Protease activities of vaginal Porphyromonas species disrupt coagulation and extracellular matrix in the cervicovaginal niche

Porphyromonas asaccharolytica and Porphyromonas uenonis are common inhabitants of the vaginal microbiome, but their presence has been linked to adverse health outcomes for women, including bacterial vaginosis and preterm birth. However, little is known about the pathogenesis mechanisms of these bacteria. The related oral opportunistic pathogen, Porphyromonas gingivalis, is comparatively well-studied and known to secrete numerous extracellular matrix-targeting proteases. Among these are the gingipain family of cysteine proteases that drive periodontal disease progression and hematogenic transmission to the placenta. In this study, we demonstrate that vaginal Porphyromonas species secrete broad-acting proteases capable of freely diffusing within the cervicovaginal niche. These proteases degrade collagens that are enriched within the cervix (type I) and chorioamniotic membranes (type IV), as well as fibrinogen, which inhibits clot formation. Bioinformatic queries confirmed the absence of gingipain orthologs and identified five serine, cysteine, and metalloprotease candidates in each species. Inhibition assays revealed that each species’ proteolytic activity can be partially attributed to a secreted metalloprotease with broad substrate specificity that is distantly related to the P. gingivalis endopeptidase PepO. This characterization of virulence activities in vaginal Porphyromonas species highlights their potential to alter the homeostasis of reproductive tissues and harm human pregnancy through clotting disruption, fetal membrane weakening, and premature cervical remodeling.


INTRODUCTION
The vaginal microbiome of healthy reproductive-age women is typically characterized by low species diversity, with Lactobacillus dominating the vaginal and ectocervical niches of the lower genital tract 1,2 . A community shift towards high species diversity, with overgrowth of anaerobic bacteria, is associated with increased risk of bacterial vaginosis (BV) 2,3 , acquisition and transmission of sexually transmitted infections [4][5][6] , preterm birth [7][8][9] , and cervical cancer [10][11][12] . Intriguingly, some women harboring a diverse cervicovaginal microbiome are healthy and asymptomatic 2 , suggesting a need to untangle how specific species contribute to poor outcomes. Among BV-associated bacteria, Gram-negative anaerobic rods corresponding to Prevotella and black-pigmented Porphyromonas species are frequently detected in vaginal samples and significantly associated with the Bacteroides morphotype from Nugent scoring 13 . While Prevotella is an abundant, species-rich, and relatively well-studied vaginal clade, comparatively little is known about Porphyromonas species inhabiting the human vagina. No Porphyromonas species is currently thought to be specific to the human urogenital tract, but P. asaccharolytica, P. uenonis, P. bennonis, and P. somerae (in decreasing order of cervicovaginal microbiome citation frequency) exhibit a preference for these niches 3,14,15 . P. asaccharolytica and P. uenonis colonize the vagina in 15-50% of healthy women and although their prevalence and abundance increase with BV, they are typically considered low abundance taxa 13,[16][17][18][19][20] . Recent studies show these species are predictors of spontaneous preterm labor 9,21 , pelvic inflammatory disease 22,23 , human papillomavirus (HPV) infections progressing to cervical neoplasia 24,25 , and uterine cancer 26,27 . Thus, an improved understanding of the functional capacity of vaginal Porphyromonas species is needed.
To date, the only Porphyromonas species that has been wellcharacterized is P. gingivalis, a low abundance species in the oral microbiome of both healthy patients and those with gingivitis 28,29 . As a keystone species driving oral (plaque) biofilm formation and periodontal disease progression 30,31 , P. gingivalis contributes to local tissue destruction directly and indirectly through the induction of inflammatory processes 32,33 . P. gingivalis can also disseminate via the bloodstream to distal infection sites such as the endocardium and joints 34,35 . During pregnancy, P. gingivalis has been isolated from the placenta and amniotic fluid of women who delivered preterm [36][37][38] , and in mouse infection models, P. gingivalis induces preterm labor via inflammatory activation of the chorioamniotic membranes 34,39,40 . Pathogenesis mechanisms contributing to these outcomes include a wide array of proteolytic activities carried out by numerous secreted proteases. Among these are the gingipain family of cysteine proteases that drive periodontal disease progression 41 , hematogenic transmission to the placenta 29,40,42,43 , and preterm labor induction in mice 40 . The gingipains degrade many extracellular matrix components, including collagen 44,45 , and amplify their effects by activating and upregulating host matrix metalloproteases (MMPs) that also degrade collagen 46,47 . Furthermore, gingipains can degrade immune factors including immunoglobulins 44,48 , complement components 49,50 , cytokines 51,52 , clotting factors 53,54 , and antimicrobial peptides 55 , giving rise to a favorable immune environment for P. gingivalis colonization.
Proteolytic activity has been previously detected in vaginal fluid from patients with BV [56][57][58] and characterized in clinical isolates of BV-associated bacteria 59,60 . In fact, collagenase (gelatinase) and caseinase activity of P. asaccharolytica (formerly Bacteroides asaccharolyticus) was previously reported in screens of Bacteroides species detected in human infections 61 and clinical isolates from reproductive tract infections 60 . However, further characterization of the enzymes responsible was not conducted, and proteolytic activity of P. uenonis has yet to be explored. Given their phylogenetic relatedness and epidemiological similarity to P. gingivalis-exhibiting high prevalence, low abundance, and association with the disease-we sought to determine whether vaginal Porphyromonas species possess the broad-acting proteolytic virulence activity of the periodontal pathogen. In this study, we show that P. asaccharolytica and P. uenonis are capable of degrading multiple extracellular matrix components in the female genital tract. Our study furthermore reveals differences between the species, suggesting preterm birth-associated P. asaccharolytica may secrete more enzymes that contribute to collagenase activity. Finally, we identify and functionally characterize a P. asaccharolytica virulence factor-a metalloprotease that is highly conserved in P. uenonis and distantly related to the PepO endopeptidase in P. gingivalis and other Porphyromonas species 62,63 . We demonstrate this protein exhibits broad-acting proteolytic capacity, which may make it a key microbial virulence factor in the pathogenesis of gynecological and reproductive health conditions.

RESULTS
Vaginal Porphyromonas species degrade type I collagen, type IV collagen, and casein using secreted proteases Collagenase activity of P. asaccharolytica and P. uenonis was evaluated using fluorescently quenched substrates (type I collagen or type IV collagen), where proteolytic digestion results in dequenching and measurable increases in fluorescence over time. Fluorometric collagenase assays confirmed that P. asaccharolytica and P. uenonis cell suspensions degrade type I collagen in a dose-dependent manner (Fig. 1a). Next, collagenase activity was measured in cell-free supernatants from P. asaccharolytica and P. uenonis, validating that both organisms secrete proteases capable of degrading type I and type IV collagen (Fig. 1b, c). Collagenase activity was further confirmed with gelatin zymography (Supplementary Fig. 2), where P. asaccharolytica supernatants produced three distinct high molecular weight zones of clearing (~85, 95, and 120 kDa), while P. uenonis supernatants generated four zones of clearing: three high molecular weight (~75, 90, and 110 kDa) and one low molecular weight (30 kDa). Due to its low-complexity tertiary structure, casein is regarded as a universal protease substrate that is highly susceptible to proteolytic degradation. Fluorometric assays revealed that supernatants from P. asaccharolytica and P. uenonis possess caseinase activity (Fig. 1d). This was further confirmed using agar-based casein degradation assays, where P. asaccharolytica, P. uenonis, and P. gingivalis cell suspensions all produced zones of clearing ( Supplementary Fig. 3). Although P. asaccharolytica and P. uenonis can degrade similar substrates as P. gingivalis 41 , P. gingivalis showed substantially higher maximum fluorescence and area under the curve for collagen degradation (Supplementary Fig. 4 and Supplementary Tables 1, 2). The area under the curve for casein degradation by P. gingivalis was also much higher than for P. asaccharolytica or P. uenonis, while the maximum fluorescence and time to maximum fluorescence for casein degradation were comparable between all three Porphyromonas species (Supplementary Fig. 4 and Supplementary Table 2). To understand if proteolytic activity might contribute to pathogenesis in the female genital tract, we evaluated whether a common commensal vaginal microbe is also capable of degrading collagen and casein. No collagenase or caseinase activity was detected from Lactobacillus crispatus cell Fig. 1 Proteolytic activity of vaginal Porphyromonas species. a Cell suspensions of P. asaccharolytica and P. uenonis at 10 7 or 10 8 CFU/reaction were incubated with fluorophore-conjugated type I collagen. Results are presented as mean ± standard error from three independent experiments performed in technical triplicate or quadruplicate. Collagen degradation was measured every 3 min by detecting the increase in fluorescence over a 2-h time course. b, c Cell-free supernatants of P. asaccharolytica and P. uenonis were incubated with fluorophore-conjugated b type I or c type IV collagen over an 18-h time course. Results are presented as mean ± standard error from seven independent experiments (type I collagen) and four independent experiments (type IV collagen) performed in technical triplicate. d Cell-free supernatants of P. asaccharolytica and P. uenonis were incubated with fluorescein (FITC)-conjugated casein over a 5-h time course. Results are presented as mean ± standard error from five independent experiments performed in technical triplicate. The mean fluorescence readings of the negative control were subtracted from experimental wells and relative fluorescence units (RFU) were plotted over time, with negative values adjusted to zero.
P. asaccharolytica and P. uenonis inhibit fibrin clot formation through fibrinogen degradation Since gingipains are known to degrade fibrinogen and exacerbate gum bleeding 41 , we next investigated whether P. asaccharolytica and P. uenonis proteases can degrade fibrinogen and impair fibrin clot formation. To evaluate direct fibrinogen degradation, cell-free media controls and cell suspensions of P. asaccharolytica or P. uenonis were incubated in the presence and absence of human fibrinogen over a 24-h time course. Samples removed at defined intervals were separated by SDS-PAGE and stained with Coomassie Brilliant Blue. When P. asaccharolytica or P. uenonis were incubated with fibrinogen, complete degradation of the fibrinogen α and β chains was observed after 18 h, while the γ chain remained intact (Fig. 2a, b). To determine whether fibrinogen degradation translates to impaired fibrin clotting, thrombininduced fibrin clot formation was measured after Porphyromonas cell suspensions were preincubated with citrated plasma. A significant delay in clot formation was observed with the highest dose of P. asaccharolytica or P. uenonis (9.0 × 10 9 CFU/reaction) compared to the no cell control (Fig. 2c, ****p < 0.0001) or a lower dose of P. asaccharolytica or P. uenonis ( Fig. 2c; ****p < 0.0001). Turbidimetry at the experimental endpoint (30 min) allowed for quantitative evaluation of final clot size. In keeping with clotting times, fibrin clot size was significantly reduced in samples exposed to P. asaccharolytica or P. uenonis at 9.0 × 10 9 CFU/reaction compared with the no cell control (Fig. 2d ****p < 0.0001) or lower doses of bacteria (Fig. 2d, ***p = 0.0003 for P. asaccharolytica and **p = 0.003 for P. uenonis). These findings were further confirmed by visual assessment of clot formation at assay endpoints (Fig. 2e).
Identification of collagenase candidates in P. asaccharolytica and P. uenonis Since our analyses confirmed previous reports that P. asaccharolytica and P. uenonis do not encode gingipain orthologs (Supplementary Note, Supplementary Fig. 1, and Supplementary Tables 3-5), we sought to identify other candidate proteins that may be responsible for the collagenolytic activity of vaginal Porphyromonas species. To identify putative collagenases amongst the 59 and 64 peptidases predicted for P. asaccharolytica and P. uenonis by the MEROPs database 64 , we cross-referenced a list of known and predicted microbial collagenases in the BRENDA enzyme information database 65 . This approach shortened the candidate peptidase list to 18 enzymes in P. asaccharolytica and 14 in P. uenonis. InterPro scans of each putative collagenase revealed proteins likely to be involved in cell wall synthesis or export machinery and narrowed the list to ten peptidases in P. asaccharolytica and nine peptidases in P. uenonis ( Supplementary  Fig. 1). Factoring in similarity to characterized collagenases, additional domains identified in InterPro, and the presence of secretion signals, a final short-list of the seven most promising candidate peptidases in each organism was generated (Tables 1, 2  and Supplementary Table 6). Intriguingly, each organism's candidate collagenases could be organized into four groups: Igcontaining serine proteases (2x), C10 cysteine proteases (2x), M13 metalloproteases (1x), and U32 proteases (2x) (Tables 1, 2). Multiple sequence alignments of candidate collagenase pairs for a given type within each strain revealed low amino acid sequence identity (29-45%, Supplementary Table 7). Conversely, BLASTP searches between species allowed for the identification of orthologous protein pairs in P. asaccharolytica and P. uenonis, with amino acid identity ranging from 56 up to 97% (Supplementary Table 7). Notably, all within-genome candidate collagenases pairs had lower amino acid identity compared to the ortholog identified in the other Porphyromonas genome (Supplementary Table 7).
Serine proteases within the short-list all contained Ig-like folds (Tables 1, 2), which are also present in the binding domain of the well-characterized collagen-degrading metalloproteases from Clostridium histolyticum (ColG, ColH) 66 . The candidate cysteine proteases contained type 9 secretion system signal domains and a SpeB domain (Tables 1, 2), indicating sequence similarity with Streptococcus pyogenes streptopain, a cysteine protease capable of cleaving host components such as fibrinogen, immunoglobulins, and complement proteins [67][68][69] . The candidate metalloproteases from P. asaccharolytica (Poras_0079) and P. uenonis (Poru_00076) each contained a catalytic collagenase domain in addition to an M13 type metallopeptidase domain and a predicted N-terminal secretion signal (Tables 1, 2). Finally, two U32 collagenases were detected in each vaginal Porphyromonas species indicating an orthologous relationship with the P. gingivalis U32 collagenase PrtC 70,71 . Of note, none of these putative U32 collagenases were found to possess secretion signals indicative of localization to the extracellular space (Tables 1, 2).
Vaginal Porphyromonas species encode metalloproteases targeting collagens, casein, and fibrinogen To further narrow down candidate enzymes responsible for collagenolytic and fibrinogenolytic activities, inhibitors of the predicted serine, cysteine, and metalloproteases (Tables 1, 2) were incorporated into functional assays. Cell-free supernatants from P. asaccharolytica and P. uenonis were incubated with type I collagen in the presence of three doses of 1,10-phenanthroline, iodoacetamide, or aprotinin to inhibit metallo-, cysteine, and serine proteases, respectively ( Fig. 3a-d and Supplementary Fig. 6). Treatment of P. asaccharolytica supernatants with 1,10-phenanthroline resulted in decreased collagenase activity, with a statistically significant reduction in max enzyme activity observed with the highest dose of inhibitor (  Fig. 6). However, when P. asaccharolytica supernatants were treated with a combination of 1,10-phenanthroline and iodoacetamide, there was a further reduction in collagenase activity compared to 1,10-phenanthroline alone (Fig. 3a, c, **p = 0.0092). There was no statistically significant decrease in the max collagenase activity secreted by P. asaccharolytica with aprotinin ( Fig. 3a, c and Supplementary Fig. 6). For P. uenonis, 1,10phenanthroline also inhibited collagenase activity (Fig. 3b, d and Supplementary Fig. 6), with a significant reduction in max collagenase activity ( Fig. 3d; 0.2 vs. 0.02 mM ***p = 0.0008; 0.2 vs. 0.002 mM **p = 0.0012). However, treatment with iodoacetamide or aprotinin did not reduce P. uenonis collagenase activity (Fig. 3b, d and Supplementary Fig. 6) and the combination treatment of 1,10-phenanthroline and iodoacetamide did not provide an additional reduction in collagenase activity or max fluorescence (Fig. 3b, d) as observed in P. asaccharolytica (Fig. 3a, c).
Next, we evaluated whether the same protease classes degrade type IV collagen and casein. For both Porphyromonas species, the metalloprotease inhibitor 1,10-phenanthroline completely abrogated type IV collagenase activity from supernatants, while serine and cysteine protease inhibitors did not significantly reduce activity ( Fig. 3e-h). P. asaccharolytica caseinase activity was similar to its type I collagenase activity (Fig. 4a, b) as proteolytic activity was reduced by treatment with 1,10-phenanthroline, and treatment with both 1,10-phenanthroline and iodoacetamide produced an additional downward shift in enzyme activity that resulted in a significant reduction in max fluorescence when compared to the no inhibitor control (Fig. 4a, c; p = 0.0413). In keeping with the type I collagen and type IV collagen degradation results, P. uenonis caseinase activity was significantly decreased by treatment with the metalloprotease inhibitor 1,10-phenanthroline (Fig. 4c, d; p = 0.018), while aprotinin and iodoacetamide treatment did not inhibit activity (Fig. 4b, d). Further to this, the 1,10-phenanthroline/iodoacetamide combination did not offer any additional reduction in protease activity (Fig. 4b, d).
Inhibitors were also incorporated into the fibrinogen degradation assay to determine whether fibrinogen is proteolyzed by the Fig. 2 Vaginal Porphyromonas species degrade fibrinogen and impair clot formation. a, b SDS-PAGE of a P. asaccharolytica and b P. uenonis incubated with human fibrinogen (+Fg), saline (no Fg), or a cell-free media control (sBHI Media + Fg). Samples were assessed for fibrinogen degradation, indicated by the absence of bands corresponding to α, β, and γ fibrinogen chains in "Porphyromonas cells + Fg" treatments compared to "no Fg" or media controls. c Time from thrombin addition to fibrin clot formation. Citrated plasma was preincubated with P. asaccharolytica, P. uenonis, or no cell controls. Experiments were performed in technical duplicate and results are presented as mean ± standard error from five independent experiments. d Quantitative evaluation of clot size via well scan optical density measurements at the experimental endpoint. Experiments were performed in technical duplicate and control reactions without thrombin were used to blank the experimental reactions. Data were normalized to the average clot size of the no cell control within each experiment. Results are presented as mean ± standard error from five independent experiments. Significance was assessed by one-way ANOVA with Holm-Sidak's multiple comparisons test (****p < 0.0001, ***p < 0.0003, **p = 0.003). e Endpoint qualitative evaluation of fibrin clots (>30 min) after clotting time assay with P. asaccharolytica (2.4 × 10 9 CFU/reaction), P. uenonis (3.0 × 10 9 CFU/reaction), or no cell control. same enzyme classes as those observed in our collagen and casein experiments. Delayed fibrinogen degradation by P. asaccharolytica was observed in the presence of 1,10-phenanthroline, but not with aprotinin or iodoacetamide (Fig. 5a), suggesting that secreted metalloproteases from P. asaccharolytica contribute to fibrinogen degradation. Fibrinogen degradation patterns from P. asaccharolytica and P. uenonis supernatants (Fig. 5) differed from the results obtained with cell suspensions (Fig. 2a-b). For both species, the α chain was only partially degraded by the supernatants (Fig. 5). For P. uenonis, the β and γ chains remained intact for the 48-hour time course (Fig. 5b). Furthermore, P. uenonis partial degradation of the α chain was not impacted by the inhibitors included in this study (Fig. 5b).
Finally, we sought to further characterize the Porphyromonas M13 metalloproteases identified in our bioinformatics inquiries (Tables 1, 2 and Fig. 6a). Genomic exploration of other Porphyromonas species detected in the urogenital tract and commonly isolated from other human body sites revealed that the M13 metalloproteases are ubiquitous (Fig. 6b). Notably, the M13 metalloprotease in P. gingivalis has been previously characterized as PepO, a secreted endopeptidase involved in host attachment/ invasion and proteolytic activation of endothelin, a potent peptide that induces vasoconstriction 62,63,72 . To determine whether the vaginal Porphyromonas M13 metalloproteases have collagenase and caseinase activity, Poras_0079 (pepO), was cloned and expressed using the myTXTL ® in vitro transcription/translation system. Expression of PepO was confirmed via SDS-PAGE with the appearance of a 76 kDa protein in PepO reactions but not in the RNA polymerase only control reactions (Fig. 7a). Fluorescent protease assays revealed that P. asaccharolytica PepO is capable of degrading casein and type I collagen, but not type IV collagen ( Fig. 7b-d). Further, the metalloprotease inhibitor 1,10-phenanthroline fully abrogated the type I collagenase and caseinase activities of PepO (Fig. 7b, c).

DISCUSSION
It is well established that mucinase activity is elevated during BV [73][74][75][76] and attributed to Gardnerella 75,77 and Prevotella species 75,78,79 . In support of this, sialidase activity has been utilized as a diagnostic marker for BV 78,[80][81][82][83] . Although most studies have focused on the degradation of mucin glycans, proteolytic activity in the vaginal fluid has also been linked with BV status 56-58 and described among predominant BV-associated bacteria 59,60 . Isolates of Prevotella bivia from women with preterm premature rupture of membranes (PPROM) were shown to secrete proteases that degrade elastin, collagen, casein, and gelatin 59 . In another study screening bacterial strains from women with PPROM, preterm labor, or puerperal infection, protease activity was confirmed in phylogenetically diverse Gram-negative and Gram-  positive organisms. This included the BV-associated bacteria Gardnerella vaginalis, Prevotella bivia (formerly Bacteroides bivius), and P. asaccharolytica (formerly Bacteroides asaccharolyticus) 60 . The authors found that P. bivia and P. asaccharolytica exhibited gelatin and casein degradation, while G. vaginalis exclusively degraded casein. Our findings confirm the collagen/casein degradation capacity of P. asaccharolytica and extend these activities to P. uenonis. Furthermore, we demonstrate that both vaginal Porphyromonas species also degrade type IV collagen, commonly found in reproductive tissues 84,85 , and describe the diverse protease types capable of coordinating these activities.
Our in vitro analyses show that the proteolytic activities of P. asaccharolytica and P. uenonis are dose-dependent across a range of 10 7 -10 9 cells/mL. With concentrations of these microbes reported at 10 3 -10 6 cells/mL in healthy women 86,87 , and up to 10 9 cells/mL in populations that include women with BV 15 , it is reasonable to expect that although vaginal Porphyromonas species are not high in relative abundance, they, like P. gingivalis, have the potential to cause disease in a protease-and communitydependent fashion 88 . To understand whether proteolytic activity is also observed among commensal bacteria that inhabit the vaginal niche, we investigated the proteolytic activity of Lactobacillus  Fig. 3 Vaginal Porphyromonas protease types display type I and type IV collagenase activity. a, b Type I collagen degradation of a P. asaccharolytica and b P. uenonis cell-free supernatants in the presence of 0.2 mM 1,10-phenanthroline, 0.4 mM iodoacetamide, 0.01 mM aprotinin or 0.2 mM 1,10-phenanthroline + 0.4 mM iodoacetamide. Results are normalized to the no inhibitor control (set to 1) and presented as mean ± standard error from four independent experiments. c, d Maximum fluorescence of c P. asaccharolytica and d P. uenonis secreted collagenase activity, normalized to the no inhibitor control and presented as mean ± standard error from four independent experiments, three independent experiments (P. asaccharolytica +0.01 mM aprotinin) or two independent experiments (P. uenonis +0.01 mM aprotinin) performed in technical triplicate. Statistical significance was assessed by one-way ANOVA and Tukey's post hoc comparison P. asaccharolytica: 1,10-phenanthroline 0.2 vs. 0.02 mM ***p = 0.0009; 1,10-phenanthroline 0.2 vs. 0.002 mM ***p = 0.0005; 1,10-phenanthroline 0.2 vs. 0.2 mM 1,10-phenanthroline +0.4 mM iodoacetamide **p = 0.0092. P. uenonis: 1,10-phenanthroline 0.2 vs. 0.02 mM ***p = 0.0008; 1,10-phenanthroline 0.2 vs. 0.002 mM **p = 0.0012. e, f Type IV collagen degradation by e P. asaccharolytica and f P. uenonis cell-free supernatants in the presence of 0.2 mM 1,10-phenanthroline, 0.4 mM iodoacetamide, 0.01 mM aprotinin, or 0.2 mM 1,10-phenanthroline + 0.4 mM iodoacetamide. Results are normalized to the no inhibitor control and presented as mean ± standard error from four independent experiments. g, h Maximum fluorescence of g P. asaccharolytica and h P. uenonis secreted collagenase activity in the presence of inhibitors. Results are presented as mean ± standard error from four independent experiments performed in technical triplicate. Statistical significance was assessed by one-way ANOVA and Tukey's post hoc comparison P. asaccharolytica: 1,10-phenanthroline 0.2 mM vs. no inhibitor *p = 0.0352.; P. uenonis: 1,10phenanthroline 0.2 mM vs. no inhibitor *p = 0.0398. The mean fluorescence readings of the negative control were subtracted from experimental wells and relative fluorescence units (RFU) was plotted over time, with negative values adjusted to zero. crispatus, demonstrating that L. crispatus is not capable of degrading type I collagen or casein. The absence of detectable proteolytic activity from a commensal Lactobacillus strain and the growing evidence for secreted proteolytic activity from BVassociated bacteria 59,60 , including vaginal Porphyromonas species, suggests that degradation of host proteins could be an important virulence trait of opportunistic pathogens in the female genital tract.
Our results demonstrate that vaginal Porphyromonas species are capable of directly degrading fibrinogen and impairing fibrin clot formation. Fibrinogen is detected in vaginal lavage fluid 89,90 and is targeted by other reproductive pathogens 91,92 . Although the implications of altered fibrinogen levels in the female reproductive tract are not clear, impaired clotting functions could have severe consequences during labor and postpartum. Fibrinogen is known to increase substantially during pregnancy 93,94 , and decreased plasma fibrinogen levels have been associated with increased severity of postpartum hemorrhage 95 . Further to this, genetic fibrinogen abnormalities are significantly associated with miscarriage, placental abruption, and postpartum hemorrhage [96][97][98][99] .
Within the female genital tract, collagens are found within the vagina, cervix, uterus, and pelvic floor and their composition and content is significantly altered throughout pregnancy and during labor 100 . Vaginal and cervical tissue is primarily composed of fibrillar type I and III collagens, with type I collagen playing a critical role in tissue integrity [101][102][103] . Type IV collagen is typically found within basement membranes and is enriched within the placenta 84 . Both type I and type IV collagens are found within chorioamniotic membranes at the maternal-fetal interface 85 . During pregnancy, cervical collagens (type I, III) undergo a shift toward increased solubility and decreased abundance, contributing to cervical softening 104-106 , while increased collagenase activity is observed during cervical ripening to prepare for dilation and parturition 107,108 . Importantly, cervical remodeling during term and preterm labor occurs via the same mechanisms, with host matrix metalloproteinases (MMPs) coordinating cervical collagen degradation 109 . Premature preterm rupture of the membrane (PPROM) has been associated with infection 110,111 , increased host MMP collagenase activity 112,113 , and decreased collagen content 114 . Furthermore, microbial collagenases can reduce the tensile strength of chorioamniotic membranes ex vivo 115 , and collagenase activity has been detected in clinical isolates from PPROM patients 59,60 . Taken together, these findings imply that host and microbial modulation of collagen within the cervix and chorioamniotic membranes could play critical roles in preterm labor and PPROM. In the present study, our findings indicate that proteases secreted by P. asaccharolytica and P. uenonis are capable of degrading both type I and type IV collagens, uncovering a possible mechanism for how these microbes contribute to the initiation of preterm labor.
Although P. asaccharolytica and P. uenonis are phylogenetically related to the opportunistic periodontal pathogen P. gingivalis and our investigations demonstrated that P. asaccharolytica and P. uenonis possess gingipain-like activities, including degradation of type I and type IV collagen, casein, and fibrinogen, gingipain homologs are absent in P. asaccharolytica and P. uenonis. Our bioinformatics inquiries identified five candidate secreted collagenases in each vaginal Porphyromonas species. Interestingly, withingenome pairs of candidate collagenases had lower amino acid identity than interspecies pairs, suggesting that within-genome pairs are paralogs that resulted from a gene duplication event prior to speciation of P. asaccharolytica and P. uenonis 116 . By incorporating protease inhibitors into our functional assays, we determined that Porphyromonas secreted metalloproteases degrade collagens (type I, IV), casein, and fibrinogen. Although P. asaccharolytica and P. uenonis possess the same candidate collagenases, only P. asaccharolytica appears to secrete both metallo and cysteine proteases in the experimental conditions used in our study. Future investigations will need to address whether pH or redox state may affect the activity of the cysteine Fig. 6 Secreted PepO metallopeptidases identified in P. asaccharolytica and P. uenonis. a Domain structure of candidate host-degrading PepO metalloproteases from P. asaccharolytica and P. uenonis. Created with BioRender.com. b 16S rRNA gene phylogeny of Porphyromonas species identified in the urogenital tract presented as a Maximum Likelihood tree rooted to Prevotella buccalis (same order, different family). Red circles on each leaf indicate the percent amino acid identity of that species' PepO ortholog compared to P. asaccharolytica PepO (Poras_0079 = 100%). The taxon previously identified as uncultivated Porphyromonas species type 1 was found to encompass two cultured, but unsequenced isolates (DNF00129 and C941, >99% 16S rRNA gene identity over 1120 nt). We identified a metagenome-assembled genome (MAG) in IMG/MER that encoded a 958 bp 16S rRNA gene fragment >99.5% identical to those from the cultured isolates DNF00129 and C941. Since this MAG represents the only genome sequence available for this species, we used it in our PepO queries, identifying an ortholog 42% identical to P. asaccharolytica PepO. As this taxon's PepO ortholog was more distantly related to P. asaccharolytica PepO than the Prevotella buccalis PepO ortholog (47% identical), Porphyromonas species type 1 may represent a novel genus or family; we, therefore, labeled this taxon Porphyromonadaceae sp. type 1. Our phylogenetic analysis, along with inquiries through the Genome Taxonomy Database furthermore indicated that vaginal isolates KA00683 and KA00676 should be designated as belonging to the species P. pasteri, with each containing a PepO ortholog 58-59% identical to that from P. asaccharolytica.
proteases. Additionally, due to the general proteolytic activity of the metallo and cysteine proteases, it is plausible that secreted proteases may degrade other proteins in the supernatants, including other proteases. Our results also indicate that P. uenonis may degrade fibrinogen using both secreted and cell surfaceassociated proteases, as P. uenonis cell suspensions degraded the α and β chains of fibrinogen, while cell-free supernatants partially degraded the α chain only. Interestingly, the secreted P. uenonis protease that contributes to partial degradation of the fibrinogen α chain is not impacted by the inhibitors included in this study, implying that the secreted fibrinogenolytic enzyme from P. uenonis is distinct from the secreted collagen and casein degrading enzymes.
Since 1,10-phenanthroline inhibited the secreted proteolytic activity of P. asaccharolytica and P. uenonis, the M13 metalloproteases (distant orthologs of P. gingivalis PepO [PgPepO]) were expected to confer the observed protease activity. Our results confirm that P. asaccharolytica PepO degrades type I collagen and casein, but not type IV collagen, suggesting there is an additional type IV collagen-degrading metalloprotease in Porphyromonas supernatants. Previous work characterizing P. gingivalis PepO revealed sequence conservation with the human endothelin converting enzyme 1 (ECE-1), which proteolytically processes inactive endothelin (big endothelin) into active endothelin. Activated endothelin peptides can induce vasoconstriction and cellular proliferation, alter vascular permeability and activate inflammatory cells 117,118 . PgPepO was confirmed to possess ECE-1 like activity, converting all three subtypes of big endothelin to active endothelin 62 . Numerous studies have also demonstrated that PgPepO plays a role in cellular invasion and intracellular survival of P. gingivalis 63,72 . However additional substrates for this endopeptidase and the functional consequences of bacterial endothelin activation have yet to be explored.
PepO orthologs have also been characterized in select Lactobacillus species: Lactobacillus lactis and Lactobacillus rhamnosus PepO can proteolyze casein, but these enzymes are either confirmed or predicted to localize to the cytoplasm 119,120 . PepO has also been explored in Streptococcus species, including Streptococcus pneumoniae and Streptococcus pyogenes (Group A Streptococci; GAS). In S. pneumoniae, PepO is detected on the bacterial cell surface and in culture supernatants. S. pneumoniae PepO binds to host cells and fibronectin, facilitating bacterial adhesion and invasion 121 . S. pneumoniae PepO also interacts with the host immune system, but there is conflicting evidence on whether this contributes to immune evasion or activation [121][122][123][124] . Surprisingly, none of these studies have shown any proteolytic Fig. 7 The P. asaccharolytica metalloprotease PepO degrades casein and type I collagen. a SDS-PAGE of control in vitro transcription/ translation (myTXTL ® ) reaction (RNA polymerase only) and PepO myTXTL ® reaction. b-d PepO myTXTL ® reactions were incubated with b fluorescein (FITC)-casein or c fluorophore-conjugated type I collagen in the presence of 0.5 mM 1,10-phenanthroline. Casein degradation was measured by detecting the increase in fluorescence over a 24-h time course. Results are presented as mean ± standard error from three independent experiments. Collagen degradation was measured by detecting the increase in fluorescence over an 18-h time course. Results are presented as mean ± standard error from five independent experiments. d PepO myTXTL ® reactions were incubated with fluorophoreconjugated type IV collagen over an 18-h time course and results are presented as mean ± standard deviation from one independent experiment. The mean fluorescence readings of the negative control were subtracted from experimental wells and relative fluorescence units (RFU) was plotted over time, with negative values adjusted to zero. activity of S. pneumoniae PepO. In GAS, PepO has been detected both in the cytoplasm and as a secreted protein 125,126 . Similar to S. pneumoniae, GAS PepO contributes to complement evasion 126 . GAS PepO also participates in regulating quorum sensing via direct degradation of peptide pheromones secreted by GAS 125 . Taken together, this body of work highlights the broad substrate specificity and diverse functionality of bacterial PepO. Our work characterizing P. asaccharolytica PepO demonstrates degradation of host extracellular matrix (type I collagen), adding another substrate to the proteolytic repertoire of PepO. The findings that PepO can degrade regulatory proteins secreted by host cells (P. gingivalis PepO: endothelin) and bacterial cells (GAS PepO: quorum sensing peptides) suggest that PepO enzymes could play a role in the dysregulation of proteolytic cascades. Further investigation is needed to better understand the substrates targeted by PepO proteins in complex mucosal body sites such as the female reproductive tract to reveal how this enzyme contributes to the pathogenesis of phylogenetically diverse bacteria.

Collagenase/caseinase assays
Cell suspensions or cell-free supernatants were tested for collagenase activity with the EnzChek Gelatinase/Collagenase assay kit (Invitrogen, Carlsbad, CA) using fluorophore-labeled DQ TM gelatin conjugate (predominantly type I collagen, Invitrogen) or a fluorophore-labeled DQ TM type IV collagen conjugate from the human placenta (Invitrogen). Cell-free supernatants were harvested from liquid cultures by centrifuging at 10,000 × g for 10 min at room temperature before the supernatant was filter-sterilized (0.

Casein plate assay
Casein plates were prepared by autoclaving three solutions: 30 g/L instant skim milk power (Pacific Dairy), 19 g/L brain heart infusion (BHI, BD Biosciences), and 30 g/L agar (Fisher Scientific). The solutions were combined in equal volume and 10 mL was added to 100 × 15 mm petri dishes (Fisher Scientific) to solidify at room temperature. Bacterial suspensions were prepared in PBS and 5 µL was spotted onto casein agar plates. Zones of clearance were measured for each spot after incubating the plates at 37°C under anaerobic conditions for 3 to 6 days.

Zymography
The total protein content of Porphyromonas cell-free supernatants was determined using a bicinchoninic acid microplate assay (BCA; Pierce, Rockford, IL). Cell-free supernatants were diluted to 8 mg/mL and 5 µL of the sample was combined with 5 µL of Novex TM Tris-Glycine SDS Sample Buffer (Invitrogen) to load 40 µg per well. Samples were separated on Novex TM 10% Zymogram Plus (Gelatin; Invitrogen) protein gels at a constant voltage of 125 V in Novex TM 1X Tris-Glycine Running Buffer (Invitrogen). After separation, gels were incubated in Novex TM 1X Renaturing Buffer for 30 min at room temperature with gentle agitation, followed by two consecutive incubations in Novex TM 1X Developing Buffer: room temperature for 30 min and 37°C for 16 h. Gels were then stained in Coomassie brilliant blue R-250 solution (Fisher Scientific) and de-stained in 5% methanol/7.5% acetic acid in distilled water. The unprocessed scan of the zymogram gel is available in the Supplementary Information ( Supplementary Fig. 8).

Clotting assays
Bacterial strains were harvested from a solid medium and suspended in 13 mM sodium citrate. Suspensions were centrifuged at 10,000 × g for 7 min at room temperature and resuspended in 13 mM sodium citrate. Duplicate cell suspensions, or cell-free controls, were incubated with 50 µL of sterile-filtered bovine plasma (Quad Five, Ryegate, MT) for 30 min at 37°C under anaerobic conditions. After incubation, samples were blinded and centrifuged at 10,000 × g for 7 min at room temperature, and 50 µL of HEMOCLOT thrombin time reagent (Hyphen Biomed, Neuville-sur-Oise, France) was added to each reaction. The clotting time for each blinded sample was estimated using a stereo microscope to directly visualize clot formation, indicated by the presence of white precipitate, tendril formation, or increased viscosity of the samples. The clotting time for samples that did not form visible clots was recorded as 1800 s. To further evaluate final clot size, samples were transferred to a clear 96-well plate and OD (405 nm) was measured for the entire well using the well area scan feature of the Synergy H1 microplate reader (BioTek).

Fibrinogen degradation assay
Porphyromonas cell suspensions (10 7 CFU/reaction), sBHI media controls, or Porphyromonas cell-free supernatants (240 µg of total protein) were incubated with 120 µg of human fibrinogen (Sigma-Aldrich, St. Louis, MO). Reaction mixtures were incubated at 37°C under anaerobic conditions or in an atmosphere of 5% CO 2 for cell suspension or cellfree supernatants, respectively. Samples from each time point (cell suspensions: 0, 2, 18, 24 h; supernatants: 0, 2, 24, 48 h) were collected, mixed 1:1 with Novex TM 2X sample buffer (Invitrogen) with dithiothreitol (DTT, Fisher Scientific), heated at 95°C for 10 min, and separated on Novex TM 10% Tris-Glycine polyacrylamide pre-cast gels (Invitrogen) at a constant voltage of 180 V. Gels were stained in 0.25% Coomassie brilliant blue R-250 solution (Fisher Scientific) and de-stained in 5% methanol/7.5% acetic acid in distilled water. Fibrinogen degradation was evaluated qualitatively by visualization of fibrinogen α chain (63.5 kDa), β chain (56 kDa), and γ chain (47 kDa) between experimental and control samples over the time-course. Gels are derived from the same experiments and were processed in parallel. Unprocessed scans of fibrinogen degradation protein gels are available in the Supplementary Information (Supplementary Figs. 9, 10).

Bioinformatic analyses
Gingipain ortholog queries. Gingipain amino acid sequences RgpA (PGN_1970; P28784), RgpB (PGN_1466; P95493), and Kgp (PGN_1728; B2RLK2) were obtained from P. gingivalis ATCC 33277 through UniProtKB. Each sequence was queried against all available P. asaccharolytica (DSM 20707; PR426713P-I) and P. uenonis (DSM 23387; 60-3) genomes using the "Selected Genomes" protein Basic Local Alignment Search Tool (BLAST; default settings) in the IMG/MER database to identify potential orthologs ( Supplementary Fig. 1A). Each gingipain amino acid (AA) sequence was also submitted to the Pfam database and all Pfam IDs and names were recorded. Gingipain Pfam IDs were searched against all available P. asaccharolytica (DSM 20707; PR426713P-I) and P. uenonis (DSM 23387 [IMG Genome ID 2585427891 and 2528311143]; 60-3) strains using the advanced gene search function in the IMG/MER database ( Supplementary  Fig. 1A). All sequences returned by these IMG/MER BLAST and Pfam ID searches were used to query P. gingivalis ATCC 33277 in a reciprocal BLAST search with the National Center for Biotechnology Information (NCBI) protein BLAST tool (Supplementary Fig. 1A). Select hits were aligned with the P. gingivalis gingipains across their full length using the EMBL-EMI Clustal Omega alignment tool 127 .
Candidate collagenase queries. Peptidases from P. asaccharolytica CCUG 7834 and P. uenonis 60-3 were identified by searching for entries with MEROPS peptidase annotations in UniProt 64,128 . For the microbial peptidases from P. asaccharolytica and P. uenonis, the following enzyme information was exported from UniProtKB for each entry: gene ontology (biological process and molecular function), MEROPs, Pfam, PANTHER, PROSITE, SMART, SUPFAM. Next, a microbial collagenase enzyme number (EC3.4.24.3) was identified in BRENDA 65 and searched against the UniProtKB database 128 , generating a list of 3417 entries corresponding to predicted and confirmed microbial collagenases. These microbial collagenase identifiers were crossreferenced against the exported peptidase information from P. asaccharolytica and P. uenonis to generate a short-list of 14-18 candidate collagenases in P. asaccharolytica and P. uenonis ( Supplementary Fig. 1B). Short-list candidates were explored in UniProt and InterPro Scan 128,129 to eliminate any proteins involved in cell wall synthesis or export machinery. Sequences were evaluated for the presence of secretion signals using SignalP v.5.0 130 or InterPro 129 . Presence of the IDs: TIGR0483, IPR026444, or PF18962 from TIGR Fam, InterPro, and Pfam, respectively, within protein C-termini was indicative of secretion via the type IX secretion system. Since the MEROPs peptidase database only contained information for P. uenonis 60-3, IMG BLAST searches were used to identify the corresponding candidate collagenase in the experimental strain used in this study, P. uenonis CCUG 48615. Sequence identity and InterPro scans were evaluated for the top hit from each BLAST search in P. uenonis CCUG 48615 (Supplementary Table 6).
Phylogenetic analysis. 16S rRNA gene sequences were collected using the Integrated Microbial Genomes & Microbiomes (IMG/MER) 131 or National Center for Biotechnology Information (NCBI) databases for each Porphyromonas species reported in the human urogenital tract by Acuna-Amador in their comprehensive review 14 . Uncultured and recently cultured Porphyromonas species were also included 15,132 . Nomenclature choices were informed by the average nucleotide identity (ANI) tool within IMG/ MER and the Genome Taxonomy Database (GTDB, Release 06-RS202, April 27, 2021) 133 . Whenever possible, type strain and near full-length sequences were selected. Multiple sequence alignment was performed with SINA aligner (v.1.2.11) 134 using Silva's Alignment, Classification and Tree Service (ACT) 135 . The phylogenetic tree was also computed through ACT using RAxML v.8.2.9 136 with the Gamma model for likelihoods. The tree was edited using the interactive Tree of Life (iTOL 137 ). The AA identity of PepO orthologs in each Porphyromonas species was queried through BLASTP searches in IMG/MER and NCBI using the P. asaccharolytica AA sequence as the query. With the exception of P. uenonis 60-3, the AA identity reported corresponds with a hit showing >98% query coverage.

Expression construct cloning and in vitro transcription/ translation
The Poras_0079 (PepO; IMG Gene ID 2504823953) DNA fragment encoding amino acid residues C21 to W682 was PCR amplified from a P. asaccharolytica DSM 20707 genomic DNA extraction (Qiagen DNeasy Blood & Tissue Kit, Germantown, MD) using the forward (5′-ATATCCA TGGCTTGTAACAAGAAGCAGGAGAATC-3′) and reverse primers (5′-ATAT CCCGGGCCAGACCACGACACGCTC-3′). The amplicon was cloned into the pTXTL-T7p14-aH plasmid (replacing alpha-hemolysin, Daicel Arbor Biosciences, Ann-Arbor, MI) using NcoI-HF and SmaI (New England Biolabs, Ipswitch, MA). The new plasmid construct (pSLP15) was transformed into Escherichia coli DH5α chemically competent cells and prepped using the QIAprep spin miniprep kit (Qiagen). Plasmid concentrations were determined using the Qubit dsDNA broad-range assay kit with a Qubit 3 fluorometer (Invitrogen). In vitro myTXTL ® reactions were prepared by combining 5 nM of pSLP15, 1 nM of pTXTL-P70a-T7rnap (expressing the T7 RNA polymerase, Daicel Arbor Biosciences), and 9 µl of myTXTL ® Sigma 70 Master Mix (Daciel Arbor Biosciences) to a final volume of 12 µL with HyClone HyPure water H 2 O (Cytiva); negative controls included 1 nM pTXTL-P70a-T7rnap and myTXTL ® Sigma 70 Master Mix only. For each reaction, 10 µL was transferred to a PCR-clean polypropylene V bottom 96well plate and covered with a silicone seal (Eppendorf, Mississauga, ON). Reactions were incubated at 29°C for 16 h and final reactions stored at −20°C. To assess proteolytic activity, control (RNAP only) and PepO reactions were incorporated into fluorescent collagenase and caseinase assays. For caseinase assays, myTXTL ® reactions were diluted 60-fold in TBS and 10 µL was added to each well. For type I and type IV collagenase assays, 10 µL of undiluted myTXTL ® reactions were added to each well. Caseinase and type I collagenase assays were also conducted in the presence of 0.5 mM 1,10-phenanthroline. Mean fluorescence readings of the negative control (substrate with RNAP myTXTL ® reaction) were subtracted from the experimental wells and relative fluorescence units (RFU) were plotted over time, with negative values adjusted to zero. The unprocessed scan of the myTXTL ® reaction protein gel is available in the Supplementary Information (Supplementary Fig. 11).

Bioinformatics and statistical analyses
Details of our bioinformatics analyses are provided in the Methods. Statistics were performed in GraphPad Prism (GraphPad, San Diego, CA) or Stata 15 (StataCorp, College Station, TX) and graphs were prepared in GraphPad. GraphPad was used to assess data variance and normality (Shapiro-Wilk test) and statistical significance was evaluated as described in figure legends. All schematic illustrations were created with BioRender. com.

Reporting Summary
Further information on research design is available in the Nature Research Reporting Summary linked to this article.

DATA AVAILABILITY
The authors confirm that the data supporting the findings of this study are available within the paper and Supplementary files. Additional data are available from the corresponding author upon reasonable request. The unprocessed protein gel scans are available in the Supplemental Information (Supplemental Figs. [8][9][10][11]. P. gingivalis ATCC 33277 gingipain protein sequences were accessed via UniProtKB using the following accession numbers: RgpA (PGN_1970; P28784), RgpB (PGN_1466; P95493), and Kgp (PGN_1728; B2RLK2). Gingipain BLASTP and Pfam searches were conducted against P. asaccharolytica and P. uenonis strains in IMG using the following IMG genome IDs: 649989985, 2504756018, 643886142, 2585427891, 2528311143. Candidate collagenases in P. asaccharolytica DSM 20707 and P. uenonis 60-3 were accessed via UniProtKB using the taxon identifiers 879243 and 596327, respectively. The protein sequence for the P. asaccharolytica PepO protease was retrieved through UniProtKB with the accession ID F4KL59. Cultured and uncultured vaginal Porphyromonas and Prevotella species' 16S rRNA gene sequences were drawn from public databases as follows: Set I -(NCBI genome accession/locus tag): P. bennonis DSM 23058 (PRJNA169760/B088DRAFT_10000), P. asaccharolytica DSM 20707 (NC_015501/Poras_R0016), P.