Veillonellaceae family members uniquely alter the cervical metabolic microenvironment in a human three-dimensional epithelial model

Bacterial vaginosis (BV) is a gynecologic disorder characterized by a shift in cervicovaginal microbiota from Lactobacillus spp. dominance to a polymicrobial biofilm composed of diverse anaerobes. We utilized a well-characterized human three-dimensional cervical epithelial cell model in conjunction with untargeted metabolomics and immunoproteomics analyses to determine the immunometabolic contribution of three members of the Veillonellaceae family: Veillonella atypica, Veillonella montpellierensis and Megasphaera micronuciformis at this site. We found that Veillonella spp. infections induced significant elevation of polyamines. M. micronuciformis infections significantly increased soluble inflammatory mediators, induced moderate levels of cell cytotoxicity, and accumulation of cell membrane lipids relative to Veillonella spp. Notably, both V. atypica and V. montpellierensis infections resulted in consumption of lactate, a key metabolite linked to gynecologic and reproductive health. Collectively our approach and data provide unique insights into the specific contributions of Veillonellaceae members to the pathogenesis of BV and women’s health.


INTRODUCTION
Human mucosae are colonized by diverse and dynamic bacterial communities that impact health and homeostasis or contribute to disease states, depending on the compositional nature of the communities 1 . In the cervicovaginal microenvironment, Lactobacillus dominance is associated with optimal reproductive health 2 . Lactic acid, a by-product of Lactobacillus fermentation, is an important metabolite in maintaining host defense and protecting the cervicovaginal mucosal epithelium from infection by sexually transmitted pathogens and opportunistic commensal bacteria 3 . Overgrowth of diverse anaerobic bacteria and depletion of Lactobacillus spp. is characteristic of bacterial vaginosis (BV) in the cervicovaginal microenvironment 4,5 . Our group and others have developed a hypothetical model suggesting that early colonizers, such as Gardnerella vaginalis and Prevotella bivia, establish biofilm scaffolds over the cervicovaginal epithelium and exhibit low pro-inflammatory properties 4,6 . This model proposes that as the biofilm matures, Lactobacillus spp. become less abundant and secondary colonizers associate with the biofilm, which results in inflammation 4 . As Lactobacillus spp. are depleted, BV-associated bacteria produce polyamines that result in increased pH 4 .
BV is a common and highly recurrent cervicovaginal disorder 7 . Additionally, BV has been associated with adverse obstetric and gynecologic outcomes, such as preterm birth, increased acquisition of sexually transmitted infections, and possibly gynecologic cancer [8][9][10][11][12][13] . It is therefore imperative to better understand the pathophysiological mechanisms of BV. Key bacteria that have been the focus of BV studies thus far include G. vaginalis, P. bivia, Sneathia amnii and Atopobium vaginae, although more studies are required in order to completely elucidate contributions of the cervicovaginal microbiome to women's health 10,14,15 . However, other vaginal commensal bacteria and BV-associated bacteria have not been studied in detail and their roles in health or BV remain obscure. For example, members of the Veillonellaceae family, such as Veillonella atypica, Veillonella montpellierensis and Megasphaera micronuciformis, are understudied bacteria that have been isolated from the female reproductive tract 16,17 .
V. atypica is commonly isolated from the oral cavity and has been studied extensively at this anatomical site [18][19][20][21][22] . In the oral microbiome, V. atypica is considered a 'bridging species' that enables the colonization of middle and late colonizers to the oral plaque biofilm, with the aid of initial colonizing species, therefore V. atypica promotes oral biofilm development [23][24][25] . In addition, V. montpellierensis, another Veillonella species, has been isolated from patients suffering from endocarditis 26 and acute otitis media infections 27 . Similar to V. atypica, M. micronuciformis has also been isolated from the oral and gut microbiomes and is implicated in gastric cancer and response to chemotherapy [28][29][30] .
M. micronuciformis and other Megasphaera spp. are typically isolated in relatively high frequency from patients with BV, along with Atopobium, Leptotrichia/Sneathia, and Eggerthella-like species 31 . In contrast, Veillonella spp. are isolated from women with and without BV and have been associated with elevated cervicovaginal pH [31][32][33] . Veillonella spp. are commonly isolated with Lactobacillus spp., G. vaginalis, and Peptostreptococcus spp 31,32 . Overall, Veillonellaceae family members have been isolated from women with BV 16,17,32,34,35 , but their contributions to BV pathogenesis and biofilm formation remain understudied. Although the Veillonella genus is broadly reported, species-level specification is lacking in sequencing datasets 31,32 . Therefore, we chose V. atypica, V. montpellierensis, and M. micronuciformis to represent Veillonellaceae family members present in the female genital tract to better understand their potential contributions to BV and the local microenvironment. The specific species and strains used in this study were isolated from the female genital The functional impact of V. atypica, V. montpellierensis, and M. micronuciformis on host response mechanisms in the human cervix, which can contribute to poor gynecologic and reproductive outcomes, has not been previously investigated. Therefore, we aimed to evaluate the immunometabolic contributions of V. atypica, V. montpellierensis, and M. micronuciformis in the cervical microenvironment to better understand the role of these species in BV and other sequalae. To achieve this, we used a robust and extensively characterized human three-dimensional (3-D) cervical epithelial cell model for the study of host-microbe interactions and infected the model with clinical isolates of Veillonellaceae family members, coupled with untargeted global metabolomics and immunoproteomics approaches. We chose to use a 3-D cervical cell model due to the clinical relevance of studying BVassociated bacteria in the context of the cervix as it relates to human papillomavirus (HPV), HIV, Chlamydia trachomatis, and Neisseria gonorrhoeae infections [36][37][38][39] . Additionally, M. micronuciformis has been isolated from women suffering from preterm birth [40][41][42] . Veillonella spp. have been implicated in secondary bacteremia in a pregnant woman, but the potential for Veillonella spp. to participate as opportunistic pathogens in the female reproductive tract is understudied 43 .
The 3-D cervical model is a robust system to study host-microbe interactions, as it accurately resembles ultrastructural features of the cervical tissue in vivo, such as apical and basal differentiation, tight junctions, microvilli, and more organotypic surfaces for the colonizing bacteria, and has been shown to recapitulate clinical findings 15,[44][45][46][47] . Herein, we applied a reductionist approach and infected the 3-D cervical model with single bacterial species to determine the individual contributions of these bacteria to the cervicovaginal microbiome that are currently lacking. This knowledge is essential to inform future experiments that involve polymicrobial infections [44][45][46][47] .

RESULTS
M. micronuciformis infections induce moderate cytotoxicity to cervical epithelial cells First, we screened cervical epithelial monolayers for cytotoxicity following infection with V. atypica, V. montpellierensis, and M. micronuciformis at multiplicities of infection (MOIs) ranging from 2 to 400. We found that M. micronuciformis induced significant cytotoxicity at all tested MOIs (6.65%, 7.96%, and 13.63%) to cervical epithelial cell monolayers (P = 0.0291, P = 0.0041, and P < 0.0001 respectively); however, cervical epithelial cytotoxicity induced by M. micronuciformis infections was still modest at the highest MOI. V. montpellierensis infections induced cytotoxicity to cervical epithelial monolayers at only the highest MOI tested (6.50%, P = 0.0359). V. montpellierensis infections did not induce significant epithelial cytotoxicity at MOIs 2-4 and 20-40 (1.84%, 3.85%, respectively). Percentage cytotoxicity of V. atypica infections resulted in similar epithelial cytotoxicity at all tested MOIs (3.99%, 3.94%, and 6.23%) (Fig. 1A). Since Veillonella and Megasphaera spp. are predicted to harbor genes encoding lactate dehydrogenase (LDH) 48,49 , we also assessed the cytotoxicity using crystal violet staining ( Supplementary Fig. 1). These results were consistent with LDH activity data with respect to cell viability measurements, indicating that potential bacterial LDH activity did not impact these data.
Next, we infected human 3-D cervical models with V. atypica, V. montpellierensis, and M. micronuciformis and confirmed colonization of the 3-D aggregates by SEM (Fig. 1B-D). Generally, V. atypica, V. montpellierensis, and M. micronuciformis sparsely colonized the 3-D cervical epithelial cells with occasional small clusters of bacteria co-localizing together on cervical cells. V. montpellierensis appeared to mostly colonize the surface of dead cells as visualized by SEM. M. micronuciformis also tended to sparsely cluster with other bacteria on cervical epithelial cell surfaces.
Infection with M. micronuciformis induces significantly elevated 3-D cervical cell pro-inflammatory responses while V. atypica significantly decreases the expression of specific epithelial barrier targets Since BV is accompanied by mild genital inflammation, remodeling of the extracellular matrix, and cell lysis 46   We utilized hierarchical clustering analysis for evaluating immune response profiles for each bacterial infection. V. atypica and V. montpellierensis exhibited similar immunoproteomic profiles and clustered separately from M. micronuciformis or PBS controls ( Fig. 2A). Overall, M. micronuciformis induced the secretion of pro-inflammatory cytokines and chemokines to a greater extent compared to V. atypica and V. montpellierensis and PBS controls. Infection with V. atypica resulted in a decrease in the concentration of PDGF-AA, VEGF, and MMP-10.
Untargeted global metabolomics analysis revealed similar metabolic profiles between V. atypica and V. montpellierensis that are distinct from M. micronuciformis We next sought to define how V. atypica, V. montpellierensis, and M. micronuciformis alter the metabolomic landscape of 3-D cervical cell culture supernatants using high-throughput global untargeted metabolomics. Principal component analysis (PCA) of global metabolomic profiles (Fig. 3A) revealed that V. atypica and V. montpellierensis metabolomes cluster together, whereas M. micronuciformis and PBS controls tended to cluster together and separately from both Veillonella groups. Relative to PBS controls, PC1 (36.8% of explained variance) scores were significantly different between V. atypica and V. montpellierensis (P < 0.0001). No significant differences were found between PBS controls and any strain tested with respect to PC2 (19.2% of explained variance) scores.
We then applied non-parametric correlation analyses (Fig. 3B) to visualize the clustering of biological replicates. We observed that metabolomic profiles of V. atypica and V. montpellierensis correlated well with each other; however, the cluster analysis failed to group individual replicates of the same species, therefore suggesting highly similar metabolic profiles among the two Veillonella spp. While in the same parent cluster, the metabolomic profiles for M. micronuciformis and PBS controls mostly separated from each other. PBS controls correlated well between individual replicates, with the exception of one infection replicate with M. micronuciformis. Overall, the correlation analysis supported the PCA results.
We found variations in lipid metabolism both in our metabolite composition analysis (  enriched pathways corresponding to lipid metabolism included glycerolipid metabolism and steroid and steroid hormone biosynthesis as modulated by V. montpellierensis and M. micronuciformis, respectively. Metabolic pathways corresponding to lipid metabolism were not significantly enriched by infection with V. atypica, therefore suggesting V. montpellierensis and M. micronuciformis influence species-specific lipid metabolic products following infection of 3-D cervical cells. V. atypica and V. montpellierensis infections resulted in significant elevation of polyamines and depletion of lactate, whereas glycerolipids were significantly elevated with M. micronuciformis infections Finally, we assessed specific metabolites that were significantly elevated or depleted in response to V. atypica, V. montpellierensis, and M. micronuciformis infections relative to PBS controls. Culture supernatants from human 3-D cervical cells infected with V. atypica and V. montpellierensis significantly accumulated polyamines, such as cadaverine (P = 0.0159 and P = 0.0037), putrescine (P = 0.0155 and P = 0.00485), agmatine (P = 0.0444 and P = 0.0170), and histamine (P = 0.0331 and P = 0.0452) (Fig. 5B).
Culture supernatants from 3-D cervical cells infected with M. micronuciformis accumulated glycolipids and sphingolipids (  shading indicates the 95% confidence interval and B Pearson's correlation analysis of metabolomics data displays how sample metabolic profiles relate to each other. Blue and red coloring indicate negative and positive correlations between samples, respectively. C Pie charts of significantly changed metabolites. The percentage of metabolites that were approached significance (P < 0.1) within each super pathway are labeled next to each slice (the total number of significantly altered metabolites for V. atypica, V. montpellierensis, and M. micronuciformis were 69, 98, 63 respectively). PCA significance (A) was tested by one-way ANOVA and Dunnett's adjustment for multiple comparisons relative to PBS; ****P < 0.0001. The significance of differences between metabolite compositions (C) was tested using a chi-squared test of trends; **P < 0.01. In addition, 2-hydroxyglutarate (an oncometabolite) was significantly elevated during infection with V. atypica and V. montpellierensis (P = 0.0308 and P = 0.0052, respectively), with V. atypica infections (P = 0.0268), and betaine was altered with V. montpellierensis infections (P = 0.0102). Urate (an HPV-associated metabolite) 51 was significantly increased with V. atypica infections (P = 0.0244).
Overall, infection of 3-D cervical epithelial cells with V. atypica or V. montpellierensis induced significant elevation of metabolites associated with a higher cervicovaginal pH and amine odor, combined with depletion of metabolites involved with polyamine biosynthesis. M. micronuciformis infections significantly elevated glycerolipids and sphingolipids that are potentially indicative of cervical epithelial cell cytotoxicity (Fig. 5).

DISCUSSION
Members of the Veillonellaceae family, V. atypica, V. montpellierensis, and M. micronuciformis have been isolated from women with BV and healthy women 2,16,17,31,32,34,40 ; however, the mechanisms employed by these bacteria that relate to BV pathogenesis and  Supplementary Fig. 3), B V. montpellierensis ( Supplementary Fig. 4), and C M. micronuciformis ( Supplementary Fig. 5), under anaerobic conditions for 24 h. D Comparison of significantly enriched metabolic pathways between infections. Color bars indicate significance, yellow represents pathways that are significantly enriched (P < 0.05), green indicates pathways that approach statistical significance (0.05 < P < 0.1), and blue indicates pathways that were not significantly enriched (P > 0.1). The super pathway that a particular sub-pathway belongs to is indicated with colored circles.
other gynecologic and obstetric sequalae are unknown. In this study, we aimed to elucidate the metabolic and inflammatory contributions of V. atypica, V. montpellierensis, and M. micronuciformis to reveal the function of these organisms in the cervical microenvironment.
To our knowledge, colonization of human organotypic genital epithelial cells by V. atypica, V. montpellierensis, or M. micronuciformis in vitro has not been previously demonstrated and visualized. Using our human 3-D cervical epithelial cell model, we demonstrated that V. atypica, V. montpellierensis, and M. micronuciformis are capable of colonizing cervical epithelial cell surfaces (Fig. 1B-D). Our human 3-D cervical epithelial cell model expresses mucins and TLRs and other physiological relevant features of parental human cervical tissue that may enhance colonization of bacteria compared to traditional 2-D monolayer cultures 46,52 .
The cytotoxic potential of Veillonellaceae members in the cervical epithelium has not been thoroughly explored. One study tested an unnamed Veillonella spp. on vaginal epithelial cell monolayers and found no evidence for cell death, though this was not quantified 53 . Our data suggest that M. micronuciformis induces moderate cytotoxicity, whereas V. atypica and V. montpellierensis do not significantly contribute to cervical epithelial cell death (Fig. 1). In this capacity, M. micronuciformis may therefore contribute to the breakdown of epithelial barrier function, in accordance with other BV-associated bacteria 4 .
To our knowledge, the pro-inflammatory host response to cervical infection with M. micronuciformis has not been previously examined in vitro, though another species under the Megasphaera genus (Megasphaera elsdenii) has been investigated 54 . In this study, dendritic cells elevated secretion of IL-1β, IL-6, IL-8, IL-12p40, and TNF-α in response to infection with M. elsdenii 54 . Additionally, M. micronuciformis was associated with genital inflammation as measured from cervicovaginal fluid in a cohort of South African adolescent females 55 . However, no studies have identified the immune response to V. atypica and V. montpellierensis in the cervicovaginal context. Our results suggest that M. micronuciformis induced a robust human 3-D cervical epithelial cell pro-inflammatory response marked by upregulated secretion of pro-inflammatory cytokines and chemokines, whereas both V. atypica and V. montpellierensis induced relatively fewer proinflammatory mediators ( Fig. 2A). We hypothesize that M. micronuciformis is more pro-inflammatory than V. atypica or V. montpellierensis in the cervicovaginal environment, which is in the accordance with the literature, although more studies are required that investigate Veillonella spp. in the cervicovaginal environment.
SCFA production is a metabolic hallmark of BV as demonstrated in clinical and in vitro studies [56][57][58] and have been linked to increased secretion of pro-inflammatory mediators 59 . We observed a higher accumulation of N-palmitoyl-sphingosine in response to infection with M. micronuciformis compared to V. atypica or V. montpellierensis, and is associated with inflammation 60 . Veillonella spp. and M. micronuciformis also induced upregulated secretion of several pro-inflammatory mediators from the 3-D cervical model, which is consistent with the literature and SCFA elevation 59 and could indicate that SCFAs are both a hallmark of BV and indicators of cervical inflammation.
Histamine, a key metabolite involved in local host responses, was significantly accumulated by infection with V. atypica and V. montpellierensis. Other studies have found that histamine can decrease the expression of tight junctional proteins (ZO-1 in nasal epithelial cells and E-cadherin in epithelial pulmonary cells) 61 . Therefore, an increase in histamine may increase epithelial permeability, although this would need to be tested in the context of the cervicovaginal epithelium to confirm the role of histamine in barrier function within the cervicovaginal mucosa.
Our metabolomics analyses revealed changes in the lipid metabolism following infections with Veillonellaceae family members. Infection with M. micronuciformis induced significant accumulation in glycerolipids and sphingolipids (Fig. 5). We also observed a significant accumulation in glycerolipids and cholesterol during human 3-D cervical cell infections with V. atypica, V. montpellierensis, and M. micronuciformis. To date, there are few studies that have reported an association with elevated glycerolipids and cell death, but sphingolipids have been related to cell death 62 . In our previous clinical study, elevation in the relative abundance of sphingomyelin and 1-stearoyl-2-docosahexaenoyl-GPC correlated with Lactobacillus depletion and genital inflammation 51 . However, additional studies are needed in order to correlate elevated glycerolipids with cell death.
Early reports that dissect the metabolic potential of Veillonella 49,63-66 revealed that Veillonella spp. utilize lactate as a sole energy source 49,64,67 . Health-associated Lactobacillus spp. secrete copious amounts of lactic acid into the cervicovaginal milieu 14,68,69 and lactic acid is depleted in severe cases of BV 3 . It is therefore possible that V. atypica and V. montpellierensis elicit their pathogenicity in the early stages of BV when lactate is still available. Our data support lactate utilization by V. atypica and V. montpellierensis (Fig. 5). Since current literature indicates lactate consumption by other Megasphaera spp. 70 , our finding could be explained by suggesting that lactate serves as a non-primary energy source for Megasphaera. Importantly, we have shown that Veillonella spp. consume lactate (key metabolite associated with cervicovaginal health) and may contribute to changes in the cervicovaginal pH 71 . Indeed, in our previous clinical study investigating the vaginal microbiome in women across cervical neoplasia, V. montpellierensis and other Veillonella spp. were enriched in women with abnormal cervicovaginal pH 33 .
Elevation of cervicovaginal polyamines is a metabolic hallmark of BV and contributes to increased pH and amine odor, both of which are key clinical symptoms of BV 72 . Polyamine synthesis by Veillonella has been documented in the oral microbiome and contributes to malodor 73 . Additionally, one clinical study demonstrated that M. micronuciformis was associated with elevated cervicovaginal putrescine 14 . Our data revealed that in the context of our human 3-D cervical epithelial cell model, both V. atypica and V. montpellierensis modulate a variety of polyamines, including cadaverine, putrescine, histamine, and agmatine, whereas polyamines were not significantly elevated by M. micronuciformis infection (Fig. 5). Additionally, in previous clinical studies the metabolite 5′-methylthioadenosine (MTA), a byproduct of polyamine synthesis, has been found to be depleted in women harboring non-Lactobacillus dominant microbiomes 72,74,75 . In our current study, MTA was also depleted in V. atypica and V. montpellierensis infections. An earlier study 14 provided clinical correlations for putrescine production by Megasphaera in a vaginal context, which is not in accordance with our data; however, the next-generation sequencing method employed was unable to identify Megasphaera at the species level. It is possible that other Megasphaera spp. present in the cervicovaginal microbiome may exert differential metabolic capabilities that influenced the observed elevation of putrescine in this study 14 . However, both the increase in polyamine synthesis and depletion of MTA observed in infections with V. atypica and V. montpellierensis suggest that these microorganisms may significantly impact polyamine synthesis in vivo and influence metabolic hallmarks of BV (e.g., putrescine and cadaverine).
Our data also revealed arginine depletion following M. micronuciformis infection. Nitric oxide, agmatine, and ornithine are by-products of arginine catabolism 5,76 . Considering neither agmatine nor ornithine was significantly altered in our analysis, perhaps the observed depletion of arginine was a result of nitric oxide production by 3-D cervical cells. Nitric oxide has been implicated as an important toxic defense molecule against pathogens 77 and has a role in inflammation and the immune response, further indicating pro-inflammatory properties of M. micronuciformis in the cervicovaginal microenvironment 78 .
In summary, lactate depletion and polyamine production observed in our 3-D human cervical cell model suggest that Veillonella spp. contribute to elevated cervicovaginal pH via consumption of a key metabolite linked to gynecologic health (Fig. 6). We found that M. micronuciformis exhibited greater proinflammatory properties compared to V. atypica and V. montpellierensis. M. micronuciformis infections also resulted in higher levels of epithelial cytotoxicity and a higher enrichment in SCFAs that can also impact the production of pro-inflammatory cytokines and chemokines or cell death (Fig. 6). In line with our working hypothesis, our results provide evidence that Veillonellaceae family members alter the local microenvironment in a distinctive fashion. As putative primary colonizers, Veillonella spp. may create a favorable environment for secondary colonizers. On the other hand, M. micronuciformis may participate as a putative secondary colonizer that induces inflammation in the cervicovaginal tract and manifest as symptoms and clinical presentation of BV, as well as other gynecologic and reproductive sequalae (Fig. 6).
All experimental model systems have their own strengths and weaknesses 79 . We acknowledge the polymicrobial nature of BV and inter-bacterial and host-microbe interactions that participate in the development and progression of the disease. In this study, we used a reductionist approach and performed mono-infections using our robust 3-D cervical cell model to dissect the individual contributions of specific Veillonellaceae members. This approach is required to first understand individual pathologic contributions from these understudied bacteria. This data will provide the foundations for future studies that utilize polymicrobial cocktails that better represent the complexity and microbe-microbe interactions present in the microbiome. Finally, our untargeted metabolomics approach is beneficial since it allows us to examine metabolic changes on a global scale. This technique is limited, however, by the relativistic nature of metabolite measurements and is therefore unable to provide absolute metabolite quantification. Future studies employing targeted metabolomics will help to directly compare concentrations of key cervicovaginal metabolites within in vivo samples and in vitro models.
Future studies on Veillonellaceae members in the context of the cervicovaginal epithelium should focus on the temporal colonization patterns of BV-associated organisms, as this will provide insights into the potential contribution to biofilm formation for these microorganisms. In addition, studies of polymicrobial "cocktails" that combine Veillonellaceae members with other vaginal bacterial isolates will provide additional insights into bacteriabacteria interactions in the context of the cervicovaginal microenvironment. In conclusion, the unique mechanisms and individual contributions of Veillonellaceae family members defined herein may promote our understanding of biofilm formation characteristic of BV and the underlying mechanisms that contribute to adverse obstetric/gynecologic health outcomes related to BV.

METHODS
Human cervical epithelial cell monolayer and threedimensional human cervical cell model culture Human cervical epithelial cells (A2EN) were grown as monolayers with keratinocyte serum-free media (Fischer Scientific) supplemented with endothelial growth factor (5 ng/ml), bovine pituitary extract (50 µg/ml), and Primocin (100 µg/ml) in the humidified atmosphere of 5% CO 2 at 37°C 46 . Primocin was not included in the media used in the bacterial infection assays outlined below. For 2-D monolayer cultures, cells were seeded into culture-treated 24-well plates to a cell density of~2 × 10 5 cells/ ml. Prior to seeding, cells were enumerated using the Countess automated cell counter (Invitrogen) and trypan blue exclusion, which were used for cytotoxicity assays.
Bacterial strains and growth conditions V. atypica strain CMW7756B and V. montpellierensis strain DNF00314 were cultured on brain heart infusion agar (Becton Dickinson) supplemented with 5% (v/v) defibrinated sheep blood (Quad Five) and 0.56 mg/l of sodium D-lactate (Sigma-Aldrich). M. micronuciformis strain DNF00954 was cultured on tryptic soy agar (Becton Dickinson) supplemented with 5% (v/ v) defibrinated sheep blood (Quad Five). Bacteria were grown at 37°C under anaerobic conditions generated using AnaeroPack sachets and anaerobic environmental chambers (Thermo Scientific). All strains were obtained from Biodefense and Emerging Infections Research Resources Repository (https://www.beiresources.org). M. micronuciformis strain DNF00954 and V. montpellierensis strain DNF00314 were isolated from women with BV, whereas V. atypica strain CMW7756B was isolated from the genital tract of a pregnant woman.

Bacterial infection of human cervical epithelial cells
Prior to the infection, bacteria were grown for 16-18 h on appropriate agar plates, resuspended in sterile Dulbecco's phosphate-buffered saline (PBS), and adjusted to an optical density at 600 nm (OD 600 ) 0.5. The viability of bacteria was confirmed using the standard plating assay. The adjusted bacterial cell suspensions were serially diluted in PBS, plated on appropriate agar media, and incubated for 96 h at 37°C under anaerobic conditions for enumeration of colony-forming units (CFU). The bacterial suspensions of OD 600 0.5 corresponded to bacterial densities of approximately 1-2 × 10 8 CFU/ml. Human cervical cell 2-D monolayers used for cytotoxicity assays were infected with a range of bacterial suspensions equivalent to OD 600 of 0.001, 0.01, and 0.1 per 1× 10 5 cervical cells/ml, which corresponded to multiplicities of infection (MOI) of 2-4, 20-40, and 200-400. Infected cervical cells were incubated for 24 h at 37°C under anaerobic conditions. In a preliminary experiment, after 24 h incubation with cervical cells, the viability of bacteria was confirmed and did not change more than 1 log when compared to the initial inocula. The 3-D cervical epithelial cells were infected with the bacterial resuspensions at an MOI of 20-40 for 24 h at 37°C under anaerobic conditions, as described above.

Lactate dehydrogenase (LDH) assay to assess cytotoxicity
The LDH assay (Invitrogen, ThermoFisher Scientific) was performed according to the manufacturer's guidelines. LDH activity was measured from cell culture supernatants by recording absorbance values at 490 and 680 nm. Spontaneous LDH activity was measured from PBS-treated cells and maximum LDH activity was measured from lysed cells. Percentage cytotoxicity was calculated by using the following formula: (bacteriainfected LDH activity − spontaneous LDH activity)/(the maximum LDH activity − spontaneous LDH activity) × 100.

Scanning electron microscopy
Three-dimensional cervical epithelial cells were infected with each bacterial species at an MOI of 200-400 for 4 h at 37°C under anaerobic conditions as described above. Samples were fixed and processed for scanning electron microscopy (SEM) as previously described 80,81 . Samples were imaged using a JSM-6300 JEOL scanning electron microscope and images were obtained with the IXRF model 500 digital processor (IXRF systems). SEM images were pseudo-colored using Photoshop 19.0 CC (Adobe).

Untargeted global metabolomics analysis of cell culture supernatants
Cell culture supernatants collected from three independent 3-D cervical cell aggregate batches for both Veillonella strains and four independent batches for M. micronuciformis infections were sent to Metabolon Inc. (Durham, NC) for untargeted global metabolomics analysis. Metabolites were resolved on a Waters ACQUITY ultra-performance liquid chromatography (UPLC) and a Thermo Scientific Q-Exactive high resolution/accurate mass spectrometer interfaced with a heated electrospray ionization (HESI-II) source and Orbitrap mass analyzer operated at 35,000 mass resolution. The sample extract was dried then reconstituted in solvents compatible to each of the four methods. Each reconstitution solvent contained a series of standards at fixed concentrations to ensure injection and chromatographic consistency. One aliquot was analyzed using acidic positive ion conditions, chromatographically optimized for more hydrophilic compounds. In this method, the extract was gradient eluted from a C18 column (Waters UPLC BEH C18-2.1 × 100 mm, 1.7 µm) using water and methanol, containing 0.05% perfluoropentanoic acid (PFPA) and 0.1% formic acid (FA). Another aliquot was also analyzed using acidic positive ion conditions; however, it was chromatographically optimized for more hydrophobic compounds. In this method, the extract was gradient eluted from the same aforementioned C18 column using methanol, acetonitrile, water, 0.05% PFPA, and 0.01% FA and was operated at an overall higher organic content. Another aliquot was analyzed using basic negative ion optimized conditions using a separate dedicated C18 column. The basic extracts were gradient eluted from the column using methanol and water, however with 6.5 mM ammonium bicarbonate at pH 8. The fourth aliquot was analyzed via negative ionization following elution from a HILIC column (Waters UPLC BEH Amide 2.1 × 150 mm, 1.7 µm) using a gradient consisting of water and acetonitrile with 10 mM ammonium formate, pH 10.8. The MS analysis alternated between MS and data-dependent MS n scans using dynamic exclusion. The scan range varied slighted between methods but covered 70-1000 m/z.
The bioinformatics system consisted of four major components, the Laboratory Information Management System (LIMS), the data extraction and peak-identification software, data processing tools for QC and compound identification, and a collection of information interpretation and visualization tools for use by data analysts. The hardware and software foundations for these informatics components were the LAN backbone, and a database server running Oracle 10.2.0.1 Enterprise Edition. Peaks were quantified using area-under-the-curve, which allows determining relative intensity of compounds among tested samples, but not the absolute concentrations.

Statistical analyses
All infections and assays were performed as at least three independent replicates. One-way ANOVA with Dunnett's adjustment for multiple comparisons was used to statistically analyze the immunoproteomics data. One-way ANOVA with Bonferroni post-hoc tests were used to test for significant differences in LDH data. Differences in metabolite pathway composition were determined by chi-squared analysis. One-way ANOVA analysis was performed using Prism v8 software (GraphPad). Hierarchical clustering analysis and heatmap visualization were performed using ClustVis 82 . Metabolite intensity values were median-scaled and logtransformed prior to performing two-tailed paired Student's t-tests (infection vs. PBS control) using the R rstatix package. Metaboanalyst 4.0 83 was used for principal component analysis (PCA), Pearson's correlation analysis, and metabolite pathway enrichment analysis. P-values of <0.05 were considered significant in all analyses. Metabolomics results were corrected for multiple testing using the false discovery rate (FDR) and q-values were reported. All error bars represent standard deviation.