Enzymatic and Molecular Characteristics of the Efficiency and Specificity of Exfoliative Toxin Cleavage of Desmoglein 1*

Exfoliative toxins (ETs) from Staphylococcus aureus blister the superficial epidermis by hydrolyzing a single peptide bond, Glu381–Gly382, located between extracellular domains 3 and 4 of desmoglein 1 (Dsg1). Enzyme activity is dependent on the calcium-stabilized structure of Dsg1. Here we further define the characteristics of this cleavage. Kinetic studies monitoring the cleavage of Dsg1 by ETA, ETB, and ETD demonstrated kcat/Km values of 2–6 × 104 m–1 s–1, suggesting very efficient proteolysis. Proteolysis by ETA was not efficiently inhibited by broad spectrum serine protease inhibitors, suggesting that the enzyme cleavage site may be inactive or inaccessible before specific binding to its substrate. Using truncated mutants of human Dsg1 and chimeric molecules between human Dsg1 and either human Dsg3 or canine Dsg1, we show that for cleavage, human-specific amino acids from Dsg1 are necessary in extracellular domain 3 upstream of the scissile bond. If these residues are canine rather than human, ETA binds, but does not cleave, canine Dsg1. These data suggest that the exquisite specificity and efficiency of ETA may depend on the enzyme's binding upstream of the cleavage site with a very specific fit, like a key in a lock.

Exfoliative toxins (ETs) from Staphylococcus aureus blister the superficial epidermis by hydrolyzing a single peptide bond, Glu 381 -Gly 382 , located between extracellular domains 3 and 4 of desmoglein 1 (Dsg1). Enzyme activity is dependent on the calcium-stabilized structure of Dsg1. Here we further define the characteristics of this cleavage. Kinetic studies monitoring the cleavage of Dsg1 by ETA, ETB, and ETD demonstrated k cat /K m values of 2-6 ؋ 10 4 M ؊1 s ؊1 , suggesting very efficient proteolysis. Proteolysis by ETA was not efficiently inhibited by broad spectrum serine protease inhibitors, suggesting that the enzyme cleavage site may be inactive or inaccessible before specific binding to its substrate. Using truncated mutants of human Dsg1 and chimeric molecules between human Dsg1 and either human Dsg3 or canine Dsg1, we show that for cleavage, humanspecific amino acids from Dsg1 are necessary in extracellular domain 3 upstream of the scissile bond. If these residues are canine rather than human, ETA binds, but does not cleave, canine Dsg1. These data suggest that the exquisite specificity and efficiency of ETA may depend on the enzyme's binding upstream of the cleavage site with a very specific fit, like a key in a lock.
Staphylococcus aureus is a frequent bacterial pathogen in human disease. Skin infections are particularly prevalent, with impetigo caused by S. aureus one of the most common infectious diseases in children (1). Approximately 30% of impetigo patients develop bullous impetigo, which is caused by S. aureus strains that produce exfoliative toxins (ETs) 1 (2,3). In bullous impetigo, blisters of the superficial epidermis occur at sites of infection. More extensive disease can occur in staphylococcal scalded skin syndrome (SSSS), in which ETs produced by a nidus of infection circulate systemically, usually in infants and young children, to produce more generalized blistering. Al-though SSSS is much less common than bullous impetigo, outbreaks that occur in neonatal nurseries have important consequences for health care delivery (2,4).
In the 1970s Melish et al. (5)(6)(7) showed that ETs produced by S. aureus injected into neonatal mice caused blister formation in skin similar to that seen in patients with bullous impetigo and SSSS. It remained unclear exactly how ETs cause this blister until it was discovered that they have structural and sequence homologies to serine proteases, at which time it was suggested that they act as proteolytic enzymes that were predicted to cleave after either a glutamic or aspartic acid (8). However, they could not be shown to cleave model polypeptide substrates. Their normal pathophysiologic substrate, if any, remained a mystery until it was realized that the blisters in bullous impetigo and SSSS show clinical and histological similarity to those of pemphigus foliaceus, an autoimmune disease in which pathogenic autoantibodies bind the desmosomal cell adhesion molecule, desmoglein 1 (Dsg1) (9). This observation led to the hypothesis that the molecular target of ETs is Dsg1. This hypothesis was validated by showing that three different ET isotypes (ETA, ETB, and ETD) specifically cleave mouse and human Dsg1 (hDsg1) but not closely related molecules such as Dsg3 or E-cadherin (9 -11). Mutation of the predicted active serine in these ETs caused loss of their proteolytic activity, confirming that they are serine proteases (12).
Dsg1 is in the desmosomal cadherin subfamily of the cadherin supergene family. Desmosomal cadherins, similar to classic cadherins, contain four amino-terminal homologous domains (called extracellular (EC) domains 1-4), of about 100 amino acid each, that are highly conserved among many cadherins (13). A fifth juxtamembrane domain (EC5) is not as well conserved within or among the cadherins. The structure of Dsg1, like other cadherins, is maintained by calcium binding that occurs between EC domains (14 -17). This calcium-stabilized structure is thought to be critical in maintaining the adhesive function of classical and, presumably, desmosomal cadherins (18 -20).
Recent studies show not only that ETs show a substrate specificity for Dsg1 but also that they hydrolyze only one peptide bond in human and mouse Dsg1 at exactly the same site, after glutamic acid 381 between EC3 and EC4 (12). Furthermore, cleavage of Dsg1 by ETs is dependent on the former's calcium-stabilized conformation (i.e. denatured or partially denatured Dsg1 cannot be hydrolyzed by ETs) (21). Thus, unlike more typical serine proteases whose specificities are defined by simple amino acid sequences and which can hydrolyze many proteins and model polypeptides, ETs show exquisite specificity for one peptide bond in one correctly folded substrate, Dsg1.
In this study, we show that in addition to being exquisitely specific for a single peptide bond in Dsg1, ETs are efficient in hydrolyzing that peptide bond. We show that the catalytic site of ETA is protected from typical serine proteases inhibitors and that domains in Dsg1 amino-terminal, but not carboxyl-terminal, to the cleavage site are necessary for its cleavage by ETA.
Finally, we show that although ETA binds to canine Dsg1 (cDsg1) it does not cleave it, but it will do so if 5 amino acids that are upstream of the cleavage site are mutated to hDsg1 residues. The results of these studies suggest that the exquisite Densitometry of stained SDS-PAGE indicates loss of hDsg1E (squares) and production of cleaved products (circles) with increasing time of digestion. D, plots of the natural log (ln) of the density of the substrate, hDsg1E, against time show good linear fits for substrate concentrations down to ϳ90% of starting concentrations, consistent with pseudo-first-order decay (i.e. velocity of decay is directly proportional to the substrate concentration at a specific starting enzyme concentration). E, the exponential constant, k obs , was calculated for appearance of products (filled circles) or disappearance of substrates (open circles) for three concentrations of ETA (see "Results"). specificity and efficiency of ETA in cleavage of Dsg1 depend on the enzyme's binding upstream of the cleavage site with the proper fit, like a key in a lock, to properly align, and perhaps activate, its catalytic site.

MATERIALS AND METHODS
Recombinant ETs-DNA encoding the ETD mature sequence was amplified by PCR using pQE-ETD-His (provided by Dr. Motoyuki Sugai, University of Hiroshima) as a template (11). PCR was also used to add nucleotides encoding the ETA signal peptide to the 5Ј-end of the ETD sequence and to add nucleotides encoding the V5 and His tags to the 3Ј-end of the ETD sequence. ETD-V5His was subcloned into the shuttle vector, pCE104 (12). pCE104 with ETD-V5His was electroporated into RN4220 S. aureus electrocompetent cells (12). We also used recombinant wild-type ETA, ETB, and ETAS195A (enzymatically active serine, at position 195 as defined by classical chymotrypsin serine protease numbering, mutated to alanine) with V5 and His tags as previously described (12). Recombinant ETs were purified on Ni 2ϩnitrilotriacetic acid columns (Qiagen, Valencia, CA) according to the manufacturer's protocol and then dialyzed against phosphate-buffered saline. Protein concentrations of ETs were estimated with a Protein Assay Kit (Bio-Rad) and on Simply Blue (a modified Coomassie G-250 solution; Invitrogen)-stained SDS-PAGE gels compared with standards and each other.
Production and ET Cleavage of Human Desmoglein 1-The entire EC domain of human recombinant Dsg1 with a His and E tag (hDsg1E) on the carboxyl terminus was produced as a secreted protein by baculovirus as previously described (22). For kinetic analysis, hDsg1E was purified from insect cell culture supernatant on an anti-E tag column (Amersham Biosciences) and eluted using 0.1 mg/ml E tag peptide in 20 mM Tris, pH 7.4, 500 mM NaCl, 1 mM CaCl 2 (TBS plus Ca 2ϩ ). The eluant was dialyzed against TBS plus Ca 2ϩ . 1 M purified hDsg1E (estimated with a protein assay kit (Bio-Rad)) was incubated with varying concentrations of ETA, ETB, and ETD for various times in TBS plus Ca 2ϩ at 25°C and then analyzed by SDS-PAGE stained with Simply Blue (Invitrogen). These gels were scanned for densitometric analysis by NIH Image software. The percentage of cleavage was calculated as a function of time, and resultant curves were analyzed by Igore Pro software (WaveMetrics, Inc., Lake Oswego, OR) (23).
1 M hDsg1E was chosen because of the previously estimated dissociation constant (K d ) of ϳ10 M, determined for the interaction of an inactive (active serine at position 195 mutated to alanine) ETA with hDsg1E (12). This value suggests a lower limit for the K m , and at a substrate (hDsg1E) concentration ϳ10-fold less than K m , the Michaelis-Menten equation predicts pseudo-first-order kinetics for any particular enzyme concentration, which was confirmed by the exponential decay of hDsg1E shown in Fig. 1.
Dsg1 Truncated Mutants and Chimeric Recombinants-PCR was used to construct cDNAs encoding various truncated hDsg1 molecules ( Fig. 3A) with E tags or E tag-glutathione S-transferase (GST) tags on their carboxyl termini. These recombinant constructs were subcloned into the baculovirus expression vector pEVmod (22). Various domainswapped chimeric molecules (Figs. 4A and 5A) between the EC domains of hDsg1 and human Dsg3 (hDsg3) (22) and between hDsg1 and mouse E-cadherin (24) were produced by PCR and subcloned into the pEV-EHis (22) cassette to add nucleotides encoding the E and His tags to their 3Ј-ends. cDsg1 cDNA (25) was cloned by reverse transcriptase-PCR using RNA from the canine cell line MCA-B1 (from Dr. Susumu Tateyama; University of Miyazaki) (26). cDNA for the EC domains of cDsg1 was subcloned into baculovirus expression vector pEV-EHis as previously described (27). cDNAs encoding truncated cDsg1 and domain-swapped chimeric molecules between the EC domains of cDsg1 and hDsg1 (Fig. 6C) were produced by PCR and subcloned into the pEV-E-GST cassette to add the E and GST tags to their carboxyl termini. The entire EC domain of hDsg3 with a His and E tag (hDsg3E) on the carboxyl terminus was produced as previously described (22). Site-directed mutagenesis of cDsg1 was performed with the Genetailor mutagenesis kit (Invitrogen) following the manufacturer's protocol to make mutants so that selected amino acids were mutated into hDsg1 equivalents (Fig. 7A). Recombinant baculovirus expression vectors were co-transfected with BaculoGold DNA (BD Biosciences) to Sf9 cells, and baculovirus in culture supernatants was used to infect High Five cells as previously described (22). Approximately, 10 nM recombinant secreted protein in the High Five culture supernatant was incubated with 1 M ETA for cleavage analysis. Degradation of hDsg1E was assayed by Western blotting with anti-E tag antibodies, as previously described (12).
Immunoprecipitation-As in methods previously described (12), supernatants of High Five insect cells transduced with baculovirus encoding hDsg1E, cDsg1E, or hDsg3E were incubated with ETAS195A and then used directly for immunoprecipitation. E tagged proteins were precipitated with anti-E tag-Sepharose (Amersham Biosciences) at 4°C for 1 h. Immunoprecipitates were washed 10 times with 1% Triton X-100 TBS plus Ca 2ϩ and then eluted with Laemmli sample buffer at 100°C.
Western Blotting-Proteins in Laemmli sample buffer were separated by 4 -20% SDS-PAGE (Bio-Rad) and then transferred to nitrocellulose sheets (Transblot; Bio-Rad). The sheets were incubated for 1 h in blocking buffer of 5% fat-free milk powder in phosphate-buffered saline. The E tag antibody conjugated with horseradish peroxidase (Amersham Biosciences) or anti-ETA sheep polyclonal antibody conjugated with horseradish peroxidase (Toxin Technology, Sarasota, FL), diluted in blocking buffer, was applied for 1 h at room temperature. After four washes with 0.1% Tween 20 in phosphate-buffered saline, the signals were detected with chemiluminescence (ECL or ECL Plus; Amersham Biosciences).

RESULTS
Kinetics of ET Cleavage of hDsg1-To investigate the efficiency of hydrolysis of Dsg1 by ETs, we measured hDsg1 cleavage by ETs as a function of time.
Purified recombinant ETA, ETB, and ETD used in these experiments showed a single band by Simply Blue staining of SDS-PAGE (Fig. 1A). Stained bands were quantified by densitometry, and the data were plotted as a function of time for analysis (Fig. 1C). Substrate consumption and product formation were fit to exponen- indicate presumptive calcium-binding amino acid residues. E381 indicates glutamic acid 381, after which ETs cleave. SP, signal peptide; Pro, propeptide sequence; E, E tag; GST, glutathione S-transferase tag. EGPYFPR and EPR are single letter amino acid codes (nonunderlined letters indicate normal hDsg1 sequence, and underlined letters indicate sequence of tag). ϩ, construct is cleaved by ETA; ϩ/Ϫ, diminished cleavage; Ϫ, no cleavage. B, Western blotting with anti-E tag of truncated hDsg1E incubated with (ϩ) or without (Ϫ) ETA. EC1-5 and EC1-4 show most efficient cleavage, but EC2-EC5 and EC2-4 also show some cleavage. EC3-EC5 was not cleaved by ETA. EC1-EC3 was no longer detected by anti-E tag after ETA incubation, suggesting that the E tag was cleaved but was not detected on the gel because of its small size. C, EC1-EC3 cleavage by ETA was confirmed with a longer tag consisting of both the E tag and GST tag on the carboxyl terminus of EC1-EC3. However, the deletion of 4 amino acids after the cleavage site (EC1-3⌬) eliminates its susceptibility to ETA. tial curves defined by Equations 1 and 2 to obtain k obs , the observed rate constant for substrate consumption or product formation as follows, where [S] represents the substrate concentration (as determined by densitometry) and [P] represents concentration of the products. Observed rate constants obtained from disappearance and formation data were in good agreement, as expected. For a given ET concentration, curves defined by first order kinetics (i.e. Equations 1 and 2) fit the data points well. This is best shown by the linear fit in Fig. 1D, in which the natural log of [S] is plotted against time. Measurements were made at three (for ETA) (Fig. 1D) or two (for ETB and ETD) ET concentrations. Values of k obs were directly proportional to ET concentrations, also consistent with pseudo-first-order reactions. The values for k cat /K m , which is a measure of enzyme efficiency, were obtained from the slopes of plots of k obs versus ET concentrations, according to Equation  and large (␣ 2 -macroglobulin) broad spectrum serine protease inhibitors to inhibit the enzymatic activity of ETA. There was minimal, if any, inhibition at the concentration tested, whereas both protease inhibitors did efficiently inhibit V8 protease at concentrations equivalent to those of ETA (Fig. 2), although an ϳ10-fold larger excess of DCI did show inhibition (data not shown). These data suggest that the catalytic site of ETA may be difficult to access in the proper alignment by some serine protease inhibitors or is relatively inactive in the absence of its proper substrate.
hDsg1 EC Domains That Are Upstream, but Not Downstream, of the Cleavage Site Are Critical for Its Hydrolysis by ETA-Previously, we have shown that the ability of ETs to cleave Dsg1 is not simply dependent on the amino acid sequence of the cleavage site but is dependent on the calciumstabilized conformation of Dsg1 (21). These results suggest that domains distant from the site of hydrolysis might influence the enzymatic efficiency of ETA. Because desmogleins have five well defined EC domains whose relationship to each other is stabilized by calcium, we first determined whether all of these domains were necessary for cleavage by testing truncated molecules of hDsg1E for their susceptibility to hydrolysis by ETA (Fig. 3, A and B). These data indicated that EC5 is not necessary for cleavage. Loss of EC1 decreases the efficiency of cleavage, yet the truncated molecule is still cleaved. Loss of EC1-EC2 prevents cleavage (Fig. 3B). After incubation with ETA, the EC1-EC3 domain can no longer be detected with anti-E tag by Western blotting (Fig. 3B), presumably because of cleavage of the E tag, which is too small to detect on these gels. To confirm the cleavage of EC1-EC3, a GST tag was added after the E tag on the carboxyl-terminal end. Incubation of this new FIG. 4. ETA cleavage of chimeric molecules between hDsg1 and hDsg3. A, schematic diagram of chimeric hDsg1 (white) and hDsg3 (gray) constructs. The numbers indicate amino acid residues counting from the initiating methionine as 1. B, Western blot with anti-E tag indicates that chimeric molecules containing hDsg1 residues 1-450 and hDsg1 residues 213-545 were cleaved, but those containing hDsg1 312-545 and hDsg1 1-309 were not. substrate, EC1-3EGST, with ETA clearly showed cleavage with the product now detectable (Fig. 3C). In this construct, the combination E and GST tag was added to hDsg1 only 4 amino acids after the cleavage site. Further analysis indicates that if these 4 amino acids were truncated, the construct was no longer susceptible to cleavage. These data show that cleavage by ETA is not dependent on intact domains EC4 and EC5.
Because truncations may result in a generally unstable overall structure, we used domain swapping between hDsg1E and hDsg3 to further test which domains in Dsg1 are critical (Fig.  4). These data suggest that sequences in hDsg1 between 213 and 450 are probably sufficient for cleavage. This conclusion was confirmed by insertion of just these sequences from Dsg1 into Dsg3 (Fig. 5). In fact, finer analysis indicated that only amino acids 214 -398 from the EC2-EC3 domain of Dsg1, inserted into Dsg3, were sufficient for cleavage (data not shown). Although sufficient in the overall structure of Dsg3, these amino acids must be in the proper overall conformation of a desmoglein, because when inserted into E-cadherin, they could no longer be hydrolyzed by ETA (Fig. 5B).
5 Amino Acids 110 Residues Upstream of the Cleavage Site in Dsg1 Are Necessary for Cleavage-To further define which particular amino acids in the EC2-EC3 region of hDsg1 are critical for cleavage, we used cDsg1 because it is highly homologous to human and mouse Dsg1 and shares identical amino acids around the cleavage site (Fig. 6A) yet is not hydrolyzed by ETA (Fig. 6, C and D). Swapping sequences of hDsg1 into cDsg1 near the EC2-EC3 junction allowed its cleavage (Fig. 6, C and E). We then determined which amino acids in this junction of cDsg1 were different from those in hDsg1 and mouse Dsg1, which, like hDsg1, is cleaved by ETs (12) (Fig. 6B). Because cDsg1 is highly homologous to the human and mouse Dsg1, there were only a limited number of candidate amino acids that might be critical to allow cleavage of cDsg1. Substitution of these amino acids from hDsg1 to cDsg1 allowed us to find 5 amino acids in hDsg1 that are critical to allow hydrolysis of cDsg1 by ETA (Fig. 7, A and B). These amino acids are about 110 residues upstream of the cleavage site (Fig. 7C).
ETA Binding and Ability to Cleave Are Independent in Canine Dsg1-We speculated that the inability of ETA to hydrolyze cDsg1 could be due to its inability to bind that substrate, as with hDsg3 (12), and substitution of amino acids from the hDsg1 sequence might restore binding and, therefore, cleavage. However, co-immunoprecipitation of an ETA with the active serine mutated to alanine with cDsg1 and hDsg1 showed that both could bind the enzyme. (A mutant must be used to prevent FIG. 5. Susceptibility to cleavage by ETA of domain-swapped molecules between hDsg1, hDsg3, and E-cadherin. A, schematic diagram of domainswapped molecules between hDsg1 (white), hDsg3 (gray), and mouse E-cadherin (E-cad) (dotted box). B, Western blotting with anti-E tag shows that hDsg3 with substitution of hDsg1 214 -450 and hDsg1 214 -398 (data not shown) were cleaved, whereas hDsg1 292-450 was not, suggesting that the junction of EC2-EC3 contains important sequences for susceptibility to cleavage by ETA. However, Ecadherin containing hDsg1 214 -450 was not cleaved, suggesting that E-cadherin does not provide the necessary overall conformation for cleavage.
hydrolysis of the hDsg1-positive control (12) (Fig. 8). These data suggest that cleavage of Dsg1 by ETA depends not only on binding but also on proper alignment and/or activation of the enzyme active site. These k cat /K m values are similar to those for the best tetrapeptide synthetic substrates for the acidic amino acid-specific endopeptidases S. aureus V8 protease and glutamic acid-specific Streptomyces griseus proteinase, which are 6 ϫ 10 3 and 9 ϫ 10 5 M Ϫ1 s Ϫ1 , respectively (28). These k cat /K m values are also greater than those obtained for the hydrolysis of the esterolytic substrate (N-t-Boc-L-Glu-␣-phenyl ester) by ETA and ETB, which were reported to range from 3 to 6 ϫ 10 3 M Ϫ1 s Ϫ1 (29). It also FIG. 6. ETA does not cleave cDsg1, but substitution of the EC2-EC3 junction from hDsg1 restores cDsg1 susceptibility to cleavage. A, amino acid sequence alignment of hDsg1, mouse Dsg1, and cDsg1 around the cleavage site of exfoliative toxins. The arrow indicates the cleavage site by ETs. Note the very well conserved sequence in hDsg1, mouse Dsg1, and cDsg1 around the cleavage site. B, amino acid sequence alignment of sequence upstream of the cleavage site. The underlined amino acids in cDsg1 show nonidentical sequence. C, comparison of hDsg1E (white), cDsg1E (gray), truncated cDsg1EC1-3EGST, and cDsg1EC1-3EGST substituted with human sequences. D, anti-E tag Western blotting shows that cDsg1E, like hDsg3E, is not cleaved by ETA. E, Western blotting with anti-E tag shows that substitution of hDsg1 amino acid residues 214 -334 in cDsg1 restored its susceptibility to cleavage by ETA, but substitution of hDsg1 amino acid residues 214 -257 in cDsg1 did not (data not shown). should be considered that the values we calculated for k cat /K m were for the hydrolysis of a large asymmetric protein, Dsg1. The interaction between two macromolecules is naturally going to be less efficient than that of an enzyme with a small substrate due to fact that only a relatively small area of each large protein is interactive (30 -32). For example, k cat /K m values for enzyme-small model peptide substrate reactions can be as high as 10 8 to 10 9 M Ϫ1 s Ϫ1 (33). The basal rate constant for productive collisions (i.e. encounters with the enzyme and substrate in the proper orientation for hydrolysis) between two large proteins in the absence of electrostatic effects is estimated at ϳ10 5 M Ϫ1 s Ϫ1 (30). The k cat /K m values for ET hydrolysis of Dsg1 approach this value. From these data, it can be appreciated that ETs are efficient enzymes. Our data also suggest that ETA is more efficient than ETB and ETD, although this observation may not reflect significant biological differences, since all of the enzymes are extremely efficient and pathological.
ETA Is Not Inhibited by Broad Spectrum Serine Protease Inhibitors-The broad spectrum serine protease inhibitors, DFP, DCI, and ␣ 2 -macroglobulin, do not efficiently inhibit ETA. These data suggest that somehow the active site of the enzyme is either inaccessible or not in the proper conformation to interact with these inhibitors and is consistent with the idea that only by binding to its physiologic substrate does the catalytic site of ETA become properly activated and/or aligned with the peptide bond in Dsg1 that it hydrolyzes. Consistent with this idea, some crystal structure analyses of ETA and ETB have suggested that their oxyanion hole is improperly formed and that the enzymes might require binding to the proper area to activate the catalytic site (8,29,34). However, one structural analysis of ETB suggests that the oxyanion hole is properly formed (35).
Amino Acids Over 100 Residues Upstream of the Cleavage Site in hDsg1 Are Required for Susceptibility to Cleavage by ETA-Previous data indicate that ETs are not simply sequence-specific proteases, like typical serine proteases such as trypsin and chymotrypsin, but are dependent on the calciumstabilized conformation of their substrate (21). These observations may account for their exquisite specificity for Dsg1. Since the EC domain of Dsg1 contains well defined subdomains, we determined which were necessary for susceptibility to cleavage by ETA. Interestingly, hDsg1 sequences in domains EC2-EC3, amino-terminal to the scissile bond, are necessary for cleavage, but cleavage is more efficient if they are in the overall structure of a desmoglein, and they are not sufficient for cleavage when inserted into the homologous region of E-cadherin. We also showed, using cDsg1, that within these domains there are 5 amino acids in hDsg1 that are critical to allow cleavage. Although highly homologous to hDsg1, cDsg1 is not cleaved by ETA, consistent with the observation that ETA does not induce blister formation in dog skin (2,36). Substitution of these 5 amino acids from hDsg1 allows its cleavage. The location of these amino acids, as determined on the homologous C-cadherin crystal structure (15), is on a loop proximal to the cleavage site (Fig. 7C), suggesting that this loop may be important in properly aligning the enzyme.
Key in Lock Model for Specificity and Efficiency of ETA-Our present data, together with previous data showing the importance of calcium-stabilized conformation of Dsg1 for its cleavage by ETA (21), suggest that the enzyme binds the substrate upstream of the cleavage site, which may align the catalytic site with the scissile bond. Furthermore, our data on resistance of ETA to serine protease inhibitors is consistent with an inactive or inaccessible catalytic site that we hypothesize is activated and/or made accessible to the scissile bond when the proper binding to the substrate occurs. Activation of ETA must require a very specific type of binding interaction, because although ETA binds to cDsg1 it does not cleave it. However, changing 5 amino acids in a loop just upstream of the cleavage site in cDsg1 allows its cleavage by ETA. The structure of this loop might allow a very specific "fit" that leads to enzyme activation and can, in this context, be thought of as similar to a key (ETA) in a lock (Dsg1).
Finally, our kinetic data indicate that either an efficient enzyme once aligned properly or, alternatively, electrostatic interactions between ETA and Dsg1 enhance the alignment beyond simple diffusion (30).
The key in lock model that explains the specificity of ETs is not unique. For example, it has also has been applied to the specificity of coagulation factors. One example is the specificity of thrombin for fibrinogen, another system involving the interaction between two large proteins (37). Thrombin exhibits a number of unique structural features that provide specificity. The deep cleft bordering the active site called the 60-insertion loop limits the accessibility of the active site. An exosite defined as a highly positively charged region away from the active site is complementary to a negatively charged region in fibrinogen. An aryl binding site requires apolar residues to occupy certain positions of substrates. Thus, the substrate must interact with all of these regions, like a key in a lock, to be hydrolyzed efficiently. Although our studies are not as detailed as those with thrombin, certain parallels can be made. Recognition of Dsg1 as a substrate is conformation-specific rather than sequence-specific and probably involves a specific exosite interaction between ETA and a region of extracellular domain 3 of Dsg1. Although the value of k cat /K m for hydrolysis of Dsg1 by ETA was not as high as that observed for thrombin hydrolysis of fibrinogen (38), Dsg1 is concentrated in vivo in desmosomes on cell surfaces. In these structures, hydrolytic rates may be much faster due to the effective concentration of the substrate.
In conclusion, ETs have evolved to efficiently and specifically hydrolyze one bond in one substrate through specific interactions with that substrate, depending on both the substrate's amino acid sequence and conformation at and distal to the cleavage site. The enzyme is thus able to efficiently and specifically target the exact molecule that allows S. aureus to spread under the stratum corneum, the major barrier of the skin.