The vaginal microbiota and behavioral factors associated with genital Candida albicans detection in reproductive-age women

Abstract

Background:

Vulvovaginal candidiasis is commonly diagnosed and has been associated in prospective studies with the acquisition of HIV. Little data is available on how the composition of the vaginal microbiota, and other risk factors, are associated with the molecular detection of Candida albicans – a common cause of vulvovaginal candidiasis.

Methods:

In a cross-sectional study, self-collected vaginal swabs were obtained from 394 non-pregnant, reproductive-age women. C. albicans was detected using PCR targeting C. albicans ITS1/2 region. Vaginal microbiota was characterized by 16S rRNA gene amplicon sequencing of the V3-V4 hypervariable regions and clustered into community state types (CSTs). Multiple logistic regression identified factors associated with C. albicans detection.

Results:

21% had C. albicans detected and 46% reported vaginal symptoms in the prior 60 days. There was a 2-fold increase in the odds of C. albicans if a woman was in a L. crispatus-dominated CST compared to CSTs with low-Lactobacillus levels (aOR: 2.05, 95% CI: 0.97-4.37). History of self-treatment with antifungals, L. crispatus relative abundance, and receptive oral sex were also significantly associated with C. albicans detection.

Conclusions:

A L. crispatus-dominated vaginal microbiota is thought to protect women from both development of bacterial vaginosis and incidence of sexually transmitted infections; however, our data suggest that L. crispatus is associated with increased C. albicans detection. Receptive oral sex may also be a risk factor for vaginal C. albicans colonization.

Short Summary:

High relative abundance of Lactobacillus crispatus, receptive oral sex and over-the-counter antifungal use were associated with genital Candida albicans detection in a cross-sectional study of reproductive-age women.

Background

Vulvovaginal candidiasis (VVC) is a commonly diagnosed vaginitis caused by species of Candida, with C. albicans being the most common etiology.(1) Symptoms are nonspecific and include vaginal discharge, soreness, irritation, burning, and dyspareunia.(1) In addition to the morbidity associated with VVC, C. albicans colonization has been associated with increased HIV acquisition in longitudinal studies.(2, 3) VVC is relatively common with 75% percent of women experiencing at least one episode of VVC in their lifetime, and 5% developing recurrent vulvovaginal candidiasis (RVVC).(4) Risk factors for VVC include reproductive age, pregnancy, hormone replacement therapy, antibiotic use, immunosuppression, uncontrolled diabetes, oral contraceptive pills, frequent sexual intercourse, and receptive oral sex.(1, 5, 6) Contributing factors for RVVC have not been as thoroughly characterized but are thought to include the former factors as well as a magnified host immune response.(4) Specifically, sequence variants of several known innate host defense genes have been associated with increased susceptibility to RVVC.(7)

Importantly, C. albicans colonization is distinguished from the clinical diagnosis of VVC; the additional clinical findings of signs or symptoms of vulvovaginal inflammation must also be present to diagnose VVC. Fungal factors associated with the pathogenesis and the development of symptomatic VVC have been described; however, host factors are not well understood.(4, 5) Cross-sectional studies have reported that C. albicans is detected in the vaginal microbiota of approximately 20% of asymptomatic women.(8, 9) A longitudinal study demonstrated that colonization is often temporary, with only 4% of women being persistently colonized over a one year period.(8) During pregnancy, rates of C. albicans colonization rise to 30%.(10) Asymptomatic C. albicans colonization often represents colonization with a much lower fungal load than is present in episodes of VVC.(11)

Prior observational studies have largely focused on understanding risk factors for VVC, in which VVC was defined by clinical assessment, detection on wet mount and/or culture, or DNA probe,(12-15) and have not explored predictors of vaginal C. albicans detection using more sensitive molecular methods like PCR, which are capable of detecting the lower fungal load which may be present in women without VVC diagnosis. Additionally, while several clinical trials have found mixed results deploying Lactobacillus probiotics in the treatment of symptomatic VVC,(16) the relationship between C. albicans and the vaginal microbiota remains unclear. We sought to understand epidemiologic factors associated with molecular C. albicans detection, and to evaluate how the vaginal microbiota and relative abundance of individual bacterial taxa differ between women with and without C. albicans detected using molecular methods.

Materials and Methods

Sample Collection and Study Design

We used secondary data and repository samples from the Vaginal Microbiota 400 Woman Study (VM400) which Ravel et al. previously described.(17) Briefly, non-pregnant, sexually-active women aged 18 to 45 representing four ethnic/racial groups (White, Black, Hispanic, and Asian) were recruited to an observational, cross-sectional study between 2008-2009 in Baltimore and Atlanta. Exclusion criteria included self-report of vaginal discharge or use of vaginal douches, sprays, wipes, or contraceptive spermicides in the prior 48 hours, irregular menstrual cycle or current menstruation, pregnancy, and use of antibiotics or antimycotics in the prior 30 days.

Participants self-collected two mid-vaginal swabs at the visit. One swab was used to characterize the composition and structure of the vaginal microbiota, and another swab was used to prepare a smear for Nugent Gram stain scoring.(18) Each participant completed a questionnaire focused on demographics, personal hygiene and sexual behaviors in the prior 60 days, and reproductive health information. All participants provided informed consent. The Institutional Review Boards at Emory University School of Medicine, Grady Memorial Hospital and the University of Maryland School of Medicine approved the protocol. Guidelines of the universities were followed in the conduct of the clinical research. The study was registered at clinicaltrials.gov under ID .

Sample Processing

Whole genomic DNA extraction was performed on the self-collected vaginal swabs in April 2009 as previously described, and DNA extracts were stored at −80°C.(17) Detection of C. albicans was performed in April 2011 using polymerase chain reaction (PCR) targeting the ITS½ region of the C. albicans genome (19) followed by gel electrophoresis. The V3-V4 regions of the 16S rRNA gene were amplified in November 2013 according to the protocol in Fadrosh et al.(20) Amplicon pooling, sequencing on an Illumina MiSeq instrument, and sequence data processing were carried out as in Holm et al.(21) Amplicon sequence variants (ASVs) generated by DADA2 were classified using the RDP Naïve Bayesian Classifier trained with the SILVA v128 16S rRNA gene database as implemented in the dada2 R package.(21) ASVs of major vaginal taxa were assigned species-level annotations using speciateIT (http://ravel-lab.org/speciateit/). Bacterial taxa present at less than 10⁻⁵ across all samples were removed and samples with fewer than 5000 reads were removed from analysis. Community State Types (CSTs) were assigned using VALENCIA, an algorithm based on similarity to the centroid of each CST as determined from a reference set of over 13,000 microbial profiles. The reference CSTs are based on relative abundances of all taxa and were determined by hierarchical clustering with Bray-Curtis dissimilarity and Ward linkage.(22) Five primary community state types (CSTs) were identified in this study; four dominated by Lactobacillus species (L. iners, L. crispatus, L. gasseri, L. jensenii), and one lacking significant numbers of Lactobacilli and characterized by higher proportions of strict and facultative anaerobic bacteria (termed CST IV). CST IV can be further broken down into 3 sub-states that generally reflect the dominance of Gardnerella vaginalis, Atopobium vaginae, and BVAB1. The sample size for this secondary data analysis was based on the 357 samples that had microbiota and Candida detection data available.

Statistical Analysis

Chi-square or fisher’s exact test, t-test, and logistic regression were used to determine the association between individual variables (CST, survey factors) and C. albicans detection. Factors collected from questionnaires that had been identified on the basis of previous literature and biologic plausibility were evaluated in analyses. It is possible that the vaginal microbiota, personal behaviors, and medication use are affected by one another, and so a multiple logistic regression model was used to evaluate the effect of each factor, taking into account the effect of other factors. Variance inflation factors (VIFs) were calculated to ensure variables in the final multivariable model were not collinear. Linear discriminant analysis (LDA) effect size (LEfSe) analysis was used to identify the vaginal bacterial taxa most likely to explain differences between samples with and without C. albicans detected.(23) Data were analyzed using SAS University Edition (SAS Institute, Inc., Cary, North Carolina).

Results

C. albicans was detected in 21% of vaginal samples (83/394). Despite the exclusion criterion of vaginal discharge at study entry, 24% reported vaginal discharge in the 60 days prior. Women who were positive for C. albicans were more likely to report use of tampons, engage in receptive oral sex, and use of over the counter (OTC) antifungals to self-treat for vaginal infections without first seeing a doctor for examination (Table 1). Neither self-reported diagnosis of VVC or bacterial vaginosis (BV) in the prior 60 days were associated with C. albicans detection. Additionally, detection was not associated with vaginal pH, reporting any vaginal symptom in the prior 60 days, age, history of pregnancy, use of hormonal contraception in the prior 60 days, number of recent sexual partners, or other forms of sexual activity in the prior 60 days.

Table 1:

Demographic and Behavioral Factors in Reproductive-Age Women With and Without C. albicans Detected

	C. albicans
Characteristic	Detected (n=83)	Not Detected (n=311)	p-valuea
Race			0.32
White or Caucasian	25 (30.1)	72 (23.1)
Black or African American	24 (28.9)	80 (25.7)
Asian	19 (22.9)	77 (24.8)
Hispanic	15 (18.1)	82 (26.4)
Sexual behaviors in past 60 days
Vaginal intercourse	66 (80.5)	227 (80.8)	0.95
Receptive oral sex	47 (58.0)	107 (39.0)	0.003
Receptive anal sex	5 (9.3)	25 (11.6)	0.62
Use sex toys	13 (17.6)	46 (17.1)	0.92
One or more partners	66 (79.5)	234 (79.0)	0.93
Contraception in past 60 days
Condom	25 (31.3)	96 (31.6)	0.96
IUD	4 (5.0)	15 (4.9)	0.99
Birth control pill	21 (26.3)	68 (22.4)	0.46
None	26 (32.1)	82 (27.0)	0.36
Vaginal infections in past 60 days
Bacterial vaginosis	3 (3.6)	10 (3.2)	0.74
Yeast infection	6 (7.2)	16 (5.1)	0.43
Vaginal symptoms in past 60 days	43 (51.8)	140 (45.0)	0.27
Tampon use	52 (62.6)	153 (49.2)	0.03
History of
Pregnancy	49 (59.0)	190 (61.7)	0.66
Douching	16 (19.5)	51 (16.6)	0.51
Lubricant use in past 60 days	13 (15.9)	49 (16.5)	0.88
Self-treatment with antifungals	25 (30.1)	45 (14.5)	0.001
Vaginal pH			0.49
≤ 4.5	46 (55.4)	159 (51.1)
> 4.5	37 (44.6)	152 (48.9)
Community State Type			0.36
IV, Low-Lactobacillus	17 (23.3)	80 (28.2)
I, L. crispatus-dominated	25 (34.2)	64 (22.5)
II, L. gasseri-dominated	4 (5.5)	17 (6.0)
III, L. iners-dominated	23 (31.5)	107 (37.7)
V, L. jensenii-dominated	4 (5.5)	16 (5.8)
Age (years)	30.3 (7.2)	30.7 (7.4)	0.68

Data are mean (standard deviation) for age and no. (%) for all other variables

^aBy chi-square or Fisher’s exact test for categorical variables and t-test for age

The heatmap in Figure 1 displays the relative abundance of the top 25 bacterial taxa associated with C. albicans-detected and -not detected samples. CST IV-A, IV-B, and IV-C were collapsed to increase sample size and because statistical modeling suggested these sub-groups were not different from each other with respect to C. albicans detection. Among samples with C. albicans detected, CST I (L. crispatus-dominated) was the most abundant CST; however, among samples without C. albicans detected, CST III (L. iners-dominated) was the most abundant CST, followed by CST IV (low-Lactobacillus) (Table 1).

Figure 1:

Heatmap displaying Candida albicans detection, vaginal bacterial taxa relative abundances, and community state types (CST) in 357 reproductive-age women. The red and grey bars indicate samples with and without C. albicans detected in the below heatmap, respectively. The top 25 most abundant taxa and relative abundance scales are indicated. CST I, II, III, and V represent bacterial communities dominated by L. crispatus, L. gasseri, L. iners, and L. jensenii, respectively. CST IV is characterized by a bacterial community with low relative abundance of Lactobacillus spp.

In multivariable modeling (Table 2), women with a L. crispatus-dominated CST had increased odds of C. albicans detection compared to a low-Lactobacillus CST IV after cofactor adjustment for receptive oral sex and history of self-treatment with OTC antifungals (aOR: 2.05, 95% CI: 0.97-4.37, n=323), a finding suggestive of a relationship but with borderline statistical significance. In the same model, antifungal use and receptive oral sex were significantly associated with C. albicans detection, with a dose response trend for increasing frequency of oral sex practice and escalating odds of detection for C. albicans (P for trend 0.03). An interaction term for race/ethnicity and CST was not significant in the multivariable model. Tampon use was not included in multivariable modeling as it was no longer significantly associated with C. albicans detection after controlling for frequency of receptive oral sex, antifungal use, and CST. Adjusting the model for any symptoms experienced in the prior 60 days did not meaningfully affect results.

Table 2:

Univariate and Multivariable Logistic Regression Models of Factors Associated With C. albicans Detection

Characteristic	ORa (95% CIb)	p-value	Adjustedc ORa (95 % CIb)	p-valued
Tampon use	1.73 (1.05-2.85)	0.03	-	-
Frequency of receptive oral sex				0.03
None	REF		REF
Less than once per week	1.74 (0.95-3.21)	0.07	1.42 (0.71-2.83)	0.32
Once or more per week	2.73 (1.48-5.03)	0.001	2.49 (1.27-4.88)	0.01
Community State Type				0.40
IV, Low-Lactobacillus	REF		REF
I, L. crispatus-dominated	1.84 (0.91-3.70)	0.09	2.05 (0.97-4.37)	0.06
II, L. gasseri-dominated	1.11 (0.33-3.71)	0.87	1.78 (0.50-6.37)	0.38
III, L. iners-dominated	1.01 (0.51-2.02)	0.97	1.25 (0.59-2.62)	0.56
V, L. jensenii-dominated	1.18 (0.35-3.96)	0.79	1.32 (0.37-4.72)	0.66
Self-treatment with antifungals	2.55 (1.45-4.48)	0.001	1.93 (1.01-3.68)	0.05

^aOdds Ratio

^bConfidence Interval

^cFull multivariable model

^dP-value for frequency of receptive oral sex is a test for trend

We also assessed individual bacteria in order to determine if there were taxa that were most likely to be associated with C. albicans detection (Figure 2). At this unadjusted taxonomic-level of modeling, LEfSe analysis identified L. crispatus, Lactobacillus coleohominis, Prevotella oris, and Prevotella veroralis as being in higher relative abundance when C. albicans was detected. Mobiluncus, Moryella, and Fusobacterium were identified as taxa in higher relative abundance when C. albicans was not detected.

Figure 2:

Linear discriminant analysis effect size: individual bacterial taxa relative abundance associated with C. albicans detection. Log(10) LDA scores are shown. Taxa associated with C. albicans being detected are shown on the right in green. Taxa associated with C. albicans not being detected are shown on the left in red.

Discussion

Vaginal C. albicans colonization is multifactorial and is influenced by both personal behaviors and a woman’s immune system.(24) Importantly, the vaginal microbiota as a whole, as well as individual bacterial taxa and microbial loads, may also play a role in the ability of C. albicans to colonize and persist in the vagina. C. albicans colonization is a public health concern as it may ultimately affect a woman’s risk for VVC or RVVC – infections which are highly prevalent and cause a great deal of discomfort and symptoms, particularly in pregnant women who have limited options for treatment.(10) In addition, studies have indicated that C. albicans may lead to increased risk for both HIV acquisition and transmission.(2, 3)

We utilized comprehensive behavioral questionnaires combined with vaginal microbiota composition and found several factors associated cross-sectionally with the molecular detection of C. albicans. We found the L. crispatus-dominated CST was associated with a 2-fold increase in the odds of C. albicans detection compared to a low-Lactobacillus state in adjusted analyses. While this finding was borderline in statistical significance (95% CI: 0.97-4.37), the LEfSe analysis also provides evidence that L. crispatus, regardless of whether it is a dominant species in a CST, is a phylotype which is statistically associated with increased C. albicans detection.

Our epidemiologic findings are supported by a number of in vitro studies that suggest a mechanistic explanation for our observed association between L. crispatus and C. albicans detection. First, BV-associated bacteria are more sensitive to lactic acid than acetic acid,(25) but C. albicans seems to be more sensitive to acetic acid than lactic acid.(26) Importantly, BV-associated bacteria produce acetic acid rather than lactic acid,(27) which offers a potential explanation for why C. albicans is negatively associated with BV and, in the LEfSe analysis, was negatively associated with Mobiluncus— a taxa commonly found in BV.(17) On the other hand, vaginal Lactobacillus spp. produce acetic acid as well as lactic acid in ratios that vary among strains.(28) Findings from Bai et al. show that L. crispatus strains have a higher lactic:acetic acid ratio than do L. iners strains,(29) thus supporting the association we identified between L. crispatus with C. albicans detection. Consequently, other CSTs (and especially CST IV) could be less hospitable to C. albicans. Second, the higher levels of lactic acid produced by L. crispatus may contribute to the ability of C. albicans to persist in the vagina by aiding in immune system evasion. When lactate is a dominant carbon source, C. albicans is able to mask β-glucan—an important pathogen-associated molecular pattern—resulting in a reduced recognition, reduced uptake, and reduced fungal-killing by macrophages which have the potential to clear C. albicans cells from the mucosal surface.(30, 31) In contrast, some in vitro and in vivo models suggest L. crispatus may ultimately protect against VVC by inhibiting C. albicans growth and hyphal formation (32) – important contributors in the transition of C. albicans from a commensal to pathogenic organism. Lastly, it is possible that L. crispatus has no role in the detection of C. albicans, and that prior findings represent spurious associations.

Other observational epidemiologic studies have also found associations between Lactobacillus identified by culture and either C. albicans as detected by culture, wet mount, or clinical diagnosis of VVC; however, most have not used molecular methods to detect and quantify Lactobacillus or Candida spp. In a prospective study of Kenyan sex workers, McClelland et al. found that presence of Lactobacillus in culture was associated with a non-significant increase in the odds of asymptomatic and symptomatic VVC defined by clinical exam and wet mount. After excluding women with concurrent BV in order to remove another potential cause of vaginal symptoms, McClelland et al. further reported that concurrent Lactobacillus was associated with an almost 4-fold increase in the odds of symptomatic VVC.(15) In another cohort of asymptomatic, sexually active, non-pregnant women, Beigi et al. detailed the presence of Lactobacillus in culture was also associated with increased odds for vaginal yeast detected on culture.(8) In one study utilizing 16S rRNA gene amplicon sequencing, VVC-only samples could not be described by a single vaginal microbiota profile, although mixed BV-VVC samples tended to have a higher relative abundance of Lactobacillus (genus-level) compared to the BV-only group.(33)

In this study, we found two primary factors from the behavioral surveys that were associated with C. albicans detection—self report of OTC antifungal use and receptive oral sex. Antifungal use may reflect Candida carriage as antifungal treatment does not always successfully eradicate Candida, even in the absence of antifungal resistance.(4) While some women remain asymptomatic during periods of Candida colonization, women with RVVC could be susceptible to additional episodes of VVC due to previously mentioned host factors. This is supported by high rates of recurrence with identical strains of Candida following cessation of long-term suppressive therapy in women with RVVC.(4) However, a self-report of OTC antifungal use may not always reflect history of VVC. Ferris et al. reported that only a fraction (34%) of women intending to self-treat with OTC antifungals had a diagnosis of just VVC upon examination, while most participants with clinical findings had either BV (19%), trichomoniasis (2%) or mixed infections (21%).(34) In addition, a recent study reported that physician-treated VVC achieved greater symptom relief as compared to self-treatment with OTC antifungals.(6)

Receptive oral sex appears to be a logical risk factor for C. albicans detection in the vagina as C. albicans is often found in the oral microbiota.(35) Masturbation with saliva and receptive oral sex increased the risk of VVC recurrence in a prospective cohort study of sexually-active women and their male partners.(14) Of note, we found a statistically significant dose response trend between increasing frequency of receptive oral sex and C. albicans detection, contributing evidence that it is likely a causal factor. The LEfSe analysis identified several species of Prevotella, which have also been found in oral samples,(36) as associated with C. albicans detection.

In regards to other behavioral risk factors for VVC, results from prior studies are mixed.(5) Some studies report that C. albicans has the ability to form biofilms on intrauterine devices (IUDs) in vitro, although a recent paper found no evidence of biofilms containing Candida microorganisms in women with VVC. (5, 37) While there is evidence for the estrogen-dependence of C. albicans,(4) results for the association between oral contraceptive pills and VVC have been inconsistent.(38) We found no association between recent use of oral contraceptive pills, condoms, or IUDs with C. albicans detection. In addition, while some studies report an association between other sexual behaviors and risk for VVC,(12, 39) we found no evidence that frequency of vaginal intercourse or receptive anal sex was associated with the molecular detection of C. albicans.

There are limitations to our study. First, cross-sectional data cannot be used to infer causal relationships, and the vaginal microbiota vary over time. Gajer and Brotman et al. have described temporal patterns in the vaginal microbiota which can be clustered into community classes (CC), and a single sample (CST) may reflect a general pattern of fluctuating in that domain (CC).(40) Second, C. albicans detection through PCR may reflect both active colonization, the temporary presence of C. albicans, or the presence of dead organisms. Lastly, this study relied on self-reported data and also excluded women who reported vaginal discharge at baseline. We did note that 24% of women reported vaginal discharge in the 60 days prior to study entry. Still, it is possible that the women with C. albicans detected could represent both asymptomatic detection and symptomatic VVC as discharge is not the only symptom of VVC.(6) Estimates of C. albicans detection in this study are consistent with other estimates of the prevalence of asymptomatic Candida colonization in other large population-based studies,(8, 9) and adjusting for reporting any symptom in the prior 60 days, which includes symptoms of VVC such as itching and irritation, did not alter the findings.

The strengths of this study are it is a large sample size with comprehensively detailed vaginal microbiota and molecularly detected C. albicans. PCR, which was employed in this study, is more sensitive than culture and DNA probe for the detection of C. albicans.(41, 42) While Candida PCR was undertaken 2 years after DNA extraction, it is unlikely that degradation would impact detection due to the small size of the target and stability of DNA.(43) This study is confirmative of prior work based on culture which suggested concurrent vaginal Lactobacillus colonization was associated with increased vaginal C. albicans colonization;(8) however, we additionally identified L. crispatus specifically as being associated with detection. Further studies should evaluate the roles the vaginal microbiota and host immune response play in the transition from asymptomatic colonization to symptomatic VVC in longitudinally collected samples, as well as the impact of duration of C. albicans colonization on VVC risk.