. Gardnerella is a prominent member of the vaginal microbiota and is almost always present in BV. BV pathogenesis models have been proposed that include Gardnerella as a keystone species in BV development that may promote colonization by other BV-associated bacteria, and which may be sexually transmitted [8, 14]. However, the high prevalence of Gardnerella species in women without BV raises questions about whether Gardnerella is a true pathogen or an opportunist.

Currently, there are limited studies examining the roles of different Gardnerella species in the development of BV and most of these studies have utilized qPCR assays developed by Balashov et al targeting 4 previously defined Gardnerella clades, yielding varying results regarding which clades are associated with BV [10, 13, 21–24] (Supplementary Table 6). It has also been noted that these assays may not detect several more recently characterized clade 2 isolates [20]. Only 1 other published study to date has utilized species-specific Gardnerella qPCR assays targeting the cpn60 gene [20]. The combination of varying study populations, use of Amsel criteria versus Nugent score for BV diagnosis, and potential underdetection of newly described Gardnerella species could all contribute to the heterogeneity of results from these studies.

In this study, we developed qPCR assays targeting the cpn60 gene of 4 Gardnerella species groups and applied them to a cross-sectional set of vaginal swabs from study participants with and without BV, as well as a longitudinal set of weekly samples. We found increased prevalence and concentrations of all 4 Gardnerella species groups in participants with BV in addition to higher numbers of species groups detected, consistent with previous study results [10, 13, 21–24]. G. vaginalis was detected in 98.0% of participants with symptomatic BV but was also detected in 50.7% of Amsel-negative participants, although usually at lower concentrations. Despite being nearly always present in BV, G. vaginalis was always detected with at least 1 other Gardnerella species group at baseline in participants with symptomatic BV, suggesting that the presence of G. vaginalis alone may not be sufficient to facilitate BV. Most BV-positive participants had G. vaginalis, G. piotii/pickettii, and either 1 or both of G. swidsinskii/greenwoodii and G. leopoldii detected, although multiple Gardnerella species could also be detected for extended periods of time before symptomatic BV was diagnosed. Many participants who were Amsel negative with 3 or more Gardnerella species groups detected had Nugent scores indicative of BV or intermediate microbiota, and more than 60% of those participants with follow-up would develop symptomatic BV within 100 days compared to 3.7% of those with 2 or fewer Gardnerella species groups detected. These data suggest that vaginal colonization with multiple Gardnerella species is a risk factor for development of BV.

Interestingly, G. swidsinskii/greenwoodii and G. leopoldii were less frequently detected together than other Gardnerella species, consistent with findings from another study [19]. We also noted that concentrations of these 2 species groups were negatively correlated in BV-positive participants, and they were almost never exclusively detected together unless 1 or both of G. vaginalis or G. piotii/pickettii was also detected. Past studies utilizing qPCR to quantify Gardnerella clades or species have not differentiated between G. swidsinskii and G. leopoldii [10, 13, 20–24]. These 2 species were previously classified within the same clade, but matrix-assisted laser desorption ionization time-of-flight (MALDI-TOF) and whole-genome sequencing have determined these to be separate species [17]. A previous study comparing these 2 species noted the appearance of a polysaccharide-like capsule in G. swidsinskii that was absent in G. leopoldii as well as differences in cytotoxicity and epithelial cell adhesion in vitro [35]. G. leopoldii was the least prevalent Gardnerella species group detected in our study population and was frequently lost in our longitudinal cohort, suggesting it may be a more transient member of the vaginal microbiota. Interestingly, G. swidsinskii/greenwoodii was the most frequently acquired species group, and acquisition events were significantly higher in participants reporting vaginal intercourse in the prior week, although this was driven by low sample numbers. More research is needed to explore how G. leopoldii and G. swidsinskii may be interacting in the vagina (such as with competitive exclusion) and whether G. swidsinskii is more likely to be sexually transmitted.

Overall, our results suggest that BV is associated with increased Gardnerella diversity and concentration, and do not suggest that these 4 Gardnerella species groups are specific for BV, although the existence of other, more virulent Gardnerella strains is possible as others have suggested [14, 15, 36]. The presence of multiple Gardnerella species was necessary for symptomatic BV to manifest in our study population. This could explain high rates of Gardnerella detection in healthy women and complicates the notion that Gardnerella is a keystone species in BV. Gardnerella has long been associated with biofilm formation, which may facilitate growth of BV-associated anaerobes such as Fannyhessea vaginae and Prevotella bivia [12, 37–39]. Biofilm formation by Gardnerella has been shown to increase in the higher pH range typically found in BV and may be suppressed by Lactobacillus species found in an optimal, non-BV state [40]. Variation in biofilm forming potential has been observed between Gardnerella isolates in vitro [36] and Gardnerella isolates are known to possess different virulence factors, such as sialidase and vaginolysin, which may uniquely contribute to BV development [41–43]. Acquisition of new Gardnerella species through sexual activity or other means could promote biofilm formation and growth of other BV-associated bacteria, leading to development of BV. However, a previous study did not see increased biofilm formation in cocultures of multiple Gardnerella clades in vitro compared to Gardnerella monocultures [44], although these cultures did not include any additional BV-associated bacteria such as F. vaginae that may contribute to biofilm formation. There have been no published studies to date exploring virulence potential in cocultures of multiple Gardnerella species.

In the 44 symptomatic BV episodes from our longitudinal cohort, Gardnerella concentrations of all 4 species groups were significantly lower in the week after metronidazole prescription for BV. However, detection below threshold of any Gardnerella species group by week 2 was uncommon, and redetection could occur within a week. Previous studies have shown that some Gardnerella strains are resistant to metronidazole [45, 46], and biofilm formation has also been shown to increase metronidazole and clindamycin resistance of Gardnerella [47, 48]. This limited activity of metronidazole against Gardnerella may contribute to BV recurrence, and more effective treatments that specifically target Gardnerella may be needed to improve treatment outcomes.

There are several limitations to this study including our ability to detect only 4 species groups within the Gardnerella genus via qPCR. We were unable to differentiate between G. vaginalis and genomospecies 2 nor between G. piotii and G. pickettii via cpn60 qPCR, although previous studies have placed these species in the same clade and subgroup [49, 50]. Likewise, we were unable to completely differentiate between G. swidsinskii, G. greenwoodii, and genomospecies 9 and 10, which were nevertheless detected with reduced sensitivity. Previous studies have placed G. greenwoodii and genomospecies 9 and 10 in subgroup D or clade 3 [49, 50], and it has been observed that this clade is not as prevalent nor abundant as the other 3 clades [13, 19, 22], although at least 1 study observed increased growth of subgroup D in the presence of other subgroups in vitro [44]. Additionally, our assays did not detect Gardnerella genomospecies 7, which has been associated with refractory BV following oral metronidazole treatment and may play a role in BV development in some women [20]. In the longitudinal cohort, weekly sampling and incomplete or minimal behavioral data, including antibiotic use other than metronidazole, limited our ability to associate acquisition and loss of Gardnerella species groups with specific behaviors. Lastly, participants enrolled in this study were more likely to have had a history of BV and this may have influenced our results through higher rates of Gardnerella detection.

Many questions remain regarding diversity within the Gardnerella genus and its contributions to vaginal health. Additional studies exploring the Gardnerella pangenome as well as in vitro characterization of different Gardnerella species and isolates will be necessary to determine the genotypic and phenotypic variation within the genus. Given the increased Gardnerella diversity seen in BV, development and utilization of molecular techniques to differentiate Gardnerella species will be vital in understanding the roles these species play in BV etiology. Unraveling the dynamics of competition and cooperation between different Gardnerella species, BV-associated anaerobes, and lactobacilli may lead to interventions that promote vaginal health.

Supplementary Data

Supplementary materials are available at The Journal of Infectious Diseases online (http://jid.oxfordjournals.org/). Supplementary materials consist of data provided by the author that are published to benefit the reader. The posted materials are not copyedited. The contents of all supplementary data are the sole responsibility of the authors. Questions or messages regarding errors should be addressed to the author.

Notes

Author contributions. D. N. F. and S. S. designed the study. M. M., S. M. S., and T. F. performed laboratory work. M. M., S. P., S. M. S., and T. F. assisted in data curation and analysis. M. M. and D. N. F. prepared the manuscript. M. M. and S. P. created graphical figures. All authors contributed to manuscript review and editing.

Acknowledgments. We thank the clinicians and study coordinators who assisted with study enrollment and sample collection, as well as all study participants who contributed to the study. We also thank D. J. Valint and Congzhou Liu for their assistance with DNA extraction of samples.

Data availability. Data available upon request.

Financial support. This work was supported by the National Institutes of Health (grant number R01 AI061628 to D. N. F.). Funding to pay the Open Access publication charges for this article was provided by the Fred Hutchinson Cancer Center.

References

Author notes