RF Lamont, Perinatology Research Branch, NICHD, NIH, DHHS, Bethesda MD and Brush Building, Level 4, 3099 John R, Detroit, MI, 48201, USA. Email email@example.com
Please cite this paper as: Lamont R, Sobel J, Akins R, Hassan S, Chaiworapongsa T, Kusanovic J, Romero R. The vaginal microbiome: new information about genital tract flora using molecular based techniques. BJOG 2011;118:533–549.
Vaginal microbiome studies provide information that may change the way we define vaginal flora. Normal flora appears dominated by one or two species of Lactobacillus. Significant numbers of healthy women lack appreciable numbers of vaginal lactobacilli. Bacterial vaginosis (BV) is not a single entity, but instead consists of different bacterial communities or profiles of greater microbial diversity than is evident from cultivation-dependent studies. BV should be considered a syndrome of variable composition that results in different symptoms, phenotypical outcomes, and responses to different antibiotic regimens. This information may help to elucidate the link between BV and infection-related adverse outcomes of pregnancy.
In normal pregnancy, the resident vaginal microbial flora is thought to provide protection against infection by a number of different mechanisms.1 In non-pregnant women, the presence of bacterial vaginosis (BV) is associated with an increased risk of upper genital tract and sexually transmitted infections2–4 and with the acquisition of HIV.5–9 In pregnancy, BV increases the risk of post-abortal sepsis,10 early miscarriage,11 recurrent abortion,12 late miscarriage,12,13 preterm prelabor rupture of membranes (PPROM),14 spontaneous preterm labor (SPTL) and preterm birth (PTB),13,15–26 histological chorioamnionitis,27,28 and postpartum endometritis.29,30 As a result, abnormal vaginal flora may predispose women to ascending colonisation of the genital tract, infiltration of the fetal membranes, microbial invasion of the amniotic cavity,31 and fetal damage.32,33 Preterm birth of infectious etiology is associated with high perinatal mortality and morbidity,34–36 and a high cost to the healthcare system.37–39 Much of our knowledge about the composition of the vaginal microbial flora comes from qualitative and semi-quantitative descriptive studies using cultivation-dependant techniques.40–44 In recent years, the development and introduction of cultivation-independent molecular-based techniques have provided new information about the composition of normal vaginal flora, as well as abnormal colonization of the genital tract, which complements existing knowledge from cultivation-dependent techniques. This review seeks to inform the busy obstetrician and gynaecologist about the background to these new developments, and what additional information they provide. These new data might help to elucidate the composition and function of normal flora, and to appreciate the microbial diversity, diagnosis, and treatment assessment of abnormal genital tract flora, which may affect the outcome of pregnancy.
The human microbiome project
During work on the human genome project, scientists estimated that the number of genes in the human genome needed to code for the proteins required to sustain human physiology would be approximately 100 000. The researchers were somewhat humbled to find only 20 000 protein coding genes, similar to that of the fruit fly. However, if one considers a human to be a superorganism (the sum of all human genes plus those of the microorganisms in, or on us), then the human genome plus the microbial genome (microbiome), which amounts to the collective genome of the symbionts, may include considerably more than 100 000 genes. This is supported by the fact that the human microbiome provides traits that the human organism has not needed to evolve for itself.45,46 In its roadmap for medical research, the National Institutes for Health (NIH) has committed $100 000 000 to the Human Microbiome Project, and plans to investigate five sites: the oral and nasal cavities, the gastrointestinal and genito-urinary tracts, and the skin. The genital tract microbiome will be studied both in health and in disease, and whether the function of the microbiome (metagenome) can be manipulated to influence physiology, and treat disease, will be investigated. The core microbiome, as well as the variable microbiome (influenced inter alia by host lifestyle, genotype, immune response, environment, pathophysiology and transient community members), will be assessed spatially and temporally, and different subsites will be assessed.47 The microbial flora on the labial surface of the teeth are different from those on the lingual surface,48 and the lower vagina, upper vagina and cervix also differ in their microbial flora.49 How this relates to the various niches in the female genital tract in pregnancy, and to the human microbiome project as a whole, is suggested in Figure 1.
Cultivation-independent techniques require us to be aware of different levels of taxonomic classification. Taxonomy is the classification of living organisms into phylogenetic groups. Phylogenetic classifications arrange organisms into groups that reflect genetic similarity and evolutionary relatedness. A taxon is a group, or level of classification, and is hierarchical, whereby broad divisions are divided up into smaller divisions such as Domains, Kingdoms, Phyla, Classes, Orders, Families, Genera and Species. Not every level is used in every discipline. The species is the basic unit of taxonomy, and many molecular-based studies use different terms such as ‘operational taxonomic units’, ‘taxons/taxa’ or ‘phylotypes’ instead of species, because the level of molecular diversity far outstrips that defined by existing microbiological or biochemical means. The human intestine demonstrates the greatest degree of diversity with over 400 phylotypes, only 20% of which are cultivatable.50
Many of the culture-independent studies focus on the detection of novel, previously uncultivated species,51–56 or are more concerned with the novel molecular techniques in themselves rather than for the microbiological information that they provide.57–60 In addition, theoretical and mathematical models have arisen that address new concepts such as ‘functional redundancy’, ‘structural diversity’, ‘interspecies interaction’, ‘mutualism’, ‘cheating’, ‘the insurance hypothesis’, ‘drivers’ and ‘passengers’.61–64
Bacterial DNA is extracted from samples, and is amplified using the polymerase chain reaction (PCR) using either universal or specific primers. This is not infallible because PCR inhibitors may be present at varying levels in individual clinical samples, and primers may preferentially amplify certain nucleic acids, and may not complement the entire bacterial kingdom.65
The commonest target for molecular identification of bacteria is the small ribosomal subunit of the 16S rRNA gene. The 16S rRNA gene is useful because it is present in all bacteria, and has regions of conserved sequence that can be targeted by universal (sometimes referred to as bacterial-domain or broad-range primers) or specific primers, yet also has areas of heterogeneity that can be used to identify bacteria or to infer phylogenetic relationships.66–71 If there is uncertainty with respect to which organisms might be present in a sample, or if a broad diversity of organisms are anticipated, then many researchers will use universal primers.53 If, on the other hand, an investigator knows what organisms to expect, or wishes to test for specific organisms, then specific primers will be used.72 Once the 16S rRNA gene has been sequenced, the variable regions can be used for species-specific PCR in a qualitative or quantitative manner. The sequences obtained are aligned and compared with large databases of 16S rRNA sequences, although diversity for vaginal microbial flora is poorly represented in these databases compared with other sites like the gastrointestinal tract.54,60
Cultivation-based studies have the disadvantage that they may fail to isolate or detect large numbers of fastidious microorganisms, or identification tests may not be available. An unknown number of species that are identified purely by molecular methods are cultivatable, but have not been identified by cultivation methods, because of a lack of phenotypic tools for the species, their lower relative titers, or inappropriate media.73
This is an important distinction, as it shifts the focus of new research from developing novel culture media to using molecular tools to identify the under-characterised colonies that already grow in current formulations. This also restricts the development of their phenotypic profiles and hence the understanding of their roles in the whole population. In addition, this suggests that before the development of molecular techniques, culture-based assessments of the vaginal microbiome did not ‘miss’ 80% of the species, but simply had no labels to impose on many colonies. As a result, these organisms were included only as members belonging to a broad phenotypic or phylogenetic branch.
Cultivation-independent techniques may show greater diversity by overcoming cultivation problems and the identification of fastidious organisms, but are limited by their tendency to sample only the most prevalent bacteria in a community, such that low-abundance or minority species are likely to be missed.74 As PCR amplification of DNA is a competitive enzymatic reaction, the 16S rRNA templates in a sample are amplified in accordance with their abundance. As a result, the 16S rRNA genes of the numerically dominant population will be the most abundant amplicons following PCR. Populations that constitute <1% of the total community (yet may still be present in concentrations of >106/g of vaginal fluid) may not be represented in such profiles, so this represents the threshold of detection. This being the case, despite their limitations, cultivation studies remain an important part of vaginal microbiology, and will need to be used in combination with cultivation-independent techniques.75 The development of the next generation of ultra-high throughput sequencing technologies (pyrosequencing) may remove an important quantitative barrier by increasing the number of reads from a gene or genome by many orders of magnitude in one experimental run.76 The power of next-generation sequencing, however, is not without limitations. Until recently, error rates using pyrosequencing were high enough to generate reading errors that indicated false microdiversity at the species or subspecies levels. However, improvements in the technology in third-generation systems now produces sequencing with error rates similar to Sanger sequencing, which should minimise this problem in future experiments. Even this improvement does not overcome the best read lengths among these techniques (approximately 350–400 bases). The variable domains of 16S rDNA, which allow the best species discrimination, are spread out over >1 kb, so only some can be captured in this run length. This limits discrimination to the genus level for some phylogenetic branches, with minimal diversity at this locus.77 This obstacle remains a stumbling block for most of the next-generation sequencing methods, but less so as read lengths are steadily improving through technical advancements. For an insight into the pros and cons of the various techniques and their limitations the reader is directed to three excellent reviews.78–80
Normal vaginal flora
Most studies, whether cultivation-dependent or -independent, give the impression that the vaginal microbial flora is static, because most studies are carried out as a ‘snapshot’ in time, and do not consider that vaginal microbial communities undergo shifts in their representation, abundance and virulence over time, and are affected by many factors.
Identification of lactobacilli to the species level using molecular-based techniques
Culture and microscopy of ‘normal’ vaginal flora typically shows a predominance of Lactobacillus species, which are believed to promote a healthy vaginal milieu by providing numerical dominance, but also by producing lactic acid to maintain an acid environment that is inhospitable to many bacteria, and which is negatively correlated with BV.81 Lactobacilli also produce hydrogen peroxide (H2O2),82 antibiotic toxic hydroxyl radicals, bacteriocins,83 and probiotics.84 Prior to molecular-based techniques, lactobacilli were generally identified only to the genus level.
In 1892, Professor Albert Döderlein85 (1860–1941) published his definitive monograph in which he recorded that cultured organisms were a source of lactic acid that could inhibit the growth of pathogens in vitro and in vivo. In 1928, Stanley Thomas86 identified Döderlein’s bacillus as Lactobacillus acidophilus, adding prophetically, that this was either a characteristic group of related species, or a species that underwent a remarkable transformation. In 1980, in keeping with Thomas’ observation, a group of organisms previously collectively known as L. acidophilus was shown to be highly heterogeneous.87 As a result, the group was divided into DNA-homologous groups that could not be distinguished biochemically,88 to form a number of separate species within the L. acidophilus complex (Box 1).89–91 The closely related species within the L.acidophilus complex are difficult to differentiate by phenotypic methods, which may account for the variation in species of lactobacilli found in different studies.92–94
Cultivation-based techniques, because they fail to detect fastidious organisms, underestimate the diversity of vaginal microbial flora, but because of deficiencies in the phenotypic identification of lactobacilli, they overestimate the diversity of Lactobacillus species in the vagina.95
Some 20 years ago, using cultivation-based phenotypic techniques, Redondo-Lopez et al.93 concluded that no two women were colonised by the same two Lactobacillus species. Using cultivation-independent techniques this now appears not to be the case, and because of their significant role in health and disease, much attention has been given to the identification of lactobacilli using genotypic means.51,65,92,96–99
Table Box 1.. Obligately homofermentative species within the Lactobacillus acidophilus complex
Lactobacillus iners: under-detected and under-appreciated
The existence of L. iners was unknown prior to 1999, but is now known to play a significant role in vaginal microbial flora. Selective media such as Rogosa or MRS agar are normally used to culture lactobacilli, so the use of cultivation-based techniques, even those followed by molecular methods, will not detect L. iners because it only grows on blood agar.91 Hence, some very important molecular-based studies failed to isolate L. iners because Rogosa or MRS agar were used, rather than blood agar.51,96,100–102
The first report of the isolation of L. iners in a woman with a normal Nugent score was in 2002.65 Cultivation-independent methods have identified L. iners, a lactic acid-producing bacterium, as one of the organisms most frequently isolated from the vagina of healthy women.62,65,98,99,103–105 In contrast to Lactobacillus crispatus, which is rarely dominant in BV,53L. iners can be detected at high levels in most subjects with and without BV,53,65,106,107 and in three studies it was the only Lactobacillus species detected in BV-positive women.52,53,106 It has been postulated that this may be because L. iners may be better adapted to the conditions associated with BV, i.e. the polymicrobial state of the vaginal flora and elevated pH.106 Alternatively, the observations could result from relative resistance of L. iners to unknown factors that lead to the demise of other Lactobacillus species during the onset of BV, or to a relative lack of antagonism of L. iners to the BV-associated anaerobes, so that their dominance predisposes the individual to acquiring BV.
Numerical supremacy of lactobacilli
Over 120 species of Lactobacillus have been identified, and more than 20 species have been detected in the vagina. Using molecular-based techniques, we now know that healthy vaginal microflora does not contain high numbers of many different species of Lactobacillus. Rather, one or two lactobacilli from a range of three or four species (mainly L. crispatus and L. iners, but also Lactobacillus jensenii and Lactobacillus gasseri) are dominant, whereas other species are rare, lower in titer, and tend to be novel phylotypes.51,52,55,62,65,98,101,102,104,107–110
In healthy Swedish women with a normal Nugent score,111 202 vaginal isolates were tested against 26 type and reference strains of Lactobacillus.98 In 18 out of 23 women, the vaginal flora was dominated by a single species of Lactobacillus, and only five women had two different species or two different strains of the same species of Lactobacillus. The only species detected were L. crispatus, L. gasseri, L. iners, and L. jensenii.98 In a follow-up study only one women was colonised by more than two Lactobacillus species, four were colonised by two different species, and 17 were colonised by one single species.99 Although this study was severely limited by the small number of colonies examined per patient (ten from blood; 100 from Rogosa agar), the exclusion of other species is in keeping with the theory of ‘competitive exclusion’,55 and the superior ability of L. iners and L. crispatus to compete with other bacteria for vaginal resources, a survival strategy known as ‘bacterial interference’.112 Alternatively, the rare coexistence of multiple dominant species of Lactobacillus could result from pre-emptive colonisation by a particular species, or from host factors that strongly influence which species are able to colonise the environment.
Cultivation-independent studies using molecular techniques have been published on different populations, such as adolescent girls,55 post-menopausal women,65,113 and women from Belgium,52,108,110 Brazil,114 Bulgaria,95 Canada,57,65,113 China,109 Germany,105 Holland,115 India,116 Italy,100 Japan,101,117 Nigeria,103,118 Sweden,98,99 both Turkey and the USA,119 the USA,54,55,58,62,96,104 African American women,106 white and black North American women,104 and a multinational group of women from seven different countries.51
Racial variation and geographical area are important,51 and different racial groups within the same geographical region have significant differences in what is the dominant vaginal organism.103 In most populations, L. crispatus is the most common dominant isolate,51–53,58,62,98,104,105 and white women are more likely to be dominated by L. crispatus and/or L. jensenii than any other species of Lactobacillus.96 A number of genetic as well as environmental factors might explain at least part of this observation. Alternatively, diet might influence the Lactobacillus species resident in the gastrointestinal tract, and hence the vagina, as the lactobacilli of the gut varies between Japanese and Western women.120–122
Production of H2O2 by lactobacilli, and the association with bacterial vaginosis
Lactobacilli differ in their ability to produce H2O2, and a reduction in the prevalence and concentration of H2O2-producing bacteria is associated with the development of BV and vaginal infections.82,123 With the introduction of molecular-based techniques, it is now possible to relate the production of H2O2 to individual species or strains of the same species of Lactobacillus, rather than the genus as a whole. Three studies using molecular-based techniques have addressed the production of H2O2 by lactobacilli.96,101,102 In Japanese women who did not have BV, L. crispatus and L. gasseri were found in the vagina of 52.7 and 20.8% of women, respectively. Lactobacillus jensenii was not detected. All strains of L. crispatus were strongly positive for H2O2, whereas 41 and 59% of L. gasseri strains were strongly and weakly positive for H2O2, respectively.101
Antonio et al.96 demonstrated that Lactobacillus species detected among 302 women with and without BV differed in their ability to produce H2O2. Lactobacillus crispatus and L. jensenii were found to colonise 32 and 23% of women, respectively, and 95 and 94% of their strains, respectively, were shown to produce H2O2. In contrast, L. gasseri and L. iners colonised 5 and 15% of women, respectively, and only 71 and 9% of their strains, respectively, produced H2O2. Not surprisingly, BV was present in 9 and 7% of women colonised by L. crispatus and L. jensenii, respectively, and in 43 and 36% of women colonised by L. gasseri and L. iners, respectively. Of the women without BV by Nugent score,111 16% had no lactobacilli present, and none of the women colonised by L. crispatus and L. jensenii had BV; however, the latter argument is circular, as the Lactobacillus morphotype is part of the Nugent score. The association between L. gasseri and BV has been confirmed in a study of homosexual women. Detection of L. gasseri was associated with a 4.2-fold increased risk of BV,124 and was attributed to a higher rectal colonisation by L. gasseri and sexual practices that increase the risk of vaginal colonisation from the rectum.125
Using culture-based techniques, Eschenbach et al.123 and McGroarty et al.126 found that 100% of L. jensenii produced H2O2, yet Nagy et al.127 found only 46% produced H2O2. Similarly, with L. acidophilus the range of species that produced H2O2 varied from 43 to 77%.123,126,127 This may have been because of the inability of biochemical assays to differentiate between the species belonging to the L. acidophilus complex. However, this led Nagy et al.127 to conclude that the ability of lactobacilli to produce H2O2 was associated with the origin of strain (whether from women with or without BV) rather than the Lactobacillus strain itself. In light of the findings of molecular-based techniques, and the current ability to identify H2O2-producing lactobacilli to species level,96 we might alternatively conclude that it is whether or not the strain/species of Lactobacillus produces H2O2 that dictates whether BV is present or absent. However, given that H2O2-producing L. gasseri are found in BV patients, albeit at lower incidence, one might also argue that in vitro production of H2O2 is only a biomarker of a protective species of Lactobacillus, not an active factor in limiting the growth of vaginal anaerobes. Indeed, it is not clear whether bacterial H2O2 production is active in the microaerobic to anaerobic vaginal flora. Other factors may include the extent to which a species can dominate numerically over its competitors, e.g. 95 versus 99.99%.
Healthy vaginal flora not dominated by lactobacilli
Using culture-independent techniques, several investigators have demonstrated that a significant proportion (7–33%) of healthy women lack appreciable numbers of Lactobacillus species in the vagina,52,58,62,103,104 which may be replaced by other lactic acid-producing bacteria such as Atopobium vaginae, Megasphaera, and Leptotrichia species.62,104 Although the structure of the communities may differ between populations, health can be maintained provided the function of these communities, i.e. the production of lactic acid, continues.55,62 Consequently, the absence of lactobacilli or the presence of certain organisms such as Gardnerella vaginalis, or species of Peptostreptococcus, Prevotella, Pseudomonas, and/or Streptococcus, does not constitute an abnormal state.58 This issue is still unclear, however, because the studies do not address whether some proportion of ‘healthy’ women are patients in transition to or from BV, or whether they have asymptomatic BV, i.e. abnormal flora but no symptoms because of genetic or other factors. Indeed, a recent molecular study further confuses the issue by pointing out that G. vaginalis may produce transient dominance in healthy women as a result of perturbations, such as an increase in pH, during menstruation.128
Vaginal probiotic lactobacilli
The first example of vaginal probiotics use we could detect was that of Stanley Thomas in 1928 following his observation that lactobacilli were absent in the presence of gonococci. He reported on two experiments, one in vitro and the other in vivo, in which the addition of a thin layer of whey broth from a culture of L. acidophilus demonstrated the eradication of Neisseria gonorrhoeae.86
Exogenous strains of lactobacilli have been suggested as a means of establishing or re-establishing normal vaginal flora. Lactobacillus fermentum and Lactobacillus rhamnosi probiotic strains have been used with poor results in urogenital infection, and this may be because they are not normally prevalent in the vagina.105,129,130 In contrast, L. crispatus might be a better choice, because it is commonly found to be numerically dominant in the healthy vagina, and 95% of the strains produce H2O2.96Lactobacillus crispatus may have a superior capacity to persist in the vagina,131 and strains of probiotic L. crispatus CTV-05 have been demonstrated to have high mean adherence to vaginal epithelial cells in vitro,132 and have established vaginal colonisation in seven out of nine women when administered vaginally.133 Combined vaginal and rectal colonisation by H2O2-producing lactobacilli is associated with a four-fold decrease in the incidence of BV.132
Using molecular-based techniques in a double-blind, randomised, single-centre study of 90 non-pregnant, sexually active women, who were free from genital infection, and had a normal Nugent score, the ability of the probiotic L. crispatus CTV-05 to establish vaginal colonisation was studied.134 The study demonstrated that L. crispatus CTV-05 established vaginal colonisation at one or more follow-up visits in 69% of women overall, and in 90% of those not already colonised with L. crispatus. Of women who were never colonised with L. crispatus CTV-05, 85% were already colonised by endogenous L. crispatus at enrollment. The authors saw this as self-regulatory rather than a shortcoming of the probiotic, and the dosage regimen did not result in overgrowth of lactobacilli, as there seemed to be a physiological adjustment to around 106 or 108 cfu/ml of vaginal fluid. They also recommended that future research should concentrate on the lactobacilli that are prevalent in the vagina,134 rather than on species such as L. fermentum and L. rhamnosus.
Abnormal vaginal flora
Abnormal vaginal flora may occur because of a sexually transmitted infection (STI), e.g. trichomoniasis, colonisation by an organism that is not part of the normal vaginal community, e.g. Streptococcus pneumoniae, Haemophilus influenzae, or Listeria monocytogenes, or by overgrowth or increased virulence of an organism that is a constituent part of normal vaginal flora, e.g. Escherichia coli. Alterations in vaginal flora do not necessarily imply disease or result in symptoms. Disease results from the interplay between microbial virulence, numerical dominance, and the innate and adaptive immune response of the host.135 The most common disorder of vaginal flora is BV. BV is a polymicrobial condition, characterised by a decrease in the quality or quantity of lactobacilli, and by a 1000-fold increase in the number of other organisms determined by cultivation-dependent techniques, particularly anaerobes Mycoplasma hominis, G. vaginalis, and Mobiluncus species. The prevalence of BV in pregnancy in the USA is 1 080 000 cases annually.136 In pregnancy, BV has been associated with early and late miscarriage,11–13 recurrent abortion,12 post-abortal sepsis,10 postpartum endometritis,30 and preterm birth.12,13,21,25,137
Bacterial vaginosis and HIV
Bacterial vaginosis is also associated with the acquisition of HIV.5–9 Healthy lactic acid-producing vaginal flora acts as a barrier against the acquisition of HIV,138,139 and is negatively correlated with BV,80 which acts as a co-factor for HIV and conversion to seropositivity.5–9 HIV-infected women with BV have higher levels of HIV viral load in genital secretions than do HIV-infected women without BV.6,140,141 BV is also associated with an increased susceptibility to other STIs, including herpes simplex virus 2, gonorrhoea, Trichomonas vaginalis, and Chlamydia trachomatis.140,142–144 Molecular-based studies suggest a trend towards increased diversity in the microbiome of HIV-positive women with BV compared with those HIV-positive women who do not have BV, suggesting that HIV infection per se is associated with changes in the diversity of genital microbiota.145
Individual bacteria as a cause of bacterial vaginosis
Many of the models necessary to demonstrate that bacteria function as mono-etiological agents require a change in the micro-environment before an infectious event is observed. Infective peritonitis is easier to induce in experimental animals if blood is instilled into the peritoneal cavity, experimental gangrene develops better if calcium chloride is implanted into muscle along with Clostridium species, and rodents will not develop vaginal candidiasis without the addition of estrogen.42 In Gardner and Dukes’146 definitive paper, the instillation of pure cultures of Haemophilus vaginalis (now known as G. vaginalis) only successfully induced vaginal colonisation in one of 13 volunteers. In contrast, 11 of 15 volunteers were successfully inoculated when the vaginal secretions of donors (screened for other genital tract infections) were instilled into the vagina. This supports the later view of others, that since vaginal secretions from donors were much more successful at causing disease than pure cultures, G. vaginalis probably acts synergistically with other organisms to cause BV.147
Atopobium vaginae: under-detected and under-appreciated
The genus Atopobium lies within the family Coriobacteriaceae, and forms a distinct branch within the phylum Actinomycetes.148 Following sequence analysis, three species formally designated Lactobacillus minutus,149Lactobacillus rimae,150 and Streptococcus parvulus,151 within the lactic acid-producing group of bacteria,152 have been reclassified as the genus Atopobium. In 1999, an organism similar but not identical to these three species was isolated from the vagina of a healthy woman in Sweden, and the organism was named Atopobium vaginae.153 Since that time, using molecular-based techniques, A. vaginae has frequently been detected in the vagina, and is found much more commonly in women with BV than in those with normal flora.52–55,59,62,72,104,106–108,110,113,114,154–159 In addition to producing lactic acid,160,161 some species of Atopobium exhibit peptidyl peptidase activity, and produce significant quantities of ammonia in other environments where sugars are a scarce source of energy.162–164 This may be why A. vaginae is found more often in the vagina of postmenopausal women who are not on hormone replacement therapy (HRT) compared with those who are taking HRT.113Prevotella bivia (formerly Bacteroides bivia) also produces ammonia, which is known to act as a substrate to promote the growth of G. vaginalis.165Atopobium vaginae is strictly anaerobic, and is very sensitive to clindamycin in vitro,156 but is highly resistant to nitroimidazoles such as metronidazole156 and secnidazole.155
High diversity of flora in bacterial vaginosis compared with normal flora
Using various molecular-based techniques and usually the Amsel clinical criteria,166 or Nugent score,111 to classify normal or abnormal flora, a number of studies have demonstrated a high diversity of organisms in women with BV compared with women with normal flora. Collectively, these studies demonstrate the presence of novel bacterial species previously unidentified using cultivation-dependent techniques.51–56 They have also demonstrated that many of these organisms have specificity for BV, and that the number of phylotypes found in association with BV is statistically significantly greater than the number detected in the presence of intermediate flora (a distinct entity in its own right)167,168 or normal flora.49,52–54,56,65,106–108,110,156 This statistic largely results from the extreme dominance of lactobacilli in healthy women, which makes detection of other species unlikely, even when they are present at levels of 100 000 or more cells per sample.
Many of these organisms will be unfamiliar to clinicians (Box 2), although for many of them, there is evidence of disease association. The renamed Atopobium parvulum, Atopobium minitum, and Atopobium rimae have been associated with dental abscesses and oral infections,150,161,164,169,170 tubo-ovarian abscesses,171 and abdominal wound infection, supporting the view that these organisms may be pathogenic to the host. Leptotrichia sanguinegens/amnionii has been reported in association with postpartum endometritis, adnexal masses, and fetal death,172–174 and has been detected in the amniotic fluid of women with PTL, PPROM, and pre-eclampsia.175–177 Also, in a study of 45 women with salpingitis and 44 controls (women seeking tubal ligation), bacterial 16S rRNA sequences were found in the fallopian tube specimens of 24% of cases, and in no controls. Bacteria phylotypes closely related to Leptotrichia species and A. vaginae were among those identified in the cases.178 In addition, Dialister pneumosintes was found as the sole agent in the blood culture from a women with suppurative postpartum ovarian thrombosis.179
In summary, these studies have demonstrated that different subjects with BV have different microbial profiles, indicating heterogeneity in the composition of bacterial taxa in women with BV. Women without BV had bacterial communities dominated by Lactobacillus species, accounting for 86% of all sequences. In contrast, women with BV did not possess a single dominant phylotype, but instead had a diverse array of vaginal bacteria, often at relatively low abundances.
Bacterial vaginosis-associated bacteria BVAB1, BVAB2, and BVAB3 in the order Clostridiales
A bacterium distantly related to Eggerthella hongkongensis (92% sequence similarity).
Molecular-based tools for the diagnosis of bacterial vaginosis
Bacterial vaginosis can be diagnosed clinically, or using composite clinical criteria,166 microscopically,111,180–185 enzymatically,186–189 chromatographically,190,191 or using qualitative or semiquantative culture methods.19 Currently, the gold standard is the Nugent score,111 but the number of methods testifies to the fact that no single test is ideal, and that they can all provide false-positive and false-negative results. Findings from molecular-based studies are now highlighting possible explanations for why diagnosis by microscopy may be inconsistent, and why molecular methods may replace them.
One of the three organisms quantified as part of the Nugent score is Mobiluncus. Several cloning and sequencing studies have only rarely identified Mobiluncus.53,58,107 Fluorescence in situ hybridization (FISH) technology has demonstrated that BV-associated bacterium 1 (BVAB1) has a curved-rod morphology,53 similar to Mobiluncus morphotypes, and it is possible that with microscopic examination of vaginal smears, Mobiluncus species may have been over-represented and mistaken for BVAB1.62,192 Alternatively, as species-specific PCR agrees with the Nugent score, Mobiluncus may be missed in universal PCR studies because it frequently falls below a threshold titre where it can be detected.79
Zhou et al. observed that the urea produced by Atopobium species was associated with halitosis,163 and that similar species of Megasphaera caused beer spoilage by turbidity, off-flavours and off-colours.193,194 They concluded that if two genera associated with malodorous metabolites can be found in the vagina of healthy women, and amines can be found in women without BV, then diagnostic techniques to diagnose BV, based on upon amine production and odour formation,166,188,191 may need to be amended.62
Microscopically, Atopobium species are gram positive, elliptical cocci, or rod-shaped organisms that occur, singly, in pairs, or short chains. The variable cell morphology of Atopobium renders it well camouflaged among the mixture of other species present in bacterial communities where the Nugent score is ≥4. Atopobium vaginae is fastidious, grows anaerobically, and forms small pin-head colonies on cultures that are easily missed.153 Although phylogenetically different from other lactic acid-producing bacteria, they are not phenotypically exceptional, and it is not difficult to see why the significance of this organism based on culture, microscopy, and phenotype may be overlooked and under-appreciated.
Using species-specific primers, the relationship between five fastidious organisms associated with BV were compared with BV diagnosed by Amsel and/or Nugent scores, and also with the individual clinical criteria of Amsel.195 The two biovars of Ureaplasma urealyticum (Ureaplasma parvum and Ureaplasma urealyticum—biovar 2) were associated with vaginal discharge and raised pH, but not with BV by either Amsel or Nugent criteria, or any of the individual Amsel clinical criteria. In contrast, with Leptotrichia sanguinegens/amnionii, A. vaginae, and BVAB1, an elevated pH >4.5 was a universal feature, and they were all associated with BV by both Amsel and Nugent criteria, and with the finding of >20% of epithelial cells as clue cells, a feature that has already been reported.53 A positive test for amine odour upon the addition of 10% solution of potassium hydroxide was significantly more likely in women testing positive for BVAB1. Douching is a recognised risk factor for BV,196 and the detection of Leptotrichia and A. vaginae was three times more likely, and BVAB1 twice as likely, when women reported douching.195
Fredricks et al.,53,72 in two linked studies, noted that some organisms or combination of organisms had high sensitivities or specificities for the diagnosis of BV using the Amsel criteria and the Nugent score (Table 1). Using quantitative real-time PCR, Menard et al. first examined qualitatively the association of individual organisms with BV diagnosed by Nugent score. To optimise the molecular methods for routine practice, an adjusted quantification was made creating a threshold of DNA copies/ml for lactobacilli and several organisms known to be associated with BV (Table 2). At a threshold of ≥108 DNA copies/ml, Lactobacillus species was predictive of normal flora (sensitivity 44%; specificity 100%). BVAB1, BVAB2, and BVAB3 alone, or in combination, had high specificity for BV diagnosed by Amsel criteria.
Table 1. The sensitivities and specificities for individual or combinations of organisms for the diagnosis of bacterial vaginosis using the Amsel criteria or the Nugent score
Table 2. Quantification of vaginal organisms for the production of bacterial vaginosis using the Nugent score158
Threshold quantification (DNA copies/ml)
AUC, area under the curve (the closer the AUC comes to 1.0, the better the bacterial count predicts BV); NPV, negative predictive value; PPV, positive predictive value; ROC, receiver operating characteristic.
The combination of A. vaginae and G. vaginalis for the diagnosis of bacterial vaginosis
As A. vaginae and G. vaginalis are frequently detected in association with BV, a number of authors using molecular-based techniques have examined the possibility of combining these two organisms as a means of diagnosing BV.52,53,59,113,154,158,159 Using DNA quantitation, 19 out of 20 BV samples had either a DNA level for A. vaginae≥108 copies/ml or G. vaginalis≥109 copies/ml, and nine out of 20 had both. The combination of an A. vaginae DNA level ≥108 copies/ml and a G. vaginalis DNA level ≥109 copies/ml demonstrated the best predictive criteria for the diagnosis of BV with excellent sensitivity (95%), specificity (99%), negative predictive value (NPV) (99%), and positive predictive value (PPV) (95%).158 When the quality and reproducibility of this combination was applied prospectively for validation of the Nugent score in 56 pregnant women, the NPV was 96% and the PPV was 99%.158
Culture-independent techniques to assess the treatment of recurrent, persistent, or resistant bacterial vaginosis
Cure of BV or improvement in symptoms following recommended treatments with metronidazole or clindamycin136 reaches 83–94% by 7–21 days of treatment.197,198 Whereas the short-term treatment response is acceptable, BV persists or recurs in 11–29% of women at 1 month,197,199,200 30% of patients relapse within 3 months, and recurrence rates may be more than 50% within a year.154,201–203 Only 48% of women will be colonised by H2O2-producing lactobacilli 70–90 days after treatment with either clindamycin or metronidazole.200,204
A number of molecular-based studies53,106,154,205,206 have addressed the problems of recurrent or persistent BV over time, and why some women with BV may be resistant to cure. A consistent finding in these studies, as women lapse in and out of being BV-positive and BV-negative, is the stability of vaginal flora requiring few phylotypes, and Lactobacillus dominance with L. crispatus and/or L. jensenii in women who remain or become BV-negative. In contrast, those who become-BV positive are commonly colonised by L. iners, with many other non-Lactobacillus species being present.53,106 Three studies have addressed the problem of persistent or recurrent BV using culture-independent techniques.154,205,206 Women with BV were appropriately treated and followed-up in a large, prospective, cross-sectional cohort study over a 12-month period. Atopobium vaginae was detected in 75% and G. vaginalis was detected in 100% of women with recurrent BV. Women in whom both organisms were detected had higher rates of recurrence of BV (83%) compared with women who had G. vaginalis alone (38%) (P < 0.001). Of relevance is the fact that >90% of a biofilm indentified on vaginal epithelial cells of women with BV was composed of A. vaginae and G. vaginalis.207,208 The biofilm may have interfered with treatment, but the authors were unable to determine whether the recurrence was the result of inadequate treatment and undetectably low levels of residual organisms after treatment (i.e. relapse), reinfection from sexual partners, which they thought was feasible, or disruption of normal flora from other exogenous factors.154
In an isolated study, several bacterial species were detected, none of which are commonly associated with BV. Organisms normally associated with BV were not detected. This atypical form of BV might explain why conventional therapy did not work, and may indicate that other antimicrobial therapies may have been more effective.205 Finally, using species-specific 16S rDNA, PCR assays targeting 17 different BV-associated bacteria in 131 women with BV, the vaginal microbiome was sampled before treatment and 1 month after for a test of the cure. Treatment was with a 5-day course of intravaginal metronidazole gel. At 1 month after treatment, BV was still present in 26% of women. The baseline detection of BVAB1, BVAB2, and BVAB3, Peptoniphilus lacrimalis, or Megasphaera phylotype was significantly associated with persistence of BV at the test of the cure. The authors concluded that pretreatment vaginal microbiology at diagnosis might define the risk of antibiotic failure.206 We anticipate that these correlations will become clearer and more meaningful when studies are repeated with quantitative PCR, instead of simply using detection/incidence.
Culture-independent studies in pregnancy
Culture-independent techniques have been used to measure prevalence, diversity, and abundance of organisms, particularly ureaplasmas in amniotic fluid,209 in association with suspected cervical insufficiency,210 preterm labour,175,211 PPROM,177,212 small-for-gestational-age babies,213 pre-eclampsia,176 and the potential for bacteria from the oral cavity to colonise amniotic fluid.214 However, apart from combining pregnant women with non-pregnant women to swell sample numbers,52,159 the information with respect to the vaginal microbiome in pregnant women is limited, particularly with respect to the outcome of pregnancy, especially preterm birth.
Using species-specific primers, Wilks et al.102 quantified the production of H2O2 by lactobacilli from swabs taken at 20 weeks of gestation from the vagina of 73 women considered to be at high risk of preterm birth, according to the Creasy criteria.215 The levels of H2O2 production varied between species of Lactobacillus. The presence of lactobacilli producing high levels of H2O2 was associated with a reduced incidence of BV at 20 weeks of gestation, and subsequent chorioamnionitis. The authors postulated that H2O2-producing lactobacilli reduced the incidence of ascending genital tract colonisation in pregnancy, which leads to infection and preterm birth.102 Unfortunately, the role of L. iners was not tested, because MRS agar culture medium was used in the study.91
In pregnant Japanese women, Tamrakar et al.117 largely confirmed the findings of Fredricks et al.53 in non-pregnant women. The prevalence of L. crispatus, L. jensenii, and L. gasseri was significantly higher, whereas that of BVAB2, Megasphaera, Leptotrichia, and Eggerthella-like bacterium were significantly lower in women with a normal Nugent score, compared with those who had BV. The prevalence of L. iners did not differ between these groups, and women with L. iners were more likely to be colonised by BVAB2, Megasphaera, Leptotrichia, and Eggerthella-like bacterium.117
In a longitudinal study of 100 pregnant women, vaginal swabs were obtained at mean gestational ages of 8.6, 21.2, and 32.4 weeks, respectively.216 In the first trimester 77 women had normal or Lactobacillus-dominated flora; 13 developed abnormal flora in the second or third trimester. When first-trimester normal flora was dominated by L. gasseri or L. iners, there was a ten-fold risk of conversion to abnormal flora. In contrast, normal flora comprising L. crispatus had a five-fold decreased risk of conversion to abnormal flora.216 This may be because L. gasseri and L. iners only produce H2O2 in a small percentage of strains.96,132
Molecular-based techniques provide important new information about vaginal microbial flora, and permit the identification of previously under-detected and hence under-appreciated organisms, such as L. iners and A. vaginae. Molecular-based techniques are not without their problems, and will not replace culture-based techniques, but, when used in combination, add greatly to our understanding of vaginal flora. In the majority of circumstances, normal vaginal flora is dominated by Lactobacillus species. In the absence of lactobacilli, normality can be maintained by other, more fastidious lactic acid-producing bacteria. In keeping with the theories of ‘competitive exclusion’ and ‘bacterial interference’, when lactobacilli dominate the vaginal flora, culture-independent techniques have demonstrated that the healthy vagina is usually colonised by only one or two dominant Lactobacillus species, mainly from L. crispatus, L. iners, L. jensenii, and L. gasseri. The dominant Lactobacillus species may differ racially or geographically, but the principle of numerical dominance persists, and may be an important defense mechanism. Without molecular-based techniques, phenotypic identification of lactobacilli is difficult, and so is normally only carried out to the genus level. The ability to identify lactobacilli to species level should enable us to gain a better understanding of the role of different species of Lactobacillus, particularly with respect to their ability to produce H2O2 and to function as a probiotic.
Molecular-based techniques indicate that there is a far greater diversity of micro-organisms associated with BV than has been evident from cultivation-dependant techniques. These diverse organisms accumulate to form different communities or profiles, which make it likely that BV is not a single entity, but a syndrome of variable composition that causes a variety of symptoms, different phenotypical outcomes, and may result in variable responses to different antibiotic regimens. Some organisms or combinations of organisms have a high specificity for BV, such that in the future, by using molecular quantification, we may better diagnose each subtype of BV, and be able to tailor treatment appropriately. The information with respect to the vaginal microbiome in pregnant women is limited, particularly with respect to preterm birth. This gap is striking, given the importance of the preterm birth problem worldwide, and must become a research and funding priority. By better understanding the vaginal microbiome during pregnancy, we may be able to predict and prevent some of the great obstetric syndromes like PPROM, preterm labor, and preterm birth, which is associated with infection and significant infant mortality and morbidity. This better understanding is imminent, with the onset of studies using much more powerful sequencing technologies.
Disclosure of interests
Contribution to authorship
All authors have contributed to the preparation and/or editing of the manuscript.
Details of ethics approval
Supported (in part) by the Intramural Research Program of the Eunice Kennedy Shriver National Institute of Child Health and Human Development, NIH, DHHS.