The Effects of Reactive Site Location on the Inhibitory Properties of the Serpin α1-Antichymotrypsin*

The large size of the serpin reactive site loop (RSL) suggests that the role of the RSL in protease inhibition is more complex than that of presenting the reactive site (P1 residue) to the protease. This study examines the effect on inhibition of relocating the reactive site (Leu-358) of the serpin α1-antichymotrypsin either one residue closer (P2) or further (P1′) from the base of the RSL (Glu-342). α1-Antichymotrypsin variants were produced by mutation within the P4-P2′ region; the sequence ITLLSA was changed to ITLSSA to relocate the reactive site to P2 (Leu-357) and to ITITLS to relocate it to P1′ (Leu-359). Inhibition of the chymotrypsin-like proteases human chymase and chymotrypsin and the non-target protease human neutrophil elastase (HNE) were analyzed. The P2 variant inhibited chymase and chymotrypsin but not HNE. Relative to P1, interaction at P2 was characterized by greater complex stability, lower inhibition rate constants, and increased stoichiometry of inhibition values. In contrast, the P1′ variant inhibited HNE (stoichiometry of inhibition = 4) but not chymase or chymotrypsin. However, inhibition of HNE was by interaction with Ile-357, the P2 residue. The P1′ site was recognized by all proteases as a cleavage site. Covalent-complexes resistant to SDS-PAGE were observed in all inhibitory reactions, consistent with the trapping of the protease as a serpin-acyl protease complex. The complete loss in inhibitory activity associated with lengthening the Glu-342-reactive site distance by a single residue and the enhanced stability of complexes associated with shortening this distance by a single residue are compatible with the distorted-protease model of inhibition requiring full insertion of the RSL into the body of the serpin and translocation of the linked protease to the pole opposite from that of encounter.

Proteins have evolved a number of different mechanisms to inhibit serine proteases (1,2). A feature common to all mechanisms is recognition of the inhibitor by the protease in a manner resembling that of a substrate. That is, S1-P1 interactions, according to standard nomenclature (3), determine the site of encounter and the specificity of the interaction. The P1 residue of a protease inhibitor is termed the reactive site. The reactive site is located within a specialized region of the inhibitor termed the reactive site loop (RSL) 1 or bait region in the case of ␣ 2 -macroglobulin.
Serpins possess a relatively large reactive site loop of ϳ22-26 residues (4). Despite the size, the reactive site appears limited to a position 16 or 17 residues from the N-terminal base of the RSL, marked by the highly conserved residue Glu-342 (4,5). The recent crystal structure of a trypsin-␣ 1 -proteinase inhibitor (␣1-PI) complex suggests an explanation for this restriction. The crystal structure showed the protease ϳ70 Å from the presumed point of encounter and residues P17-P3 of the RSL inserted into the face of the inhibitor as the 4th strand of a 6-strand A-␤-sheet structure (6). At the interface region between the two proteins, the P1 residue of the serpin was covalently tethered to the protease through an ester bond with Ser-195. Active site structure, especially around the P1 residue of the serpin, was highly distorted, giving rise to the impression that the protease was trapped in a deformed-uncatalytic state by compression against the serpin. These observations imply that a length of about 17 residues was required to translocate the covalently tethered protease to the opposite end of the inhibitor where it was denatured by being forced against the rigid body of the serpin.
Although a number of spectroscopic studies using site-specific fluorescent labels support the location of the inhibited protease at the pole opposite from that of encounter (7,8), other studies employing fluorescent labels (9), monoclonal antibodies (10,11), and synthetic peptides resembling the RSL (12) place the enzyme closer to, if not at, the site of encounter. The latter studies suggest that only partial insertion of the RSL is required for inhibition. Among this group is a spectroscopic study similar to those that find the protease at the pole opposite of encounter (9). As suggested by O'Malley and Copperman (9), controversy over protease location and the extent of RSL insertion raises the possibility for multiple sites on the serpin capable of mediating inhibition.
Partial insertion models cannot easily explain the need for a large RSL and the restricted location of the reactive site within the RSL. Also not as evident from partial insertion models is why interaction at sites in the vicinity of P1 (typically P7-P3) produce cleavage and irreversible inactivation of the serpin instead of inhibition (13)(14)(15)(16)(17). To define the inhibition/substrate properties of the RSL more precisely, Zhou et al. (18) moved the reactive site of ␣1-PI Pittsburgh 1-3 residues closer or further from the beginning of the RSL (Glu-342). The results demonstrated a sharp change in inhibition/substrate properties depending on the direction of the change and the distance from Glu-342 (18). These findings were interpreted as support for the full-insertion protease distortion mechanism implied by the crystal structure (6).
In this study, the effect on inhibition of relocating the reactive site of the serpin ␣ 1 -antichymotrypsin (ACT) one residue closer or further from Glu-342 was evaluated using three different proteases. The three enzymes, bovine chymotrypsin, human chymase, and human neutrophil elastase (HNE) were studied because each emphasizes a different aspect of proteaseserpin interactions. Chymotrypsin is inhibited rapidly by ACT but forms relatively unstable complexes (19). Analysis of the chymotrypsin-ACT complex located the protease near the site of encounter (9), suggesting that inhibition of chymotrypsin by ACT may have RSL requirements different from those described by Zhou et al. (18). The chymotrypsin-like protease, chymase, forms a stable complex with ACT but is inhibited slowly and with an unusually high apparent SI of 4 (20,21). HNE is not inhibited by ACT even though the P1 residue is Leu-358. Rather, HNE cleaves the RSL of ACT at several different positions near the P1 site, thereby inactivating the inhibitor (17). We show that changing the reactive site by one residue in either direction has profound and different effects on all aspects of serpin protease reactions. The patterns observed were similar to those described by Zhou et al. (18), suggesting that full insertion is a general requirement of protease inhibition by serpins.

EXPERIMENTAL PROCEDURES
Materials-Peptide NA substrates were obtained from Bachem (Philadelphia, PA). HNE was obtained from Elastin Products (St. Louis, MO) or Athens Research Products (Athens, GA), and chymotrypsin was from Worthington (Lakewood, NJ). Fast Q-, SP-, and heparin-Sepharose were obtained from Amersham Biosciences. Gelcode Blue was from Pierce.
Recombinant Production of Serpins and Proteases--ACT and ACT variants were expressed in Escherichia coli and purified from the cellular supernatants (19). Variants were produced by the overlap-extension PCR method of Ho et al. (22). To purify ACTs, supernatants were separated from cellular debris by centrifugation and then applied to a 50-ml Fast Q-Sepharose column equilibrated in 10 mM MOPS (pH 7.5), 0.05 M NaCl. The column was washed, and the serpin was eluted with a gradient of 0.05-0.4 M NaCl. All ACTs eluted at ϳ0.3 M NaCl. ACTs were identified by SDS-PAGE or by the inhibition of chymotrypsin. Fractionated ACT was then dialyzed overnight against 0.05 M NaCl, 10 mM MOPS (pH 6.8) and applied to a 50-ml SP-Sepharose column. The column was eluted with a gradient of 0.05-0.4 M NaCl. ACT eluted at 0.2 M NaCl. Fractions containing ACT were pooled, and solid NaCl was then added to a final concentration of 1.0 M NaCl. The preparation was then concentrated by pressure dialysis, dialyzed against 1.0 M NaCl, 0.01 M MOPS (pH 6.8), and stored at Ϫ80°C. Human chymase was expressed as a pseudo-zymogen using a baculovirus-insect cell system and then purified as previously described (23). The primary structure of chymase differed from that of native chymase by three residues in positions that do not affect biochemical or structural properties (23,24).
Characterizations of Serpin-Protease Interactions-SI values were determined by titration of protease activity. Titrations were performed by the addition of increasing amounts of inhibitor to a fixed amount of protease. Protease-inhibitor reactions were in a total volume of 25-50 l of solution containing 0.5-1.0 M protease. Aliquots of 1-5 l were typically removed for assay in 250 -500 l of substrate solution. The amount of protease in each aliquot was sufficient to produce a ⌬A 410 nm / min of 0.1-0.15 in the absence of inhibitor. Reactions of inhibitors with chymase, chymotrypsin, and HNE were performed at 25°C in 1.0 M NaCl, 0.1 M Tris-HCl (pH 8.0), 0.01% dodecyl maltoside, 0.5 M NaCl, 0.1 M Tris-HCl (pH 8.0), 0.01% dodecyl maltoside, and 0.5 M NaCl, 0.1 M HEPES (pH 7.5) or 0.1 M Tris-HCl (pH 8.0), 0.01% dodecyl maltoside, respectively.
Inhibition rate constants were measured under pseudo-first order conditions, where [I] 0 ϭ [E] 0 ϫ SI ϫ 10. Measurements were made using the progress curve method (25) or by monitoring the time-dependent loss of activity. Incubation conditions were the same as those of titrations except for the presence of substrate (and final 9% Me 2 SO) in progress curve measurements. Chymase was measured using 0.5 mM Suc-AAPF-NA (K m ϭ 0.8 mM) or 0.05 M Suc-AEPF-7-amido-4-methylcoumarin (K m ϭ 0.5 mM). Chymotrypsin was measured using 0.1 mM Suc-AAPF-NA (K m ϭ 0.05 M). Elastase was measured in 1 mM MeO-Suc-AAPV-NA (K m ϭ 0.15 M). In the second method, used only for chymotrypsin, aliquots were removed, and residual activity was measured by dilution into assay media containing a saturating concentration of substrate.
Rate constants for complex breakdown were determined by following the return of enzyme activity at 25°C. Complexes were formed at high protease and serpin concentrations with the serpin in excess. Thus complex formation was rapid, and virtually all free enzyme was inhibited. These conditions limit proteolysis of the serpin-complexed protease; proteases inhibited by serpins may become highly sensitive to proteolysis by excess or exogenous protease (26,27). Return of enzyme was monitored after a 250 -500-fold dilution of the complex into a substrate-containing media. Breakdown of chymotrypsin-serpin complexes were monitored in a solution of 0.5 M NaCl, 0.1 M Tris-HCl (pH 8.0), 0.01% dodecyl maltoside, 9% Me 2 SO, 2 mM Suc-AAPF-NA (Ͼ20 ϫ K m ), and breakdown of chymase-serpin complexes were in 1.0 M NaCl, 0.1 M Tris-HCl (pH 8.0), 0.01% dodecyl maltoside, 9% Me 2 SO, 1.0 mM Suc-AVPF-NA (7 ϫ K m ). Conversion of substrate to product was followed continuously for ϳ18 h in a thermostatted spectrophotometer.
SDS-PAGE of Protease-Inhibitor Reactions-Reactions were typically performed in a total volume of 25 l, and proteins were resolved by discontinuous gel electrophoresis using a Bio-Rad mini gel system. 12% running gels were prepared according to Laemmli (28). Reactions were performed under the buffer conditions for titrations. Incubation times covered at least five half-lives, assuming second order conditions. Reactions were stopped by the addition of inhibitors PMSF or MeO-Suc-AAPV-chloromethyl ketone, and samples were denatured in an SDS-DTT solution with or without heating at 90°C for 10 min. DTT was not used in the denaturation of chymotrypsin-serpin reactions; this omission prevents dissociation of chymotrypsin into its three peptide components. Protein bands were visualized employing Gelcode Blue.
In certain studies, reactions were stopped instantaneously by the addition of trichloroacetic acid plus carrier RNA (10% final trichloroacetic acid concentration). Samples were allowed to precipitate in the cold for 1-2 h and then centrifuged to collect the precipitate. Precipitates were washed twice with ice-cold EtOH and then resolubilized in the SDS-DTT sample buffer and denatured by heating.

N-terminal Sequence Analysis of HNE-rACT-RSP1Ј
Complexes-Reactions contained an 8-fold excess of inhibitor in a solution of 0.1 M HEPES, 0.4 M NaCl, and 0.01% dodecyl maltoside. After completion, products were denatured by SDS-DTT without heating and resolved on a 12% SDS gel. Bands were electroblotted onto polyvinylidene difluoride paper and visualized with Coomassie Brilliant Blue. The complex band was cut from the paper and subjected to 15 rounds of Edman degradation. Deconvolution of the data with knowledge of the full sequences of HNE and rACT-RSP1Ј allowed for identification of the N termini of each protein and RSL cleavage sites.
The above procedure for identifying the reactive site was based upon an empirical observation in this laboratory. We have found that protease-ACT complexes treated with SDS-DTT without heating retain a significant fraction of the serpin C-terminal fragment produced during complex formation. We suspect that this resistance to complete denaturation reflects the enhanced stability of serpins when the reactive site is cleaved. Data Analysis-Data was analyzed using the fitting routines of Igor Pro from Wavemetrics.

RESULTS
Design of Inhibitors-The RSL structure of ACT variants with reactive sites at the P2 position (16 residues from Glu-342) and the P1Ј position (18 residues from Gly-342) are shown in Table I. The P2 reactive site variant, rACT-RSP2, was produced by mutating Leu-358 to Ser. This mutation makes Leu-357-Ser-358 the default reactive site. The P1Ј reactive site variant rACT-RSP1Ј was produced by replacing the sequence P2-P2Ј of rACT by Ile-Thr-Leu-Ser. This relocates Leu to P1Ј and maintains the sequence of the surrounding residues. As a result of these mutations, position 358 is either Ser or Thr and, therefore, should not be recognized.
Characterization of the Interaction of Proteases with Reactive Loop Variants-The parameters, SI, k inh , and complex stability, describing the inhibition of chymase, chymotrypsin, and HNE with variants having the reactive site relocated to P2 or P1Ј are reported in Table II. Also shown for comparison is the previously published parameters for the inhibition of the same proteases by rACT. The results show that each reactive site variant is functional, inhibiting at least one of the proteases being analyzed. Chymotrypsin and chymase were inhibited by rACT-RSP2, and HNE was inhibited by rACT-RSP1Ј.
Inhibition of Chymotrypsin-like Proteases by rACT and rACT-RSP2-Titrations of chymotrypsin and chymase hydrolytic activity by rACT-RSP2 and chymotrypsin hydrolytic activity by rACT are shown in Fig. 1A. All titrations demonstrated a linear loss of hydrolytic activity; only the chymotrypsin-rACT titration extrapolated to an end point con-sistent with an SI of 1. Endpoints for the inhibition of chymotrypsin and chymase by rACT-RSP2 were at [I] 0 /E] 0 values of 3 and 12, respectively, Both SI values are significantly higher than 1 and 4, observed for the inhibition of each protease by rACT (Table II). SIs Ͼ 1 are due to the degradation of the inhibitor by a reaction in competition with the inhibition reaction.
SDS-PAGE analysis of the reaction of rACT-RSP2 with chymase (panels A and B) and chymotrypsin (panel C) are shown in Fig. 2. Lane 1 in each gel is the control, demonstrating the migration of active inhibitor (I) and PMSF-inhibited protease. Lanes 2-4 show the banding patterns of reactions produced at [I] 0 /E] 0 that range from below to above the SI established by titration. Bands consistent with the formation of covalent (1:1) protease-serpin complex (CM) are evident in reactions where [I] 0 /E] 0 Ն SI. When [I] 0 /E] 0 is below the SI (Fig. 2, A-C, lanes 2), degradation of the protease-serpin complex is observed to variable degrees. Degradation is due to proteolysis of the complex, protease, and/or serpin components by excess protease. Such proteolysis does not result in the release of active protease (21,26,29). The patterns of protease disappearance (chymase or chymotrypsin) and that of intact inhibitor appearance over the [I] 0 /E] 0 range reflect the high SI values observed for each protease in titrations. Cleaved serpin (CL) is a consequence of the cleavage pathway producing SI Ͼ 1. The relative staining intensities of the complex and cleaved inhibitor bands also are consistent with the high SI values.
The close migration of intact and cleaved inhibitor in gels A and C indicate a minimal difference in mass. The small difference is consistent with hydrolysis within the RSL located in all serpins ϳ40 -50 residues from the C terminus. Gels A and B represent the same chymase-rACT-RSP2 reaction, only differing in the temperature used for SDS denaturation. Gel B, in which SDS denaturation is performed at 25-37°C instead of 90 -100°C, shows enhanced resolution of cleaved and intact inhibitor bands. The better separation confirms the extraordinarily high SI for the chymase-rACT-RSP2 reaction. The enhanced resolution is likely related to incomplete denaturation of cleaved inhibitor due to the greater structural stability of this serpin form (30,31).
Sites of RSL cleavage were identified by MALDI-MS. This technique can be used to measure the mass of the 4 -5-kDa C-terminal fragment generated in production of serpin-protease complex and cleaved serpin (29,32). In Table III, the size of the C-terminal fragment(s) produced in the reactions of chymase with rACT, a P3Ј variant of rACT, and rACT-RSP2 are reported. Experimentally determined masses are reported be-

Summary of the inhibition characteristics for the reaction of rACT and rACT-reactive site variants with
human chymase, chymotrypsin, and HNE Reactive site variants refer to rACTs with the reactive site relocated to the P2 (rACT-RSP2) or P1Ј (rACT-RSP1Ј) position as reported in Table  I. SI, k inh and t1 ⁄2 bkdn (half-life for breakdown of protease-serpin complex) were measured as described under "Experimental Procedures" and "Results." low the closest mass calculated from the amino acid sequence. The cleavage products formed in the chymase-rACT and chymase-rACT-L361V reactions indicate two primary sites of en-counter at the P1 and P2 residues of the RSL. A potential chymase cleavage site at P3Ј-P4Ј does not appear to be a primary site because its removal does not affect the SI (Table III). The only fragment detected for the chymase-rACT-RSP2 reaction corresponds to cleavage between Leu-357-Ser-358, defining this site as both the reactive site and the cleavage site producing the SI Ͼ 1. Given the similar substrate specificity of chymase and chymotrypsin, exclusive interaction with the P2 site also likely explains the SI increase for the chymotrypsin-rACT-RSP2 reaction. k inh values measured for the reactions of chymotrypsin and chymase with rACT-RSP2 are reported in Table II. k inh is markedly decreased for both reactions relative to that obtained with rACT. The decrease in the efficiency of inhibition of chymotrypsin (500-fold decrease) was greater than that observed for chymase (6-fold).
The stability of complexes was determined by monitoring the return of enzyme activity after a large dilution of preformed complex into substrate (peptide-NA)-containing solution as described under "Experimental Procedures." The release of protease was measured by following the progress of NA accumulation at A 410 nm . The progress curve data were numerically differentiated to obtain the plots shown in Fig. 3. Only data for chymotrypsin-rACT and chymotrypsin-rACT-RSP2 are shown.
Comparable chymase-serpin complexes demonstrated negligible release of protease over the 18-h period of monitoring. Rate constants for the breakdown of the chymotrypsin-serpin complexes were determined by fitting the differentiated data to a single exponential function (solid line) with 100% release of the protease as the end point.
As shown in the Fig. 3, the chymotrypsin-rACT-RSP2 complex was ϳ20 -40-fold more stable than the chymotrypsin-rACT complex. The half-life values estimated from the rate constants were 100 -200 h and 5-6 h, respectively. The half-life for the chymotrypsin-rACT complex is comparable with the half-life of 7 h previously reported using a discontinuous method of assay (19). To ensure that the difference in breakdown rates was not influenced by the difference in inhibitor concentrations used to form each complex, a study similar to  Table I. Enzyme activity is reported as fractional activity. To calculate this value, data points were divided by the y-intercept obtained from regression analysis of the original titration. Two sets of data (circles and squares) shown for HNE-rACT-RSP1Ј titration were obtained after 1 and 24 h of incubation. Panel A also shows titration of chymotrypsin with rACT for comparison. that shown in Fig. 3 was performed using identical concentrations of both inhibitors. Similar results were obtained.
These results demonstrate that shortening the distance between the beginning of the RSL (Glu-342) and reactive site by 1 residue (17 to 16 residues) does not abolish the ability of the serpin to inhibit proteases and can improve complex stability. On the other hand, relocation of the reactive site to the P2 position decreased k inh and increased SI values, suggesting that this site is sub-optimal with respect to reaction parameters determining inhibitory efficiency.

MALDI-MS analyses identifying the C-terminal fragments produced by the reaction of chymase and HNE with rACT and variants
The RSL of ACT is located approximately 40 -50 residues from the C terminus. Therefore, cleavage within the RSL produces polypeptide fragments of approximately 5000 Da. Reactions with SI Ͼ1 may produce single or multiple fragments depending on whether the substrate site is same as or different from the reactive site. Shown below are the P6-P4Ј sequences of rACT and rACT variants, the calculated mass for a peptide fragment extending from the indicated residue to the C terminus, and the experimentally measured masses of fragments produced upon reaction of chymase or HNE with each serpin. initiated immediately after a 100-fold dilution of free and inhibitor-treated chymase into reaction buffer containing a saturating concentration of substrate. The progress curve for inhibitor-treated chymase demonstrates a lag before steady state hydrolysis is attained. The lag indicates that chymase had reacted with the inhibitor but in an unstable manner. SDS-PAGE monitoring of a chymase-rACT-RSP1Ј reaction performed at an [I] 0 /[E] 0 near 1 demonstrates the transient formation of a covalent serpin-protease complex immediately after mixing (Fig. 5A) and the rapid turnover of this complex to free enzyme and cleaved inhibitor. The slow, but catalytic turnover of inhibitor is demonstrated in the time courses where [I] 0 /[E] 0 was increased to 10 (Fig. 5B). The wider separation of cleaved (CL) from intact inhibitor (I) in panel B was due to the modification in the SDS denaturation process, as described above. The site cleaved by chymase was identified as Leu-359 -Ser-360 (P1Ј-P2Ј relative ACT) using MALDI-MS (Table III).
The formation of a transient complex was specific to chymase. Chymotrypsin interaction with rACT-RSP1Ј did not demonstrate a lag when incubations were diluted into substrate solution (Fig. 4B). Rapid cleavage of rACT-RSP1Ј demonstrated by SDS-PAGE indicated that the inhibitor was recognized by chymotrypsin (data not shown).
These results suggests that lengthening the distance between the beginning of the RSL and the reactive site by a single residue (17 to 18 residues) effectively abolishes the ability of rACT to form a stable complex with a protease. The above results were not due to the poor or inappropriate folding of rACT-RSP1Ј, as will be shown below.
Inhibition of Neutrophil Elastase by rACT-RSP1Ј-Unexpectedly, as reported in Table II, HNE was effectively inhibited by rACT-RSP1Ј. Titration of HNE activity with rACT-RSP1Ј demonstrated a linear loss in hydrolytic activity with an SI of ϳ4 (Fig. 1B). Over a 24-h period of incubation, there was no change in the titration, suggesting that HNE was virtually irreversibly inhibited at all values of [I] 0 /[E] 0 . Consistent with the production of a stable complex, SDS-PAGE showed the formation of a covalent complex that did not break down over a 24-h period of monitoring (Fig. 5C). The magnitude of k inh , 14, 000 M Ϫ1 s Ϫ1 , for the HNE-rACT-RSP1Ј reaction was the highest measured for the reactive site variants.
MALDI-MS and N-terminal sequence analysis were used to identify the reactive site for the HNE-rACT-RSP1Ј reaction. Reaction products analyzed by MALDI-MS revealed cleavage products consistent with hydrolysis of the RSL at several sites (Table III). Cleavage at P6-P5 and P4-P3 have been reported previously for the reaction of HNE with rACT (17) and variants of rACT (29). Because HNE is not inhibited by rACT, these sites likely do not mediate inhibition. Two other fragments not previously observed revealed masses consistent with cleavage at P2-P1 and P1Ј-P2Ј. Peak heights of analyses suggest that the fragment corresponding to cleavage at P2-P1 was the most abundant reaction product. N-terminal sequence analysis of HNE-rACT-RSP1Ј complexes resolved on SDS-PAGE demonstrated four sequences. Two sequences were the N termini of each protein, and two corresponded to cleavage within the RSL. One RSL-derived sequence (H-TLXLVETRTIVRFN-OH) was produced by cleavage of the P2-P1 peptide bond, whereas the other was produced by cleavage of the P4Ј-P5Ј peptide bond. Cleavage at P4Ј-P5Ј has been observed by MALD-MS in HNE reactions with rACT variants capable of inhibiting HNE by interaction at the P1 site (17,29). Cleavage at this site was variable and is believed to occur after complex formation by exposure of the site to enzyme not yet inhibited (29).
The finding of P2-P1 cleavage and no other by both N-terminal sequencing of the complex and by MALDI-MS indicate that the P2-Ile residue of rACT-RSP1Ј is the site mediating inhibition. Cleavage at P6-P5, P-4-P3, and P1Ј-P2Ј are likely responsible for the SI Ͼ 1.

DISCUSSION
The current study evaluates the effect of reactive site relocation on the interaction of ACT with chymotrypsin, chymase, and HNE. This study parallels a recent report using the interaction of ␣1-PI Pittsburgh with factors Xa and thrombin as a model system (18). In the study by Zhou et al. (18), the reactive site was relocated nearer or further from the base of the RSL, Glu-342, by inserting Ala residues (Add-1, Add-2 variants) or removing RSL residues (delete 1-3 variants) at sites 2-3 residues preceding the reactive site. In our study, the method of reactive site relocation to the P2 or P1Ј position of ACT did not affect the overall length of the RSL. Relocation was accomplished by mutation of reactive site residues rather than by adding or deleting residues (Table I). Despite the difference in the design of variants, similar effects on the inhibition parameters of complex stability, k inh and SI, were observed with changes in reactive site location. The agreement between these two studies suggests that ␣1-PI and ACT follow a similar mechanism of inhibition.
Complex Stability-Reaction at the P2 site, located 16 residues from Glu-342, produced inhibitory complexes that were as stable or more stable than those produced at P1. Chymotrypsin-rACT-RSP2 complexes were ϳ20 -40-fold more stable than chymotrypsin-rACT complexes. The stability of the HNE-rACT-RSP1Ј complex formed by interaction of HNE with the P2-Ile residue was even more striking considering the relative instability of complexes produced by interaction with numerous ACT-reactive site variants maintaining the reactive site at P1 (17,29).
The Glu-342-reactive site distance cannot be reduced to less than 15 residues without loss of inhibitor activity. We have shown previously that HNE cleaves ACT at P6-P5 and P4-P3 bonds in a substrate-like manner (17). Similarly HNE cleavage of peptide bonds 12-14 residues from Glu-342 have been shown to inactivate several other serpins including C1 inhibitor (15), antiplasmin (13), and antithrombin (16). Zhou et al. (18) produced three RSL-shortened mutants (Delete 1-3), placing the reactive site at P4, P3, and P2, 14 -16 residues from Glu-342. Although interaction of thrombin at P2 (Delete-1) produced stable inhibition similar to that observed in our study, reaction at P3 and P4 produced only cleaved serpin. Factor Xa was somewhat less sensitive to loop shortening, demonstrating stable inhibition with Delete-1 and 2 but not Delete-3. Taken together these observations strongly indicate that the placement of a functional reactive site has a lower limit of 15 or 16 residues from Glu-342.
In contrast to the apparent increase in complex stability associated with the P2 site, relocating the reactive site to P1Ј led to complete loss or diminished inhibitory ability. Chymotrypsin and HNE appeared to recognize the P1Ј site only as a cleavage site. The interaction of chymase with the P1Ј site demonstrated transient inhibition (t1 ⁄2 Ͻ 5 min), defined by the rapid formation and then disappearance of an SDS-stable complex. Similar transient inhibition was observed for the interaction of Xa with the Add-1 (reactive site at P1Ј) variant of ␣1-PI Pittsburgh (18). Inhibition of thrombin by the Add-1 variant was more stable, demonstrating a half-life for enzyme return of 10 h compared with 14 weeks (virtually irreversible) with native ␣1-P1-Pittsburgh. The Add-2 variant behaved as a transient inhibitor of thrombin, supporting a general trend toward rapid complex breakdown upon increased Glu-342-reactive site distance.
The abrupt changes in complex stability observed by us and by Zhou et al. (18) upon relocation of the reactive site suggests that serpin functionality is dependent on the reactive site being located 15-17 residues from Glu-342. This finding is consistent with the recent crystal structure of a protease-serpin complex (6) and studies proposing protease distortion as the mechanism of inactivation (18,26,(33)(34)(35). The crystal structure showed a distorted protease located at the pole opposite that of encounter and insertion of the RSL from P17-P3 into ␤-sheet A. Full insertion of the RSL into ␤-sheet A appeared to underpin the distortion process. The embedding of the RSL forced the close apposition of both proteins, resulting in strain on Ser-195 and distortion of protease around the "rigid" body of the cleaved serpin. In this insertion-distortion model, inhibition would be dramatically reduced if the Glu-342-reactive site distance was too long to provide strain on the protease after full insertion (Ն18 residues) or too short to complete the insertion process (Յ15 residues). The improved stability observed for proteaseserpin complexes upon shortening the Glu-342-reactive site distance to 16 instead of 17 residues would be consistent with greater strain and distortion of the protease.
The dependence of complex stability on reactive site location was not the same for all proteases. This variation may be related to active site structure influencing the extent of distortion (18,34,35). It was suggested by Zhou et al. (18) that the stability of thrombin with the Add-1 variant was due to its unusually deep active site cleft. The deep cleft is primarily produced by the large size of two loops termed 60 and 147. Based on the backbone structures of chymotrypsin, chymase, and HNE shown in Fig. 6 (24,36,37), chymase appears to have the deepest active site cleft, and HNE appears to have the most shallow. A three-residue insertion makes the 37 loop of chymase more prominent than the corresponding loops of the other proteases, and the 145 loop is not cleaved like that of chymotrypsin. These features may explain the virtual irreversible inhibition of chymase upon interaction with reactive sites located at P1 and P2 and the transient inhibition upon the interaction at P1Ј. Chymotrypsin did not form a stable complex at P1, did not form even a transient complex at P1Ј, and only approached the stability of chymase at P2. HNE has the shallowest-appearing active site cleft and virtually no 170 -180 loop. The shallowness of the active site cleft may explain why reaction at the P2 site of rACT-RSP1Ј produced a highly stable HNE-serpin complex, whereas numerous other rACT variants with the reactive site at P1 produced relatively unstable complexes (17,29). An additional distinguishing feature of the chymase structure, possibly enhancing its susceptibility to distortion, is the absence of a disulfide bond between residues 191 and 220. Residues 191 and 220 join the two major sequence  (24), chymotrypsin (49), and HNE (50) were rendered using Rasmol. Accession numbers for the crystal structures are 1PJP, 1HNE, and 4CHA, respectively. The loops forming the walls of the active site cleft and the S1 pocket are presented in color (24,37,51). The 37 loop is in green, the 60, 70 -80, 99, and 147 loops (autolysis loop in chymotrypsin) are in blue, and 174 loop is in orange. The loops producing the S1 pocket, 189 -195 and 213-228, are in red. Specific residues are presented over some loops as well; all numbering is based on homology to chymotrypsin as used in the above crystal structures. Ser-195 is shown as a black stick model in each structure. segments, 189 -195 and 213-228, that form the S1 pocket. In chymase, the cystine is replaced by Phe-191 and Ala-220.
As stated in the introduction, not all studies of serpin-protease complexes find the protease at a position consistent with full insertion. Studies implying partial insertion of the RSL do not provide a mechanism for protease inactivation. In the absence of an alternative mechanism for complex stability, it is difficult to understand the correlation provided here between complex stability and a length of RSL residues just appropriate for translocation of the protease to the pole opposite that of encounter. This correlation also argues against the suggestion that stable complexes can be produced at different locations.
Stoichiometry of Inhibition-The increases in SI values for chymase and chymotrypsin upon interaction with the P2 site (see Table II) reproduces a similar trend to that reported by Zhou et al. (18). The mechanism primarily assumed for generation of an SI Ͼ 1 involves a competition for the serpin acyl protease between deacylation on the hydrolytic pathway and the initiation of RSL insertion on the inhibition pathway (38 -41). This mechanism is derived mainly by analogy with suicide substrates (38,39) and from studies involving mutation of the hinge region near Glu-342 (42,43). It does not straightforwardly explain the current observation pertaining to reactive site relocation. A reactive site located at P2 could pose difficulty for completion of the insertion process at the pole of the serpin, where distortion of the acyl enzyme is presumed to occur. An enzyme translocated on a shortened arm might not align properly with the pole, thereby requiring increased time for completion of the insertion process and subsequent distortion of the protease.
Another finding related to SI values is that the unusually high SI for the chymase-rACT reaction (SI ϭ 4) is due at least in part to a partitioning of the reaction between the P1 and P2 sites (Table III). MALDI-MS analysis of chymase-rACT reactions suggested that both sites are recognized by chymase. The SI for the reaction of chymase with rACT-RSP2 was greater than 10, whereas the SI for the reaction with a variant having a P2 site not recognized by chymase, rACT-Val357Leu (32), was about 1.7. Interaction with each site at an approximately equal rate, as will be discussed subsequently, would produce an SI ϭ 4.
Inhibition Rate Constants-Inhibition through reaction with the P2 site of ACT decreased k inh values. The decrease was more dramatic for chymotrypsin (500-fold) than chymase (6fold). Because both reactions exhibited increased SI values (Table II), the recognition of the P2 site is somewhat greater than that estimated by k inh . Considering the SI, recognition of the P2 site by chymotrypsin is still poor relative to the P1 site of rACT. On the other hand, the high SI (12) for the chymase-rACT-RSP2 reaction suggests that the P2 site is recognized by chymase at a rate comparable with that of the P1 site. As pointed out above, approximately equal recognition of P2 and P1 could explain the high SI of 4 for the chymase-rACT reaction. The most efficient inhibition reaction was that of HNE (k inh ϭ 14,000 M Ϫ1 s Ϫ1 ) with the P2-Ile of rACT-RSP1Ј (see Table II). Because rACT-RSP1Ј does not inhibit either chymotrypsin or chymase and ACT does not inhibit HNE, the modifications to produce rACT-RSP1Ј effectively changed the specificity of ACT. Decreases in k inh of 3-and 36-fold were observed for the interaction of ␣1-PI Pittsburgh Delete-1 variant with thrombin and factor Xa, respectively. The decreased efficiency of inhibition observed for variants constructed with the reactive site at P2 may reflect the fact the ␣1-PI and ACT evolved to optimize only one reactive site 17 residues from Glu-342. Naturally occurring serpins with a reactive site 16 residues from Glu-342 are highly efficient inhibitors (13, 44 -48), exhib-iting k inh values similar to those of serpins, with the reactive site located 17 residues from Glu-342. This includes serpins monocyte/neutrophil elastase inhibitor 1 (44), and human cytoplasmic antiproteinase (45) which have reactive sites of different specificity at positions 16 and 17 residues from Glu-342.
In summary, this study provides a rationale for the apparent limitation of the reactive site to positions 16 or 17 residues from Glu-342 in naturally occurring serpins. The mechanism of inhibition implied by this limitation is that of full insertion, translocation of the protease 70 Å, and inactivation by distortion. The agreement between our results and those of a parallel study utilizing a different serpin and different proteases (18) supports the overall application of the above mechanism to all serpins. The underlying cause for the effect of reactive site location on SI and k inh values remain to be determined.