Protozoan-Viral-Bacterial Co-Infections Alter Galectin Levels and Associated Immunity Mediators in the Female Genital Tract

Co-infections with sexually transmittable pathogens are common and more likely in women with disturbed vaginal bacteriome. Among those pathogens, the protozoan parasite Trichomonas vaginalis (TV) is most common after accounting for the highly persistent DNA viruses human papillomavirus (HPV) and genital herpes. The parasitic infection often concurs with the dysbiotic syndrome diagnosed as bacterial vaginosis (BV) and both are associated with risks of superimposed viral infections. Yet, the mechanisms of microbial synergisms in evading host immunity remain elusive. We present clinical and experimental evidence for a new role of galectins, glycan-sensing family of proteins, in mixed infections. We assessed participants of the HIV Epidemiology Research Study (HERS) at each of their incident TV visits (223 case visits) matched to controls who remained TV-negative throughout the study. Matching criteria included age, race, BV (by Nugent score), HIV status, hysterectomy, and contraceptive use. Non-matched variables included BV status at 6 months before the matched visit, and variables examined at baseline, within 6 months of and/or at the matched visit e.g. HSV-2, HPV, and relevant laboratory and socio-demographic parameters. Conditional logistic regression models using generalized estimating equations calculated odds ratios (OR) for incident TV occurrence with each log10 unit higher cervicovaginal concentration of galectins and cytokines. Incident TV was associated with higher levels of galectin-1, galectin-9, IL-1β and chemokines (ORs 1.53 to 2.91, p <0.001). Galectin-9, IL-1β and chemokines were up and galectin-3 down in TV cases with BV or intermediate Nugent versus normal Nugent scores (p <0.001). Galectin-9, IL-1β and chemokines were up in TV-HIV and down in TV-HPV co-infections. In-vitro, TV synergized with its endosymbiont Trichomonasvirus (TVV) and BV bacteria to upregulate galectin-1, galectin-9, and inflammatory cytokines. The BV-bacterium Prevotella bivia alone and together with TV downregulated galectin-3 and synergistically upregulated galectin-1, galectin-9 and IL-1β, mirroring the clinical findings of mixed TV–BV infections. P. bivia also downregulated TVV+TV-induced anti-viral response e.g. IP-10 and RANTES, providing a mechanism for conducing viral persistence in TV-BV co-infections. Collectively, the experimental and clinical data suggest that galectin-mediated immunity may be dysregulated and exploited by viral–protozoan–bacterial synergisms exacerbating inflammatory complications from dysbiosis and sexually transmitted infections.

Co-infections with sexually transmittable pathogens are common and more likely in women with disturbed vaginal bacteriome. Among those pathogens, the protozoan parasite Trichomonas vaginalis (TV) is most common after accounting for the highly persistent DNA viruses human papillomavirus (HPV) and genital herpes. The parasitic infection often concurs with the dysbiotic syndrome diagnosed as bacterial vaginosis (BV) and both are associated with risks of superimposed viral infections. Yet, the mechanisms of microbial synergisms in evading host immunity remain elusive. We present clinical and experimental evidence for a new role of galectins, glycan-sensing family of proteins, in mixed infections. We assessed participants of the HIV Epidemiology Research Study (HERS) at each of their incident TV visits (223 case visits) matched to controls who remained TV-negative throughout the study. Matching criteria included age, race, BV (by Nugent score), HIV status, hysterectomy, and contraceptive use. Non-matched variables included BV status at 6 months before the matched visit, and variables examined at baseline, within 6 months of and/or at the matched visit e.g. HSV-2, HPV, and relevant laboratory and socio-demographic parameters. Conditional logistic regression models using generalized estimating equations calculated odds ratios (OR) for incident TV occurrence with each log 10 unit higher cervicovaginal concentration of galectins and cytokines. Incident TV was associated with higher levels of galectin-1, galectin-9, IL-1b and chemokines (ORs 1.53 to 2.91, p <0.001). Galectin-9, IL-1b and chemokines were up and galectin-3 down in TV cases with BV or intermediate Nugent versus normal Nugent scores (p <0.001). Galectin-9, IL-1b and chemokines were up in TV-HIV and down in INTRODUCTION Co-infections with taxonomically diverse sexually transmittable pathogens are common and more likely in women with disturbed vaginal bacterial communities (Onderdonk et al., 2016;Brown and Drexler, 2020). Among those pathogens, the extracellular protozoan parasite Trichomonas vaginalis (TV) is most common after accounting for human papillomavirus (HPV) and genital herpes (Workowski, 2015). TV accounts for close to half of the annual incidence of nearly half-a-billion sexually transmitted infections (STIs) according to WHO estimates (WHO, 2012;Rowley et al., 2019). Adding to the public health care burden, TV is associated with adverse pregnancy outcomes, high-risk HPV genotypes and cancer (Fichorova, 2009;Yang et al., 2020), as well as with HIV acquisition (Cu-Uvin et al., 1999;Van Der Pol et al., 2008) and transmission (Kissinger et al., 2009). Bacterial vaginosis (BV), a common syndrome of disturbed vaginal microbiota, goes handin-hand with TV infection (Onderdonk et al., 2016) and is also associated with HIV acquisition (Atashili et al., 2008), shedding and transmission (Low et al., 2014) and with HPV infection (Oh et al., 2015;Yang et al., 2018;Liang et al., 2019;Brusselaers et al., 2019;Norenhag et al., 2019) and lower rates of HPV clearance (King et al., 2011). BV affects over 21 million (~30%) women of child-bearing age in the US alone (CDC, 2020). Mixed TV-BV infections are common (Brotman et al., 2010), and in HIV-infected women they can dramatically increase HIV shedding in the genital tract, with adjusted odds ratios (OR) as high as 18.63 (95% CI 6.71-51.72) when comparing women with TV and BV to those with neither (Fastring et al., 2014).
Despite the strong epidemiologic evidence for protozoanbacterial-viral synergisms in evading host immunity in the genital tract mucosa, the molecular mechanisms facilitating coinfections remain largely unknown (Malla et al., 2014). To gain insights into innate immunity breakdown by microbial synergisms, we conducted a nested case-control study utilizing participants from the large HIV Epidemiology Research Study (HERS), which enrolled 1,310 US reproductive age women and followed them over 7 years collecting comprehensive information on STI acquisition and relevant laboratory and socio-demographic variables (Cu-Uvin et al., 1999;Mayer et al., 2003;Tohill et al., 2004;King et al., 2011). We turned our attention to the galectin family of glycan-binding proteins (Sato et al., 2009) as innate immunity mediators yet underexplored in the human cervicovaginal environment despite growing experimental evidence of their involvement in HIV (Lanteri et al., 2003;Ouellet et al., 2005;Mercier et al., 2008;Sato et al., 2012) and TV pathogenesis (Okumura et al., 2008;Fichorova et al., 2016;Heiss et al., 2016). We tested the hypothesis that TV and mixed TV-BV and protozoan-viral (TV-HIV, TV-HPV) infections will be associated with specific patterns of altered galectin levels, which in turn would correlate with altered mediators of innate immunity, e.g. cytokines and chemokines in the cervicovaginal secretions. For the first time in this study we investigated experimentally whether concurrent exposure to TV and BV bacteria dysregulates galectin expression and whether the host-protozoan-bacterial interactions are modified by the protozoan viral endosymbiont Trichomonasvirus (TVV), which is commonly carried by the vaginal isolates of the parasite (Fichorova et al., 2017).

The HERS Cohort
The HERS cohort was started in 1992 before highly active antiretroviral therapy (HAART) became available. A total of 871 HIV-infected women and 439 high-risk HIV-uninfected women between the ages of 16-55 years were enrolled in four U.S. urban centers (Cu-Uvin et al., 1999;Mayer et al., 2003;Tohill et al., 2004). Women with a clinical diagnosis of AIDS were ineligible for enrollment. Women were followed prospectively with comprehensive clinical and laboratory examination every 6 months for up to 7 years. At each visit, a complete pelvic exam was performed that included collection of cervicovaginal lavage (CVL) and vaginal swabs. Polymorphonuclear (PMNs) and mononuclear (MNC) white blood cells (WBC) were enumerated per five oil-immersion microscopy fields. TV, Neisseria gonorrhea, Chlamydia trachomatis, HSV-2 and HPV were diagnosed as described before (Smith et al., 1997). The Nugent Gram-staining scoring system was used to define normal (score 0-3), altered (score of 4-6) and BV microbiota (score 7-10). A strength of the study was that the Nugent scoring for all samples was performed in a single well-qualified central laboratory (Dr. J. Sobel's laboratory at Wayne State University, Detroit, MI). Pregnancies, but not breastfeeding, were recorded. As part of a structured interview, women were asked to identify themselves as currently using pill/ oral contraceptive, Norplant, or Depo-Provera; if they answered "yes" to any of these, they were listed as taking hormonal contraceptives. Socioeconomic status was measured at enrollment by monthly income, source of income, education, employment, and health insurance coverage and type.

Matched Case-Control Nested Study Of Incident TV Infection
We defined cases as all HERS participants with documented acquisition of TV who had a CVL collected at a TV+ visit (confirmed by culture), following a prior 6-month visit at which all tests for TV were negative. Each of these TV positive visits post a TV negative visit is referred to as "incident TV infection". Some women transitioned from TV negative to TV positive status more than once. All such episodes were captured. Women who were negative in all tests for TV recorded throughout study follow-up served as controls. Each incident TV visit was randomly matched 1:1 to a visit by a control participant by race (White, Hispanic/ Latina, Black/African American, Native American, Asian) and by the following visit-specific matching criteria: age (within 10 years), HIV status (positive/negative), BV by Nugent categorization, hormonal contraceptive use, and hysterectomy status. We were able to identify 169 women with 223 incident TV infections and 147 control women with 223 matched TV-negative visits. We compared the women selected for the nested case-control study to the overall HERS cohort by all matched variables as well as a number of unmatched socio-behavioral and clinical laboratory parameters. The case-control sample was found to be well-representative of the overall cohort, which provided confidence for the generalizability of the nested case-control findings to the overall study population (Supplementary Tables 1, 2).

Isogenic TV Strains
A TV isolate that carries the endosymbiont Trichomonasvirus (strain 347v+) and its isogenic derivative strain (347v−) cured from the virus were obtained from Dr. John Alderete (Washington State University) (Provenzano et al., 1997). The status of Trichomonasvirus (TVV) infection of each isolate was confirmed as described and reported (Fichorova et al., 2012). Parasites were cultured in modified Diamond's medium supplemented with 10% heat-inactivated horse serum (HyClone Laboratory) and iron, as reported earlier (Gilbert et al., 2000).

Vaginal Bacteria
Lactobacillus gasseri, L. crispatus, L. jensenii, Gardnerella vaginalis and Prevotella bivia were originally isolated by vaginal swabs from women participating in various vaginal microbiota research studies (Onderdonk et al., 1987;Delaney and Onderdonk, 2001). These isolates were identified using phenotypic characteristics and established criteria (Manual of Clinical Microbiology, Washington (DC): ASM Press; 1995), and identification was confirmed using the Microbial Identification System for long chain fatty acid analysis (MIDI Inc., Newark, DE). Atopobium vaginae (ATCC BAA-55) was acquired from the American Type Culture Collection. L. crispatus and L. gasseri were chosen as common homeostatic Lactobacillus species representative of the healthy vaginal microbiota that are noninflammatory in contrast of the most common BV associated bacteria P. bivia, G. vaginalis and A. vaginae (Fichorova et al., 2013;Anahtar et al., 2015). Although also commonly found in the human vagina, L. iners was not included in our experimental homeostatic Lactobacillus panel because it has been associated both epidemiologically and causally with disturbed immune homeostasis and vaginal inflammation (Anahtar et al., 2015).

Co-Infection Model
TV-BV bacteria co-infection was conducted as previously described (Fichorova et al., 2013). In brief, vaginal epithelial cells grown to confluency were first colonized with bacteria for 24 h followed by removal of cell culture supernatants along with non-adherent bacteria. The colonized epithelial cells were then incubated for 24 h with cell culture medium control, TV 347v+ or TV 347v−. After 24 h incubation under conditions mimicking the vaginal microenvironment (Mitsubishi AnaeroPack, Fisher), cell culture supernatants were collected for assessment of galectin and cytokine levels while cells were harvested for viability assessment by Trypan blue.

Statistics
The subset of women in the HERS cohort selected for the casecontrol study were compared to those not in this study using t-tests and Fisher Exact tests (Supplementary Tables 1, 2). All immune mediators were log 10 -transformed prior to analysis, and values below the lower limit of detection were set to the lower limit. Due to the use of matched data in this study, odds ratios (OR) are used when comparing the cases and controls. ORs and 95% confidence intervals (CI) for each 1 log 10 unit higher concentration of the immune mediator with incident T. vaginalis (TV) were calculated by conditional logistic regression. To account for potential correlation among multiple longitudinal visits from the same woman, we fitted our conditional logistic regression models using generalized estimating equations (GEE) and employed robust standard errors throughout the analysis. This approach allows for consistent parameter estimation even when the correlation structure is incorrectly or incompletely specified (Liang and Zeger, 1986). Using separate models for each covariate to avoid collinearity, we also used conditional logistic regression to examine the OR associations between log unit higher immune mediator levels and incident TV when visits were stratified on BV, HIV, and other matched and non-matched variables, as well as to compare ORs within levels of these variables.
Correlations between the log 10 -transformed concentrations of immune mediators simultaneously measured in CVLs from all visits were calculated using Pearson's product moment correlation.
Within the TV positive subgroup only, we used linear regression to evaluate the association between levels of the inflammatory markers as a continuous outcome and several covariates. Again, to account for a possible correlation between multiple visits within a woman, we used generalized estimating questions and employed robust standard errors to construct 95% confidence intervals and calculate p-values (Liang and Zeger, 1986). Covariates examined within the TV positive group were the matched variables as well as unmatched variables that appeared to differ between the TV positive and TV negative women (Supplementary Tables 1, 2) as well as between TV positive and TV negative visits (Supplementary Table 4), including Nugent score categories, HIV, HPV, and HSV status, presence of genital tract WBC, smoking and alcohol use. The latter analysis was not done for the control visits, since they were selected to be similar to the matched incident-TV visits on confounders and therefore we could not consider them representative of the broader population of any TVnegative women.
The size of OR and 95% CI were similar in Black women who represented the majority of the women with incident TV selected for our case-control sample (148/169, 88%) ( Table 1) as well as the majority of overall HERS cohort participants (736/1310, 59%) and women infected with TV at baseline and throughout the study (445/566, 79%) ( Supplementary Tables 1, 2).
Higher Galectin-1 and -9 Are Associated With Incident TV Co-Infections With BV, HIV, HPV or HSV, While Galectin-3 Is Negatively Associated With Incident TV Co-Infections With HIV And Normal Bacterial Flora Separate conditional logistic regression models examined the association between each one log 10 higher level of immune mediator and incident TV when stratified by co-infection status at the case-control matched visit ( Figures 1A-D).
An interaction was observed between TV and concurrent abnormal vaginal microbiota ( Figure 1A). Incident TV was associated with higher RANTES, MIP-3a, MIP-1b, MCP-1 and galectin-1 only when BV-positive, and higher IP-10-only when positive for BV or Nugent score 4-6. Higher galectin-9, IL-1b and IL-8 were significantly associated with incident TV in each Nugent categories; however, the 95% CI for the OR shifted closer to 1.0 within each Nugent category shift closer to normal vaginal flora. The relationship between galectin-3 and incident TV was inverted, with significant negative association only when normal microbiota was present. The concurrent HIV positive status had less impact on the association between incident TV and immune mediators with the most notable exception of galectin-3 which was inversely associated with incident TV in the HIV-positive but not HIVnegative visits ( Figure 1B).
Higher levels of galectin-9, IL-1b, IL-8, RANTES and galectin-1 were significantly associated with incident TV in both HPV-positive and -negative visits while higher IP-10 was associated with incident TV in HPV negative visits only ( Figure 1C).
Higher levels of galectin-9, IL-1b and IL-8 were significantly associated with incident TV in both HSV negative and HSV positive visits; however, higher levels of RANTES and IP-10 and lower levels of galectin-3 were associated with incident TV in the HSV positive visits only ( Figure 1D).

Co-Infections, Cervicovaginal Leukocytes, Hysterectomy, And Socio-Behavioral Variables Affect Levels Of Galectins And Inflammatory Mediators At Incident TV
To identify factors that may affect levels of immune mediators and galectins associated with TV incidence, we examined nonmatched socio-behavioral and clinical laboratory parameters at baseline, 6 months prior to matched visit and at the matched 446 visits comparing cases to controls (Supplementary Table 4).
At the matched visit, incident TV was more likely to test positive versus negative for cervicovaginal MNCs (OR = 2.33; Median and inter-quartile ranges (IQR) of back-transformed geometric means of concentrations of immune mediators and odds ratios (OR) and 95% confidence intervals (CI) for incident T. vaginalis (TV) with each 1 log 10 unit higher level of immune mediator are calculated from conditional logistic regression models fitted using generalized estimating equations. Significantly different ORs are bolded and indicated by **p < 0.01 and ***p < 0.001. To further identify factors that may be driving the levels of galectins and inflammatory mediators among the TV incident visits only, we performed regression analysis with focus on matched variables (Nugent score, HIV status and hysterectomy) and non-matched variables that differed at the matched and 6-month prior visits. Only factors showing at least one significant association with the immune mediators are shown in Figure 2.
The co-presence of PMNs and MNCs compared to PMNs alone was associated with significantly higher levels of all immune mediators (p <0.001) except galectin-3, IP-10 and MIP-3a. The absence of any WBC was associated with significantly lower levels of all immune mediators including galectins-1 and -9 (p <0.001) but not galectin-3 which did not reach significance. These data suggested that WBC contribute to higher levels of galectin-1 and galectin-9 in incident TV ( Figure 2D).

Bacterial-Protozoan-Viral Synergisms Upregulate Galectin-1, -9 And Interleukin-1b, Downregulate Galectin-3 And Selectively Alter Chemokine Expression
To test the causality underlying epidemiologic TV-BV interactions, we applied an established experimental human infection model ( Figure 3A). In order to investigate the effect of mixed infection, we infected bacteria-colonized epithelial cells with TVV-positive parasites and their TVV-cured isogenic counterparts (Provenzano et al., 1997).
The presence of bacteria modified the effects of both virusinfected and cured protozoa and vice versa ( Figure 3B). The coinfection with TVV+TV and P. bivia synergistically upregulated  D) and hysterectomy (G) status at the time of immunologic assessment, and by cigarette smoking (E) and alcohol consumption (F) during the 6 months preceding the immunologic assessment. Differences were assessed using ordinary least squares linear regression (ANOVA) with *p < 0.05, **p < 0.01, and ***p < 0.001.

DISCUSSION
This study is the first to identify galectins as a molecular basis for protozoan-viral-bacterial synergisms at the mucosal interface. In our cohort, higher galectin-1 and galectin-9 levels correlated with markers of cervicovaginal inflammation, marked by higher levels of IL-1b and chemokines e.g. IL-8. In TV infected women they were associated with vaginal dysbiosis/ BV and with presence of monocytes in the mucosal secretions, which are additional risk factors for HIV acquisition and transmission (Alfano and Poli, 2002). Prior clinical studies have shown higher levels of IL-1b and IL-8 in women with concurrent BV and TV as compared to women with neither infection (Cauci and Culhane, 2007); however, we show for the first time higher levels of galectin-1, RANTES, and IP-10 to be associated with incident TV in women with BV only.
We provide new experimental evidence that galectin-1 and galectin-9 are synergistically upregulated in host epithelial cells by virus-infected TV and the BV pathobiont P. bivia, emphasizing the need to study the individual complements of the disturbed vaginal microbiome as drivers of immune imbalances. Our data suggest molecular mechanisms by which BV and particularly P. bivia may be driving susceptibility to . Then the bacteria-colonized epithelial cells were infected with TV protozoa harboring Trichomonas vaginalis virus (TVV+) or a derivative TV strain that was cured from the virus (TVV−). Bars represent mean and SD from triplicate cultures in one of three experiments. P values are from ANOVA with Bonferroni multiple comparison test. X, *p < 0.05; XX, **p < 0.01; XXX, ***p < 0.001, different from medium (med) control. Fichorova et al. Galectins in Vaginal Polymicrobial Interactions Frontiers in Cellular and Infection Microbiology | www.frontiersin.org August 2021 | Volume 11 | Article 649940 persistent TV infection. By upregulating epithelial galectin-1 expression P. bivia may contribute to the amplified galectin-1 levels in the natural TV-BV coinfection and thereby facilitate the adherence of TV to epithelial cells (Okumura et al., 2008), which is mediated by galectin-1 binding to the protozoan surface lipophosphoglycan (LPG) . The increased levels of galectin-1 and -9 also offer a molecular basis for bacterial-protozoan synergism with HIV and other viruses. Galectin-1 facilitates HIV attachment to macrophages (Ouellet et al., 2005;Mercier et al., 2008), and in turn, HIV-1 amplifies this effect via altering cell surface glycosylation (Lanteri et al., 2003). Galectin-9 also promotes HIV entry into host cells (Bi et al., 2011). Higher levels of galectin-9 may contribute to CD8+ T cell exhaustion thorough biding to Tim-3 (Merani et al., 2015). Knocking down galectin-9 in mouse models improved clearance of HSV infections (Merani et al., 2015) and we saw trend of higher galectin-9 levels in the TV-HSV co-infection but this association did not reach significance.
Individual BV-associated organisms upregulated IL-1b, IL-8, MIP-3a, IP-10 and RANTES in our in-vitro model and our experiments with isogenic virus-positive and virus-negative TV parasites supported previous findings of bacterial synergisms with soluble LPG and molecular domain and with cell-free endosymbiont protozoan viruses, leading to upregulation of IL-1b (P. bivia) and IL-8 (G. vaginalis, A. vaginae) (Fichorova et al., 2013). We have also shown that the genomic dsRNA originating from the protozoan TVV viruses can initiate viral stress response in human vaginal and cervical epithelial cells through TLR3/IRF3 signaling (Fichorova et al., 2012). The TLR3/IRF3 pathway activation can explain the galectin-9 upregulation by TVV as shown with a dsRNA viral mimic in vascular cells (Imaizumi et al., 2007). The stimulation of an antiviral immune response by TVV+TV alone includes anti-viral immune mediators e.g. IP-10, MIP-1b and RANTES (this study and (Fichorova et al., 2012). However, this potentially protective immune alarming function of the vaginal epithelium was dampened in the presence of TV co-infection with the BVsignature bacterium P. bivia providing a plausible mechanism for hampering HPV clearance in TV-BV co-infections. In the HERS study BV but not TV alone was associated with increased odds of prevalent and incident HPV and delayed HPV clearance (King et al., 2011). The strong (over two logs) synergistic upregulation of galectin-9 by TVV+TV and P. bivia and higher levels of galectin-9 which we observed in TV-HPV co-infections may provide an additional boost to persistence of high-risk HPV genotypes since galectin-9 expression has been significantly correlated with the presence of HPV type 16 or 18 in cervical cancer tissue (Punt et al., 2015). The innate anti-viral role of galectin-9 secreted in the cervicovaginal secretions and how it may be hijacked by vaginal pathogens is yet to be elucidated. Further studies are needed to investigate galectin-1 and galectin-9 mediated signaling in TV-BV-HPV interactions, to what extend they may be protective and whether our finding of suppressed innate immune responses in mixed HPV-TV infections versus TV alone relates to susceptibility to cervical cancer observed in women with TV (Gram et al., 1992;Viikki et al., 2000;Misra and Singh, 2006;Depuydt et al., 2010;Rodriguez-Cerdeira et al., 2012).
Furthermore, our study suggests for a first time a distinct role for galectin-3 which was adversely associated with TV infection both clinically and experimentally. Clinically, we observed that with higher soluble levels of galectin-3 in their cervicovaginal secretions women had significantly lower OR of having incident TV when having normal Nugent score (OR = 0.03, 95% CI = 0, 0.36) or co-infection with HIV (OR 0.39, 95% CI = 0.16-0.96) or HSV (OR = 0.22, 95% CI = 0.06-0.74) (Figure 1). These results supported that lower galectin-3 levels are associated with higher likelihood of TV incidence but that this association depends on bacterial and viral co-infection status. We have previously shown that the parasite alone, in the absence of endosymbiont virus and bacteria, depletes extracellular galectin-3 levels, which benefits the parasite survival by suppressing chemokine-driven host immune defenses . We now show experimentally that galectin-3 levels were further synergistically suppressed in TV co-infection with P. bivia (Figure 3), which provides a mechanism for clinically reduced cervicovaginal galectin-3 levels in TV-BV co-infected women compared to TV alone observed in our study ( Figure 2). Galectin-3 stimulates innate and adaptive immunity (Rabinovich et al., 2004;Stowell et al., 2008), and thus lowering the extracellular levels of galectin-3 would be expected to decrease the clearance or resistance to both bacterial and protozoan infection. HIV on the other hand hijacks cell-associated galectin-3 by driving its expression via the HIV tat protein (Fogel et al., 1999), which appears to supports the initial stage of HIV infection (Schroder et al., 1995) and viral budding (Wang et al., 2014). The TV parasite may not interfere with this initial stage of HIV infection because we have shown that experimentally it decreases extracellular, but not cell-associated, galectin-3 .
The role of P. bivia in the polymicrobial syndrome of vaginal immunity breakdown requires special attention. Women with high relative abundance of P. bivia in vaginal metagenome analysis were shown to be 19 times more likely to have a proinflammatory vaginal cytokine profile and~13 times more likely to acquire HIV, compared to all others in a South African cohort (Eastment and McClelland, 2018). Moreover, P. bivia is the single BV organism linked to HPV gene expression and metastatic cervical cancer (Lam et al., 2018). Our experimental model confirmed the causative role of this Gram-negative anaerobe as a modifier of anti-viral and anti-protozoan host immune response. Further studies are needed to investigate the role of galectin-1, -3. -9 and other galectins in BV-TV-HIV-HPV co-infections and their specific interaction with P. bivia.
Lastly, we observed suppressed vaginal immunity, including galectins-1 and -9, in women with incident TV who smoked or consumed alcohol. It is possible that tobacco or alcohol substances have a direct immunosuppressive effect that contributes to the higher risk of incident TV diagnosed in HERS participants reporting their use. Emerging clinical evidence implicates smoking in supporting BV-associated microbial communities, and cessation of smoking was suggested as a plausible pre-requisite for restoring healthy Lactobacillus dominance, providing a pathway for a causative effect on vaginal innate immunity (Brotman et al., 2014). Alternatively, smoking and alcohol consumption may be correlates of other risk factors, such as lower socio-economic status, stress and depression capable of acting as immunity modifiers, which we were unable to investigate in our sample. A further limitation of our study is that it included predominantly Black women with or at high risk for HIV and only non-pregnant women and we had limited information on methods of contraception. Thus, we were unable to assess galectins in polymicrobial infections in the context of race, ethnicity and associated effect modifiers, as well as in the context of pregnancy, breastfeeding and hormonal contraceptive use, previously associated with altered cervical immunity in women with BV, TV and viral STIs (Morrison et al., 2014). The results presented here may help assess the power needed for future studies to be conducted in diverse populations to address the role of socio-demographic factors and health disparities in galectin-mediated immunity underlying susceptibility to mixed protozoan-viral-bacterial infections.

DATA AVAILABILITY STATEMENT
The datasets for this study are publicly available through the Centers for Disease Control and Prevention (CDC) where the HERS database is maintained. A written request reviewed by CDC is required. To obtain details on how to request access contact the corresponding author Dr. Fichorova at rfichorova@bwh.harvard.edu.

ETHICS STATEMENT
The studies involving human participants were reviewed and approved by the Institutional Review Board (IRB) at each participating institution at HERS' four study sites (Bronx, NY; Baltimore, MD; Detroit, MI; and Providence, RI), and at the CDC and Brigham and Women's Hospital. Written informed consent to participate in this study was provided by the participants' legal guardian/next of kin.

AUTHOR CONTRIBUTIONS
RNF conceived the case-control and experimental study and drafted the manuscript. AKD provided the statistical method and analysis. RNF and HSY contributed to immunologic data acquisition. All authors contributed to the data interpretation and provided a critical review and approval of the manuscript.