Unique Vaginal Microbiota That Includes an Unknown Mycoplasma-Like Organism Is Associated With Trichomonas vaginalis Infection

Abstract

Background. The prevalence of Trichomonas vaginalis infection is highest in women with intermediate Nugent scores. We hypothesized that the vaginal microbiota in T. vaginalis–infected women differs from that in T. vaginalis–uninfected women.

Methods. Vaginal samples from 30 T. vaginalis–infected women were matched by Nugent score to those from 30 T. vaginalis–uninfected women. Equal numbers of women with Nugent scores categorized as normal, intermediate, and bacterial vaginosis were included. The vaginal microbiota was assessed using 454 pyrosequencing analysis of polymerase chain reaction–amplified 16S ribosomal RNA gene sequences. The 16S ribosomal RNA gene sequence of an unknown organism was obtained by universal bacterial polymerase chain reaction amplification, cloning, and sequencing.

Results. Principal coordinates analysis of the pyrosequencing data showed divergence of the vaginal microbiota in T. vaginalis–infected and T. vaginalis–uninfected patients among women with normal and those with intermediate Nugent scores but not among women with bacterial vaginosis. Cluster analysis revealed 2 unique groups of T. vaginalis–infected women. One had high abundance of Mycoplasma hominis and other had high abundance of an unknown Mycoplasma species. Women in the former group had clinical evidence of enhanced vaginal inflammation.

Conclusions. T. vaginalis may alter the vaginal microbiota in a manner that is favorable to its survival and/or transmissibility. An unknown Mycoplasma species plays a role in some of these transformations. In other cases, these changes may result in a heightened host inflammatory response.

Increasingly, Trichomonas vaginalis has become recognized as a sexually transmitted infection (STI) with a number of consequences other than symptomatic vaginal discharge. These include adverse pregnancy outcomes [1, 2], pelvic inflammatory disease [3, 4], and, most importantly, acquisition of human immunodeficiency virus (HIV) infection [5–7]. Recent prospective studies have shown that antecedent bacterial vaginosis, as defined by the Nugent score, is a risk factor for the acquisition of T. vaginalis [8, 9]. Further evidence of the interaction of T. vaginalis and the vaginal microbiota comes from a recent study, which showed that the standard 2-g single-dose metronidazole treatment for T. vaginalis infection failed more frequently in HIV-infected women with bacterial vaginosis than in HIV-infected women without bacterial vaginosis [10]. A proposed explanation for this finding was that a vaginal microbiota in which bacterial vaginosis was determined to be present on the basis of Nugent score provided a protective environment for T. vaginalis, perhaps by partial inactivation of the drug. Although it is often stated in the literature that T. vaginalis infection is associated with abnormal vaginal microbiota, the precise nature of this relationship has seldom been investigated. In 1992, Hillier et al observed that T. vaginalis was significantly more common in pregnant women with an intermediate Nugent score than in pregnant women with a normal or bacterial vaginosis–associated Nugent score [11]. This finding seems to have been largely forgotten, since it is seldom mentioned in the T. vaginalis literature. Taken together, these observations led us to hypothesize that the vaginal microbiota of T. vaginalis–infected women differs significantly from that of women without T. vaginalis. Here, we confirm in nonpregnant women that T. vaginalis is strongly associated with an intermediate Nugent score and that the vaginal microbiota of many women with T. vaginalis infection is indeed unique. Moreover, we show that T. vaginalis infection is very strongly associated with a previously unknown, uncultivated member of the genus Mycoplasma.

MATERIALS AND METHODS

Study Population and Specimen Collection

Briefly, subjects were 400 unselected women attending the New Orleans STI clinic from 2003 to 2004 who were enrolled in a study of the role of Mycoplasma genitalium in endocervicitis. A detailed description of the patient population has been published elsewhere [12]. All participants underwent a pelvic examination. During the examination, a speculum was inserted into the vagina, and 4 vaginal swab specimens were obtained and analyzed as follows: (1) Gram staining and determination of vaginal pH by use of a vaginal secretion smear and indicator paper (ColorpHast Indicator Strips [pH 4.0–7.0], EMD Chemicals, Gibbstown, NJ); (2) detection of T. vaginalis, clue cells, and white blood cells on a wet mount and detection of yeast and performance of the “whiff test” by use of a 10% KOH slide preparation; (3) detection of T. vaginalis, using the InPouch culture system (BioMed, White City, OR); (4) detection of M. genitalium. The laboratory technician performing the Nugent score had >12 years of experience and was blinded to the microbiota analysis results. After cleaning the face of the cervix, endocervical swab specimens were obtained for detection of M. genitalium, Chlamydia trachomatis, and Neisseria gonorrhoeae. The study was approved by the LSU Health Sciences Center Institutional Review Board, and informed consent was obtained from all study subjects. Complete data sets were available from 394 women, who are referred to hereafter as the “endocervicitis cohort.”

Laboratory Methods

Nucleic Acid Amplification Assays

C. trachomatis and N. gonorrhoeae were detected by a strand displacement amplification assay following the manufacturer's instructions (ProbeTec, Becton Dickinson, Sparks, MD). DNA was extracted, and an M. genitalium–specific polymerase chain reaction (PCR) was performed as described in detail by Mena et al [13], except that a dot blot was used for amplicon detection.

Identification of Bacterial Taxa in Vaginal DNA Specimens

Stored vaginal DNA extracted as described above for the detection of M. genitalium underwent 454 pyrosequencing of PCR-amplified 16S ribosomal RNA (rRNA) gene segments to identify bacterial taxa (operational taxonomic units or OTUs). PCR amplification of the V4–V6 region of the 16S rRNA gene, pyrosequencing, postprocessing, and quality checking of the resulting reads were performed as previously described [14, 15]. A total of 238 074 reads with an average length of 487 nucleotides were used in the analysis. A BLAST analysis was performed against a local database, using a 97% sequence similarity threshold. The database contained 283 16S rRNA gene sequences previously detected in vaginal specimens either by us or by others as reported in the literature. Lists of National Center for Biotechnology Information reference numbers, literature references, and the sequences in this database are available on request. Following the BLAST search, 25 767 reads (10.8%) were <97% homologous to sequences in our vaginal microbiota database. Taxonomic assignment of these reads was performed using the RDP Classifier set at an 80% confidence threshold [16]. Of these, 11 018 had a confidence threshold of <80% for classification at the genus level and were excluded from the analysis. By use of this approach, 122 OTUs were identified and form the basis for the vaginal microbiota analyses described below, with the exception of the principal coordinates analysis.

Cloning and Sequencing

Universal bacterial 16S rRNA gene primers (17F: GTTTGATCCTGGCTCAG; and 1492R: GGTTACCTTGTTACGACTT) were used to amplify nearly complete 16S rRNA gene sequences from 3 vaginal DNA specimens in which pyrosequencing analysis had revealed the presence of an unknown, highly abundant Mycoplasmataceae sequence. The PCR amplicons were cloned using a TOPO TA cloning kit in accordance with the manufacturer's instructions (Life Technologies, Carlsbad, CA) and were sequenced by an outside contractor (Davis Sequencing, Davis, CA). Sequences were aligned using Sequencher, version 4.10.1(Gene Codes, Ann Arbor, MI).

Analytic and Statistical Methods

T. vaginalis infection was defined as a T. vaginalis–positive result of an InPouch culture of a vaginal swab specimen, and Nugent scores were categorized as normal (scores of 0–3), intermediate (scores of 4–6), and bacterial vaginosis (scores of 7–10).

Principal coordinates analysis of the vaginal microbiota in T. vaginalis–infected and T. vaginalis–uninfected women in each Nugent score category (normal, intermediate, and bacterial vaginosis) was performed using the QIIME pipeline [17]. Sequences were clustered at 97% similarity, using UCLUST [18]; were aligned using PyNast [19] and the Greengenes core reference alignment [20]; and were classified using the RDP Classifier 2.2 [16]. To perform UniFrac analysis, a phylogenetic tree (not shown here) was created using FastTree 2.1.3 [21], and β-diversity was calculated using unweighted UniFrac [22] after normalization by random subsampling to a common depth on the basis of the number of reads in the data sets [23]. Analyses of similarity [24] were performed using QIIME to test whether or not bacterial community composition of T. vaginalis–infected and T. vaginalis–uninfected specimens in each Nugent score category differed significantly.

Vaginal bacterial communities were clustered on the basis of the relative percentage abundance of OTUs in each specimen, using complete linkage clustering with Euclidian distance. Only OTUs that were present in at least 1 specimen at an abundance of ≥1% were included in this analysis. A heat map based on the cluster analysis was generated using a locally developed pyrosequencing data analysis pipeline as previously described [25].

Differences between T. vaginalis–uninfected and T. vaginalis–infected women (eg, ethnicity and clinical characteristics) were assessed using χ² analysis or the Fisher exact test, for categorical characteristics, and t tests, for continuous measures (eg, age).

The average number of reads was lower in the T. vaginalis–uninfected group (1497) than in the T. vaginalis–infected group (6483). Consequently, the number of reads was included as a covariate in regression analyses of OTUs in T. vaginalis–uninfected and T. vaginalis–infected women. Logistic regression methods were used to assess differences in OTU prevalence.

RESULTS

Relationship of T. vaginalis and Other STIs to the Nugent Score in the Endocervicitis Cohort of Women

Figure 1A shows the relationship between T. vaginalis prevalence and the composition of vaginal microbiota as defined by Nugent score. T. vaginalis prevalence was significantly greater in the intermediate group (45%) as compared to the normal (16%) and bacterial vaginosis (18%) groups (P = .0001 for both comparisons). In contrast, the prevalence of C. trachomatis, N. gonorrhoeae, and M. genitalium was relatively constant across the Nugent score groups (Figure 1B). Additionally, we found that the rates of several STI behavioral risk factors (≥2 sex partners in the last year, ≥2 sex partners in the last 2 months, and a new sex partner within the last 2 months) did not vary across the Nugent score groups (data not shown).

Figure 1.

Prevalence of sexually transmitted infections relative to Nugent score. A, Percentage of women infected with Trichomonas vaginalis, by Nugent score category. B, Percentage of women infected with Chlamydia trachomatis (Ct), Neisseria gonorrhoeae (Ng), and Mycoplasma genitalium (Mg), by Nugent score category.

Patient Specimens Selected for Vaginal Microbiota Analyses by Pyrosequencing of 16S rRNA Genes

To explore the relationship of the vaginal microbiota to T. vaginalis infection in a relatively small sample that would allow adequate assessment of cases in all 3 Nugent score categories, 30 samples from T. vaginalis–infected women were matched by Nugent score to 30 samples from T. vaginalis–uninfected women. Ten pairs of matched samples were selected for each of the 3 Nugent score categories. The characteristics of the patients from whom these samples were obtained are summarized in Table 1. T. vaginalis–infected women were younger than T. vaginalis–uninfected women, but the difference was not significant. The population was >90% African American. Concurrent STIs were more common among T. vaginalis–infected women, but the difference was significant only for M. genitalium (27% vs 7%; P = .04). A total of 122 OTUs were identified in the 60 specimens by pyrosequencing. One vaginal specimen in the T. vaginalis–uninfected group was aberrant in that 99% of the sequences in the pyrosequencing library were Neisseria species. In contrast, the abundance of Neisseria species in the pyrosequencing libraries of 9 other women who tested positive for N. gonorrhoeae by the ProbeTec assay was ≤0.03%. Therefore, this specimen was not included in the analyses described below.

Table 1.

Characteristics of the Matched Patient Populations Selected for Comparison of the Vaginal Microbiota of Women With and Those Without Trichomonas vaginalis Infection

Variable	T. vaginalis Infection (n = 30)	No T. vaginalis Infection (n = 30)	Pa
Demographic characteristic
Age, y	24.7	27.1	.18
African American ethnicity	28 (93.3)	29 (96.7)	1.0
STI behavioral risk factor
≥2 sex partners last 2 months	9 (30.0)	6 (20.0)	.37
≥2 sex partners last 12 months	17 (56.7)	15 (50.0)	.61
New sex partner in last 2 months	8 (26.7)	6 (20.0)	.54
Past history of STI
T. vaginalis	8 (26.7)	12 (41.4)b	.23
Bacterial vaginosis	6 (20.0)	13 (43.3)	.05
Chlamydial infection and/or gonorrhea	18 (60.0)	19 (63.3)	.79
Concurrent bacterial STI
C. trachomatis	10 (33.3)	7 (23.3)	.39
N. gonorrhoeae	7 (23.3)	3 (10.0)	.17
M. genitalium	8 (26.7)	2 (6.7)	.04
Positive vaginal yeast culture	1 (3.3)	1 (3.3)	1.0c
Nugent score	4.7	4.7	NDd

Data are no. (%) of women or means.

Abbreviations: C. trachomatis, Chlamydia trachomatis; M. genitalium, Mycoplasma genitalium; N. gonorrhoeae, Neisseria gonorrhoeae; STI, sexually transmitted infection.

^a By χ² analysis, unless otherwise indicated.

^b One patient had missing data.

^c By the Fisher exact test.

^d Not done (ND) because women with and women without T. vaginalis infection were matched on the basis of Nugent score.

Influence of T. vaginalis Infection on the Prevalence of Specific Bacterial OTUs in Vaginal Specimens

We compared the prevalence of all 122 individual OTUs in the 30 T. vaginalis–infected women and the remaining 29 T. vaginalis–uninfected women. Only OTUs that differed significantly are shown in Table 2, and all 16 were more common among T. vaginalis–infected women.

Table 2.

Prevalence of Individual Operational Taxonomic Units (OTUs) Among Women With and Women Without Trichomonas vaginalis Infection

Bacterial OTUa	T. vaginalis Infection Status, Women, No. (%)		Adjusted Pb
	Infection (n = 30)	No Infection (n = 29)
Bulleidia	16 (53)	8 (28)	.045
Campylobacter	4 (13)	2 (7)	.039
Dialister	28 (93)	22 (76)	.049
Fusobacterium	12 (40)	4 (14)	.011
Gemella	22 (73)	10 (34)	.021
Hallella	13 (43)	4 (14)	.012
Mnola	19 (63)	1 (3)	.026
Megasphaera 2	22 (73)	11 (38)	.019
Parvimonas	23 (77)	5 (17)	.002
Peptoniphilus	24 (80)	14 (48)	.010
Peptostreptococcus	24 (80)	15 (52)	.008
Prevotella 123b-95	7 (23)	2 (7)	.036
Prevotella rRNA247	9 (30)	1 (3)	.006
Prevotella T05-04	9 (30)	4 (14)	.047
Sneathia	27 (90)	14 (48)	.032
Sutterella	8 (27)	1 (3)	.044

^a Of the 122 OTUs identified in this study, only those with a P value of < .05 for differences are shown here. A complete list all OTUs analyzed is available on request.

^b Adjusted by logistic regression analysis for the number of reads available for each specimen.

One of the OTUs that was significantly more prevalent in T. vaginalis–infected specimens was Mnola (Table 2), a previously unrecognized Mycoplasma organism. This sequence was detected in 63% of T. vaginalis–infected specimens (19/30) and in only 3% of T. vaginalis–uninfected specimens (1/29; P = .026). PCR, cloning, and sequencing revealed 13 nearly full-length (1472 bp) Mnola 16S rRNA gene sequences in vaginal specimens whose pyrosequencing libraries showed a high abundance of Mnola reads. A BLAST search using this Mnola sequence (GenBank accession no. JX508800) confirmed that the organism is a member of the genus Mycoplasma (Figure 2). The closest match (94% sequence identity) is to the 16S rRNA gene of an uncultivated Mycoplasma found in a bovine rumen specimen [26]. In contrast, the Mnola 16S rRNA gene sequence is only 85% similar to the closest human pathogen, M. genitalium, and only 78% similar to Mycoplasma hominis. A Spearman correlation analysis of the relative abundances of M. hominis and Mnola among all 59 specimens found them to be inversely correlated (rho = 0.32; P = .014).

Figure 2.

Neighbor-joining tree analysis of 16S ribosomal RNA gene sequences showing the phylogenetic position of the Mnola sequence within the Mycoplasma genus. Members of the Mycoplasma genus, along with representatives of the Mollicutes class (Anaeroplasma, Spiroplasma, and Acholeplasma genera) were included for comparison. The tree was derived using nearly full-length (approximately 1470 nucleotides) 16S ribosomal RNA gene sequences obtained from GenBank. It was constructed by the neighbor-joining method, using the MEGA 5 software package. Bootstrap values are shown at the branch points as percentages of 1000 analyses. The scale bar shows the number of base substitutions per site. GenBank accession numbers are shown before the organism names of each of the operational taxonomic units included in the tree.

Principal Coordinates Analysis

Principal coordinates analysis of the bacterial communities in T. vaginalis–infected and T. vaginalis–uninfected women and of T. vaginalis–infected and T. vaginalis–uninfected women, stratified by Nugent score category, is shown in Figure 3. Figure 3A suggests that there is no clear difference between the composition of microbial communities in T. vaginalis–infected women and T. vaginalis–uninfected women in general. However, as can be seen in Figure 3B–D, principal coordinates analysis of women stratified by Nugent score categories showed that, while there was no separation of the T. vaginalis–infected and T. vaginalis–uninfected communities among women with bacterial vaginosis (Figure 3B) and increased separation among the communities with intermediate Nugent scores (Figure 3C), there was complete separation among the communities with normal Nugent scores (Figure 3D). By analysis of similarity, the difference between T. vaginalis–infected women and T. vaginalis–uninfected women among those with normal Nugent scores and those with intermediate Nugent scores were significant (P = .001 and P = .015, respectively).

Figure 3.

Principal coordinates analysis of Trichomonas vaginalis–infected specimens (blue dots) and T. vaginalis–uninfected specimens (red dots). There were 30 T. vaginalis–infected women and 28 T. vaginalis–uninfected women in this analysis. One T. vaginalis–uninfected woman with a Nugent score–intermediate specimen that had a low read count (95 reads total) was not included, as was another specimen from a T. vaginalis–uninfected woman in which 99% of the reads were Neisseria species. The halos surrounding the dots represent the degree of variation among jackknifed replicates. The panels show data for all T. vaginalis–infected women and all T. vaginalis–uninfected women (A), data for women with Nugent score–determined bacterial vaginosis (B), data for women with intermediate Nugent scores (C), and data for women with normal Nugent scores (D). By analysis of similarity, P values for the difference in bacterial community distance between T. vaginalis–infected and T. vaginalis–uninfected women with Nugent scores indicating normal findings, intermediate findings, and bacterial vaginosis were P = .001, .015, and .433, respectively.

Cluster Analysis

A cluster analysis was performed to further visualize the association between T. vaginalis and the composition of vaginal microbiota. As can be seen in Figure 4, there are 2 main clusters, A and B. Each is divided into subclusters, termed A1–A4 and B1–B4, respectively. Twenty-eight specimens were included in cluster A, and 31 were included in cluster B. All specimens in cluster A were classified by Nugent score as normal or intermediate. All Nugent score–confirmed bacterial vaginosis specimens were found within cluster B. However, cluster B also included 4 specimens with normal Nugent scores and 7 with intermediate Nugent scores. The 30 T. vaginalis–infected specimens were evenly divided between the 2 clusters. Subclusters A2 and A4 had high abundances of Lactobacillus crispatus and Lactobacillus iners, respectively. Between these 2 subclusters, only 2 were T. vaginalis infected. In contrast, all specimens in subclusters A1 and A3 were T. vaginalis infected. These 2 subclusters had very distinctive microbiota. M. hominis was present in all 7 A1 cases, with abundances ranging from 4% to 54% (median, 9%). In contrast, the abundance of M. hominis in the 52 non-A1 specimens ranged from 0% to 12% (median, 0.2%). Mnola was present in 4 of the A1 specimens, with abundances ranging from 1% to 14%. Of the 6 specimens in subcluster A3, M. hominis was present in only 2. In contrast, Mnola was present in all of these cases, with abundances ranging from 9% to 84% (median, 55%). As noted above, M. genitalium was more common among T. vaginalis–infected women, but it appeared to be evenly distributed among the heat map subclusters (data not shown).

Figure 4.

Heat map analysis. The dendrogram at the top of the heat map was generated using complete linkage clustering with Euclidean distance, based on the relative percentage abundance of taxa (operational taxonomic units [OTUs]) in each specimen. Columns represent vaginal specimens, and rows represent bacterial OTUs in this list Pv is an abbreviation of Prevotella. Each cell in the heat map is colorized on the basis of the relative percent abundance of an OTU; a higher abundance is indicated by brighter green. The first bar under the dendrogram reflects the Trichomonas vaginalis infection status of all participants. The second bar represents the Nugent score (NS) category for each case. The case identifier for each case is at the bottom of the figure.

Clinical Correlates With the Heat Map Subclusters

An exploratory analysis suggested that signs of inflammation were more common in the 7 women with cluster A1 microbiota. Therefore, data for these 7 women were compared to data for the 23 T. vaginalis-infected women with non-A1 microbiota. Table 3 shows that vaginal erythema was seen in 6 of 7 women with A1 microbiota (86%) as compared to 6 of 22 (27%) with non-A1 microbiota (P = .011). Copious vaginal discharge, purulent discharge appearance, and cervical petechiae were all more common among women with A1 microbiota, but the differences were not significant.

Table 3.

Comparison of Clinical Examination Findings Between Women in Heat Map Subcluster A1 and All Other Women With Trichomonas vaginalis Infection

Clinical Finding	Examination Finding, Women, Proportion (%)		Pa
	Subcluster A1	All Others
Vaginal mucosa erythema	6/7 (86)	6/22 (27)b	.011
Cervical petechiae	5/7 (71)	6/23 (26)	.068
Copious vaginal dischargec	6/7 (86)	9/23 (39)	.08
Purulent vaginal discharged	3/7 (43)	3/23 (13)	.075

^a By the 2-tailed Fischer exact test.

^b One patient had missing data.

^c Defined as discharge visible at the vaginal introitus prior to insertion of the speculum.

^d Defined as thick, yellow-green secretions.

DISCUSSION

Hillier et al [11] reported in 1992 that Nugent score–intermediate pregnant women were significantly more likely to have T. vaginalis infection, and our data from 394 nonpregnant women attending an STD clinic strongly support this largely forgotten observation. Moreover, we found that this association was unique to T. vaginalis among the STIs assessed in our study and that STI risk behaviors did not play confounding roles. These observations suggest that T. vaginalis might be more prevalent among women with intermediate Nugent scores because of unique shifts in the vaginal microbiota.

As shown in Table 2 there are a number of OTUs known to be associated with bacterial vaginosis, including Peptostreptococcus, Megasphaera-type2, and Sneathia, that are more frequently detected in vaginal specimens from T. vaginalis–infected women [27–31]. The organism most strongly associated here with T. vaginalis is Parvimonas, a well-known oral pathogen associated with dental root canal infections [32, 33]. A previously unknown Mycoplasma organism, Mnola, is also strongly associated with T. vaginalis. Thus, there are clear differences in the composition of vaginal bacterial communities between T. vaginalis–infected women and T. vaginalis–uninfected women.

Principal coordinates analyses (Figure 3) suggest that these differences in the vaginal microbiota of T. vaginalis–infected women and T. vaginalis–uninfected women occur primarily among those with normal and intermediate Nugent scores. Cluster analysis provides further insight into the relationship between T. vaginalis and the vaginal microbiota (Figure 4). Nugent score–intermediate samples were evenly distributed between clusters A and B. All samples from women with Nugent score–confirmed bacterial vaginosis were located in cluster B, while most of the Nugent score–normal samples (16/20) were in cluster A. Fifty percent of the T. vaginalis cases were present in both clusters. Samples in 2 subclusters, A1 and A3, were all T. vaginalis infected and both contain only Nugent score intermediate and normal cases. The most abundant Mycoplasma species in cluster A1 was M. hominis, while Mnola was the most abundant species in A3. Thus, there appear to be 2 unique T. vaginalis–associated microbiota subclusters, which differ with respect to the dominant Mycoplasma species. Cohort studies of women followed for up to a year with quarterly sampling have shown that Nugent score–determined bacterial vaginosis at a given visit significantly increased the risk of incident T. vaginalis infection at a subsequent visit [8, 9]. To a lesser degree, this is true of Nugent score–intermediate microbiota, as well. These observations, combined with our data, suggest the following hypothesis: women with Nugent score–confirmed bacterial vaginosis and a diverse microbiota and a subset of women with an intermediate Nugent score and a microbiota resembling that of women with bacterial vaginosis are at increased risk for incident T. vaginalis infections. In a subset of women, the parasite effects changes in the microbiota, resulting in environments (eg, subclusters A1 and A3) that are presumably more favorable for its long-term survival and/or its transmissibility. The fact that all women with subcluster A1 and A3 specimens were T. vaginalis infected argues against the alternative hypothesis that these microbiota simply increase one's susceptibility to T. vaginalis infection. If this were the case, detection of at least a few T. vaginalis–uninfected women with these microbiota would have been expected, but there were none. The theory that T. vaginalis actively changes the vaginal microbiota in some women could explain the observed association of T. vaginalis with intermediate Nugent scores. Population-based Nugent score surveys have shown that either women with normal Nugent scores or women with bacterial vaginosis predominate, with Nugent score–intermediate women composing only 12%–24% of the total [8, 11, 34–36]. According to our hypothesis, in populations with a high incidence of T. vaginalis infection, the shift in the vaginal microbiota would result in increased numbers of women with intermediate or normal Nugent scores, but there would be a disproportionately greater effect on the former, given the smaller size of this Nugent score category among T. vaginalis–uninfected women.

M. hominis is a well-recognized component of bacterial vaginosis microbiota and quantitative M. hominis nucleic acid amplification assays (NAATs) are predictive of bacterial vaginosis [37]. The association of T. vaginalis and M. hominis has been reported, but stratified analysis of this association by Nugent score category has not been reported [38–40]. In vitro studies have shown that M. hominis is taken up by T. vaginalis and is able to survive within cytoplasmic vacuoles [41, 42]. Moreover, M. hominis significantly affects the metabolism of T. vaginalis [43] and may increase its pathogenicity [44]. In this study, M. hominis was present in 80% of 30 T. vaginalis–infected women, compared with 45% of 29 T. vaginalis–uninfected cases, a nonsignificant difference after adjustment for read numbers. All 7 women with subcluster A1 microbiota had M. hominis, and its abundance in these cases was relatively high. Although the numbers were small, clinical findings in women belonging to subcluster A1 suggest that their vaginal microbiota were associated with a heightened proinflammatory state. This may be important because it is known that T. vaginalis is a risk factor for HIV acquisition, as are local vaginal proinflammatory states in general [6, 7, 45, 46]. If further studies support our findings and if the A1 vaginal microbiota is stable, it could be that this group of women is at especially high risk for HIV infection.

Mnola, a member of the Mycoplasma genus described here for the first time, was rarely present in the absence of T. vaginalis but was present in 63% of T. vaginalis–infected women. The ability of this organism to replicate in vitro is unknown, but it is likely to be poor as it has not yet been identified in multiple past culture studies of vaginal flora, using specialized methods for the growth of Mycoplasma organisms. Given the strength of the association between Mnola and T. vaginalis, it may be that Mnola is an obligate symbiont of T. vaginalis. In this study, we did not use a NAAT for the detection of T. vaginalis, so it is possible that the single Mnola-positive/T. vaginalis culture–negative specimen was actually positive for T. vaginalis. If in fact Mnola and T. vaginalis have a dependent relationship, specimens with very high Mnola abundances should have among the highest abundance of T. vaginalis. Future in vitro and in vivo studies that use quantitative NAATs for T. vaginalis and Mnola will be necessary to investigate these questions further.

There are some limitations to our study. The total number of specimens evaluated is relatively small, and larger numbers of T. vaginalis–infected cases will be necessary to support the heat map cluster assignments made here and to determine the prevalence of this unique microbiota in unselected T. vaginalis–infected women. However, similarities between cluster and subcluster assignments of the T. vaginalis–uninfected women with those of previous publications provide support for the validity of the cluster assignments as presented here [14, 47]. In addition, this is a cross-sectional study, so the temporal stability of the unique T. vaginalis–associated microbiota described here is unknown. Longitudinal cohort studies will be needed to address this question. Finally, NAAT detection of T. vaginalis would have modestly increased the number of T. vaginalis–infected cases available for analysis in this study. Future studies should include such an assay. A recent vaginal microbiota study by Brotman et al did so but suffered from having only 11 T. vaginalis–infected women in the analysis [48].

In summary, we have shown in this cross-sectional study of women with or without T. vaginalis infection who were matched by Nugent score that a significant number of T. vaginalis–infected women have a unique vaginal microbiota in which M. hominis and/or Mnola, a previously unknown Mycoplasma species, are prominent. We hypothesize that T. vaginalis is the driving force behind the formation of this unique microbiota and that this transformation of the vaginal environment enhances its survival and/or transmissibility. T. vaginalis–associated microbiota featuring increased M. hominis abundance may induce an inflammatory host response, which could put women at higher risk for HIV infection. Further research will be needed to address these interesting and potentially important questions.