|Home | About | Journals | Submit | Contact Us | Français|
Knowledge of the abundance of bacterial species in vaginal communities will help us to better understand their role in health and disease. However, progress in this field has been limited because quantifying bacteria in natural specimens is an arduous process. We developed quantitative real-time PCR (qPCR) assays to facilitate assessments of bacterial abundance in vaginal specimens and evaluated the utility of these assays by measuring species abundance in patients whose vaginal floras were clinically described as normal, intermediate, or bacterial vaginosis (BV) as defined by Nugent's criteria. The qPCR measurements showed that Lactobacillus species were predominant in normal vaginal specimens and that high Lactobacillus crispatus and Lactobacillus jensenii abundance was specific to normal specimens, while Lactobacillus iners abundance was high in all categories including BV. The abundances of all non-Lactobacillus species were higher in BV specimens than in normal specimens. Prevotella species were prevalent in all specimens and represented a high percentage of total species in BV specimens. qPCR assays can be a useful tool for describing the structure of vaginal communities and elucidating their role in health and disease.
Vaginal bacterial communities are composed of mixtures of diverse species, and the relative abundance of these species in part determines urogenital health and disease in women (22). It is generally acknowledged that vaginal communities predominated by Lactobacillus species are normal and healthy while communities predominated by other genera, such as Gardnerella vaginalis, are abnormal and unhealthy (36). The latter condition essentially defines a poorly understood syndrome known as bacterial vaginosis (BV). While BV can be asymptomatic and benign in some women, it is a common cause of malodorous vaginal discharge for many. Moreover, BV flora is of concern because it is associated with an increased risk of adverse sequelae, such as preterm birth (8, 24), postoperative complications in women (40), enhanced risk of acquiring sexually transmitted infections (31), and increased shedding of HIV (11). Treating BV has not proven effective for the prevention of these adverse events possibly due to the fact that standard BV treatment results in high failure and relapse rates (25, 29). Furthermore, while suspected pathogens such as G. vaginalis have been implicated, no agent or factor has been identified as the cause of BV, despite experimental (10) and epidemiological (28) evidence that suggests that BV is transmissible (10). Because of all the uncertainties surrounding this syndrome, BV has been described as a microbiological and clinical enigma (16, 17).
Failure to understand the microbiology specific to BV is perhaps not surprising given that the basic ecology of the genitourinary microbiota, namely, the composition, relative abundance, and temporal fluctuations of vaginal species, are poorly understood. This lack of knowledge is highlighted by recent cultivation-independent broad-range PCR surveys, which show that there are scores of species and genera common to the vaginal environment that had not been recognized, and many of these represent new, as-yet-uncultivated species (18, 22, 33, 43, 46, 47). Broad-range PCR surveys have also revealed patterns in species composition across vaginal specimens which support the concept of defining multiple types of normal and BV-like communities, as some researchers have proposed (12, 44, 47). It is possible that one type of vaginal community might be associated with cases of BV that appear to arise endogenously (6, 34), while a different type might be associated with cases of BV that appear to arise via sexual transmission (10, 28). Broad-range PCR surveys (14, 18, 33, 38) have also shown that the presence of newly discovered vaginal species, such as Atopobium vaginae, and unknown Clostridiales and Megasphaera spp. might be diagnostic of BV (18). Subsequent studies, using PCR assays targeting individual vaginal species, have demonstrated that the mere presence of well-known BV-associated species, such as G. vaginalis, is relatively nonspecific for diagnosis of BV (19). Other research has shown that species-specific quantitative real-time PCR (qPCR) assays that incorporate species concentration cutoff values can greatly improve the performance of PCR assays for diagnoses of BV (30). qPCR assays have shown that the concentrations of bacteria such as Prevotella spp., Atopobium spp., and G. vaginalis, among others, are elevated in BV while the concentrations of Lactobacillus spp. are reduced (5, 20, 45). Such assessments of species abundance may be essential to describing the ecology of microbial communities (26) and understanding their roles in health and disease.
We developed qPCR assays to assess the abundance of 19 vaginal species that we and others commonly detect in PCR surveys of vaginal flora, eight of which have not yet been cultured. We used these assays along with qPCR assays for “total” bacteria and “total” Prevotella spp. to measure the relative abundance of bacteria in vaginal specimens from 37 patients clinically diagnosed as having either normal, intermediate, or BV floras.
Vaginal swab samples for this study were collected over the course of several years (2002 to 2004) in the New Orleans sexually transmitted diseases (STD) clinic following a Louisiana State University Health Sciences Center international review board (IRB)-approved protocol. The primary purpose of the protocol was to investigate microbiologic associations with endocervicitis among young women without a recent history of antibiotic use presenting for routine STD assessment. The data presented here are based on a subsample of the 400 women enrolled in this study. The women were carefully assessed for sexual risk behaviors by history and physical findings by speculum pelvic examination. Amsel's criteria (2) and Nugent scores (32) were determined for each patient specimen. Briefly, Amsel's criteria assess vaginal fluid pH, consistency of vaginal discharge, amine odor upon addition of KOH to vaginal fluid, and the presence of bacteria adhering to epithelial cells (clue cells). The presence of at least three of these criteria supports the diagnosis of BV. Nugent scores are derived from a microscopic analysis of vaginal swab smears. This method assesses the abundance and morphology of vaginal bacteria in a standardized fashion. A score of 0 to 3 is considered normal, 4 to 6 is intermediate, and 7 to 10 is BV. Vaginal smears were read by a technician with 15 years of experience assigning Nugent scores. We initially selected vaginal samples from six women with normal vaginal floras who had Nugent scores of 0 and exhibited none of the Amsel's criteria and from nine women with BV who had Nugent scores of 10 and exhibited all four Amsel's criteria. We believed that the development of qPCR assays would be facilitated by first working with specimens whose microflora would be most likely to differ significantly. We subsequently added 22 specimens from women with Nugent scores ranging from 2 to 9. Amsel's criteria were not considered in selecting these cases. All samples were from women who had negative vaginal yeast cultures. The median age of these patients was 26 years. Ninety-five percent were African American. Seventy-three percent complained of a vaginal discharge and 21% of vaginal itching. A history of douching was present in 54%, and 60% gave a history of receiving oral sex. This was a group of women at high risk for STDs, and their past histories of infection were as follows: trichomoniasis, 32%; gonorrhea, 32%; and chlamydial infection, 40%. None had a history of HIV infection, but 43% had had BV in the past.
Vaginal DNA was extracted from swab specimens by using a High Pure PCR template preparation kit (Roche Molecular Diagnostics, Penzberg, Germany). Plasmids containing 16S rRNA genes of vaginal bacterial species were obtained from Escherichia coli clone libraries generated in previous studies of vaginal communities (15). Known concentrations of plasmids were used as template 16S rRNA genes to measure the detection limit of the qPCR assays and to generate standard curves for quantifying assay results. The species names and GenBank reference numbers corresponding to these plasmids are listed in Table Table1.1. Plasmid DNA was extracted using a QIAprep Spin miniprep kit (Qiagen, Valencia, CA). Plasmid DNA and DNA extracted from vaginal specimens were quantified using a TBS-380 fluorometer (Turner Biosystems, Sunnyvale, CA) with Quant-iT PicoGreen double-stranded DNA (dsDNA) reagent (Invitrogen Inc., Carlsbad, CA). Plasmid copy numbers were determined with the aid of an online copy number calculator (www.uri.edu/research/gsc/resources/cndna.html). Standard curves were generated using serial 10-fold dilutions (100 to 107 copies) of plasmids. The range of slopes for the qPCR assays was from −3.1 to −3.6 (85 to 110% efficiency), and linearity (r2) values were all ≥0.99.
All qPCR assays were SYBR green-based and were performed on an iCycler iQ real-time detection system (Bio-Rad, Hercules, CA). The reaction mixture (50 μl) contained 1× iQ-SYBR green PCR supermix (Bio-Rad), 0.5 μM final concentration of forward and reverse primers, and 1 μl template DNA (10 ng). Temperature cycling for all assays was 95°C for 2 min, followed by 40 cycles at 95°C for 1 min, 62 to 65°C for 1 min (Table (Table1),1), and 72°C for 1 min. Fluorescence was measured at the final step of each cycle. Following amplification, melting curve data were obtained by slow heating at 0.5°C increments from 55 to 95°C, temperature was held for 10 s at each step, and fluorescence was acquired at each temperature increment. The PCR amplificates were visualized in 1.4% agarose gels stained with ethidium bromide. Melt curves and agarose gel analysis were used to assess the formation of nontarget PCR products. For each assay, vaginal specimens and plasmid standards were run in duplicate, and the average values were used to calculate bacterial concentration. Negative (no DNA) controls were run with every assay to check for contamination. Assay results were expressed as 16S rRNA gene copies per 10 ng of vaginal DNA.
PCR primers targeting the 16S rRNA genes of 19 vaginal bacterial species were designed using the computer program PRIMROSE (3) and Integrated DNA Technologies PrimerQuest software. The specificity of each primer was checked using BLASTn analysis (1). “Species” were arbitrarily defined as groups of 16S rRNA genes that share ≥98% similarity (39). Nineteen species-specific qPCR assays were developed (Table (Table1).1). We also used PRIMROSE to design a PCR primer set to measure the abundance of a broad range of potentially important vaginal Prevotella species recently shown to be abundant in vaginal specimens (33). GenBank accession numbers for the sequences used to develop the Prevotella primers are AY958798, AY958868, AY958879, AY959091, AY959133, AY959135, AY959212, L16475, L16476, L16483, AF414821, AB108826, and AJ011683. Probe match analysis at the RDP web site (9) indicated that the primer set complements these recently described vaginal Prevotella (33) and other Prevotella sequences originating from studies of vaginal specimens. Probe match analysis also indicated that the forward primer complemented 3,107 (62%) and the reverse primer 4,602 (93%) of the 4,948 full-length 16S rRNA sequences listed in the genus Prevotella.
The primer set complements 12 16S rRNA gene sequences outside the family Prevotellaceae, five in the phylum Firmicutes, and seven in the phylum Bacteroidetes. GenBank references indicated these are uncultivated bacteria. Probe match analysis also showed that the Prevotella primer set complements 41 sequences in the family Prevotellaceae that are not grouped within the genus Prevotella; five are in the genus Hallella, and 34 are in a group labeled “unclassified” Prevotellaceae. All but one of these 41 Prevotellaceae sequences represented uncultivated bacteria, and none originated from analyses of vaginal floras. A PCR primer set (4) targeting a broad range of bacterial phyla was used as a positive control to test for PCR inhibition by possible contaminants in vaginal DNA extracts and as a means of assessing the relative abundance of “total” bacteria in vaginal specimens.
To measure assay specificity, each qPCR primer pair was checked for cross-reactivity against each individual nontarget plasmid listed in Table Table1.1. Each cross-reactivity test was performed using 107 copies of nontarget plasmid as template in the PCR in the absence of target DNA. Under these conditions, we noted that there was a weak fluorescence signal in some assays at high threshold cycle (Ct) values. These high Ct values corresponded to those we observe when the assays are run using very small amounts (1 to 10 copies) of target plasmid 16S rRNA genes in the qPCR. The Leptotrichia amnionii qPCR assay exhibited the lowest specificity in our cross-reactivity tests; in this case, 107 copies of nontarget but closely related Sneathia sanguinegens 16S rRNA genes produced a weak fluorescence signal with a Ct value equivalent to those of ~10 copies of target L. amnionii 16S rRNA genes. We used the results of the cross-reactivity tests to define the specificity of each qPCR assay as the minimum number of target 16S rRNA gene sequences that can be detected without cross-reactivity with 107 copies of nontarget 16S rRNA genes template. We set this minimum number 5- to 10-fold higher than the concentration at which nonspecific product formation was observed in the cross-reactivity tests. In the case of the L. amnionii example mentioned above, cross-reactivity was observed at a Ct value equivalent to those of ~10 copies of target 16S rRNA genes, thus the minimum number of target 16S rRNA gene molecules that can be detected without cross-reactivity with 107 nontarget 16S rRNA genes molecules was set at ≥100 (Table (Table1).1). To assess whether the assays exhibited cross-reactivity in situ, we sequenced ≥75% of the qPCR amplificates from vaginal specimens for each assay. Emphasis was placed on sequencing amplificates detected in low abundance since cross-reactivity tests suggested that nonspecific amplification is more probable in these cases. Sequencing was performed by a commercial vendor (Davis Sequencing, Inc., Davis, CA). The sequences of all PCR products were ≥99% similar to those of the targeted species.
Graphs were prepared, Mann-Whitney U tests and Fischer's exact tests were performed, and P values and odds ratios were obtained using Prism V4.0c for Macintosh (GraphPad Software Inc., San Diego, CA).
The qPCR results indicated that Lactobacillus species were predominant in all but 1 of the 13 patients with normal specimens, while non-Lactobacillus species were predominant in all but one of the 16 BV patients (Table (Table2).2). Among the six patients with Nugent scores of 0, lactobacilli represented ≥98% of the species assayed in five and 89% in the sixth (Table (Table2).2). Lactobacillus iners was the predominant species in eight patients with normal specimens, while Lactobacillus crispatus and Lactobacillus jensenii were predominant in three and one patients with normal specimens, respectively (Table (Table2).2). All three L. crispatus dominated specimens had Nugent scores of 0. L. crispatus and L. jensenii were only present in high concentration (>105) in normal specimens, and were never in high concentration in intermediate or BV specimens (Fig. (Fig.1).1). In contrast, L. iners concentrations were high (>105) in all patient categories (Fig. (Fig.1).1). Moreover, L. iners was the only species detected in every specimen (Table (Table3).3). Lactobacillus gasseri was the least prevalent Lactobacillus species surveyed, detected in only two specimens, both normal, and was not predominant in any specimen (Table (Table22).
The qPCR results showed that the median concentrations of all non-Lactobacillus species, as well as total bacteria and the total Prevotella spp., were all more than 10-fold higher in BV specimens than in normal specimens (Fig. (Fig.1),1), and with the exception of Mobiluncus curtisii, the differences in concentration were significant (P < 0.05). The median concentrations of non-Lactobacillus species were also higher in intermediate specimens than in normal specimens, except for Megasphaera type 2, Mycoplasma hominis, Mobiluncus mulieris, and M. curtisii (Fig. (Fig.1).1). The prevalence of non-Lactobacillus species was also higher in BV specimens (Table (Table3),3), as was the average number of non-Lactobacillus species detected per specimen—15.5 for BV, 9.5 for normal, and 12.0 for intermediate specimens. Of note was the fact that the average number of species detected was lower (7.2) in specimens with the lowest possible Nugent score (0) than in normal specimens with Nugent scores of 1 to 3 (11.4), (P = 0.044, 95% CI for difference: −8.4 to −0.14). Among the eight intermediate specimens (Nugent scores of 4 to 6), L. iners was present in all cases and predominated (73% to 95%) in six (Table (Table2).2). G. vaginalis was also present in all eight intermediate specimens and was the second most abundant species in these specimens (Table (Table2).2). The abundance of other organisms in intermediate cases is summarized in Table Table2.2. Among BV specimens, BVAB1 (bacterial vaginosis-associated bacteria 1), an uncultivated bacterium in the order Clostridiales, had the highest median concentration of any individual species (Fig. (Fig.1)1) and was the most abundant species detected in nine of the 16 BV specimens. Seven of the specimens predominated by BVAB1 had Nugent scores of 10 (Table (Table2).2). G. vaginalis and Megasphaera type 1 had the second and third highest median concentrations in BV specimens, respectively (Fig. (Fig.1),1), followed by Prevotella buccalis and Atopobium vaginae, and all were detected in every BV specimen (Table (Table3).3). BVAB2 (also in the order Clostridiales), Peptostreptococcus sp., Eggerthella sp., and Leptotrichia amnionii were detected in almost all BV specimens (Table (Table3)3) but at low concentrations relative to the other non-Lactobacillus species assayed (Table (Table2).2). Sneathia sanguinegens was present in low concentrations in all of the BV cases. While the individual species P. buccalis was less abundant than other non-Lactobacillus species in BV specimens (Table (Table2),2), as a group, members of the genus Prevotella had a higher median concentration (Fig. (Fig.1),1), and their percent abundance in BV patients was higher than all of the individual assayed species combined (Table (Table22).
Although non-Lactobacillus species were in relatively low abundance in normal specimens (Fig. (Fig.1),1), they were detected in a high percentage of these specimens (Table (Table3).3). G. vaginalis was detected in more normal specimens (85%) than any other non-Lactobacillus species (Table (Table3).3). G. vaginalis also had the highest median concentration of any non-Lactobacillus species in normal specimens (Fig. (Fig.1).1). A. vaginae, Eggerthella sp., BVAB1, P. buccalis, and Megasphaera type 1 were all detected in over 50% of normal specimens (Table (Table3).3). In fact, there was no significant association between the presence of G. vaginalis, BVAB1, and Megasphaera type 2 and clinical diagnosis of BV (Table (Table4).4). However, the concentrations of these three species were significantly higher in BV specimens (P < 0.005). Thus, concentration values above a minimum “threshold” level could be used to significantly increase discrimination between normal and BV patients (Table (Table4).4). Moreover, the use of non-Lactobacillus species threshold values improved the ability to discriminate BV from normal specimens with the exception of Mobiluncus curtisii (Table (Table44).
Since all methods to detect and identify bacteria in natural environments have inherent biases, similar patterns in bacterial species composition observed by independent investigators by use of different analysis methods lend credence to inferences about microbial community structure. The limitations of using qPCR to characterize bacterial communities in situ have been well described (13, 27, 37). The most obvious of these limitations is that all nontargeted species are completely ignored. Nevertheless, comparisons between assessments of vaginal communities provided by different methods can be made. In general, our qPCR results indicate that vaginal microbial communities of clinically healthy patients are predominated by Lactobacillus spp. while those of BV patients are predominated by non-Lactobacillus spp. This general description is congruent with those of other qPCR (5, 20, 30, 45) and broad-range PCR assessments of vaginal communities (22, 33, 38).
In clinically normal vaginal flora, we found that a relatively recently recognized vaginal Lactobacillus species, L. iners, was predominant in most cases (8 of 13). L. iners was also the predominant species in four of eight women with intermediate Nugent scores. Moreover, L. iners was the only species among those we studied that was detected in all (n = 37) patients (Table (Table2).2). This observation is consistent with recent studies that, taken together, indicate that L. iners is possibly the most prevalent and abundant Lactobacillus species in the vagina (7, 22, 41, 42, 47). In contrast to L. crispatus and L. jensenii, L. iners is common and abundant in patients whose vaginal community includes high concentrations of non-Lactobacillus species, such as in BV (44). Other studies of perturbed vaginal flora suggest that L. iners might be a transitional species, colonizing after disturbances to the vaginal environment (15, 23). Though the role of L. iners in vaginal bacterial community structure is not clear at this point, it no doubt plays a major role since it is one of the most common single species in vaginal specimens regardless of Nugent score or the presence of Amsel's criteria for BV.
Our use of qPCR to assess the quantitative relationships of 19 common vaginal microbial species resulted in several potentially important observations concerning the role of anaerobic Gram-negative rods in vaginal communities. In our patient population, Prevotella spp. were present in every case regardless of the Nugent score. Moreover, among the 16 women with Nugent scores of 7 to 10, Prevotella spp. as a group dominated the flora in every case. In a culture-based survey of vaginal flora in 171 pregnant women, Hillier et al. (21) found Prevotella spp. in the majority of all patients, although women with Nugent scores consistent with BV were more likely to harbor this organism and to be colonized by higher concentrations. In a study using the broad-range 16S rRNA gene amplification, cloning, and sequencing approach, Oakley et al. (33) recently found that Prevotella sequences were the most common clones in vaginal libraries of 21 specimens from women clinically defined as having BV. Moreover, these investigators identified 21 different operational taxonomic units (OTUs) (species) in the genus Prevotella in both their study and a review of vaginal sequences in the NCBI and RDP databases. Additionally, Spear et al. (38) used pyrosequencing to describe the vaginal communities of women with and without BV as defined by Nugent score and found Prevotella spp. to be one of the most common OTUs present. It seems likely that vagina-specific Prevotella spp. collectively may play a key role in vaginal communities.
We observed nine BV cases in which BVAB1 was the most abundant individual species representing from 44 to 90% of the sequences detected by qPCR. Six of these were among the group of nine cases selected for inclusion in this study based on Nugent scores of 10 and presence of all four of the Amsel's criteria; thus, these bacterial communities are overrepresented in this sample of women with BV. These cases could represent a BV community subset with different pathogenic consequences for the host. Similar BVAB1 predominant communities were evident in broad-range PCR-based clone library studies of vaginal flora (18, 22). Among the remaining BV specimens, L. iners was predominant in one, A. vaginae in two, and G. vaginalis in four. An important unknown is how stable these vaginal microbial communities are over time. It could well be that BVAB1 prominence in the nine communities noted above is only a transient phenomenon and that there are no consistent differences in BV communities in our population, at least in so far as the major components as measured in our study are concerned.
Sha et al. (35) described the use of qPCR for A. vaginae and M. hominis to identify a group of HIV-infected women who were more likely to shed HIV in their vaginal secretions. Menard et al. (30) recently described the use of qPCR to define quantitative threshold values for A. vaginae and G. vaginalis, which were then combined as a diagnostic tool for BV that was much more specific than merely finding the presence of these organisms. Our data are consistent with theirs in showing that A. vaginae and G. vaginalis are present at higher concentrations in BV patients and that these organisms were present in a large proportion of patients that had normal Nugent scores. We extended their observations by defining BV both by the Nugent and the Amsel criteria and performing separate analyses for both definitions (Table (Table4).4). The results were essentially the same, regardless of which BV definition was used. Our results also suggest that there are likely to be other combinations of organisms whose relative concentrations could serve as an indicator for diagnosis of BV. In fact, quantitation of total bacterial 16S rRNA genes could probably be used as an indicator of BV. However, the present study was not designed to answer such questions. Studies using larger unselected patient samples will be needed for this purpose.
There are a few limitations of this study. First, the number of samples is relatively small, which limits statistical analysis. Second, the samples studied were not randomly chosen but were selected based on Nugent scores and Amsel's criteria, with an overweighting of very low (0) and very high (10) Nugent scores. Therefore, these data cannot be interpreted as being representative of the distribution of various vaginal bacterial communities in a general population of women. Our findings need to be verified by a larger study of women randomly selected from representative populations of women.
In summary, in our population of predominately African-American women at high risk for STIs, the most prevalent vaginal species was L. iners, and in 14 of 21 specimens with Nugent scores of <7, it was the predominant species. By using primers that amplify a broad range of species in the genus Prevotella, we found that these organisms were abundant in the flora of all women with BV (Nugent scores of ≥7) and all patients harbored at least low levels of bacteria belonging to this genus. Among BV patients, especially those with Nugent scores of 10, a significant subset had relatively high concentrations of BVAB1. The significance of these findings will require longitudinal studies to ascertain the stability of vaginal flora and, eventually, studies to determine which BV communities impose significant health risk to their human hosts. We conclude that quantitative molecular methods are necessary to fully characterize human vaginal microbial communities.
This study was supported by an LSUHSC Translational Research Award and by The Research Institute for Children, New Orleans, LA.
Published ahead of print on 19 March 2010.