Endocervical Regulatory T Cells Are Associated With Decreased Genital Inflammation and Lower HIV Target Cell Abundance

Regulatory T cells (Tregs) play important roles in tissue homeostasis, but few studies have investigated tissue Tregs in the context of genital inflammation, HIV target cell density, and vaginal microbiota in humans. In women from Nairobi (n=64), the proportion of CD4+ CD25+ CD127low Tregs in the endocervix correlated with those in blood (r=0.31, p=0.01), with a higher Treg frequency observed in the endocervix (median 3.8 vs 2.0%, p<0.0001). Most Tregs expressed FOXP3 in both compartments, and CTLA-4 expression was higher on endocervical Tregs compared to blood (median 50.8 vs 6.0%, p<0.0001). More than half (34/62, 55%) of participants displayed a non-Lactobacillus dominant vaginal microbiota, which was not associated with endocervical Tregs or CD4+ T cell abundance. In a multivariable linear regression, endocervical Treg proportions were inversely associated with the number of elevated pro-inflammatory cytokines (p=0.03). Inverse Treg associations were also observed for specific cytokines including IL-1β, G-CSF, Eotaxin, IL-1RA, IL-8, and MIP-1 β. Higher endocervical Treg proportions were associated with lower abundance of endocervical CD4+ T cells (0.30 log10 CD4+ T cells per log10 Treg, p=0.00028), with a similar trend for Th17 cells (p=0.09). Selectively increasing endocervical Tregs may represent a pathway to reduce genital tract inflammation in women.


INTRODUCTION
Genital inflammation defined by elevated levels of proinflammatory cytokines in mucosal secretions has been associated with~3-fold increased HIV acquisition and diminished topical PrEP efficacy in women (1,2). Additional work by our group has demonstrated that elevated cervicovaginal cytokines coincide with higher numbers of HIV target cells and a decrease in proteins associated with mucosal barrier function, providing a potential mechanism linking genital inflammation with HIV acquisition (3). In this model, inflammatory cytokines recruit immune cells, which in turn secrete proteins that contribute toward mucosal tissue damage and further cytokine release, perpetuating a cycle of impaired barrier function and more target cell recruitment for HIV to establish a localized infection (4). To disrupt this cycle, it is critical to understand the mechanisms that regulate inflammation and its pathogenic effects, which might contribute to interventions that can reduce the risk of HIV acquisition. While more attention has been directed toward the activation and memory status of CD4+ T cell subsets such as Th17 cells in the context of tissue inflammation in the cervix (5)(6)(7)(8)(9), little is known about regulatory T cells (Tregs) in this tissue, and the roles they might play in reducing inflammation and maintaining tissue homeostasis.
Tregs play a vital role in limiting excess immune responses and maintaining tissue immune tolerance and homeostasis (10). Tregs are generally characterized on the basis of their origin, with natural Treg (nTreg) developing in the thymus and induced Treg (iTreg) derived from conventional CD4+ T cells in blood (11,12). Tregs are defined by their memory and activation status, their phenotype, defined by expression of soluble and cell surface markers linked to their function, such as CTLA-4, PD-1, TGF-b, IL-10, and IL-35, their antigen specificity, and their relationship to other T helper subsets (13). While iTregs often predominate in tissues and mature through interactions with cytokines and commensal flora, the general consensus is the provenance of tissue Tregs is dependent on the tissue examined (14,15). Therefore, the balance and distribution of Tregs could have important implications for inhibition of specific inflammatory processes in tissues.
Fewer studies have focused on endocervical Tregs, with most focusing on their role in pregnancy/fertility outcomes (16)(17)(18). This is in part because Treg frequency may be influenced by sex hormones; with FOXP3 expression being increased in the ovulatory phase of the menstrual cycle (19), to increase odds of implantation. At the onset of fetal development, FOXP3+ Tregs migrate from peripheral blood to the maternal/fetal interface and mediate immune suppression (19). Three nTreg and iTreg subsets; CD25 hi FOXP3+, PD1 hi IL-10+, and TIGIT+ FOXP3 dim were observed in the decidua and these contribute to tolerance during pregnancy (20).
Taken together, these data suggest there are many knowledge gaps regarding Tregs and the genital inflammatory milieu. We hypothesized that increased frequency of cervicovaginal Tregs would correspond to a reduction of genital inflammation. We tested this hypothesis amongst female sex workers in Nairobi, correlating endocervical Tregs to inflammatory surrogates of HIV susceptibility and adverse reproductive health outcomes. We found that cervicovaginal Tregs were associated with decreased genital inflammation and reduced frequencies of HIV target cell abundance, suggesting that increasing this cell population may represent a mechanism for better control of genital inflammation.

Participant Characteristics
HIV uninfected women were enrolled from the Sex Workers Outreach Program (SWOP) clinic in Nairobi (n=64, Table 1). The median age was 28 years (IQR: [25][26][27][28][29][30], and median time in sex work was 3 years (IQR: 2-6). Most (86%) reported sex work as their primary mode of income, and nearly three-quarters (72%) had completed secondary or tertiary education. Participants reported a median of one sex act with both repeat (IQR: 0-2) and casual clients (IQR: 0-3) in the week prior to enrolment. Condom use was quite common with casual and repeat clients, with >90% reporting always using condoms with these types of sexual partners. Just under half of participants (42%) reported having a husband or boyfriend, and 'always' condom use was much lower in this context (15%). Over half (53%) reported vaginal douching, and 67% reported use of enhancing substances during sex. Prior pre-exposure prophylaxis (PrEP) access was relatively common at 55% (35/ 64), and approximately one-quarter (17/64, 27%) had reported

Endocervical Treg Associations With Behavioral and Reproductive Variables
We next correlated Tregs with sexual behavioural, demographic and reproductive health variables. Endocervical Tregs correlated inversely with duration of sex work (r=-0.277, p=0.027, Table 2).
The frequency of sex with repeat (r= -0.35, p=0.005) but not casual (r=0.038, p=0.766) sex partners correlated inversely with cervical Treg. History of stillbirth, miscarriage or spontaneous abortion was associated with lower frequencies of cervical Tregs (p=0.043). Lower Tregs were observed amongst women who used DMPA for contraception, compared to other methods, while the use of oral pills as contraception trended toward more endocervical Tregs; neither comparison was statistically significant. No other sociodemographic, clinical or behavioural associations with Tregs were observed.

Cervicovaginal Microbiome Profiles and Associations With Tregs
The relative abundance of cervicovaginal microbial communities was characterized by 16S rRNA sequencing of menstrual cup cell pellets. A majority of participants had a non-Lactobacillus dominant profile (n = 34, 55%, Figure 2A). Unsupervised hierarchical clustering differentiated the OTUs into three clusters, with Lactobacillus dominant as branch 1 (n = 28, 45%), a mixture Sneathia, Prevotella, and Gardnerella as branch 2 (n = 25, 40%), and a predominance of Gardnerella as branch 3 (n = 9, 15%). The most common Lactobacillus species in branch 1 was L. iners. The alpha diversity of branches 2 and 3

Endocervical Tregs and Genital Inflammation
We have previously defined genital inflammation based on the number of pro-inflammatory cytokines in the upper quartile (1,2). In this study, we quantified cytokine concentrations in cervical secretions collected via menstrual cups and found that approximately one-quarter (24%, n=15/62) had ≥ 3 proinflammatory cytokines in the upper quartile. We modeled number of elevated cytokines as a continuous variable. Although the number of elevated genital cytokines was not associated with endocervical Tregs prior to adjustment, an inverse correlation was observed in an adjusted linear regression model with endocervical Treg as the outcome (beta = -0.07, 95% CI: -0.13 to -0.01, p = 0.028, Table 3). To better understand which cytokines in the multiplex panel were most associated with endocervical Treg, we correlated each cytokine as a log 10 pg/ml with normalized Treg frequency; six cytokines were inversely associated with Treg in a multivariable analysis (Table 3), the strongest of which were the pro-inflammatory cytokine IL-1b (p = 0.00019) and its related antagonist IL-1RA (p = 0.005), the growth factor G-CSF (p = 0.00022), and the chemokines IL-8 (p = 0.02), MIP-1b (p = 0.031) and Eotaxin (p = 0.002). Clear step-wise decreases in concentration of each of these cytokines was observed by endocervical Treg tertile ( Figure 3).

Endocervical Tregs and HIV Target Cells
We also measured associations between Tregs and other T cell subsets in the same endocervical samples, particularly cells that have been described as highly susceptible to HIV infection (22). In the endocervix, higher Treg proportions were associated with lower abundance of CD3+ T cells (data not shown), CD4+ T cells, and Th17 cells defined by CCR6 and CD161 expression ( Figure 4). Endocervical CD4+ T cell abundance was inversely associated with endocervical Treg frequency in both unadjusted (p=0.000099) and adjusted models (beta = -0.27, 95% CI: -0.41 to -0.13, p=0.00028, Table 3). Similar associations were observed in models that adjusted for sexual exposure to different types of partners (not shown). The mean number of endocervical CD4+ T cells declined step-wise, from 3.1 to 2.7 to 2.3 log 10 CD4+ T cell abundance, a nearly 10-fold reduction in the number of CD4+ T cells from highest to lowest Treg tertile. Endocervical Th17 abundance showed an inverse trend with endocervical Tregs in unadjusted models (p=0.054), and remained as a similar trend in multivariable analysis (p=0.092).

DISCUSSION
Despite efforts to reduce HIV incidence using the currently available prevention, UNAIDS estimates that over 1.5 million new infections continue to occur annually, further underscoring the need to develop new tools meet UN targets to reduce HIV incidence by 90% by 2030 (23). Many of the biological factors implicated in HIV transmission have genital inflammation as a central feature (4). While the causes and consequences of inflammation have been the focus of many studies, less attention has been paid to natural regulatory mechanisms that might restrain inflammatory responses despite exposure to inflammatory stimuli. Here we show that Tregs are more frequent in the endocervix than the blood, and display high expression of both FOXP3 and CTLA-4. Furthermore, their frequency correlates inversely with both the concentrations of inflammatory cytokines and the abundance of HIV target cells in the genital tract. While it is not possible to determine causal relationships in this cross-sectional study, one interpretation of these data is that increased Tregs in tissue are able to control the inflammatory responses generated by epithelial and/or immune cells in response to inflammatory stimuli. If confirmed, inducing Treg using interventions such as low dose IL-2 could reduce genital inflammation (24,25), with important implications for preventing STIs and improving reproductive health. Epidemiological associations of endocervical Tregs involved reproductive health variables, in particular a history of miscarriage  or stillbirth. Although the study was not designed for this purpose, the inflammation-reducing potential of Treg could be investigated further, as inflammation has been associated with adverse reproductive health outcomes in addition to HIV (26). Sexual intercourse with repeat but not casual clients was also associated inversely with endocervical Tregs. The reason for this difference is unclear and requires further investigation. We were unable to evaluate condom use, as this is reported to be quite high (>90% 'always' with both client types). Understanding natural variation in endocervical Tregs between women is an important further research objective.
Our findings add to the literature around reproductive tract Tregs outside of the context of reproductive variables. Similar to previous work, we find Treg are proportionally increased in the endocervix compared to the blood, by approximately two-fold (27). Interestingly, with respect to pathogens, a balance is required between Tregs that protect against excess inflammation, facilitating protective immunity [as is the case with HSV-2 (28,29)] versus those that can be co-opted to limit host immunity as a persistence strategy, as has been suggested for gonorrhea (30). In the HIV acquisition context, the finding that women whose CD4+ T cell populations contained proportionally higher Treg had lower concentrations of inflammatory cytokines and HIV target cells suggests that Tregs may play a protective role by limiting HIV's ability to cross the mucosal barrier and establish local infection, both thought to be critical events in transmission success. Experiments to test this hypothesis formally are needed.
While Tregs may protect against HIV by limiting inflammation, they also regulate frequencies of Th17 cells (31), which we and others have shown to be preferential targets for HIV in the mucosa (5,7,8). In the present study, we observed an inverse trend between endocervical Treg and Th17; a larger, prospective study that induces Treg would be needed to determine the impact of increasing Treg on Th17 cells in this tissue. For statistical analysis, see Table 3.  Table 3.
We did not observe any associations between endocervical Treg and vaginal microbiota. While this needs to be confirmed in a larger, prospective study, one explanation is that Treg regulate the inflammatory response to non-optimal, bacterial vaginosis (BV) associated bacteria like Sneathia, Gardnerella, and Prevotella, rather than their presence in the mucosa. This could be an important consideration, as bacterial vaginosis, a condition marked by depletion of vaginal health associated Lactobacillus, is associated with increased risk of STI acquisition (32)(33)(34)(35), reproductive complications (36)(37)(38), and is highly recalcitrant to existing treatments (39), so modulating the detrimental effects of bacteria could be an alternative or additional approach to minimize the impact of BV. Similar to what we have previously shown in a cohort of adolescent girls and young women in Mombasa, Kenya (40), our current data also indicates that involvement in sex work might be associated with non-optimal vaginal microbiomes. Most women in SWOP displayed a non-Lactobacillus microbial profile characterized predominantly by Sneathia, Prevotella, and Gardnerella. Previous studies have shown that Lactobacillus crispatus induces an anti-inflammatory phenotype that included increased Tregs in the female reproductive tract (FRT) (41). However, this could not be confirmed in this study due to the infrequency of these 'optimal' bacteria in our study population. These data underscore the need to improve vaginal and reproductive health in key populations at risk of HIV acquisition.
Our study had some important limitations. The crosssectional design and modest sample size were sufficient to test associations between Treg and other variables including those related to inflammation, but cannot infer causation and provide limited opportunity to stratify the data extensively. Literature on tissue Tregs is expanding, and our flow cytometry panel was unable to fully phenotype emerging subsets, such as those coexpressing RORC, which regulate Th17 cells in the gut (42), or that co-express GATA3, which are important in repairing fibrosis in the skin. We did not collect ectocervical biopsies in this study, and Tregs in this tissue may differ from those recovered using endocervical cytobrushes.
In summary, our data provide important insights into mechanisms that may be critical in regulating inflammation and its pathogenic effects. Since the causes of genital inflammation are quite broad, with sources including sexual intercourse (40), intravaginal practices, bacterial/viral sexually transmitted infections (STIs) (43,44), and non-Lactobacillus vaginal microbiota associated with bacterial vaginosis (45), one advantage of inducing regulatory mechanisms to reduce inflammation is that this regulation could protect against inflammation regardless of the cause. The strongest immune associations with endocervical Treg included overall CD4+ T cell abundance, which are believed to be required for HIV to establish itself in the mucosa, and a number of cytokines including those that attract immune cells to the site. Further exploration of associations between Treg and other immune cell subsets may be critical to understand precise mechanisms by which these effects are manifested. These data suggest that inducing Tregs in the FRT might limit inflammation, and by extension, reduce HIV acquisition risk and improve reproductive health. This hypothesis would need to be evaluated in a clinical trial that also assessed safety and immunological/ microbiological endpoints in parallel.

Study Population and Ethics Statement
Female HIV sero-negative women (n=68) accessing care at Sex Workers Outreach Program (SWOP) clinic, Nairobi City, were enrolled in the study. The study was approved by the Kenyatta National Hospital -University of Nairobi Ethics and Research Committee. Prior to enrolment, written informed consent was obtained from all volunteers. A comprehensive questionnaire was then administered to establish each participant's sociodemographic, reproductive and behavioural characteristics.

Sample Collection and Processing
Cervicovaginal secretions were collected by insertion of a menstrual cup (SoftCup ™ ) into the vagina during administration of the study questionnaire for at least 1 hour. On removal, the menstrual cup was transferred to a 50ml tube and kept on ice. In the laboratory, the Soft Cup secretions were resuspended in a 1:4 ratio in cold phosphate buffered saline (PBS), centrifuged, supernatant aliquoted in cryovials and the pellet resuspended in RNA-Later. Both the supernatant and pellet were stored -80°C for cytokine and microbiome profiling respectively. Paired blood (10ml), in Acid Citrate Dextrose (ACD) tubes and cervical cytobrushes were collected from all enrolled volunteers for peripheral blood mononuclear cells (PBMC) and Cervical Mononuclear Cells (CMC) isolation respectively. Two cytobrushes were serially inserted into the endocervix, rotated through 360°transferred into a conical tube containing 3ml of PBS and placed on ice for processing within 2 hours of collection. In the laboratory, the tube containing cytobrushes was vigorously vortexed to dissociate the cells and mucus from the brushes, rinsed with R-10 medium (10% Foetal Bovine Serum (FBS) in RPMI containing penicillin and streptomycin) and filtered through 100-micron nylon cell strainer (Becton Dickinson) fitted into a 50 ml tube. Cells were washed twice by centrifugation at 1600rpm and pellet resuspended in washing buffer (PBS 2% FBS). PBMCs were isolated by Ficoll density separation and reconstituted in washing buffer. Ex vivo phenotyping of regulatory T cells (Tregs) and Th17 cells by flow cytometry was then performed on fresh CMC and PBMC.

Microbiome Analysis
We performed 16S rRNA sequencing of bacterial DNA isolated from menstrual cup cell pellets according to Illumina's 16S metagenomics sequencing library preparation guide using the MiSeq Reagent Kit v3 (600-cycle) and MiSeq sequencing platform according to the manufacturer's instructions (https:// support.illumina.com/documents/documentation/chemistry_ documentation/16s/16s-metagenomic-library-prep-guide-15044223-b.pdf). DNA was extracted using the DNeasy Blood & Tissue kit (Qiagen, Hilden, Germany). Prior to DNA isolation, cell pellets were homogenized in approximately 200 µL of PBS using pre-filled silica bead tubes (VWR, Radnor, Pennsylvania) and a Bead Ruptor Elite homogenizer (OMNI International, Kennesaw, Georgia) using the following settings: 3 cycles of bead beating at 6 m/s for 45 seconds followed by a 2 min dwell period. PCR amplification verification and library validation was done using a TapeStation electrophoresis system and D1000 ScreenTape reagents (Agilent, Santa Clara, California). Libraries were quantified using a Qubit 4 Fluorometer and the Qubit dsDNA HS assay kit (ThermoFisher Scientific, Waltham, Massachusetts). The QIIME2 platform was used to process sequence data, perform quality control, and generate a table representing the number of reads that mapped to specific operational taxonomic units (OTUs).

Statistical Analysis
Descriptive statistics including medians and interquartile ranges, and proportions, were used to summarize continuous and categorical variables. These clinical, sociodemographic, behavioural and reproductive health variables were also correlated to the proportion of endocervical Treg and Th17, and the abundance of Treg, Th17, and CD4+ T cells, using Spearman rank correlation and Chi-squared tests. Vaginal microbiome data, using genera relative proportions, were grouped based on unsupervised hierarchical Euclidean clustering, and results were visualized with R (package 'ggplot2', Wickham, v3.3.2, 2016), and correlated to Treg proportions. Centered log ratio (CLR) transformations of proportional cellular data (Cell proportions and bacterial relative proportions) were performed in R (package 'clr', Pierrot, v0.1.2, 2019) before Spearman rank correlation calculations with PRISM (v9.0.0). Unmapped bacterial sequences were removed from analysis, and the lowest abundance 52 taxa were grouped into "other" for visualization (average=0.4±0.7% per sample).
Associations with other immune variables were tested using multivariable linear regression, with log 10 transformed endocervical Treg proportions as the outcome. Separate models were used for each immune variable (number of cytokines, log 10 concentration of genital cytokines, and log 10 endocervical cell counts), adjusting for covariates associated with inflammation and/or Tregs, either in this study or in the literature. Models were adjusted for duration of sex work and not age, as the former was associated with endocervical Treg. Immune variables were graphed by Treg tertile for visualization purposes, but analysed as continuous variables. Graphing and visualization was performed using Prism version 9 (GraphPad Software, LLC), with additional analyses performed using R and SPSS v. 27.0.

DATA AVAILABILITY STATEMENT
The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found below: NCBI, PRJNA760281.

ETHICS STATEMENT
The studies involving human participants were reviewed and approved by Kenyatta National Hospital and the University of Manitoba. The patients/participants provided their written informed consent to participate in this study.

ACKNOWLEDGMENTS
We thank all the study participants and the clinical staff at the SWOP City clinic in Nairobi, Kenya who made the study happen. We acknowledge the Canadian Institutes of Health Research CIHR PJT-148796) and the Manitoba Medical Service Foundation (MMSF 8-2019-06) for their generous funding support. LRM is supported by a CIHR New Investigator Award.