INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

The proteolytic enzymes of Porphyromonas gingivalis are widely recognized as important virulence factors of this organism. In addition to enabling it to access essential nutrients, they may also perturb host defense and tissue homeostasis mechanisms by degrading a range of host proteins including plasma proteinase inhibitors and immunoglobulins, dysregulating coagulation, complement, and kallikrein/kinin cascade pathways (7, 29), interfering with cellular functions (14), and degrading periodontal tissue components directly (2, 27, 30) and indirectly (28). These mechanisms may all contribute to the role of P. gingivalis as a major causative organism in human periodontal disease (26, 33). In animal model systems using subcutaneous inoculation, greater protease activity has also been associated with increased virulence of P. gingivalis (12, 18).

The trypsin-like enzyme activity of this bacterium, which has been the focus of much research, is now known to be due to a mixture of proteases with individual specificity for arginine and lysine residues (21). These proteases are associated with membrane vesicles and may also be released extracellularly. Partially purified bacterial fractions with proteolytic activity have been shown to degrade basement membrane collagen, elastin, and fibronectin (27, 30) and to stimulate the secretion of collagenase and plasminogen activator by cultured gingival fibroblasts, thereby inducing the host cells to degrade their own pericellular matrix (31). Such matrix degradation may lead to the marked loss of connective tissue integrity which is typical of destructive periodontal disease.

Cells bind to extracellular matrix components via interaction with integrin surface receptors which are linked through their intracellular domains to the cytoskeleton. Integrin receptors and their ligands are known to be targets for binding by a number of pathogens which exploit this group of molecules in order to adhere to and/or invade host cells (13, 19). We have previously shown that components of the culture supernatant of P. gingivalis W83 can damage human gingival fibroblast (HGF) integrin-substrate interactions, with the ₅ and ₁ integrin subunitsthe receptor for fibronectinbeing considerably more susceptible than _V and ₃the receptor for vitronectin (24). These effects were reduced by heating the supernatant, implicating heat-labile proteins such as bacterial proteases. Here we report similar effects with the supernatant from the virulent P. gingivalis strain W50 but not the nonpigmented avirulent variant (W50/BE1) and, using purified arginine-specific proteases from strain W50, examine their site and mechanism of action on HGF in vitro.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Bacterial culture and supernatant preparation. P. gingivalis W50 and W50/BE1 were grown in brain heart infusion broth supplemented with hemin (5 mg liter¹) in an atmosphere of 80% N₂, 10% H₂, and 10% CO₂ at 37°C for 6 days. Culture supernatants obtained by centrifugation (11,000 × g for 20 min at 5°C) were sterilized by passage through 0.2-µm-pore-size cellulose acetate membranes and stored at 70°C.

1

Treatment of supernatant with protease inhibitor. Samples of filter-sterilized culture supernatant were monitored for arginine-specific protease activity, and the enzyme activity was then irreversibly inactivated by incubation for 30 min at 4°C with 1 mM N-p-tosyl-L-lysine chloromethyl ketone (TLCK) in the presence of 10 mM -mercaptoethanol and CaCl₂ (23). Excess inhibitor was removed by dialysis using 50 mM sodium acetate buffer (pH 5.3). Further samples of supernatant, not treated with inhibitor, were dialyzed as controls, and the protease assay was repeated. The enzyme activity of the inhibitor-treated supernatant was compared with that of the dialyzed control supernatant.

Purification of arginine-specific proteases. The arginine-specific proteases (RI and RIA) from strain W50 were purified as described previously (23), using a combination of ammonium sulfate precipitation, affinity chromatography on arginine-agarose, and ion-exchange chromatography. Both RI and RIA were stored at 4°C in 50 mM sodium acetate buffer and neutralized with 1 N NaOH immediately before use with cultured fibroblasts. Where indicated, samples were treated with TLCK as described above.

HGF culture and treatment with P. gingivalis preparations. HGF obtained from clinically healthy gingival tissue were grown at subconfluent concentrations on 13-mm-diameter glass coverslips as previously described (24). The culture medium was Dulbecco's modified Eagle's medium (DMEM; Life Technologies Ltd., Paisley, Scotland) supplemented with 10% (vol/vol) fetal bovine serum (Globepharm Ltd., Surrey, England), penicillin (50 IU/ml), streptomycin (50 µg/ml), and amphotericin B (0.25 µg/ml) (Life Technologies). After incubation overnight to allow attachment, cells were exposed for 1 h at 37°C to doubling dilutions of W50 (1/2 to 1/256) and W50/BE1 (1/2 to 1/64) culture supernatant, TLCK-treated W50 culture supernatant (1/2 to 1/16), RI (protease activity 1.6 to 0.006 U/ml) and RIA (protease activity 1.4 to 0.175 U/ml), RI and RIA with or without 5 mM L-cysteine, and RI and RIA with or without TLCK treatment. Cell monolayers exposed to purified proteases were prewashed in serum- and supplement-free DMEM, and doubling dilutions made in this medium as the presence of serum protein interfered with the activity of purified proteases. Each experiment was repeated at least three times using different HGF cell lines.

Immunofluorescence studies. The mouse monoclonal antibody against the ₁ integrin subunit (clone DF7; Affiniti Research Products Ltd., Exeter, England) was used in all experiments. Additional primary antibodies, used where indicated in Results, were mouse monoclonal antibodies against ₅ and _V integrin subunits (P1D6 and VNR 147, respectively; Life Technologies), ₃ subunit (clone BB10, purified; Chemicon International Ltd., Harrow, England), and fibronectin (clone Fn-3; Cymbus Bioscience Ltd., Southampton, England). Localization of the purified proteases was examined by using rabbit antiserum raised against P. gingivalis W50 whole cells (PgWC) (9) and monoclonal antibody 1A1 (6), which reacts only with the  component of the RI heterodimer (8). All primary antibodies were diluted in PBS (pH 7.4) containing 0.05% (wt/vol) sodium azide and 5% (vol/vol) human serum and used at 1/100 except PgWC, which was diluted to 1/20,000. Omission of the primary antibodies served as the negative controls. After washing in PBS, cells used for single-labelling studies were incubated for 60 min with a 1/200 concentration of biotin-conjugated goat anti-mouse antibody (B0529; Sigma Chemical Company Ltd., Poole, England) or swine anti-rabbit antibody (code EO353; Dako Ltd., High Wycombe, England) as appropriate, followed by a 1/100 concentration of streptavidin-Texas red complex (RPN 1233; Amersham International, Little Chalfont, England) for 60 min. Double-labelling studies with mouse primary antibodies against integrins or fibronectin combined with PgWC antibody used fluorescein isothiocyanate-conjugated goat anti-mouse immunoglobulin G (Fab specific) (F5262; Sigma) at 1/50 dilution in addition to the biotin-conjugated swine anti-rabbit secondary antibody followed by streptavidin-Texas red as described above. All coverslips were mounted on glass slides by using Immumount [Life Sciences International (UK) Ltd., Hampshire, England) containing 1,4-diazabicyclo [2,2,2] acetone (2.5 mg/ml) to reduce fading and viewed with a Nikon Microphot FXA microscope equipped for epifluorescence with excitation filters of 460 to 500 and 510 to 560 nm.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Comparative effects of culture supernatant from P. gingivalis W50 and W50/BE1 on ₁ integrin in HGF. HGF incubated for 1 h with strain W50 culture supernatant displayed a dose-dependent loss of ₁ integrin staining, with control levels reached at or below a 1/256 concentration. In contrast, culture supernatant from W50/BE1 produced only minor loss of ₁ integrin even at the highest concentration used (1/2), and control levels were seen at a 1/32 concentration (Fig. 1). The arginine-specific protease activities of undiluted W50 and W50/BE1 culture supernatants were 3.75 and 0.95 U/ml, respectively, and W50 supernatant demonstrated greater loss of ₁ integrin staining than W50/BE1 at all comparable enzyme activities.

View larger version (17K): [in this window] [in a new window]   FIG. 1.   1 staining in HGF incubated for 1 h with culture supernatant from strains W50 () and W50/BE1 (). The maximum concentration of each supernatant used was a 1/2 dilution in DMEM. The arginine-specific enzyme activities of doubling dilutions of the culture supernatants are shown on a log2 scale. Integrin staining was graded as described in Materials and Methods. The results represent the mean integrin grade ± standard error (SE) from five separate experiments for W50 and three for W50/BE1. Integrin staining in control cultures incubated in DMEM is indicated by the dotted line.

Action of RI and RIA on integrin staining. RI is an ~110-kDa heterodimer containing two subunits: an  component with protease activity and a  component of similar size which possesses adhesin-like properties (8, 23). RIA is a monomer of only the catalytic  chain. Incubation of HGF with each of these purified proteases caused dose-dependent losses of ₁ integrin staining (Fig. 2). However, at comparable arginine-specific protease activities, RI was considerably more active that RIA, with control integrin levels reached at activities of 0.03 U/ml for RI, compared with 0.36 U/ml for RIA, indicating a 12-fold increase in activity of RI. In certain experiments, a final concentration of 5 mM L-cysteine was included in the HGF culture medium during incubation with the purified proteases. This enhanced the ₁ integrin loss between two- to fourfold compared with cultures from which L-cysteine was omitted.

View larger version (16K): [in this window] [in a new window]   FIG. 2.   1 integrin staining in HGF incubated for 1 h with RI (solid line) and RIA (dashed line) in DMEM without supplements. 1 integrin staining, as described previously, is plotted against arginine-specific enzyme activity on a log2 scale. The results represent the mean integrin grade ± standard error (SE) from five separate experiments.

Localization of P. gingivalis components in HGF cultures treated with TLCK-inactivated proteases. As RI differs from RIA only in the possession of a  chain in addition to the catalytic  chain, this indicated that the  component contributed to the greater loss of integrin expression that we observed with RI. The sites of binding of these proteases to HGF were then examined by incubating cells with TLCK-treated RI and RIA, in which all proteolytic activity had been inhibited, followed by immunofluorescent labelling with antibodies 1A1 and PgWC. With monoclonal antibody 1A1, which binds to the  chain of RI (8), RI-treated cells showed a linear staining pattern along the long cytoplasmic axis (Fig. 3A) typical of that seen with ₁ and ₅ integrins. In addition, a network of strong intercellular staining, which was particularly prominent at low magnification, was visible across the cell monolayer. In contrast, RIA-treated cells displayed only diffuse, nonspecific staining (Fig. 3B) comparable with that which occurred with control cells cultured in DMEM alone, confirming that no  chain was associated with these cells. With the PgWC antibody raised against whole P. gingivalis W50 cells, RI-treated HGF showed a staining pattern (Fig. 3C) similar to that observed with antibody 1A1, indicating that it was detecting mainly the adhesin domain of RI, whereas RIA treatment resulted in patchy staining consisting mainly of granular deposits with short linear bands (Fig. 3D) which were much less intense than those seen with RI.

View larger version (69K): [in this window] [in a new window]   FIG. 3.   HGF incubated for 1 h with TLCK-treated RI (A and C) or TLCK-treated RIA (B and D) and stained with antibody 1A1 (A and B), which recognizes only the adhesin component of RI, or PgWC antibody against whole-cell components (C and D). The nuclear staining seen in panels A and B was the result of nonspecific staining which occurred with certain batches of biotin-conjugated anti-mouse secondary antibodies. Magnification, ×750; bar = 10 µm.

View larger version (99K): [in this window] [in a new window]   FIG. 4.   HGF incubated for 1 h with TLCK-treated RI (A to I) or TLCK-treated RIA (J to L). The first column shows cells stained with the PgWC antibody, and the second column shows the same cells double labelled with mouse antibodies against 1 integrin (B), V integrin (E), and fibronectin (H and K). The final column shows cells multiply exposed to both labels. RI-treated cells showed a comparable linear distribution of both P. gingivalis components and 1 integrin along the cytoplasm (C), whereas the peripheral distribution of V integrin mainly around the tips of the cell differed from the more central staining with PgWC (F). Exact colocalization of fibronectin and P. gingivalis components in RI-treated cells is indicated by the yellow, generally linear, cytoplasmic staining (I), whereas the weaker staining of P. gingivalis components in RIA-treated cells resulted in a predominantly green staining due to the fibronectin (L). Identical exposure periods were used for panels I and L. The patterns of staining with 5 and 3 integrin subunits (not shown) resembled those illustrated for 1 and V respectively. Magnification, ×750; bar = 10 µm.

Comparative effects of RI and RIA on degradation of fibronectin in HGF cultures. Having demonstrated the precise colocalization of TLCK-inactivated RI with fibronectin in contrast with the less intense, patchy association seen with RIA, we then compared the action of the active enzymes on degradation of the cell-associated fibronectin network. HGF cultures were incubated for 1 h with doubling dilutions of RI and RIA as already described. We observed a dose-dependent loss of fibronectin which closely followed the pattern of loss of ₁ integrin staining seen in Fig. 2, with RI showing considerably greater activity than RIA at all corresponding protease activities. At 0.2 to 0.4 U of arginine-specific enzyme activity per ml, loss of fibronectin staining occurred with RI (Fig. 5A), whereas staining was clearly visible in cells treated with RIA at the same enzyme activity (Fig. 5B) and remained visible at 1 U/ml, although less prominent than in untreated control cultures (Fig. 5C).

View larger version (38K): [in this window] [in a new window]   FIG. 5.   HGF stained to demonstrate fibronectin after incubation for 1 h with RI (A) or RIA (B), both at 0.2 U of arginine-specific enzyme activity per ml, compared with control cells in DMEM (C). RI-treated cells show only remnants of fine granular cytoplasmic staining and often an indistinct nucleus, in contrast with the clear fibronectin network associated with RIA-treated and control cells. Following removal of cells, the fibronectin matrix was totally disrupted after exposure to RI at 0.2 U/ml (D), whereas a clear fibronectin network was retained in the presence of a similar concentration of RIA (E) and in the DMEM control (F). Magnification, ×750; bar = 10 µm.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

This study has investigated a mechanism by which P. gingivalis might target the damaging effects of its arginine-specific enzymes within periodontal tissues. We have shown that major disruption of fibronectin-integrin interactions in HGF is due to the adhesin-mediated, targeted delivery of the catalytic domain of RI to specific sites on the cells.

Microorganisms and their products need to adhere to host cell surfaces and/or invade cells to establish an infection. Many have evolved strategies involving binding of microbial adhesins to cell receptors or extracellular matrix molecules to facilitate their interaction with the host (19). Binding to fibronectin has been extensively demonstrated by many microbial pathogens, including staphylococci, streptococci, Candida albicans, and Treponema pallidum. Other microorganisms exploit interactions with -chain integrins leading to internalization (13). Invasin, an outer membrane protein which binds to several different ₁ integrin receptors promoting uptake, has been widely studied in enteropathogenic Yersinia species. In addition, adherence of some bacteria to host tissue is mediated by fimbriae. Interestingly, the binding of P. gingivalis fimbriae to human fibroblasts and extracellular matrix components has recently been shown to be enhanced by arginine-specific protease treatment, resulting in the exposure of hidden binding sites on host tissues (15, 16).

The enhanced activity displayed by the RI heterodimer compared with the monomeric RIA with similar protease activity, which we report, indicates that the adhesin chain of RI is important in the loss of both ₅ and ₁ integrin subunits of HGF. The exact colocalization of RI with fibronectin and its major ₅₁ receptor demonstrated in HGF cultures would provide the bacterium with a method of targeting its protease component to specific susceptible regions within the host tissue. A similar targeting mechanism may operate in P. gingivalis-mediated hemagglutination (8). However, as RIA, which contains only the protease domain, retains some ability to localize to fibronectin, the possibility that an additional recognition site exists on the catalytic  chain should be considered.

The use of TLCK-treated purified proteases has allowed their sites of action on HGF to be demonstrated in the absence of the normal proteolytic degradation. In the presence of active protease, RI caused loss of the cell-associated fibronectin network under conditions in which ₅₁ integrin receptor loss was also observed. In contrast, comparable concentrations of RIA caused considerably less destruction of fibronectin and loss of the ₅₁ integrin receptor. Internalization of certain integrins, including ₅₁, can occur rapidly in the absence of the corresponding ligands (4) and is enhanced if the integrins are cross-linked (11). As ligand binding is known to anchor receptors at the cell surface, it is likely that the protease-mediated destruction of fibronectin that we report is the primary event which leads to the loss or lack of detection of its ₅₁ receptor.

The failure of P. gingivalis to localize at sites of _V and ₃ integrins (the vitronectin receptor) and the resistance of these integrins to destruction by arginine-specific proteases, even after 24 h incubation, is in contrast with the colocalization with and susceptibility of the ₅₁ fibronectin receptor. As the vitronectin receptor is susceptible to destruction by P. gingivalis culture supernatant after prolonged incubation (24), this would appear to be mediated by a factor(s) other than arginine-specific proteases.

The failure to demonstrate any very significant enhancement of activity of RI or RIA on HGF integrin loss in the presence of cysteine was unexpected and contrasts with the rapid and marked increase in hydrolysis of synthetic substrates by both proteases on addition of cysteine (23). It is possible that sufficient reducing conditions exist locally to activate the protease. However, the fibronectin trails attached to the glass surface after removal of the cells with 0.02% EDTA remained susceptible to destruction by RI in the absence of cysteine. Hence, localized reducing conditions, if they exist, are clearly not cell dependent. The precise mechanisms of action of RI and RIA therefore require further investigation.

Our initial experiments which compared the ₁ integrin loss with W50 and W50/BE1 culture supernatants indicated that the former is considerably more active, even when used at a comparable arginine-specific protease activity. This apparent discrepancy could be due to the existence of five biochemically distinct forms of arginine-specific proteases (RI, RIA, RIB, RIIA, and RIIB) (22). All contain the catalytic  chain and therefore contribute to the BApNA activity, but the distribution of the isoforms differs between W50 and W50/BE1 (5) and the relative effects of the different enzymes on integrins are not known. Furthermore, W50/BE1 is a pleiotropic mutant with multiple differences from the parent strain (17, 25), which could contribute to the different effects seen with the two culture supernatants.

Fibroblasts are anchorage-dependent cells, and cell attachment is required for normal growth and proliferation. Specific extracellular ligand-integrin interactions initiate intracellular signalling events which regulate the cell phenotype. Thus, any disruption of this interaction is likely to have major consequences for the cell. As well as degrading extracellular matrix components directly, P. gingivalis factors can also stimulate host cells to break down their own matrix. A proteinase from P. gingivalis culture medium has been reported to stimulate collagen degradation by cultured epithelial cells (3) via a mechanism involving activation and processing of latent host matrix metalloproteinases MMP-1, MMP-3, and MMP-9 (10). Degradation products of fibronectin also signalled changes in metalloproteinase gene expression in rabbit synovial fibroblasts (32). Similar mechanisms may have contributed to the degradation of cell surface and matrix glycoproteins in cultured gingival fibroblasts described by Uitto et al. (31). The specific binding of P. gingivalis adhesins to extracellular matrix proteins such as fibronectin, described here, may represent one mechanism by which this organism can target its protease and is thus likely to have a significant impact on the functioning of host cells resident in connective tissue. As the adhesin domain of P. gingivalis proteases binds to laminin and fibrinogen in addition to fibronectin (20), it is likely that the catalytic effects of the proteases are also targeted at other specific cells and tissues. These possibilities are under further investigation.