Abstract
Background. We sought to describe the temporal relationship between vaginal microbiota and human papillomavirus (HPV) detection.
Methods. Thirty-two reproductive-age women self-collected midvaginal swabs twice weekly for 16 weeks (937 samples). Vaginal bacterial communities were characterized by pyrosequencing of barcoded 16S rRNA genes and clustered into 6 community state types (CSTs). Each swab was tested for 37 HPV types. The effects of CSTs on the rate of transition between HPV-negative and HPV-positive states were assessed using continuous-time Markov models.
Results. Participants had an average of 29 samples, with HPV point prevalence between 58%–77%. CST was associated with changes in HPV status (P < .001). Lactobacillus gasseri–dominated CSTs had the fastest HPV remission rate, and a low Lactobacillus community with high proportions of the genera Atopobium (CST IV-B) had the slowest rate compared to L. crispatus–dominated CSTs (adjusted transition rate ratio [aTRR], 4.43, 95% confidence interval [CI], 1.11–17.7; aTRR, 0.33, 95% CI, .12–1.19, respectively). The rate ratio of incident HPV for low Lactobacillus CST IV-A was 1.86 (95% CI, .52–6.74).
Conclusions. Vaginal microbiota dominated by L. gasseri was associated with increased clearance of detectable HPV. Frequent longitudinal sampling is necessary for evaluation of the association between HPV detection and dynamic microbiota.
Keywords: 16S rRNA gene analysis, human papillomavirus (HPV), transition model, vaginal microbiota
Persistent detection of human papillomavirus (HPV) is associated with increased risk for dysplasia and cervical malignancy but the factors affecting HPV persistence are not fully understood [1]. There is mounting evidence that the bacterial communities that inhabit the vagina serve as an important line of defense against pathogens, including sexually transmitted infections (STIs) [2]. Vaginal Lactobacillus spp. are known to provide broad-spectrum protection through their production of lactic acid [3], bacteriocins (bactericidal proteinaceous molecules) [4], antagonistic bacteriocin-like substances [5], and biosurfactants [6], and in their ability to adhere to mucus, which form barriers against pathogens [7] and disrupt biofilms [8]. Further, Motevaseli et al recently demonstrated that vaginal lactobacilli exert cytotoxic effects on cervical tumor cells in vitro, independent of pH and lactic acid [9].
Numerous epidemiologic studies have suggested an association between cervical HPV infections and bacterial vaginosis (BV)–associated vaginal microbiota [10–16]. BV is a condition which is prevalent in approximately 29% of U.S. women [17] and is characterized by decrease in Lactobacillus spp. with concomitant increase in diversity and numbers of anaerobic bacteria, including species of Gardnerella, Prevotella, and Clostridiales [18]. A meta-analysis based on mostly cross-sectional studies demonstrated a positive association between BV and cervical HPV [10]. Two large longitudinal studies showed that BV was significantly associated with incident cervical HPV detection [11, 12]. A few studies have also suggested delayed clearance of HPV when BV was prevalent [12, 13]. The aforementioned studies determined BV and vaginal microbiota status based on Gram stain analysis of vaginal smears, which relies on the subjective morphologic assessment by a microscopist, Amsel's clinical criteria, or clue cell findings on Papanicolaou-stained slides. Just 1 recent study utilized 16S rRNA gene analysis to evaluate the association between vaginal bacterial community composition and HPV prevalence in a cross-sectional study of Korean women [15]. The authors found that participants who were HPV positive had a lower proportion of Lactobacillus spp. than HPV-negative women. In addition, the authors identified Fusobacteria, including Sneathia spp. as key bacteria associated with HPV detection.
We have previously reported high variability in the short-term temporal dynamics of both vaginal bacterial communities [19] and HPV detection [20]. To better understand the possible mutual influence of vaginal microbiota and HPV variability, we describe in this study the temporal relationship between vaginal microbiota characterized by 16S rRNA gene analysis and transitions between HPV positivity and HPV negativity within a woman over a 16-week period.
METHODS
Study Population and Specimen Collection
We utilized archived samples from a prior longitudinal study of reproductive-age women conducted in Baltimore, MD. The study design and participant population have been previously described [19, 21]. In brief, 39 reproductive-age women were enrolled in a vaginal douching cessation study and self-collected midvaginal swabs twice weekly for 16 weeks (n = 32 women with complete microbiota profiles, 937 samples with microbiome data, and 930 samples with HPV results). The parent study was designed as a douching cessation study in which participants were instructed to continue with their usual vaginal douching practices for the first 4 weeks and then to cease all douching practices in the following 12 weeks [21]. The majority of women (71%) reported using a vaginal douche product 1 or 2 times during that first month-long observation period [21]. Participants reported douching, other feminine hygiene practices, and sexual behaviors on daily diaries and demographic information on questionnaires at clinic-based visits. Swabs and diaries were submitted to the study coordinator weekly, and the swabs were stored in freezer archive at −80°C.
Participant identification numbers (ID#s) reported in the figures were double-blinded prior to data release and the new ID#s were assigned based on the dominant community state type (CST) in each woman's series of samples [19]. The bacterial sequence data are available in the small read archive (SRA) (accession number: SRA026073), and the metadata associated with the sequence data are available in dbGap (dbGap study number: phs000261) [19].
Ethical approval was obtained from the Institutional Review Boards of the Johns Hopkins University School of Medicine and the University of Maryland School of Medicine.
Composition of Vaginal Bacterial Communities and HPV Detection
DNA extraction from vaginal swabs, polymerase chain reaction amplification of the V1-V2 hypervariable regions of 16S rRNA genes using primers barcoded 27F and 338R [22], and pyrosequencing were previously published [19]. DNA extracted from the swabs were tested for 37 types of HPV DNA using the Roche Linear Array HPV Genotyping Test according to the manufacturer's instructions. HPV positive was defined as any HPV detected. High-risk HPV (hrHPV) was defined by positivity for any of HPV types 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, or 68.
Statistical Analysis
The effects of time-dependent covariates on the rate of transition between HPV-negative to HPV-positive state (and vice versa) were assessed using continuous-time multistate Markov models fit using maximum likelihood [23]. The models were proportional hazards models of times to transition of HPV state conditioned on the current HPV state. Specifically, CST was modeled as a time-varying exposure of HPV status transition at later times (where subsequent HPV status was tested as early as 3 days after CST). This approach ensured that the exposure assessment occurred before the outcome assessment.
The multistate model simultaneously assessed transitions from the HPV-negative to HPV-positive states and transitions from HPV-positive to HPV-negative states. When participants were HPV negative, they were at risk to transition to HPV positive. When participants were HPV positive, they were at risk to transition to HPV negative. Therefore, each participant can contribute multiple transitions to the analysis. If many participants do not transition, this will be reflected by low estimated transition rates, and long average sojourns (length of time) in a given state. Therefore, a woman who is persistently HPV negative will be at risk for becoming HPV positive for a long duration, whereas a woman who is persistently HPV positive will be at risk for becoming HPV negative for a long duration. For both HPV-positive and -negative states, the outcome is time-to-transition, and the measures estimated are transition rates, which are hazard rates for the event of transitioning. The measure of association is the transition rate ratio, analogous to the hazard rate ratio. All P values are reported unadjusted for multiple comparisons [24]. Model fit was assessed using a Pearson-type goodness of fit test [25]. All analyses were performed using the msm package in R version 3.0.2 [26].
Covariates
Time-varying exposures such as sexual behaviors were assessed from the daily diaries 1 or 2 days prior to the sample defining the HPV outcome. Hormonal contraception was assessed at the baseline visit, and changes in contraception were queried at follow-up visits (no changes reported). A phase variable reflecting douching observation and douching cessation phases of the parent study in accordance with the original study design (see above) [21] were also included. Potential confounders were assessed in the model prior to HPV sampling to ensure that they were confounders, not mediators. These models account for the time spent in each HPV state and the rate of transitioning to and from HPV positivity. Covariates were assumed to have a log-linear relationship with transition rates. Models were adjusted for age categories, hormonal contraceptive use, vaginal sex (with or without condom), and menses using normalized menstrual cycle times. The length of menstrual cycles within and between women was normalized to 28 days as previously described [19] by rescaling the intervals from the first day a woman reported bleeding to the first day of the next menstrual cycle. Lubricant use, vaginal douching, and use of thong underwear were assessed and excluded from the model because they were colinear with the included sexual behavior variables (vaginal sex and condom use).
RESULTS
The average age of participants was 37 years (SD, 8.1; range, 22–53). Fifty percent of women (n = 16) self-reported their race to be African American and 40% (n = 13) reported white. Eighty-four percent (n = 27) of participants reported being in a monogamous heterosexual relationship. None of the women reported same-sex relationships. When evaluating the differences between women who were persistently HPV negative for 16 weeks (n = 5) compared to women who had at least 1 HPV-positive sample during the study observation (n = 27), there were no statistically significant differences in demographic factors (Table 1). Age was of borderline statistical significance in its association with observed HPV positivity over the 16-week interval, with all 13 women age 40 or older having HPV detected.
Table 1.
Comparison of Demographics Between HPV-Positive and HPV-Negative Women Enrolled in a 16-Week Prospective Study With Twice-Weekly Sampling, Baltimore, MD, 2005–2007, n = 32
| HPV Negative During Study Observation (n = 5) |
HPV Positive During Study Observation (n = 27) |
P Value* | |||
|---|---|---|---|---|---|
| n | % | n | % | ||
| Age (years) | .13 | ||||
| 18–29 | 1 | 20.00 | 5 | 18.52 | |
| 30–39 | 4 | 80.00 | 9 | 33.33 | |
| ≥40 | 0 | 0.00 | 13 | 48.15 | |
| Race/ethnicity | .78 | ||||
| Black | 2 | 40.00 | 14 | 51.85 | |
| White | 3 | 60.00 | 10 | 37.04 | |
| Hispanic | 0 | 0.00 | 1 | 3.70 | |
| Other | 0 | 0.00 | 2 | 7.41 | |
| Educational attainment | .22 | ||||
| Less than high school | 0 | 0.00 | 4 | 14.81 | |
| High school graduate (or GED) | 1 | 20.00 | 14 | 51.85 | |
| More than high school | 4 | 80.00 | 9 | 33.33 | |
| Hormonal contraception | .32 | ||||
| None | 3 | 60.00 | 21 | 77.78 | |
| Hormonal contraceptive | 1 | 20.00 | 5 | 18.52 | |
| IUD | 1 | 20.00 | 1 | 3.70 | |
| Monogamous relationship | 5 | 100.00 | 22 | 81.48 | .56 |
| Baseline vaginal douchinga | .32 | ||||
| 1–2 times per month | 3 | 60.00 | 19 | 70.37 | |
| 3–4 times per month | 0 | 0.00 | 5 | 18.52 | |
| ≥5 times per month | 2 | 40.00 | 3 | 11.11 | |
| Ever smoked | 1 | 20.00 | 14 | 51.85 | .34 |
| Alcohol use | .66 | ||||
| Never | 1 | 20.00 | 9 | 33.33 | |
| 1–4 times per month | 4 | 80.00 | 14 | 51.85 | |
| 1–4 d per week | 0 | 0.00 | 4 | 14.81 | |
| History of genital herpes or ulcers | 1 | 20.00 | 1 | 3.85 | .30 |
| Lifetime number of sexual partners | .64 | ||||
| 1–5 | 3 | 60.00 | 11 | 42.31 | |
| ≥6 | 2 | 40.00 | 15 | 57.69 | |
| Community class, dominant bacterial community over the 16 wks observationb | .23 | ||||
| DA, dominated by CST IV-A | 0 | 0.00 | 3 | 11.11 | |
| DB, dominated by CST IV-B | 1 | 20.00 | 8 | 29.63 | |
| LC, dominated by L. crispatus | 2 | 40.00 | 3 | 11.11 | |
| LG, dominated by L. gasseri | 1 | 20.00 | 1 | 3.70 | |
| LI, dominated by L. iners | 1 | 20.00 | 12 | 44.44 | |
Abbreviations: CST, community state type; CST IV-A, low Lactobacillus community; CST-IVB, low Lactobacillus community with high proportions of the genera Atopobium; GED, General Educational Development; HPV, human papillomavirus; IUD, intrauterine device.
a Because the parent study was a vaginal douching intervention study, women were instructed to continue their usual feminine hygiene practices in the first 4 weeks of observation and were followed for 12 additional weeks of douching cessation.
b Community class was defined by Gajer and Brotman et al [19] by longitudinal clustering of CST profiles over 16 weeks of observation.
* P values were determined by Fisher's exact test.
Among the 32 women participating in the study, 5 were always HPV negative throughout the 16-week observation (Table 1), 2 were positive for 1 HPV subtype during the follow-up period, and 25 were positive for 2 or more HPV subtypes during the follow-up period. The maximum number of subtypes for which a woman was positive over the whole follow-up period was 13; the median (interquartile range [IQR]) number was 3.5 (2–5). At any one observation, the maximum number of subtypes for which a woman was positive was 11; the median (IQR) number was 1 (0–2). The most commonly detected subtypes were HPV 62, 54, and 45. All subtypes were detected at least once, except for 26, 40, 58, and 64.
In prior work utilizing samples from this study [19], our research group demonstrated that the vaginal microbiota when characterized by 16S rRNA gene sequence analysis clustered into 6 CSTs: 4 dominated by either L. crispatus (termed CST I), L. gasseri (CST II), L. iners (CST III), or L. jensenni (CST V), and 2 that lacked substantial numbers of Lactobacillus spp. (CST IV-A and IV-B). CST IV-A comprised a variety of anaerobic bacteria (including Streptococcus and Prevotella), whereas CST IV-B harbored higher proportions of the genera Atopobium, Gardnerella, and Prevotella among others. While herein we identify each CST by the dominant bacteria for ease of notation, it is important to note that each CST represents a community of bacteria that is clustered together based on diversity and relative abundance of all bacterial phylotypes detected. Figure 1 displays the bacteria that cluster with each CST and the proportion of HPV-positive samples within each CST. CSTs with the greatest proportion of HPV-positive samples were the low-Lactobacillus CST IV groups (71%) and the L. iners–dominated CST III (72%), compared to 45% in L. crispatus–dominated CST I and 12% in L. gasseri–dominated CST II. The number of hrHPV samples within a CST for each woman is shown in Figure 2 and also depicts the greater proportion of hrHPV-positive samples among women in CST III and CST IV-B.
Figure 1.
Heatmap of relative percent abundance of bacterial taxa found in the vaginal samples of 32 reproductive-age women who contributed 930 samples in a longitudinal study. Each vertical line represents 1 sample. Bacterial community state type (CST) was described by Gajer and Brotman et al [19], and relative abundance is indicated by the key. HPV positivity is also shown across the top of the figure. Abbreviations: CST, community state type; HPV, human papillomavirus.
Figure 2.
Number of hrHPV-positive samples per participant and stratified by CST. During the longitudinal study, each participant collected up to 33 samples and could have samples in numerous CSTs [19]. This figure displays the number of hrHPV-positive and -negative samples by each CST. ID#s were double blinded, and reassignment was based on the dominant CST observed in each woman's series of samples. Abbreviations: CST, community state type; hrHPV, high-risk human papillomavirus; ID#s, identification numbers.
The longitudinal samples, which were self-collected every 3–4 days, revealed tremendous variability in HPV-positive samples detected and CST dynamics [19] (Figure 3). Participants had an average of 29 (range, 25–33) samples tested for HPV, with point prevalence ranging from 58% to 77% and 16-week period prevalence of 84%. In a model of the rate of transition between HPV-negative and HPV-positive states, in which CST was evaluated at the sample just prior to the sample tested for HPV, CST was significantly associated with remission of HPV (P = .008), but not with new detection of HPV (P = .10). CST IV-A may be associated with higher transition rates from HPV-negative to HPV-positive state compared to CST I (L. crispatus–dominated), but the finding was not statistically significant (adjusted transition rate ratio [aTRR], 1.86, 95% confidence interval [CI], .52–6.74; P value, .34) (Table 2). CST II (L. gasseri–dominated) was associated with the fastest HPV remission rate (HPV positive to no detection) and CST IV-B was associated with the slowest HPV remission rate compared to the CST I (aTRR, 4.43, 95% CI, 1.11–17.70 and aTRR, 0.33, 95% CI, .12–1.19, respectively) (Table 3).
Figure 3.
Dynamics of vaginal CSTs and HPV detection in 32 women over 16 weeks. Double-blinded participant ID#s are indicated on the right. Each symbol represents 1 sample in the time series. The absence of a symbol indicates a missing HPV sample. The CST in the time series for each woman are color coded and the HPV status by symbol according to the key above. Participants are clustered by bacterial community class, which groups together women of similar bacterial community patterns over time [19]. Abbreviations: CSTs, community state types; HPV, human papillomavirus; ID#s, identification numbers.
Table 2.
Effects of Vaginal Microbiota on the Rate of Transition Between HPV-Negative to HPV-Positive States Assessed Using Continuous-Time Markov Models
| TRR | P Value | 95% CI |
aTRRa | P Value | 95% CI |
|||
|---|---|---|---|---|---|---|---|---|
| Vaginal bacterial CST 3 d prior to HPV sampling, dominant bacteria | ||||||||
| CST I, L. crispatus | REF | … | … | … | REF | … | … | … |
| CST II, L. gasseri | 1.13 | .85 | .31 | 4.15 | 0.34 | .21 | .06 | 1.85 |
| CST III, L. iners | 1.94 | .12 | .85 | 4.46 | 1.79 | .22 | .71 | 4.51 |
| CST IV-A, low Lactobacillus | 4 | .007 | 1.46 | 11.01 | 1.86 | .34 | .52 | 6.74 |
| CST IV-B, low Lactobacillus | 1.06 | .92 | .37 | 3.04 | 0.76 | .62 | .26 | 2.24 |
| Ethnicity/race | ||||||||
| White | REF | … | … | … | REF | … | … | … |
| African American | 1.06 | .85 | .57 | 1.98 | … | … | … | … |
| Hispanic or other | 11.16 | .002 | 2.49 | 49.94 | … | … | … | … |
| Age | ||||||||
| <30 | REF | … | … | … | REF | … | … | … |
| 30–39 | 2.45 | .08 | .91 | 6.58 | 0.96 | .98 | .02 | 43.7 |
| ≥40 | 2.27 | .11 | .83 | 6.25 | 3.36 | .55 | .06 | 173.11 |
| Hormonal contraception | ||||||||
| None | REF | … | … | … | REF | … | … | … |
| Hormonal contraceptive use | 0.27 | .09 | .06 | 1.24 | 0.68 | .87 | .01 | 67.3 |
| IUD | 0.4 | .16 | .11 | 1.45 | 1 | .99 | .1 | 10.57 |
| Study phase | ||||||||
| Nondouching phase | REF | … | … | … | REF | … | … | … |
| Douching intervention phase | 0.71 | .33 | .35 | 1.41 | 0.74 | .5 | .32 | 1.75 |
| Time-varying factors 1 day prior to HPV sampling | ||||||||
| Vaginal intercourse | ||||||||
| None | REF | … | … | … | REF | … | … | … |
| With condom | 0.18 | .26 | .01 | 3.51 | 0.35 | .32 | .04 | 2.74 |
| No condom | 0.83 | .65 | .37 | 1.87 | 1.37 | .52 | .52 | 3.56 |
| Lubricant use | 0.002 | .62 | <<.001 | >>999 | … | … | … | … |
Abbreviations: aTRR, adjusted transition rate ratio; CI, confidence interval; CST, community state type; HPV, human papillomavirus; IUD, intrauterine device; REF, reference; TRR, transition rate ratio.
a The model also adjusted for normalized menstrual cycle time.
Table 3.
Effects of Vaginal Microbiota on the Rate of Transition Between HPV-Positive to HPV-Negative States Assessed Using Continuous-Time Markov Models
| TRR | P Value | 95% CI |
aTRRa | P Value | 95% CI |
|||
|---|---|---|---|---|---|---|---|---|
| Vaginal bacterial CST 3 d prior to HPV sampling, dominant bacteria | ||||||||
| CST I, L. crispatus | REF | … | … | … | REF | … | … | … |
| CST II, L. gasseri | 5.19 | .01 | 1.45 | 18.5 | 4.43 | .03 | 1.11 | 17.7 |
| CST III, L. iners | 0.55 | .13 | .25 | 1.19 | 0.79 | .65 | .28 | 2.2 |
| CST IV-A, low Lactobacillus | 0.98 | .97 | .33 | 2.86 | 2.38 | .17 | .69 | 8.17 |
| CST IV-B, low Lactobacillus | 0.3 | .01 | .12 | .77 | 0.33 | .09 | .12 | 1.19 |
| Ethnicity/race | ||||||||
| White | REF | … | … | … | … | … | … | … |
| African American | 1.14 | .68 | .61 | 2.12 | … | … | … | … |
| Hispanic or other | 1.35 | .68 | .32 | 5.72 | … | … | … | … |
| Age | ||||||||
| <30 | REF | … | … | … | REF | … | … | … |
| 30–39 | 0.88 | .8 | .32 | 2.39 | 1.74 | .36 | .53 | 5.69 |
| ≥40 | 1.22 | .69 | .45 | 3.28 | 1.57 | .5 | .42 | 5.93 |
| Hormonal contraception | ||||||||
| None | REF | … | … | … | REF | … | … | … |
| Hormonal contraceptive use | 0.07 | 0 | .01 | .38 | 0.19 | .01 | .05 | 0.7 |
| IUD | 0.1 | .17 | .003 | 2.71 | 0.37 | .45 | .03 | 4.89 |
| Study phase | ||||||||
| Nondouching phase | REF | … | … | … | REF | … | … | … |
| Douching intervention phase | 0.8 | 0.55 | 0.39 | 1.65 | 0.64 | .27 | .3 | 1.41 |
| Time-varying factors 1 day prior to HPV sampling | ||||||||
| Vaginal intercourse | ||||||||
| None | REF | … | … | … | REF | … | … | … |
| With condom | <<0.001 | 0.97 | <<0.001 | >>999 | 0.57 | .66 | .05 | 6.83 |
| No condom | 0.94 | 0.86 | 0.48 | 1.85 | 1.11 | .79 | .52 | 2.35 |
| Lubricant use | 0.005 | 0.005 | <<0.001 | >>999 | … | … | … | … |
Abbreviations: aTRR, adjusted transition rate ratio; CI, confidence interval; CST, community state type; HPV, human papillomavirus; IUD, intrauterine device; REF, reference; TRR, transition rate ratio.
a The model also adjusted for normalized menstrual cycle time.
Although there were only 21 observations of vaginal lubricant use in the longitudinal study, there were no observations where lubricant use was accompanied by HPV onset. Other behavior variables, including vaginal douching and specific sexual behaviors, could not be evaluated because of correlation with the variable for vaginal sex. Of interest, the multivariable model also showed that hormonal contraceptive (HC) use was associated with slower clearance rate of detectable HPV compared to women reporting no hormonal contraception (aTRR, 0.19, 95% CI, .05–.70) (Table 3).
Based on longitudinal clustering of CST profiles over 16 weeks of observation [19], the women who were persistently HPV negative (Figures 2 and 3) for 16 weeks were most often in CST IV-B (1 woman), L. crispatus–dominated CST I (2 women), L. gasseri–dominated CST II (1 woman), and L. iners–dominated CST III (1 woman). Eleven women were persistently HPV positive (4 participants were most often in CST IV-B, 1 in L. crispatus–dominated CST I, and 6 in L. iners–dominated CST III).
DISCUSSION
In this longitudinal study of twice-weekly vaginal sampling, we previously found that HPV detection was extremely variable over very short intervals [20], and in the current analysis, we found women with vaginal microbiota dominated by L. gasseri (CST II) had a faster clearance rate of detectable HPV. Samples that exhibited a low relative abundance of Lactobacillus spp. (as in CST IV) or that were dominated by L. iners (CST III) had the greatest relative proportion of HPV-positive samples, suggesting that these 2 states may represent an environment that increases the risk of HPV acquisition and/or persistence. Our data are also suggestive that CST IV-A may also be associated with increased risk to transitioning to a HPV-positive state. Numerous longitudinal studies have documented both fluctuations in HPV detection [20, 27–31] and remarkable dynamics in vaginal bacterial communities [19, 32, 33]. This work underscores the importance for prospective and frequent sampling in order to capture the complex temporal changes of the vaginal ecosystem and to evaluate the factors that are associated with changes in detectability of HPV.
The mechanisms by which the vaginal microbiota might influence HPV detection are not known. Lactobacillus spp. have been shown to produce several microbicidal factors, including lactic acid [3]. Lactic acid is more effective than acidity alone in controlling overgrowth of bacteria and preventing STIs [9, 34, 35]. Our results showing a higher detectability of HPV in women with a low relative abundance of Lactobacillus spp. are consistent with observations from a large cohort study that reported higher HPV detection, risk of multiple infections, low-grade intraepithelial lesions, and Chlamydia trachomatis infection among women with elevated pH [16]. Our data on L. iners are similar to other investigators who have found that L. iners is unusual among lactobacilli in that it may not be as protective as other Lactobacillus-dominated microbiotas, and some strains are associated with instability of vaginal microbiota [19, 36]. Future studies should focus on elucidating how vaginal microbiota might affect the epithelium and immune microenvironment of the lower genital tract and the mechanisms governing these associations.
Our multivariable model also indicated that HC use was associated with reduced clearance of HPV, suggesting a possible influence of female sex hormones on host response to HPV infection. Prior work has investigated the cross-sectional association between HC and prevalent HPV [37] with the general assessment that there is no association. In contrast, Ghanem et al found an intriguing association between HC and a type-specific HPV-16 detection [38], and Marks et al reported an increased risk of HPV persistence in a longitudinal study of combined low-dose oral contraceptives users [39]. Because HC is thought to decrease the risk of BV [40], future studies may more closely examine the complex temporal relationships between the vaginal microbiota, sex hormones, and HPV infection.
Although the data were too sparse to model, it is of note that there were no observations where lubricant use was accompanied by HPV onset or remission. Wounding or microtrauma during intercourse could be a risk factor for both new HPV infection and reactivation so it is plausible that vaginal lubricant might lower the risk for such trauma and could provide protection. However, a few studies have begun to demonstrate that lubricant use may be associated with increased BV [21] and STI risk [41, 42], potentially by altering the composition of the vaginal microbiota, although there is little in vivo data available.
Licensure of the HPV vaccines (Gardasil and Cervarix) has put a great deal of focus in the field of cervical cancer prevention on vaccination as a primary method for reducing cervical cancer incidence and preventing disease. Despite these exciting scientific breakthroughs, many low-resource settings will be unable to benefit from the vaccine due to economic and logistic barriers. Consequently, it remains important to identify alternative, cost-effective, evidence-based intervention strategies that can be implemented to reduce cervical cancer incidence. Improving our understanding of the vaginal microbiota offers opportunities for maximizing a woman's first line of defense and to develop low-cost and practical interventions to reduce susceptibility to HPV.
To our knowledge, this is the first report of the longitudinal dynamics of vaginal microbiota and HPV detection. Strengths of this study include frequent sampling for a long duration (twice weekly for 16 weeks) in combination with high-throughput genomics to characterize the dynamics of the vaginal microbiota and its relation to HPV detection. Weaknesses of the study include self-sampling of midvaginal swabs instead of clinician-collected cervical specimens; however, several studies have demonstrated that vaginal swabs for HPV detection approximate well those of clinically obtained specimens from the cervix [27, 43, 44]. Investigators have also shown that self-collected vaginal swabs were as appropriate as physician-collected swabs of the midvagina for microbiota molecular characterization [45]. The small sample size limits some of our inferences, although numerous findings did reach statistical significance. It will be important to replicate these results in larger studies with similar frequency of sampling. Because of the already complex nature of this analysis, we elected to analyze the HPV data as a binary outcome instead of a quantitative value. Future studies will expand this work by evaluating the association between the microbiota and risk of individual and multiple infections, high-risk oncogenic HPV types, HPV strains associated with genital warts, and changes in HPV viral load, and including other important factors such as smoking [46] and HC type. While future studies designed to look more carefully at the complex temporal relationships between vaginal CSTs and variable detection of multiple types of HPV infections [20] will be critical for a better understanding of the natural history of HPV, the analysis of CST relationships to transitioning between a completely negative to any HPV-positive state presented here is useful in addressing a broader question of whether probiotic therapies may have a global impact on reducing HPV risk. Based on concerns that a HPV-positive sample may reflect deposition by sex partner rather than a true infection, we included a variable in our model that controlled for vaginal sex with and without a condom, and although not statistically significant, condom use did show greater protection. Next, while it is theoretically possible that the microbiota and associated inflammatory milieu and vaginal proteome may modify the efficiency of HPV amplification in the linear array assay, it is unlikely that this effect would be pathogen specific. In addition, CSTs were obtained using a clustering algorithm that takes into account the relative abundances of all phylotypes detected in a sample [19]. While we found statistically significant findings using the CSTs as defined, these CSTs could certainly be split out further, creating new and novel CSTs. Applying categories to a sample may result in residual confounding, but such categories are necessary for modeling and the practice of epidemiology. It should also be noted that the CSTs are based on relative abundance of bacteria and not actual concentrations of bacterial species. Lastly, it is unknown whether the rapid shifts in HPV detection represent frequent infection/clearance/reinfection events or variable control of persistent infection at levels below the limit of detection of our assay [20]. If the transitions measured here are representative of temporary loss and gain of immunological control of HPV infection, understanding the micro-environmental factors associated with these transitions may lead to important new interventions to reduce clinically relevant HPV persistence (ie, long-term detectable HPV infection).
We hypothesize that rational selection of probiotics—a personalized approach that selects the most appropriate probiotic strains for a woman given her microbiome's patterns—would be most effective for women's health and protection from STIs. In 1 pilot study of 54 women with HPV-positive low-grade squamous intraepithelial lesions (the cytological manifestation of productive HPV infection), daily consumption of a probiotic drink containing L. casei Shirota appeared to promote the clearance of HPV-cytologic abnormalities but did not result in a statistically significant reduction of HPV detection [47]. Studies have shown that probiotics can colonize the vagina following oral intake [48], and there is an association between rectal bacteria and the vaginal bacteria [49, 50]. It will be interesting to determine in future studies whether a probiotic intervention can reduce productive viral infection and drive HPV into a clinically “latent” state.
In summary, prior work has indicated differences in HPV prevalence between single-time and multiple-time-point HPV detection assays [27–31], indicating a fluctuation in the detection of HPV. The vaginal microbiome may play a significant role in the control of latent HPV infection, defense against incident HPV acquisition, or increased clearance of HPV, and future work should investigate whether probiotic and/or prebiotic approaches reduce HPV burden. Such studies could reveal the mechanisms by which the vaginal microbiota, as a community or through the actions of specific bacteria, provides protection to HPV.
Notes
Financial support. This work was supported by National Institutes of Health grants K01-AI080974 (Brotman), K25-AG034216 (Shardell), UH2-AI083264 (Ravel and Forney), and R03-AI061131 (Zenilman).
Potential conflicts of interest. All authors: No potential conflicts of interest.
All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.
References
- 1.Schlecht NF, Kulaga S, Robitaille J, et al. Persistent human papillomavirus infection as a predictor of cervical intraepithelial neoplasia. JAMA. 2001;286:3106–14. doi: 10.1001/jama.286.24.3106. [DOI] [PubMed] [Google Scholar]
- 2.Martin DH. The microbiota of the vagina and its influence on women's health and disease. Am J Med Sci. 2012;343:2–9. doi: 10.1097/MAJ.0b013e31823ea228. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Boskey ER, Cone RA, Whaley KJ, Moench TR. Origins of vaginal acidity: high D/L lactate ratio is consistent with bacteria being the primary source. Hum Reprod. 2001;16:1809–13. doi: 10.1093/humrep/16.9.1809. [DOI] [PubMed] [Google Scholar]
- 4.Aroutcheva A, Gariti D, Simon M, et al. Defense factors of vaginal lactobacilli. Am J Obstet Gynecol. 2001;185:375–9. doi: 10.1067/mob.2001.115867. [DOI] [PubMed] [Google Scholar]
- 5.Ocana VS, Pesce De Ruiz Holgado AA, Nader-Macias ME. Characterization of a bacteriocin-like substance produced by a vaginal Lactobacillus salivarius strain. Appl Environ Microbiol. 1999;65:5631–5. doi: 10.1128/aem.65.12.5631-5635.1999. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Reid G, Heinemann C, Velraeds M, van der Mei HC, Busscher HJ. Biosurfactants produced by Lactobacillus. Methods Enzymol. 1999;310:426–33. doi: 10.1016/s0076-6879(99)10033-8. [DOI] [PubMed] [Google Scholar]
- 7.Boris S, Barbés C. Role played by lactobacilli in controlling the population of vaginal pathogens. Microbes Infect. 2000;2:543–6. doi: 10.1016/s1286-4579(00)00313-0. [DOI] [PubMed] [Google Scholar]
- 8.McMillan A, Dell M, Zellar MP, et al. Disruption of urogenital biofilms by lactobacilli. Colloids Surf B Biointerfaces. 2011;86:58–64. doi: 10.1016/j.colsurfb.2011.03.016. [DOI] [PubMed] [Google Scholar]
- 9.Motevaseli E, Shirzad M, Akrami SM, Mousavi AS, Mirsalehian A, Modarressi-Hossein M. Normal and tumor cervical cells respond differently to vaginal lactobacilli, independent of pH and lactate. J Med Microbiol. 2013;62(Pt 7):1065–72. doi: 10.1099/jmm.0.057521-0. [DOI] [PubMed] [Google Scholar]
- 10.Gillet E, Meys JF, Verstraelen H, et al. Bacterial vaginosis is associated with uterine cervical human papillomavirus infection: a meta-analysis. BMC Infect Dis. 2011;11:10. doi: 10.1186/1471-2334-11-10. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Watts DH, Fazzari M, Minkoff H, et al. Effects of bacterial vaginosis and other genital infections on the natural history of human papillomavirus infection in HIV-1-infected and high-risk HIV-1-uninfected women. J Infect Dis. 2005;191:1129–39. doi: 10.1086/427777. [DOI] [PubMed] [Google Scholar]
- 12.King CC, Jamieson DJ, Wiener J, et al. Bacterial vaginosis and the natural history of human papillomavirus. Infect Dis Obstet Gynecol. 2011;2011:319460. doi: 10.1155/2011/319460. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Guo YL, You K, Qiao J, Zhao YM, Geng L. Bacterial vaginosis is conducive to the persistence of HPV infection. Int J STD AIDS. 2012;23:581–4. doi: 10.1258/ijsa.2012.011342. [DOI] [PubMed] [Google Scholar]
- 14.Gao W, Weng J, Gao Y, Chen X. Comparison of the vaginal microbiota diversity of women with and without human papillomavirus infection: a cross-sectional study. BMC Infect Dis. 2013;13:271. doi: 10.1186/1471-2334-13-271. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Lee JE, Lee S, Lee H, et al. Association of the vaginal microbiota with human papillomavirus infection in a Korean twin cohort. PLOS ONE. 2013;8:e63514. doi: 10.1371/journal.pone.0063514. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Clarke MA, Rodriguez AC, Gage JC, et al. A large, population-based study of age-related associations between vaginal pH and human papillomavirus infection. BMC Infect Dis. 2012;12:33. doi: 10.1186/1471-2334-12-33. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Koumans EH, Sternberg M, Bruce C, et al. The prevalence of bacterial vaginosis in the United States, 2001–2004; associations with symptoms, sexual behaviors, and reproductive health. Sex Transm Dis. 2007;34:864–9. doi: 10.1097/OLQ.0b013e318074e565. [DOI] [PubMed] [Google Scholar]
- 18.Fredricks DN, Fiedler TL, Marrazzo JM. Molecular identification of bacteria associated with bacterial vaginosis. N Engl J Med. 2005;353:1899–911. doi: 10.1056/NEJMoa043802. [DOI] [PubMed] [Google Scholar]
- 19.Gajer P, Brotman RM, Bai G, et al. Temporal dynamics of the human vaginal microbiota. Sci Transl Med. 2012;4:132ra52. doi: 10.1126/scitranslmed.3003605. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Liu SH, Cummings DA, Zenilman JM, Gravitt PE, Brotman RM. Characterizing the temporal dynamics of human papillomavirus DNA detectability using short-interval sampling. Cancer Epidemiol Biomarkers Prev. 2013;23:200–8. doi: 10.1158/1055-9965.EPI-13-0666. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Brotman RM, Ghanem KG, Klebanoff MA, Taha TE, Scharfstein DO, Zenilman JM. The effect of vaginal douching cessation on bacterial vaginosis: a pilot study. Am J Obstet Gynecol. 2008;198:628–7. doi: 10.1016/j.ajog.2007.11.043. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.McKenna P, Hoffmann C, Minkah N, et al. The macaque gut microbiome in health, lentiviral infection, and chronic enterocolitis. PLOS Pathog. 2008;4:e20. doi: 10.1371/journal.ppat.0040020. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Kalbfleisch JD, Lawless JF. The analysis of panel data under a Markov assumption. J Amer Stat Assoc. 1985;80:863–71. [Google Scholar]
- 24.Goodman SN. Multiple comparisons, explained. Am J Epidemiol. 1998;147:807–12. doi: 10.1093/oxfordjournals.aje.a009531. discussion 15. [DOI] [PubMed] [Google Scholar]
- 25.Titman AC, Sharples LD. A general goodness-of-fit test for Markov and hidden Markov models. Stat Med. 2008;27:2177–95. doi: 10.1002/sim.3033. [DOI] [PubMed] [Google Scholar]
- 26.Christopher J. Multi-state models for panel data: the msm package for R. J Stat Softw. 2011;38:1–28. [Google Scholar]
- 27.Brown DR, Shew ML, Qadadri B, et al. A longitudinal study of genital human papillomavirus infection in a cohort of closely followed adolescent women. J Infect Dis. 2005;191:182–92. doi: 10.1086/426867. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Moscicki AB, Palefsky J, Smith G, Siboshski S, Schoolnik G. Variability of human papillomavirus DNA testing in a longitudinal cohort of young women. Obstet Gynecol. 1993;82:578–85. [PubMed] [Google Scholar]
- 29.Crum CP, Beach KJ, Hedley ML, et al. Dynamics of human papillomavirus infection between biopsy and excision of cervical intraepithelial neoplasia: results from the ZYC101a protocol. J Infect Dis. 2004;189:1348–54. doi: 10.1086/382956. [DOI] [PubMed] [Google Scholar]
- 30.de Villiers EM, Wagner D, Schneider A, et al. Human papillomavirus DNA in women without and with cytological abnormalities: results of a 5-year follow-up study. Gynecol Oncol. 1992;44:33–9. doi: 10.1016/0090-8258(92)90008-7. [DOI] [PubMed] [Google Scholar]
- 31.Woodman CBJ, Collins SI, Young LS. The natural history of cervical HPV infection: unresolved issues. Nat Rev Cancer. 2007;7:11–22. doi: 10.1038/nrc2050. [DOI] [PubMed] [Google Scholar]
- 32.Srinivasan S, Liu C, Mitchell CM, et al. Temporal variability of human vaginal bacteria and relationship with bacterial vaginosis. PLOS ONE. 2010;5:e10197. doi: 10.1371/journal.pone.0010197. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Schwebke J, Morgan S, Weiss H. The use of sequential self-obtained vaginal smears for detecting changes in the vaginal flora. Sex Transm Dis. 1997;24:236–9. doi: 10.1097/00007435-199704000-00009. [DOI] [PubMed] [Google Scholar]
- 34.O'Hanlon DE, Moench TR, Cone RA. In vaginal fluid, bacteria associated with bacterial vaginosis can be suppressed with lactic acid but not hydrogen peroxide. BMC Infect Dis. 2011;11:200. doi: 10.1186/1471-2334-11-200. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Alakomi HL, Skytta E, Saarela M, Mattila-Sandholm T, Latva-Kala K, Helander IM. Lactic acid permeabilizes gram-negative bacteria by disrupting the outer membrane. Appl Environ Microbiol. 2000;66:2001–5. doi: 10.1128/aem.66.5.2001-2005.2000. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Macklaim JM, Gloor GB, Anukam KC, Cribby S, Reid G. At the crossroads of vaginal health and disease, the genome sequence of Lactobacillus iners AB-1. Proc Natl Acad Sci USA. 2011;108(suppl 1):4688–95. doi: 10.1073/pnas.1000086107. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Vaccarella S, Herrero R, Dai M, et al. Reproductive factors, oral contraceptive use, and human papillomavirus infection: pooled analysis of the IARC HPV prevalence surveys. Cancer Epidemiol Biomarkers Prev. 2006;15:2148–53. doi: 10.1158/1055-9965.EPI-06-0556. [DOI] [PubMed] [Google Scholar]
- 38.Ghanem KG, Datta SD, Unger ER, et al. The association of current hormonal contraceptive use with type-specific HPV detection. Sex Transm Infect. 2011;87:385–8. doi: 10.1136/sextrans-2011-050005. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Marks M, Gravitt PE, Gupta SB, et al. Combined oral contraceptive use increases HPV persistence but not new HPV detection in a cohort of women from Thailand. J Infect Dis. 2011;204:1505–13. doi: 10.1093/infdis/jir560. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40.Riggs M, Klebanoff M, Nansel T, Zhang J, Schwebke J, Andrews W. Longitudinal association between hormonal contraceptives and bacterial vaginosis in women of reproductive age. Sex Transm Dis. 2007;34:954–9. [PubMed] [Google Scholar]
- 41.Moench TR, Mumper RJ, Hoen TE, Sun M, Cone RA. Microbicide excipients can greatly increase susceptibility to genital herpes transmission in the mouse. BMC Infect Dis. 2010;10:331. doi: 10.1186/1471-2334-10-331. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.Fuchs EJ, Lee LA, Torbenson MS, et al. Hyperosmolar sexual lubricant causes epithelial damage in the distal colon: potential implication for HIV transmission. J Infect Dis. 2007;195:703–10. doi: 10.1086/511279. [DOI] [PubMed] [Google Scholar]
- 43.Wright TC, Jr, Denny L, Kuhn L, Pollack A, Lorincz A. HPV DNA testing of self-collected vaginal samples compared with cytologic screening to detect cervical cancer. JAMA. 2000;283:81–6. doi: 10.1001/jama.283.1.81. [DOI] [PubMed] [Google Scholar]
- 44.Gravitt PE, Lacey JV, Jr, Brinton LA, et al. Evaluation of self-collected cervicovaginal cell samples for human papillomavirus testing by polymerase chain reaction. Cancer Epidemiol Biomarkers Prev. 2001;10:95–100. [PubMed] [Google Scholar]
- 45.Forney LJ, Gajer P, Williams CJ, et al. Comparison of self-collected and physician-collected vaginal swabs for microbiome analysis. J Clin Microbiol. 2010;48:1741–8. doi: 10.1128/JCM.01710-09. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 46.Castellsague X, Munoz N. Chapter 3: Cofactors in human papillomavirus carcinogenesis—role of parity, oral contraceptives, and tobacco smoking. J Natl Cancer Inst Monogr. 2003;31:20–8. [PubMed] [Google Scholar]
- 47.Verhoeven V, Renard N, Makar A, et al. Probiotics enhance the clearance of human papillomavirus-related cervical lesions: a prospective controlled pilot study. Eur J Cancer Prev. 2012;22:46–51. doi: 10.1097/CEJ.0b013e328355ed23. [DOI] [PubMed] [Google Scholar]
- 48.Reid G, Charbonneau D, Erb J, et al. Oral use of Lactobacillus rhamnosus GR-1 and L. fermentum RC-14 significantly alters vaginal flora: randomized, placebo-controlled trial in 64 healthy women. FEMS Immunol Med Microbiol. 2003;35:131–4. doi: 10.1016/S0928-8244(02)00465-0. [DOI] [PubMed] [Google Scholar]
- 49.Marrazzo JM, Fiedler TL, Srinivasan S, et al. Extravaginal reservoirs of vaginal bacteria as risk factors for incident bacterial vaginosis. J Infect Dis. 2012;205:1580–88. doi: 10.1093/infdis/jis242. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 50.Antonio MA, Rabe LK, Hillier SL. Colonization of the rectum by Lactobacillus species and decreased risk of bacterial vaginosis. J Infect Dis. 2005;192:394–8. doi: 10.1086/430926. [DOI] [PubMed] [Google Scholar]