Zebrafish as a new model to study effects of periodontal pathogens on cardiovascular diseases

Porphyromonas gingivalis (Pg) is a keystone pathogen in the aetiology of chronic periodontitis. However, recent evidence suggests that the bacterium is also able to enter the bloodstream, interact with host cells and tissues, and ultimately contribute to the pathogenesis of cardiovascular disease (CVD). Here we established a novel zebrafish larvae systemic infection model showing that Pg rapidly adheres to and penetrates the zebrafish vascular endothelium causing a dose- and time-dependent mortality with associated development of pericardial oedemas and cardiac damage. The in vivo model was then used to probe the role of Pg expressed gingipain proteases using systemically delivered gingipain-deficient Pg mutants, which displayed significantly reduced zebrafish morbidity and mortality compared to wild-type bacteria. In addition, we used the zebrafish model to show efficacy of a gingipain inhibitor (KYT) on Pg-mediated systemic disease, suggesting its potential use therapeutically. Our data reveal the first real-time in vivo evidence of intracellular Pg within the endothelium of an infection model and establishes that gingipains are crucially linked to systemic disease and potentially contribute to CVD.

a model system to study human infectious diseases and virulence of bacterial pathogens such as Enterococcus faecalis 17 , Salmonella typhimurium 18 , Psudomonas aeruginosa 19 and Staphylococcus aureus 20 . One of the main advantages for using zebrafish to study infectious diseases is the possibility of non-invasive real-time imaging in vivo at the cellular level. Since larval zebrafish are transparent the fate of fluorescently-labelled bacteria after injection into zebrafish larvae can be monitored and their co-localization with fluorescently-tagged host cells examined, making these models an attractive alternative to other vertebrate systems where disease development cannot be visualized 17,21 .
In this study we used a zebrafish larvae infection model to study the mechanisms of Pg systemic pathogenicity. Using gingipain knock-out mutants and protease inhibitors we identify, for the first time, the crucial role that these proteases play during systemic infection, and show the ability of these bacteria to cross the vasculature and disseminate into surrounding tissues in vivo. Data presented here demonstrate that the zebrafish model is suitable to study both the host response to Pg infection as well as bacterial virulence and provides a powerful optically and genetically accessible tool to study the systemic pathogenicity of oral bacteria.

P. gingivalis is pathogenic for zebrafish larvae in a systemic infection model. To determine
whether Pg W83 strain is virulent in a systemic infection model, zebrafish larvae were injected at 30 hours post-fertilization (hpf) and survival monitored up to 72 hours post-infection (hpi). We observed a dose-and temperature-dependent increase in mortality that was significantly (p < 0.05) different to PBS-injected controls at 48 and 72 hpi (Fig. 1a,b). A LD 50 was established when 5 × 10 4 CFU Pg W83 were injected into zebrafish for 48 hpi at 30 °C and so these conditions were used throughout the remainder of the study. We noted that zebrafish, remaining viable after infection with W83, presented with visible signs of systemic disease that increased with severity overtime compared to controls. In particular, zebrafish displayed severe pericardial and yolk sac oedemas as early as 24 hpi along with eye abnormalities and lordtic spine curving (Fig. 1c,d). In addition, W83-infected embryos developed marked cardiac injury as early as 24 hpi. The atrium and ventricle were separated forming thin, elongated chambers within an oedemic appendage and the cardiac tissue displayed extensive damage upon Figure 1. Systemic infection of zebrafish larvae with Pg W83 leads to fish mortality and morbidity in a dose-and temperature-dependent manner. Kaplan-Meier survival plots of zebrafish larvae infected with increasing doses of Pg W83 strain at 30 °C (a) or temperature-dependent survival of zebrafish larvae infected with 4 × 10 4 CFU W83 strain (b). At least three individual experiments were performed (n ≥ 20 larvae per group in each experiment). Comparisons between survival curves were made using the log rank test. Levels of dead, oedematous or healthy Pg W83-infected larvae for up to 72 h upon injection with 5 × 10 4 CFU Pg W83 (c). Development of the oedema disease phenotype in zebrafish larvae infected with Pg W83 or PBSinjected controls over 72 h. Black arrows indicate pericardiac and yolk sac oedemas, red arrows indicate eye abnormalities and blue arrowhead indicates spine curving in W83-infected larvae. At least three individual experiments were performed (n ≥ 20 larvae per group in each experiment). Scale bars = 500 in (d).
histological analysis (Fig. 2a). Further tissue damage was apparent within the yolk sac and cranial regions that displayed evidence of tissue necrosis (Fig. 2b). P. gingivalis W83 disseminates rapidly in zebrafish larvae, infecting endothelial cells and crossing the vasculature to invade surrounding tissues. We next performed real-time, confocal in vivo imaging of kdrl:memRFP transgenic zebrafish embryos (red-labelled endothelial cells) infected systemically with fluorescently labelled Pg W83 to determine its cellular targets and if the bacterium could traverse the endothelium. Pg was disseminated rapidly in zebrafish and was found associated and co-localised with the inter-segmental endothelial cells at 1 hpi (Fig. 3) providing evidence of adhesion to the endothelium and intracellular localization in vivo. In addition, Pg was observed outside the red-labelled endothelial cells indicating that bacteria had bound to and crossed the vasculature before entering into the surrounding tissue (Fig. 3C). As well as the inter-segmental vessels we observed Pg localization in other highly vascularized tissues including the eye, tail and yolk sac at 1 and 24 hpi ( Supplementary Fig. 1). We also observed strong Pg adhesion to cardiac tissues where bacteria were not dislodged from tissue despite the intracardiac blood flow ( Fig. 3D and Supplementary Movie 1). To visualize the innate immune response to the systemic infection we dual-labelled Pg with the pH sensitive dye, pHrodo (non-fluorescent until it is acidified within phagosomes where upon it fluoresces red) and fluorescein to visualize phagocytosed bacteria. After 1 hpi phagocytes had bound and ingested Pg and many bacteria were localized within phagosomes (Fig. 3E). Phagocytosis was observed in all tissues examined at this time point and was increased at 24 hpi which was associated with a concurrent decrease in bacterial load ( Supplementary Fig. 1).
W83 is the most virulent strain of P. gingivalis in the zebrafish larvae infection model. Previous studies have shown that strains of Pg isolated from patients with periodontitis exhibit different levels of pathogenicity in both in vitro and in vivo models 22,23 . We selected three extensively studied strains of Pg (W83, ATCC 33277 and FDC 381) and compared their virulence in the zebrafish larvae systemic infection model. Kaplan-Meier plots show that strain W83 was the most virulent in terms of mortality causing significantly more death (p < 0.01) than ATCC33277 and FDC 381 by 72 hpi (Fig. 4a). In contrast, the virulence of ATCC33277 and 381 were not significantly different from each other, causing less than 20% mortality of infected larvae by 72 hpi (Fig. 4a). This inter-strain variability was also observed for the development of disease symptoms in infected embryos where infection with W83 rapidly led to oedemas in the vast majority of infected larvae at 24 hpi with an associated mortality rate of more than 50% of larvae by 72 hpi (Fig. 4b-d). Similarly, 80 ± 14% of ATCC33277 infected larvae had developed oedemas at 24 hpi but the effect appeared to be less progressive than for W83 as only 16 ± 9% of larvae were killed by 72 hpi (Fig. 4b-d). Pg strain FDC 381 was the least virulent in terms of inducing both systemic disease and mortality. Indeed, some of the larvae displaying oedema were able to recover from this condition by 72 hpi (Fig 4b-d).

Mortality of P. gingivalis-infected zebrafish larvae depends on the viability of the pathogen and its active cell surface components.
To determine if microbial viability was essential for virulence, zebrafish larvae were injected with viable, heat-killed (HK) or formalin-fixed (FF) Pg W83 and mortality assessed. Kaplan-Meier analysis showed that HK or FF treatment of Pg markedly attenuated bacterial virulence causing a significant reduction (p < 0.0001) in zebrafish mortality after 72 hpi when compared to zebrafish injected with viable Pg (Fig. 5a). Although there was no difference in the percentage of zebrafish larvae killed with the different treatments after 24 h, more (88 ± 8%) zebrafish displayed oedemas compared to 42 ± 9% for FF and only 13 ± 5% for HK Pg (Fig. 5b). In zebrafish larvae injected with viable bacteria, almost all fish developed progressive disease symptoms that eventually led to the death of 56 ± 5% of larvae by 72 hpi (Fig. 5c,d). In contrast, the majority of the HK-infected embryos remained disease-free over a 72 h period (p ≤ 0.05, Fig. 5b-d) with only a minority experiencing oedema or mortality. A similar effect was observed for FF Pg-infected larvae although this group displayed consistently more oedema than the HK Pg-treated group over a 72 h period ( Fig. 5b-d). As well as ensuring loss of viability heat treatment and fixation may also affect the activity of virulence factors, in particular those associated with enzyme activity. We found that treatment with elevated temperatures of 60 °C almost abolished the activity of both lysine (Kgp) and arginine (RgpA and RgpB) specific gingipains, whereas fixation of bacteria with 10% formalin reduced the activity of these gingipains by 60% (lysine specific) and 90% (arginine specific) respectively (Supplementary Fig. 2) indicating that these proteins may be important in mediating virulence.

Gingipains are crucial virulence factors of P. gingivalis in zebrafish systemic disease. To exam-
ine whether gingipains play a role systemic disease, zebrafish larvae were infected with wild-type (WT) parental strain W83, single (Δ Kgp, Δ RgpA or Δ RgpB), double (Δ K/Ra) or triple (Δ K/Ra-b) gingipain-deficient mutants. Removal of any of the gingipains caused an increase in the survival rate of infected larvae compared to WT W83 and this effect was additive with respect to the double or triple mutants whose mortality rates were similar to PBS controls (p ≤ 0.01 for Δ K/Ra and p ≤ 0.0001 for Δ K/Ra-b; Fig. 6a). With the exception of the Δ K/Ra-b triple mutant, zebrafish larvae injected with any of the mutated strains of Pg caused significant oedema in the first 24 hpi although there was little mortality (Fig. 6b). By 48 and 72 hpi (Fig. 6c) the levels of oedema remained the same; however, larvae mortality was increased to 60% for WT W83 Pg and 40% for lysine-deficient mutant (Δ Kgp). Levels of larvae mortality were slightly, but not significantly, reduced upon infection with the arginine-deficient mutants Δ RgpA and Δ RgpB compared to Δ Kgp. The Δ K/Ra-deficient mutant produced mortality levels of ~10% after 72 hpi whilst gingipain triple mutants (Δ K/Ra-b) were virtually avirulent, producing a larvae phenotype similar to PBS controls (Fig. 6c,d).
To confirm these findings we used a specific inhibitor of arginine and lysine gingipains (KYT) 24 on W83 Pg to directly inhibit the proteolytic activity before systemic injection. The amount of inhibitor required to completely inhibit gingipain activity was determined in vitro before use in vivo (data not shown). We found that infection with W83 + KYT significantly (p < 0.001) reduced larvae mortality by 48 and 72 hpi compared to those injected without inhibitor (Fig. 7a-c). However, the presence of gingipain inhibitor did not influence the number of larvae displaying signs of oedema compared to WT-infected Pg over a 72 h period (Fig. 7a-c). Comparisons between survival curves were made using the log rank test. Differences between groups displaying oedema were measured using One-way ANOVA after normality was assured.

Discussion
There is increasing evidence to suggest that Pg influences the progression of chronic systemic conditions such as cardiovascular disease 25,26 . In support of this Pg has been found in atherosclerotic plaques 27,28 and has been found to influence myocardial infarction in a murine model 29 . In this study we used zebrafish larvae, an in vivo model system that has been successfully used to investigate host-pathogen interactions 30 to examine the effects of Pg on the circulatory system. We used a well-established systemic infection route 20,31 along with genetically engineered zebrafish larvae that allowed us to examine, for the first time, the interaction of Pg with the vasculature at the cellular level in vivo.
We found that systemic infection with W83, a highly virulent strain of Pg compared to other strains tested, lead to dose-dependent disease severity and mortality in larvae 32,33 . A similar dose-dependent disease effect has been previously observed using a Drosophila melanogaster infection model albeit via a non-systemic route 34 . In addition, we observed extensive damage to the heart and surrounding tissue associated with pericardial oedemas. Microbial-mediated pericarditis in humans often leads to pericardial oedema and DNA from Pg along with that from other oral bacteria have been isolated from pericardial fluid and in endocarditis lesions 35 suggesting that, although exaggerated, the disease process in zebrafish may reflect those in humans.
Using real-time in vivo confocal imaging we observed Pg both closely associated to and within endothelial cells of the vasculature, confirming in vitro data that Pg can adhere and be internalized by endothelial cells 36 . In addition, we found Pg deep within tissues showing that once internalized Pg are able to survive within endothelial cells and be released into tissues in vivo 37,38 . Li et al. showed in vitro that Pg can be transmitted intracellularly from cell to cell as a mechanism of spreading within the endothelium, and that these bacteria were able to replicate once within newly infected cells and then be released for further uptake 39 . Although we could recover viable Pg from infected zebrafish, we were are unable to determine if the bacterium replicated in vivo and so further studies are required to resolve this issue.
Although innate immune cells rapidly internalized and phagocytosed bacteria we also observed marked numbers of Pg in the circulatory system that was associated with significant damage and oedema in several tissues, most noticeably the pericardium and yolk sac. Moreover, this pathology was dependent upon the presence of gingipain proteases as lysine and arginine gingipain-deficient strains were both significantly less virulent, suggesting both types of gingipains can mediate tissue damage. Indeed, gingipain-null (Δ K/Ra-b)-infected zebrafish displayed no disease symptoms. Pg W83 (a). Percentage of dead, oedematous or healthy Pg W83-infected larvae after 24, 48 and 72 hpi respectively (b-d). Larvae were injected with viable, 60 °C HK or 10% FF Pg W83 at 5 × 10 4 CFU. At least three individual experiments were performed (n ≥ 20 larvae per group in each experiment). ***p ≤ 0.001. Comparisons between survival curves were made using the log rank test. Differences between groups displaying oedema were measured using Oneway ANOVA after normality was assured.

Figure 5. Heat and formalin treatment of Pg W83 decreases its pathogenicity in vivo. Kaplan-Meier survival plots of zebrafish larvae infected with viable, heat-killed (HK) or formalin-fixed (FF)
Scientific RepoRts | 6:36023 | DOI: 10.1038/srep36023 The importance of gingipains in causing periodontal as well as vascular damage in murine models has been previously highlighted 40,41 and more recently gingipains were found to mediate myocardium damage in a murine model of Pg-aggravated myocardial infarction 29 . Several elegant studies have shown that either the lysine or arginine-specific gingipains mediate tissue damage by cleaving and therefore inactivating many important host extracellular matrix molecules including laminin, fibronectin and collagen, leading to loss of cell-matrix contacts that are crucial for tissue integrity and cell survival 42 . Moreover, this tissue destruction is exacerbated by Pg-mediated activation of matrix metalloproteinases, in particular MMP3 43 and MMP9 44 . As well as disrupting cell-matrix interactions, there is in vitro evidence that gingipains also disturb endothelial cell-to-cell contacts. Sheets et al. found that lysine and arginine gingipains cleaved N-and VE-cadherin as well as integrin β 1 expressed by monolayers of human microvascular endothelial cells leading to cell detachment along with caspase-mediated and caspase-independent apoptosis 45,46 . Given this data it is plausible that the increased vascular permeability leading to oedema we observed in systemically infected zebrafish may be due to gingipain-mediated damage of adhesion molecules leading to loss of endothelial adherin, gap and tight junction integrity. This may be further aggravated by reduced endothelial contacts with underlying connective tissue or support cells in more mature vessels. The model established here will pave the way for further studies to measure the effect of Pg and its virulence determinants on vascular leakage using a range of zebrafish lines open to the field.
Recently, small peptide analogs (KYT-1 and KYT-36) that are selective inhibitors of Kgp and Rgp, respectively 47,48 have been developed. These inhibitors have been shown to suppress the pathogenic potential of Pg in murine models 24 . Similarly, we found that zebrafish systemically infected with WT Pg W83 along with KYT1 and KYT36 displayed significantly reduced mortality and disease symptoms, suggesting not only that these inhibitory molecules show therapeutic promise but that the zebrafish infection model may be used to screen other therapeutic agents and strategies aimed at Pg or other bacterial-mediated systemic diseases. In all experiments WT Pg or its gingipain mutants were injected into 30 hpf zebrafish larvae at 5 × 10 4 CFU. Black arrow indicates oedemas in WT Pg W83-infected larvae. Scale bar = 500 μ m. Comparisons between survival curves were made using the log rank test. Differences between groups displaying oedema were measured using One-way ANOVA after normality was assured.
Scientific RepoRts | 6:36023 | DOI: 10.1038/srep36023 In summary, this study reveals that zebrafish larvae represent a powerful novel vertebrate model system to examine the role of Pg and potentially other periodontal bacteria in cardiovascular disease or endocarditis. Additionally, we demonstrated for the first time that Pg gingipains are crucially important in mediating systemic disease. We believe that this in vivo model system will provide an amenable tool to further examine the mechanism of Pg-mediated systemic disease at the cellular and molecular level.

Methods
Bacteria, zebrafish strains and culture conditions. All strains of Pg and Danio rerio used in this study are listed in Table 1. Wild-type Pg W83 strain and its isogenic mutants were maintained on Schaedler (BTL Spolka, Poland) anaerobe blood agar (ABA) plates supplemented with 0.5 mg/L menadione sodium bisulfate, 0.25 g/L L-cysteine-HCl, 5 mg/L hemin, 5% (v/v) defibrinated horse blood and antibiotic (1 μ g/ml tetracycline or 5 μ g/ml erythromycin) when appropriate. For liquid cultures, strains were grown in hemin, menadione and L-cysteine supplemented Schaedler broth. ATCC33277 and FDC 381 strains were maintained on fastidious anaerobe agar (Lab M, Bury, UK) supplemented with 5% (v/v) oxylated horse blood (Oxoid, Basingstoke, UK). For growth in liquid cultures these strains were grown in brain-heart infusion broth (BHI; Oxoid) supplemented with 5 g/L yeast extract (Oxoid), 5 mg/L hemin and 0.5 mg/L menadione. Agar plates or liquid cultures were incubated at 37 °C in an anaerobic chamber (80% N 2 , 10% CO 2 and 10% H 2 ). For experiments, strains were grown in broth overnight at 37 °C in anaerobic conditions after which fresh cultures were set to an optical density (OD 600 ) equal to 0.1 and cultured until late log phase. Bacteria were harvested by centrifugation at 6,000 × g for 10 min, washed with PBS and resuspended at 5 × 10 9 CFU/ml. For formalin fixation, Pg was resuspended in 10% PBS-buffered formalin and incubated overnight at 4 °C. For heat killing, bacteria were resuspended in PBS and incubated in a water bath at 60 °C for 30 min. The loss of bacterial viability was confirmed by culture on ABA plates. For inhibition of gingipain activity, Pg (OD 600 = 1.0) were pretreated with KYT-1 or KYT-36 inhibitors 24 at a final concentration of 2 μ M and bacteria were incubated for 30 min at 37 °C in anaerobic conditions; after which, bacteria were washed once with PBS, resuspended at 5 × 10 9 CFU/ml and injected into zebrafish larvae. Zebrafish maintenance and experimental work was performed in accordance with UK Home Office regulations and UK Animals (Scientific Procedures) Act 1986. Ethical approval was given by the University of Sheffield Local Ethical Review Panel. London wild-type (LWT) inbred zebrafish larvae were obtained from The Bateson Centre, University of Sheffield. The transgenic kdrl:memRFP strain was used to enable visualization of the endothelium as previously described. All larvae were maintained in E3 medium at 28.5 or 30 °C according to standard protocols and monitored for up to 4 days post-fertilization (dfp).

Microinjection of P. gingivalis onto zebrafish embryos.
Pg were injected systemically by direct inoculation into the common cardinal vein (Duct of Cuvier) of dechorionated larvae at 30 hpf as previously described with PBS used as control 20 . Briefly, tricaine-anesthetized larvae were embedded and positioned in a solution of 3% (w/v) methylcellulose (Sigma Aldrich) in E3 medium on glass slides. Larvae were injected individually using micro-capillary needles loaded with known concentrations of Pg and were monitored for disease development and survival for up to 72 hpi unless otherwise stated. Zebrafish viability was assessed by examining the presence of a heart beat and blood flow within the circulation. Live imaging of anesthetized zebrafish larvae was performed using spinning disc confocal microscopy (Perkin Elmer). At least 20 zebrafish were injected per treatment group in all experiments.
Evaluation of gingipain activity. To evaluate gingipain proteolytic activity Pg cultures were adjusted to OD 600 1.0. Ten microliters (for Rgp) and 20 μ L (for Kgp) of bacterial suspension were pre-activated in TNCT activity buffer (50 mM Tris-HCl pH 7.5, 5 mM CaCl 2 , 150 mM NaCl, 0.05% Tween-20 supplemented with 10 mM L-cysteine-HCl neutralized with 10 mM NaOH) for 5 minutes at room temperature in a total volume of 200 μ L in a 96-well plate. Substrate (0.4 μ L, final concentration 200 mM) was added and the rate of its hydrolysis was recorded at 405 nm using a spectrophotometer as described previously 49 . N-(p-Tosyl)-Gly-Pro-Lys-p-nitroanilide and benzoyl-arginine-p-nitroanilide (BApNA) (both from Sigma Aldrich) were used as substrates specific for Rgps and Kgp, respectively.
Histological processing of zebrafish. Infected and control zebrafish were fixed in 10% PBS-buffered formalin for at least 24 h and then paraffin wax-embedded. Tissue sections (5 μ m) were cut, dewaxed in xylene, rehydrated through a series of alcohols to water and stained with haematoxylin and eosin (H&E). Sections of zebrafish were then dehydrated, mounted in DPX and images taken using an Olympus BX51 microscope and a IIIu camera with associated Cell^D software (Olympus, GmbH, Münster, Germany).

Labeling of P. gingivalis and fluorescence microscopy.
To visualize bacteria and to monitor phagocytosis in vivo, Pg were fluorescently labeled with fluorescein-5-EX and/or pHrodo red succinimidyl ester (Life Technologies, Paisley, UK) respectively. Prior to injection, a 200 μ L bacterial suspension of Pg (in PBS; pH 9) was simultaneously mixed with 0.25 μ L fluorescein-5-EX (10 mg/ml) and/or 0.25 μ L of pHrodo red (1.7 mg/ml) for 30 min at 37 °C in the dark with shaking (100 rpm). Bacteria were washed twice with PBS (pH 8) to remove excess dye, resuspended in PBS (pH 7.4) to an OD 600 5.0 (~5 × 10 9 CFU/ml) and 2 nL injected into each zebrafish. At 2 hpi, six infected larvae were anesthetized, mounted in 1% low-melting-point agarose in E3 medium on coverslips. Fluorescent bacteria were visualized using either a spinning disc laser scanning confocal microscope (Perkin Elmer) and images acquired and processed using Volocity software (Improvision, Perkin Elmer) or a Zeiss Z1 light-sheet microscope with Zen Black 2014 software.
Data and statistical analysis. All data presented is from at least 3 independent experiments with results expressed as the mean ± SD. Survival data were evaluated using the Kaplan-Meier method and comparisons between individual curves were made using the log rank test. Differences between groups displaying oedema were measured using One-way ANOVA after normality was assured. All analysis was performed using Graphpad Prism v6.0 (GraphPad, La Jolla, CA) and statistical significance was assumed if p < 0.05.

Strains
Genotype/Description Reference