Porphyromonas gingivalis induces penetration of lipopolysaccharide and peptidoglycan through the gingival epithelium via degradation of coxsackievirus and adenovirus receptor

Porphyromonas gingivalis is a major pathogen of human periodontitis and dysregulates innate immunity at the gingival epithelial surface. We previously reported that the bacterium specifically degrades junctional adhesion molecule 1 (JAM1), causing gingival epithelial barrier breakdown. However, the functions of other JAM family protein(s) in epithelial barrier dysregulation caused by P. gingivalis are not fully understood. The present results show that gingipains, Arg‐specific or Lys‐specific cysteine proteases produced by P. gingivalis, specifically degrade coxsackievirus and adenovirus receptor (CXADR), a JAM family protein, at R145 and K235 in gingival epithelial cells. In contrast, a gingipain‐deficient P. gingivalis strain was found to be impaired in regard to degradation of CXADR. Furthermore, knockdown of CXADR in artificial gingival epithelium increased permeability to dextran 40 kDa, lipopolysaccharide and peptidoglycan, whereas overexpression of CXADR in a gingival epithelial tissue model prevented penetration by those agents following P. gingivalis infection. Together, these results suggest that P. gingivalis gingipains breach the stratified squamous epithelium barrier by degrading CXADR as well as JAM1, which allows for efficient transfer of bacterial virulence factors into subepithelial tissues.


| INTRODUCTION
Periodontal diseases are chronic infectious diseases caused by complex actions of periodontal bacteria in oral biofilm and one of the most common infectious diseases affecting humans (Eke, Dye, Wei, Thornton-Evans, & Genco, 2012). In periodontal tissues, subgingival epithelium responds to microbial infection, which helps to harmonise innate immunity (Dixon, Bainbridge, & Darveau, 2004). Lipopolysaccharide (LPS), a large molecule consisting of lipid and polysaccharide located in the outer membrane of Gram-negative bacteria, and Hiroki Takeuchi and Shunsuke Yamaga contributed equally to this work. peptidoglycan (PGN), a carbohydrate backbone found in cell walls of most bacteria, are classified as pathogen-associated molecular patterns (PAMPs) and recognised by the innate immune system (Kawai & Akira, 2001).
P. gingivalis, a periodontal pathogen, can impair the host immune defense system and promote inflammation (Lamont, Koo, & Hajishengallis, 2018). This bacterium secretes Arg-specific (RgpA and RgpB) and Lys-specific (Kgp) cysteine proteases, which demonstrate abundant proteolytic activities (Nakayama, Kadowaki, Okamoto, & Yamamoto, 1995;Potempa, Pike, & Travis, 1995). Hence, host proteins targeted and degraded by gingipains are potentially crucial factors that influence the pathogenesis of periodontitis. We previously showed that JAM1 as a gingival epithelial barrier protein is degraded by gingipains, which causes breakdown of the epithelial barrier by allowing increased epithelial permeability to PAMPs (Takeuchi, Sasaki, Yamaga, Kuboniwa, & Amano, 2019). Whether other JAM family proteins are targeted by periodontal pathogens has not been elucidated.
In the present investigation, JAM family proteins involved in barrier dysfunction caused by P. gingivalis were screened using a threedimensional (3D) multilayered tissue model. The results identified CXADR as an additional JAM family protein specifically targeted by the pathogen. Furthermore, it was shown that gingipains degrade CXADR at R145 and K235, which increases gingival epithelial permeability and allows LPS and PGN transmission. Together, the findings presented here help to better understand the process by which the gingival epithelial barrier is disturbed during the pathogenesis of periodontitis.

| P. gingivalis gingipains degrade CXADR in gingival epithelial cells
To elucidate the JAM1-independent mechanism of gingival epithelial barrier function, eight JAM family proteins (Ebnet et al., 2004) were screened for their expression in immortalised human gingival epithelial (IHGE) cells. For this purpose, reverse transcriptional polymerase chain reaction (RT-PCR) analysis was performed, which showed JAM2, CLMP and CXADR expressed in IHGE cells in addition to JAM1 (Figure 1a), while only negligible expression of JAM3, JAM4, CD2 and ESAM was observed in IHGE cells. Next, whether gingipains degrade these JAM family proteins in cells at the endogenous protein level was examined. IHGE cells were infected with P. gingivalis ATCC33277 wild type (WT) or the gingipain-deficient mutant KDP136 (Δkgp ΔrgpA ΔrgpB) at a multiplicity of infection (MOI) of 100. Decreased levels of JAM1, CXADR and CLMP were detected in IHGE cells infected with P. gingivalis WT, while the gingipain-deficient mutant had negligible effects (Figure 1b). In contrast, JAM2 was not degraded by P. gingivalis up to 1 hr after infection. These results suggest that gingipains specifically degrade JAM1, CXADR and CLMP.
To examine the roles of CXADR and CLMP expression in association with gingival epithelial permeability, IHGE cell lines stably expressing small hairpin RNA (shRNA) against JAM1 (shJAM1 #110 and shJAM1 #508), CXADR (shCXADR #38 and shCXADR #317) and The effects of secreted gingipains on CXADR were examined in the same manner as in our previous study of the impact of gingipains on JAM1 (Takeuchi et al., 2019). P. gingivalis culture supernatant was collected and administered to IHGE cells, and then 1 hr later, a decreased level of CXADR in those cells was detected. On the other hand, the gingipain-deficient mutant showed a negligible effect ( Figure S1). These results indicate that CXADR is degraded by secreted gingipains.
Next, a morphological examination of the association of P. gingivalis with CXADR in IHGE cells was conducted using laser confocal microscopy. At 1 hr after infection with P. gingivalis WT, CXADR signal intensity was reduced in the surface area of IHGE cells  (Nishiguchi, Yoshida, Matsusaki, & Akashi, 2011). Tissue samples were infected with either P. gingivalis WT or the gingipain-deficient mutant, and then CXADR degradation in deeper epithelium was analysed with confocal microscopy. At 2 hr after infection, CXADR signals in tissues infected with P. gingivalis WT were reduced, even three to four layers below the surface ( Figure 3b). In contrast, no changes in signals were seen in those infected with the gingipain-deficient mutant, thus suggesting that gingipains invade deeply into human gingival epithelial tissues and degrade CXADR.

| CXADR localizes in plasma membrane after cleavage at N-terminal signal peptide
We speculated that P. gingivalis gingipains degrade CXADR localized on the plasma membrane but not efficiently in intracellular space.
CXADR has a predicted signal peptide followed by immunoglobulin (IG)-like domains and a transmembrane domain (Bergelson et al., 1997;Tomko, Xu, & Philipson, 1997). In order to test our speculation, localization of CXADR in IHGE cells was visualised before and after cleavage of its signal peptide. A plasmid-encoding Myc-tagged human influenza haemagglutinin (HA)-inserted CXADR was created ( Figure S2), and then localization of the chimeric protein in IHGE cells was examined with confocal microscopy. To monitor CXADR localization, IHGE cells were stained with HA-tagged enhanced green fluorescent protein (HA-EGFP) ( Figure S3a) or Myc-tagged HAinserted CXADR ( Figure S3b) using the anti-HA antibody, with or without permeabilisation. We were able to label the HA-EGFP protein with the anti-HA antibody in the permeabilised cells but not in the non-permeabilised IHGE cells, suggesting that the cytosolic proteins are not stained without permeabilisation. However, we were able to label the HA-inserted CXADR protein with the anti-HA antibody even without permeabilisation, suggesting that the ectopic protein of HAinserted CXADR was properly transported to the plasma membrane.
In order to trace the pathway(s) of CXADR, examinations of IHGE cells expressing Myc-tagged HA-inserted CXADR protein and each organelle marker, EGFP-SEC61β (located in the membrane of endoplasmic reticulum) and TOMM20 (located in the outer membrane of mitochondria, not included in the endomembrane system), were performed with confocal microscopy. Co-localization of anti-Myc and anti-HA signals with SEC61β ( Figure S4), but not TOMM20 ( Figure S5), was detected, indicating transport of CXADR via an endomembrane pathway. Additionally, phalloidin-stained actin in the peripheral area of IHGE cells was found co-localized with anti-HAlabelled CXADR but not the Myc signal ( Figure S6). These results indicate that the signal peptide of CXADR is cleaved at the endoplasmic reticulum and transported to the plasma membrane and that gingipains are capable of targeting the mature form of CXADR at the cell surface.
2.3 | P. gingivalis, but not Streptococcus gordonii or Fusobacterium nucleatum, degrades CXADR To examine whether P. gingivalis degrades both immature and mature CXADR, IHGE cells expressing Myc-tagged HA-inserted CXADR were infected and the kinetics of Myc-and HA-tagged CXADR analysed. In consideration of the specific gingipain property of the substrate, an HA tag was utilised as a marker of mature CXADR, as K and R residues are not included in the HA amino acid sequence (YPYDVPDYA).
In addition, immature and mature CXADR in immunoblots were distinguished based on the N-terminal Myc tag. As shown in Figure S7a, P. gingivalis infection resulted in a decreased amount of mature CXADR stained with anti-HA, while the level of immature CXADR labelled with anti-Myc was not changed, suggesting that the mature form is targeted by P. gingivalis. To examine the effects of another strain, IHGE cells were infected with P. gingivalis TDC60 isolated from a severe periodontal lesion (Watanabe et al., 2011). At 1 and 2 hr after infection, the mature form of CXADR was markedly reduced ( Figure S7b), indicating that other P. gingivalis strains are capable of degrading CXADR.
F. nucleatum and S. gordonii are human oral bacteria and participate in establishment of mixed-species communities, leading to periodontopathic biofilm formation (Kolenbrander et al., 2002). In order to examine their effects on CXADR, IHGE cells expressing Myctagged HA-inserted CXADR were separately infected with those species. As demonstrated in Figure S7c,d, both S. gordonii and F. nucleatum failed to reduce the level of HA-inserted CXADR at 2 hr F I G U R E 2 Coxsackievirus and adenovirus receptor (CXADR) loss increases epithelial barrier permeability. (a) Schematic representation of culture insert system. Immortalised human gingival epithelial (IHGE) cells expressing the indicated shRNA were cultured in culture inserts. A fluorescein isothiocyanate (FITC)-conjugated tracer was added to cell culture medium in the upper compartment. At 30 min after administration, tracer transmission from the upper to lower compartment was analysed by spectrometry. (b-d) Immunoblot analysis for detection of knockdown of indicated proteins in IHGE cells. (e-g) Permeability to FITC-labelled dextran 40 kDa (e), Porphyromonas gingivalis lipopolysaccharide (LPS) (f) and P. gingivalis peptidoglycan (PGN) (g) in IHGE cells stably expressing indicated shRNA. Results are shown as fold change relative to control cells expressing shLuc (M ± SD, n = 8). *p < .05, one-tailed Dunnett's test after infection, suggesting that CXADR cannot be degraded by either of these organisms.

| CXADR R145 and K235 responsible for degradation by P. gingivalis
In order to examine which residue(s) are responsible for degradation by P. gingivalis gingipains, plasmids encoding deleted and mutated CXADR were constructed, and structural analysis was performed ( Figure S8a). The plasmids were transfected into IHGE cells and then infected with P. gingivalis. To detect residue(s) targeted by gingipains, the R and K residues were replaced with H, a basic amino acid, in point mutation constructs. Figure S8b shows

| CXADR prevents penetration of LPS and PGN through gingival epithelium
Although it was previously reported that CXADR mediated homotypic cell adhesion in Chinese hamster ovary cells (Cohen et al., 2001), the barrier function of CXADR in human squamous epithelium remains F I G U R E 3 Degradation of coxsackievirus and adenovirus receptor (CXADR) in IHGE cells infected with Porphyromonas gingivalis gingipains. (a) Confocal microscopic analysis for detection of CXADR in immortalised human gingival epithelial (IHGE) cells infected with P. gingivalis wildtype (WT) or Δkgp ΔrgpA ΔrgpB mutant. Cells were fixed, then stained with 4 0 ,6-diamidino-2-phenylindole (DAPI) (cyan) and anti-CXADR (yellow). Bars, 10 μm. (b) Gingival epithelial tissues were infected with P. gingivalis WT or Δkgp ΔrgpA ΔrgpB mutant. Tissues were stained with anti-CXADR (green) and Alexa Fluor 568-conjugated phalloidin (magenta) and then analysed using confocal microscopy. Bars, 30 μm unknown. To examine the role of CXADR expression in gingival epithelium permeability, a 3D-tissue model stably expressing shLuc or shCXADR was generated, and then permeability to a fluorescent probe was assayed (Figure 4a). Knockdown of CXADR expression was confirmed by confocal microscopy (Figure 4b). Figure 4c shows that the 3D-tissue model expressing shRNA against CXADR had greater permeability to FITC-dextran 40 kDa than shLuc-expressing cells, suggesting CXADR involvement in molecular flux along with dextran 40 kDa via gingival epithelium.
In addition, increased permeability to FITC-labelled P. gingivalis in IHGE cells expressing shCXADR #317 or shJAM1 #508 was examined with confocal microscopy. As shown in Figure S9a, localization of JAM1 in an IHGE monolayer expressing shJAM1 was negligible, whereas that was demonstrated in IHGE cells expressing shCXADR.
In the same manner, the expression level of CXADR in IHGE monolayers expressing shCXADR was negligible but remained in IHGE cells expressing shJAM1 ( Figure S9b). Thus, it was shown that CXADR and JAM1 are not interdependent in regard to localization in gingival epithelial cells.
To assess the interdependence of CXADR and JAM1 for localization of each in gingival epithelium, a 3D-tissue model of IHGE cells expressing shCXADR and shJAM1 was generated, then the localization of deeply seated CXADR and JAM1 was analysed using confocal microscopy. Immunoblot results confirmed the expression levels of shCXADR and shJAM1 expressed in IHGE cells ( Figure 5a). As shown in Figure 5b

| Gingipains penetrate gingival epithelial barrier by degrading CXADR
We previously reported that JAM1 degradation by gingipains is required for gingipain penetration in gingival epithelial cells (Takeuchi et al., 2019).
In order to examine the CXADR-dependent mechanism of penetration by gingipains, a two-layered cell culture model that can detect penetration of gingipains from the upper to lower space was utilised (Figure 6a).
Six hours after administration of bacterial culture media, the level of HAinserted CXADR, but not Myc-tagged CXADR, was decreased in the cells of the lower layer treated with P. gingivalis WT in a greater amount as compared to treatment with the gingipain-deficient mutant (Figure 6b).
Next, we examined whether overexpression of CXADR, the expression level of which compensates CXADR degradation by P. gingivalis, blocks loss of CXADR in gingival epithelial cells. For this experiment, IHGE cells overexpressing CXADR were cultured in a two-layered model, with localization of CXADR in cells of the lower layer infected with P. gingivalis monitored. When IHGE cells overexpressing CXADR were cultured in medium that included P. gingivalis for 30 min, CXADR proteins were detected at nearly the same level as in the non-treated IHGE cells ( Figure S10), which suggested that degradation can be compensated for by overexpression of CXADR in this system. In line with those findings, at 6 hr after administration, an increased level of HA-inserted CXADR was detected in the lower layer when IHGE cells overexpressing CXADR were present in the upper layer (Figures 6b and S11), indicating that degradation of CXADR by gingipains is important for permeation of gingival epithelial cells by proteases.

| Degradation of CXADR by P. gingivalis causes penetration of LPS and PGN
Finally, a 3D-tissue model using IHGE cells overexpressing CXADR was generated (Figure 7a). Marked amounts of CXADR were found to F I G U R E 6 Coxsackievirus and adenovirus receptor (CXADR) involved in penetration of Porphyromonas gingivalis gingipains through immortalised human gingival epithelial (IHGE) cells. (a, b) Schematic representation of culture insert model (a). Gingival epithelial wild-type (WT) cells or those stably expressing Myc-tagged haemagglutinin (HA)-inserted CXADR were cultured in the upper compartment, with IHGE cells stably expressing Myc-tagged HA-inserted CXADR cultured in the lower compartment on a coverslip. Culture supernatant of P. gingivalis WT or Δkgp ΔrgpA ΔrgpB mutant was administered to the cells. At 6 hr after administration, IHGE cells in the lower compartment were stained with anti-Myc (green) and anti-HA (magenta) and analysed using confocal microscopy (b). Bars, 10 μm. See also Figure S11 F I G U R E 7 Porphyromonas gingivalis degrades coxsackievirus and adenovirus receptor (CXADR) of gingival epithelium, allowing penetration by lipopolysaccharide (LPS) and peptidoglycan (PGN). (a, b) Schematic representation of 3D culture model (a) and confocal microscopic images (b) of 3D tissue models with CXADR detected in tissues (wild-type [WT] or overexpressing CXADR) infected with P. gingivalis for 30 min. Tissues were stained with anti-CXADR (green) and Alexa Fluor 568-conjugated phalloidin (magenta). Bars, 30 μm. (c-g) Permeability to fluorescein isothiocyanate (FITC)-labelled dextran 40 kDa (c), P. gingivalis LPS (d), P. gingivalis PGN (e), Escherichia coli LPS (f) and Staphylococcus aureus PGN (g) in gingival epithelial tissues (WT or overexpressing CXADR) infected with P. gingivalis. Results are calculated as fold change relative to uninfected WT cells (M ± SD, n = 7). *p < .05, one-tailed t test (closed testing procedure) remain following infection with P. gingivalis (Figure 7b). In order to examine permeability, tissues were treated with FITC-labelled dextran 40 kDa, P. gingivalis LPS or P. gingivalis PGN. At 30 min after administration, decreased permeability to FITC-labelled dextran 40 kDa ( Figure 7c), P. gingivalis LPS (Figure 7d), P. gingivalis PGN (Figure 7e), E. coli LPS (Figure 7f) and S. aureus PGN (Figure 7g) was detected in tissues overexpressing CXADR. These results show that degradation of CXADR by P. gingivalis is a key factor for allowing penetration by LPS and PGN through gingival epithelium.

| DISCUSSION
Based on results obtained in this study, we propose a model to demonstrate gingival epithelial barrier dysfunction caused by P. gingivalis ( Figure 8). CXADR and JAM1 were found localized in phalloidinstained plasma membrane regions in 3D tissue ( Figure 5b) and are considered to function together as a gingival epithelial barrier. By degrading not only JAM1 but also CXADR, P. gingivalis efficiently allows penetration of LPS, PGN and gingipains through gingival epithelium and then into deeper periodontal tissues. In this manner, it is likely that P. gingivalis can breach the epithelial barrier of gingival tissues.
We were able to generate gingival epithelium even with knockdown of CXADR and JAM1 (Figure 5b), indicating that neither is necessary for gingival epithelium reconstruction. Hence, the present cellaccumulation technique is considered useful for defining the function of JAMs in other types of epithelium as well.
HA-inserted CXADR was also generated for use as a probe in gingival epithelial cells, which enables detection of amino acid residues targeted by gingipains. P. gingivalis gingipains degrade CXADR at K235 ( Figure S6) and JAM1 at R234 (Takeuchi et al., 2019). The hinge region between the C-terminal IG-like domain and transmembrane domain, possessing CXADR K235 and JAM1 R234, does not contain a consensus sequence for N-glycosylation or a secondary structure, including helix, β-sheet and turn (Uniplot, JAM1: https://www. uniprot.org/uniprot/Q9Y624; CXADR: https://www.uniprot.org/ uniprot/P78310). It has been shown that the N-terminal IG-like domain of CXADR and JAM1 is responsible for homodimerisation (Mandell, McCall, & Parkos, 2004;Patzke et al., 2010), thus gingipains have a function to effectively separate the N-terminal region away from the transmembrane domain.
Single knockdown of CXADR or JAM1 was found to increase permeability to dextran 40 kDa in gingival epithelial tissues at nearly the same level (Figure 5c), while double knockdown of those resulted in a greater increase of permeability as compared to knockdown of either ( Figure 5c). Upregulated CXADR expression has been shown to be inhibited by the AP-1 inhibitor (Azari et al., 2011) and that of JAM1 messenger RNA by the NF-κB inhibitor (Chung et al., 2019). Thus, the multiple transcription pathways of CXADR and JAM1 may provide gingival epithelium a robust defense against external bacterial stress.
The present findings showed that CXADR or JAM1 overexpression rescues gingival epithelial tissues infected with P. gingivalis from permeability ( Figure 7) (Takeuchi et al., 2019), indicating that the barrier function of each protein against P. gingivalis is complementary, but not interdependent. Gingival epithelial function to control barrier F I G U R E 8 Proposed model of transfer of bacterial virulence factors by Porphyromonas gingivalis gingipains through gingival epithelium. Permeability of gingival epithelial tissues coxsackievirus and adenovirus receptor (CXADR) (magenta) and junctional adhesion molecule 1 (JAM1) (cyan) is not interdependent. P. gingivalis gingipains degrade CXADR and JAM1, leading to increased permeability to gingipains and other factors. Subsequently, gingipains become translocated to deeper epithelium for additional degradation of CXADR and JAM1, thus allowing lipopolysaccharide (LPS) and peptidoglycan (PGN) to penetrate the gingival epithelium and reach subepithelial tissues permeability is an issue that should be further investigated in association with the etiology of periodontal disease.
When considering the results of this study, at least two questions remain that should be considered in the future. One is whether other major pathogens can specifically degrade CXADR and JAM1.
Another important question is regarding the relationship between protein structure and degradation. In our previous investigation, we showed that CXADR and JAM1, but not Claudin (CLDN) 1 or CLDN 4, were degraded by P. gingivalis in gingival epithelial cells (Takeuchi et al., 2019). JAM family proteins possess a single transmembrane domain, while CLDN family proteins are members of the tetra-spanstransmembrane family and possess two extracellular loops (Tsukita et al., 2001). Accordingly, it will be necessary to examine other proteins for gingipain-dependent degradation, which are potentially involved in regulation of the gingival epithelial barrier without an extracellular loop.

| EXPERIMENTAL PROCEDURES
4.1 | Bacteria and cell culture P. gingivalis ATCC 33277 was purchased from the American Type Culture Collection. P. gingivalis TDC60 was kindly provided by Kazuyuki Ishihara (Tokyo Dental College) and P. gingivalis KDP136 (Δkgp ΔrgpA ΔrgpB) by Koji Nakayama (Nagasaki University) (Shi et al., 1999). P. gingivalis, S. gordonii ATCC 35105 and F. nucleatum subsp. nucleatum ATCC 25586 were grown using methods previously described (Takeuchi et al., 2019). IHGE cells were kindly provided by Shinya Murakami (Osaka University) (Murakami et al., 2002). The P. gingivalis bacterial culture supernatant was prepared as described in a previous report (Takeuchi et al., 2019). IHGE cell cultures and epithelial barrier function assays were performed as previously reported (Takeuchi et al., 2019).

| Antibodies, reagents and experimental equipment
Antibodies and reagents, and the experimental equipment used in this study are shown in Table S1. Immunoblotting and immunocytochemistry were performed according to methods previously reported (Takeuchi et al., 2019).

| Epithelial barrier function assay
The epithelial barrier function assay was performed as previously described (Takeuchi et al., 2019). Briefly, the in vitro epithelial permeability assay to assess barrier function was performed with 12-well cell culture inserts (353,180;Corning). Preparation of the FITClabelled tracer was done according to a method previously described (Takeuchi et al., 2019). When IHGE cells in the upper compartment reached 100% confluence, 20 μl of FITC-dextran, FITC-LPS or FITC-PGN was added to the upper compartment of the insert. After a 30-min incubation, the medium was collected from the lower compartment, and fluorescence intensity was determined using a microplate reader (1,420 ARVO X; PerkinElmer).

| Plasmids
The empty vectors used in this study are shown in Table S1. The plasmid-encoding Myc-tagged HA-inserted CXADR was constructed by cloning PCR products amplified from IHGE cells into pCMV-Myc (Clontech) using exogenously added EcoRI and KpnI sites. To produce HA-inserted CXADR, the DNA sequence of HA-tag was inserted into CXADR using fusion PCR. Plasmids encoding HA-tagged CXADR deletion mutants and point mutations were constructed by PCR.

| Generation of cell line stably expressing CXADR
A plasmid-encoding CXADR was constructed by cloning PCR products amplified from the pCMV plasmid into pMRX-IRES-Puro (Takeuchi et al., 2019). The pMRX-IRES-Puro-CXADR plasmid was used for overexpression of cDNA in IHGE cells. IHGE cells stably expressing CXADR were selected, as previously described (Takeuchi et al., 2019).

| Statistical analysis
p values were determined by use of t-test or Dunnett's test with the Excel application (Microsoft), with p < .05 considered to indicate significance.