Association Between Trichomonas vaginalis and Vaginal Bacterial Community Composition Among Reproductive-Age Women

Abstract

Objectives 

Some vaginal bacterial communities are thought to prevent infection by sexually transmitted organisms. Prior work demonstrated that the vaginal microbiota of reproductive-age women cluster into 5 types of bacterial communities; 4 dominated by Lactobacillus species (L. iners, L. crispatus, L. gasseri, L. jensenii) and 1 (termed community state type (CST) IV) lacking significant numbers of lactobacilli and characterized by higher proportions of Atopobium, Prevotella, Parvimonas, Sneathia, Gardnerella, Mobiluncus, and other taxa. We sought to evaluate the relationship between vaginal bacterial composition and Trichomonas vaginalis.

Methods 

Self-collected vaginal swabs were obtained cross-sectionally from 394 women equally representing 4 ethnic/racial groups. T. vaginalis screening was performed using PCR targeting the 18S rRNA and β-tubulin genes. Vaginal bacterial composition was characterized by pyrosequencing of barcoded 16S rRNA genes. A panel of 11 microsatellite markers was used to genotype T. vaginalis. The association between vaginal microbiota and T. vaginalis was evaluated by exact logistic regression.

Results 

T. vaginalis was detected in 2.8% of participants (11/394). Of the 11 T. vaginalis-positive cases, 8 (72%) were categorized as CST-IV, 2 (18%) as communities dominated by L. iners, and 1 (9%) as L. crispatus-dominated (P = 0.05). CST-IV microbiota were associated with an 8-fold increased odds of detecting T. vaginalis compared with women in the L. crispatus-dominated state (OR: 8.26, 95% CI: 1.07–372.65). Seven of the 11 T. vaginalis isolates were assigned to 2 genotypes.

Conclusion 

T. vaginalis was associated with vaginal microbiota consisting of low proportions of lactobacilli and high proportions of Mycoplasma, Parvimonas, Sneathia, and other anaerobes.

In Brief

T. vaginalis was associated with vaginal microbiota 69 consisting of low proportions of lactobacilli and high proportions of Mycoplasma, Parvimonas, Sneathia and other taxa.

The National Health and Nutrition Examination Survey from 2001 to 2004 found that 3.1% of US women were positive for Trichomonas vaginalis genital infections,¹ making it a more pervasive sexually transmitted infection (STI) than Neisseria gonorrhea and Chlamydia trachomatis combined.² In 1999, Cates et al estimated an incidence of 5 million new cases of trichomoniasis annually in the United States.³ Racial disparity in trichomoniasis rates are particularly striking with non-Hispanic black women having over 10-fold higher prevalence rates (13.3%) than non-Hispanic white (1.3%) or Mexican American (1.8%) women.¹ In a longitudinal study of 268 adolescents in Indianapolis, IN, incident cases of T. vaginalis were detected in 23% of participants and reinfection episodes occurred in 31% of those individuals.⁴ Although trichomoniasis is a treatable STI,⁵ it continues to represent a critical public health issue as >85% of cases are asymptomatic,¹ and the protozoan is associated with numerous adverse outcomes, including preterm delivery of low-birth weight infants⁶ and increased susceptibility to and transmission of HIV infection.^7,8 It has been estimated that 746 new HIV cases among US women each year can be attributed to enhanced acquisition of HIV by underlying T. vaginalis infection.⁹

Lactic acid producing bacteria in the vagina, made up in large part by Lactobacillus sp., are thought to play a key protective role in preventing urogenital infection^10,11 by lowering the vaginal pH, generating bacteriostatic and bacteriocidal compounds,¹² and by possible competitive exclusion of nonindigenous organisms that include opportunistic and overt pathogens. Ravel et al¹³ recently demonstrated by high-throughput culture-independent molecular analysis that the vaginal bacterial communities of North American women can be classified into 5 community state types (CSTs); 4 of which are dominated by 1 of Lactobacillus iners, L. crispatus, L. gasseri, or L. jensenii. The remaining group (termed CST-IV) comprised 27% of the women in the study and was characterized by higher vaginal pH and greater relative abundance of strict anaerobic organisms, including Atopobium, Prevotella, Dialister, Gardnerella, Megasphaera, Peptoniphilus, Sneathia, Eggerthella, Aerococcus, Finegoldia, and Mobiluncus. Our overall hypothesis is that the 5 different vaginal bacterial communities differ in terms of their ability to fend off colonization by pathogens.

In the current study, we sought to compare the vaginal microbiota of T. vaginalis-positive and T. vaginalis-negative women using cultivation-independent methods based on the analysis of 16S rRNA genes. This molecular epidemiologic study, using high-throughput cultivation-independent methodologies, is able for the first time to decipher how vaginal bacterial communities differ among T. vaginalis cases. Prior work on the association of vaginal microbiota and STIs have focused on cultivation-based studies of lactobacilli and microscopy evaluation of Gram stained smears.^14–18 Cultivation-based approaches are hindered as many microbial species resist cultivation in the laboratory¹⁹ and Gram stain provides only morphologic information.²⁰ In addition, we genotyped the T. vaginalis-positive samples using species-specific microsatellite (MS) markers, and successfully assigned 7 of the 11 isolates to the 2 types that characterize the parasite’s global population structure.²¹

MATERIALS AND METHODS

Clinical Study Design

This report used secondary analysis and repository samples from the study by Ravel et al of 394 North American women representing 4 ethnic/racial groups (white, black, Hispanic, and Asian), which has been previously described.¹³ Briefly, between 2008 and 2009, nonpregnant, women aged 12 to 45 with regular menstrual cycles were recruited to an observational, cross-sectional study in 1 of the 2 US cities (Atlanta or Baltimore). Women were ineligible if they reported vaginal discharge in the prior 48 hours, pregnancy, use of intravaginal products or sexual activity in the prior 48 hours, use of antibiotics or antimycotics in the prior 30 days, significant medical or gynecological conditions (e.g., posthysterectomy or immunocompromised status such as congenital, acquired or secondary to medication), or current menstrual bleeding or vaginal discharge. Participants underwent a detailed interview in a private office, which was self-administered or administered by a female interviewer. The interview was focused on personal hygiene, sexual and contraceptive history, and demographic information. Participants self-reported racial classification using predefined categories. Data on self-reported racial identification were collected because of observed disparities in STI rates between US racial groups.¹

Using validated protocols,^22,23 participants self-collected 2 midvaginal swabs using the Elution-swab system (Copan) that were respectively used to prepare a smear for Nugent Gram stain scoring²⁰ and to characterize the composition and structure of the resident bacterial communities using pyrosequencing of barcoded 16S rRNA genes.¹³ The Nugent scoring criteria reflects the relative abundance of large Gram-positive rods (lactobacilli), Gram-negative and Gram-variable rods and cocci (i.e., Gardnerella vaginalis, Prevotella, Porphyromonas, and peptostreptococci), and curved Gram-negative rods (Mobiluncus).²⁰ This technique assesses the relative numbers of morphotypes, allowing a coarse classification of certain vaginal microbes. Nugent scores reflect the range of vaginal microbiota states. A score of 0 to 3 is designated as normal, 4 to 6 as an intermediate state, and 7 to 10 as a high Nugent score indicative of bacterial vaginosis diagnosis.²⁰ Participants also self-collected and reported vaginal pH using the VpH glove (Inverness Medical). Study staff confirmed the pH measurement and also took a digital photograph of the pH strip adjacent to the color scale.

The study was approved by Institutional Review Boards at the University of Maryland School of Medicine and the Emory University School of Medicine. All participants provided written informed consent. The study was registered at clinicaltrials.gov under ID NCT00576797.

T. vaginalis Screening and Primers

Whole genomic DNA extraction was performed on the self-collected vaginal swabs as previously described.¹³T. vaginalis was detected by polymerase chain reaction (PCR)-based methods using 2 primer sets, TV16Sf-2/TV16Sr-2 (a 323 bp amplicon) targeting the 18S rRNA gene²⁴ and TVK3/TVK7 (a 300 bp amplicon) targeting a DNA repeat region of the T. vaginalis genome.²⁵ A third primer set, BTub3/bkmt (a 195 bp amplicon), targeting the ß-tubulin gene was only used on 4 discordant samples and confirmed the absence of T. vaginalis.²⁴ All PCRs were performed as previously reported, and amplicons were visualized by gel electrophoresis.

T. vaginalis Genotyping

T. vaginalis-positive samples were genotyped using 11 MS loci, as previously described.²⁶ This subset of the published 21 MS panel consists of the most informative set of markers that allow distinction between type 1 and type 2 parasites, the 2 groups that T. vaginalis parasites are found to cluster into world-wide,²¹ and is ideal for use with clinical samples in which low levels of target parasite DNA are a confounding factor. All genotyping PCR reactions were performed in duplicate, and size polymorphisms were measured by capillary gel electrophoresis on an ABI 3,130xl sequencer. GeneMapper 4.0 (ABI, Foster City, CA) was used to score MS allele sizes, and all peaks were manually edited. Discrepant allele calls were repeated for a third time. Type assignment was then performed using the Bayesian clustering program, STRUCTURE 2.2.²⁷

Statistical Analysis

Fisher exact test and exact logistic regression were used to determine the association between vaginal bacterial CST¹³ and the detection of T. vaginalis. Factors collected from questionnaires that had been identified on the basis of previous literature and biologic plausibility were evaluated in univariable analyses. The limited number of observed T. vaginalis-positive cases (11 women) would not allow for further multivariable odds ratio (OR) analyses. To assess differences in the bacterial community structure between T. vaginalis cases and controls, we also used distance-based redundancy analysis (dRDA)²⁸ and a permutation test to assess significance of the results. Data were analyzed using STATA/SE 10.0 for Windows (Stata Corporation, College Station, TX), and the dRDA was performed using the Vegan package implemented in R (R Foundation for Statistical Computing, Vienna, Austria).^29,30

RESULTS

Of the 394 women, 11 (2.78%) were positive for T. vaginalis. Among the T. vaginalis-positive cases, 8 (72%) were detected among women belonging to the low-lactobacilli CST-IV group, whereas 2 (18%) and 1 (9%) were among women with bacterial communities dominated by L. iners and L. crispatus, respectively (P = 0.05, Table 1), reflecting the differences in classification of bacterial CST by T. vaginalis positivity. Figure 1 displays the proportions of bacterial taxa found among the 383 T. vaginalis-negative cases in comparison with the 11 T. vaginalis-positive study participants.

TABLE 1: Characteristics of Trichomonas vaginalis Positive and Negative Participants in a Cross-Sectional Study of 394 Women, Baltimore, MD and Atlanta, GA, 2008–2009

Figure 1: Heatmap of relative percent abundance of bacterial taxa found in the vaginal bacterial communities of 394 reproductive age women (A) and of the 11 T. vaginalis positive cases (B). Vaginal bacterial CST is indicated as described by Ravel et al.¹³ Nugent Gram stain scores, pH measurements and T. vaginalis positivity are shown for each of the 394 samples.

A multivariate analysis of variance table resulting from dRDA analysis (data not shown) indicated that there were statistically significant differences in bacterial community structure within CST-IV between the T. vaginalis-positive and T. vaginalis-negative samples (permuted P = 0.013, number of permutations = 10,000). Among women assigned to CST-IV, Mycoplasma and, to a lesser degree, Parvimonus and Sneathia were associated with T. vaginalis positivity, and Streptococcus, Prevotella, L. iners and Atopobium were associated, although not as strongly, with T. vaginalis-negative controls. However, only 2.5% of the total variation between the communities of CST-IV is explained by T. vaginalis, indicating that T. vaginalis is not a robust predictor of community structure in this study or there were too few cases observed. For detailed information on the composition of vaginal communities of the T. vaginalis-positive cases, see supplementary information in Ravel et al¹³ for the following double-masked identification numbers: S023, S070, S102, S103, S104, S109, S234, S255, S307, S329, S383.

Of the 11 T. vaginalis cases, 10 (91%) were observed among black women, 1 from an Asian woman (9%) and none among the Hispanic or white women (Table 1, differences between case and control, P = 0.00002). In addition, none of the T. vaginalis-positive women reported hormonal contraceptive use versus 28% of controls (P = 0.04). The proportions in the paired levels of vaginal pH were also significantly different (P = 0.01), with pH higher (>4.5) in 90% of cases compared with 46% of T. vaginalis-negative women. pH is also displayed in (Fig. 1).

In exact logistic regression modeling, CST-IV was associated with a significant 8-fold increased odds for detection of T. vaginalis compared with women in the L. crispatus CST (OR: 8.26, 95% CI: 1.07–372.65) (Table 2). Similarly, high Nugent score was associated with a 9-fold increased odds for the presence of T. vaginalis compared with women with low Nugent scores (OR: 9.53, 95% CI: 1.77–95.71).

TABLE 2: Odds of Trichomonas vaginalis Detection by Various Factors, Baltimore, MD and Atlanta, GA, 2008–2009 (N = 394 Women)

Seven of the 11 T. vaginalis-positive samples were successfully genotyped at 6 or more MS loci. Three isolates were found to cluster within type 1 (S023, S102, S103), and 4 isolates within type 2 (S070, S104, S234, and S329), and none of the 7 isolates appeared to be made up of more than 1 genotype. No statistically significant associations between isolate type and participant characteristics were found, including vaginal bacterial CST, pH, Nugent Gram stain score, age, race/ethnicity, use of hormonal contraception, or number of sexual partners (data not shown).

COMMENT

Our study revealed a disproportionate burden of T. vaginalis among women whose vaginal bacterial communities were determined by molecular analysis to be composed of higher proportion of members of the genera Mycoplasma, Parvimonas, Sneathia, and other taxa (Fig. 1). These samples concomitantly displayed relatively low abundance of lactobacilli. A limitation of our study is that we can only report on the association between vaginal bacterial communities and the presence of T. vaginalis infection; we cannot establish in this cross-sectional study design whether the observed vaginal CST occurred before or after T. vaginalis acquisition. However, we can hypothesize that a highly plausible mechanism for a causal relationship is the lack of production of lactic acid by lactobacilli. O’Hanlon et al demonstrated inhibition of bacteria by lactic acid, and we can hypothesize that lactic acid may play a protective role against T. vaginalis as well.¹¹ Prior longitudinal studies have shown a significant association between vaginal microbiota (as determined by Amsel’s clinical criteria for bacterial vaginosis or Gram stain analysis) and T. vaginalis acquisition.^15,16,31 Our data provide the critical impetus that a large, longitudinal study with frequent sampling is needed to determine whether certain types of vaginal bacterial communities predispose women to acquisition of STIs or whether the observed CSTs are the result of the STI. Clinical intervention on the former could reduce risk for STI acquisition.

In this report, we also sought to detect whether specific correlations between the vaginal microbiota and T. vaginalis genotype could be made; however, with the low number of T. vaginalis positive samples genotyped, we failed to find any significant associations. It has been demonstrated that globally T. vaginalis clusters into 2 genetically distinct types, type 1 and type 2, which appear to be equally geographically widespread.²¹ Future studies with larger sample sizes should be considered.

Our finding of an association between T. vaginalis and CST-IV may be a result of confounding based on two observations. First, black women have higher prevalence rates for T. vaginalis¹; and second, CST-IV is more commonly observed among black women.¹³ Ravel et al previously reported that black women made up 39% of women in CST-IV,¹³ and among the T. vaginalis-positive cases in our study, 90% were black women. However, it is also possible that the racial disparity in T. vaginalis observed in our study might reflect a STD transmission core. Laumann et al modeled bacterial STD transmission in a nationally representative dataset and demonstrated that blacks’ higher STI rates could be partially explained by the patterns of sexual networks and segregated partner choices (assortative mating).³² The latter would suggest that the higher T. vaginalis prevalence rate among black women in the current study indicates that women with communities with relatively low-lactobacilli abundance (CST-IV) could be at increased risk for T. vaginalis acquisition on exposure.

Strengths of this study include a large, racially diverse study population, validated protocols for self-sampling,^22,23 and the use of molecular techniques to detect both T. vaginalis and vaginal bacterial CSTs. Published PCR sensitivities for T. vaginalis range from 81% to 97%, exceeding the sensitivities that have been reported for wet mount evaluation, Papanicolaou test, DNA probe, or culture.⁵

Limitations of the study include the recruitment of only women who reported no vaginal discharge and the limited sample size of 11 T. vaginalis cases out of a study of 394 women. Although statistically significant, the OR estimate for the low-lactobacillus CST-IV should be taken cautiously owing to the wide confidence intervals likely due to the small number of observed cases. There were no T. vaginalis cases detected among the L. gasseri and L. jensenii CSTs that may reflect protection by these Lactobacillus spp. or limitations of the sample size. In addition, a limitation of PCR is that viable and nonviable organisms cannot be distinguished. Participants reported they had not taken an antibiotic in the prior 30 days, and one of the T. vaginalis-positive cases indicated on survey questioning that she had a T. vaginalis infection treated in the prior 60 days. Previously treated infections could potentially have resulted in a positive PCR result without viable organisms, although dead cells and free DNA are unlikely to persist. Lastly, Gardnerella vaginalis could be under represented when using the primer set 27F and 338R,¹³ however, there are biases with any primer set.

It is now recognized that the vaginal microbiota play a major role in maintaining the reproductive health of women. A recent study by Gatski et al reported that high Nugent Gram stain score was associated with early failure of metronidazole single-dose treatment for trichomoniasis among HIV-positive women.³³ In addition, differences in vaginal bacterial community composition constitute a significant variable that has not been emphasized in current research on risk factors for STI acquisition. Future research should focus on functional differences of the various vaginal bacterial communities, including the dominant lactobacilli species (and strains), and their ability to protect against urogenital pathogens. Longitudinal studies of vaginal microbiome and STIs are critical in order to determine temporality and causality. All women recruited to the current study did not report vaginal discharge, highlighting further that T. vaginalis is often an asymptomatic infection and suggests that screening could increase awareness and reduce transmission. Rapid screening tests for T. vaginalis⁵ are available to medical practitioners, and self-collected vaginal swabs can be used for T. vaginalis detection.³⁴ In addition, use of self-collected midvaginal swabs for microbiome analysis has been validated^22,23 and provides feasibility for field-based longitudinal studies of STI incidence.¹⁶ An increased understanding of the vaginal microbial ecosystem could lead to more effective and personalized strategies for the prevention of genital T. vaginalis and other STIs.