Association between vaginal microbiota and vaginal inflammatory immune markers in postmenopausal women

Abstract

Objective:

In premenopausal individuals, vaginal microbiota diversity and lack of Lactobacillus dominance is associated with greater mucosal inflammation, which is linked to higher risk of cervical dysplasia and infections. It is not known if the association between the vaginal microbiota and inflammation is present after menopause, when the vaginal microbiota is generally higher-diversity and fewer people have Lactobacillus dominance.

Methods:

This is a post-hoc analysis of a subset of postmenopausal individuals enrolled in a randomized trial for treatment of moderate-severe vulvovaginal discomfort which compared vaginal moisturizer, estradiol, or placebo. Vaginal fluid samples from 0, 4 and 12 weeks were characterized using 16S rRNA gene sequencing (microbiota) and MesoScale Discovery (vaginal fluid immune markers: IL1b, IL1a, IL2, IL6, IL18, IL10, IL9, IL13, IL8, IP10, MIP1a, MIP1b, MIP3a). Global associations between cytokines and microbiota (assessed by relative abundance of individual taxa and Shannon index for alpha, or community, diversity) were explored, adjusting for treatment arm, using linear mixed models, principal component analysis, and GLMM MiRKAT.

Results:

A total of 119 individuals with mean age 61 were included. At baseline, 29.5% of participants had a Lactobacillus-dominant vaginal microbiota. Across all timepoints, alpha diversity (Shannon index, p=0.003) was highly associated with immune markers. Individual markers that were associated with Lactobacillus dominance were similar to those observed in premenopausal people: IL10, IL1b, IL6, IL8, (FDR<0.01), IL13 (FDR=0.02) and IL2 (FDR=0.09). Over 12 weeks, change in alpha diversity was associated with change in cytokine concentration (Shannon, p=0.018), with decreased proinflammatory cytokine concentrations observed with decreasing alpha diversity.

Conclusions:

In this cohort of postmenopausal individuals, Lactobacillus dominance and lower alpha diversity were associated with lower concentrations of inflammatory immune markers, as has been reported in premenopausal people. This suggests that after menopause, lactobacilli continue to have beneficial effects on vaginal immune homeostasis, despite lower prevalence.

Introduction

Associations between the vaginal microbiome, local inflammation and disease acquisition and persistence have been described, primarily in premenopausal people. Many outcomes associated with high-diversity vaginal microbial communities, such as human papilloma virus (HPV) persistence and progression to dysplasia^1,2, or increased risk for sexually transmitted infection (STI) acquisition^3,4, are also relevant for postmenopausal people. However, comparatively little is known about how the postmenopausal vaginal microbiome impacts the vaginal mucosa and mucosal immune responses, nor the impact of leveraging microbial changes for sexual health.

In premenopausal people, both greater diversity of the vaginal microbiome and lower levels of Lactobacillus are associated with inflammation ^5,6. Inflammation of the female reproductive tract is linked to a higher risk of STIs and cervical dysplasia ^3,7–9. Some studies have shown differences in vaginal fluid immune markers and immune cell populations between pre- and postmenopausal people, suggesting a change in immune function after menopause ^10,11. It is unclear whether the inflammatory associations present in premenopausal people are also observed after menopause.

Estrogen levels decrease during the menopausal transition, leading to decreased glycogen storage in vaginal epithelial cells and decreased prevalence of Lactobacillus ^12,13. There is clearly a relationship between estrogen and the vaginal microbiome after menopause. Previous studies have demonstrated that postmenopausal people using vaginal estradiol ^14,15 and oral menopausal hormone therapy^16–18 are more likely to have vaginal microbiota dominated by Lactobacillus and lower vaginal microbial diversity¹⁹. Despite this, menopausal individuals do not meet criteria for bacterial vaginosis at higher rates ^20,21 and it is unusual for those with high Nugent scores to have a vaginal community comprised of bacteria associated with bacterial vaginosis (BV)²².

A link between the vaginal microbiome and local inflammation has not been clearly characterized in postmenopausal people. The inherent differences in vaginal mucosal homeostasis before and after menopause makes it difficult to extrapolate findings from the premenopausal to postmenopausal populations. We aimed to establish whether the associations between vaginal microbial populations and local inflammation observed in premenopausal cohorts were also present in a postmenopausal cohort.

Methods

This is a post-hoc analysis of a subset of postmenopausal individuals with moderate-severe vulvovaginal discomfort enrolled in a randomized trial of Vagifem 10μg tablet with placebo gel, placebo tablet plus Replens moisturizer, or dual placebo of tablet and gel ²³. In brief, the trial took place over 12 weeks at two centers: Kaiser Permanente Washington Health Research Institute in Seattle and University of Minnesota in Minneapolis. Eligible participants were at least 2 years since last menses, age 45–70 years, with at least 1 moderate to severe vulvovaginal symptom of itching, pain, irritation, dryness or pain with vaginal penetration. Individuals with other explanations for symptoms were excluded (e.g. vaginal infections, lichen sclerosus). Visits were conducted at 0, 4, and 12 weeks after enrollment for collection of vaginal swabs and cervicovaginal lavage (CVL). Vaginal swabs were used for wet mount, pH measurement, and microbiota analysis. CVL was used to measure soluble immune markers.

Samples from a subset of the initial study population underwent analysis of microbiota and vaginal fluid immune markers ^14,24. Participants were only selected for the subset analysis if the participant used at least 80% of study product doses, and the participant was not diagnosed with infectious vaginitis. For the subset analysis, within each arm we chose 20 people who had a decrease in symptom severity of 2 or more points, and 20 people who had less improvement, or worsening of symptoms. Vaginal swabs were clinician-collected via speculum exam, and the vaginal microbiome was characterized using amplicon sequencing of the V3-V4 region of the 16S rRNA gene on the Illumina MiSeq instrument (Illumina, San Diego, CA) as previously described ^24,25. Inflammatory markers were assessed using an immunoassay (MesoScale Discovery) for the following vaginal fluid immune markers: IL1β, IL1α, IL2, IL6, IL18, IL10, IL9, IL13, IL8, IP10, MIP1α, MIP1β, MIP3α ²⁴.

Microbiome analysis focused on relative abundance of individual taxa as well as within-sample microbial diversity (alpha diversity) as measured by Shannon index. Cytokines were assessed individually using a linear mixed model (correcting for multiple comparisons); by principal component (PC) analysis; and globally using GLMM MiRKAT. Principal components were calculated from the distance matrix of cytokines and then additional information was incorporated using a kernel matrix using GLMM MiRKAT ²⁶. Loadings of cytokine contribution to each principal component were obtained using Euclidean distance matrix of cytokines. Global associations between microbiota and inflammatory cytokines were explored, adjusting for treatment arm, using linear mixed models or GLMM MiRKAT. Statistical analysis was carried out using R. Where indicated, p-values were corrected for multiple comparison testing using FDR (false discovery rate).

Results

One hundred nineteen participants had complete immune and microbiota data and were included in this post-hoc analysis; 102 of these participants had complete data across all three timepoints. The participants were divided evenly between study arms, 40 from the estradiol + placebo group, 39 from the moisturizer + placebo group, and 40 from the dual placebo group. The mean age of the included participants was 61 years (Table 1).

Table 1:

Demographics of participants included in analysis. Demographics are also broken down by vaginal microbiota status at baseline visit, either Lactobacillus dominant or non-dominant. P-values obtained via either Wilcoxon (continuous) or Fisher exact (categorical) tests.

	Lactobacillus dominant at baseline visit (N=31)	Lactobacillus non-dominant at baseline visit (N=74)	P-value (Wilcoxon or Fisher exact test)
Age	60 [57.5, 62]	61 [58, 64]	0.225
Years since LMP	9.3 [6, 14.3]	10.6 [6.6, 16.2]	0.637
Sexually active			0.423
Yes	23 (74.2%)	61 (82.4%)
No	8 (25.8%)	13 (17.6%)
Ethnicity			0.003
White	24 (77.4%)	71 (96%)
Black	4 (12.9%)	0 (0%)
American Indian or Alaska Native	1 (3.2%)	0 (0%)
Asian	2 (6.5%)	2 (2.7%)
Other (missing)	0 (0%)	1 (1.4%)

At baseline, the proportion of participants with Lactobacillus-dominant microbiota was 29.5% (31/105); of these, 6 had Lactobacillus crispatus dominance (5.7% of the cohort). Participants who did not have Lactobacillus-dominant microbiota at baseline were more likely to be White, but otherwise there were no significant demographic differences in age, years since menopause, or sexual activity (Table 1).

Inflammatory markers in the participants across all timepoints were condensed into a single variable using cytokine principal component analysis, with the first Principal Component (PC1), representing the most variation in the data, capturing 32% of the variance and the second Principal Component (PC2), representing the second most variation in the data, capturing an additional 23% of the variance. PC1 separated samples in women who were Lactobacillus-dominant at week 0 from women who were not (Figure 1). Cytokines IL6, IL8, IL1b, and IL1a most strongly contributed to the negative direction of PC1, which was associated with non-Lactobacillus dominance. Conversely, IP10 and MIP3a drove a positive PC1 but were also the strongest contributors to PC2 (Figure 1).

Figure 1:

Biplot representation of cytokine contributions to PC1 (x axis) and PC2 (y axis), along with all datapoints for each participant over the three timepoints in the study. Colored dots represent the baseline visit (visit 0). Blue represents Lactobacillus dominant at baseline, while red represents non-Lactobacillus dominant at baseline. Arrows are a vector representation of the contribution of each cytokine to PC1 and PC2. Cytokines IL6, IL8, IL1B, and IL1a most strongly contributed to the negative direction of PC1. A positive PC1 represents a less inflammatory signature, as demonstrated here based on pro-inflammatory cytokines aligning with negative PC1. IP10 and MIP3a drove a weakly positive PC1 but were also the strongest contributors to PC2.

Across all timepoints, bacterial alpha diversity was highly associated with a global assessment of immune markers from GLMM MiRKAT, after adjusting for treatment arm (p = 0.003; Figure 2). Of the top 40 taxa associated with PC1, 5 were Lactobacillus. All Lactobacillus species were positively associated with PC1 (meaning lower levels of the markers most strongly contributing to PC1, such as IL1b, IL6, IL8, IL1a) (Figure 3).

Figure 2:

Stacked barplot representation of the 15 taxa present at an abundance of at least 1% throughout the dataset, as well as “other” taxa that represent <1% of bacteria in the dataset, with each bar representing a sample. Per-sample representation of cytokine Principal Component PC1 and Shannon diversity (within sample diversity) are plotted beneath the barplot. Samples from all timepoints and all participants are included and are ordered by increasing relative abundance of Lactobacillus species. Within-sample bacterial diversity (alpha diversity, measured by Shannon index) was associated with cytokine PC1 (p = 0.003), after adjusting for treatment arm. High PC1 and low Shannon diversity are associated with higher Lactobacillus dominance, as demonstrated by the stacked barplot.

Figure 3:

Taxa that were significantly associated with PC1 of cytokines, by a kernel FDR cutoff of <0.2. Taxa are in alphabetical order and colored by positive vs negative correlation with PC1. All Lactobacillus species were positively associated with PC1 (meaning lower concentrations of inflammatory cytokines most strongly contributing to PC1, such as IL1b, IL6, IL8, IL1a).

Several inflammatory markers had significantly different levels between participants with Lactobacillus dominance (n=31) and participants with more diverse communities (n=74) (Figure 4): IL10 (p < 0.001, FDR < 0.01), IL1b (p < 0.001, FDR < 0.01), IL6 (p < 0.001, FDR < 0.01), IL8 (p < 0.001, FDR < 0.01), IL13 (p = 0.002, FDR = 0.02) and IL2 (p = 0.01, FDR=0.09).

Figure 4:

Comparison of each cytokine analyzed according to Lactobacillus dominance (Lactobacillus dominant n = 31; Lactobacillus non-dominant n = 74). Significance is determined by Wilcoxon test via GLMM adjusted for multiple samples per participant and by arm. Boxplots show median and IQR. Significance symbols are defined as follows: ns p ≥ 0.1; * p < 0.1; ** p < 0.05; *** p < 0.01 (individual cytokine p values are FDR corrected; original p value used for PC1). The following inflammatory markers were statistically significantly different between these groups after correcting for multiple comparisons: IL10 (p < 0.001, FDR < 0.01), IL1b (p < 0.001, FDR < 0.01), IL6 (p < 0.001, FDR < 0.01), IL8 (p < 0.001, FDR < 0.01), IL13 (p = 0.002, FDR = 0.02) and IL2 (p = 0.011, FDR=0.09).

Over the 12 weeks of the study, change in bacterial alpha diversity (within sample diversity measured by Shannon index) was associated with global change in cytokines while adjusting for study arm (p=0.018). A diverse community at week 0 shifted to Lactobacillus-dominant community at week 12 in 18 people (17.5% of cohort), but this transition was not associated with a significant change in PC1. Conversely, a Lactobacillus-dominant community at week 0 shifted to a diverse community at week 12 in 5 people (4.9% of cohort), again with a nonsignificant change in cytokine PC1 (Supplemental Figure 1). There were no statistically significant differences in paired individual cytokines levels from week 0 to week 12 in the small number of participants whose vaginal microbiome Lactobacillus dominant status changed over the course of the study (Supplemental Figure 1). However, we did find reduced inflammatory cytokine levels, with increased PC1 (p=0.019) over the study period in the 26 participants who maintained a Lactobacillus dominant microbiota throughout the twelve weeks of the study (Supplemental Figure 1). Within these participants who remained Lactobacillus dominant throughout the study, pH was significantly associated with IL8 concentrations (Supplemental Figure 2a, p = 0.02), while Shannon diversity (Supplemental Figure 2b) was not statistically significantly associated with immune marker concentrations. These exploratory longitudinal analyses are underpowered to show a clear link between the microbiome and inflammatory markers within participants over time and suggest larger cohorts are needed to fully explore how the vaginal microbiome correlates over time with genital inflammation in an individual.

Discussion

It is well established that the vaginal microbiome in premenopausal people is associated with the local mucosal inflammatory milieu. In particular, higher bacterial diversity and lack of Lactobacillus dominance is associated with higher concentrations of several pro-inflammatory cytokines and chemokines. Inflammation, in turn, is thought to be the mechanistic link to the poor reproductive health outcomes associated with dysbiosis in numerous observational studies ^3,5,6,8,9. The vaginal microbiome and local inflammation may mediate genitourinary health in postmenopausal people, but comparatively little is understood about the postmenopausal vaginal microbiome’s relationship to local inflammation. The inherent differences between the pre- and post-menopausal vaginal tissue, host physiology, and commensal bacteria make extrapolation from the premenopausal literature tenuous.

In this small cohort of postmenopausal individuals, we show that both Lactobacillus dominance and lower bacterial alpha diversity are associated with lower concentrations of inflammatory immune markers, similar to the premenopausal population. Postmenopausal people with higher diversity microbiomes or who lacked Lactobacillus dominance had higher concentrations of pro-inflammatory cytokines. The cytokines driving this signal, such as IL-1a, IL-1b, IL10, IL8, and IL6, are similar to the cytokines previously reported to be associated with the vaginal microbial composition in premenopausal people ^5,27. Similar to findings in premenopausal people, IP-10 was not associated with dysbiosis ⁵.

These preliminary findings suggest that after menopause, lactobacilli may continue to have beneficial effects on mucosal immunity by contributing to – or at least associating with – a lower local inflammatory state. These findings also suggest that increased alpha diversity of the vaginal microbiota is associated with pro-inflammatory cytokines in post-menopausal individuals. Although the high-diversity vaginal microbial community after menopause is often different than that seen with bacterial vaginosis, this suggests that the post-menopausal vaginal ecosystem may still benefit from Lactobacillus dominance and lower bacterial diversity. Together, these findings suggest that although Lactobacillus dominance may not be “normal” after menopause, it could represent a “healthy” vaginal microenvironment associated with a lower inflammatory state.

This analysis represents a study of associations, albeit in the setting of a randomized trial. A prospective randomized study modifying the microbiome would be the optimal design to understand causal relationships regarding the vaginal microbiome, however as yet there is not a reliable intervention to do this. Additionally, this is a posthoc analysis of a subset of trial participants, and so may have insufficient power to detect small associations. Additionally, we only followed people for 12 weeks after the start of intervention; a longer study may be necessary to fully assess whether shifts in the microbiome within a participant correlate with inflammatory changes. Finally, due to the small sample size, we were unable to assess whether estrogen treatment is an effect modifier for the relationship between vaginal microbiota and inflammatory markers.

The data presented here have been analyzed previously to explore other links between estrogen, the vaginal microbiome, inflammation, and genitourinary symptoms of menopause. We previously found that people randomized to vaginal estrogen had increased Lactobacillus relative abundance and decreased diversity in the vaginal microbiota and that even within participants who started with a low-diversity vaginal microbiome, vaginal estrogen decreased vaginal pH ¹⁴. However, vaginal estrogen did not improve symptoms more than other treatments ^23,28. We further showed that the vaginal microbiome was not associated with decreased severity of the most bothersome symptom of menopause over the course of this study ²⁴.

The analysis presented here suggests that the link between the vaginal microbiome aberrations from Lactobacillus dominance and inflammation is preserved compared to the premenopausal population. Establishing the link between Lactobacillus dominance and decreased vaginal inflammation in the postmenopausal population, as we have shown here, may have several implications for postmenopausal genitourinary health beyond genitourinary symptoms of menopause. Further investigation is required to assess whether associations between vaginal microbiota and outcomes established in the premenopausal population, such as risk for acquisition of STIs or risk of cervical dysplasia, are also present in postmenopausal people with higher diversity vaginal microbiota. Other gynecologic endpoints particularly relevant to the postmenopausal population that may be influenced by the vaginal microbiome and local inflammation include mucosal wound healing (for example, after hysterectomy) and urinary incontinence²⁹.

Further studies are needed to describe the relationship between bacterial species prevalence (notably Lactobacillus), microbial diversity, mucosal inflammatory markers, and health outcomes in postmenopausal people. The clinical importance of the microbiome and mucosal inflammation in the premenopausal setting, combined with the findings presented here of similar relationships between these biomarkers in postmenopausal people, suggests the potential utility of modulating bacteria to reduce mucosal inflammation in the postmenopausal population.

Conclusion

In this small cohort of postmenopausal people, a post hoc analysis of clinical trial data showed that a minority had a Lactobacillus-dominated vaginal microbiome but that Lactobacillus dominance and low bacterial diversity was associated with lower inflammatory markers in the genital tract. The association between a diverse, Lactobacillus-depleted vaginal microbiome and local inflammation has previously been documented in the premenopausal population; an altered microbiome has been associated with adverse gynecologic outcomes in the premenopausal context. The data presented here are a first step toward examining whether this may also be true in the postmenopausal vagina.