Bacterial vaginosis-driven changes in vaginal T cell phenotypes and their implications for HIV susceptibility

Abstract

Bacterial vaginosis (BV) is a dysbiosis of the vaginal microbiome that is prevalent in reproductive-age women worldwide. Adverse outcomes associated with BV include an increased risk of sexually acquired Human Immunodeficiency Virus (HIV), yet the immunological mechanisms underlying this association are not well understood. To investigate BV driven changes to cervicovaginal tract (CVT) and circulating T cell phenotypes, participants with or without BV provided vaginal tract (VT) and ectocervical (CX) tissue biopsies and peripheral blood mononuclear cells (PBMC). Immunofluorescence analysis of genital mucosal tissues revealed a reduced density of CD3⁺CD4⁺CCR5⁺ cells in the VT lamina propria of individuals with compared to those without BV (median 243.8 cells/mm² BV− vs 106.9 cells/mm² BV+, p=0.043). High-parameter flow cytometry of VT biopsies revealed an increased frequency in individuals with compared to those without BV of dysfunctional CD39⁺ conventional CD4⁺ T cells (Tconv) (median frequency 15% BV− vs 30% BV+, p_adj=0.0331) and tissue-resident CD69₊CD103⁺ Tconv (median frequency 24% BV− vs 38% BV+, p_adj=0.0061), previously reported to be implicated in HIV acquisition and replication. Our data suggests that BV elicits diverse and complex VT T cell alterations and expands on potential immunological mechanisms that may promote adverse outcomes including HIV susceptibility.

Introduction

Bacterial vaginosis (BV) is characterized by a vaginal dysbiosis in which the normally Lactobacillus-dominated microbiome is replaced by one with higher bacterial diversity, including increased concentrations of anaerobic bacteria (1). BV results in symptoms of vaginal inflammation, discharge, and discomfort, and is estimated to affect 23-29% of reproductive-age women worldwide (2), making BV the most common cause of vaginal symptoms leading patients to seek medical care (3). BV is also associated with several adverse medical outcomes (1, 2, 4–11) including preterm delivery (12), chorioamnionitis (13), endometritis (14), and an increased risk of Human Immunodeficiency Virus (HIV) acquisition (5, 11). The issue of HIV acquisition risk is of particular concern in parts of the world where BV and HIV are individually prevalent, as is the case in sub-Saharan Africa (15). Although many studies have investigated immune system alterations associated with BV (1, 5–7, 10, 11, 16–19), the link between cervicovaginal tract (CVT) T cells, which may mediate the relationships between BV and adverse outcomes, is not well understood. Here, we sought to better understand the impact of BV on CVT mucosal and systemic immune responses in Kenyans and how these changes may alter HIV susceptibility.

In the CVT, innate and adaptive immune cells line the epithelium and lamina propria to provide immunological surveillance (20). Notably, T cell populations in the CVT maintain distinct phenotypes from those circulating in the blood (21, 22). These include tissue-resident memory T cells (T_RM) bearing CD69 with or without CD103 (23). CD69 is a marker of activation (24) that has a unique role in maintaining tissue residency by binding S1PR1 and preventing egress from the tissue (25–27). CD103 further promotes tissue residency by binding to E-cadherin (28). T_RM are retained in the mucosal tissue and have the ability to readily respond to pathogens in situ (21, 29–31). Upon recognition of their cognate antigen, T_RM use their “sensing and alarm function” to produce pro-inflammatory cytokines and chemokines that rapidly recruit and activate other immune cells, thereby potentiating the tissue immune response while also executing their specific effector and cytotoxic functions (32–35).

Given that CVT mucosal tissue is the site where, in the context of heterosexual transmission, HIV exposure first occurs, activated CD4⁺ T cells expressing the HIV coreceptor CCR5 in the CVT have been thought to be prime targets for initial HIV infection (36). Studies using a mouse model demonstrated that the introduction of BV-associated bacteria into the vagina led to an increased quantity of CVT CD4⁺ T cells expressing the activation marker CD44 and HIV coreceptor CCR5 (7), suggesting dysbiosis of normal vaginal flora may lead to an increase of HIV target cells. In humans, studies of immune responses in context of BV have characterized immune cells collected from the CVT lumen via minimally invasive sampling techniques such as cervicovaginal lavage, vaginal swabs, cervical swabs, or cytobrushes. These studies have identified an increase in CCR5⁺CD4⁺ T cells in the CVT lumen of BV+ individuals (37–40), without addressing whether this observation is true in the CVT deeper tissue layers. In addition to the detection of HIV target cells in the CVT lumen of individuals with BV, studies have shown a connection between BV and the presence of pro-inflammatory cytokines in the CVT, most reliably IL-1β (7, 37, 41, 42); adding further support to the idea that proinflammatory mechanisms underlie the link between BV and increased HIV susceptibility. However, the relationship between BV and an increase in CD4⁺ T cell activation is inconsistently observed (43, 44), suggesting more complex mechanisms may be contributing to increased HIV susceptibility in individuals with BV. Further characterizing BV driven immune modulations in the unique immune environment of the CVT is paramount to understanding mechanisms underlying increased HIV susceptibility in BV+ individuals.

To comprehensively evaluate immune alterations associated with BV, we utilized tissue cryopreservation (45) and high-parameter, high-throughput flow cytometry to compare immune cells isolated from CVT tissue biopsies and blood collected from individuals with versus without BV. Given the dense T cell populations within CVT tissue (46), cells isolated from CVT mucosal biopsies may better represent the mucosal tissue immune environment than those isolated from the CVT lumen by cytobrush or other methods. Additionally, improved flow cytometry technology allows us to broadly evaluate immune cells in the context of BV for phenotypic markers not previously characterized.

Here, we report detailed T cell characteristics of vaginal tract (VT) and ectocervix (CX) mucosal tissues and peripheral blood mononuclear cells (PBMC) from the same participants, comparing those individuals with to those without BV. Additionally, we analyzed CVT fluids and serum for local and systemic cytokines and chemokines in individuals with versus without BV. While we did not find an increase in CD3⁺CD4⁺CCR5⁺ HIV target cells in the CVT of BV+ individuals, we did observe VT T cells with dysfunctional phenotypes and altered expression of soluble mediators in CVT fluid from BV+ individuals. Our findings highlight additional mechanisms not previously entertained that may underlie increased HIV susceptibility associated with BV.

Results

Study population characteristics

Of the 245 Kinga Study participants with analyzed samples, 212 participants met the criteria to be included in the BV flow cytometry analysis. 64 participants contributed samples when they were BV+ by Nugent score (46 at study enrollment and 18 at study exit). The BV− comparator group was comprised of samples contributed by the remaining N=148 participants who had normal flora at enrollment. Overall, our study population was young (57% were <30 years of age), sexually active (median of 8 unprotected sex acts in the prior month), with 40% using hormonal contraceptives, 39% HSV-2 seropositive, and 17% had a sexual partner living with HIV (Table 1). Analysis for soluble immune factors was performed on a subset of these participants (N=130 BV−, N=57 BV+). Similarly, immunofluorescent imaging was performed on a smaller subset (N=55 VT BV−, N=18 VT BV+, N=44 CX BV−, N=16 CX BV+).

Table 1 –

Demographic and clinical data from 212 participants included in this analysis.

Demographic and Clinical Characteristics

BV Status	Overall N = 2121	Negative N = 1481	Positive N = 641
Age
<=24	69 (33%)	51 (34%)	18 (28%)
25-29	50 (24%)	31 (21%)	19 (30%)
30-34	35 (17%)	27 (18%)	8 (13%)
35-39	29 (14%)	18 (12%)	11 (17%)
40-49	26 (12%)	20 (14%)	6 (9.4%)
>=50	3 (1.4%)	1 (0.7%)	2 (3.1%)
HIV Exposure
HIV exposed (HIV serodifferent couple)	37 (17%)	25 (17%)	12 (19%)
HIV unexposed (HIV negative, concordant couple)	175 (83%)	123 (83%)	52 (81%)
HSV-2 Status
Indeterminate	9 (4.2%)	8 (5.4%)	1 (1.6%)
Seronegative	118 (56%)	91 (61%)	27 (42%)
Not done	2 (0.9%)	1 (0.7%)	1 (1.%)
Seropositive	83 (39%)	48 (32%)	35 (55%)
Hormonal Contraceptive Use
Had menses in last 35 days	93 (45%)	64 (44%)	29 (48%)
Secondary amenorrhea	32 (15%)	21 (14%)	11 (18%)
Uses hormonal contraceptives	82 (40%)	62 (42%)	20 (33%)
Missing	5	1	4
Number of unprotected vaginal intercourse in last 30 days	8 (3,12)	8 (4,12)	8 (1,12)

¹n (%); Median (IQR)

Reduced abundance of CCR5⁺ HIV target cells in the VT lamina propria in individuals with BV

It has been reported that T cells are distributed throughout the epithelium and lamina propria tissue compartments in the CVT with great abundance in the lamina propria space bordering the epithelium (46). To quantify how CD3⁺CD4⁺CCR5⁺ HIV target cell density in both the epithelium and lamina propria is affected by BV, we performed immunofluorescent staining on CX and VT tissue sections from individuals with versus without BV. The epithelium and lamina propria were visualized by H&E staining (Fig 1a) or were labeled with immunofluorescent antibodies against CD3, CD4, and CCR5 and counterstained with DAPI (Fig 1b) for T cell density quantification. We found the density of CD3⁺ and CD3⁺CD4⁺ cells in the VT lamina propria were not significantly different between BV+ and BV− individuals (Fig 1c); however, we observed a significant reduction in CD3⁺CD4⁺CCR5⁺ HIV target cells in the VT lamina propria (median 243.8 cells/mm² BV− vs 106.9 cells/mm² BV+, p = 0.043; Table S2.1) in samples from BV+ compared to BV− individuals (Fig 1c), with minimal differences observed in the VT epithelium (Fig 1d). In contrast, no significant differences in CD3⁺, CD3⁺CD4⁺, or CD3⁺CD4⁺CCR5⁺ cell densities were observed in the lamina propria or epithelium of CX tissue sections from BV+ individuals compared to BV− individuals (Fig 1e–f, Table S2).

Figure 1. BV is associated with a decreased abundance of CD3⁺CD4⁺CCR5⁺ cells in the VT lamina propria.

(a) Representative hematoxylin and eosin-stained vaginal tract tissue section imaged with brightfield microscopy. (b) Representative immunofluorescent stained tissue section from the same vaginal tract (VT) biopsy as Figure 1a. DAPI stain is shown in blue, the CD3 stain is shown in red, the CD4 stain is shown in cyan, and the CCR5 stain is shown in green. All fluorescent signals overlayed. (c-f) Comparison of the cell density of CD3+ T cells, CD3+/CD4+ T cells or CD3+/CD4+/CD5+ HIV target cells between BV negative and BV positive samples in the VT Lamina Propria (c), VT Epithelium (d), CX Lamina Propria (e) and CX Epithelium (f). Wilcoxon T-test was performed for each comparison shown. “ns” labeled comparisons are not significant with p > 0.05.

CD8⁺ T cells are enriched in the VT of individuals with BV

To investigate the CVT mucosal T cell population in greater detail, we performed high-parameter flow cytometry on cells isolated from cryopreserved CX, VT, and PBMC samples. The majority of CD45⁺ cells in the VT and CX are CD3⁺ T cells (median > 75%; Fig 2a, Table S3.1–2) demonstrating the importance of T cells in immunosurveillance of the CVT. More significant phenotypic T cell differences were observed when VT samples from BV+ vs BV− participants were compared, and fewer differences were observed in either the CX or PBMC (Table S3), underscoring that major alterations in T cell populations in the context of BV may be specific to the VT. More specifically, we evaluated CD3⁺ T cell frequency as a fraction of CD45⁺ cells and found significantly greater CD3⁺ T cell frequency in VT samples (median frequency 77% BV− vs 83% BV+, p_adj = 0.0485; Table S3.1) from BV+ compared to BV− individuals (Fig 2a). The fraction of CD3⁺ T cells with a conventional CD4⁺ T cell (Tconv) phenotype (CD4⁺CD25⁻; Figure S1) among total CD45⁺ cells was not significantly altered by BV status (Fig 2b), but the frequency of CD8⁺ T cells among CD45⁺ cells was significantly increased (median frequency 30% BV− vs 38% BV+, p_adj = 0.0041; Table S3.1) in the VT of individuals with vs without BV (Fig 2c), suggesting that CD8⁺ T cells are enriched in VT samples from BV+ compared to BV− individuals and may be driving a total increase in VT T cells in BV+ individuals. When evaluating the frequency of Tconv cells as a fraction of total CD3⁺ T cells, we observed a significant decrease in VT samples (median frequency 44% BV− vs 32% BV+, p_adj = 0.018; Table S3.1) from BV+ individuals compared to BV− individuals (Fig 2d). Consequently, we observed a significantly increased CD8⁺ T cell fraction of total T cells in VT samples (median frequency 38% BV− vs 44% BV+, p_adj = 0.0075; Table S3.4) from BV+ vs BV− individuals (Fig 2e). The enrichment of CD8⁺ T cells in the VT of BV+ individuals appears to be associated with higher total T cell frequency, as well as a shift in CD4⁺/CD8⁺ T cell ratio in the VT of BV+ individuals.

Figure 2. CD8⁺ T cell abundance increases in the VT tissues of individuals with BV.

Flow cytometry was used to quantify the proportions of T cells within different tissue sites as indicated. (a) The frequency of CD3⁺ among the total CD45⁺ population in ectocervix (CX), vaginal tract (VT), and peripheral mononuclear blood cells (PBMC) samples in BV− and BV+ individuals. The frequency of Tconv (CD3⁺CD4⁺CD25⁻ cells) (b) and CD3⁺CD8⁺ (c) among the total CD45⁺ population in CX, VT, and PBMC samples in BV− and BV+ individuals. The frequency of Tconv (d) and CD8⁺ (e) among the total CD3⁺ population in CX, VT, and PBMC samples in BV− and BV+ individuals. Adjusted rank regression T test performed to compare frequencies in each tissue between BV− and BV+ individuals. Adjusted p-value displayed when p ≤ 0.05, and the p-value is labeled “ns” for not significant when adjusted p > 0.05.

Few phenotypic alterations in CVT or circulating CD8⁺ T cells in individuals with BV

To further investigate the enriched CD8⁺ T cell subset in the VT of BV+ individuals, we evaluated CD8⁺ T cells for differences in activation markers comparing BV+ vs BV− individuals. Overall, we found very few differences in CD8⁺ T cell phenotypes in the CX, VT, or PBMC of individuals with vs without BV (Table S3.4–6). Notably, we observed a significantly lower frequency of the activated phenotype (47) HLA-DR⁺CD38⁺ CD8⁺ T cells in the VT (median frequency 2.0% BV− vs 0.9% BV+, p_adj = 0.0309; Table S3.4) of BV+ compared to BV− individuals (Fig 3a), possibly indicating a reduction in CD8⁺ T cell activation in dysbiotic VT tissues. Additionally, we evaluated Granzyme B as a measure of the cytotoxic potential of CD8⁺ T cells (48) and found no significant differences in the frequency of Granzyme B⁺ CD8⁺ T cells (Fig 3b) or any other CD8⁺ T cell activation marker comparing BV+ and BV− individuals, (Table S3.4–6).

Figure 3. Few phenotypic alterations in CVT or circulating CD8⁺ T cells in individuals with BV.

Flow cytometry was used to examine CD8⁺ T cell phenotypes within different tissue sites as indicated. The frequency of HLA-DR⁺CD38⁺ (a), Granzyme B⁺ (b), CD69⁺CD103⁺ (c), and CD101⁺ (d) among the total CD8⁺ T cell population in CX, VT, and PBMC samples in BV− and BV+ individuals. Adjusted rank regression T test performed to compare frequencies in each tissue between BV− and BV+ individuals. Adjusted p-value displayed when p ≤ 0.05, and the p-value is labeled “ns” for not significant when adjusted p > 0.05.

Next, we evaluated CD103⁺CD69⁺ co-expression on CD8⁺ T cells as markers indicative of T_RM (24, 46). We hypothesized that CD8⁺ T_RM would play a critical role in immune modulations associated with BV but found that CD69⁺CD103⁺ expression on CD8⁺ T cells was not differentially expressed in BV+ vs BV− individuals in any of the tissue compartments (Fig 3c). In addition to canonical T_RM markers CD69 and CD103, we also stained for CD101, another marker associated with the tissue memory T cell signature (23). Similar to CD69 and CD103 expression, we observed no differences in the frequency of CD101⁺ CD8⁺ T cells in BV+ vs BV− individuals (Fig 3d). Our data suggests that, while CD8⁺ T cells comprise the majority of T cells in the VT of BV+ individuals, there is little difference in their cytotoxic potential, tissue-resident signature, or other functional markers compared with CD8⁺ T cells from BV− individuals.

Conventional CD4⁺ T cells display signs of dysfunction in VT of individuals with BV

To evaluate whether the reduced Tconv frequency among total T cells in the VT of BV+ individuals (Fig 2d) was related to differences in markers of effector function, we examined a wide range of phenotypic markers on Tconv, as we did for CD8⁺ T cells (Table S3.7–9). We found CD39⁺ frequency significantly increased on Tconv cells in VT samples from BV+ individuals (median frequency 15% BV− vs 30% BV+, p_adj = 0.0331; Table S3.7), with minimal differences observed in the CX and PBMC samples (Fig 4a). CD39 is a marker of metabolic stress (49) and may contribute to Tconv dysfunction in the VT of BV+ individuals.

Figure 4. BV is associated with an increase in Tconv with reduced functional potential as well as an increase in the fraction of Tconv with a T_RM phenotype in the VT.

Flow cytometry was used to examine Tconv cell phenotypes within different tissue sites as indicated. (a) The frequency of CD39⁺ (a), T cell factor 1 (TCF-1⁺) (b), CD69⁺CD103⁺ (c), and CD101⁺ (d) among the total Tconv T cell population in CX, VT, and PBMC samples in BV− and BV+ individuals. Adjusted rank regression T test performed to compare frequencies in each tissue between BV− and BV+ individuals. Adjusted p-value displayed when p ≤ 0.05, and the p-value is labeled “ns” for not significant when adjusted p > 0.05.

Additionally, we found a statistically significant decrease in the frequency of TCF-1⁺ Tconv in the VT of BV+ compared to BV− individuals (median frequency 30% BV− vs 24% BV+, p_adj = 0.041; Table S3.7), with no significant differences in the CX or PBMC (Fig 4b). TCF-1 is a stabilizing transcription factor implicated in the regulation of progenitor potential and promotion of T cell fate specification (50–56). Thus, our data suggests that Tconv cells in the VT of BV+ individuals have a reduced capacity to undergo differentiation or serve as proliferative progenitor cells.

Conventional CD4⁺ T cells display increased markers of tissue residency in VT samples from individuals with BV

Next, we analyzed tissue resident phenotypes in Tconv cells in all three sample types. We found a significant increase in CD69⁺CD103⁺ frequency of Tconv cells in VT from BV+ compared to BV− individuals (median frequency 24% BV− vs 38% BV+, p_adj = 0.0061; Table S3.7), with no differences observed in the CX or PBMC (Fig 4c). The frequency of CD69⁺CD103⁺ Tconv was enriched in both the VT and CX compared to PBMC, where it was negligible, as previously reported (21).

We also stained Tconv cells for CD101 to evaluate the frequency of CD101⁺ Tconv cells as another indication of the tissue residency signature. We found that CD101⁺ Tconv cells were significantly increased in the VT from BV+ compared to BV− individuals (median frequency 27% BV− vs 38% BV+, p_adj = 0.0118; Table S3.7) in contrast to the CX and PBMC (Fig 4d). CD101⁺ Tconv frequency was also higher in the VT and CX compared to the PBMC regardless of BV status, supporting a role for CD101 in mediating tissue localization (Fig 4d). These data suggest that BV promotes tissue-resident Tconv phenotypes in the VT of BV+ individuals.

Th17 cells display increased markers of tissue residency in VT samples from individuals with BV

Lineage-specific Th17 Tconv (Th17) cells have been shown to play a role in immune responses to extracellular bacteria (57–59). To investigate the role that vaginal bacterial dysbiosis has on Th17 cell modulation, we first analyzed the frequency of Th17 cells (defined as CD161⁺CCR6⁺ Tconv cells (60)) among total Tconv cells and found that BV+ individuals did not have increased frequency of this phenotype in any of the tissue types compared to BV− individuals (Fig 5a). However, when we evaluated Th17 cells for specific phenotypes, we found that the frequency of tissue-resident (CD69⁺CD103⁺ expressing) Th17 cells among total Th17 cells was significantly increased in the VT of BV+ compared to BV− individuals (median 49% BV− vs 59% BV+, p_adj = 0.0059, Table S3.7) (Fig 5b). In addition, the frequency of Th17 cells expressing CD101 was also significantly increased in the VT of BV+ compared to BV− individuals (median 47% BV− vs 65% BV+, p_adj = 0.0024, Table S3.7) (Fig 5c). The frequency of Th17 cells expressing the activation marker and HIV co-receptor CCR5 was not significantly different between BV− vs BV+ individuals, although notably, the median frequency of Th17 cells expressing CCR5 in the VT of BV+ individuals was 97% (Table S3.7) (Fig 5d). Therefore, Th17 cells almost invariably express the HIV coreceptor, CCR5, coinciding with previous reports that Th17 cells are viable targets for HIV transmission (61). These results demonstrate that BV driven increased tissue-resident or CD101-expressing VT Th17 cells may present another mechanism for increased HIV susceptibility in individuals with BV.

Figure 5. Increased fractions of VT Th17 cells have a T_RM phenotype in individuals with BV.

Flow cytometry was used to examine Th17 cell phenotypes within different tissue sites as indicated. (a) The frequency of Th17 T cells, defined as CD161⁺CCR6⁺ Tconv, among the total Tconv T cell population in CX, VT, and PBMC samples in BV− and BV+ individuals. (b) The frequency of CD69⁺CD103⁺ (b), CD101⁺ (c), and CCR5⁺ (d) among the total Th17 T cell population in CX, VT, and PBMC samples in BV− and BV+ individuals. Adjusted rank regression T test performed to compare frequencies in each tissue between BV− and BV+ individuals. Adjusted p-value displayed when p ≤ 0.05, and the p-value is labeled “ns” for not significant when adjusted p > 0.05.

Regulatory T cells may inhibit immune activation in the VT of individuals with BV

Regulatory T cells (Tregs) contribute to the homeostasis of the immune environment by suppressing immune activation (62). To further investigate inhibitory mechanisms associated with immune dysfunction, we characterized Tregs (CD3⁺CD4⁺CD25⁺CD127^lowFoxp3⁺, Fig S1) in individuals with or without BV (Table S3.10–12). Treg frequency among all CD45⁺ cells was not significantly different between BV− vs BV+ in any of the tissues examined (Fig 6a). Upon evaluation of Treg phenotypes, we found a significant increase in CXCR3⁺ Tregs in VT samples from BV+ compared to BV− individuals (median frequency 32% BV− vs 74% BV+, p_adj = 0.0039, Table S3.10), with no differences observed in CX or PBMC samples (Fig 6b). CXCR3 is associated with recruitment via chemokine signaling and activation (63) that may contribute to increased suppressive function of Tregs. We also analyzed Tregs for phenotypes differentially expressed in the VT Tconv population of BV+ individuals. We observed no significant differences in CD69⁺CD103⁺ Treg or CD101⁺ Treg frequency in VT, CX, or PBMC samples from BV− vs BV+ individuals (Figs 6c–d), nor any significant differences in other activation markers examined in the VT of BV− vs BV+ individuals (Table S3.10).

Figure 6. Regulatory T cell expression of CXCR3 in the VT is associated with BV.

Flow cytometry was used to examine regulatory T cell phenotypes within different tissue sites as indicated. (a) The frequency of regulatory T cells (Tregs; CD3⁺CD4⁺CD25⁺CD127^lowFoxp3⁺) among the total Tconv T cell population in CX, VT, and PBMC samples in BV− and BV+ individuals. (b) The frequency of CXCR3⁺ (b), CD69⁺CD103⁺ (c), CD101⁺ (d) among the total Treg population in CX, VT, and PBMC samples in BV− and BV+ individuals. Adjusted rank regression T test performed to compare frequencies in each tissue between BV− and BV+ individuals. Adjusted p-value displayed when p ≤ 0.05, and the p-value is labeled “ns” for not significant when adjusted p > 0.05.

BV promotes the expression of cytokines and chemokines in the CVT

To investigate the soluble immune factors associated with individuals with BV versus those without BV, we analyzed 71 cytokines and chemokines in serum and CVT fluid collected via menstrual cup. In the menstrual cup fluid, 61 different cytokines and chemokines met the criteria for continuous analysis, while 10 cytokines and chemokines were analyzed as a dichotomous outcome (detected/not detected). In the serum, 52 different cytokines and chemokines met the criteria for continuous analysis, 17 cytokines and chemokines were analyzed using the dichotomous outcome, and 2 cytokines were detectable in all individuals but had greater than 20% of samples out of range high so were excluded from the analysis (Table S4). We observed significantly higher concentrations of Leukemia Inhibitory Factor (LIF), IL-23, IL-12p70, M-CSF, and Eotaxin in the menstrual cup fluid collected from participants with versus without BV (Fig 7a). Additionally, we observed significantly lower concentrations of CXCL9 and thymus and activation regulatory chemokine (TARC) (Fig 7a) in participants with BV versus those without BV. In our analysis of serum cytokines, we observed no significant differences in quantifiable cytokines (Fig 7b). For cytokines that did not meet the criteria for quantitative analyses, SDF-1α/β was detected significantly more frequently in menstrual cup fluid collected from BV− vs BV+ participants (Fig 7c). While we observed significantly greater concentrations of IL-12p70 in cervicovaginal fluid from BV+ individuals, we observed a significantly lower proportion of detection of serum IL-12p70 in the presence versus absence of BV (Fig 7d), suggesting a distinct role for IL-12p70 in the CVT immune environment compared to circulation. Overall, differentially expressed cytokines were primarily observed in menstrual cup samples, as opposed to serum samples, highlighting that BV has the greatest effects on the local CVT immune environment.

Figure 7. BV drives unique cytokine production in CVT fluid.

Luminex was used to quantify cytokines and chemokines from CVT fluid or serum samples. Adjusted difference (BV positive - BV negative) for CVT fluid cytokines (a) and serum (b) that met the criteria for quantification of cytokine concentrations (greater than or equal to 80% of samples were detectable). Adjusted odds ratio (BV positive/BV negative) for CVT fluid cytokines (c) and serum (d) that did not meet the criteria for quantification of cytokine concentrations (less than 80% of samples were detectable). The 95% confidence interval is shown for all comparisons. Significant results when adjusted p≤0.05 comparing BV− vs BV+ are colored in orange and non-significant differences p>0.05 comparing BV− vs BV+ are purple.

Discussion

A critical fact underlying the HIV pandemic is that T cells serve as a double-edged sword: acting as targets for HIV infection and replication, while also mediating anti-viral functions required to fight the virus (64, 65). The findings presented here suggest that previous explanations for the observed increased risk of HIV infection in the context of BV may have focused too heavily on the target cell side of that sword without appreciating BV’s impact on altering T cell functionality.

In contrast to prior studies that identified increased CD3⁺CD4⁺CCR5⁺ HIV target cells in samples from the CVT lumen (7, 39, 40), our analysis of tissue sections revealed that CD3⁺CD4⁺CCR5⁺ HIV target cells are significantly reduced in the VT lamina propria of BV+ individuals. This seeming contradiction may be explained by the recognition that our tissue samples capture changes in cell constituents of the vaginal lamina propria that would be missed by sampling only the CVT lumen. Our results underscore that the abundance of CD3⁺CD4⁺CCR5⁺ HIV target cells within VT tissues may not be the primary mechanism for the observed increase in HIV susceptibility in the context of BV. Our application of high-parameter, high-throughput flow cytometry on cells isolated from VT and CX tissue biopsies as well as PBMCs documents that BV may alter T cell functionality in the VT mucosa in addition to identifying alternate HIV target cells that may be relevant for HIV susceptibility.

We observed an increased frequency of CD4⁺ Tconv cells expressing CD39 in the VT of BV+ individuals. The potential dysfunction of this T cell phenotype is suggested by prior studies showing CD39⁺ T cells have a decreased response to vaccines and an increased likelihood of undergoing apoptosis (49), raising the possibility that in the context of BV, VT CD4⁺ Tconv may exhibit increased intrinsic immunoregulation leading to reduced effector potential. This could lead to increased HIV susceptibility through an inability to mount an effective anti-viral response upon exposure to the virus.

We also observe a significant reduction of TCF-1⁺ CD4⁺ Tconv in the VT of BV+ individuals (Fig 4b). TCF-1 is necessary for follicular T cell differentiation (50–52) and has been shown to play a role in Th17 responses by inhibiting IL-17 production (55). Fewer progenitor CD4⁺ T cells implies that more T cells are terminally differentiated in the VT which may limit lineage-specific differentiation upon secondary exposure.

We also found that CD101⁺ Tconv cells, and specifically CD101⁺ Th17 cells, are more frequently observed in the VT of BV+ individuals (Fig 4d, Fig 5c). Less is known about the function of CD101⁺ T cells in comparison to the other phenotypes analyzed in this study, but CD101⁺ T cells have been associated with decreased expansion after adoptive transfer in mice (66), suggesting CD101⁺ Tconv cells are inhibited and less likely to undergo proliferation. Additionally, CD101 is associated with exhausted T cell phenotypes that result from chronic activation and can promote immune dysfunction (67–69). Thus, the increased abundance of CD101⁺ Tconvs and Th17 cells in the VT of BV+ individuals may also contribute to a dysfunctional Tconv immune response upon secondary exposure to HIV in the VT of BV+ individuals. Given the often recurrent and persistent nature of BV (44), chronic immune activation from repeated antigen exposure caused by BV-associated microbiota may lead to the progression of these observed dysfunctional phenotypes that could promote HIV susceptibility by limiting antiviral T cell function, expansion, and differentiation.

In addition to CD4⁺ T cells expressing CCR5, it has been shown that CD69⁺CD4⁺ T cells are associated with increased susceptibility to HIV infection in the CVT mucosal tissue (70). This suggests that the enrichment of T_RM in the VT of BV+ individuals may mediate increased HIV acquisition in the context of BV. In particular, the enriched Th17 T_RM population of the VT (Fig 5b), which highly expresses CCR5, provide a suitable target for HIV acquisition in BV+ individuals, consistent with previous reports that Th17 cells serve as primary target cells during vaginal SIV infection (71). In addition to the role CD4⁺ T_RM in CVT mucosa play in HIV infection, they also act as reservoirs for HIV viral replication which can occur during early HIV infection (72). The increased frequency of Tconv T_RM, and in particular Th17 T_RM, may not just greatly contribute to susceptibility to primary HIV infection, but may also promote viral replication in the VT of BV+ individuals.

Further supporting the hypothesis that Th17 cells play a role in BV-associated immune modulations, our cytokine analysis revealed significant increases in IL-23, IL-12p70, M-CSF, and LIF collected in CVT fluid samples from BV+ individuals (Fig 7a). IL-23 and IL-12p70 can affect a variety of pro-inflammatory responses, one being the enhancement of Th17-mediated inflammation (73, 74). IL-23 and IL-12p70 may activate Th17 responses in the context of BV. The activation of Th17 cells may then lead to the formation of Th17 T_RM. Additionally, M-CSF, produced by macrophages, has been shown to promote Th17 differentiation of CD4⁺ memory T cells (75), which may also contribute to the expanded Th17 T_RM pool in the VT of BV+ individuals. LIF, on the other hand, inhibits inflammatory responses including Th17-mediated immune responses (76). LIF may be produced in response to the activation of Th17 cell responses to limit overall Th17 cell differentiation. The reduction of TCF-1⁺ progenitor Tconv cells and the production of LIF in the VT could limit the differentiation of progenitor T cells to Th17 cells which could explain why we observe comparable frequencies of total Th17 cells in individuals with or without BV.

While CD8⁺ T cells were observed more frequently in the VT of BV+ individuals (Fig 2c), we only found one phenotypic difference in these cells that was associated with BV: a reduction in HLA-DR⁺CD38⁺ CD8⁺ T cells (Fig 3a). This suggests that although VT CD8⁺ T cells are abundant during BV, they are not overly active. Activated CD8⁺ T cells are thought to play a role in controlling HIV after infection (77, 78). The inhibition of CD8⁺ T cell activation observed in individuals with BV may relate to increased Tconv dysfunction, as CD4⁺ Tconv cells are known to enhance CD8⁺ T cell effector function (79). The interactions, or possibly lack thereof, between CD8⁺ T cells and dysfunctional Tconv cells in the context of BV, may thus also impact CD8⁺ T cell activation after HIV exposure and contribute towards increased HIV susceptibility in BV+ individuals.

In addition to the inhibitory function observed on Tconv cells, Tregs serve to extrinsically restrain T cell responses via active suppressive mechanisms. Expression of the chemokine receptor and activation marker CXCR3 is increased on VT Tregs in BV+ individuals. CXCR3 plays a role in T cell trafficking and is activated by the chemokines CXCL9 (MIG), CXCL10 (IP-10), and CXCL11 (I-TAC) (63). In our CVT fluid cytokine analysis, we observed a significant reduction of one of the ligands for CXCR3, CXCL9 (Figure 7a), as well as a trend towards lower concentrations of CXCL10. While other studies have also reported a reduction in local CXCL9 and CXCL10 in the CVT in BV+ individuals (19, 80–82), it is unclear how to interpret this reduction in chemokine ligands along with our finding that their CXCR3 chemokine receptor is increased on Tregs. One possible explanation is that chemokine production within the deeper tissue compartment that may be involved in recruiting cells expressing CXCR3 may not be captured in the CVT lumen. Nonetheless, activated CXCR3⁺ Tregs in the VT may inhibit immune activation and contribute to dysregulated phenotypes observed in the VT of BV+ individuals.

In our chemokine analysis, we observed a significant reduction of TARC in CVT fluid samples from BV+ individuals (Fig 7a). TARC has been associated with the recruitment of Th2 cells (83, 84), which also rely on TCF-1 to promote Th2 differentiation and regulation of function (54). Th2 cells are typically associated with allergic and anti-parasitic responses (85) and have been rarely studied in the CVT. While they are not thought to contribute a significant role in BV associated immune responses, the reduction of TCF-1 and TARC in BV+ individuals does suggest that Th2-specific responses may be reduced in the context of BV. The significantly increased expression of Eotaxin that we identified in CVT fluid samples of BV+ individuals (Fig 7a) may arise to compensate for reduced Th2-specific responses that recruit eosinophils (86). Eotaxin is primarily produced by epithelial and endothelial cells and enhances the recruitment of eosinophils (87). Eosinophils are also not typically involved in bacterial-specific immune responses and this likely has little impact on HIV susceptibility but this does further support the notion that T cell mechanisms are more dysfunctional in BV+ individuals.

In our CVT fluid cytokine analysis, we also observed increased detection of SDF-1α/β in BV+ individuals (Fig 7c). As a natural ligand for CXCR4, an HIV coreceptor (88–90), increased SDF-1α/β expression might be expected to increase HIV resistance by decreasing CXCR4 expression. Therefore, if SDF-1α/β plays a role in promoting HIV susceptibility in individuals with BV, the mechanism for this is unclear. Finally, we observed no significant difference in IL-1β in CVT samples from BV+ vs BV− individuals. Although IL-1β is frequently identified as increased in CVT fluid with BV, this is not a universal observation (91, 92). The factors that mediate altered expression of soluble immune mediators during BV have not been elucidated.

Our study had limitations: first, we focused on characterizing diverse T cell markers and a large panel of soluble immune mediators in the context of BV infection, leaving the relationship of BV with other components of the immune response unevaluated. One justification for limiting our evaluation to characteristics of CD3⁺ T cells is that these constitute the predominant lymphocyte population in the CVT mucosa (Fig 2a). A second limitation was that we collected 3-mm tissue biopsies that often captured low amounts of rare T cell subsets such as Tregs. This reduced our power to evaluate the role of such subsets in immune responses to BV. A final limitation of our analysis is that it was performed as an exploratory, hypothesis-generating analysis, and as such, used nominal P values adjusted for confounders but not discounted for multiple comparisons (138 immune cell subsets comparisons, and 71 soluble mediators comparisons). Thus, our findings should be confirmed through studies employing hypothesis-driven testing.

In summary, our study performed the most comprehensive evaluation to date of CVT tissue T cell subsets associated with BV. This data shows BV driven immune alterations have the most profound impact on the VT T cells and changes at this site may be most responsible for increased HIV susceptibility in BV+ individuals. Specifically, we identify dysfunctional T cell subsets including CD39⁺ Tconv and decreased progenitor TCF-1⁺ Tconv in the VT of BV+ individuals that could alter host response and contribute to increased HIV susceptibility by limiting the antiviral capabilities of VT T cells. Furthermore, we found the enrichment of Tconv T_RM, and specifically the Th17 T_RM subset, in the VT of BV+ individuals that may serve as targets for HIV infection and replication. Confirmation of these findings and elaboration on molecular mechanisms may identify novel immune interventions to reduce the risk of adverse outcomes associated with BV, including increased risk of HIV infection.

Methods

Participants, samples, and data collection

Samples and data for this analysis came from the Kinga Study (Clinicaltrials.gov ID# NCT03701802) which enrolled a total of 406 heterosexual Kenyan couples from Oct. 2018 through Dec. 2019 to evaluate how exposure to sexually transmitted infectious agents alters genital mucosal immune responses. Among these couples, 110 were HIV serodifferent (defined as a person living with HIV (PLWH) and their heterosexual HIV-exposed partner), with the remaining 298 couples involving partners who were both HIV-seronegative at enrollment. HIV serodifferent couples were excluded if, prior to enrollment, the PLWH had initiated antiretroviral therapy (ART) with resulting suppressed HIV viral load, or the HIV-exposed partner had initiated tenofovir-based pre-exposure prophylaxis (PrEP). ART and PrEP were provided to enrolled PLWH or HIV-exposed partners, respectively, upon enrollment.

Seven sample types were requested from all Kinga Study participants at enrollment and 6-month follow-up visits for evaluation of immune responses: cervicovaginal fluid via Softcup^® (menstrual cup); two 3-mm vaginal tissue biopsies with one cryopreserved for immunofluorescent imaging, and one cryopreserved for flow cytometry; two 3-mm ectocervical tissue biopsies processed in parallel to the vaginal biopsies; fractionated peripheral blood mononuclear cells; and serum were all collected (Note: genital samples and CVT fluid collection were deferred if participants were actively menstruating and participants were encouraged to return to clinic when not menstruating for collection of all sample types at the same timepoint). Swabs for BV testing were collected prior to biopsies, and BV was assessed by Nugent score with 0-3 defined as normal flora, 4-6 as intermediate flora, and 7-10 classified as BV (93). Additional demographic, epidemiologic, clinical, and sexual behavior data (including self-reported frequency of vaginal sex, condom use, and hormonal contraceptive use) were collected at all visits.

Laboratory testing at the enrollment and 6-month follow-up visits included HIV serology by Determine HIV 1/2 Rapid diagnostic Test (RDT) (Abbott Laboratories, Inc., Abbott Park, IL, USA) with confirmation of positive results with First Response RDT (Premier Medical Corporation Ltd., Kachigam, India), and further confirmation of RDT results using 4th generation HIV-1-2 Ag/Ab Murex EIA assay (DiaSorin, Inc., Kent, UK). Herpes simplex virus type 2 (HSV-2) serology was performed with HerpeSelect 2 ELISA IgG test (Focus Technologies, Inc., Cypress, CA) using an index value cut-off of ≥3.5 to improve test specificity (94–98).

To identify samples for the current cross-sectional analysis of the effects of BV on immune responses we focused on 245 participants living without HIV who were identified as female at birth. These consisted of all 44 participants who may have been exposed to HIV by their enrolled heterosexual sexual partners, and 201 whose enrolled heterosexual sexual partner was without HIV and whose data could support a variety of analyses, including the analysis of BV mediated immune responses. Immunofluorescent imaging was performed on enrollment samples from a subset who were HSV-2 seronegative; had HIV-uninfected partners; and if the participant was BV− by Nugent score, were also BV− by Amsel criteria (99) to prioritize clean comparison groups. All immunologic testing was performed blinded to participant exposure data.

Analysis of tissue samples by immunofluorescent microscopy

Sample collection:

Biopsies were collected using baby Tischler forceps at the lateral vaginal wall and the ectocervical os, embedded in optimal cutting temperature (OCT) compound and cryopreserved on dry ice.

Laboratory Analysis:

Fresh frozen tissue biopsies in OCT compound were sectioned 8 μm thick using a cryostat and placed onto frosted microscope slides. Serial sections were used for hematoxylin and eosin (H&E) staining and immunofluorescent (IF) staining. H&E stained slides were used to identify lamina propria and epithelium and to evaluate tissue integrity to ensure tissue sections met the criteria for analysis (intact lamina propria and epithelium).

Tissue sections for IF staining were fixed in acetone at −20°C for five minutes, dried for thirty minutes at room temperature, and rehydrated in tris-buffered saline with 0.05% tween (TBST). Slides were quenched using 3% H₂O₂ for twenty minutes at room temperature, rinsed with TBST, and endogenous binding sites were blocked using 1:10 fish gelatin blocking agent (Biotium, Cat# 22010) diluted in TBST for one hour. Unconjugated mouse anti-human CCR5 antibody (clone MC-5 (100) provided by M. Mack) at optimal dilution (0.3 μg/ml) was added for one hour, washed with TBST, and then goat anti-mouse antibody conjugated with horseradish peroxidase (polyclonal, Invitrogen, Cat# B40961) was added for thirty minutes. After washing with TBST, a solution of Tris Buffer (100mM, MilliporeSigma, Cat# 648315), optimally diluted tyramide (AF488, Invitrogen, Cat# B40953, diluted 1:100), and H2O2 (diluted 1:67,000) was then used for immunofluorescence and signal amplification. Slides were washed with TBST and a combination of optimally diluted, directly conjugated anti-CD3 (AF594, Clone UCHT1, R&D Systems, Cat# FAB100T, diluted 1:50) and anti-CD4 (AF647, Clone RPA-T4, BioLegend, Cat# 300520, diluted 1:25) antibodies were added for incubation overnight in the 4°C fridge. Slides were then rinsed in TBST and stained with DAPI (Invitrogen, Cat# D1306, diluted 2.5 ng/ml) for five minutes. Slides were washed with PBS, saturated briefly for ten seconds in ammonium acetate, saturated for ten minutes in copper sulfate, and then saturated again briefly for ten seconds in ammonium acetate. Slides were thoroughly rinsed in dH₂O, dried, mounted with Prolong Gold (Invitrogen, Cat# P36930) and a cover slip, and allowed to set for 24 hours.

Tissue sections were imaged within 48 hours of staining using a 20x/0.8 Pan-APOCHROOMAT air objective on a Zeiss Axio Imager Z2 microscope as part of a TissueFAXS system (TissueGnostics; Vienna, Austria) equipped with an X-cite 120Q lamp (Excelitas), and with DAPI (49000 ET), EGFP (49002 ET), Texas Red (49008 ET), and Cy5 (490006 ET) filter sets. Images were acquired using an Orca-Flash 4.0 camera (Hamamatsu) with TissueFAXS 7.1 software (TissueGnostics).

Tissue section image analysis

We quantified the density of CD3⁺, CD3⁺CD4⁺, and CD3⁺CD4⁺CCR5⁺ cells in the epithelium and lamina propria of CX and VT sections from individuals with or without BV. Images acquired using the TissueFAXS system were exported as multi-channel 16-bit tiff files of stitched regions. Cell segmentation, region of interest (ROI) selection (lamina propria vs epithelium), marker intensity thresholding, and statistical analyses were performed with a custom graphical user interface (GUI) developed in MATLAB (R2022b). Nucleus segmentation was performed by finding the regional maxima of the grayscale DAPI signal, followed by morphological thickening constrained by the binary DAPI mask. Threshold intensities for the markers of interest were set manually through the GUI. The boundary between the lamina propria and epithelium regions was segmented based on nuclear density, and further adjusted manually through the GUI. Finally, for each region (lamina propria and epithelium), the number of single-, double-, and triple-positive cells, as well as the density (number of cells per mm²) were extracted. Codes and an extended description of the GUI can be found at github.com/FredHutch/Kinga_Study_BV_MacLean.

Analysis of tissue samples by high-parameter flow cytometry:

Sample collection:

Biopsies were collected using baby Tischler forceps biopsies either at the lateral vaginal wall or the ectocervical os, placed in a cryovial containing fetal bovine serum at 4°C, and transported to the lab. In the lab, dimethyl sulfoxide (DMSO) was added to the cryovial to a final concentration of 10%, cryopreserved overnight at −80°C, and transferred to liquid nitrogen for long-term storage.

Peripheral blood mononuclear cells (PBMCs) were isolated from acid-citrate dextrose (ACD) whole blood by centrifugation and resuspension in Dulbecco’s Phosphate Buffered Saline (DPBS). Cells were centrifuged over Ficoll-Histopaque (Sigma-Aldrich, Cat# 10771) at room temperature. The buffy coat was collected and washed twice in DPBS at 4°C before live cells were counted on a hemacytometer with trypan blue. Cells were pelleted and resuspended in 10% DMSO in fetal calf serum at 4°C at a concentration of either 5-7 x 10⁶ cells/mL, or 10-15 x 10⁶ cells/mL. Cryovials with 1 mL of these PBMC suspensions were placed in controlled rate freezing equipment (Mr. Frosty) and stored overnight at −80° C. Cells were subsequently transferred to liquid nitrogen for long term storage and shipped in liquid nitrogen dry shippers to University of Washington for further analysis.

Tissue Processing for Flow Cytometry:

Cryopreserved PBMCs, ectocervical biopsies, or vaginal biopsies in liquid nitrogen were quickly thawed and transferred to 10% FBS (PBMCs) or 7.5% FBS (VT/CX biopsies) complete RPMI media (RP10 or RP7.5). PBMCs were spun down and resuspended in RP10. After sitting in warmed RP7.5 for 10 minutes, biopsies were incubated at 37 degrees C for 30 minutes with Collagenase II (Sigma-Aldrich, 700 U/mL) and DNase I (Sigma-Aldrich, 400U/mL) in RP7.5. Biopsies were then passed through a 100-mm cell strainer using a plunger to disrupt the tissue, washed, and resuspended in phosphate-buffered saline (PBS). PBMCs and cells isolated from tissues were stained immediately after processing for flow cytometry. Cryopreserved PBMCs from a healthy control donor (Seattle Area Control Cohort (SAC)) were used as a longitudinal technical control for all flow cytometry acquisitions (data not shown).

Flow Cytometry:

Immediately following isolation, cells were incubated with UV Blue Live/Dead reagent (Table S1) in PBS for 30 mins at room temperature. After washing, cells were then stained with biotinylated CXCR3 (Table S1) in 0.5% FACS buffer for 20 mins at room temperature. Cells were washed and then stained extracellularly with antibodies (Table S1) diluted in 0.5% FACS buffer and brilliant staining buffer (BD Biosciences, Cat# 563794) for 20 mins at room temp. Cells were fixed with Foxp3 Transcription Factor Fixation/Permeabilization buffer (ThermoFisher, Cat# 00-5521-00) for 30 mins at room temp. Cells were washed and then stained intracellularly with antibodies diluted in 1X Permeabilization Buffer (ThermoFisher, Cat# 00-8333-56) for 30 mins at room temp. Cells were then resuspended in 200 ml 0.5% FACS buffer and stored at 4 degrees until ready for use. Antibodies were titrated and used at optimal dilution. Staining was performed in 5-ml polystyrene tubes (Falcon, Cat# 352054). Analysis was performed using Flowjo software and we required at least 25 cells for analysis of phenotypes associated with the parent cell type and further downstream gating and analysis (Fig S1).

Cytokine and chemokine sample collection, and processing

Cervicovaginal fluid:

The menstrual cup was inserted into the vaginal canal beneath the cervix at the beginning of the study visit and remained in place for a minimum of 15 and a maximum of 60 minutes while other visit procedures were being conducted. Sample collection was deferred if the participant was menstruating. The participant was encouraged to move around while the menstrual cup was in place. Once removed, the menstrual cup and all CVT fluid contents were placed in a 50 mL sterile collection tube and transported to the lab on ice. CVT fluids were collected in the 50 mL collection tube by centrifuging at 1500 rpm (320xg) for 10 min at 4°C. The mucous and fluids were gently mixed and 200-300 μl was aliquoted with a 1000 mL pipet tip into each cryovial and stored at −80°C.

Serum:

Whole blood was collected in an SST vacutainer, allowed to clot for less than 2 hours, and centrifuged at 1300g for 15 minutes. Serum was aliquoted and stored at −80°C.

Soluble mediator assays:

Serum and menstrual cup (CVT fluid) aliquots were shipped on dry ice to Eve Technologies (Calgary, Alberta, Canada). All samples were measured upon the first thaw. CVT fluid samples were weighed and diluted in PBS at a ratio of 1g sample to 1mL PBS. CVT fluid samples were vortexed for 2 minutes and up to 500 μl was loaded into 0.2 μm low-bind spin filters (MilliporeSigma, Cat# CWLS01S03) and centrifuged at ~14,000 x g for 15 mins to remove the mucus from the samples prior to analysis. Levels of cytokines and chemokines from CVT fluid and serum samples were measured using the Human Cytokine Array/Chemokine Array 71-403 Plex Panel (Eve Technologies, HD71).

Statistics

For purposes of data analysis, BV+ was defined by Nugent score of 7-10. BV− was defined as normal flora by Nugent score 0-3. For flow cytometry and soluble mediator analyses, to maximize the size of our BV+ group, we used enrollment visits of participants who were BV+ at enrollment, augmented with the six-month visit from any additional participants who were BV+ at the six-month exit visit. Thus, our BV+ group consisted of one time point per individual, but in some cases, it was enrollment while for others it was at exit. Among the remaining individuals, enrollment visits from those with normal flora at enrollment served as the BV− reference group. For immunofluorescent imaging, all specimens were from enrollment.

Cell densities of CD3⁺, CD3⁺CD4⁺, and CD3⁺CD4⁺CCR5⁺ cells in tissue sections were compared between individuals with versus without BV using the Wilcoxon rank sum test. To estimate adjusted differences in the percentage of T-cells with specified markers from flow cytometry data by BV status, we used rank-based regression, a nonparametric method robust to outliers (101). To investigate associations of soluble mediators in serum and CVT fluid samples with BV status, we first determined whether, for each mediator, at least 80% of the specimens generated a quantifiable level (vs out of quantifiable range or missing for another reason). If at least 80% of results were quantifiable, we imputed out of range values by randomly selecting a value between the lowest observed value and half the lowest observed value (if out of range low), or by selecting the largest observed value (if out of range high). We then compared log₁₀-transformed levels of the mediator from individuals in BV+ vs BV− groups using a t-test and estimated differences in mean log cytokine concentrations from both serum and CVT fluid samples using linear regression. If <80% of the levels for a mediator were quantifiable, we categorized the mediator as detected vs not detected and estimated the odds ratio for the effect of BV on mediator detection using logistic regression. For both flow cytometry and soluble mediator data, results from VT and CX samples were adjusted for hormonal contraceptive use (yes, no and menstruating, no and amenorrheal), HSV-2 serology (positive, negative, indeterminate), HIV exposure (HIV status of sexual partner), and number of unprotected sex acts in the last thirty days (continuous), while results from PBMC samples were adjusted for hormonal contraceptive use (yes, no and menstruating, no and amenorrheal), only. In this exploratory work, we did not adjust p-values for multiple testing. Statistical analyses were conducted using R version 4.3.3.

Study approval

All participants provided written informed consent using documents reviewed and approved by the University of Washington institutional review board, and the Scientific and Ethics Review Unit of the Kenya Medical Research Institute.

Data availability

All supporting data are included in the supplemental figures and tables.

Supplementary Material

Supplement 1

Supplementary Figure 1. Flow cytometry gating to determine phenotype expression on individual cells. Cells were gated by forward scatter height (FSC-H) and forward scatter area (FSC-A) to filter out non-singlet cells. Then cells were gated by side scatter area (SSC-A) and forward scatter area (FSC-A) to isolate lymphocytes. Lymphocytes were analyzed by live/dead and CD45 expression. Live CD45⁺ cells were then analyzed for CD3 expression. CD3⁺ cells were then analyzed for CD4 and CD8 expression. CD8⁺CD4⁻ were analyzed for the phenotypes listed. CD4⁺CD8⁻ were analyzed for CD25 and CD127 expression. CD25⁻ cells were defined as conventional CD4⁺ T cells (Tconv). CD25⁺CD127^low were analyzed for Foxp3 expression. Foxp3⁺ among the CD25⁺CD127^low population were defined as regulatory T cells (Tregs). Tregs were analyzed for the phenotypes listed. Tconv were analyzed for the phenotypes listed. In addition, Tconv were analyzed for T-bet expression which defined Th1 cells, and CCR6⁺CD161⁺ co-expression which defined Th17 cells. Th1 and Th17 cells were also analyzed for expression of the phenotypes listed. For each gating analysis, samples were only continued for further analysis if the parent gate had a minimum of 25 cells.

Supplement 2

Table S1. Antibodies used for high parameter flow cytometry on biopsies and peripheral blood mononuclear cells.

Kinga Study Team:

University of Washington: International Clinical Research Center

Jairam R Lingappa (co-Principal Investigator and Protocol Chair), Justice Quame-Amaglo (study coordinator), Harald Haugen, Elena Rechkina, Daphne Hamilton, Matthew Ikuma, Marie Bauer, Xuanlin Hou, Zarna Marfatia, Katherine Thomas, Corinne Mar, Adino Tesfahun Tsegaye

Fred Hutchinson Cancer Center:

Jennifer Lund (co-Principal Investigator), Jessica Graham, Jessica Swarts, Sarah Vick, Nicole Potchen, Irene Cruz Talevara, Lakshmi Warrier, Finn MacLean, Elizabeth McCarthy

Kenya Medical Research Institute: Thika Partners in Health Research and Development (and Jomo Kenyatta University [JKUAT])

Nelly R. Mugo (Site Principal Investigator), Kenneth Ngure (Site Investigator), Catherine Kiptinness (site coordinator), Bhavna H. Chohan (Site Laboratory Director), Nina Akelo, Charlene Biwott, Stephen Gakuo, Elizabeth Irungu, Marion Kiguoya, Edith Kimani, Eric Koome, Solomon Maina, Linet Makena, Sarah Mbaire, Murugi Micheni, Peter Michira, Jacinta Nyokabi, Peter Mogere, Richard Momanyi, Edwin Mugo, Caroline Senoga, Mary Kibatha, Jelioth Muthoni, Euticus Mwangi, Philip Mwangi, Margaret Mwangi, Charles Mwangi, Stanley Mugambi Ndwiga, Peter Mwenda, Grace Ndung’u, Faith Njagi, Zakaria Njau, Irene Njeru, John Njoroge, Esther Njoroge, John Okumu, Lynda Oluoch, Judith Achieng Omungo.