Serine and Cysteine Proteases Are Translocated to Similar Extents upon Formation of Covalent Complexes with Serpins

CrmA is a “cross-class” serpin family inhibitor of the proapoptotic serine protease, granzyme B, as well as cysteine proteases of the caspase family. To determine whether crmA inhibits these structurally diverse proteases by a common conformational trapping mechanism, we mapped the position of the protease in crmA complexes with granzyme B or caspase-1 by fluorescence perturbation and fluorescence resonance energy transfer (FRET) analyses of site-specific fluorophore-labeled crmAs. A reactive loop P6 NBD label underwent similar large fluorescence enhancements (>200%) either upon reactive loop cleavage by AspN protease or complex formation with granzyme B or caspase-1, consistent with the insertion of the cleaved reactive loop into sheet A in both types of crmA-protease complexes. NBD labels on the noninserting part of the reactive loop docking site for protease (P1′ residue) or midway between the two ends of sheet A (helix F residue 101) showed no significant perturbations due to protease complexation. By contrast, labels at positions 68 and 261, lying at the end of sheet A most distal from the reactive loop, showed marked perturbations distinct from those induced by AspN cleavage and thus ascribable to granzyme B or caspase-1 proximity in the complexes. Substantial FRET between protease tryptophans and 5-dimethylaminonaphthalene-1-sulfonyl-labeled crmAs occurred in protease complexes with crmAs labeled at the 68 and 261 positions, but not the P1′ position. These results suggest that granzyme B and caspase-1 are inhibited by crmA by a common mechanism involving full reactive loop insertion into sheet A and translocation of the protease to the distal end of the sheet as previously found for inhibition of other serine proteases by serpins.

The serpins are an ever expanding superfamily of proteins that are found in all types of organisms and thus thought to have an early evolutionary origin (1,2). These proteins have in common a conserved ϳ350-residue core globular domain consisting of three ␤-sheets and eight to nine ␣-helices and are known to function in a vast array of physiologic processes (3). Most members of the family inhibit and thereby regulate the activity of specific proteolytic enzymes by forming stable stoichiometric complexes with the enzymes through an exposed reactive center loop. The mechanism of this inhibition is unusual from the point of view of it being dependent on conformational changes in the serpin and protease and involving a branched pathway leading to alternate substrate and inhibitor modes of reaction, i.e. the mechanism resembles that of a suicide substrate inhibitor (4). Another most unusual feature of serpins is their ability to inhibit proteases from more than one mechanistic or structural classes, a property that has been termed "crossclass inhibition" (5).
The first serpin shown to possess such cross-class inhibition was the cowpox virus serpin, cytokine response modifier A or crmA (6). CrmA was shown to inhibit the proinflammatory cysteine protease, interleukin-1␤ converting enzyme, also known as caspase-1, and subsequently other caspases in the apoptotic pathway, as well as the proapoptotic serine protease, granzyme B (7,8). This ability to inhibit two structurally and mechanistically distinct families of proteases was an unprecedented finding in view of the strict specificity toward a single mechanistic and structural class of protease that had been exhibited by other known protein protease inhibitor families (9). Subsequently, other serpins were shown to have this same ability to inhibit cysteine proteases of either the papain or caspase families and this ability was found to be engineerable into serpins by making the reactive loop sequence compatible with the substrate recognition determinants of a cysteine protease (10,11).
An outstanding question that remains to be answered is whether serpins utilize the same conformational change-dependent mechanism to inhibit both serine and cysteine classes of protease. In the established mechanism for serpin inhibition of serine proteases, the serpin and protease form an initial Michaelis-type complex that then converts to an acyl-intermediate complex in which the protease catalytic serine is covalently linked to the serpin P1 residue and the reactive loop is cleaved at the P1-P1Ј bond. The serpin subsequently undergoes a major conformational change wherein the N-terminal end of the cleaved loop inserts into the center of ␤-sheet A, dragging the tethered protease along ϳ70 Å to the opposite end of the molecule where the protease is inactivated by conformational deformation (12,13). Whether cysteine proteases are inactivated by the same conformational change-driven mechanism is a germane question because of some fundamentally different properties of the reactions of serpins with the two types of protease. Whereas serpin reactions with serine proteases typically result in the formation of equimolar SDS-stable complexes that represent stable acyl-intermediates, the reactions with cysteine proteases do not produce such SDS-stable complexes and are observable only under native nondenaturing conditions (5,8,10). Moreover, studies of the mechanism by which a serpin inhibits a nontarget cysteine protease revealed that inhibition arose from autolysis of complexed protease by free protease, suggesting a possible variation from the acyl-intermediate trapping mechanism demonstrated for serine protease inhibition (14).
In the present study, we have addressed the question of whether serpins inhibit cysteine proteases by the same conformational change-dependent mechanism established for inhibition of serine proteases. To do this, we have applied the fluorescence perturbation and FRET 2 approaches used in previous studies to map the position of the protease in a serpin-serine protease complex (15,16) to localize the protease in complexes of crmA with its serine and cysteine protease targets, granzyme B and caspase-1. This was accomplished by engineering single cysteine variants of crmA and labeling the variants with fluorophore probes to provide site-specific reporter groups for localization of the two target proteases in crmA-protease complexes. Our findings show that the protease is translocated to similar positions at the distal end of sheet A in stable crmA complexes with both granzyme B and caspase-1 and support similar conformational change-dependent mechanisms for serpin inhibition of the two classes of protease.

EXPERIMENTAL PROCEDURES
Proteins-Granzyme B was purified from cytotoxic T lymphocytes as described (17). Concentrations of the protease were determined by the bicinchoninic acid assay (Pierce Biotechnology) and confirmed to be in close agreement (Ͻ15%) with values determined from the 280-nm absorbance using an extinction coefficient calculated from the amino acid sequence (18). The stoichiometric inhibition of the enzyme by crmA and formation of an SDS-stable complex (see "Results") implied that the enzyme was fully functional. Caspase-1 was generously provided by Dr. Nancy Thornberry, Merck, Rahway, NJ, or the enzyme was expressed in Escherichia coli, refolded, and purified according to Thornberry et al. (19,20). Briefly, p10 and p20 subunits of caspase-1 were separately expressed in E. coli as inclusion bodies using plasmids supplied by Dr. Kathy McCusker (Merck) (20). The inclusion bodies were dissolved in 6 M guanidine-HCl and an equimolar mixture refolded in the presence of a YVKD-aldehyde affinity matrix (19,20) as previously described (21). Enzyme bound to the affinity matrix was eluted with free YVAD-aldehyde, activated with hydroxylamine, and finally converted to a glutathione disulfide adduct according to the protocol detailed previously (21). Enzyme was stored at Ϫ80°C in HSC buffer (0.1 M Hepes, 0.1% CHAPS, 0.1 mM EDTA, 20% sucrose, pH 7.5). Caspase-1 was activated just prior to use by incubation with HSC buffer containing 5-10 mM DTT. The concentration of functional caspase-1 was determined by active-site titration as described below. Recombinant wild-type crmA and a variant crmA in which all nine cysteines were mutated to serine were expressed in E. coli as inclusion bodies, refolded, and purified as in previous studies (22), with the inclusion of a final Mono-Q chromatography step to ensure the isolation of monomeric crmA free of polymers as described (23). Single-cysteine crmA variants were engineered by mutagenesis procedures identical to those used to create the variant with all cysteines mutated to serine as described (23). CrmA concentrations were determined from the absorbance at 280 nm using a calculated extinction coefficient of 31,900 M Ϫ1 cm Ϫ1 from the amino acid composition (18).
Experimental Conditions-Characterization of the kinetics and stoichiometry of caspase-1 reactions with crmA variants were conducted in HSC buffer containing 5 mM DTT at 25°C except where noted. Measurements of fluorescence emission spectra of labeled crmAs in the absence and presence of caspase-1 were performed in HSC buffer lacking sucrose but containing 5 mM DTT. All experiments characterizing the reactions of crmAs with granzyme B or with AspN protease were performed in 20 mM sodium phosphate, 0.1 M NaCl, 0.1 mM EDTA, 0.1% PEG 8000, pH 7.4, at 25°C.
Active-site Titration of Caspase-1-Solutions of YVAD-chloromethylketone (Bachem) at concentrations ranging from 0 to 100 nM were made up in HSC buffer lacking DTT by dilution from a stock solution in 0.1 mM HCl. Freshly activated caspase-1 was then added to a final concentration of 100 nM. The final DTT concentration in the incubation was 0.5 mM. The concentration of the stock chloromethylketone solution was determined from the tyrosine absorbance at 276 nm after dilution into HSC buffer without sucrose using an extinction coefficient of 1450 M Ϫ1 cm Ϫ1 . After incubation for 1-2 min, residual enzyme activity was measured by addition of YVAD-AMC substrate (25 M final concentration) in HSC buffer lacking sucrose and monitoring the initial linear rate of substrate hydrolysis fluorimetrically using excitation and emission wavelengths of 380 and 440 nm, respectively. Plotting of the residual activity versus inhibitor concentration and fitting to a straight line by linear regression analysis yielded the functional caspase-1 concentration from the end point on the abscissa axis. That the end point was not affected by any significant inactivation of the chloromethylketone by DTT during the time of reaction was shown by the finding that when similar reactions were conducted at twice the enzyme concentration to increase the reaction rate with the chloromethylketone titrant, indistinguishable results were obtained. Separate measurements of the inactivation of 1 nM caspase-1 by a large molar excess of chloromethylketone (10 nM) (pseudo-first order conditions) in caspase-1 buffer containing 0.5 mM DTT by quenching aliquots into substrate and measuring residual enzyme activity as above indicated a second-order rate constant of 1.3 ϫ 10 6 M Ϫ1 s Ϫ1 , consistent with completion of the reaction during the incubation time allowed. From the active-site concentration, a turnover number of 80 min Ϫ1 was measured for reaction of caspase-1 with 25 M YVAD-AMC in HSC buffer without sucrose at 25°C based on the caspase-1 (p20-p10) 2 tetramer concentration (two active-sites/tetramer), a value indistinguishable from that calculated from the published specific activity of active-site titrated caspase-1 with YVAD-AMC (20). The iodoacetamido-fluorophore dissolved in Me 2 SO was added to ϳ10 -20 M protein to a final concentration of 400 M (2% final Me 2 SO) and the reaction was allowed to proceed on ice for 4 h. The reaction was then quenched with 0.4 -1 mM DTT. Excess fluorophore was removed by gel filtering through a PD10 column followed by extensive dialysis against 20 mM sodium phosphate, 0.1 M NaCl, 0.1 mM EDTA, pH 7.4. In some cases, the labeled protein was further subjected to chromatography on Mono-Q according to the procedure used to purify unlabeled protein. The extent of fluorophore labeling was determined from the ratio of NBD or dansyl absorbance to the protein absorbance at 280 nm with the latter corrected for the absorbance contribution of the fluorophore. Extinction coefficients of 25,000 M Ϫ1 cm Ϫ1 for NBD at 478 nm and 5700 M Ϫ1 cm Ϫ1 for dansyl at 330 nm were used for these calculations.
Stoichiometry of CrmA-Protease Interactions-To a fixed concentration of protease (20 -200 nM) was added varying concentrations of crmA up to an inhibitor/protease ratio ranging from 1 to 20. Following incubation at 25°C for 60 min for granzyme B reactions and 5 min for caspase-1 reactions, residual enzyme activity was measured by addition of 25 M YVAD-AMC or WEHD-AMC substrates in HSC buffer without sucrose but containing 5 mM DTT (caspase-1 reactions) or 50 M IETD-AMC substrate in sodium phosphate buffer (granzyme B reactions). The initial rate of substrate hydrolysis was measured fluorimetrically using excitation and emission wavelengths of 380 and 440 nm, respectively. Turnover numbers for substrate hydrolysis were calculated by calibrating the fluorometer with known concentrations of AMC product.
Kinetics of CrmA-Protease Interactions-The kinetics of inhibition of caspase-1 by crmA were measured by continuous assays in which 0.1 nM caspase-1 was added to a solution of 5-500 nM crmA and 100 M WEHD-AMC substrate in HSC buffer plus 5 mM DTT and the decrease in the rate of substrate hydrolysis was monitored with time until a steady-state end point rate was reached. Inhibition progress curves were fit by a single exponential plus linear function to determine the observed pseudo first-order rate constant (k obs ) for the inhibition reaction. Values for k obs were corrected for the competitive effect of the substrate by multiplying by the factor, 1 ϩ [substrate]/K m , using a measured K m of 1.8 Ϯ 0.1 M for hydrolysis of the substrate under the same conditions (the K m determination employed substrate concentrations ranging from 0.3 to 30 K m ). Inhibition of granzyme B by crmA was measured by incubating 20 -100 nM enzyme with 500 -600 nM inhibitor and removing aliquots at different times into 100 M IETD-p-nitroanilide or 50 M IETD-AMC substrates for spectrophotometric or fluorometric measurement of residual enzyme activity from the initial rate of substrate hydrolysis at 405 or 440 nm emission/380 nm excitation, respectively. Inactivation progress curves were fit by a single exponential decay function with zero end point to obtain k obs . Second-order rate constants for association of crmA and enzymes (k a ) were obtained by dividing k obs by the crmA concentration.
Fluorescence Perturbation Studies-Perturbation of the fluorescence of NBD-labeled crmAs by complexation with proteases or by reactive loop cleavage with AspN protease (Sigma) was assessed by excitation of 200 -400 nM NBD-labeled crmA at 480 nm and measuring the emission spectrum between 500 and 600 nm in 5-nm steps (5-10-s integrations) with an SLM 8000 spectrofluorometer. The fluorescence changes induced by a single addition of a 7% molar ratio of AspN/crmA or two successive additions of a 1.2-fold and then a ϳ0.5-fold molar excess of granzyme B over crmA based on measured inhibition stoichiometries were continuously monitored at 540 nm to an end point value (1000 -2000 s for completion of either AspN or granzyme B reactions) after which the emission spectrum was measured. The effects of reaction with caspase-1 were determined by titrating in substoichiometric levels of the enzyme and monitoring the fluorescence change after each addition for several minutes to yield a stable end point value. After no further fluorescence change was induced, a final emission spectrum was taken. For labeled crmAs that did not undergo significant changes in fluorescence upon reaction with proteases, the time allowed for complete reaction was based on the measured association rate constants. All spectra in a given experiment were taken at least twice and averaged. Background signal was subtracted by taking a buffer spectrum and all spectra taken after treatment with proteases were corrected for dilution (Ͻ10%). Experiments with each labeled crmA and protease were performed two or more times and after normalizing spectra by a factor yielding a value of 1 for the maximum fluorescence emission of native crmA, the results of independent experiments were averaged.
FRET Studies-FRET between protease tryptophans (donor) and dansyl-labeled serpin (acceptor) due to complex formation was measured by taking emission spectra of 200 nM labeled serpin before and after protease treatment as in previous studies (15). To selectively excite protein tryptophan residues and minimize tyrosine excitation, an excitation wavelength of 292 nm was used and spectra were taken from 300 to 600 nm. To control for any fluorescence changes of the acceptor fluorophore due to the serpin conformational change induced by protease reaction, the effect of AspN cleavage of the dansyl-labeled serpins on the emission spectrum was determined. Dansyl-labeled crmAs were reacted with a single addition of 7% AspN, with two successive additions of a molar excess of granzyme B over the serpin or several additions of caspase-1 from substoichiometric to a molar excess over the serpin based on the measured stoichiometry of inhibition (SI) as in the experiments with the NBD-crmAs. The time dependence of dansyl fluorescence changes due to protease treatment was continuously monitored at 500 nm to ensure completion of the reaction after each addition. When minimal fluorescence change was observed, sufficient time was allotted for reactions based on measured association rate constants. Corrections were made for the background buffer signal and sample dilution. Each emission spectrum obtained before and after treatment with protease was taken at least twice and at least two independent experiments with each labeled crmA and protease were performed and the spectra averaged after normalization, as with the NBD-crmA spectra. The effect of direct perturbation of the dansyl fluorophore in labeled crmAs by protease complexation was evaluated by exciting at 330 nm and scanning the fluorescence emission spectrum from 400 to 600 nm.
Isolation of CrmA-Caspase-1 Complexes-Complexes of labeled crmAs with caspase-1 were formed by incubating 100 -200 pmol of caspase-1 in HSC buffer plus 5 mM DTT with a suffi-cient molar excess of inhibitor over protease to result in complete protease inactivation (ϳ10% greater than the SI ratio). The reaction mixture was then injected onto a Superdex 200 column equilibrated and run with HSC buffer without sucrose as described previously (21). Elution of crmA-protease complex and cleaved/ unreacted crmA peaks was monitored by protein fluorescence detection using excitation and emission wavelengths of 280 and 340 nm, respectively. Peaks were pooled and concentrated and emission spectra of equivalent concentrations of complex and cleaved crmA peaks were compared with unreacted labeled crmA based on NBD or dansyl absorbance. ϳ5 g from each of the two peaks were rerun separately to assess the extent of any cross-contamination between peaks. CrmA-caspase-1 complex peaks were found to be 70 -90% pure.

Strategy for Mapping the Protease Interaction Site in CrmA-Protease
Complexes-Single cysteine variants of recombinant crmA were engineered and site specifically labeled with fluorophore reporter groups for the purpose of mapping the position of the protease in crmA complexes with serine and cysteine proteases by fluorescence perturbation or FRET analyses. A set of five crmA variants were prepared by first mutating all nine cysteines of wild-type crmA to serine and then engineering single cysteines at positions 68, 101, 261, 298, and 304. The first three positions are equivalent to those previously used for labeling the serpin, ␣ 1 -proteinase inhibitor (␣ 1 PI), to map the site of protease interaction in ␣ 1 PItrypsin complexes (15,16). They are located on helix F (residue 101) and on two loops at the edge of sheet A most distal from the reactive loop, i.e. the strand 2A-helix E loop (residue 68) or the strand 5A-helix I loop (residue 261) (Fig. 1). The last two of the chosen positions, 298 and 304, correspond to the P6 and P1Ј residues of the protease recognition sequence in the reactive center loop of the serpin and are normally cysteine. Thus, the preparation of these variants required changing only eight of the cysteine residues of the wild-type serpin to serine. The chosen sites provided reporter groups at or midway between the two poles of the serpin where serine proteases have been localized in initial docking and final covalent serpin-protease complexes (Fig. 1).
Functional Properties of the CrmA Variants-The five single cysteine variants were all specifically labeled at near stoichio-FIGURE 1. Sites of fluorophore labeling in single cysteine variants of crmA viewed on the structures of the noncovalent ␣ 1 PI-S195A trypsin complex and the covalent ␣ 1 PI-elastase complex. The five amino acids of crmA that were changed to cysteine and labeled with NBD or DANS fluorophores are shown as the equivalent ␣ 1 PI residues in a ribbon diagram of the noncovalent ␣ 1 PI-S195A trypsin docking complex (PDB code 1OPH) and in the ␣ 1 PI-elastase covalent complex (PDB code 2D26) in space filling representation (red). The serpin is colored yellow and the protease in cyan. The N-terminal P1-P15 residues of the reactive loop is colored blue and the protease active-site serine is shown in space-filling representation in green. Residue 304, the P1Ј residue in the reactive loop, is not visible in the covalent complex structure and is depicted as a circle in its expected position. Position 298 of crmA corresponds to the P7 reactive loop residue of ␣ 1 PI that is exposed at the top of the serpin in the noncovalent complex and resides on strand 4 of sheet A in the covalent complex structure. This change in location results from the insertion of the reactive loop into the center of sheet A and the concomitant dragging of the covalently linked protease to the distal end of sheet A following cleavage of the P1-P1Ј bond and acylation of the P1 residue by the protease active-site serine. metric levels with iodoacetamido-NBD (IANBD) or iodoacetamidoethylamino-dansyl (IAEDANS) fluorophores (0.7-1 mol of fluorophore/mol of protein). The cysteine-free crmA variant retained an inhibitory activity toward the serine and cysteine protease targets, granzyme B and caspase-1, similar to that of the wild-type serpin (not shown), indicating that the cysteines were not important for activity. This finding was in keeping with the x-ray structure of the cysteine-free crmA variant showing that the modified cysteine residues are exposed on the protein surface and not in positions expected to form disulfide cross-links in the wild-type protein (22). Labeling of the single cysteine crmA variants with fluorophores did not affect their ability to function as inhibitors of granzyme B and caspase-1, but did increase the extent to which the crmAs reacted as substrates of these proteases in most cases. Granzyme B reactions were the least affected by labeling of the crmA variants, notable exceptions being the 298 cysteine variant that showed both a ϳ4-fold increase in the stoichiometry of inhibition and a comparable decrease in the association rate constant (k a ) and the 304 cysteine variant that showed a decrease in k a of ϳ10and ϳ2-fold increase in inhibition stoichiometry. The reduction in k a for the latter labeled variant is consistent with the expected steric interference of the P1Ј label with the initial docking of the protease with the serpin reactive loop as was found previously for the reaction of a P1Ј-NBD plasminogen activator inhibitor-1 (PAI-1) with tissue plasminogen activator (26). More pronounced effects of the labeling were found on the interaction of the crmA variants with caspase-1. Whereas the SI was increased modestly for the 101 NBD and 304 NBD variants, it was increased somewhat more for the 261 NBD variant and substantially more (ϳ10-fold) for the 68 NBD variant, indicating that the labeling enhanced a substrate mode of reaction of the crmAs with caspase-1, especially for the 68 NBD variant. The increases in SI values were paralleled by comparable decreases in k a values for reaction of the labeled crmA variants with caspase-1. As a result, similar values were found for the product of k a and SI, reflecting the corrected rate of reaction through the inhibitory pathway (4). This finding suggested that the principal effect of the labeling was to slow the rate of the conformational change that stabilizes and kinetically traps the acyl-intermediate complex, thereby increasing the fraction of the acyl-intermediate that is able to deacylate and complete the substrate reaction cycle.
SDS-PAGE analyses of the reactions of unlabeled wild-type crmA and NBD-labeled crmA variants with a molar excess of granzyme B and caspase-1 showed that all crmAs were fully reactive with the two proteases, the crmA bands being completely converted to high molecular weight complexes and cleaved serpin in the case of granzyme B or just to cleaved serpin in the case of caspase-1 (Fig. 2). All crmAs were fully transformed to an indistinguishable cleaved form by treatment with catalytic levels of AspN protease (not shown), presumably reflecting specific cleavage of the P1 Asp residue in the reactive loop of the crmAs. The failure to observe high molecular weight covalent complexes of the crmAs with caspase-1 is consistent with previous reports that crmA-caspase-1 complexes are dissociated when denatured in SDS (5). Complex bands were observable for crmA-caspase-1 reactions on native PAGE gels in all cases in keeping with previous reports (5). Together, these results indicated that the cysteine-free unlabeled crmA and all single cysteine fluorophore-labeled crmAs were fully functional with the two target proteases.
Fluorescence Changes Accompanying NBD-labeled CrmA Interactions with Proteases-To map the position of the protease in crmA complexes with granzyme B and caspase-1, the NBD-labeled crmAs were reacted with levels of the proteases sufficient to cause complete reaction of the crmA and the changes in the fluorescence emission spectrum of the NBD reporter group was determined. The fluorescence changes resulting from protease cleavage at the P1 Asp bond were also directly evaluated by treating the crmAs with AspN protease. Reactions of the NBD-labeled crmAs with proteases were followed by continuously monitoring changes in fluorescence until an end point was achieved or allowing sufficient time for complete reaction based on measured association rate constants. The 101 NBD and 304 NBD crmAs showed minimal or no significant changes in the NBD fluorescence emission spectrum upon reaction with granzyme B or caspase-1 as was also the case for the reaction with AspN protease (Fig. 3). In contrast to these results, the perturbations in the fluorescence of NBD at the 298 position induced by the proteases were quite substantial, although similar to the changes induced by AspN cleavage (Fig. 3). The NBD fluorescence emission was thus enhanced from 3-to 4-fold by reaction with either the two target proteases or AspN, indicating that these changes were largely the result of the reactive loop conformational change induced by protease cleavage in the loop. In this conformational change the N-terminal end of the cleaved loop inserts into the center of ␤-sheet A, causing residue 298 to become buried beneath helix F, as is evident from the x-ray structure of cleaved crmA (22). The fluorescence changes are similar to those found with other serpins labeled with NBD in this or an adjacent res- idue in the loop and can be accounted for by the burial of the solvent-exposed NBD fluorophore attached to residue 298 in native crmA following the cleavage-induced conformational change (24,25). Small differences in the magnitude of the NBD fluorescence enhancement result from the nature of the attached protease, in keeping with previous findings (26). Significant changes in NBD fluorescence were also induced by protease reaction with crmAs labeled in the 68 and 261 positions at the distal end of sheet A, but unlike the changes induced in the 298 variant, these changes were dissimilar to the changes induced by AspN cleavage in the reactive loop (Fig. 3). The NBD fluorescence emission spectrum was enhanced 1.2-1.5fold and significantly blue-shifted upon reaction of the 261 NBD crmA with granzyme B and caspase-1, whereas reaction with AspN resulted in a quenching of the emission spectrum and no significant spectral shift. Reaction of 68 NBD crmA with granzyme B resulted in a ϳ2-fold enhancement of the emission intensity and a blue-shift. By contrast, reaction with AspN or caspase-1 resulted in smaller 1.3-1.7-fold enhancements in the fluorescence emission spectrum without any significant shift in the emission maximum. That the spectral changes accompanying reaction of 68 NBD crmA with caspase-1 principally reflected cleavage at the P1-P1Ј bond and minimally the formation of a crmA-caspase-1 complex was suggested by the large SI observed for the reaction of 68 NBD crmA with caspase-1. The fluorescence changes resulting from caspase-1 or AspN reaction therefore reflect similar perturbations of the fluorophore at residue 68 due to the reactive loop conformational change with the different intensity changes accounted for by the different sites of cleavage in the crmA reactive loop and consequent shorter length of the inserted loop for AspN (P1-P2) versus caspase-1 (P1-P1Ј) cleavage.
Because of the significant contribution of a substrate mode of reaction of the 68 NBD and 261 NBD crmAs with caspase-1, most pronounced for the 68 NBD crmA reaction, it was necessary to separate the crmA-protease complex from cleaved crmA to resolve the effects of direct protease perturbation of the NBD fluorophore from those due to the global conformational change induced by reactive loop cleavage. This separation was accomplished by exploiting the difference in molecular size between the free and complexed serpin through chromatography on a Superdex 200 size-exclusion column (21). Equivalent concentrations of the crmAcaspase-1 complex and cleaved crmA peaks from the chromatography separation were compared with unreacted crmA based on NBD absorbance. The identity of these peaks as protease-complexed and cleaved crmA has been demonstrated in recently published work (21). The 261 NBD crmA-caspase-1 complex showed a large ϳ3-fold enhancement and blue-shift in the NBD emission spectrum relative to the unreacted crmA, qualitatively similar to the changes induced in the labeled crmA upon reaction with caspase-1. By contrast, the isolated cleaved crmA peak was marginally perturbed from the native crmA spectrum, similar to the effects of AspN reaction. Likewise, the 68 NBD crmAcaspase-1 complex peak displayed an emission spectrum that was enhanced ϳ1.8-fold and blue-shifted, similar to the changes observed upon reaction of the labeled crmA with granzyme B. By contrast, the cleaved crmA peak showed an enhancement in emission intensity but no blue-shift, similar to the changes observed with AspN cleavage and those induced by the predominant substrate reaction with caspase-1. The isolated complexes of other labeled crmAs with caspase-1 showed emission spectra not distinguishably different from the spectra of their cleaved crmA peaks (not shown), in keeping with the similar effects of AspN and protease reactions with these labeled crmAs on their NBD emission spectra. These results indicated that granzyme B and caspase-1 induced NBD fluorescence changes in the 68 NBD and 261 NBD crmAs that could not be accounted for by the reactive loop conformational change and which therefore must have resulted from direct perturbation of the fluorophore by the complexed proteases due to their proximity. Fluorescence emission spectra of NBD-labeled crmA variants are shown before (solid line) and after complete reaction with catalytic AspN protease (long dash-short dash) or with a molar excess of granzyme B (short dash-dot) or caspase-1 (dotted) based on measured SIs. For 68 NBD crmA and 261 NBD crmA variants, crmA-caspase-1 complexes were resolved from cleaved crmA generated in the reaction by Superdex chromatography and the observed spectrum of an equivalent concentration of the isolated complex is shown (solid circles). Reaction conditions and details on acquisition of spectra are given under "Experimental Procedures." Fluorescence changes were monitored in each case to an end point value. Each spectrum except that for the isolated complex represents the average of spectra obtained from two or more experiments with error bars representing the range of values or standard deviations. All spectra were normalized using a factor that yielded a value of 1 for the maximum fluorescence of native crmA.
FRET Mapping of the Protease Interaction Site in CrmA-Protease Complexes-To confirm the results of the fluorescence perturbation analyses, we sought to provide independent FRET analyses that localized the interaction sites for the protease in crmA complexes with granzyme B and caspase-1. The three crmA variants with cysteines residing at opposite poles of the serpin, i.e. the 68 Cys, 261 Cys, and 304 Cys variants, were chosen for these analyses and labeled with an IAEDANS fluorophore to determine whether FRET could be detected between the protease tryptophans (donor) and the serpin dansyl fluorophore (acceptor). Stoichiometries of inhibition were increased to a similar extent by dansyl labeling as they were by NBD labeling ( Table 1) especially in the case of caspase-1 reactions, again necessitating the chromatographic separation of the crmAcaspase-1 complex from cleaved crmA to resolve the effects of protease binding from those of cleavage. Granzyme B produced substantial FRET when complexed with the 68 DANS crmA variant, increasing the dansyl fluorescence emission spectrum intensity ϳ4-fold and inducing a blue-shift in the spectrum. In contrast, AspN cleavage caused a quenching of dansyl fluorescence (Fig. 4). Whereas no FRET was detected when the 68 DANS crmA was reacted with caspase-1, as anticipated due to cleaved crmA being the dominant product of the reaction, a substantial FRET enhancement of Ͼ2-fold and blue-shift of the spectrum was observed for the isolated 68 DANS crmAcaspase-1 complex as compared with isolated cleaved 68 DANS crmA or unreacted native 68 DANS crmA. The 261 DANS crmA variant similarly showed substantial FRET as evidenced by 2.5-3.5-fold enhancements and blue-shifts upon reaction with granzyme B or when comparing the isolated complex formed with caspase-1 with unreacted crmA, whereas AspN treatment caused a modest quenching of the dansyl fluorescence (Fig. 4). Similar FRET but of reduced magnitude was observed when the 261 DANS crmA was reacted with caspase-1 without separation of the reaction products, consistent with the expected amount of cleaved crmA formed in the reaction. That the observed fluorescence changes in the 68 DANS and 261 DANS crmA complexes with granzyme B and caspase-1 reflected FRET between protease tryptophans and the crmA dansyl group was supported by the finding that no FRET was observable when the 304 DANS crmA was reacted with granzyme B or the complex with caspase-1 was isolated nor were any changes detected when the labeled crmA was cleaved by AspN protease (Fig. 4). Moreover, direct excitation of the dansyl fluorophore in 68 DANS and 261 DANS crmAs showed blue-shifts of the dansyl emission spectra in crmA-caspase-1 complexes relative to the unreacted labeled crmAs but minimal intensity changes (Ͻ10%) due to proximity effects of the complexed protease. These results are in agreement with the fluorescence perturbation studies and support the conclusion that the protease is localized in similar positions at the distal end of sheet A in crmA complexes with both serine and cysteine proteases.

DISCUSSION
These studies were undertaken to determine whether serpins inhibit cysteine proteases by the same translocation and deformation mechanism that has been demonstrated for inhibition of serine proteases (12,13). That there might be significant differences in the mechanism of inhibition of these two mechanistically distinct classes of protease has been suggested by the fact that serpin complexes with cysteine proteases are not SDS-stable and therefore may not be stabilized as acyl-intermediate complexes as is the case with serpin complexes with serine proteases (5,8,10). Indeed, examples of serpins inhibiting proteases through a noncovalent lock FIGURE 4. FRET analysis of dansyl-labeled crmA-protease complexes. FRET between protease tryptophans (donor) and crmA dansyl reporter groups (acceptor) was assessed from increases in dansyl fluorescence emission resulting from reaction of the labeled crmA with protease when protease tryptophans were excited at 292 nm. Dansyl emission spectra are shown of labeled native crmA before (solid line) and after complete reaction with granzyme B (short dash-dot) or with caspase-1 (dotted). Intramolecular FRET between serpin tryptophans and the dansyl label resulting from the serpin conformational change induced by reactive loop cleavage was tested by analyzing changes in dansyl emission resulting from AspN cleavage in the reactive loop (long dashshort dash). For all three labeled variants analyzed, crmA complexes with caspase-1 were resolved from cleaved crmA generated in the reaction by Superdex chromatography and the emission spectrum of the isolated complex was determined at an equivalent dansyl concentration (solid circles). Other details are provided under "Experimental Procedures." Spectra were normalized by a factor that gave a value of 1 for the maximum dansyl fluorescence of native crmA. Error bars represent the range or standard deviation from at least two experiments. and key type interaction that does not involve proteolysis or triggering of the serpin conformational change have been demonstrated (3). A variant type of mechanism for inhibition of cysteine proteases without trapping the acyl-intermediate has also been proposed in the case of a serpin inhibiting a nontarget cysteine protease (14). That crmA possesses the machinery for conformational trapping of the acyl-intermediate is suggested from the ability of the serpin to inhibit granzyme B, or of P1 arginine variants of crmA to inhibit thrombin and factor Xa, by forming SDS-stable complexes (7,23). Moreover, two x-ray structures of reactive loop-cleaved forms of crmA confirm that crmA is capable of undergoing the serpin conformational change that is an essential part of the acyl-intermediate trapping mechanism (22,27). The signature feature of this conformational change, the insertion of the cleaved reactive loop as a central strand into the major ␤-sheet, sheet A, is thus clearly evident from these structures. However, the ability to undergo this conformational change is by itself insufficient evidence that a serpin shows inhibitory function as is evident from studies of several noninhibitory serpins (3). That the conformational change plays a role in crmA inhibition of its cysteine protease targets, the caspases, is suggested by the loss of inhibitory function in a crmA variant in which the conformational change was blocked by mutation of the P14 hinge residue from Thr to Arg (28). Our use of the fluorescence perturbation and FRET mapping approaches, applied successfully in previous studies to localize the protease in the ␣ 1 PI-trypsin complex (15,16) and in PAI-1serine protease complexes (29) provided a direct means of assessing whether crmA uses a similar conformational trapping mechanism to inhibit the principal cysteine protease target, caspase-1. Because crmA is also an inhibitor of the serine protease, granzyme B, and forms an SDS-stable complex with the enzyme, a comparison of the complexes formed with both proteases provided an opportunity to assess the similarities and differences between crmA complexes with cysteine and serine proteases. Our findings clearly support the utilization of a protease translocation mechanism for crmA inhibition of both granzyme B and caspase-1 that resembles the mechanism shown for serpin inhibition of serine proteases (3). The two positions at the distal end of sheet A that were shown to be directly perturbed by protease and to show the largest FRET between protease and serpin in the ␣ 1 PI-trypsin complex were found to be similarly perturbed in crmA complexes with granzyme B and caspase-1 and to show substantial FRET. By contrast fluorophores placed in the reactive loop docking site or midway between this docking site and the distal end of sheet A showed no direct fluorescence perturbation or FRET by either of the proteases examined. Greater FRET was observed in the crmA-granzyme B complex when the dansyl fluorophore was at position 68 than at position 261, whereas the opposite was true for the crmA-caspase-1 complex or for the ␣ 1 PI-trypsin complex in previous studies (15). However, the dependence of FRET on the sixth power of the distance of separation of donor and acceptor fluorophores together with the different number and location of acceptor tryptophans in the proteases implies that these observed differences in FRET cannot represent large differences in the positioning of the proteases at the distal end of sheet A of the serpin. Our recent finding that in the crmAcaspase-1 complex the p10 subunit is dissociated also does not affect the conclusions drawn about the location of caspase-1 in the complex because the active site-containing p20 subunit remains covalently linked to crmA and responsible for the observed fluorescence perturbation and FRET in the labeled crmA-caspase-1 complexes (21).
The effects of AspN cleavage of the labeled crmA variants provided a means of distinguishing fluorescence changes due to reactive loop insertion into sheet A and the consequent global changes in serpin conformation associated with this insertion from those caused by direct perturbation of the fluorophore due to the proximity of the protease (15,16,29). Similar to past studies, the reactive loop 298 residue, which represents P6 in crmA but is P7 in most other serpins due to the shorter crmA reactive loop, was most responsive to this conformational change as a result of its burial beneath helix F when the cleaved loop inserts into sheet A (24). The similar changes induced in this label in the two crmA-protease complexes confirm that the inhibition of these two proteases involves reactive loop cleavage and insertion into sheet A in a manner similar to that induced by the nontarget protease, AspN. However, reproducible differences in the intensity change induced by the proteases and AspN imply that the nature of the protease tethered to the P1 residue at the end of the inserted loop can affect the conformational change-induced fluorescence perturbation. Whereas the labeling of the crmAs did not prevent them from acting as inhibitors of granzyme B and caspase-1, there were significant effects of the labels on the fraction of the serpin reacting as a substrate that differed for the two target proteases. The two positions at the distal end of sheet A that are the closest to the protease in both crmA-protease complexes significantly increased the fraction of crmA reacting as a substrate in the reaction with caspase-1, whereas most labeled positions produced comparatively modest increases in a substrate mode of reaction with granzyme B. Such findings support the overall conclusions of this study that caspase-1 is translocated to the distal end of sheet A and is trapped as an acyl-intermediate through active-site distortion because the labels at the distal end of the sheet most markedly interfere with the distortion and allow a greater fraction of the acyl-intermediate to deacylate. PAI-1 labeled at the same position also showed enhanced SI values with proteases, in particular with thrombin (29). The smaller effects of the labels on the conformational trapping of granzyme B is similar to what was found for the reaction of trypsin with ␣ 1 PI labeled at these positions, implying that the different effects result from a more favorable rate of conformational trapping of the protease relative to the rate of deacylation for these serpin-protease pairs. Interestingly, unlabeled crmA forms complexes with caspase-1 that are markedly less stable than the complexes formed with granzyme B based on the much higher rates of dissociation of caspase-1 than granzyme B complexes (21,23). This may be due to effects of the distortion on the noncovalent interfaces in the caspase-1 (p20-p10) 2 tetramer, which are weakened in the complex such that dissociation of the p10 subunit is promoted (21). Further studies will be required to determine the extent to which active-site distortion and the weakening of subunit interface contacts account for the reduced stability of the crmA-caspase-1 complex.
If serpins trap cysteine proteases by a similar conformational trapping mechanism involving protease translocation, then it becomes important to explain why these complexes dissociate during SDS-PAGE unlike serpin complexes with serine proteases. A reasonable explanation is the greater susceptibility of the thiol ester linkage between serpin and cysteine protease in the acyl-intermediate to hydrolysis as compared with the serine ester linkage formed with a serine protease (10). Under native conditions the thiol ester may be stabilized by its lack of exposure to water and solvent nucleophiles. Denaturation by SDS and boiling fully exposes this bond that would be expected to be rapidly hydrolyzed by nucleophiles present in standard SDS-PAGE buffer systems. Indeed, the baculovirus p35 inhibitor of caspases forms an SDS-stable covalent thiol ester complex with caspase-8 under milder denaturing conditions of lower pH and temperature (30), implying that conditions that stabilize the thiol ester could allow visualization of a crmA-caspase-1 complex. However, even employing such milder conditions of denaturation, we have been unable to prevent dissociation of crmAcaspase-1 complexes, implying a more labile thiol ester bond in these complexes. The existence of a labile covalent thiol ester bond in crmA-caspase-1 complexes is supported by our finding that these complexes are rapidly dissociated by the nucleophile, hydroxylamine, as are serpin complexes with serine proteases, but the kinetic stability of the complexes under physiologic conditions is considerably less than that of serpin-serine protease complexes (21).