A Potential Role for Multiple Tissue Kallikrein Serine Proteases in Epidermal Desquamation*

Desquamation of the stratum corneum is a serine protease-dependent process. Two members of the human tissue kallikrein (KLK) family of (chymo)tryptic-like serine proteases, KLK5 and KLK7, are implicated in desquamation by digestion of (corneo)desmosomes and inhibition by desquamation-related serine protease inhibitors (SPIs). However, the epidermal localization and specificity of additional KLKs also supports a role for these enzymes in desquamation. This study aims to delineate the probable contribution of KLK1, KLK5, KLK6, KLK13, and KLK14 to desquamation by examining their interactions, in vitro, with: 1) colocalized SPI, lympho-epithelial Kazal-type-related inhibitor (LEKTI, four recombinant fragments containing inhibitory domains 1–6 (rLEKTI(1–6)), domains 6–8 and partial domain 9 (rLEKTI(6–9′)), domains 9–12 (rLEKTI(9–12)), and domains 12–15 (rLEKTI(12–15)), secretory leukocyte protease inhibitor, and elafin and 2) their ability to digest the (corneo)desmosomal cadherin, desmoglein 1. KLK1 was not inhibited by any SPI tested. KLK5, KLK6, KLK13, and KLK14 were potently inhibited by rLEKTI(1–6), rLEKTI(6–9′), and rLEKTI(9–12) with Ki values in the range of 2.3–28.4 nm, 6.1–221 nm, and 2.7–416 nm for each respective fragment. Only KLK5 was inhibited by rLEKTI(12–15) (Ki = 21.8 nm). No KLK was inhibited by secretory leukocyte protease inhibitor or elafin. Apart from KLK13, all KLKs digested the ectodomain of desmoglein 1 within cadherin repeats, Ca2+ binding sites, or in the juxtamembrane region. Our study indicates that multiple KLKs may participate in desquamation through cleavage of desmoglein 1 and regulation by LEKTI. These findings may have clinical implications for the treatment of skin disorders in which KLK activity is elevated.

Proteases (final concentration of 12 nM for KLK1, KLK5, KLK13, KLK14, and NE or 30 nM for KLK6) were preincubated with LEKTI fragments (0 -60 nM), SLPI (0 -1.2 M), and/or elafin (0 -1.2 M) in 10 l of optimal buffer at 25°C (for LEKTI reactions) or 37°C (for SLPI and elafin reactions) with gentle agitation for different time points (10 min for KLK1 and KLK6, 5 min for KLK6, 1 min for KLK5, and 10 s for KLK14 with LEKTI fragments; 30 min for KLK1, KLK6, and KLK13, 5 min for KLK6 and KLK14, and 1 min for NE with SLPI and elafin). The mixtures were subsequently added to 90 l of optimal buffer containing several fixed AMC peptide concentrations ranging from 4 to 3000 M within polystyrene microtiter plate wells. Initial rates of protease-mediated peptide hydrolysis were monitored by measuring free AMC fluorescence on the Wallac 1420 Victor 2TM fluorometer (PerkinElmer Life Sciences) with excitation and emission filters of 380 and 480 nm, respectively, at 1-min intervals for 20 min at 37°C. Protease-free reactions, for each substrate concentration, were used as negative con-trols, and the background counts obtained were subtracted from each value. A standard curve was constructed using known concentrations of AMC to convert rates of reaction from AMC fluorescence counts/min to free AMC produced/ min. The slope of the resultant AMC standard curve was 19.184 AMC fluorescence counts/nM AMC.
The rate changes (nanomolar AMC/min) of inhibited and control reactions were determined from the velocity plots. Activities were expressed relative to control incubations from which inhibitors were excluded. Average IC 50 values, i.e. the inhibitor concentration required for 50% inhibition of protease activity, were determined by non-linear regression analysis using Prism (Version 4.0, GraphPad, San Diego, CA). Michaelis-Menten parameters (K m and V max ), the equilibrium inhibition constant (K i ), and inhibitory mechanisms were determined by linear and non-linear regression analysis using the Enzyme Kinetics Module 1.1 (Sigma Plot, SSPS, Chicago, IL). All experiments were performed in triplicate and repeated at least twice.
Effect of KLKs on SLPI Activity-Prior to measuring the inhibitory activity of SLPI against NE, individual KLKs were incubated with SLPI at several molar ratios (i.e. 1:5, 1:20, and 1:50) for different time points (0, 8, and 24 h) at 37°C with gentle agitation. SLPI was also incubated alone for each time point, as a positive control. Samples from each time point were subsequent assayed for SLPI activity against NE, at NE to SLPI molar ratios of 1:1.25, 1:2.5, 1:5, and 1:10, using AAPV-AMC (0.1 mM) as a substrate. Free AMC fluorescence was measured on the Wallac 1420 Victor 2TM fluorometer for 20 min, as described above. NE-free reactions were used as negative controls, and background fluorescence was subtracted from each value. All experiments were performed in triplicate. (Note: the hydrolysis of AAPV-AMC by KLKs was not observed (data not shown).) Pre-treatment and in Vitro Proteolysis of DSG1/Fc by KLKs-Prior to incubation with KLKs, DSG1/Fc (50 ng/l, ϳ0.5 M) was incubated at 37°C with gentle agitation in the presence of either: 1) Tris-buffered saline (TBS; 20 mM Tris base, 140 mM NaCl, pH 7.6) for 1 h, 2) TBS containing 5 mM EGTA for 1 h, 3) TBS containing 1 mM CaCl 2 for 1 h, 4) TBS with 5 mM EGTA for 1 h, and then treated with TBS including 1 mM CaCl 2 for 1 h, 5) TBS with 1 mM CaCl 2 for 1 h, followed by the addition of 5 mM EGTA for 1 h. DSG1/Fc also was pre-treated as above in 0.1 M sodium acetate buffer (pH 5.4) in a total reaction volume of 70 l. Proteolysis of DSG1/Fc was subsequently studied via incubation of individual KLKs (50 ng) with pre-treated DSG1/Fc (500 ng) in a final volume of 20 l, in either TBS or sodium acetate, for various time intervals (0, 15 min, 30 min, 1 h, 2 h, 4 h, 8 h, and 24 h) at 37°C with shaking. Reactions were stopped by freezing in liquid nitrogen. Proteins were separated by reducing SDS-PAGE and visualized by silver staining, as described above.
Effect of Calcium and EGTA on KLK Activity-KLK1, KLK5, KLK6, KLK13, and KLK14 (12 nM) were pre-treated with either: 1) TBS (pH 7.6), 2) sodium acetate buffer (pH 5.4), 3) TBS with 1 mM CaCl 2 , 4) sodium acetate buffer with 1 mM CaCl 2 , 5) TBS containing 1 mM CaCl 2 and 5 mM EGTA, and 6) sodium acetate buffer supplemented with 1 mM CaCl 2 and 5 mM EGTA, at a final volume of 100 l in microtiter plate wells for 10 min at 37°C with gentle agitation. Next, the appropriate AMC substrates for each KLK were applied to the wells at a final concentration of 0.5 mM. Free AMC fluorescence was measured on the Wallac 1420 Victor 2TM fluorometer for 20 min, as described above. KLK-free reactions were used as negative controls, and background fluorescence was subtracted from each value. All experiments were performed in triplicate.
N-terminal Sequencing-N-terminal sequence analysis was performed to identify KLK cleavage sites within SLPI and DSG1/Fc. KLKs (250 ng) were incubated with SLPI or DSG1/Fc (2.5 g) for different time points as specified by the proteolysis experiments, separated by reducing SDS-PAGE, and transferred to a polyvinylidene difluoride membrane (Amersham Biosciences), previously immersed in 100% methanol, at 30 V for 1 h. After the transfer, the membrane was removed and rinsed with de-ionized water, stained with Coomassie Blue R-250 (0.1% solution in 40% methanol) for 5 min, and de-stained for 5 min in 50% methanol. The membrane was then thoroughly washed with de-ionized water and air-dried. SLPI and DSG1/Fc fragments were excised and subjected to automated N-terminal Edman degradation consisting of five cycles of Edman chemistry on a ABI 492 Procise cLC sequencer (Applied Biosystems, Foster City, CA), followed by analysis of resultant phenylthiohydantoinamino acid residues on an high-performance liquid chromatography column.

Multiple KLKs Are Inhibited by rLEKTI Fragments-Because
KLKs represent likely physiological targets of LEKTI we tested rLEKTI fragments spanning all 15 domains for their inhibitory activity against the trypsin-like serine proteases KLK1, KLK5, KLK6, KLK13, and KLK14, in vitro.
Among the KLKs examined, KLK1 activity was uniquely unaffected after incubation with all rLEKTI fragments ( Fig. 1), even in the presence of an 8-to 10-fold molar excess of rLEKTI or after different preincubation times (0 -30 min). Furthermore, cleavage of rLEKTI fragments by KLK1 was not detected under the conditions used for kinetic inhibition assays (see Fig. 3). These results are in agreement with prior studies that demonstrate a lack of inhibition of the porcine ortholog of KLK1 by native and recombinant LEKTI domain 6 (37,55).
The remaining KLKs were inhibited by rLEKTI fragments with IC 50 and K i values in the nanomolar range (Table 1 and Fig. 1) via noncompetitive or mixed-type inhibitory mechanisms, as determined by Eadie-Hofstee plots ( Fig. 2 where the slope is equal to ϪK m and the y-intercept is equal to V max ) and corresponding kinetic parameters (V max , K m ; Tables 2-5). The overall rank order for KLK inhibition by rLEKTI fragments based on IC 50 and K i values from lowest to highest was: KLK5 Ϸ KLK14 Ͻ KLK6 Ͻ KLK13, where values obtained for KLK5 and KLK14 values were consistently several orders of magnitude lower than those for KLK6 and KLK13 (Table 1). rLEKTI(1-6) was the most potent inhibitor for all KLKs tested (exclusive of KLK1), particularly for KLK5 and KLK14, with K i values of 2.35 nM and 0.22 nM and IC 50 values of 2.43 nM and 0.0051 nM, respectively. The concentration-dependent inhibition of KLKs by rLEKTI(1-6) is shown in Fig. 1A. As evident by the decrease in V max and increase in K m for KLK5, KLK6, KLK13, and KLK14 in the presence of increasing rLEKTI(1-6) concentrations, shown graphically via the Eadie-Hofstee plot ( Fig. 2A), the inhibitory mechanisms were classified as mixed, i.e. containing both competitive and non-competitive components (Table 2). rLEKTI(6 -9Ј) and rLEKTI(9 -12) also exerted strong mixedtype inhibitory activity against KLK5 and KLK14, with K i values ranging from 2.75 to 10.26 nM (Tables 3 and 4 and Fig. 2C). These data are in general accordance with that of Schechter et al. (40), who recently published K i values of 3.5-5.5 nM for the inhibition of KLK5 by rLEKTI(6 -9Ј) and rLEKTI(9 -12) at pH 8.0. In contrast to KLK5 and KLK14 inhibition, KLK6 and KLK13 activity was inhibited to a lower extent and in a noncompetitive manner by rLEKTI(6 -9Ј) and rLEKTI (9 -12), as evidenced by the dose-response curves (Fig. 1, B and C) and the increase in V max values accompanied by relatively constant K m values in the presence of increasing rLEKTI(6 -9Ј) and rLEKTI(9 -12) concentrations (Tables 3 and 4 and Fig. 2B), respectively. Both KLK6 and KLK13 were also less effectively inhibited by rLEKTI(6 -9Ј) and rLEKTI(9 -12) than by rLEKTI(1-6) ( Table 1). For example, the K i values for the inhibition of KLK13 by rLEKTI(6 -9Ј) and rLEKTI(9 -12) were  Table 1.  FEBRUARY 9, 2007 • VOLUME 282 • NUMBER 6
KLK5 was the only KLK tested to be inhibited by rLEKTI (12)(13)(14)(15) (Tables 1 and 5 and Figs. 1D and 3D), indicative of a stringent or unique specificity for this rLEKTI fragment. The observed lack of inhibitory activity of rLEKTI (12)(13)(14)(15)) against the other KLKs tested was unexpected based on the primary LEKTI sequence, which contains highly favorable P1-Arg active site residues within domains 12-14. Thus, the general inactivity of this fragment could be 1) attributed to the presence of a P1-Lys residue within domain 15, which is not highly preferred by KLK5, KLK6, KLK13, or KLK14 (45,(52)(53)(54) and is supported by the finding that rLEKTI domain 15 does not inhibit KLK5 (39) or 2) possibly due to an unfavorable tertiary structure (e.g. suboptimal backbone angles) within P3-P3Ј regions, which form the canonical binding loops that interact with serine proteases. In the case of KLK14, the lack of detectable inhibition by rLEKTI(12-15) may be due to the rapid proteolysis and possible inactivation of this inhibitory fragment by KLK14 (Fig. 3D). We have recently identified a putative KLK14 cleavage site within LEKTI domain 15, i.e. Arg 1034 2 Gln 1035 (LEKTI numbering according to GenBank TM accession number NP_006837), by phage display substrate technology (47), which would give rise to a ϳ6.5-kDa fragment of rLEKTI (12)(13)(14)(15). Due to its classification as a reversible, tight-binding, "standard" mechanism inhibitor, LEKTI acts as highly specific protease substrate, which mediates inhibition through cleavage of a reactive site bond by the cognate protease (56,57). Typically, hydrolysis of the inhibitor is extremely slow and does not proceed to completion (57). Thus, the stability of rLEKTI fragments against KLK proteolysis and of KLK/rLEKTI complexes was also investigated (Fig. 3). Relative rates of proteolysis were determined upon examination of the decrease in intensity of intact rLEKTI bands over time. rLEKTI fragments were generally resistant to digestion by KLKs, with the exception of the complete hydrolysis of: rLEKTI(6 -9Ј) by KLK6 after 24 h (Fig. 3B), rLEKTI(9 -12) by KLK6 and KLK14 after 4 h (Fig. 3C) and rLEKTI(12-15) by KLK14 after 1 h (Fig. 3D).

Tissue Kallikreins as Desquamation-related Proteases
rLEKTI(1-6) and rLEKTI(6 -9Ј) were digested at a slower rate by KLK5 (Fig. 3, A and B) as was rLEKTI(12-15) by KLK13 (Fig.  3D). Notably, rLEKTI(1-6) was not fully degraded by any KLK, suggesting that it may have a longer half-life than other rLEKTI fragments (data not shown), which may be due to a unique domain conformation. With respect to the KLKs that were inhibited by rLEKTI fragments, most did not recover their activities even after 24 h, except for the few examples described above. Taken together, rLEKTI fragments exhibited both permanent (pseudoirreversible) and temporary (reversible) behavior, depending on the KLK/LEKTI fragment combination. Temporary inhibition of trypsin by native and recombinant LEKTI domain 6 was also observed, as evidenced by trypsin digestion of the inhibitor at the reactive site after a 1-h incubation, indicative of a short-lived, unstable trypsin-LEKTI complex (55). This instability was attributed to the presence of only two of three disulfide bridges that characterize Kazal-type inhibitors within the LEKTI domain 6 (55). Furthermore, our results suggest that rLEKTI fragments function as intact, multidomain fragments and need not be processed into single domains to be active, as previously postulated (31). KLK Activity Is Not Inhibited by SLPI or Elafin-SLPI (58) and elafin/SKALP (59) are present in the stratum corneum and are implicated in desquamation. For instance, both SLPI and elafin can inhibit corneocyte shedding in vitro and inhibit KLK7 activity (41), some NS patients do not display increased epider-    FEBRUARY 9, 2007 • VOLUME 282 • NUMBER 6 mal serine protease activity, indicating that other SPI are present to attenuate serine protease function (25), and elafin is overexpressed in the epidermis of NS patients (50,60). This prompted us to test the hypothesis that additional KLKs may be targeted by these two inhibitors. However, unlike the effect of elafin and SLPI on KLK7, no demonstrable inhibitory activity was found against any KLK studied under the experimental conditions chosen, even at a 100-fold molar excess of either inhibitor (data not shown). These results are likely explained by the presence of P1-Leu 72 and P1-Ala 84 within the reactive sites of SLPI (61) and elafin (62), respectively, both of which are unfavorable P1 amino acids for the KLKs tested. (Note: the numbering starts from the mature SLPI sequence and the elafin preproprotein sequence according to GenBank TM accession numbers NP_003055 and NP_002629, respectively). Similarly, lack of inhibition of KLK3 (also known as prostate-specific antigen) by SLPI has also been reported (63). Proteolytic Processing of SLPI by KLKs-The incubation of SLPI with individual KLKs at a 26:1 molar ratio for 24 h resulted in the generation of 2 or 4 SLPI degradation products, depending on the KLK (Fig. 4). Intact SLPI was visualized as a ϳ12-kDa band and contains two C-terminal whey acidic protein (WAP) domains, whereby the first possesses antimicrobial activity (64) and the second harbors inhibitory activity (65). KLK14 was able to completely degrade intact SLPI after 24 h; lower hydrolysis rates were exhibited by KLK1, followed by KLK6, KLK5, and KLK13. Due to cleavage at peptide bond Arg 20 -Tyr 21 within the first WAP domain, all KLKs produced two SLPI fragments of 9.6 kDa (S1; Tyr 21 -Ala 132 ) and 2 kDa (Ser 1 -Arg 20 ) (Table 6). An additional cleavage site was identified for KLK6 and KLK14 within the second WAP domain at Leu 72 -Met 73 , which generated another two SLPI fragments of 5.8 kDa (S2; Tyr 21 -Leu 72 ) and 3.5 kDa (S3; Met 73 -Ala 132 ). Although P1-Leu is not a preferred amino acid for either KLK, the observed cleavage may be due to the relative exposure and susceptibility of the reactive site Leu 72 -Met 73 bond to protease attack and/or the importance of other residues, in addition to P1, to protease specificity.

Tissue Kallikreins as Desquamation-related Proteases
The presence of KLK6 and KLK14 cleavage sites within the second WAP domain of SLPI, which contains the reactive site P1-Leu 72 for protease inhibition, impelled us to examine whether KLKs could inactivate SLPI and thereby diminish its inhibitory activity against NE. Upon incubation with both KLK6 and KLK14, SLPI activity was decreased to only 90% of control (i.e. SLPI alone). However, this may not be significant compared with cathepsins L and S, which have been shown to completely abrogate SLPI activity against NE in a 1-h time period (66). The reason(s) for this disparity are not clear, but may be attributed to the location and number of the cathepsin and KLK cleavage sites within the second WAP domain. Cathepsins cleave SLPI at two sites, i.e. Thr 68 -Tyr 69 and Met 73 -Leu 74 , and may affect the structure of the intervening region, Tyr 69 -Met 73 , which is involved in binding SLPI target proteases (66,67). In contrast, KLK6 and KLK14 cleave solely after  indicate KLK-generated SLPI fragments. SLPI fragments subjected to N-terminal sequence analysis are labeled "S1-S7" on the right, and the results are listed in Table 6. A schematic representation of KLK-mediated SLPI cleavage is shown on the right. Molecular mass standards (kilodaltons) are indicated on the left. the active site P1-Leu 72 and do not disrupt the protease binding site or significantly alter the structure of SLPI. Further stabilization of KLK6-and KLK14-cleaved SLPI conformation is likely provided by the four disulfide bridges within the WAP domain, which thereby facilitates SLPI inhibition of NE, even in the presence of these two KLKs. Moreover, cleavage at P1-R 20 by all KLKs tested could possibly abrogate the antibacterial activity of SLPI. The consequences of KLK-mediated digestion of SLPI are to be determined.
KLK-mediated Proteolysis of DSG1-A prerequisite for stratum corneum desquamation is the degradation of the desmosomal cadherin, DSG1 (8,9). KLKs are implicated in DSG1 proteolysis due to: 1) their colocalization with DSG1 in the stratum corneum, 2) the in vitro digestion of DSG1 within epidermal extracts by KLK5 (24), and 3) the correlation between decreased DSG1 levels and elevated KLK activity within the epidermis of NS patients (25). Given this, we examined whether DSG1 could serve as a substrate for KLK1, KLK6, KLK13, and KLK14, in addition to KLK5, in vitro. DSG1/Fc, a chimeric protein formed of the five extracellular domains of DSG1 followed by a peptide linker and the Fc region of human IgG 1 , was incubated with KLKs for increasing time periods at pH 5.4, within the acidic pH range of the stratum corneum (68), and at pH 7.6, within the optimal pH range for KLK activity (Fig. 5). In all cases, no degradation of DSG1/Fc was observed in the absence of KLKs (Fig. 5, A-D, lane 2). Relative KLK cleavage efficiencies were assessed by the decrease in intact DSG1/Fc band intensity over time. As expected, higher KLK activity was generally observed at pH 7.6 rather than pH 5.4, and despite their similar specificity for P1-Arg, each KLK produced a distinct DSG1/Fc cleavage pattern and cleaved at unique sites ( Fig. 5 and Table 6). No detectable digestion of DSG1/Fc by KLK13 was found at either pH tested, even after 24 h of incubation (data not shown). KLK1 showed minimal activity against DSG1/Fc at pH 5.4, which was slightly elevated at pH 7.6, allowing for the determination of two putative cleavage sites at K 197 -Ile 198 and K 445 -Asn 446 ( Table 6).
The remaining KLKs demonstrated comparably higher DSG1/Fc cleavage rates, with a rank order of KLK6 Ͻ KLK5 Ͻ Ͻ KLK14, from lowest to highest efficiency. As indicated in Fig.   5C, KLK6 generated three major DSG1/Fc fragments of ϳ66, 55, and 50 kDa, the latter two of which were denoted 6A and 6B and resulted from the cleavage at Lys 368 -Ala 369 and R 422 -Tyr 423 (Table 6). In contrast, KLK5 induced a relatively faster decrease in intact DSG1/Fc and generated a multitude of DSG1/Fc degradation products. The N-terminal sequence of four major DSG1/Fc fragments of ϳ66, 45, 36, and 25 kDa (Fig.  5, A-D, respectively) was identified, and permitted us to determine three putative cleavage sites at Arg 146 -Val 147 , Arg 242 -Asp 243 , and Arg 422 -Tyr 423 (Table 6). These results agree with those of Caubet et al. (24), who demonstrated efficient KLK5mediated degradation of DSG1 within epidermal extracts at pH 5.4. Undoubtedly, DSG1/Fc was the best substrate for KLK14. Complete digestion of intact DSG1/Fc was demonstrated in the presence of KLK14 after 30 min at pH 5.4 and 5 min at pH 7.6 ( Fig. 5D). Two major cleavage products of ϳ40 and 37 kDa (14A and 14B, respectively) were obtained, and were not further degraded after a 24-h incubation period (data not shown). The N terminus of both 14A and 14B fragments was Ser 529 -Ser-Glu-Pro-Gly-Asn, suggestive of cleavage at the Tyr 528 -Ser 529 peptide bond. The P1-Tyr specificity of KLK14 is not surprising as we previously determined that KLK14 is a dual specificity enzyme with both trypsin and chymotrypsin-like activity (47).
Given that calcium is essential for the stabilized conformation and adhesive function of cadherins such as DSG1 (4) we addressed whether the presence/absence of calcium would affect the susceptibility of DSG1 to proteolysis by KLKs. To this end, DSG1/Fc was pre-treated with 1 mM CaCl 2 and/or 5 mM EGTA, a chelator with high selectivity for Ca 2ϩ ions in two buffer systems, TBS (pH 7.6) and sodium acetate (pH 5.4). As a control, DSG1/Fc was also pre-treated in buffer only, i.e. in the absence of CaCl 2 and EGTA, as in the time-course experiments. A significant effect on KLK activity by CaCl 2 and EGTA was not observed (data not shown). As shown in Fig. 6, the presence of either EGTA or CaCl 2 did not influence the efficiency of DSG1/Fc digestion by KLKs, under the experimental conditions chosen. Hence, the calcium-stabilized conformation of DSG1 may not be required for KLK-mediated cleavage. This is also the case for other serine proteases such as trypsin and V8 proteinase, but not for exfoliative toxins, which can only cleave

DISCUSSION
Stratum corneum desquamation is a tightly regulated process, orchestrated by the concerted function of serine proteases and SPI within the intercorneal matrix. To date, serine protease action in the stratum corneum has been linked to two members of the human kallikrein family: KLK5 and KLK7. Yet, besides KLK5 and KLK7, it is now recognized that an additional nine KLKs are expressed in the uppermost epidermal layers (26). Among these, the stratum corneum levels of six KLKs have been quantified (29), and one, namely KLK14, has been recently purified from the stratum corneum in its active form (30). The present study is the first to support a role for kallikreins other than KLK5 and KLK7, namely KLK6, KLK13, and KLK14, as desquamatory enzymes, as indicated by their efficient digestion of DSG1 and/or their potent inhibition by LEKTI, in vitro.
Unregulated desquamation drives and sustains the phenotype of NS, an inherited skin disorder in which the serine protease/SPI balance is tipped in favor of serine proteases, due to mutations in the gene encoding LEKTI. This study is the first to profile the inhibitory capacity of all 15 inhibitory LEKTI domains against multiple KLKs, summarized in Fig. 7. Our in vitro data indicate that KLK5, KLK6, KLK13, and KLK14 are significantly inhibited by rLEKTI fragments. To establish the in vivo significance of protease-inhibitor interactions, the determination of in vitro kinetic parameters (e.g. K i ) as well as the in vivo inhibitor concentration are generally required (71,72). Although in vivo LEKTI stratum corneum concentrations are currently unknown, in the case of KLK-LEKTI interactions, the resultant low K i constants (in the nanomolar range), which indicate the formation of relatively stable KLK-LEKTI complexes and Notably, the individual rLEKTI fragments utilized in the present study and by Schechter et al. (40) display distinct inhibitory specificities against the KLKs examined thus far (Fig. 7). Generally, LEKTI domains 1-8 can potently inhibit multiple KLKs; yet domains 9 -12 exhibit a higher specificity toward KLK5 and KLK14, whereas domains 12-15 can only inhibit KLK5. This suggests that each domain or fragment may have one or more preferred protease targets in vivo. Although these rLEKTI fragments may not be fully representative of physiological LEKTI forms, accumulating studies suggest that post-translational processing of LEKTI occurs in vitro and in vivo. For instance, LEKTI domains 1, 5, and 6 have been isolated from human blood filtrates (37,73), and several multidomain LEKTI fragments ranging from 30 to 68 kDa are present in the conditioned media of normal primary human keratinocytes (38,48,74,75), indicating that both single and multidomain fragments of LEKTI occur naturally.
Inhibition of KLK activity by rLEKTI fragments occurred via non-competitive or mixed-type mechanisms. Jayakumar et al. (49) previously proposed that the non-competitive inhibitory action of rLEKTI(6 -9Ј) could be due to the interaction between 1) the canonical loop of one inhibitory domain with the catalytic site of the protease and 2) the binding of a second inhibitory domain to an allosteric site within the protease. Similar reasons could account for the observed inhibitory mechanisms of rLEKTI on KLKs. Further studies are warranted to clarify these findings.
SLPI and elafin/SKALP are implicated in the regulation of desquamation, yet neither inhibitor could attenuate the activity of any KLK studied. Putative targets of SLPI and elafin inhibitory action within the skin may include KLK7 (41) and elastase I, which is expressed by cultured primary keratinocytes (76) and is known to interact with these inhibitors in other tissues, in vivo. Moreover, KLK1 activity was not attenuated by any of the SPIs studied. These results were particularly unexpected for  Table 6). Arrows indicate KLK protein bands. Molecular mass standards (kilodaltons) are shown on the left. Note that the protein band corresponding to KLK5 is masked by KLK5-generated DSG1/Fc fragments.  FEBRUARY 9, 2007 • VOLUME 282 • NUMBER 6 LEKTI, which harbors P1-Arg in the majority of its inhibitory domains, a residue that is highly favored by KLK1 (51). However, KLK1 possesses a strong preference for large hydrophobic residues at P2 (77,78) and for basic residues at P3 (79). Thus, the presence of P2-Thr/Pro and P3-Cys within LEKTI inhibitory domains may explain, in part, the resistance of KLK1 to inhibition by LEKTI. The regulation of KLK activity, particularly KLK1, by other SPIs in the epidermis, i.e. ␣ 2 -macroglobulin-like-1 (42), protein C inhibitor (80), and plasminogen activator inhibitor-2 (81), may be worth examining in the future.

Tissue Kallikreins as Desquamation-related Proteases
As a (corneo)desmosomal adhesion molecule, DSG1 is essential for the maintenance of epidermal integrity (4), as evidenced by various inherited and acquired skin diseases that target DSG1 (82). Previous studies have shown that KLK5 can digest DSG1 in vitro (24) and that an association between premature DSG1 degradation and elevated KLK activity in the epidermis of NS patients exists (25). In the work presented here, we provide evidence for the direct cleavage of DSG1 by KLK1, KLK5, KLK6, and KLK14 and identify a number of putative cleavage sites, in vitro. Four KLK cleavage sites occur within the first three extracellular cadherin repeat domains (EC1-ECIII), which mediate heterophilic binding to desmocollins (e.g. desmocollin-1) (83). Among these sites, two reside within calcium binding motifs, responsible for the calcium-dependent conformation needed for proper DSG1 function. Interestingly, the KLK14 cleavage site is located a few amino acids N-terminal to the transmembrane domain, which would result in shedding of the entire ectodomain of DSG1. Thus, based on the location of KLK cleavage sites, the coordinated action of several KLKs might destabilize DSG1 structure and abolish adhesion to desmocollins on neighboring corneocytes. Moreover, extracellular DSG1 proteolytic fragments generated by KLKs may further perturb intercellular adhesion by competing with intact DSG1 and desmocollin-1 molecules for binding, as shown for soluble E-cadherin ectodomain fragments (84). Ultimately, KLK action could lead to corneocyte dyshesion during normal epidermal turnover but can result in overdesquamation and an impairment of epidermal barrier function when it is unopposed as in NS (25,(31)(32)(33).
Given the above, the results of our study may have important therapeutic implications for the treatment of skin disorders characterized by dysregulated KLK activity. We and others (39,40) have shown that LEKTI has a potent inhibitory capacity for several KLKs. Hence, the administration of LEKTI as a therapeutic agent, to balance its deficiency in NS and/or of other KLK-specific inhibitors, to attenuate unopposed KLK activity observed in other diseases, could result in the restoration of epidermal integrity, permeability barrier function and a reduction of pro-inflammatory responses.
It has been postulated that the clinical severity of NS is correlated with the location of mutations within the SPINK5 gene (31). That is, NS patients with mutations near the 3Ј-end of SPINK5 encode LEKTI forms with a higher number of inhibitory domains than those whose mutation reside near the 5Ј-end. Accordingly, the degree of serine protease/SPI imbalance in the former will be lower than in the latter, resulting in a moderate increase in serine protease activity and hence, a less severe disease phenotype (31). As aforementioned, rLE-KTI fragments display distinct inhibitory profiles against KLKs (Fig. 7). Hence, the lack of specific LEKTI domains in NS may lead to the elevation in the activity of only certain KLKs (e.g. deficiency of domains 12-15 may cause an increase in KLK5 activity specifically). Taken together, our data could allow for the development of tailored therapeutic strategies for subgroups of NS patients with distinct SPINK5 mutations and corresponding clinical severities. For instance, NS patients with downstream mutations may benefit from the administration of LEKTI (12)(13)(14)(15) or another inhibitor that targets KLK5 specifically, whereas patients with upstream mutations may require the suppression of multiple KLKs via LEKTI(1-6) or a more general SPI.
In conclusion, our results clearly demonstrate that multiple KLKs are potently inhibited by LEKTI and are able to efficiently degrade DSG1, in vitro. This study further substantiates the role of KLK5 as a desquamatory protease and also reveals that KLK6, KLK13, and particularly KLK14 may contribute to epidermal turnover and homeostasis under normal and pathological conditions. Accordingly, the suppression of KLK activity by LEKTI and/or other specific inhibitors may represent novel therapeutic strategies for the treatment of various skin diseases.