Bacterial vaginosis is the most common cause of vaginitis and vaginal discharge, with a prevalence of up to 29% of women [1].This infection is associated with multiple serious health outcomes, including human immunodeficiency virus and other STI acquisition, human papillomavirus infection [2],post-abortion pelvic inflammatory disease, post-cesarean endometritis, low birth weight, preterm labor, and preterm delivery [3,4]. A high rate of recurrence is among the greatest clinical challenges with BV, reportedly very high despite standardized treatment with oral or intravaginal metronidazole or clindamycin, frustrating patients and clinicians [5,6]. Treatment failure results from diagnostic error, re-infection from sexual partners, the presence of antimicrobial resistance genes, non-genetic determinants of resistance, and drug resistance of the BV-infected biofilm [7]. A complete and accurate description of the composition of the complex polymicrobial vaginal microbiome can only be obtained by multiplex next-generation sequencing (NGS) testing, thereby assisting with the clinical challenge of BV [8]. These tests have recently shown that BV consists of a spectrum of multiple bacterial community clusters [8-10] dominated by Gardnerellavaginalis, Lactobacillus iners, or anaerobic combinations with or without Lactobacillus spp. in varying proportions [10]. Also, multiple significant BV-associated bacteria had previously escaped notice with all other detection methods such as cultures; undetected were Atopobiumvaginae and A. parvulum, Lactobacillus iners and other Lactobacillus spp., and multiple Lachnospiraceaespp (BVA1, BVA2, and BVA3) [11].

In addition to identifying all incident bacterial species, 16S ribosomal RNA gene sequencing (NGS) allows numerous other improvements compared with existing methods of detection [12]. Advantages include measurement of the relative proportion and absolute number of bacteria present (both pathogens and normal commensals), as well as identification of antimicrobial resistancedetermining genes for the individual patient (precision medicine). In the following case studies, drawn from clinical experience with more than 2000 cases diagnosed with NGS, the utility of NGS testing in management of individual patient quandaries in BV was demonstrated. In each case, NGS provided novel information on the spectrum of infection and drug resistance. This study was approved by the Institutional Review Board of American International Biotechnology LLC, Richmond, Virginia.

An otherwise-healthy 39 year-old with a prior history of recurrent BV treated empirically with metronidazole presented with a two-week history of irritation and thin watery grey-white vaginal discharge.Amsel criteria as modified by Gutmanet al.[13] were positive for BV. Vaginal swab for NGS testing was obtained, but she was empirically placed immediately on metronidazole. NGS results subsequently showed recurrent BV, with 56% mixed anaerobes (Atopobiumparvulum, Dialistermicraerophilus, Enterococcus faecalis, Megasphaeramicronuciformis,Prevotelladisiens, Prevotellatimonensis, Sneathiasanguinegens) 22% Gardnerellavaginalis, 20% Lactobacillus iners, 3% Atopobiumvaginae, and antimicrobial drug resistance for clindamycin and macrolides (ermTR). Metronidazole was continued for a total of ten days.

She returned 3 months later with recurrent BV symptoms, with positive modified Amsel criteria. NGS testing revealed recurrent BV: 40% Gardnerellavaginalis, 46.5% mixed anaerobes (Enterococcus faecalis, sulfureus,and villorum; and Streptococcus oralis, mitis, and pseudopneumoniae); 12% Lactobacillus “other” (non-crispatus and non-iners); and 2.5% Mycoplasma hominis. Antimicrobial drug resistance remained positive for clindamycin and macrolides (ermTR). Metronidazole treatment was given again for a total of ten days.

Discussion and conclusion of case 1

Empirical treatment of BV with metronidazole and/or clindamycin has one of the lowest success rates for patient management in all of microbiology, with 50%- 80% recurrence within 12 months [14-16]. In a controlled clinical trial of 14-days of metronidazole for symptomatic BV, Schwebe and Desmond reported 38% failure within one week and 57% within three weeks [17]. These findings caused Eschenbach to conclude that “…treatment of BV… is in a sorry state…” [18].

He noted that antibiotic cure with metronidazole was greater than 90% when it was first introduced, but subsequent decades have seen this drop to 50-80%, with most contemporary reports at the low end of this range: “How can cure rates steadily decrease over 30 years?” Similarly, Austin et al. reported an increase in baseline clindamycin resistance over two decades [19].

Carr et al. argued that incomplete diagnosis of BV was clinically and economically inefficient, resulting in both empirical under-treatment and over-treatment, costly outcomes that cause unnecessary side effects, delayed diagnosis, heightened risk of AMR (antimicrobial resistance), and increased likelihood of recurrence [20] (Figure 1).

Figure 1: Vaginal microbiome: normal reference standard in non-gravid women during reproductive years (left); at presentation with BV (center); and with subsequent recurrent BV (right) in Case 1. Normal reference standard consists of lactobacilli with a predominance of L. crispatus . At presentation, BV was characterized by predominantly G. vaginalis and mixed anaerobes, with a minority of L. iners . After treatment that failed to account for drug resistant genes and the entire spectrum of pathogens present, recurrentBV was characterized predominantly by G. vaginalis (increased number) and mixed anaerobes, with loss of L. iners .

Molecular diagnostic testing of 16s ribosomal RNA with next-generation sequencing (NGS) provides a comprehensive description of the vaginal bacterial microbiome and offers promise as a rapid method for targeted genetic identification of multiple AMR genes simultaneously. [21]. NGS testing allowed evaluation of resistance genes in non-culturable bacteria, facilitating study of many of the most abundant BV-associated bacteria found only recently with advanced molecular diagnostics, including Atopobiumvaginae, Atopobiumparvulum, Lactobacillus iners and other Lactobacillus spp., and multiple Lachnospiraceaespp (BVAB1, BVAB2, and BVAB3) [22].

Van de Wijgert et al. suggest that NGS will eventually replace Nugent scoring, the current diagnostic gold standard for BV [10]. It is also possible that NGS testing will change the treatment of BV, driving the use of narrow spectrum agents likely to cause less harmful disruption of the vaginal flora; intermittent therapy; or targeted combination therapy.

A 23 year-old otherwise-healthy woman presented with a one-week history of symptoms of bacterial vaginosis (irritation, discomfort, thin discharge); Amsel criteria as modified by Gutmanet al. [13] were positive for BV. Vaginal swab for NGS testing was obtained, but she was empirically placed immediately on metronidazole (5-nitroimidazole). NGS results subsequently showed no evidence of BV, with the microbiome dominated by L. crispatusand L. iners; AMR gene determination was negative. Metronidazole was discontinued(Table 1). She returned 3 months later with recurrent BV symptoms, with positive modified Amsel criteria. NGS testing revealed mixed infection with BV and Candida albicans. AMR genes that were previously negative had converted to positive (resistance) to 5-nitroimadazoles for BV treatment and triazoles (e.g., fluconazole) for Candida treatment (Table 1).

Table 1: Results of Antimicrobial Resistance-Determining Genes at Presentation of Bacterial Vaginosis and at Recurrence: Case 2.

Discussion and conclusion of case 2

In the initial visit, the patient received metronidazole for only a few days until the NGS test was reported as negative, but it is possible or even likely that this medication caused profound alterations in the vaginal flora and the emergence of AMR that was observed in the follow-up visit. Mayer et al. recently reported that “…under antibiotic pressure, the vaginal microbiome undergoes rapid shifts in a matter of hours…” [23].

They noted that anaerobes cleared more rapidly than Gardnerella vaginalis , but BV-associated bacteria quickly re-emerged after cessation of therapy, and suggested the possibility of intermittent prophylactic treatment (Figure 2).

Figure 2: Vaginal microbiome at initial presentation (left) and at recurrence with BV (right) in Case 2. At presentation, NGS revealed normal microbiome consisting of lactobacilli , especially L. rispatus . After unnecessary empirical treatment that could have been avoided with NGS, there was a profound alteration of the microbiome with predominance of BV-associated bacteria and near-complete loss of lactobacilli ; Candida was also present (not shown).

A major goal in tackling antimicrobial resistance (AMR) is rapid and targeted detection and diagnosis of infection and resistance for patient benefit, infection control, surveillance, and prudent antibiotic stewardship, including identification of alternative drugs unencumbered by the presence of AMR genes [24]. Major initiatives were recently launched for controlling AMR: the U.S. National Action Plan for Combating Antiobiotic Resistant Bacteria (March, 2015) and the World Health Organization to Tackle Antimicrobial Resistance (May 2015).

According to Mullaney, “…antibiotic resistance is a major threat to human health and well-being.” He noted that the only way to effectively combat AMR is to understand the complete range of different resistance genes that allow bacteria to resist antibiotics; to do this, the entire microbiome needs to be investigated, and NGS is the only method that provides for such an investigation [25].

We recently reported that antimicrobial resistance (AMR) genes are common in the vaginal microbiome, especially in women with BV when compared with symptomatic women without BV; there was an increase in overall abundance (74.3% vs. 22.1%, respectively) and number of AMR genes (mean, 1.5 vs. 0.3, respectively) [21]. There were varying levels of susceptibility for the two primary drugs used to treat BV, with the highest level of AMR genes identified (61.8%) with clindamycin, one of the lincosamides.

AMR gene analysis is in widespread clinical use in management of other infectious diseases, with reported sensitivity and specificity of 72.3-100% and 55-100%, respectively [24]. Concordance was high between the phenotypic method (MIC determination) and the presence of the ermX gene for Corynebacterium isolated from the nasal mucosa, noting that this was the most common mechanism of resistance [26]. According to the Clinical and Laboratory Standards Institute, “…tests for mecA… are the most accurate… for prediction of resistance to oxacillin [27].” AMR gene analysis was superior to routine methods for determination of low-level drug resistance for Mycobacterium tuberculosis [28].

An otherwise-healthy 34 year-old woman presented with a one-week history of watery vaginal discharge and irritation. Amsel criteria as modified by Gutmanet al. [13]. were positive for BV. Vaginal swab for NGS testing was obtained, but she was empirically placed immediately on metronidazole. NGS results subsequently showed BV, with 54% BVAB1, 41% Lactobacillus iners, 4% mixed bacteria (Enterococcus avium, hermanniensis, hirae, and villorum), and 1% Lactobacilli “other” (non-crispatus, non-iners). In addition, significant co-infection with Candida albicans was identified (fungal and parasitic DNA sequence reads are not included in the percentages of bacterial reads). Antimicrobial drug resistance (MDR1 gene) was positive for triazoles (fluconazoles). Metronidazole treatment was continued for a total of ten days.

Discussion and conclusion of case 3

This case illustrates two important issues with BV: (1) G. vaginalis is neither necessary nor sufficient for diagnosis; and (2) Mixed infections are a frequent and confounding management problem with BV. Prior to the discovery of BVAB1 in 2005 [29], this patient may have been classified as having pure yeast vaginitis, with high probability of incomplete treatment, persistence of BV, and subsequent recurrence.

How often is G. vaginalis missing in BV? Srinivasanet al. found a prevalence of 3% non- G. vaginalisBV in symptomatic women in an STI clinic in Seattle [9], similar to the 5% described among women in a pregnancy follow-up clinic [30]. If one relies on current methods of BV identification that rely on only G. vaginalisfor diagnosis (e.g., BD Affirm VPIII) it would have given a false-negative result in this case.

How often G. vaginalisis present in women without BV? The prevalence in virginal women was 28% [31] to 45% [32], but these studies were criticized for relying on self-reporting of sexual status [33]. Schwebkeet al. found G. vaginalisin 38.5% of asymptomatic women with a “healthy” vaginal microbiome (predominance of Lactobacilli and a Nugent score of 0-3) [34]. After pregnancy, G. vaginaliswas observed in 47.3% of women [30]. In symptomatic women without BV, the prevalence of G. vaginaliswas 20% (STI clinic in London, with bacterial assessment by culture) [35]. Vaginal inoculation by pure culture of G. vaginalisproduced symptoms of BV in 8% [36] and 41% of women, compared with 73% inoculated with BV-infected secretions [36] (Figure 3).

Figure 3: Vaginal microbiome at initial presentation with BV in Case 3. Candida albicans coinfection was also identified (not shown). See Figure 1, left, for normal reference standard. Recurrence reveals an otherwise-typical case of BV with mixed infection, but lacking G. vaginalis.

The frequent presence of G. vaginalisin asymptomatic women is an accepted finding, but its significance is uncertain. Does this represent incipient infection, resolving infection in transition phase, a low-virulence strain that usually acts as a normal commensal and only manifests virulence with loss of Lactobacilli and/or when combined with another factor such as the biofilm, or a normal-virulence strain held in check by the microbiome and other environmental factors? Importantly, G. vaginaliselicits a modest immune response similar in magnitude to Lactobacilli, in contrast with the robust response from Atopobiumvaginae and other pathogens [37].

It is critical to consider the possibility of mixed infections in patients with presumptive BV that may include STI such as Chlamydia to avoid the serious threat of pelvic inflammatory disease or other complications. The prevalence of vaginal colonization with Candida spp.is higher in women without BV than in those with BV, but the opposite is true for T. vaginalis[10]. Using NGS, we were unable to confirm these findings, but differences in results may be owing to different patient populations or insufficient number of patients to identify significant variance [21]. We also found that almost one-fourth of symptomatic women had co-existent Candida spp. (after excluding clinical cases of pure vulvovaginal candidiasis), usually Candida albicans, with no difference between BV and non-BV samples (p = 0.8). Trichmonasvaginalis and STIs were rare, precluding comparison of prevalence.

Next-generation sequencing (NGS) provides comprehensive description of the complex polymicrobial vaginal microbiome and associated antimicrobial resistance-determining genes, facilitating personalized diagnosis and therapy for patients with complicated bacterial vaginosis, STIs, and vaginitis; it will likely replace the nowinadequate vaginal culture in the near future.