Distinct Effects of the Cervicovaginal Microbiota and Herpes Simplex Type 2 Infection on Female Genital Tract Immunology

Summary

High-diversity vaginal microbiota was associated with inflammatory cytokines, while herpes simplex virus type-2 (HSV-2) infection correlated with CD4⁺ T-cell numbers and a distinct cytokine profile. This suggests that HSV-2 infection and the genital microbiota may influence human immunodeficiency virus susceptibility through independent biological mechanisms.

Abstract

Background.

Genital inflammation is a key determinant of human immunodeficiency virus (HIV) transmission, and may increase HIV-susceptible target cells and alter epithelial integrity. Several genital conditions that increase HIV risk are more prevalent in African, Caribbean, and other black (ACB) women, including bacterial vaginosis and herpes simplex virus type-2 (HSV-2) infection. Therefore, we assessed the impact of the genital microbiota on mucosal immunology in ACB women and microbiome-HSV-2 interactions.

Methods.

Cervicovaginal secretions and endocervical cells were collected by cytobrush and Instead Softcup, respectively. T cells and dendritic cells were assessed by flow cytometry, cytokines by multiplex enzyme-linked immunosorbent assay (ELISA), and the microbiota by 16S ribosomal ribonucleic acid gene sequencing.

Results.

The cervicovaginal microbiota of 51 participants were composed of community state types (CSTs) showing diversity (20/51; 39%) or predominated by Lactobacillus iners (22/51; 42%), L. crispatus (7/51; 14%), or L. gasseri (2/51; 4%). High-diversity CSTs and specific bacterial phyla (Gardnerella vaginalis and Prevotella bivia) were strongly associated with cervicovaginal inflammatory cytokines, but not with altered endocervical immune cells. However, cervical CD4⁺ T-cell number was associated with HSV-2 infection and a distinct cytokine profile.

Conclusions.

This suggests that the genital microbiota and HSV-2 infection may influence HIV susceptibility through independent biological mechanisms.

Despite the scale of the global human immunodeficiency virus (HIV) pandemic, less than 1 in 1000 vaginal exposures to HIV results in transmission [1], generally following infection of a single mucosal target cell [2]. Parameters such as cervicovaginal mucus, an intact epithelium, mucosal antibodies, and certain antimicrobial peptides (AMPs) play important defensive roles [3], but the host mucosal immune system is a double-edged sword. Vaginal levels of several AMPs, including cathelicidin LL-37 and the α-defensins, correlate with enhanced HIV acquisition [4], as do proinflammatory cytokine/chemokine levels [5]. The mechanism by which genital inflammation increases HIV risk is at least 2-fold. First, mucosal inflammation recruits activated CD4⁺ T cells [6–8], which express higher levels of the HIV coreceptor chemokine receptor 5 (CCR5) [9] and are preferential targets for HIV [10]. In addition, inflammation is associated with disruption of the genital epithelial barrier [7].

Bacterial sexually transmitted infections (STIs) increase genital proinflammatory cytokine levels [11] and HIV susceptibility by 3–5 fold [12, 13], while asymptomatic herpes simplex virus type-2 (HSV-2) infection increases HIV risk by 3-fold [14] in the context of increased mucosal CD4⁺ T-cell numbers [15, 16] but unaltered genital cytokines [17]. Therefore, diverse genital pathogens increase HIV susceptibility, at least in part through alterations in the genital mucosal immune milieu.

In addition to intermittent incursions by pathogens, all mucosal surfaces have a distinct resident microbiota [18]. Host–microbiome interaction is critical for numerous essential host functions [19], and may also be very relevant to HIV susceptibility. Bacterial vaginosis (BV), characterized by increased bacterial diversity, an overabundance of anaerobes, and a paucity of Lactobacillus spp. [20] is associated with increased HIV risk [21] and elevated genital proinflammatory cytokine levels [6, 11], potentially induced by specific BV-associated bacterial species [6].

African, Caribbean, and other black (ACB) women from sub–Saharan Africa (SSA) and North America have a particularly high risk of HIV infection [22], and the per-contact risk of HIV infection is approximately 6-fold higher in a woman from SSA [23]. Possible biological explanations include a higher prevalence of HSV-2 [24] and BV [25] in ACB communities. Therefore, the purpose of the current study was to assess the interactions between the cervicovaginal microbiota, genital immunology, and HSV-2 infection in ACB women.

METHODS

Participant Enrolment and Inclusion Criteria

ACB women were recruited from the Women’s Health in Women’s Hands Community Health Centre, Toronto, Canada. All participants provided informed written consent, and the study protocol was approved by the HIV Research Ethics Board at the University of Toronto. Exclusion criteria included infection by HIV-1/2, Neisseria gonorrhoeae, or Chlamydia trachomatis; age under 16 years; pregnancy; or self-report of a genital infection within the previous 3 months.

Study Protocol and Sampling

All women visited the clinic between days 10–18 after the last day of bleeding of their previous menstrual period. A questionnaire, blood, urine, and vaginal swab were collected. Undiluted cervicovaginal secretions were self-collected for 1–2 minutes using an Instead Softcup (Evofem, San Diego, CA), diluted in phosphate-buffered saline, and cryopreserved (–80°C). Two endocervical cytobrushes were sequentially inserted into the cervical os, rotated 360°, and kept on ice in Roswell Park Memorial Institute medium with 10% heat-inactivated fetal bovine serum (Sigma-Aldrich, Carlsbad, CA), 100 mg/mL streptomycin, 100 U/mL penicillin, and 1X-GlutaMAX-1 (Gibco, Grand Island, NY) media; filtered (100 μm), washed, and aliquoted for staining. Peripheral blood mononuclear cells (PBMCs) were isolated by Ficoll-Hypaque density centrifugation.

Coinfection Diagnostics

Vaginal smear Gram stains were assessed using Nugent criteria [20]. First-void urine was used to test for N. gonorrhoeae and C. trachomatis by nucleic acid amplification test (ProbeTech Assay, BD, Sparks, MD). HSV-2 serostatus was determined with HerpeSelect gG-1/2 ELISA (Focus Technologies, Cypress, CA; adjusted 3.5 threshold) [26].

Assays of Soluble Mucosal Immune Factors

Cervicovaginal levels of granulocyte-macrophage colony-stimulating factor (GM-CSF), interleukin (IL)–1α, IL-8, monocyte chemoattractant protein-1 (MCP-1), monokine induced by interferon-γ (MIG), macrophage inflammatory protein 3α (MIP-3α), regulated on activation, normal T cells expressed and secreted (RANTES), IL-10, IL-17, IL-1β, IL-6, interferon-inducible protein 10 (IP-10), macrophage inflammatory protein 1β (MIP-1β), and tumor necrosis factor α (TNF-α) were assayed using the Meso Scale Discovery (Rockville, MD) electrochemiluminescent ELISA (Supplementary Methods). ELISA was used to quantify secretory leukocyte protease inhibitor (R&D Systems, Minneapolis, MN), human neutrophil peptides 1–3 (HNP1–3), and LL-37 (HyCult Biotechnology, Uden, Netherlands), and human beta defensin-2 (HBD-2; Peprotech, Rocky Hill, NJ) (Supplementary Methods).

DNA Extraction and Sequencing

Total DNA was extracted from genital secretions by enzymatic cell wall digest and bead beating [25, 27]. Amplification of the V3-V4 regions of the 16S ribosomal ribonucleic acid (rRNA) gene used a dual-barcode system with fusion primers 338F and 806R [28]. Amplicons were sequenced on Illumina MiSeq (Illumina, San Diego, CA) using the 300-base-pair paired-end protocol. Sequences were processed, assigned a species/genera, and clustered in community state types (CSTs) (Supplementary Methods) [27]. Bacterial load was quantified using a previously validated in-house quantitative polymerase chain reaction assay [29], and the relative abundance of identified taxa and CST assignments for each sample are provided in Supplementary Table S1 (SRA362820).

Immune-Cell Phenotyping

Cervical cells and PBMCs were stained with 2 monoclonal antibody panels: (1) α4-fluorescein-isothiocyanate (FITC; Miltenyi Biotec, Bergisch Gladbach, Germany), CD4-ECD (Beckman Coulter, Marseille, France), CCR5-phycoerythrin (CCR5-PE), β7-allophycocyanin (β7-APC), CD38-AlexaFluor700, antigen D–related human leukocyte antigen (HLA-DR)-APC–cyanine 7 (Cy7), cluster of differentiation (CD)69-eFluor450 (BD Biosciences, Franklin Lakes, NJ), Live/Dead Aqua (Invitrogen), CD25-PerCP-Cy5.5, CD39-PE-Cy7, and CD3-eFluor650 (eBiosciences, San Diego, CA); and (2) blood dendritic cell antigen 2 (BDCA-2)-FITC (Miltenyi), CD207-PE (Beckman Coulter), dendritic cell–specific intercellular adhesion molecule-grabbing non-integrin (DC-SIGN)–peridinin chlorophyll protein (PerCP)–Cy5.5, CD206-APC (BD Biosciences), CD83-streptavidin, CD123-PE-Cy7, CD11c-AlexaFluor700, CD14-AlexaFluor780, CD1a-v450, CD3-eFluor650 (eBiosciences), and Live/Dead Aqua (Invitrogen). Cells were enumerated with a BD LSR-2 flow cytometer (BD Systems) and analyzed by a blinded researcher using FlowJo 9.3.2 software (Tree Star, Ashland, OR). Metadata are available in Supplementary Tables S2/S3.

Statistical Analyses

Intergroup differences were assessed using Fisher exact test for categorical variables or an independent samples T test or 1-way analysis of variance (ANOVA) (Tukey post-hoc testing) for continuous variables. Cytokine levels and number of cells/cytobrush were normalized through log₁₀-transformation. Bivariate correlations were assessed using the Pearson correlation. Elevated genital proinflammatory cytokines (“genital inflammation”) was predefined as having ≥3/7 proinflammatory cytokines (IL-1α, IL-8, MIP-3α, RANTES, IL-1β, MIP-1β, and TNF-α) in the upper quartile [7]. The association between specific phyla and genital inflammation was determined using general additive models and ordinary logistic regression linear and spline models. Principal component analysis (PCA) was used to segregate immune parameters into linearly uncorrelated principal components, with a PCA score ≥0.3 considered meaningful. Statistical analyses were performed using SPSS version 22 software (IBM, New York, NY) and the statistical package R (R Foundation for Statistical Computing, Vienna, Austria).

RESULTS

Participant Demographics

In total, 52 ACB women met inclusion criteria and were enrolled. The 16S rRNA gene amplification failed for 1 participant, who was excluded. Insufficient cells were present on the cytobrushes of 5 participants, who were included in all other analyses; likewise, there were insufficient samples for the assessment of AMPs in 3 participants. The median participant age was 34 years (range, 20–66; Table 1). Regular menstrual cycles were reported by 43/51 (84.3%) participants; 11.8% (6/51) were currently using oral hormonal contraception; 35.3% (18/51) reported having sex in the past week, with 38.9% (7/18) using a condom. Vaginal washing was reported by 11.8% (6/51) of participants. The prevalence of clinical conditions was: BV 13.7% (7/51; Nugent score ≥7), intermediate vaginal flora 29.4% (Nugent score 4–6; 15/51), HSV-1 infection (90.2%; 46/51), HSV-2 seropositivity (60.8%; 31/51), any vaginal human papillomavirus infection (33.3%; 17/51), and yeast infection (13.7%, 7/51).

Table 1.

Demographic Characteristics of Participants

Variable		All Participants (N = 51)	CST-I/II (n = 9)	CST-III (n = 22)	CST-IV (n = 20)
Age, median (range)		34 (20–66)	44 (27–57)	33.5 (20–66)	33 (25–61)
Postmenopausal, n (%)		8 (15.7)	3 (33.3)	2 (9.1)	3 (15.0)
Combined oral contraceptive, n (%)		6 (11.8)	1 (11.1)	3 (13.6)	2 (10.0)
Vaginal sex during past week, n (%)		18 (35.3)	2 (22.2)	11 (50.0)	5 (25.0)
	Condom usea	7 (38.9)	1 (50.0)	3 (27.3)	3 (60.0)
Intravaginal practices [douching, n (%)]		6 (11.8)	2 (22.2)	3 (13.6)	1 (5.0)
Vaginal Gramb	BV, n (%)	7 (13.7)	0 (0)	1 (4.5)	6 (30.0)c
	Intermediate, n (%)	15 (29.4)	1 (11.1)	2 (9.1)	12 (60.0)c
	Normal, n (%)	29 (56.9)	8 (88.9)	19 (86.4)	2 (10.0)c
HSV-1 prevalence, n (%)		46 (90.2)	8 (88.9)	21 (95.5)	17 (85.0)
HSV-2 prevalence, n (%)		31 (60.8)	5 (55.6)	13 (59.1)	13 (65.0)
HPV prevalence, n (%)		17 (33.3)	1 (11.1)	6 (27.3)	10 (50.0)c
Yeast prevalence, n (%)		7 (13.7)	1 (11.1)	5 (22.7)	1 (5.0)

All participants tested negative for HIV-1/2, Neisseria gonorrhoeae, or Chlamydia trachomatis and were sampled between days 10–18 after the last day of bleeding of their previous menstrual period.

Abbreviations: BV, bacterial vaginosis; CST, community state type; HPV, human papillomavirus; HSV, herpes simplex virus

^aAmong 18 participants reporting sex during the past week.

^bBased on Nugent score of 7–10 (BV), 4–6 (intermediate) or <4 (normal). BV analyses performed excluding the intermediate group, and intermediate analyses performed excluding the BV group.

^c P < .05 when compared to CST-I/II.

The Vaginal Microbiota: Diversity, Bacterial Load, and Clinical Validation

Participant genital microbiota grouped into 4 distinct community state types (CSTs; Figure 1): 13.7% (7/51) CST-I (most often dominated by L. crispatus), 3.9% (2/51) CST-II (L. gasseri), 43.1% (22/51) CST-III (L. iners), and 39.2% (20/51) CST-IV (characterized by a paucity of Lactobacillus and a diversity of anaerobes and strict anaerobes). No participants had a L. jensenii–dominant microbiome. Estimates of total cervicovaginal bacterial load did not vary between CST groups: CST-I (mean, 9.13 log₁₀), CST-II (8.56), CST-III (8.91), and CST-IV (9.10; P = .658), and total cervicovaginal bacterial load did not correlate with the Shannon diversity index (r = 0.108; P = .450). However, diversity was higher in participants classified as CST-IV (mean, 2.37) compared with CST-I/II (0.73; P < .001) and CST-III (0.66; P < .001; ANOVA P < .001). For subsequent analyses CST-I and CST-II were pooled, as both are considered to represent “healthy” vaginal microbiota, while CST-III may be more prone to transition into CST-IV [27, 30].

Figure 1.

Community state type distribution and bacterial phylotypes. Distribution of vaginal bacterial community state types in the overall study cohort (N = 51) (A), and a heat map of the relative abundance of the 30 most common bacterial phylotypes within the study cohort with the Shannon diversity index and Nugent score indicated (B).

CST distribution by 16S rRNA gene analysis was strongly correlated with Nugent score: 6/7 women with BV (Nugent score ≥7) fell within CST-IV (85.7%, 6/7), 1 within CST-III, and none in CST-I/II (likelihood ratio [LR] = 8.2; P = .016). Furthermore, 18/20 (90%) of participants in CST-IV had a Nugent score ≥4, compared to 11.1% and 13.6% within CST-I/II and III, respectively (LR = 32.93; P < .001). Total cervicovaginal bacterial load negatively correlated with age (r = –0.282; P = .045) and participants with BV (Nugent score ≥7) had a higher overall bacterial load (mean = 9.47 log₁₀) compared to those with no BV (Nugent score ≤3; mean = 8.93 log₁₀; P = .028). No other demographic factors assessed were associated with bacterial load, CST, or Shannon diversity index (all P > .05; data not shown).

The Cervicovaginal Microbiota and Soluble Genital Immune Factors

Of the 14 genital cytokines assayed, 3 were detectable in all participants (IL-1α, IL-8, MCP-1), and others in a subset: GM-CSF (25/59), MIG (57/59), MIP-3α (46/59), RANTES (49/59), IL-10 (18/59), IL-17 (39/59), IL-1β (57/59), IL-6 (51/59), IP-10 (46/59), MIP-1β (32/59), and TNF-α (15/59). Our primary analysis assessed the relationship of the genital microbiota with the presence or absence of genital inflammation (predefined based on cytokine levels [7]). Overall, 13/51 (25.5%) of participants demonstrated genital inflammation: 45% (9/20) of participants within CST-IV, 18.2% (4/22) within CST-III, and 0% (0/9) within CST-I/II (LR = 9.5; P = .024, Figure 2A). Genital inflammation was associated with microbiome diversity (mean Shannon diversity index = 2.02 vs 1.11; P = .019), but not vaginal bacterial load (mean = 9.24 log₁₀ vs 8.92; P = .078).

Figure 2.

Genital cytokines and the cervicovaginal microbiome. A, Association of elevated levels of vaginal proinflammatory cytokines with the cervicovaginal microbiome, particularly CST group IV. LR and associated P value indicated above the graph. Specific association of some individual cytokines (IL-1α and IL-1β (B, C) but not others (MIP-3α) (D) with the cervicovaginal microbiome. One-way ANOVA P values indicated above the y-axis of each graph and, if significant, post-hoc Tukey P values indicated above each pair-wise comparison. Solid line indicates the mean, and dotted lines indicate the lower limit of quantification of the assay.

Abbreviations: ANOVA, analysis of variance; CST, community state type; IL, interleukin; LR, likelihood ratio; MIP-3α, macrophage inflammatory protein 3α.

The association of genital inflammation with the cervicovaginal microbiome was particularly driven by elevated levels of the cytokines IL-1α, IL-1β, GM-CSF, and IL-10 within CST-IV (Figure 2). Interestingly, participants within CST-III had elevated levels of the chemokine IP-10 compared with both CST-I/II and CST IV (each P ≤ .01), with similar trends seen for MIG (P = .11 and 0.02, respectively). Furthermore, the presence of Gram stain–defined BV (Nugent ≥7; N = 7/51) was associated with elevated levels of IL-1α (3.88 vs 3.00 log₁₀ pg/mL; P = .002), but reduced MIP-3α (0.59 vs 1.51 log₁₀ pg/mL; P < .001), MCP-1 (1.50 vs 2.21 log₁₀ pg/mL; P = .015), RANTES (0.78 vs 1.15 log₁₀ pg/mL; P = .039), IP-10 (1.83 vs 2.64 log₁₀ pg/mL; P = .001), MIG (2.73 vs 3.43 log₁₀ pg/mL; P = .038) and MIP-1β (1.26 vs 1.75 log₁₀ pg/mL; P < .001) when compared to those with no BV (Nugent ≤7; N = 29/51).

Increased Shannon diversity was associated with elevated levels of IL-1α (r = 0.619; P < .001), GM-CSF (r = 0.301; P = .032), IL-10 (r = 0.382; P = .006), and IL-1β (r = 0.477; P < .001), and decreased levels of MCP-1 (r = –0.324; P = .020), MIG (r = –0.454; P = .001), and IP-10 (r = –0.537; P < .001). Only IL-1α concentrations correlated with total vaginal bacterial load (r = 0.380; P = .006), and no other associations were demonstrated between proinflammatory cytokines and demographic or clinical parameters. CST groups were not associated with differences in AMP levels, although total vaginal bacterial load and Shannon diversity index correlated with levels of secretory leukocyte protease inhibitor (r = 0.520; P < .001 and r = –0.312; P = .031, respectively) and HBD-2 (r = 0.373; P = .008 and r = –0.302; P = .037).

In terms of individual bacterial phylotypes, genital inflammation was associated with a decreased relative abundance of L. crispatus (P = .027), and with increased relative abundances of G. vaginalis (P = .005), P. bivia (P = .02), Aerococcus christensenii (P = .035), and Peptoniphilus harei (P = .045). No association was observed between genital inflammation and the relative abundance of L. iners (P = .99; Figure 3). Similar associations were seen with absolute abundances (data not shown).

Figure 3.

Genital inflammation and the relative abundance of specific bacterial phyla. General additive models and ordinary logistic regression linear and spline models were fit to inflammation status and each phylotype’s log₁₀ relative abundance for all phylotypes that were detected in at least 50% of all samples (n = 14 phylotypes). The resulting plots of inflammation risk as a function of log₁₀-transformed relative abundance of Lactobacillus crispatus (A), L. iners (B), Gardnerella vaginalis (C), and Prevotella bivia (D) are shown; the number of participants with detectable levels is shown above each. Horizontal gray lines indicate the estimated inflammation risk at the lowest detected relative abundance; if the 95% confidence region (shaded) does not contain this line, the inflammation risk is significantly associated with the relative abundance of the given phylotype.

Endocervical CD4⁺ T-Cell Subsets

Cervicovaginal CST was not associated with differences in the overall cervical CD4⁺ T-cell number (Supplementary Figure 1A), nor the proportion of genital or blood CD4⁺ T cells expressing CCR5, CD69, CD38/HLA-DR, or α4β7 integrin (Supplementary Figure 1B–1D). Specifically, CD4 numbers/subsets were not associated with CST-III, CST-IV, pooled CST-III and CST-IV, vaginal bacterial load, Shannon diversity index, the relative abundance of individual bacterial phylotypes (data not shown), or Gram stain–defined BV or intermediate vaginal flora.

While not associated with the cervicovaginal microbiota, endocervical CD4⁺ T-cell numbers did correlate with elevated genital levels of the specific cytokines/chemokines IL-8 (r = 0.489, P = .001; Figure 4A), MIP-3α (r = 0.428, P = .003; Figure 4B), and IL-6 (r = 0.365, P = .013). Interestingly, these cytokines tended to be distinct from those that had been associated with the cervicovaginal microbiota; specifically, IL-1α, IL-1β, GM-CSF, IL-10, IP-10, and MIG. Endocervical CD4⁺ T-cell numbers also strongly correlated with levels of the antimicrobial peptides LL-37 (r = 0.529; P < .001; Figure 4C) and HNP1-3 (r = 0.485; P = .001; Figure 4D).

Figure 4.

Correlation of specific cytokines and host defense peptides with cervical CD4⁺ T-cell numbers. Absolute number of endocervical CD4⁺ T cells/cytobrush correlated with the concentration of IL-8, MIP-3α, LL-37, and HNP1–3 (A–D) in undiluted cervicovaginal secretions collected by Instead Cup. Normalization was achieved through log₁₀ transformation of CD4⁺ T-cell numbers, and the Pearson product-momentum correlation was performed with the r and P values indicated.

Abbreviations: IL, interleukin; HNP1–3, human neutrophil peptides 1–3; MIP-3α, macrophage inflammatory protein 3α.

HSV-2 infection was associated with an increased cervical CD4⁺ T-cell number (mean 2.9l og₁₀ vs 2.5 log₁₀ in HSV-2–uninfected women; P = .028), as well as proportion of CD4⁺ T cells expressing CCR5 (48.6% vs 37.9%; P = .040). Cervical T-cell populations were not associated with other demographic variables (data not shown).

Endocervical Dendritic Cell Populations

Associations were assessed between the cervicovaginal microbiota and monocyte-derived dendritic cells (mDCs), monocytes, plasmacytoid dendritic cells (pDCs), and Langerhans cells (LCs). Endocervical pDCs, defined by the coexpression of BDCA-2 and CD123, were too infrequent for interpretation (median, 15 cells/cytobrush). No associations were apparent between CST, bacterial load, Shannon diversity index, phylotypes, or demographic variables and the absolute number of endocervical mDCs, monocytes, or LCs, or the proportion of these DC subsets expressing the C-type lectin receptors langerin, mannose receptor (MR), or DC-SIGN (data not shown). However, among participants reporting sexual intercourse within the past week, there was a positive correlation between the number of days since vaginal intercourse and the number of endocervical Langerhans cells (r = 0.506; P = .032), with more recent sex associated with reduced Langerhans cell numbers.

Distinct Immune Associations of HSV-2 Infection and the Cervicovaginal Microbiota

Overall, cervicovaginal CST-IV was associated with elevated proinflammatory cytokines (but not cells), while HSV-2 infection was associated with elevated cervical CD4⁺ cells (without cytokine alterations). Because many mucosal immune parameters were intercorrelated, immune alterations associated with vaginal CSTs and HSV-2 infection were further explored using PCA. Two PCA models were run: the first incorporated the Shannon diversity score as a marker of microbial diversity in the vaginal microbiota, log₁₀-transformed cervicovaginal cytokine levels (n = 14), and cervical CD4⁺ T-cell numbers (either total CD4⁺ cells or CCR5⁺CD4⁺ cells per cytobrush); the second model included log₁₀-transformed cytokine levels, cervical CD4⁺ T-cell numbers and 2 binary variables (HSV-2 infection and presence of CST-IV).

As expected, Shannon diversity was much lower in participants with CST-I/II and CST-III than with CST-IV (0.73 and 0.66 vs 2.37; P < .0001). In the first PCA model, 3 principal components were responsible for 76% of the dataset variance: the first was dominated by multiple cytokines (PCA score >0.3 for TNF-α, GM-CSF, IL-10, IL-1α, IL-1β, IL-17, IL-8, RANTES, MIP-1β, and MIG) and included Shannon diversity (0.54), the second only consisted of Shannon diversity (0.81) and IL-1α levels (0.37), and the third included cervical CD4⁺ cells (0.80) and the cytokines MIP-3α (0.81), IL-6 (0.74), MIP-1β (0.43), and IL-8 (0.37) (data not shown). Incorporating CCR5⁺CD4⁺ cells in the place of total cervical CD4⁺ cells yielded very similar results (data not shown). In the second PCA model, 3 principal components were responsible for 72% of the dataset variance, with even clearer divisions: the first component included multiple cytokines and CST-IV, the second component only included CST-IV (0.63) and IL-1α (0.38), and the third only HSV-2 status (0.67) and cervical CD4⁺ cell number (0.84). Therefore, there were quite distinct immune and clinical associations of cervical CD4⁺ T-cell numbers and of cervicovaginal microbiota diversity.

DISCUSSION

In keeping with previous studies of black women from both the United States and South Africa [6, 25], we found that most HIV-uninfected ACB women from Toronto either had a vaginal microbiota dominated by L. iners (CST-III), or lacking Lactobacillus spp. and characterized by increased diversity (CST-IV). These 2 CSTs were associated with the presence of genital inflammation, and also with elevated levels of several individual proinflammatory cytokines. Specifically, CST-IV was strongly associated with elevated IL-1α and IL-1β, and to a lesser degree GM-CSF and IL-10, while CST-III was associated with elevated IP-10 and MIG. Interestingly, levels of the latter 2 chemokines were reduced in women with CST-IV, with even more substantial reductions of these and other chemokines in women with BV by Nugent score, an unexpected finding that may merit further investigation. Further, genital inflammation was associated with specific bacterial phyla, most strikingly with a decreased relative and absolute abundance of L. crispatus, and increased relative and absolute abundances of several individual anaerobic bacterial phyla, including G. vaginalis and P. bivia. Neither CST-III/IV, BV, nor intermediate microbiota (defined by Nugent score) were associated with alterations in the number or phenotype of endocervical CD4⁺ T cells, endocervical dendritic cell numbers, or their subsets. We did not observe any association between vaginal CST and HSV-2 infection, despite the significant epidemiological synergy that exists between BV and HSV-2 infection [31–33].

The preexisting mucosal immune milieu at the site of sexual HIV exposure is a key determinant of HIV acquisition risk, with relative immune quiescence being associated with reduced risk and genital inflammation with enhanced HIV susceptibility [34]. Mucosal inflammation may increase HIV risk in at least 2 ways. Proinflammatory cytokines such as IL-1α and TNF-α are produced after stimulation of innate immune receptors on both epithelial cells and local dendritic cells [35], and directly disrupt mucosal epithelial barrier function both in vitro and in vivo, potentially facilitating HIV entry [7, 36]; furthermore, inflammation is associated with an influx of activated CD4⁺ T cells that fuels local HIV replication [8]. Our analyses now suggest that these 2 mechanisms can operate independently of each other. Vaginal CST-IV was associated with mucosal proinflammatory cytokines, particularly the cytokine IL-1α that is produced via microbial stimulation of Toll-like receptor 2 and Toll-like receptor 4 on local antigen presenting cells [6, 37], without alterations in CD4⁺ T-cell targets. However, HSV-2 infection was associated with increases in mucosal CD4⁺ T-cell subsets without alterations in local proinflammatory cytokines; while some cytokines/AMPs were associated with increased CD4⁺ T cells (MIP-3α, IL-6, IL-8, LL-37, and HNP1-3), these were generally distinct to cytokines associated with CST-IV. Together, these results suggest that HSV-2 and vaginal CST-IV enhance HIV susceptibility through independent mucosal immune mechanisms.

These findings have clear relevance for the ACB community, where HSV-2 prevalence is high [38] and vaginal CSTs associated with inflammation are more common [6, 38]. Over 92% of our study participants (48/52) were either HSV-2 seropositive or had a CST-III/IV vaginal microbiota, and 50% (26/52) demonstrated both. Whether having both conditions will synergistically increase HIV susceptibility is not known. There was a nonsignificant trend to increased endocervical CD4⁺ T-cell subsets in participants with both conditions, but we could not comment on potential enhanced HIV susceptibility due to alterations in the epithelial barrier, such as breakdown of tight junctions, resulting in increased permeability. However, even if such synergy exists, it is not clear if/how this could be translated into effective HIV prevention strategies. While vaginal proinflammatory cytokines are significantly reduced 4 weeks after standard clinical treatment of BV [39], the clinical ability to alter the vaginal microbiota in the long term is limited. Specifically, clinical BV recurs soon after standard metronidazole treatment in one-third of women, and altered microbiota (by Nugent score) recurs in almost two-thirds of women [40], demonstrating that clinically induced “normalization” of the microbiota tends to regress to an individual’s mean. While the presence of rectal Lactobacillus spp. is associated with reduced rates of BV [41], the success of oral probiotics as BV treatment or prevention has been mixed [42, 43]. Therefore, better ways to restore and maintain a Lactobacillus-dominated microbiota are urgently needed.

These findings reflect expected levels of vaginal cytokines [44], and are broadly consistent with prior work demonstrating that vaginal dysbiosis is associated with a proteomic profile characterized by elevated proinflammatory cytokines and altered AMPs [45]. Prior studies have also associated Nugent-defined BV with increased proinflammatory cytokines, particularly IL1α and IL1β, as well as decreased levels of IP-10 [11, 46]. However, while vaginal proinflammatory cytokines have been associated with increased mucosal CD4⁺ T-cell numbers [6, 7], a direct association between the vaginal microbiome and cervical CD4⁺ T cells has not been shown [6].

Limitations to our study include a modest sample size of just over 50 participants. While this reduces our ability to definitively demonstrate the lack of association between the vaginal microbiota and cervical HIV target cells, the fact that we still see strong cellular associations with HSV-2 infection clearly demonstrates that a vaginal microbiota association with genital target cells (if it exists) must be less robust. Further, while no association was seen between chronic HSV-2 infection and the vaginal microbiota, we could not assess the impact of HSV-2 reactivation or acute HSV-2 infection on the microbiome or genital immune parameters. This study only enrolled participants from the ACB community, reducing our power to assess the immune impact of vaginal CSTs dominated by Lactobacillus spp. other than L. iners (CST-I, -II, -V). Our analysis of genital cellular immunity assessed cellular phenotype rather than function, due to limitations in the number of cells collected by cytobrush sampling. Whether cervical T cells in women with a vaginal CST-IV demonstrate altered function is an interesting question that will require future study.

In summary, we found that the cervicovaginal microbiota in ACB women from Toronto was predominated by CSTs associated with genital inflammation. Interestingly, the mucosal immune impact of this dysbiosis was quite distinct from that of HSV-2 infection, correlating with increased proinflammatory cytokines but not with increased HIV target cell populations.

Supplementary Data

Supplementary materials are available at The Journal of Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.

Notes

Acknowledgments. The investigators acknowledge the kind assistance of the staff at the Women’s Help in Women’s Hands Clinic, and the time and cooperation of all study participants.

Financial support. This work was supported by the Canadian Institutes of Health Research (R. K., grants no. TMI-138656 and OCH-131579; B.S., studentship) and Ontario HIV Treatment Network (T. J. Y., studentship). R. K. is supported by a University of Toronto–Ontario HIV Treatment Network (OHTN) Endowed Chair in HIV Research. J. R., B. M., M. S. H., and P. G. are supported in part by the National Institute of Allergy and Infectious Diseases of the National Institutes of Health (awards no. U19AI084044 and R01AI116799).

Potential conflicts of interest. All authors: No reported conflicts. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.