Comparative Properties of Two Cysteine Proteinases (Gingipains R), the Products of Two Related but Individual Genes ofPorphyromonas gingivalis *

Proteolytic enzymes produced byPorphyromonas gingivalis are important virulence factors of this periodontopathogen. Two of these enzymes, referred to as arginine-specific cysteine proteinases (gingipains R), are the product of two related genes. Here, we describe the purification of an enzyme translated from the rgpB/rgp-2 gene (gingipain R2, RGP-2) and secreted as a single chain protein of 422 residues. The enzyme occurs in several isoforms differing in pI, molecular mass, mobility in gelatin zymography gels, and affinity to arginine-Sepharose. In comparison to the 95-kDa gingipain R1, a complex of catalytic and hemagglutinin/adhesin domains, RGP-2 showed five times lower proteolytic activity, although its activity on various P1-arginine p-nitroanilide substrates was generally higher. Gingipains R amidolytic activity, but not general proteolytic activity, was stimulated by glycyl-glycine. However, in cases of limited proteolysis, such as the inactivation of α-1-antichymotrypsin, glycyl-glycine potentiated inhibitor cleavage. In contrast, α-1-proteinase inhibitor was not inactivated by gingipains R and only underwent proteolytic degradation during boiling in reducing SDS-polyacrylamide gel electrophoresis treatment buffer. Similarly, native type I collagen was completely resistant to cleavage by gingipains but readily degraded after denaturation. Together, these data explain much of the controversy regarding gingipains structure and substrate specificity and indicate that these enzymes function asP. gingivalis virulence factors by proteolysis of selected target proteins rather than random degradation of host connective tissue components.

Periodontal disease, the major cause of tooth loss in the general population of industrial nations (1,2), is a chronic inflammatory disorder of tissues directly supporting the root of the tooth in alveolar bone sockets. The disease is characterized by bone resorption, loss of tooth attachment, and formation of periodontal pockets infested with specific bacteria (3). Although more than 400 distinct bacterial taxa have been identified in dental plaque samples, very few can be correlated with the development and/or progression of the specific clinical variants of periodontitis (2). Adult onset disease, the most common form of periodontitis, is strongly associated with infections by Porphyromonas gingivalis (4), and this is further substantiated by the finding that infection by this bacterium initiates periodontitis in primates (5) and can be avoided by animal immunization with P. gingivalis antigens (6).
P. gingivalis, a Gram-negative, anaerobic, nonmotile, nonsporing short rod elaborates a multiplicity of virulence factors involved in invasion, tissue destruction, and evasion of host defenses, all of which enable this bacterium to colonize within its ecological niche of either the gingival sulcus or periodontal pocket (reviewed in Ref. 7). Among these factors, the primary focus of research has been on proteolytic enzymes that are produced in large quantities by this bacterium and could add to its pathogenicity. Indeed, it was shown that these proteinases can directly or indirectly degrade constituents of the periodontal tissues, destroy host defense elements, dysregulate coagulation, complement, and kallikrein-kinin cascades. Thus, they could be responsible for most of the clinical hallmarks of periodontitis (reviewed in Ref. 8). Until recently, proteinases belonging to two catalytic classes and produced by P. gingivalis have been identified. A serine proteinase with prolyl peptidase activity was found to be associated with the bacterial cell surface (9), whereas large amounts of cysteine proteinases with trypsin-like activity were detected in either a soluble form in culture media or combined with whole bacteria or their membranous fragments, including vesicles (10).
Because of convenient assay systems with chromogenic substrates, trypsin-like enzymes were targets for numerous attempts at purification, but only recently has it become apparent that two distinct proteinases are responsible for this activity. One enzyme, described as an Arg-X specific proteinase, was purified from the culture media of P. gingivalis HG66, either as a single chain 50-kDa proteinase (RGP-1) 1 (11) or a high molecular mass (95 kDa) complex of a 50-kDa catalytic domain with hemagglutinins/adhesins (12). Recently, molecular cloning and characterization of its gene revealed that all components of the 95-kDa gingipain-R (HRGP) are created by proteolytic processing of a single polyprotein (13). Similarly, a Lys-X specific proteinase (gingipain K, KGP) was separated as a 105-kDa protein complex composed of a unique catalytic domain (60 kDa) associated with hemagglutinins/adhesins (12). Since the first purified enzyme shared some properties with clostripain, it was named as a gingipain (P. gingivalis clostripain) so that, following recommendations by the IUB, these proteinases are referred to as gingipain-R and gingipain-K to account for their unique specificity.
The construction of isogenic P. gingivalis mutants deficient in the gingipain-R gene has shown unequivocally that these enzymes are pivotal virulence factors (14) and, therefore, a perfect target for the development of specific potentially therapeutic inhibitors. However, before such compounds can be designed and tested as alternative treatments to reduce adult periodontitis, the biochemical characterization of rigorously purified gingipain-R forms is absolutely necessary. In this report, we examine distinct forms of this enzyme that have been found to be products of closely related but discrete genes. In addition, we clarify a controversy regarding: (i) putative collagenolytic activity, (ii) synthetic substrate specificity, (iii) inhibition by plasma proteinase inhibitors, and iv) stimulation of activity by glycyl-glycine.

Methods
Cultivation of Bacteria-The strain of P. gingivalis (HG66) was a gift of Dr. Roland Arnold (University of North Carolina, Chapel Hill). The cells were grown in 200 ml of broth containing 6.0 g of Trypticase Soy broth (Difco), 2.0 g of yeast extract, 1 mg of hemin, 200 mg of cysteine, 20 mg of dithiothreitol, and 0.5 mg of menadione (all from Sigma) anaerobically, at 37°C for 48 h in an atmosphere of 85% N 2 , 10% CO 2 , 5% H 2 . The culture was used to inoculate 5 liters of the same broth, which was then incubated anaerobically, at 37°C for about 48 -60 h until the late stationary phase of bacteria growth (final absorbance at 660 nm Ͼ 2).
Proteinase Purification-The initial steps of gingipain R2 (RGP-2) purification were performed according to the methods designed for HRGP and KGP isolation (12). Briefly, the cell-free culture fluid was precipitated with acetone, and the protein pellet was redissolved in 20 mM Bis-Tris, 150 mM NaCl, 0.02% NaN 3 , pH 6.8, containing 1.5 mM 4,4Ј-dithiopyridine disulfide. The solution was then dialyzed, first against the above buffer (one change), followed by two changes with Bis-Tris/NaCl buffer supplemented with 5 mM CaCl 2 but without 4,4Јdithiopyridine disulfide. The dialyzed fraction was clarified by centrifugation (40,000 ϫ g, 2 h), concentrated by ultrafiltration (Amicon PM-10 membrane), and applied to a Sephadex G-150 column equilibrated with Bis-Tris/NaCl buffer. The column was developed at a flow rate of 30 ml/h with three peaks of activity on BAPNA being found. The highest molecular mass activity peak was used for the purification of HRGP, exactly as described by Pike et al. (12), whereas the low molecular mass peak (50 kDa), having the majority of the activity on BAPNA, was pooled, concentrated, extensively dialyzed against 50 mM Bis-Tris, 1 mM CaCl 2 , pH 6.5, and loaded onto a DE-52 cellulose (Whatman) column (1.5 ϫ 20 cm), equilibrated with Bis-Tris/CaCl 2 buffer at a flow rate of 20 ml/h. The column was washed until the A 280 nm base line fell to zero, followed by application of a gradient of 0 -200 mM NaCl in a total volume of 200 ml. Fractions (5 ml) were assayed for activity on BAPNA. Some activity was found in the void volume of the column, but the major peak was eluted at about 100 mM NaCl. The latter activity was pooled, dialyzed, and applied to an arginine-Sepharose column (1.5 ϫ 30 cm, 50 ml) equilibrated with 50 mM Tris, 1 mM CaCl 2 , pH 7.4, with 0.02% NaN 3 at a flow rate of 20 ml/h. The column was washed with buffer until activity on BAPNA fell below 20 mOD/min/l, and the remaining enzyme then eluted with 0.5 M NaCl. Five pools of activity obtained in this step (Fig. 1), nonadsorbed (A), retarded (B-D), and eluted with NaCl (E) were collected, concentrated, and dialyzed against 25 mM Bis-Tris, pH 6.3. Different isoforms of gingipain R2 were obtained from these pools by chromatofocusing on a mono-P column (Pharmacia, fast protein liquid chromatography system) equilibrated with 25 mM Bis-Tris, pH 6.3, using a pH gradient developed with 50 ml of 10ϫ diluted Polybuffer 74 (Pharmacia) adjusted to a pH of 4.0.
Enzyme Activity Assays-Routinely, amidolytic activities of gingipains R were measured with L-BAPNA (1 mM) in 1.0 ml of 0.2 M Tris-HCl, 0.1 M NaCl, 5 mM CaCl 2 , 10 mM L-cysteine, pH 7.6, at 37°C. After a specific time of incubation, the reaction was stopped by addition of 0.05 ml of glacial acetic acid, and the O.D. at 405 nm then measured against a blank sample containing no proteinase. To determine the effect of glycyl-glycine on RGP activity, buffer (0.1 M Tris-HCl, 5 mM CaCl 2 , 10 mM cysteine, pH 7.6) was supplemented to a desired dipeptide concentration by mixing with 1.0 M glycyl-glycine, 5 mM CaCl 2 , 10 mM cysteine, pH 7.6. To compensate for any change of the ionic strength to the control without glycyl-glycine, appropriate volumes of 0.1 M Tris, 0.2 M NaCl, 5 mM CaCl 2 , 10 mM cysteine, pH 7.6, were added. General proteolytic activity was measured with 1.0% (w/v) azocasein (15), whereas the amount of active enzyme in each batch of gingipain was determined by active site titration using FFRck (16).
Kinetics Studies-The k cat and K m values were measured at 21°C using substrates at concentrations ranging from 0.005 to 2 mM, with a final concentration of active site titrated enzyme of 3.2 nM in 0.1 M Tris-HCl, 5 mM CaCl 2 , 10 mM cysteine, 200 mM glycyl-glycine, pH 7.6. In the absence of glycyl-glycine, 12.5 nM gingipain was used in 0.2 M Tris-HCl, 5 mM CaCl 2 , 10 mM cysteine, 0.1 M NaCl, pH 7.6. The assay was performed in a total volume of 0.2 ml on microplates coated with albumin (20). To 0.05 ml of substrate, 0.15 ml of enzyme solution was added with a multichannel pipette, and the initial turnover rate at 12 different substrate concentrations was recorded at 405 nm using a micro plate reader (Molecular Devices, V max ). The K m and k cat values were calculated using Hyperbolic Regression Analysis, a program written by J. S. Easterby (University of Liverpool, UK) and obtained through shareware. Stimulation of gingipain amidolytic activity on different substrates by glycyl-glycine (stimulation factor) was determined at substrate concentrations at least five times higher than the K m value.
Protein Derivatization, Chemical and Enzymatic Fragmentation, and Peptide Purification-RGP-2 was denatured in 6 M-guanidine HCl, reduced with dithiothreitol, and S-carboxymethylated or S-pyridylethylated using protocols provided by Applied Biosystems. After rapid desalting on a PD-10 column (Pharmacia) equilibrated with 50 mM ammonium bicarbonate, pH 7.8, the modified protein solution was lyophilized. The derivatized protein was subjected to digestion with cyanogen bromide (in 70% w/v formic acid) at room temperature in the dark for 18 h at a 1:1000 molar ratio with respect to methionine. Enzymatic fragmentation was performed with L-1-tosylamido-2-phenylethyl chloromethyl ketone-treated trypsin and Glu-C endopeptidase. Generated peptides were separated by reverse phase high pressure liquid chromatography using a Vydac protein C18-10 column (4 ϫ 30 mm). The peptides were eluted with 0.1% trifluoroacetic acid and acetonitrile containing 0.08% trifluoroacetic acid using a gradient from 0 to 80% acetonitrile over 60 min. Peptides were monitored at 220 nm and collected manually.
For active site cysteine residue determination, RGP-2 was first labeled with biotin-⑀-aminocaproyl-Phe-Pro-Arg-chloromethylketone (Hematologie Technologies, Inc., Essex Junction, VT) and then S-pyridylethylated and subjected to proteolytic fragmentation with trypsin. Biotinylated peptide was purified on avidin-agarose (Sigma) and analyzed for both amino acid composition and sequence as described previously (13).
Sequence and Amino Acid Analysis-Sequence analysis was performed with an Applied Biosystems 477A pulsed liquid phase sequencer with on-line phenylthiohydantoin analysis using an Applied Biosystems 120A high pressure liquid chromatography system operated according to the manufacturer's recommendations. For amino acid analyses, purified RGP-2 isoforms and peptides were hydrolyzed for 24 h at 110°C in 6 N HCl containing 0.1% phenol. The tubes were evacuated and flashed with nitrogen several times before hydrolysates were applied to a Beckman System 6300 high performance amino acid analyzer with a Hewlett Packard 3390A integrator.
Molecular Mass Determination-The molecular masses of the purified isoforms of RGP-2 were estimated by SDS-PAGE using the Tris-HCl/Tricine buffer system (17) and the method of Laemmli (21) on 10 or 12.5% separating gels. Accurate molecular mass measurements were performed on all major isoforms of RGP-2 employing matrix-assisted laser desorption ionization, with mass spectra acquired using a Vestec matrix-assisted laser desorption ionization linear time-of-flight mass spectrometer (Perspective Biosystems).

RESULTS
Enzyme Purification-Gel filtration chromatography of the acetone precipitate obtained from culture fluids yielded a protein and activity profile (against BAPNA and Bz-L-Lys-pNA) identical to that reported previously (12). This time, however, the low molecular mass active fraction containing the majority of the activity against BAPNA was used for further purification. Anion-exchange chromatography of this fraction resulted in the separation of two peaks of activity, one that did not bind to DE-52 cellulose (V o ) and a second major activity that eluted at 100 mM NaCl concentration and contained 95% of the total activity. The activity present in the V o fraction was not due to the column overloading because, during its rechromatography, all activity still passed unretarded through the column. The chromatographic behavior of this fraction resembles that of the 50-kDa gingipain R1 (RGP-1) purified by Chen et al. (11).
The major peak of activity was subjected to affinity chromatography on an arginine-Sepharose column, with three active fractions being obtained ( Fig. 1): nonbound (A), retarded (B-D), and bound (E). Again, the chromatographic behavior of each fraction was checked by rechromatography, and it was shown that regardless of whether proteinases were activated with reducing agent or blocked at the reactive site with tosyl-Llysine chloromethyl ketone, each fraction was eluted from the column in a specific manner, apparently because of differing affinities for arginine residues. All gingipains in the collected fractions had the same molecular mass except pool A, which, besides containing the 48.3-kDa band of the native proteinase, also had a protein band of about 43 kDa (Fig. 1, inset). The N-terminal amino acid sequence was the same for both proteins, suggesting that the 43-kDa band represented a C-terminally truncated enzyme. Curiously, and despite similar or identical apparent molecular masses, all gingipain fractions differed in specific activity (Table I), the most active being in fractions B and C.
Chromatofocusing Separation of Gingipain Isoforms-Each gingipain subfraction was subjected to chromatofocusing, with four enzyme isoforms being purified (I, II, III, IV) (Fig. 2). Pool A contained an enriched source of isoform I, as well as an inactive protein component. In pool B, isoform II was predominant, and in pool C, isoforms II and III prevailed (Fig. 2). In pool D, the proportion of gingipain isoforms was shifted toward isoform III, whereas in pool E, isoform IV predominated, with an additional form eluting from the column at a pH lower than any other form (data not shown).
Molecular Mass, pI, and Zymography Analysis-All isoforms had identical N-terminal sequences as well as the same electrophoretic mobility in SDS-PAGE (Fig. 2, inset A), equivalent to a molecular mass of 48.3 Ϯ 0.4 kDa as determined using laser densitometry scanning of several gels. The masses of 48,019 Da (isoform I-B), 48,031 Da (isoform II-B), 48,138 kDa (isoform I-C), 48,078 Da (isoform II-C), 48,256 Da (isoform III-C), and 48,369 Da (isoform C-IV) were determined by matrix-assisted laser desorption ionization-time-of-flight, and these results are in keeping with those from SDS-PAGE. In addition to differences in molecular mass, the various isoforms also differed in isoelectric points that were determined to be 5.21, 5.09, 5.03, and 4.96 for isoforms from I to IV, respectively ( Fig. 2, inset B). In contrast, the pI of HRGP was found to be 3.74 (data not shown). The difference in pI of RGP-2 isoforms correlated with their mobility in gelatin zymography gels (Fig.  2, inset C) in that forms with the higher pI were more retarded in the gel. From comparison of band patterns in crude and partially purified RGP-2 preparations, it is apparent that isoform II (pI 5.09) is the major form of this gingipain released by P. gingivalis HG66.
Structure of RGP-2-The gene encoding RGP-2 has been cloned and sequenced from two different strains of P. gingivalis (GenBank accession nos. D64081 and U85038). It encodes a polypeptide chain with a 227-amino acid prepropeptide followed by a catalytic domain of 507 residues. The calculated molecular mass of 55 kDa for the catalytic domain was 7 kDa larger than its mass determined by mass spectroscopy and SDS-PAGE, indicating a truncation at the C terminus. The differences in the molecular mass among different RGP-2 isoforms were small (48,148 Ϯ 135 Da), implying that the polypeptide chain of mature RGP-2 was truncated around residue 433 (calculated mass 48, 118 Da) (Fig. 3). The amino acid analysis of the isoforms failed to show any significant differences among them and consistently indicated that the various polypeptide chains were a few amino acid residues shorter than anticipated from molecular mass analysis (data not shown). To clarify this discrepancy, the primary structure of RGP-2 was determined by sequence analysis of CNBr and proteolytic enzyme (trypsin and V8 protease)-generated fragments of C-pyridylethylated protein. Except for a seven-amino acid residue peptide ( 396 DG-KAIIK), all other fragments of RGP-2 could be purified and sequenced, indicating that RGP-2 is composed of 422 amino acid residues with the C-terminal sequence -Gly-Tyr-Asn-Lys-COOH apparently being formed through truncation of the gingipain polypeptide chain by a Lys-specific proteinase (Fig. 3B). Amino acid analysis confirms such a structure (Fig. 3A), although its calculated molecular mass is about 1,000 Da lower than determined from the mass spectroscopy analysis. This discrepancy is most likely because of some amino acid side chain modifications that additionally lead to formation of the RGP-2 isoforms. Apparently, these variations had no effect on enzyme properties, as in screening experiments it was found that the isoforms were indistinguishable with regard to stability, pH optima, kinetic characteristic, and proteolytic activity. Therefore, in experiments described below comparing properties of RGP-2 and HRGP, either fraction B from the arginine-Sepharose chromatography step or purified isoform II was used.
It has been suggested that rgp-1 and rgp-2/rgpB genes were created by domain-specific rearrangement between the two Arg-gingipain-encoding genes in P. gingivalis, possible through nonreciprocal recombination (22). Indeed, the primary structure of RGP-2 is generally identical to the catalytic domain of HRGP, although both polypeptides differ considerably within the C-terminal region (Fig. 3). This identity also applies to the region containing the active site cysteine residue, whose exact location in HRGP has been a matter of controversy (13,23). This prompted us to determine the position of this residue in  2. Chromatofocusing of RGP-2 on mono-P. The concentrated fraction C of RGP-2 from the arginine-Sepharose column, with buffer changed to 20 mM Bis-Tris, pH 6.3, by ultrafiltration, was loaded on a mono-P column equilibrated with the same buffer. The pH gradient was developed with 50 ml of Polybuffer 74 with pH adjusted to 4. Activity peaks eluted at the same pH after chromatofocusing of RGP-2 fractions A, B, and C were considered to be the same isoform of the enzyme, pooled together, concentrated by ultrafiltration with buffer exchange to 20 mM Bis-Tris, 1 mM CaCl 2 , 0.02% NaN 3 , pH 6. RGP-2 by specific biotinylation of the active site residue using biotin-⑀-aminocaproyl-Phe-Pro-Arg-chloromethylketone (24). The labeling was found at Cys-244 (data not shown), in agreement with the observation of Nishikata and Yoshimura (23).
Stability-Both of the RGP-2 isoforms as well as HRGP showed very similar pH stability profiles. In the absence of cysteine, all were found to be stable over a pH range of 4.5-8.5 over several hours; however, in its presence the enzymes only remained stable for similar time periods at pH values between 7.5 and 8.5, losing activity rapidly outside this pH range. In contrast to Ca 2ϩ , which stabilized enzyme activity, glycyl-glycine considerably enhanced the inactivation rate. Gingipains in neutral pH in buffers containing Ca 2ϩ (1-5 mM CaCl 2 ) were stable for months in ice and indefinitely if stored frozen at Ϫ20°C.
Kinetic Characteristics-The optimum pH for the hydrolysis of p-nitroanilide substrates was 8.0 for both RGP-2 and HRGP. In contrast, with protein substrates, such as azocasein, the maximum proteolytic activity for HRGP was at pH 7.5, whereas RGP-2 showed a very broad and flat spectrum of proteolytic activity in the pH range from pH 5.0 (50% activity) to a maximum at pH 9.5 (100%). At a higher pH, proteolytic activity decreased sharply, apparently because of enzyme instability.
Without reducing agents, the activities of purified RGP-2 and HRGP were below 1% of their activity in the presence of 10 mM cysteine, the most effective reducing agent for activation of gingipains. Although low levels of cysteine were able to activate both proteinases, 1 and 5 mM were required for the development of full proteolytic activity for RGP-2 and HRGP, respec-tively. A similar effect was exerted by glutathione and cysteamine, with maximum activity at concentrations higher than 5 mM. In contrast, dithiothreitol and 2-mercaptoethanol were less effective gingipain activators, requiring a high concentration (100 mM) for full HRGP activation (60 -80% of enzyme activity in presence of 10 mM cysteine). Maximum RGP-2 proteolytic activity (80 -90% of cysteine-activated proteinase) was obtained with a 1-5 mM concentration of these reducing agents, with the enzyme being denatured at higher values.
The gingipains amidolytic activity against BAPNA was also strongly dependent on cysteine and cysteamine concentration, with the effect being most profound for activation of HRGP (Fig. 4, A and B). The increase of amidolytic activity of gingipains could also be achieved in the presence of glycyl-glycine, and at constant reducing agent levels it was clearly dependent on the dipeptide concentration. However, gingipains general proteolytic activity was not affected by glycyl-glycine (Fig. 4C).
To further investigate gingipains specificity and the effect of glycyl-glycine on their activities, the K m and k cat values for several commercially available arginine-containing substrates were determined in the presence and absence of this dipeptide (Fig. 5). In the absence of glycyl-glycine, the K m was in the range of 9.5-56.3 M and 13.9 -128 M for HRGP and RGP-2, respectively. In comparison, the K m values for substrates with a P 2 glycine residue (I and IX-XVI) were generally higher for RGP-2. No such correlation was apparent for HRGP. Altogether, the k cat values were slightly higher for RGP-2 than for HRGP. In the presence of glycyl-glycine, both the K m and k cat increased in parallel for the same substrate, without much

FIG. 3. Amino acid composition of RGP-2 (A) and comparison of the primary structure of RGP-2 and HRGP catalytic domain (B).
The amino acid composition of RGP-2 was compared with the composition calculated from the RGP-2 primary structure determined at the protein level (sequence in capital letters in panel B) or inferred from the nucleotide sequence encoding the entire catalytic domain (numbers in brackets). In panel B, the identical amino acids in the HRGP catalytic domain are marked by bullets. A single gap (Ϫ) in HRGP sequence was introduced to improve alignment. An underlined fragment of RGP-2 has not been sequenced, and it is assumed that it is covalently modified either on a Lys or Asp residue. The fragment of 55 residues (434 -489) in the C-terminal region of the initial polypeptide chain of RGP-2 and 40-residue (433-473) fragment of HRGP catalytic domain share no similarity. These fragments were excluded from alignment and are marked with dots. effect on the k cat /K m ratio, which is a measure of enzyme catalytic efficiency. Comparison of the k cat /K m ratio revealed that both RGPs have similar general catalytic potency on the substrates tested, with Z-Arg-Arg-pNA and Leu-Thr-Arg-pNA being, respectively, the least and the most sensitive substrates for gingipains. In general, beside Z-Arg-Arg-pNA, substrates with P 2 proline residue (compounds XIX-XXI) were relatively poor substrates for both gingipains, with the exception of Tos-Gly-Pro-Arg-pNA (XXII).
At substrate concentrations significantly higher than the K m , the increase in gingipains activation by glycyl-glycine (stimulation factor) was very closely correlated to the change of the k cat value (data not shown). In general, the substrate turnover stimulation by glycyl-glycine was about 2-fold higher for HRGP than RGP-2, and for both gingipains it was dependent on peptide structure. For substrates with P 2 hydrophobic residues (Fig. 5, compounds II, V, VI, XVII, and XVIII), the stimulation factor was the lowest. Also, stimulation was relatively low for substrates with hydrophobic or positively charged residues at P 3 . In contrast, a P 3 Glu (compounds XIV, XV, and XX) seemed to correlate with glycyl-glycine stimulation of gingipains amidolytic activity.
Comparison of Gingipain Proteolytic Activity-Using equimolar amounts of HRGP and RGP-2, it was found that the general proteolytic activity, using azocasein or azocoll, of HRGP was five times higher than that of RGP-2, despite the fact that both enzymes were equally active on synthetic substrates (data not shown). Although proteolytic degradation of azocasein and azocoll was not affected by glycyl-glycine, the dipeptide enhanced proteolytic inactivation of ␣ 1 -Achy (Fig. 6). The acceleration was not a result of a glycyl-glycine effect on serpin structure as antichymotrypsin inactivation by Staphylococcus aureus metalloproteinase, trypsin, and KGP was unchanged in the presence of the dipeptide (data not shown). In comparison with general proteolytic activity, antichymotrypsin inactivation occurs through a single peptide bond cleavage at the P 7 Ј residue (-Val-Glu-Thr-Arg * Ile-) within the reactive site loop. The sequence of the cleavage site resembles the structure of substrates whose turnover is enhanced by glycyl-glycine to the highest degree. On the other hand, neither HRGP nor RGP-2 was able to inactivate ␣ 1 -PI, in keeping with the fact that there is no arginine residue within the reactive site loop of this inhibitor. These data, however, contradict a report that a cysteine proteinase from P. gingivalis, apparently identical with one of the RGP forms described in this report, degrades ␣ 1 -PI (25).
P. gingivalis elaborates collagenolytic activity, and this activity has, at least partially, been suggested to be due to gingipain-like arginine-specific cysteine proteinases (25,26). To compare the collagenolytic activity of gingipains described in this report, type I collagen from human placenta was incubated with KGP, HRGP, and RGP-2 for 24 h at 25 and at 37°C. Although P. gingivalis W50 vesicles and culture medium of strain HG66 possessed significant collagenolytic activity (Fig.  7, lanes 2 and 3), no collagen degradation by purified gingipains was observed by SDS-PAGE if the reaction was stopped by addition of FFRck and if the sample was first boiled in nonreducing sample buffer ( Fig. 7; lanes 4 -6, 10, 13, and 18). However, if pretreatment with inhibitor was omitted ( Fig. 7; lanes  12, 15, 16, and 19) or inhibitor treated sample was directly boiled in reducing conditions ( Fig. 7; lanes 11, 14, and 17) extensive degradation of this substrate was observed even without any preincubation of collagen with gingipains. Exactly the same results were obtained using type I collagen from calf and guinea pig skin or rat and kangaroo tail (data not shown). The ineffectiveness of the inhibitor treatment without preboiling in nonreducing conditions is apparently due to the fact that the high concentration of dithiothreitol in sample buffer activates the remaining enzyme, which in latent state is not inhibited by FFRck. Therefore, it is obvious that collagen degradation occurred during the initial period of sample heating, apparently because of its instant unfolding in the presence of SDS, whereas gingipain R remained active for a longer time. In addition, ␣ 1 -PI degradation could only be observed in SDS-PAGE if the sample was not pretreated with FFRck prior to boiling in reducing conditions (data not shown), and this may explain the conflicting results described above (25).
Without proper precaution, gingipains undergo fragmentation during boiling in SDS-reducing agent treatment buffer, with KGP being least susceptible to autodegradation (Fig. 8A). Interestingly, the catalytic domain of HRGP is apparently more susceptible to degradation than any other component of this proteinase-hemagglutinin/adhesin complex, as determined by laser densitometry scanning of the gel (data not shown).
Polypeptide components of the HRGP complex show unusual electrophoretic behavior in different SDS-PAGE systems (Fig.  8B). In gel systems with Tricine as tracing ion, the HRGP complex resolves into four major polypeptides of 49.5 kDa (catalytic domain), 43 kDa, 23.3 kDa, and 15.9 kDa. In the Laemmli system, however, the catalytic domain comigrates with a 43-kDa adhesin domain, as was confirmed by Western blotting using specific antibodies (Fig. 8, C and D). In addition, the lower molecular mass band of HRGP divided into two bands of molecular mass of 17 and 16 kDa. In contrast, the KGP band pattern was unchanged in different electrophoresis systems. DISCUSSION Several genes for arginine-specific proteases of P. gingivalis have been identified, including agp (P. gingivalis 381) (27), rgp-1 (HG66) (13), prtR (W50) (28), prpR1 (W50) (29), and prtH (W58) (30). Despite some significant differences in nucleotide sequence, size, and physical organization, all of these genes are believed to represent a single genetic locus that is either strain variable or has suffered recombination between repeat regions during or prior to cloning (31,32). In addition, the second locus encoding a similar proteinase (rgpB and rgp-2 from ATCC 33277 and HG66, respectively) has been cloned and sequenced  (22,33), and it is now apparent that the arginine-specific activity of P. gingivalis is derived from the products of just two genes (14). Several different forms of such enzymes have already been purified, but due to the lack of structural data, it is impossible to predict from which gene they were derived (for review see Ref. 8). One of the exceptions is the proteinases, referred to as gingipains R, purified from P. gingivalis HG66 (11,12) and their counterparts isolated from W50 (34 -36). It is now unambiguous that both the 50-kDa RGP-1 and high molecular mass forms of the enzyme (HRGP) are derived from the initial translation product of the rgp-1 gene (13). In contrast, the enzyme described in this report (RGP-2) is evidently encoded by a second gene, rgp-2.
For a considerable length of time, our group confused the 50-kDa RGP-1 with RGP-2. After the report on RGP-1 purification and characterization had been published (11), we modified the original procedure to improve enzyme yield; however, by this new procedure we obtained RGP-2, a fact we were not aware of until both rgp-1 and rgp-2 genes were cloned and sequenced. This means that RGP-2, not the 50-kDa RGP-1 originally isolated (11), was used in all work published since 1993.
In contrast to other strains of P. gingivalis where gingipain activity is predominantly membrane associated (10), RGP-2, together with HRGP and KGP, constitutes the majority of soluble protein secreted into the culture media by strain HG66 (Fig. 1, inset). Thus, this makes the strain an excellent source of easily purified soluble proteinases and, as shown after the first step of purification (gel filtration on Sephadex G-150), large amounts (Table I) of nearly homogenous RGP-2 can be obtained. However, when purified in this way, RGP-2 shows considerable heterogeneity in its affinity for arginine-Sepharose, apparently due to the occurrence of discrete isoforms. The RGP-2 isoforms have the same primary structure, and comparison to the amino acid sequence inferred from the nucleotide sequence of the rgp-2 gene revealed that the mature protein is truncated at its C terminus. Apparently, the polypeptide chain  6 and 7) were prepared either by enzyme treatment with 50 M FFRck, boiling in nonreducing sample buffer and then reboiling in 5% dithiothreitol before loading on gel (lanes 2, 4, and 6), or by direct heating in reducing conditions (lanes 3, 5, and 7). B-D, Laemmli 12% gel system. Samples of enzymes were prepared by inhibitor treatment and initial boiling in nonreducing conditions. Lane 9, RGP-2; lane 10, HRGP; lane 11, KGP. C and D, gingipains were Western blotted using chicken antibodies against the N-terminal fragment of gingipains catalytic domains (C) or against the 40-kDa hemagglutinin/adhesin domain of KGP. Before being blocked with albumin, membranes were transiently stained with Ponceau S, and protein bands were equivalent to the gingipains catalytic domain or 44-kDa hemagglutinin/adhesin domain marked. of RGP-2 is proteolytically processed at Lys 422 , indicating involvement of a lysine-specific proteinase in this process. Presently, it is unknown in which cell compartment this processing takes place or if there is any pathophysiological meaning, but it is tempting to speculate that it may occur on bacterial cell surfaces leading to proteinase(s) release into culture medium. Protein shedding from bacterial surfaces by endogenous microbial proteinases has been recently recognized. For example, the Streptococcus pyogenes cysteine proteinase (streptopain) releases several biologically active polypeptides from bacterial cell surfaces, including C5a peptidase, which likely contributes to the pathogenicity and virulence of this bacterium (37).
The polypeptide structures of the RGP-2 isoforms are evidently nearly identical. Therefore, any differences in pI, molecular mass, affinity to arginine, and electrophoretic mobility in zymography gel are likely to be due to variations within a 650 -1000-Da moiety attached to the polypeptide chain. The nature of this modification is presently unknown, but we hypothesize that it may be due to attachment to a part of a lipopolysaccharide moiety because a post-translationally lipidmodified form of RGP, which migrates to the 70 -80-kDa region on SDS-PAGE, has been detected with monoclonal antibodies to lipopolysaccharide (35). This assumption is further supported by the fact that in strains W50, W83, OMGS 100 and ATCC 33277, the 70 -80 kDa form of enzyme is exclusively membrane associated, while being absent in membranous fraction of HG66, indicating its likely release in modified forms into the media (10) in this latter strain. The lipopolysaccharidemodified enzymes have been purified from culture media of P. gingivalis W50 (36) and, with regard to primary structure and gene origin, they seem to correspond to both RGP-2 and the catalytic domains of HRGP from HG66.
The numerous attempts to purify arginine-specific proteinases from different strains and cellular fractions of P. gingivalis in most cases have led to the separation of 43-55-kDa proteins (for review see Ref. 8). Only in the case of a 53-kDa enzyme from W50, referred to as PrpRI, was it shown that it is a noncovalent dimer of a catalytic domain (␣-chain) and the same size ␤-chain (35). The corresponding enzyme purified from HG66, 95-kDa RGP-1 (HRGP), occurs as a mixture of three isoforms in which the 50-kDa catalytic domain is associated with either HGP44, a counterpart of the ␤-chain of PrpRI, or with fragments (HGP15, HGP17, and HGP27) derived from the C-terminal portion of the initial polyprotein encoded by rgp-1 (8,13). Regardless of the gel system, the catalytic domain of HRGP and RGP-2 have the same electrophoretic mobility, and in Tricine-SDS-PAGE the catalytic chain of HRGP is separated from the HGP44 (␤-chain) hemagglutinin/adhesin domain (Fig. 8). In the Laemmli system, however, regardless of the acrylamide concentration, both chains migrate with exactly the same mobility, equivalent to a molecular mass of 49.5 Ϯ 4.5 kDa. Using Western blotting, as described in Fig. 8, it was also determined that in other strains (W50, W83, ATCC 33277, 381, and OMGS 100), the catalytic domain of HRGP (␣-chain) and hemagglutinin/adhesin domain HGP44 (␤-chain) have the same mobility in the SDS-PAGE Laemmli system, although they migrate separately in the Tricine-SDS-PAGE (data not shown). Therefore, it is likely that from those strains a form of Arg-specific gingipain consisting of the catalytic domain and HGP44 was purified, but its composite structure was not perceived due to abnormal electrophoretic behavior. The observed comigration of the catalytic and hemagglutinin domain resolves considerable confusion regarding the hemagglutinin activity of purified arginine-specific cysteine proteinases. Because it was unequivocally shown that neither RGP-2 nor the HRGP catalytic domain (50-kDa RGP-1) have hemagglutinin activity (38), this activity associated with purified proteinases (39) was apparently the result of the heterodimeric structure of the enzyme. Therefore, the term hemagglutining protease (23) should be used with caution, at least until the direct involvement of the proteinase catalytic domain in erythrocyte recognition is verified experimentally.
In many respects, RGP-2 and HRGP show very similar properties including stability, activation by various reducing agents, and pH optimum for the hydrolysis of p-nitroanilide substrates, and this is not surprising when one considers the high degree of identity of their catalytic chains. Also, specificity against synthetic substrates with P1 arginine residue, expressed as the k cat /K m ratio, was similar for both forms of RGP, and no clear preference was observed for particular amino acid residues at the P2 and/or P3 position, as described previously by others (25,26). This contradiction is apparently due to the fact that earlier conclusions on enzyme specificity were suggested after examining the turnover rate of only a limited number of substrates.
One unique property of RGP is the stimulation of their amidolytic activity by amino acids and dipeptides, particularly glycyl-glycine (40). This feature was successfully utilized to detect P. gingivalis in subgingival bacterial plaque by a simple assay of glycyl-glycine stimulated activity against BAPNA in samples of crevicular fluid collected from discrete periodontitis sites (20). In the present work, it was also shown that dipeptide stimulation can be used as a tool to distinguish between RGP-1 and RGP-2 related forms, as activity of the former enzyme against BAPNA was stimulated 18-fold in the presence of 200 mM glycyl-glycine in comparison with RGP-2, whose activity was increased only 7-fold.
The stimulation of RGP activity is due to an increase of the k cat value, but the enzyme catalytic potency measured as the ratio k cat /K m remains unaffected because of a parallel increase in K m (Fig. 5). Apparently, in the presence of glycyl-glycine, nonproductive, high affinity binding of the substrate arginine side chain is eliminated. This explanation is further supported by the observation that the stimulation factor for given substrates strongly depends on its peptide composition. In addition, dipeptides do not have any effect on the general proteolytic activity of RGPs (Fig. 4C) or on the inactivation rate of cathepsin G. 2 In contrast, however, the rate of ␣ 1 -Achy inactivation was enhanced by glycyl-glycine, indicating that for certain macromolecular substrates of gingipains, where a single Arg-Xaa peptide bond is cleaved, dipeptides may have a stimulatory effect as was previously noticed for C3a degradation by RGP-1 (41). The above observations rejuvenate the hypothesis suggested earlier (40) that dipeptide stimulation of RGP activity may represent bacterial adaptation to the natural environment, which is potentially reached in glycine-containing peptides released from degraded collagen fibers.
Despite the fact that HRGP and RGP-2 are equally active on synthetic substrates, HRGP general proteolytic activity is 5-fold higher, and this is particularly noticeable in ␣ 1 -Achy inactivation. In addition, both enzymes also differ considerably in pH optimum for proteolytic activity. Whereas HRGP was most active in slightly alkaline pH, RGP-2 was functional in a broad pH range with a maximum at pH 9.5. This feature may represent an important adaptation of the P. gingivalis proteolytic system to the environment and explain why both genes are maintained in all strains so far examined (36).
One of the clinical hallmarks of periodontitis is degradation of a periodontal ligament, the fibrous attachment of the tooth inserting into cementum of the root and into the alveolar bone.
The main collagen types within the periodontium ligament, and gingival connective tissue as well, are types I, III, and V, with the type I being the most abundant (3). Although degradation of all three types of collagen by proteinases produced by periodontopathogens may significantly contribute to the pathology of the disease, only type I collagen degradation has been extensively studied. For years, P. gingivalis collagenolytic activity was a source of controversy since some authors attributed the ability to degrade type I collagen to trypsin-like proteinases, whereas others found no such activity associated with these enzymes (for review see Ref. 8). More recently, collagenolytic activity was directly linked to purified proteinases that undoubtedly represent one form of RGP (25,26). In this report, we were unable to show degradation of type I collagen even by large amounts of purified gingipains, although collagenolytic activity was found in crude preparations (Fig. 7). However, if a sample of gingipain with collagen was boiled directly in reduced SDS-PAGE sample buffer, without prior inhibitor treatment, total protein degradation was observed. Apparently, during the denaturation process, the proteinase-impregnable structure of collagen was lost faster than gingipain activity, and denatured collagen (gelatin) became a substrate whose degradation was enhanced by SDS (42), temperature, and a highly reducing environment. Other examples of artificial protein degradation by RGPs during boiling in the reducing SDS-PAGE sample buffer include those involving ␣-1-PI, cystatin C, and ␣-2-macroglobulin. The first inhibitor sustains its antielastase activity even after long incubation with large amounts of RGP, but if the sample is not treated with FFRck before SDS-PAGE, the inhibitor is totally degraded, apparently during sample denaturation (data not shown). Similarly, cystatin C in its native conformation is only cleaved at a single peptide bond near the N terminus (Arg 6 -Leu 7 ) by RGP but is totally degraded during boiling in SDS (43). Gingipains, themselves, also undergo extensive autodegradation, as shown during sample preparation for SDS-PAGE (Fig. 8A).
The list of potential substrates of proteinases elaborated by P. gingivalis is long and includes proteinase inhibitors, immunoglobulins, iron transporting/sequestering proteins, coagulation, fibrinolytic, complement, and kallikrein/kinin cascade proteins, proteins of extracellular matrix, and bactericidal proteins and peptides (for references see Ref. 8). Certainly, degradation of some of these proteins occurs in physiological conditions, but others may just be an artifact associated with sample denaturation in a highly reducing condition of SDS-PAGE sample buffer. Therefore, the data on protein degradation by P. gingivalis crude fractions and purified trypsin-like proteinase assessed by SDS-PAGE should be treated with caution.
In summary, from accumulating data it is becoming apparent that the pathophysiological function of gingipains is not just a random degradation of host proteins to provide nutrients for this asaccharolytic bacterium but rather a sophisticated utilization of host systems toward the benefit of the pathogen. In many cases, this purpose can be achieved by limited proteolysis at basic residues of factors involved in blood clotting and fibrinolysis, kinin generation, complement activation, and proteinase-activated receptors and signaling. Indeed, gingipains were found to be able to affect each of these otherwise tightly regulated cascades (41, 44 -48). 3 In addition, gingipains R seem to carry on significant housekeeping functions that are important for bacterial cell physiology and that are also based on limited hydrolysis at strictly selected Arg residues, including proteolytic processing of bacterial precursor proteins (profimbrilin) (49,50), the 75-kDa major outer membrane protein (50), and progingipain K (51), in this way being responsible for fimbriae assembly and hemagglutinin activity expression by P. gingivalis.