Prevotella timonensis degrades the vaginal epithelial glycocalyx through high fucosidase and sialidase activities

ABSTRACT Bacterial vaginosis (BV) is a polymicrobial infection of the female reproductive tract. BV is characterized by replacement of health-associated Lactobacillus species by diverse anerobic bacteria, including the well-known Gardnerella vaginalis. Prevotella timonensis, and Prevotella bivia are anerobes that are found in a significant number of BV patients, but their contributions to the disease process remain to be determined. Defining characteristics of anerobic overgrowth in BV are adherence to the mucosal surface and the increased activity of mucin-degrading enzymes such as sialidases in vaginal secretions. We demonstrate that P. timonensis, but not P. bivia, strongly adheres to vaginal and endocervical cells to a similar level as G. vaginalis but did not elicit a comparable proinflammatory epithelial response. The P. timonensis genome uniquely encodes a large set of mucus-degrading enzymes, including four putative fucosidases and two putative sialidases, PtNanH1 and PtNanH2. Enzyme assays demonstrated that fucosidase and sialidase activities in P. timonensis cell-bound and secreted fractions were significantly higher than for other vaginal anerobes. In infection assays, P. timonensis efficiently removed fucose and α2,3- and α2,6-linked sialic acid moieties from the epithelial glycocalyx. Recombinantly expressed P. timonensis NanH1 and NanH2 cleaved α2,3 and α2,6-linked sialic acids from the epithelial surface, and sialic acid removal by P. timonensis could be blocked using inhibitors. This study demonstrates that P. timonensis has distinct virulence-related properties that include initial adhesion and a high capacity for mucin degradation at the vaginal epithelial mucosal surface. Our results underline the importance of understanding the role of different anerobic bacteria in BV. IMPORTANCE Bacterial vaginosis (BV) is a common vaginal infection that affects a significant proportion of women and is associated with reduced fertility and increased risk of secondary infections. Gardnerella vaginalis is the most well-known BV-associated bacterium, but Prevotella species including P. timonensis and P. bivia may also play an important role. We showed that, similar to G. vaginalis, P. timonensis adhered well to the vaginal epithelium, suggesting that both bacteria could be important in the first stage of infection. Compared to the other bacteria, P. timonensis was unique in efficiently removing the protective mucin sugars that cover the vaginal epithelium. These results underscore that vaginal bacteria play different roles in the initiation and development of BV.

outcomes including preterm birth (4).BV is diagnosed according to the Amsel criteria, namely, high vaginal pH (>4.5), detection of thin discharge, an odor of amines after addition of potassium hydroxide, and the presence of "clue cells" in vaginal secretions (5).Bacterial Gram staining followed by the Nugent score test is also used to diagnose BV (6,7).Vaginal secretions in BV patients contain enzymes that are capable of degrading the protective mucus layer, including mucinases and sialidases, which can also be used for diagnostics (8,9).
G. vaginalis is the most well-studied BV-associated anerobe.Due to its ability to adhere to the vaginal epithelium and tolerate small amounts of oxygen, it is proposed to be an initial anerobic colonizer that can replace resident Lactobacillus species (16).G. vaginalis can use glycogen, a carbon source that is abundant at the vaginal epithelium, (17) and degrades the protective mucus layer through the production of sialidases (18,19).G. vaginalis also secretes vaginolysin (VLY), a cytotoxin capable of killing epithelial cells (20).However, not all G. vaginalis strains are sialidase-positive, and G. vaginalis is also found in healthy women (21,22).Therefore, G. vaginalis may require other species for BV initiation.Prevotella bivia, for example, produces ammonia, which stimulates the growth of G. vaginalis (23).Such synergistic relationships between different vaginal anerobes with different pathogenic properties most likely drive BV development.
The pathogenic potential of Prevotella species in BV is understudied in comparison to that of G. vaginalis.Previous research mainly focused on P. bivia, the most commonly isolated Prevotella species during BV (15).However, recent studies demonstrate that P. timonensis is also often found in women with BV (24,25), and recently the new name Hoylesella timonensis was proposed for this species (26).Due to the high similarity of their 16S rRNA genes, many studies could not discriminate between different Prevotella spp.(27).Vaginal Prevotella spp. in general have been associated with increased cytokine levels in the cervicovaginal fluid (28,29).Other reports suggest that Prevotella spp.have a role in biofilm formation and mucus degradation (18,27).We have shown that P. timonensis, but not P. bivia, induces a strong proinflammatory response through dendritic cell activation (30) and increases HIV-1 uptake by Langerhans cells, turning these cells into HIV-1 reservoirs (31).Prevotella spp.have also been associated with sialidase activity in vaginal secretions of BV patients (8).P. bivia exhibits sialidase activity that targets the vaginal mucus layer (18) and leads to increased adhesion of other BV-associated bacteria, including A. vaginae (32).P. timonensis also exhibited sialidase activity and altered mucin expression in the human endometrial epithelial cell line HEC1-A (27).However, the role played by the different Prevotella strains in BV is currently not clear.In this study, we determined the pathogenic properties of P. timonensis compared to that of P. bivia and other BV-associated bacteria, focusing on bacterial interactions with human cells and glycans.We conclude that P. timonensis has unique virulence-related traits that might play an important role during initiation and develop ment of BV.

Prevotella timonensis adheres to vaginal and endocervical cells
Attachment to the vaginal epithelium is thought to be the first step toward replacement of commensal Lactobacillus species and colonization by bacterial anerobes (Fig. 1A).We investigated the extent to which different BV-associated bacteria can attach to vaginal and endocervical cells and included commensal L. crispatus as a control.The vaginal cell line VK2/E6E7 and endocervical cell line End1/E6E7 were grown to a fully confluent monolayer, followed by incubation with L. crispatus, G. vaginalis, P. timonensis, or P. bivia at a multiplicity of infection (MOI) of 10 in anerobic conditions.After 18 hours, the percentage of attached bacteria was determined by colony counting.We observed that commensal L. crispatus adhered well to both VK2/E6E7 and End1/E6E7 cell lines, at 66% and 81% of the total bacterial inoculum, respectively, while P. bivia was the least adherent bacterium, with 12% attachment to VK2/E6E7 cells and 2% to End1/E6E7 cells.G. vaginalis and P. timonensis showed comparable intermediate binding, with adhesion percentages varying between 15% and 40% (Fig. 1B).
In an independent set of experiments, we assessed bacterial binding to cell surfaces by using fluorescence in situ hybridization (FISH).We designed specific fluorescently labeled peptide nucleotide acid (PNA) probes for P. timonensis (PT-Cy3) and P. bivia (PB-Cy3).We used a previously reported PNA probe for G. vaginalis (Gard162-AF488) and a general 16S probe for L. crispatus (EUB338-AF488).Bacteria were adhered to coated glass slides to test the specificity of the PNA probes, and all four probes showed good correlation with the 4′,6-diamidino-2-phenylindole (DAPI) signal (Fig. S1).We then infected confluent VK2/E6E7 and End1/E6E7 cells with bacteria at a MOI 50 for 18 hours in anerobic conditions.The infected epithelial monolayers were stained with the FISH probes and wheat germ agglutinin (WGA) to visualize the epithelial surface.We again found that G. vaginalis and P. timonensis attached more effectively to the epithelial surface compared to P. bivia (Fig. 1C and D).L. crispatus showed strong adherence to both cell lines (Fig. 1D and E).The Gram-positive L. crispatus required a specific permeabi lization buffer to achieve efficient labeling of the bacteria with the PNA probe, which led to increased WGA staining of the bacteria in addition to the epithelial cells (Fig. 1C and  D).Together, these colony counting and FISH experiments demonstrate that P. timonensis can adhere to the surface of vaginal and endocervical monolayers, to comparable levels as the well-known BV-associated pathogen G. vaginalis.

P. timonensis does not cause cell cytotoxicity and does not induce major inflammatory responses
We next investigated the cytotoxic and inflammatory potential of P. timonensis after adhesion to VK2/E6E7 and End1/E6E7 cell lines.We infected confluent epithelial monolayers with the selected bacteria at an MOI of 10 and 100 for 18 hours in anerobic conditions and measured LDH release, an indicator of cellular cytotoxicity.As previously described, G. vaginalis was highly cytotoxic, resulting in an LDH release of approximately 70% of the maximum release of the total monolayer (Fig. 2A and B).Incubation with P. timonensis, P. bivia, or L. crispatus for 18 hours did not result in increased LDH release compared to uninfected cells (Fig. 2A and B).
To determine whether the different vaginal bacteria trigger an inflammatory response, we incubated confluent epithelial monolayers with bacteria at MOI 10 and 100 for 18 hours in anerobic conditions and measured the mRNA expression of the cytokines IL-1β, IL-8, CCL5, and CCL20 using quantitative RT-PCR.Compared to the commensal L. crispatus, only G. vaginalis significantly increased IL-1β, IL-8, and CCL20 expressions in both VK2/E6E7 and End1/E6E7 cells (Fig. 2C through E).CCL5/RANTES, a chemoattractant of T lymphocytes and monocytes (33), was slightly but significantly upregulated in endocervical cells incubated with P. timonensis and P. bivia, but not G. vaginalis (Fig. 2F).These results suggest that despite its extensive attachment to the epithelial surface, P. timonensis does not induce a strong inflammatory response in vaginal or endocervical cells.

Utilization of glycogen and mucins as carbon sources by BV-associated bacteria
Vaginal and cervical epithelial cells produce high amounts of glycogen, which is deposited onto the epithelium once epithelial cells are shed and lysed and can serve as a carbon source for the resident vaginal bacteria (34).We investigated whether our selected BV-associated bacteria could utilize glycogen for growth.Carbohydrates were removed from each bacterium-specific medium and supplemented with 0.5% glycogen.P. timonensis, P. bivia, and G. vaginalis were all able to grow on glycogen (Fig. 3A through  C).Interestingly, G. vaginalis reached a higher OD in the basal media supplemented with glycogen compared to the complete specific media, demonstrating a preference for glycogen as the carbon source (Fig. 3C).Akkermansia muciniphila, a member of the intestinal microbiota known to degrade mucins, did not grow on glycogen (Fig. 3D), supporting the notion that glycogen is a preferred carbon source for vaginal-associated bacteria.
The cervicovaginal mucus that covers the vaginal and endocervical epithelium facilitates uterine lubrication and microbial clearance (35).We assessed whether the BVassociated bacteria could use mucins as a carbon source by supplementing the basal media with 0.5% purified porcine gastric mucins (PGM).A. muciniphila grew well on mucins (Fig. 3D), but P. bivia and G. vaginalis did not exhibit increased growth on mucins compared to the basal medium without carbohydrates (Fig. 3B and C).P. timonensis showed a small but significant increase in growth in the mucin-containing medium compared to the basal medium without carbohydrates (Fig. 3A), suggesting that P. timonensis might degrade and utilize mucins.

The genome of P. timonensis predicts a high O-glycan degradation potential
To determine the genetic potential of P. timonensis and the other BV-associated bacteria to degrade different carbon sources, we sequenced the P. timonensis, P. bivia, G. vaginalis, and A. muciniphila strains used in this study (sequences deposited in PRJEB67799).The genomes were analyzed for the presence of carbohydrate-active enzymes (CAZymes) using the dbCAN2 meta server pipeline for automated CAZyme annotation (36).Only CAZyme genes that were predicted by at least two out of three annotation tools were selected.Detected CAZyme genes included glycoside hydrolases (GH), carbohydrate esterases (CE), glycosyl transferases (GT), carbohydrate-binding modules (CBM), and auxiliary activities (AA).The mucin degrader A. muciniphila presented the highest amount of putative CAZy domains (155 ORFs), followed by P. timonensis with 104 ORFs, P. bivia with 71 ORFs, and G. vaginalis with 40 ORFs (Fig. 4A).
We next examined these results to identify candidate enzymes for degradation of specific substrates.Mucins have polypeptide backbones that are decorated by complex O-linked glycan structures that require sequential degradation by glycoside hydrolases with high specificity (Fig. 4B).Within the glycoside hydrolase category, several genes encoding predicted sialidases (GH33 class) and fucosidases (GH29 and GH95 classes) were detected in the genomes of all four bacteria (Fig. 4C).Furthermore, P. timonensis and A. muciniphila possessed a great number of predicted α/β-galactosidases, α/β-Nacetylgalactosaminidases, and α/β-N-acetylglucosaminidases, enzymes that hydrolyze the glycosidic linkages underlying the terminal sialic acids and fucoses (Fig. 4C).Many of these putative CAZymes contained a signal peptide, suggesting that the proteins may be translocated to the bacterial surface or secreted into the environment (Fig. 4C).
Sialic acids and fucoses cap mucin O-glycan structures and are the first monosacchar ides that need to be removed for further mucin degradation (37).The fucosidase family consists of the retaining fucosidases (GH29) and inverting fucosidases (GH95).The P. timonensis genome uniquely encoded a predicted GH95 enzyme in addition to three GH29-containing fucosidases.The P. bivia and G. vaginalis genomes encoded two and one predicted GH29 enzymes, respectively (Fig. 4D).The P. timonensis genome encoded two predicted sialidases with a GH33 domain and signal peptides with different domain structures.The NanH1 sialidase is predicted to be 412 amino acids in length, and NanH2 is a much larger protein with 1,029 amino acids.Both P. timonensis sialidases have predicted signal peptides and are therefore likely exported and/or secreted enzymes.The two P. timonensis sialidases have been biochemically characterized in a recent manuscript that is highly complementary to our study (38).The G. vaginalis genome encoded two predicted sialidases (NanH1 and NanH3, NanH2 was not present in our Gardnerella strain), and a single GH33 sialidase (NanH) was predicted for P. bivia (Fig. 4E).Earlier studies showed that G. vaginalis NanH2 and NanH3, but not NanH1, had high activity toward 4-methylumbelliferyl N-acetyl-α-D-neuraminic acid (4-MU-Neu5Ac) and bovine submaxillary mucin (39).These studies underpin the importance of studying enzyme activity.Altogether, our genomic analysis suggests that P. timonensis has a larger repertoire of potential mucin-degrading enzymes compared to the other BV-associated bacteria, including multiple fucosidases and sialidases.

P. timonensis displays high fucosidase and sialidase activity on the bacterial surface and in the supernatant
Sialidase and fucosidase activities in bacteria are often associated with pathogenic behavior as the removal of terminal monosaccharides from the mucin O-glycan structure promotes further degradation by exposing underlying glycans and the mucin peptide backbone that is sensitive to proteases (37).To assess the presence of sialidase and fucosidase activities in our vaginal bacterial strains, we performed culture-based assays.We also included Bacteroides fragilis as a positive control, an intestinal bacterium that is sometimes associated with vaginitis (40) and pelvic inflammatory disease (41) and is known to have sialidase and fucosidase activities.Bacteria were grown over night, followed by centrifugation, to separate the pellet from the supernatant fraction.To determine fucosidase and sialidase activities, both fractions were incubated with fluorescent substrates, and the fluorescence produced by each enzyme was measured.
No fucosidase activity was detected for G. vaginalis and L. crispatus, which was surprising as the G. vaginalis genome does encode a GH29 fucosidase (Fig. 4D).P. timonensis, P. bivia, and B. fragilis displayed high fucosidase activity in the bacterial pellet.In addition, fucosidase activity was detectable in the supernatants of P. timonensis and P. bivia, but only reached statistical significance in the case of P. timonensis compared to the fucosidase-negative G. vaginalis supernatant (Fig. 4F).For sialidase activity, the highest cell-bound activity could be measured for P. timonensis and P. bivia, followed by G. vaginalis and B. fragilis.Both P. timonensis and P. bivia sialidase activities were significantly higher than those of G. vaginalis.Of the supernatant fractions, only that of P. timonensis contained detectable sialidase activity, suggesting that this bacterium secreted or released sialidases into the medium under the conditions tested (Fig. 4G).

P. timonensis sialidase and fucosidase activity leads to O-glycan degradation at the vaginal surface
Next, we determined the O-glycan-degrading capacity of vaginal bacteria at the vaginal epithelial surface.Vaginal VK2/E6E7 monolayers were incubated with bacteria for 18 hours and stained with lectins to detect different mucin glycan structures including fucoses (UEA-1), α2,3 sialic acids (MAL-II), and α2,6 sialic acids (SNA).Visualization by confocal microscopy demonstrated that all glycan structures were present on the vaginal epithelial surfaces in the absence of bacteria (Fig. 5A through C, top panels).UEA-1 staining was most strongly reduced after incubation with P. timonensis, G. vaginalis, or recombinant fucosidase enzyme, but incubation with P. bivia and L. crispatus also changed the UEA-1 staining pattern (Fig. 5A).As genomes of L. crispatus strains ST1 and AB70 do not contain any predicted G29-or G95-domain fucosidase (our own analysis), we conclude that the altered staining pattern is not only related to fucosidase activity but could be, in part, the result of incubation of the epithelium with live bacteria.We quantified the UEA-1 signal intensity in all the different conditions and concluded that P. timonensis and G. vaginalis most significantly reduced UEA-1 staining compared to the untreated vaginal epithelium (Fig. 5D).Interestingly, G. vaginalis significantly reduced UEA-1 staining in this experimental setup, while G. vaginalis cultures grown in the CMM medium did not display fucosidase activity (Fig. 4F).Perhaps, this indicates that the predicted G. vaginalis single fucosidase gene 05980 (Fig. 4D) is induced by the interaction with vaginal epithelial cells.
For sialic acids, incubation with P. timonensis strongly and significantly decreased the staining for α2,3 sialic acids (MAL-II) and α2,6 sialic acids (SNA) on the vaginal epithelial surface (Fig. 5B through D).Incubation with G. vaginalis reduced MAL-II staining, but not SNA, but the reduction was less compared to that of P. timonensis (Fig. 5B through D).P. bivia and L. crispatus did not significantly reduce sialic acid staining (Fig. 5B through D).Overall, these results show that P. timonensis efficiently removes sialic acids and fucoses from the vaginal epithelium.
To investigate if the two P. timonensis sialidases (PtNanH1 and PtNanH2) could remove sialic acids from the epithelial surface, we incubated the vaginal monolayers with recombinantly expressed and purified PtNanH1 and PtNanH2.Incubation with either enzyme led to a significant reduction of both MAL-II and SNA (Fig. 6A through C).Next, vaginal epithelial monolayers were incubated with P. timonensis in the presence of either the broad sialidase inhibitor DANA or zanamivir, an inhibitor that was found in the accompanying study to be effective toward P. timonensis sialidases.In the presence of DANA or zanamivir, removal of α2,3 sialic acids and α2,6 sialic acids from the vaginal epithelial surface by P. timonensis was significantly reduced, demonstrating efficient inhibition of the bacterial sialidases, highlighting the role of these enzymes in glycan degradation (Fig. 6D through F).In conclusion, the BV-associated bacterium P. timonensis has a high potential for O-glycan degradation at the vaginal epithelial mucosal surface through a diverse array of glycosyl hydrolases, including highly active sialidases.

DISCUSSION
BV, one of the most common pathological conditions in women of different ages, increases susceptibility to sexually transmitted infections and negatively impacts fertility and quality of life.Unlike the health-associated vaginal microbiome, which is dominated by Lactobacillus species, BV is characterized by a polymicrobial infection of different anerobes including G. vaginalis, A. vaginae, and different Prevotella species (Fig. 1A).High sialidase activity can be detected in the vaginal discharge of women with BV (8), and persistence of sialidase-positive bacteria is a risk factor for subclinical intrauterine infections and preterm birth (42).
Thus far, G. vaginalis and P. bivia were considered to be the main producers of sialidases in the cervicovaginal environment (43,44).Our study, together with a complementary study by Pelayo et al. (38), demonstrates that P. timonensis has high sialidase activity and should be considered among the bacteria that play a pivotal role in the initiation and progression of BV.We found that P. timonensis had the highest sialidase activity of the BV-associated bacterial strains tested.In addition to cell-bound sialidase activity, P. timonensis was the only bacterium with detectable sialidase activity in the culture supernatant (Fig. 4G).It remains to be established whether P. timonensis sialidases are (in part) secreted or if they are released by proteolytic activity, as has been suggested for G. vaginalis enzymes (39).After attachment to vaginal epithelial cells, P. timonensis removed the majority of surface α-2,3-linked and α-2,6-linked sialic acids (Fig. 5B through  C), and the two identified P. timonensis sialidases (PtNanH1 and PtNanH2) were highly active at removing sialic acids from the vaginal epithelial cell surface (Fig. 6D through F).Notably, the sialidase activity of P. timonensis at the vaginal epithelial surface could be blocked with DANA and zanamivir inhibitors (Fig. 6A through C).In addition to sialidase activity, P. timonensis also displayed fucosidase activity in culture and during attachment to the vaginal epithelium (Fig. 4F and 5A).Metagenomics and metatranscriptomics analyses on bacterial vaginosis samples demonstrated that P. timonensis NanH2 is highly abundant in clinical samples (38).Sialidases and fucosidases are essential enzymes that can initiate degradation of mucin O-glycan structures of secreted and epithelium-bound mucins.It has been previously shown that the removal of sialic acids renders mucins more vulnerable to further degradation by glycosyl hydrolases and proteases, leading to exposure of the underlying epithelium (45).A recent study also demonstrated that recombinant sialidases of Gardnerella species efficiently desialylated vaginal glycans and treatment of VK2 cells with recombinant Gardnerella NanH2-induced pathways of cell death, differentiation, and inflammatory responses (46).In our experiments, P. timonensis did not induce a strong epithelial proinflammatory response (Fig. 2), despite its high sialidase and fucosidase activities at the vaginal epithelial surface.Perhaps, this suggests that P. timonensis has the means to suppress proinflammatory pathways while interact ing with the epithelium.We conclude that sialidases are important virulence factors that can contribute to the establishment and development of BV and that P. timonensis is a more crucial pathogen in BV than previously established.
Investigating bacterial nutritional preference for cervicovaginal mucus and glycogen is important to understand how different members of the vaginal microbiome thrive in this unique environment.Sialidases and fucosidases are crucial for bacterial growth on mucin (47).Besides these enzymes, P. timonensis also encoded a wide array of other mucin-degrading enzymes (Fig. 4A and C) and showed a small but significant growth on mucins as the sole carbon source (Fig. 3A through D).G. vaginalis and P. bivia encode fewer mucin-degrading enzymes (Fig. 4A and C) and did not grow in mucin-enriched media (Fig. 3B and C).For these experiments, we used pig gastric mucus (PGM) contain ing 5-N-glycolylneuraminic acid (Neu5Gc) (48).This mucus might be less suitable for human microbiota as human mucus does not contain Neu5Gc, and it was previously suggested that G. vaginalis is not capable of degrading Neu5Gc (44).Therefore, future experiments should be conducted with human (vaginal) mucus to conclusively establish the growth capacities of the different vaginal microbiota on cervicovaginal mucus.Glycogen is a large, highly branched D-glucose polymer that is abundant in vaginal tissue (49) but present in reduced levels in women with BV (50).Several Lactobacillus spp.have been shown to directly use glycogen for growth (51,52).In this study, we show that P. timonensis, G. vaginalis, and P. bivia were all able to grow on glycogen as the single nutrient source, which is in line with the presence of α-glucosidases in their genomes (17,34).Overall, glycogen utilization seems to be a shared trait of vagina-associated bacteria, indicating their metabolic adaptation to the vaginal environment.
Adhesion to the cell epithelium is a crucial step in BV, and many studies in the field indicate a stepwise disease progression with primary and secondary bacterial colonizers (Fig. 1A).L. crispatus demonstrated the highest capacity to adhere to vaginal epithelial cells (Fig. 1B and C).The mechanisms by which BV-associated bacteria, such as G. vaginalis, displace adherent Lactobacillus species from the vaginal epithelial surface are not yet fully understood.Antibiotic use, sexual intercourse, menstruation, and sexually transmitted infections can lead to microbiota changes, physical disruption, pH alterations, hormonal changes, and inflammation, which can all impact Lactobacillus attachment (53).
Subsequently, disruptions in the Lactobacillus biofilm can lead to attachment and proliferation of other bacteria, including G. vaginalis and P. timonensis.G. vaginalis is an important early colonizer as this bacterium can adhere to the vaginal epithelium and potentially form a biofilm (16).Other anerobic bacteria such as P. bivia and A. vaginae can join the G. vaginalis biofilm as secondary colonizers (32,54).In the current study, we demonstrate that P. timonensis, similar to G. vaginalis but unlike P. bivia, can efficiently bind to both vaginal and endocervical cells (Fig. 1B, D and E).Previously, it has been shown that P. timonensis can induce elongated microvilli in a 3D endometrial epithelial cell model, and it was speculated that these changes might induce increased adhesion of this species and of other secondary colonizers (27).Based on these combined observa tions, we propose that P. timonensis may be an initial colonizer of the vaginal epithelium and does not require an established G. vaginalis biofilm.After attachment, the high sialidase and fucosidase activities of P. timonensis at the vaginal epithelial surface remove the protective terminal glycans of the glycocalyx, likely creating new binding sites for secondary colonizers (46,55), which can lead to bacterial colonization of the upper parts of the FRT (56).
While P. timonensis is perhaps an initial colonizer during BV, it does not contribute to cytotoxicity nor does it induce a proinflammatory response in a similar manner to G. vaginalis.Only G. vaginalis induced high LDH release by both vaginal and endocervical cells, while P. timonensis, P. bivia, and L. crispatus were not cytotoxic (Fig. 2A and B).To induce cytotoxicity, G. vaginalis expresses the cytotoxin vaginolysin (vly) (20) and also utilizes membrane vesicles (57).Cytotoxicity might be an important aspect of G. vaginalis virulence, as strains isolated from women with BV were more cytotoxic than non-BV isolates (22).In our infection experiments with single species of bacteria, G. vaginalis strongly induced the expression of IL-1β, IL-8, or CCL20 in the vaginal and endocervical epithelium, while P. timonensis, P. bivia, and L. crispatus did not significantly induce proinflammatory cytokine response (Fig. 2C through E), which was in line with previous reports that investigated single species (27,58).These data are in line with strong epithelial proinflammatory responses that were previously observed for G. vaginalis and A. vaginae (58,59) and demonstrate that P. timonensis by itself is not equally pathogenic.When interacting with dendritic cells, P. timonensis does induce a strong proinflammatory immune response (30), and we previously demonstrated that P. timonensis is unique in increasing HIV-1 uptake by Langerhans cells, turning these cells into HIV-1 reservoirs (31).Because in vivo observations in cervicovaginal fluid indicate increased cytokine levels when Prevotella spp.are present in the vagina (28,29), the contributions of different Prevotella species to proinflammatory responses in more complex polymicrobial infections remain to be established.
This study provides evidence that the understudied vaginal bacteria P. timonensis has pathogenic properties that could support primary colonization of the female reproduc tive tract in BV.A complementary publication demonstrates that these bacteria and their sialidases are abundant in clinical BV samples (38).Unlike G. vaginalis, the virulence traits of P. timonensis do not include cell cytotoxicity nor triggering of a strong proin flammatory response by the epithelium but rather a strong and previously unappreci ated capacity to degrade the protective epithelial mucus layer through sialidase and fucosidase activities.We also demonstrate that the P. timonensis sialidase activity at the vaginal epithelial glycocalyx can be efficiently inhibited by small-molecule inhibitors.For G. vaginalis, it was previously demonstrated that a sialidase inhibitor also reduced cellular invasion (60).The application of sialidase inhibitors in BV treatment might therefore be an interesting novel therapeutic approach to reduce bacterial adhesion, invasion, and mucosal damage.

MATERIALS AND METHODS
Cell lines, bacterial strains, and culture conditions VK2/E6E7 (ATCC, CRL-2616) and End1/E6E7 (ATCC, CRL-2615) cells were routinely grown and maintained as indicated in the supplemental material.The bacterial strains used in this study are listed in Table 1.Bacterial media and growing conditions can be found in the supplemental material.

Adhesion assays
VK2/E6E7 and End1/E6E7 cells were seeded in a 12-well plate and grown until full confluency.Cells were infected with overnight bacterial cultures at an MOI of 10 for 18 hours in anerobic conditions.Serial dilutions from the supernatant and cell suspen sions were plated in their specific plate media.Colonies were counted to calculate the percentage of adherent bacteria, as described in the supplemental methods.

Peptide nucleotide acid (PNA) probe in silico design
PNA probes for the specific detection of P. timonensis or P. bivia were designed using the protocol described in detail in the supplemental materials.The resulting PNA probes were named PT-Cy3 and PB-Cy3.The Gard162-AF488 (61) and EUB338-AF488 (62) PNA probes used in this study have been previously described.All probes are listed in Table 2.

Cytotoxicity assays
VK2/E6E7 and End1/E6E7 cells were grown until full confluency in 96-well plates.Overnight cultures of P. timonensis, P. bivia, G. vaginalis, and L. crispatus were used to infect the cells at an MOI of 10 and 100 for 18 hours under anerobic conditions.The presence of released LDH in the supernatant was assessed using the Cytotox 96 Non-Radioactive Cytotoxicity Assay (Promega, G1780).The extended protocol can be found in the supplemental material.

Reverse transcription-quantitative polymerase chain reaction (RT-qPCR)
Non-confluent VK2/E6E7 and End1/E6E7 cells were infected with G. vaginalis, P. timonensis, or P. bivia at an MOI of 10 for 18 hours at 37°C in anerobic conditions.RNA was extracted and treated with DNase I. Primers used in the RT-qPCRs can be found in Table S1.All cycle thresholds were averaged from triplicate reactions and normalized to the housekeeping gene TMEM222.Fold changes were calculated using the delta-delta Ct method.Additional details of bacterial infection and RT-qPCR protocols can be found in the supplemental material.

Mucin and glycogen growth assays
Mucins were purified from commercially available porcine gastric mucins (PGM, Sigma-Aldrich, M2378) as described in the supplemental material.Bacterial cultures of P. timonensis, P. bivia, G. vaginalis, and A. muciniphila were diluted to OD 600 = 0.02 in their respective medium without carbohydrates.In a 24-well plate, diluted bacterial cultures were mixed 1:1 with their corresponding medium without carbohydrates, supplemented with purified mucins or glycogen (Sigma-Aldrich, 10901393001) at a final concentration of 0.5%(wt/vol).As a positive control, complete medium with carbohydrates was used.The cultures were incubated in anerobic conditions for up to 56 hours at 37°C.During incubation, the absorbance of 100 µL of each culture was measured at 595 nm with the FLUOstar Omega microplate reader at 24, 32, 48, and 56 hours for P. timonensis and A. muciniphila, and at 16, 24, 40, and 48 hours for the faster-growing bacteria P. bivia and G. vaginalis.

Bacterial whole-genome sequencing and CAZyme analysis
Bacterial DNA was isolated and sequenced using Nanopore technology.Detailed information can be found in the supplemental material.Predicted bacterial protein sequences were used to analyze the presence of carbohydrate-active enzymes using the CAZy database and dbCAN3 meta server.CAZymes identified with at least two out of three tools (HMMER: dbCAN, DIAMOND: CAZy, and HMMER: dbCAN_sub) were considered for further analysis.

Enzymatic activity assays
Fucosidase and sialidase activities were measured in overnight bacterial pellets and concentrated supernatants using fluorogenic substrates 4-methylumbelliferyl α-L-fucopyranoside and 4-methylumbelliferyl N-acetyl-a-D-neuraminic acid sodium salt, respectively.The detailed enzyme assay protocols can be found in the supplemental material.

Cloning, heterologous expression, and purification of P. timonensis sialidase genes
Full-length sialidase genes were amplified from genomic DNA purifications using primer pairs depicted in Table S2.The resulting gene products were assembled into the pET28a expression vector and transformed into DH5α Escherichia coli chemically competent cells.Confirmed plasmids were transformed into E. coli BL21 (DE3) for protein expression.A more extensive detailed protocol can be found in the supplemental material.

FIG 1
FIG 1 Prevotella timonensis can adhere to the vaginal and endocervical epithelium.(A) Schematic representation of the different microbial communities of the vaginal epithelium in the healthy state and during the development of bacterial vaginosis.(B) Percentage of adhesion of L. crispatus (LC), P. timonensis (PT), P. bivia (PB), and G. vaginalis (GV) to VK2/E6E7 and End1/E6E7 cells assessed by quantification of colony-forming units (CFUs).The graph represents the average and SEM of at least three to four independent experiments.Statistical test: one-way ANOVA with Dunnett's correction compared to L. crispatus in each cell line.*P < 0.05; **P < 0.01; ****P < 0.0001.(C, D) Fluorescence in situ hybridization (FISH) in combination with confocal microscopy of L. crispatus, P. timonensis, P. bivia, and G. vaginalis adhesion to (C) VK2/E6E7 and (D) End1/E6E7 cells stained for wheat germ agglutinin (WGA) and using peptide nucleotide acid (PNA) probes.For each bacterium, the corresponding PNA signal is shown in red, cell surface in cyan (WGA), and 4′,6-diamidino-2-phenylindole (DAPI) in white.White scale bars represent 20 µM.

FIG 4
FIG 4 High mucin degradation potential in Prevotella timonensis.(A) Abundance of predicted carbohydrate-active enzymes (CAZymes) families in the sequenced genomes of our P. timonensis, P. bivia, G. vaginalis, and A. muciniphila strains.(B) Schematic representation of a mucin glycoprotein molecule with a protein backbone and diverse O-glycan structures.Target sites for different classes of glycosyl hydrolases are depicted.(C) Number of identified O-glycan-targeting CAZymes in the genomes of the sequenced P. timonensis (PT), P. bivia (PB), G. vaginalis (GV), and A. muciniphila (AM) strains.(D-E) Domain architecture of the predicted (D) fucosidases and (E) sialidases of the designated bacteria.The displayed domains are identified by HMMER, Diamond, and Signal IP 6.0 tools and drawn to scale.(F) Fucosidase and (G) sialidase activities measured in bacterial pellets and supernatants of the different BV-associated bacteria and B. fragilis as the positive control.Abbreviations: PT (P.timonensis); PB (P.bivia); GV (G.vaginalis); BF (B.fragilis); and LC (L.crispatus).The graph represents the average ± SEM of four independent experiments.Statistical test: one-way ANOVA with Tukey's HSD correction.ns: not significant; *P < 0.05; ** P < 0.01; *** P < 0.001; ****P < 0.0001.

TABLE 1
Overview of bacterial strains used in this study a n/a, not applicable.Research Article mBioSeptember 2024 Volume 15 Issue 9 10.1128/mbio.00691-24 13 Briefly, confluent monolayers of VK2/E6E7 and End1/E6E7 cells were infected with G. vaginalis, P. timonensis, or P. bivia at an MOI of 50 for 18 hours at 37°C in anaerobic conditions.Cells were washed and stained with wheat germ agglutinin-663 (WGA-633, Invitrogen, W21404).Then, cells were fixed and stained with 1000 nM EUB338-AF488 probe, 600 nM Gard162-AF488 probe, 200 nM PT-Cy3, or 200 nM PB-Cy3 probe in hybridization buffer for 2 hours at 50°C in a humidity chamber.Slides were washed, stained with DAPI, and mounted for imaging on a Leica SPE-II confocal microscope.Additional details of the FISH staining protocol can be found in the supplementary materials.

TABLE 2
Sequences of the PNA/FISH probes