Gingipains of Porphyromonas gingivalis Affect the Stability and Function of Serine Protease Inhibitor of Kazal-type 6 (SPINK6), a Tissue Inhibitor of Human Kallikreins*

Periodontitis, a chronic inflammation driven by dysbiotic subgingival bacterial flora, is linked on clinical levels to the development of a number of systemic diseases and to the development of oral and gastric tract tumors. A key pathogen, Porphyromonas gingivalis, secretes gingipains, cysteine proteases implicated as the main factors in the development of periodontitis. Here we hypothesize that gingipains may be linked to systemic pathologies through the deregulation of kallikrein-like proteinase (KLK) family members. KLKs are implicated in cancer development and are clinically utilized as tumor progression markers. In tissues, KLK activity is strictly controlled by a limited number of tissue-specific inhibitors, including SPINK6, an inhibitor of these proteases in skin and oral epithelium. Here we identify gingipains as the only P. gingivalis proteases responsible for SPINK6 degradation. We further show that gingipains, even at low nanomolar concentrations, cleaved SPINK6 in concentration- and time-dependent manner. The proteolysis was accompanied by loss of inhibition against KLK13. We also mapped the cleavage by Arg-specific gingipains to the reactive site loop of the SPINK6 inhibitor. Moreover, we identified a significant fraction of SPINK6-sensitive proteases in healthy saliva and confirmed the ability of gingipains to inactivate SPINK6 under ex vivo conditions. Finally, we demonstrate the double-edge action of gingipains, which, in addition, can activate KLKs because of gingipain K-mediated proteolytic processing of the zymogenic proform of KLK13. Altogether, the results indicate the potential of P. gingivalis to disrupt the control system of KLKs, providing a possible mechanistic link between periodontal disease and tumor development.

The family of kallikrein proteases (KLKs) 3 contains 15 enzymes that are expressed in many different tissues, suggesting a range of physiological and pathophysiological functions. This diversity of functions is exemplified by the involvement of kallikreins in such processes as normal skin desquamation, psoriatic lesion formation, tooth development, neural plasticity, cancer, and Alzheimer disease (for a review, see Ref. 1). Moreover, kallikreins function in wound healing and proteinase-activated receptor signaling, they process antimicrobial peptides in the epithelial areas, and they affect the stability of some growth factors (2,3).
KLKs are secreted as inactive proenzymes (pro-KLKs). With the exception of pro-KLK4, all zymogen forms of pro-KLK are then activated in the extracellular milieu by trypsin-like cleavage of the pro-peptide after either an Arg or Lys residue. The activation cleavage is exerted by the kallikrein itself or by still unknown endogenous proteases and constitutes a key mechanism regulating KLK activity in tissues (4).
Kallikrein-specific inhibitors, including some members of the serine protease inhibitor of Kazal-type (SPINK) family, control the activity of KLKs released into tissues (5). To date, at least 10 genes encoding SPINKs were identified in the human genome, and of them, SPINK5, 6, and 9 were reported to control the activity of different KLKs in different tissues. SPINK5 is detected ubiquitously in epithelial tissues and is composed of 15 protease Kazal-type inhibitory domains that inhibit KLK5 and KLK7. It is well known that mutations in the SPINK5 gene result in elevated activity of epidermal kallikreins and the development of Netherton syndrome, a severe skin disease manifested through abnormal skin desquamation, keratinization, hair follicle defects, and loss of skin barrier function (6). The expression of other SPINK family inhibitors is more tissue-specific. SPINK9 is exclusively found in the palms and soles of the feet (7,8), and it seems to exclusively inhibit KLK5. In contrast, strongly conserved SPINK6 has a broad spectrum of action that controls kallikrein activity (KLK4, KLK5, KLK12, KLK13, and KLK14) not only in the skin (8,9) but also in other tissues. This includes the salivary glands, where SPINK6 inhibits KLK5, 6, and 7, and the main salivary kallikrein, KLK13, which comprises 50% of the overall kallikrein activity in the glands (10). Because the reported activity of KLK13 includes degradation of extracellular matrix proteins (11), it is of no surprise that SPINK6 is cross-linked to fibronectin. This increases the local concentration of the inhibitor and protects the extracellular matrix from KLK-mediated damage (12).
An imbalance between SPINK proteins and proteases may cause severe diseases such as pancreatitis, celiac disease, Netherton syndrome, skin barrier defects, and cancer (7,(13)(14)(15)(16)(17)(18). Similarly, illegitimate, excessive activation of pro-KLKs by bacterial proteases accompanied by degradation/inactivation of SPINKs may result in damage of an infected connective tissue. This is a likely scenario to occur during progression of periodontitis.
Evidence accumulated to date has strongly implicated the contribution of proteolytic enzymes of subgingival plaque bacteria to the pathogenicity of periodontal disease, the most widespread chronic inflammatory condition in the human host. Importantly, in recent years, periodontal disease has been identified as a significant factor in the development of other systemic diseases, including rheumatoid arthritis, cardiovascular disease, aspiration pneumonia, preterm births, and low birth body weight in infants (19).
The key pathogen implicated in the development of chronic periodontitis is an anaerobic bacterium, Porphyromonas gingivalis. P. gingivalis secretes three related cysteine proteases, referred to as gingipains, that constitute its main virulence factors. Two gingipains are specific for Arg-Xaa peptide bonds (HRgpA and RgpB), whereas Kgp cleaves after a Lys residue (20). Collectively, gingipains digest a broad spectrum of host proteins to provide the pathogen with nutrients for growth (21). More importantly, gingipains are involved in the disruption of host defense inflammatory reactions and hinder P. gingivalis clearance by the immune system (22,23). This is accomplished by hijacking proinflammatory signaling pathways via cleavage and activation of the proteinase-activated receptor 2 (PAR-2) on human neutrophils (24). Gingipains also affect many host proteolytic systems, affecting the homeostasis of the organism via complex interactions with host proteins, as exemplified by the activation of the kallikrein/kinin pathway and the resultant increase in vascular permeability (25)(26)(27). However, activation of tissue prokallikreins, especially locally abundant KLK13, with simultaneous degradation of this protease endogenous inhibitor have not been investigated. This is of significant interest because recent discoveries implicate that infection with oral pathogens may promote tumor progression via facilitation of the epithelial-to-mesenchymal transition and an increase in the malignancy of the tumor (28). Therefore, analysis of the interaction network of pathogen proteases and tumor-related proteinases of the KLK family may provide valuable insights into the mechanistic background of oral infectioncancer correlation.
Because SPINK6 is the main inhibitor of salivary kallikreins, especially KLK13, the inactivation of this molecule was our main focus in the context of P. gingivalis proteinase-mediated disruption of the KLK-inhibitor balance. Here we present, to our knowledge, the first report of the interplay of the tissue kallikrein system with P. gingivalis proteases, further adding to our understanding of the mechanisms of tissue disruption by this widespread oral pathogen.

Results
SPINK6 Is Degraded by P. gingivalis Cultures-P. gingivalis produces many proteolytic enzymes, and, in addition to gingipains, it secretes several other proteases that may interact with SPINK6. Therefore, to investigate the ability of gingipains and other proteases to degrade SPINK6, the inhibitor was preincubated with wild-type P. gingivalis W83 in the presence or absence of gingipain-specific inhibitors or with the triple gingipain knockout mutant W83 ⌬kgp⌬rgpA⌬rgpB , completely devoid of gingipain activity (Fig. 1). The wild-type W83 strain completely degraded SPINK6 over the course of the experiment, as indicated by the absence of SPINK6 immunoreactivity on the Western blot of the samples. Preincubation of the W83 strain with Kyt inhibitors that are gingipain-specific, reversible, and do not affect the activity of other proteases, at least at the concentration used (29), protected the majority of SPINK6 from degradation ( Fig. 1). Some cleavage of SPINK6 was observed, apparently due to residual activity of the gingipains in the presence of Kyt inhibitors, which were used at low micromolar concentration (5 M). Exclusive gingipain involvement in degradation was confirmed by the finding that SPINK6 incubation with the triple gingipain mutant W83 ⌬kgp⌬rgpA⌬rgpB did not affect the stability of the inhibitor. Moreover, when SPINK6 was incubated with another cysteine proteinase of P. gingivalis, the streptopainlike TPR, no degradation was observed, even in the presence of a 100 nM concentration of the protease (data not shown). Taken together, these results indicate that wild-type bacteria efficiently degrade SPINK6 and that three gingipains (Kgp, RgpB, and HRgpA) are the enzymes responsible for this activity.
Gingipains Degrade SPINK6 in a Time-and Concentrationdependent Manner-P. gingivalis gingipains degraded SPINK6 in a concentration-dependent (Fig. 2, A-F) and time-dependent (Fig. 2, G-L) dependent manner. The Lys-specific gingipain Kgp (20 nM) degraded SPINK6 nearly completely, as visualized by silver-staining SDS-PAGE gels. By contrast, SPINK6 appeared intact on non-reducing gels following incubation with Arg-specific gingipains, even at relatively high enzyme concentrations (10 nM RgpB and 50 nM HRgpA). However, SDS-PAGE under reducing conditions revealed the accumulation of the lower molecular weight proteolysis product during incubation with 2 nM RgpB and 10 nM HRgpA.
Similarly, the time course analysis showed progressive degradation by Kgp (50 nM) with no significant differences between reducing and non-reducing conditions. After 45 min, only traces of the native inhibitor could be detected by SDS-PAGE. Incubation with the Arg-specific gingipains led to the accumulation of the nicked inhibitor, represented by two polypeptide chains resolved only by reducing SDS-PAGE. For both RgpB (2 nM) and HRgpA (20 nM), a 30-min incubation was long enough for almost complete cleavage of the inhibitor. Cumulatively, these results are congruent with the gingipain specificities and the SPINK6 structure, which contains six Lys residues scattered throughout the polypeptide chain and only a single Arg residue at the P1 position in the reactive site loop of the inhibitor stabilized by three disulfide bridges.
Gingipain-treated SPINK6 Is Inactive as a KLK13 Inhibitor-The reactive site loop of SPINK6 is stabilized by three disulfide bridges that hold in close proximity the P1 Arg and P1Ј Glu residues. Because Kazal inhibitors nicked at the P1-P1Ј bond may retain inhibitory activity (30), we tested the ability of gingipain-treated SPINK6 to inhibit KLK13 (Fig. 3). The results indicated that the apparently stable, Rgps-nicked SPINK6 displayed no inhibitory activity when incubated in a 2:1 molar ratio with KLK13, in contrast to the intact inhibitor, which diminished KLK13 activity by ϳ90%.
Limited Proteolysis of SPINK6 by Gingipains-To further characterize SPINK6 interactions with gingipains, the cleavage sites within the inhibitor were identified. As suggested by SDS-PAGE analysis, reverse-phase HPLC chromatography of Kgptreated SPINK6 (Fig. 4A) confirmed the complete fragmentation of the inhibitor. Six discrete fragments of the SPINK6 polypeptide chain generated by Kgp were identified, resulting from the four Lys-Xaa peptide bond cleavages as identified by MS/MS analysis (Fig. 4G). As expected, none of the degradation products showed any inhibitory activity toward the KLK13 protease (Fig. 4D).
The pattern of degradation products for the Arg-specific HRgpA and RgpB differed significantly from that of Kgp. Five main peaks were identified (Fig. 4, B and C), and none of them contained the unmodified inhibitor. All of the products were consistent with a single cleavage after Arg 42 . Importantly, the SPINK6 structure is stabilized by three disulfide bridges that keep the polypeptide fragments together after cleavage with gingipains. However, 2 mM L-cysteine required for gingipains activity reduced the disulfide bridges in the SPINK6 molecule already destabilized by a proteolytic cleavage. In the case of multiple cleavages of the inhibitor polypeptide chain by Kgp, there were no intact bridges left to keep the fragments together. Conversely, a single cleavage at Arg 42 by Rgps generated predominantly the double-chain, full-length molecule, as demonstrated by the ϩ18-Da mass shift compared with the native unmodified inhibitor. However, in an unstable two-chain structure of SPINK6, disulfide bridges holding the molecule together must be partially reduced, yielding N-(SPINK6 1-42 ) and C-terminal (SPINK6 43-80 ) fragments. It can be speculated that the rearrangement of the disulfide bridges in these fragments during sample preparation for HPLC analysis led to two different molecular conformations of identical mass (MS/MS data) but differed in the retention times on HPLC (Fig. 4, H and I). Nonetheless, the fragmentation of the inhibitor chain seems to be required, as the intact inhibitor was unaffected by 2 mM L-cysteine used in the assay and eluted as a single peak, retaining the full inhibitory activity.
To test for residual activity among the cleavage products, the inhibitory potential of all of the Kgp-, RgpB-, and HRgpA-derived products was evaluated (Fig. 4, D-F). Peak areas were used to calculate the amount of peptide in each peak. Reactions were carried out under more rigorous conditions than previous assays, using a 10:1 molar ratio of the given inhibitor fragment to KLK13 (compared with the 2:1 molar ratio in the experiments above). The N-terminal SPINK6 1-42 and the C-terminal SPINK6 43-80 fragments were completely inactive as inhibitors, whereas the full-length, ϩ18-Da, single hydrolysis-modified peptide displayed 50% inhibitory activity of the control SPINK6 when incubated with KLK13 under the same reaction conditions. This result indicates that this product is a relatively weak KLK13 inhibitor, as a 10-fold molar excess did not fully inhibit the target enzyme compared with native SPINK6, which inhibited 90% of the protease activity at a 2:1 molar ratio in the experiments above.
SPINK6 Acts as a Potentially Important Inhibitor in Human Saliva-Human saliva displays a significant, trypsin-like proteolytic activity. As SPINK6 remains cross-linked to the ECM and provides localized inhibitory protection, the fraction of saliva activity susceptible to the SPINK6 control was estimated. Hence, a saliva mixture from five healthy individuals was titrated with recombinant SPINK6 using the Boc-Val-Pro-Arg-AMC fluorogenic substrate. Saliva (5 l) was diluted to a final volume of 200 l (40-fold dilution), and proteolytic activity was measured in the presence of 0 -100 nM SPINK6 (Fig. 5B). In addition, the susceptibility of the saliva of each individual was tested separately (Fig. 5A). The 50 nM SPINK6 was able to inhibit ϳ60% of trypsin-like activity in each of the individual saliva samples. Moreover, the Arg-specific proteinase activity in the saliva mixture decreased in a SPINK6 concentration-dependent manner until 75% reduction. Nonetheless, ϳ25% of the trypsin-like activity in the tested sample remained unaffected by SPINK6. The overall SPINK6-sensitive activity was determined as ϳ1 M in the undiluted saliva by SPINK6 active site titration (data not shown). This result indicates that inactivation of SPINK6 could render a major fraction of protease activity uncontrolled, with the potential to disrupt signaling pathways and damage tissues.
Kgp Activates pro-KLK13-The majority of the kallikrein family proteases contain an Arg or Lys residue at the profragment cleavage site (31) and, therefore, may be susceptible to activation by gingipains. Because KLK13 contains a Lys residue at the activation site, the ability of Kgp, a Lys-specific gingipain, to generate the mature form of the protease was tested (Fig. 6). Increasing concentrations of Kgp were incubated with pro-KLK13, and KLK13 activity was measured using the Boc-Val-Pro-Arg-AMC substrate (Fig. 6A), which is not hydrolyzed by Kgp. To ensure that no HRgpA/RgpB contamination affected the results, the assay was performed in the presence of the Rgps inhibitor Kyt-1. As shown in Fig. 7A, Kgp activated proKLK13 in a concentration-dependent manner, with a maximal effect at 5 nM Kgp. Further increases in Kgp concentration (50 and 100 nM) resulted in a gradual loss of activity that could be the result of degradation of the mature KLK13 by high concentrations of gingipains. Nonetheless, the activity was fairly stable, leaving ϳ80% of maximal KLK13 activity even at 100 nM Kgp. Similarly, the active site labeling by FP-biotin indicated the proper binding of the biotinylated inhibitor, as presented in Fig. 6B. Increasing concentrations of Kgp resulted in the appearance of the band, corresponding to the active KLK13. Already at 0.1 nM Kgp concentration, binding of the FP-biotin was visible, and with further increases up to 30 nM Kgp, the band was saturated. Finally, the activation of KLK13 by Kgp was confirmed by the removal of the N-terminal His tag attached to the KLK13 profragment (Fig. 6C). Similarly, the Western blotting analysis using anti-His tag antibody revealed a decrease in the signal with increasing Kgp concentration, resulting in near-complete disappearance of the respective band at 50 nM Kgp concentration. This indicates that increasing concentrations of Kgp indeed facilitate the removal of the N-terminal profragment in the nanomolar range of concentrations and, coupled with the activity and FP-biotin labeling results, confirm the ability of Kgp to activate KLK13 and to release a functional protease. This indicates that low nanomolar levels of Kgp activate proKLK13, disrupting the protease-inhibitor balance already affected by SPINK6 proteolytic inactivation by gingipains. Gingipains Degrade SPINK6 in the Presence of Saliva-Finally, the in vivo environment contains not only saliva but numerous other proteins, possibly acting as competing substrates for gingipains. Therefore, the degradation of SPINK6 was analyzed in the presence of 25% saliva (a mixture from five healthy volunteers). SPINK6 (10 g) was incubated for 1 h with 50 nM gingipains in the presence of the 25% saliva mixture diluted in reaction buffer. Samples were analyzed by Western blotting with anti-His tag antibodies. Additionally, the residual inhibitory activity of the degradation mixture was verified using the KLK13 inhibition assay (Fig. 7A). Western blotting results (Fig. 7B) confirmed the superior activity of the Arg-specific HRgpA and RgpB, which were able to completely degrade 10 g of SPINK6 in the presence of 25% saliva. Similarly, the presence of HRgpA or RgpB blocked inhibition by SPINK6 because Ͼ60% of KLK13 activity in saliva was restored. Kgp again was proven to be the least effective, with residual levels of SPINK6 detected by Western blotting and only 15% of KLK13 activity restored. These data indicate that Arg-specific gingipains may specifically target SPINK6 in the complex protein mixture, whereas Kgp degradation could be regarded as a rather nonspecific attack on the exposed Lys residues, limited by the presence of other saliva-derived proteins.

Discussion
Our results indeed confirm the ability of P. gingivalis to affect the kallikrein-inhibitor balance, adding to the well known pathological interactions of this bacterium with the protease control system of the host. Biochemical and mutant strain analyses using healthy saliva mixtures in vitro showed that gingipains were the sole proteases responsible for this activity. Gingipain concentrations in the nanomolar range were effective, and a combination of gingipains was found to be most efficient. Moreover, the Lys-specific Kgp not only degraded the inhibitor but was also capable of activating the proform of KLK13, as opposed to Arg-specific HRgpA and RgpB.
This important observation indicates that gingipains indeed possess the potential to activate the main kallikrein proteinase, KLK13, present in human saliva (32). Additionally, our results confirm the ability of gingipains (both Lys-and Arg-specific) to activate other members of KLK family of proteases, an obser- FIGURE 5. The majority of trypsin-like proteinase activity in healthy saliva is sensitive to SPINK6 inhibition. A, saliva of each healthy individual (5 l, n ϭ 5) was diluted 40ϫ in reaction buffer, and 50 nM SPINK6 was added to each sample. Activity was measured using the fluorescent trypsin substrate Boc-Val-Pro-Arg-AMC. The results were normalized, and activity in the SPINK6-untreated sample was set as the 100% control (ctrl) for each individual. Significantly different means were identified using Mann-Whitney test. B, the saliva mixture from healthy individuals (5 l) was diluted 40ϫ in reaction buffer and incubated with increasing concentrations of recombinant SPINK6. Residual trypsin-like protease activity was measured using the Boc-Val-Pro-Arg-AMC substrate and was plotted as percent of control sample activity (no inhibitor added). Data are presented as mean Ϯ S.D. vation pursued in our future work. 4 Moreover, the complex protein mixture present in saliva did not block the inactivation of SPINK6 by Arg-specific gingipains, indicating a targeted activity of these pathogen-derived proteinases.
Additionally, our results indicate a substantial level of salivary trypsin-like activity, of which 75% is susceptible to SPINK6-mediated inhibition. This allows the estimation of the total KLK activity in saliva approximating 1 M, an amount left uncontrolled after the SPINK6 inactivation. Taken together, these data indicate that the kallikrein-inhibitor balance is one of the targets of P. gingivalis proteases in producing inflammation.
Gingipains are well established virulence factors of P. gingivalis and are often described as an archetype of protease involvement in bacterial virulence. These aggressive enzymes, with strict P1 specificity toward Arg (HRgpA and RgpB) or Lys (Kgp) residues, target a number of essential components of human immunity, coagulation cascade, and regulatory pathways (33). Their impact on host protease-inhibitor systems has been reported previously, as gingipains inactivate the ␣ 1 -prote-ase inhibitor (34), secretory leukocyte peptidase inhibitor (35), and, as reported previously by our group, elafin (36). The common feature of the aforementioned inhibitors is that all three, working in concert, control the proteolytic activity released from neutrophils and, thus, prevent damage of connective tissue during inflammation. This report provides evidence that gingipains can efficiently degrade SPINK6, a Kazal-type inhibitor of the tissue kallikrein family of proteinases. To our knowledge, this is the first example of the gingipain-mediated disruption of a system that controls the activity of these tissue-specific enzymes.
All three gingipains are capable of inactivation of SPINK6. As the sequence of SPINK6 (Fig. 8) contains six Lys residues, it is not surprising that Lys-specific gingipain K was able to fragment the protein into small peptides unable to inhibit the target enzyme. More importantly, however, cleavage by Arg-specific HRgpA and RgpB also led to the loss of inhibitor activity. SPINK6 contains only a single Arg residue, located in the P1 position of the reactive site loop. Moreover, the disulfide bridges in the polypeptide stabilize the inhibitor against dissociating into separate fragments (Fig. 8). Several reports indicate that these inhibitors can maintain their function after cleavage of the P1-P1Ј peptide bond, possibly with lowered efficiency (37). Indeed, as mass spectrometry analysis confirmed that the main product of HRgpA or RgpB degradation of SPINK6 was the nicked inhibitor with the reactive site loop held together by disulfide bridges, and this polypeptide retained partial inhibitory activity. Two other peptides were identified, fragments SPINK6 1-42 and SPINK6  , both with the oxidized disulfide bridges, and were inactive as proteinase inhibitors. Each fragment eluted from an HPLC column with two distinct retention times and with the oxidized ("bound") disulfide bridge state. Therefore, as the MS/MS sequence analysis confirmed the identity of these peptides, we hypothesize that the distinct elution times for each of the peptide fragments correspond to the different disulfide bridge configurations within the given fragment.
As cysteine proteinases, gingipains are dependent on reducing agents in the reaction buffer, and, thus, it is possible that reorganization of the disulfide bridges is the result of the initial reduction of the protein. Nevertheless, our data confirm full activity of SPINK6 under optimized experimental conditions with buffer containing 2 mM L-cysteine (data not shown). Moreover, all control samples in the experiments were incubated in reaction buffer and, as confirmed by HPLC and MS, maintained full activity with no changes in the structure or disulfide bridges. Therefore, we conclude that reorganization of the disulfide bridge pattern is most likely a result of internal instability in the cleaved inhibitor. This model is in agreement with the standard mechanism of inhibition where a target protease forms a pocket tightly binding the inhibitor and preventing its free dissociation, hence stabilizing the cleaved form in the binding cleft. By contrast, hydrolysis by gingipain R may allow release of the internal tension of the molecule by disulfide bridge reorganization. As mentioned above, this hypothesis is supported by the detected state of disulfide bridges in the isolated peptides that, despite 2 mM L-cysteine in the reaction buffer, were all oxidized. FIGURE 7. Gingipains affect SPINK6 integrity and function in the presence of saliva. SPINK6 (10 g) was incubated with a 25% saliva mixture from five healthy donors in the presence of the indicated gingipain (50 nM). After 1 h, the reaction was stopped by addition of gingipain-specific inhibitors and analyzed. A, the ability of the reaction products to inhibit KLK13 was evaluated in 25% saliva in the presence of gingipain inhibitors. A 2-fold molar excess of the SPINK6 reaction products was added to KLK13 solution, and activity was measured using fluorescent monitoring of the Boc-VPR-AMC substrate. B, Western blotting analysis with anti-His tag antibodies shows that HRgpA and RgpB fully degraded SPINK6. Kgp was less active, leaving a small fraction of the inhibitor intact. Ctrl, control. Statistical significance: *, p Ͻ 0.01; **, p Ͻ 0.05.
The inactivation of SPINK6, a potent, tightly binding inhibitor of kallikrein family proteinases (10), is an important observation. SPINK6 is present in epithelial tissues, including skin and gingiva, in two forms: fully soluble and transglutaminated by its N-terminal glutamine to extracellular matrix proteins, like fibronectin (data not shown and Ref. 12). This diversification of the inhibitor forms provides protection for the extracellular matrix, locally increasing the inhibitory capacity because of covalent linkage of the active inhibitor to the ECM proteins. As suggested for other tightly binding tissue inhibitors of proteases (e.g. elafin (38)), the protective system is constituted by the initial response molecule, a tightly binding, locally produced inhibitor, which later transfers the protease molecule onto the final, serum-derived, irreversible inhibitor (either serpin or ␣ 2 -macroglobulin). Hence, the tightly binding inhibitor provides the necessary quick response and protects essential components of the tissue. The fact that gingipains not only inactivate SPINK6, as presented here, but are also capable of direct degradation of ECM proteins (39) likens them to an enzymatic arsenal that is capable of both inactivation of host protease inhibitors and degradation of the ECM matrix itself.
Tissue kallikreins are a family of 15 extracellular serine proteases with trypsin-and chymotrypsin-like specificity. Their functions are not well defined, but recent studies indicate that kallikreins 5, 7, and 14 regulate epithelial desquamation (40); KLK3, 7, and 13 may activate plasminogen; and KLK2, KLK3, KLK6, KLK7, and KLK14 may degrade ECM proteins (as reviewed in Refs. 41 and 42). Moreover, KLK5 and KLK7 are enzymes that process both precursor and mature forms of antimicrobial cathelicidine hCAP18 in the skin (3), and KLK5 and KLK14 regulate inflammation through activation of proteinase-activated receptors (43). Furthermore, our results implicate KLK13, one of the main salivary kallikreins, as an important enzyme in the regulation of epithelial wound healing, most likely via processing of the growth factor precursors. 4 Additionally, kallikreins have been implicated in numerous reports as important factors and biomarkers in the development and progression of tumors (as reviewed in Ref. 44).
Interestingly, the development of periodontitis and bacterial infections of the oral cavity have been linked to an increased risk of oral cancer (45,46). More recent reports further strengthen the correlation between P. gingivalis infection and oral and stomach cancer development in vitro and in vivo (28,47). This effect is usually attributed to the direct effect of bacterial toxins and virulence factors promoting the development of chronic inflammation. Our results propose a novel explanation of the periodontitis-cancer correlation. The inactivation of a kallikrein-specific inhibitor may be beneficial for bacteria, as it affects inflammation, tissue reconstruction, and antimicrobial-peptide processing; however, the long-term increase in kallikrein activity could eventually lead to the development of oral cancer via activation of growth factors, facilitation of chronic inflammation, and promotion of metastasis by ECM matrix remodeling.
Our report provides evidence for the involvement of P. gingivalis gingipains in destabilization of the regulation of KLK activity through inactivation of SPINK6, the epithelium-specific inhibitor of kallikreins. This result not only adds to the known tendency of P. gingivalis to disrupt the proteolytic balance in infected tissues but also provides a possible mechanistic link between the development of periodontitis and oral cancer progression. As the role of the salivary kallikreins in the development of oral cancer is not known, the reported data provide a basis for further investigation of P. gingivalis infection and kallikreins in carcinogenesis and tumor metastasis in the oral cavity.

Experimental Procedures
The Pichia pastoris expression system was obtained from Invitrogen. Gingipains and proKLK13 were purified as described previously (48,49). As HRgpA and Kgp were purified from whole culture medium, cross-contamination was possible. Therefore, the Arg gingipain-specific (Kyt-1) and Lys gingipain-specific (Kyt-36) (29) small molecule, tightly binding inhibitors were used in all gingipain experiments to ensure inhibition of undesired cross-reactivity in the samples. All reagents were obtained from Sigma-Aldrich (Taufkirchen, Germany) unless otherwise indicated.
Expression and Purification of Recombinant SPINK6 -A DNA sequence encoding SPINK6 (GenBank accession no. GQ504704.1) was amplified from keratinocyte cDNAs in three consecutive PCR reactions that introduced an EcoRI cleavage site and histidine tag-and enterokinase cleavage site-coding sequence at the 5Ј end and a NotI recognition site at the 3Ј end. The final amplicon was ligated into the pPIC9 vector following the ␣-factor secretion sequence and used to transform Escherichia coli strain Dh5␣. Positive clones were sequenced to verify the sequence, and the plasmid was linearized with SalI (Thermo Scientific) and electrotransformed into P. pastoris strain GS115 to obtain the Mut ϩ phenotype. Following addition of 1 M sor-  (24 -1) , Lys 37 , Lys 58 , and Lys 71 (numbering of the full-length SPINK6 according to the Uniprot database). The hydrolysis after Lys (24 -1) is located at the enterokinase cleavage site in the N-terminal extension of the protein (gray). All other Kgp recognition sites are within the sequence of the native SPINK6 protein. HRgpA and RgpB hydrolyze the single peptide bond on the C-terminal side of Arg 42 . An additional unassigned hydrolysis site at Gln 53 is marked (asterisk) and may be an artifact of the MS analysis. The disulfide bond scaffold binds N-terminal and C-terminal regions of the molecule, hence dissociation of the SPINK6 1-42 and SPINK6 43-72 requires reorganization of the disulfide bridges.
bitol, cells were plated on MD plates (1.34% yeast nitrogen base, 0.00004% biotin, and 2% glucose) and grown for 48 h at 30°C. The Mut ϩ phenotype was verified by the ability to grow in the presence of methanol. Positive recombinants were chosen for expression cultures. BMGY medium (1% yeast extract, 2% peptone, 1.34% yeast nitrogen base, 1% ammonium sulfate, 0.0004% biotin, 100 nM potassium phosphate buffer (pH 6.0), and 1% glycerol) was inoculated with a single colony and incubated at 30°C for ϳ16 h with shaking (250 rpm), and, after reaching A 600 ϭ 4.0, the cells were harvested by centrifugation (5 min, 3000 RCF) and resuspended in BMMY medium (1% yeast extract, 2% peptone, 1.34% yeast nitrogen base, 1% ammonium sulfate, 0.0004% biotin, 100 nM potassium phosphate buffer (pH 6.0), and 0.5% methanol) at A 600 ϭ 1.0 to induce expression. Cultures were incubated for 72 h at 30°C with shaking at 250 rpm for protein production. Every 24 h, aliquots of 100% methanol were added to maintain 0.5% final concentration. Cells were then harvested by centrifugation (15 min, 3000 RCF), and the supernatant was collected, filtered, and mixed with 50 mM sodium phosphate buffer (pH 7.5), 2.5 M NaCl, and 25 mM imidazole at a ratio of 4:1 (supernatant:buffer, v:v). Finally, equilibrated nickel-Sepharose resin (GE Healthcare) was added and incubated with the supernatant for 16 h at 4°C with gentle agitation. Histidine-tagged protein was eluted with increasing concentrations of imidazole. The fraction of interest was then acidified by adding TFA to a final concentration of 0.1%. The sample was loaded onto an equilibrated, semipreparative HPLC column (Jupiter C18, 5 m, 10/100 mm, Phenomenex). Protein was eluted using a 25-35% gradient of phase B (0.08% TFA, 80% acetonitrile) in 17 column volumes using an ÄktaMicro chromatography unit (GE Healthcare). Fractions containing the inhibitor were collected and dried using a CentriVap concentrator (Labconco) to remove the solvent.
Degradation of SPINK6 by Bacterial Cultures-P. gingivalis wild-type W83 and W83 ⌬kgp⌬rgpA⌬rgpB strains were cultured overnight in Schaedler medium supplemented with hemin (1 mg/ml), vitamin K (0.5 mg/ml), and cysteine (0.05 mg/ml) under anaerobic conditions (90% N 2 , 5% CO 2 , 5% H 2 ) at 37°C. Overnight cultures were diluted into fresh medium to A 600 ϭ 0.1 and cultured as above until reaching A 600 ϭ 1.0. Subsequently, 200 ng of SPINK6 was added to 1 ϫ 10 6 bacteria in culture medium and incubated under anaerobic conditions for 1 h at 37°C. The reaction was terminated by addition of SDS-PAGE sample buffer containing DTT and boiling for 5 min. The products of the reaction were separated by SDS-PAGE, and proteins were electrotransferred onto a PVDF membrane (Immobilon-P, Millipore). The membrane was blocked with 3% skim milk in TTBS buffer (50 mM Tris-HCl (pH 7.5), 500 mM NaCl, and 0.05% Tween 20) for 1 h at 37°C. The blocked membrane was incubated overnight at 4°C with primary antibody (1:1000, goat anti-human SPINK6, purified as described before (8)) and then washed four times with TTBS and incubated with a secondary antibody (HRP-conjugated anti-goat IgG (1:30,000, catalog no. A8919, Sigma) for 1 h at room temperature. Pierce ECL (Thermo Fisher Scientific) was used as the chemiluminescence substrate, and images were developed using Medical X-Ray-Film Blue (Agfa).
37°C. The reaction was stopped by acidification of the samples with TFA (0.1% final concentration (v/v)). Samples (100 l) were loaded onto a column equilibrated with 0.1% TFA (AERIS C18 peptide, 3.6 m, 4.6/150 mm, Phenomenex). Elution was performed using a 10 -80% gradient of phase B (0.08% TFA, 80% acetonitrile) in 19 column volumes. Collected fractions were dried using a CentriVap concentrator (Labconco). Each fraction was resuspended in TNET buffer to a final concentration of 1 M based on the peak integration data. The sample (20 nM final concentration) was then mixed with KLK13 (2 nM final concentration) and incubated in TNET buffer for 15 min at 37°C (100-l total volume). Subsequently, 40 M substrate was added (100 l), and fluorescence intensity was measured as described above. The final mixture contained 10 nM SPINK6 or its degradation products, 1 nM KLK13 (10:1 molar ratio), and 20 M substrate in a final volume of 200 l.
MS Analysis of the Degradation Products-Peptide fractions were dried by evaporation, redissolved in 30% methanol supplemented with 0.1% (v/v) formic acid, and analyzed using an HTC Ultra ETD II mass spectrometer (Bruker) equipped with an electrospray ionization source and an ion trap analyzer. The samples were directly injected using a syringe pump (KD Scientific) at a flow rate of 180 l/h and analyzed in positive ion mode with a capillary voltage of 3.5 kV, a nebulizer pressure of 10 psi, a drying gas flow of 5 liters/min, and an ion source temperature of 300°C. Spectra were acquired in MS and MS/MS modes in the range of 100 -3000 m/z with three different fragmentation types: collisioninduced dissociation, collision-induced dissociation/electron transfer dissociation, or electron transfer dissociation/proton transfer reaction. MS data were processed manually using the DataAnalysis TM 4.0 and Biotools TM 3.2 software.
Activation of pro-KLK13 with Kgp-Kgp was activated in TNCT buffer supplemented with 10 mM L-cysteine for 15 min at 37°C. A sample of 200 nM pro-KLK13 (expressed in P. pastoris) was incubated with 0, 0.5, 1, 5, 10, 50, or 100 nM Kgp in TNCT containing 2 mM L-cysteine and 5 M Kyt-1 for 1 h at 37°C. The reactions were terminated by addition of Kyt-36 (5 M) to inhibit Kgp. After 15-min incubation, 40 M N-t-Boc-Val-Pro-Arg-AMC substrate was added, and fluorescence intensity was recorded as described above. The final mixture contained 100 nM pro-KLK13/KLK13, 1 nM Kgp, 1 M Kyt-1, 1 M Kyt-36, and 20 M substrate in TNET buffer (200-l total volume). Simultaneously, to verify the formation of the functional active site of KLK13 and to monitor removal of the N-terminal His tag, a separate reaction was performed. Pro-KLK13 (2 g, 3.5 M) was incubated with 0, 0.1, 5, 10, 20, 40, or 50 nM of activated Kgp in TNCT buffer containing 2 mM L-cysteine for 1 h at 37°C. The reactions were terminated as described above. These samples were divided, and the first set of samples (0.5 g of proKLK13) was incubated with 1 M FP-biotin (Santa Cruz Biotechnology Inc.) for 30 min at 37°C. The second set containing 1 g of proKLK13 was used for an anti-His tag Western blot. Both sets of samples were separated by SDS-PAGE and electrotransferred onto a PVDF membrane (Immobilon-P, Millipore). Nonspecific binding sites were blocked with 5% BSA in TTBS. The FP-biotin-labeled samples were incubated with HRPconjugated streptavidin (Sigma-Aldrich) diluted 1:25,000 in TTBS. Similarly, the second set of samples was developed using a monoclonal, HRP-conjugated anti-His tag antibody (catalog no. A7058, Sigma-Aldrich) diluted 1:20,000 in TTBS. After incubation, the antibodies were washed, and bands were visualized as described above.
Titration of Trypsin-like Activity in the Saliva-Saliva collected from five healthy individuals was pooled. The mixture was centrifuged for 30 min at 12,000 ϫ g, and 5 l of supernatant was diluted 10-fold in 45 l of TNET buffer on a microtiter plate. To wells containing diluted saliva, 50 l of solution containing 0, 2,8,20,40,80,200, or 400 nM SPINK6 was added. After a 15-min preincubation, 100 l of 40 M N-t-Boc-Val-Pro-Arg-AMC substrate was added, and residual trypsin-like activity was measured. In addition, separate saliva samples from healthy donors were incubated with 50 nM SPINK6 and analyzed, as above, to provide information about individual variations in SPINK6-sensitive protease content.
Degradation of SPINK6 in the Presence of Saliva-Saliva was clarified by centrifugation as described above and diluted 2-fold with sterile PBS. Gingipains were preactivated in TNCT buffer supplemented with 10 mM reduced L-cysteine (Sigma-Aldrich) for 15 min at 37°C. To eliminate potential cross-contamination of Rgp with Kgp and vice versa, Kyt inhibitors (2 M final concentration) were added to eliminate any contaminating activity: Kyt-1 to the Kgp sample and Kyt-36 to the HRgpA and RgpB samples. SPINK6 (10 g) was incubated with each gingipain (50 nM final concentration) in the presence of 25% saliva (a mixture of five healthy donor samples) for 1 h in 37°C in TNCT buffer containing 2 mM cysteine. The reaction was stopped by adding Kyt-1 or Kyt-36 (2 M final concentration) into samples with HRgpA/RgpB or Kgp, respectively, and incubating at 37°C for 5 min. The integrity of SPINK6 was determined by both Western blotting analysis and by its ability to inhibit the enzymatic activity of KLK13. Samples separated by SDS-PAGE were transferred onto PVDF (Bio-Rad). Nonspecific binding sites were blocked with 5% skim milk in TTBS, followed by incubation with a monoclonal HRP-conjugated anti-His tag antibody (catalog no. A7058, Sigma-Aldrich) diluted 1:20,000 in TTBS. Alternatively, a 2-fold molar excess of the reaction products (as estimated by the original SPINK6 amount in the sample) was added to a 10 nM KLK13 solution in 0.1 M Tris-HCl (pH 7.5) supplemented with 2 M Kyt-1 and Kyt-36. Kallikrein activity was measured by monitoring the fluorescence intensity as described above.