Formation of the covalent chymotrypsin.antichymotrypsin complex involves no large-scale movement of the enzyme.

alpha(1)-Antichymotrypsin is a member of the serine proteinase inhibitor, or serpin, family that typically forms very long-lived, enzymatically inactive 1:1 complexes (denoted E*I*) with its target proteinases. Serpins share a conserved tertiary structure, in which an exposed region of amino acid residues (called the reactive center loop or RCL) acts as bait for a target proteinase. Within E*I*, the two proteins are linked covalently as a result of nucleophilic attack by Ser(195) of the serine proteinase on the P1 residue within the RCL of the serpin. This species is formally similar to the acyl enzyme species normally seen as an intermediate in serpin proteinase catalysis. However, its subsequent hydrolysis is extremely slow as a result of structural changes within the enzyme leading to distortion of the active site. There is at present an ongoing debate concerning the structure of the E*I* complex; in particular, as to whether the enzyme, bound to P1, maintains its original position at the top of the serpin molecule or instead translocates across the entire length of the serpin, with concomitant insertion of RCL residues P1-P14 within beta-sheet A and a large separation of the enzyme and RCL residue P1'. We report time-resolved fluorescence energy transfer and rapid mixing/quench studies that support the former model. Our results indicate that the distance between residue P1' in alpha(1)-antichymotrypsin and the amino terminus of chymotrypsin actually decreases on conversion of the encounter complex E.I to E*I*. These results led us to formulate a comprehensive mechanism that accounted both for our results and for those of others supporting the two different E*I* structures. In this mechanism, partial insertion of the RCL, with no large perturbation of the P1' enzyme distance, is followed by covalent acyl enzyme formation. Full insertion can subsequently take place, in a reversible fashion, with the position of equilibrium between the partially and fully inserted complexes depending on the particular serpin-proteinase pair under consideration.

␣ 1 -Antichymotrypsin is a member of the serine proteinase inhibitor, or serpin, family that typically forms very long-lived, enzymatically inactive 1:1 complexes (denoted E*I*) with its target proteinases. Serpins share a conserved tertiary structure, in which an exposed region of amino acid residues (called the reactive center loop or RCL) acts as bait for a target proteinase. Within E*I*, the two proteins are linked covalently as a result of nucleophilic attack by Ser 195 of the serine proteinase on the P1 residue within the RCL of the serpin. This species is formally similar to the acyl enzyme species normally seen as an intermediate in serpin proteinase catalysis. However, its subsequent hydrolysis is extremely slow as a result of structural changes within the enzyme leading to distortion of the active site. There is at present an ongoing debate concerning the structure of the E*I* complex; in particular, as to whether the enzyme, bound to P1, maintains its original position at the top of the serpin molecule or instead translocates across the entire length of the serpin, with concomitant insertion of RCL residues P1-P14 within ␤-sheet A and a large separation of the enzyme and RCL residue P1. We report time-resolved fluorescence energy transfer and rapid mixing/quench studies that support the former model. Our results indicate that the distance between residue P1 in ␣ 1 -antichymotrypsin and the amino terminus of chymotrypsin actually decreases on conversion of the encounter complex E⅐I to E*I*. These results led us to formulate a comprehensive mechanism that accounted both for our results and for those of others supporting the two different E*I* structures. In this mechanism, partial insertion of the RCL, with no large perturbation of the P1 enzyme distance, is followed by covalent acyl enzyme formation. Full insertion can subsequently take place, in a reversible fashion, with the position of equilibrium between the partially and fully inserted complexes depending on the particular serpin-proteinase pair under consideration. 1 is a member of the serine pro-teinase inhibitor, or serpin, family that typically forms very long-lived, enzymatically inactive 1:1 complexes (denoted E*I*) with its target proteinases. Serpins share a conserved tertiary structure in which an exposed region of amino acid residues (called the reactive center loop or RCL) acts as bait for a target proteinase. For ACT, this loop extends from residues 342 to 367, denoted P17-P9Ј, in which, following the nomenclature of Schechter and Berger (1), the scissile bond cleaved by the targeted proteinase chymotrypsin (Chtr) is between the P1 and P1Ј residues. The native serpin structure is unusual in that it is metastable and is considered to be in a stressed (S) conformation. The cleaved serpin released from the complex is much more stable than the intact serpin (2,3) and is considered to be in a relaxed (R) conformation. Noteworthy among many structural differences between the S and R conformations (4) is the A ␤ sheet, which is converted from a five-strand sheet into an antiparallel six-strand sheet by the insertion of RCL residues P1-P14 as strand s4A. As a result of this insertion, the P1 and P1Ј residues are separated by 70 Å.
There is now very good evidence that within E*I* the two proteins are linked covalently; an acyl enzyme has formed following attack by the nucleophilic Ser 195 of the serine proteinase on the P1 residue of the serpin (5)(6)(7). This species is formed rapidly (5) and is formally similar to the acyl enzyme species normally seen as an intermediate in serine proteinase catalysis. However, its subsequent hydrolysis is extremely slow, resulting in the observed inhibitory effect of serpins. This "trapped" acyl enzyme results from structural changes within the enzyme leading to distortion of the active site, which was inferred from proteolysis (8,9) and NMR (10) studies in solution and recently confirmed by a crystal structure of the trypsin*antitrypsin* complex (11). Within this structure, the antitrypsin is in the R conformation, and the trypsin attached to the P1 residue has translocated from the top of the serpin molecule, defined as the position of the RCL in the intact, active serpin, across its entire length to the bottom of the molecule.
The trypsin*antitrypsin* structure provides strong support for one side in what has been an ongoing debate concerning the extent of RCL insertion within the E*I* complex, whether full insertion (R conformation), as first proposed by Wright and Scarsdale (12), or partial insertion, as suggested by Whisstock et al. (13). Earlier experimental results supporting each point of view are summarized in Stone et al. (14). More recently, studies employing fluorescence resonance energy transfer (FRET) (15) and donor-donor energy migration (16) have given results supporting the full insertion model, whereas monoclonal antibody binding studies to defined serpin epitopes (17,18) favor the partial insertion model. These continuing disagreements raise questions as to whether the structure of the E*I* complex may be different for different serpins and to what extent the time scale employed in measurement affects the structure observed, i.e. whether the partially inserted form may be an intermediate on the way to full insertion (19).
Here, we report time-resolved stopped-flow FRET and rapid mixing/quench studies of E*I* formation between ␣ϪACT, derivatized with a fluorescence donor at the P1Ј position, and ␦ϪChtr, derivatized with a fluorescence acceptor at the amino terminus. Our results demonstrate multistep formation of the E*I* complex in which the distance separating the two fluorophores is inconsistent with formation of the fully inserted R conformation. In consequence, we propose a mechanism for E*I* formation that rationalizes the apparent disagreements mentioned above. The chief features of this mechanism are that: (a) partial insertion of the RCL occurs with the P1-P1Ј peptide linkage intact; (b) acyl enzyme formation occurs following partial insertion and is not concomitant with full insertion; and (c) the R conformation forms reversibly, but not necessarily rapidly, from the partially inserted conformation.
Construction, Expression, and Purification of rACTs-S359C-rACT was constructed using sequence overlap expression polymerase chain reaction and the ACT expression vector described previously (20,22). The internal primers coding for the Ser 3 Cys mutation are as follows: 5Ј-CCCTCCTTTGTGCATTAGTGGAGACA-3Ј and 5Ј-GCACAAAGGA-GGGTGATTTTGAC-3Ј (the mutation sites are in boldface). The polymerase chain reaction product, representing the entire coding region, was cut with BstXI, gel-purified, and inserted in the correct reading orientation in pZMS. Full gene sequencing confirmed a single codon change. S359C-rACT and rACT were purified to homogeneity as described earlier (22).
Preparation of Derivatized S359C-rACT-S359C-rACT (2 M in 50 mM Tris, 50 mM KCl, pH 8.3) was reacted with a 100-fold excess of BMMC for 12 h at 0°C in an amber-colored reaction vessel. Samples were concentrated and buffer-exchanged using Amicon, Centriprep-30, and Centricon-30 concentrators (Amicon, Inc.). The stoichiometry of (7-methoxycoumarin)-methyl (MCM) labeling was determined by absorbance measurements. Total bound MCM concentration was determined using ⑀ 330 ϭ 13,000 M Ϫ1 cm Ϫ1 . Total protein concentration was determined using ⑀ 280 ϭ 39,000 M Ϫ1 cm Ϫ1 , calculated as the sum of the coefficients of the unlabeled protein (36,000 M Ϫ1 cm Ϫ1 , calculated from an ACT solution standardized by Bradford analysis) and the MCM group, determined using the A 280 /A 330 ratio for BMMC. Purified protein had a calculated stoichiometry of 1.05 MCM/protein. As shown earlier (22,23), the unique Cys residue in wild-type rACT, Cys 237 , is buried in the hydrophobic core of the molecule and has negligible reactivity under the reaction conditions used for derivatization.
Preparation of Derivatized ␦-Chymotrypsin-Zymogen was derivatized prior to activation, permitting specific labeling of the unique amino terminus present in the proenzyme. Chymotrypsinogen (19 M in 20 mM sodium P i , pH 7) was reacted with a 6-fold excess of 7-diethylaminocoumarin-3-carboxylic acid, DEACA-succinimidyl ester at 27°C for 30 to 150 min in a foil-covered reaction vessel. The reaction was quenched by adding a 0.1 volume of 1.5 M hydroxylamine, pH 8.5, with stirring at room temperature for 1 h. The sample was then washed and concentrated using Amicon and Centricon-10 concentrators. Total bound DEACA concentration was determined using ⑀ 432 ϭ 44,000 M Ϫ1 cm Ϫ1 (24). The ⑀ 280 of labeled protein, 58,000 M Ϫ1 cm Ϫ1 , was calculated as the sum of the extinction coefficients of the unlabeled protein (50,000 M Ϫ1 cm Ϫ1 ) (25) and the DEACA group (8,000 M Ϫ1 cm Ϫ1 , as determined by the A 280 /A 432 ratio for DEACA). The proenzyme had a calculated stoichiometry of 0.50 -1.2 DEACA/protein depending on the time of incubation employed. All of the kinetics experiments reported in this paper were conducted with 0.75 DEACA/Chtr, prepared using a 30-min incubation time. Labeled proenzyme (8 M) was activated by incubation with trypsin (0.4 M) at 27°C in 100 mM Tris, pH 7.6, until maximum activity was achieved (ϳ90 min). The sample was then quenched with TLCK (final concentration, 2 mM), concentrated, and buffer-exchanged using Amicon, Centriprep-10, and Centricon-10 concentrators (Amicon, Inc.).
Characterization of Derivatized S359C-rACT-Second-order rate constants for inhibition k i and stoichiometry of inhibition (SI) were determined for all serpin-proteinase pairs. Inhibition rate constants were determined by incubating equimolar concentrations of enzyme and inhibitor under second-order conditions and removing aliquots for residual enzyme activity determination as described earlier (20). SI values were determined by densitometric analysis of SDS-PAGE gels by comparing the intensity of cleaved serpin following complex formation with proteinase with the band intensity of unreacted intact serpin (23).
Mass Spectral Analysis-Electrospray ionization was performed on a Micromass Platform LC Electrospray mass spectrometer (Micromass ® UK, division of Waters Corp., Milford, MA) at the Mass Spectrometry Facility of the Department of Chemistry at the University of Pennsylvania. MALDI-TOF (matrix-assisted laser desorption ionization timeof-flight mass spectrometry) was performed on a VG Tofspec (Fisons Instruments, Danvers, MA) at the Protein Chemistry Laboratory in the Medical School of the University of Pennsylvania.
Oxidative Liberation of the Amino-terminal Fragment Cys 1 -Leu 13 from ␦-Chtr-Oxidation of chymotrypsin with performic acid results in cleavage of the disulfide links in the protein and allows for the separation of the amino-terminal fragment (␣-chain) from the rest of the molecule (27,28). ␦-Chtr or DEACA-␦-Chtr was suspended in formic acid (88% in H 2 0), and a 2-fold volume of performic acid (prepared by adding 1/20 volume 30% H 2 0 2 to formic acid followed by room temperature incubation for 1 h (29, 30)) was added, bringing the final concentration of enzyme to 20 mg/ml. Following incubation on ice for 15 min, the reaction mixture was brought to pH 5 by the addition of 10 volumes of 1 M Tris, pH 10, leading to virtually complete precipitation of residues Ile 16 -Asn 245 (27,28), as shown by subsequent HPLC analysis of the supernatant (see below). The precipitate was removed by centrifugation and redissolved in formic acid, and the amount of incorporated probe was determined by A 430 measurement. The supernatant was subject to RP-HPLC (Rainin C-18 column, Microsorb-MV TM , 50 ϫ 4.6 mm, 300 Å, 5 m). Following equilibration with 0.1% trifluoroacetic acid in water, peptides were eluted with the following gradient of acetonitrile (also containing 0.1% trifluoroacetic acid): 0 -10 min, 0 -20%; 10 -30 min, 20 -30%; 30 -90 min, 30 -40%; 90 -95 min, 40 -100%; 100 -105 min, 100%. HPLC data were analyzed using Turbochrom Navigator software from PerkinElmer Life Sciences. The amounts of ␣-chain were estimated from peak areas at 215 nm, corrected for elution yields of both the modified (23.4 Ϯ 1.8%) and unmodified (71.1 Ϯ 5.8%) forms, taking into account the contributions to A 215 from peptide bonds, side chains (31), and fluorescent probe.
Stopped-flow Fluorescence-Stopped-flow fluorescence emission traces were acquired using an Applied Photophysics SX.18MV stoppedflow spectrofluorometer with excitation at 330 and 430 nm. Emission was monitored at 400 nm (the emission maximum of MCM-S359C-rACT) or 475 nm (the emission maximum of DEACA-␦-chymotrypsin). Traces were fit to a three-step kinetic model using Pro-Kineticist software (Applied Photophysics Ltd.).
Steady-state Fluorescence Spectra-Serpin-proteinase complexes were formed at 21°C in 20 mM sodium P i , 10 mM EDTA, pH 7, by the addition of 5 M proteinase (50 l) to 20 M serpin (50 l) and incubation for 15 min. Static fluorescence spectra were acquired using a PerkinElmer LS50 luminescence spectrometer with excitation at 330 nm.
Rapid Mixing Quenched Flow-Rapid quenched-flow kinetic studies were carried out using a KinTek Chemical-Quench-Flow Model RQF-3 machine followed by densitometric analysis on SDS-PAGE as described previously (22). SDS-PAGE analysis was performed on aliquots of the acid-quenched reaction mixture using gels containing 12% polyacrylamide. Prior to analysis, all samples were precipitated with freshly prepared 20% trichloroacetic acid, incubated on ice 30 min, centrifuged and redissolved in 10 l of 20 mM sodium P i , pH 7, to which 2 l of 1 M Tris base was added to neutralize the excess trichloroacetic acid present in the quenched samples. Phenylmethylsulfonyl fluoride at a final concentration of 2 mM was also added to rapidly inactivate any renatured chymotrypsin. Gels were stained overnight in GelCode® Blue Stain Reagent (Pierce), destained overnight in water, then dried prior to densitometric analysis. Data were fit to a three-step kinetic scheme using Hopkinsim, version 1.7 (32).
Measurement of FRET Efficiency-Energy transfer was determined by the acceptor enhancement method as well as by the decrease in donor fluorescence (33). In calculating efficiencies, observed values of fluorescence were corrected for the inner filter effect (Equation 1) (33), in which A EX is the absorption of all species at D1 and A EM is the absorption of all species at the emission wavelength ( D2 ). Such corrections varied from 3 to 19% for both acceptor and donor fluorescence (see below).
The decrease in donor efficiency (35) was calculated by Equation 3, in which F DA is the fluorescence of the donor/acceptor pair, with irradiation at the excitation wavelength of the donor and detection at the emission maximum of the donor ( D1 D2 ϭ 330 nm excitation, 400 nm emission) and F D is the fluorescence of the donor in the absence of acceptor: i.e. MCM-S359C-rACT/unlabeled chymotrypsin, using the same excitation/emission pair.
Calculation of Distance-The measured efficiency is a function of the distance between donor and acceptor, R, equal to [E Ϫ1 Ϫ 1] 1/6 r o , where R O , the Förster distance at which the efficiency is 50%, is calculated from Equation 4.
In this equation, J, the overlap integral that relates the degree of spectral overlap between the emission spectrum of the MCM group covalently bound to serpin and the absorption spectrum of the DEACA group attached to proteinase, was determined to be 4.52 ϫ 10 Ϫ14 M Ϫ1 cm Ϫ1 nm 4 using the method of Conrad and Brand (36); D , the quantum yield of the donor in the absence of acceptor, was determined to be 0.023 for labeled serpin complexed to unlabeled proteinase by comparison with quinine sulfate as standard (37); n, the index of refraction of the solvent, is assumed to be 1.33 (38); and 2 is the orientation factor, describing the relative orientation in space of the transition dipoles of the donor and acceptor. 2 may be set equal to 0.667 for donors and acceptors that randomize by rotational diffusion prior to energy transfer or to 0.476 assuming that a range of static donor-acceptor orientations exists that does not change during the lifetime of the excited state (33). Using either of these values gives similar values for R O of 23.5 Å ( 2 ϭ 2/3) or 21.0Å ( 2 ϭ 0.476).
Modeling-Modeling of the MCM-S359C-rACT⅐DEACA-␦-Chtr complex was performed using the program Quanta version 98.1111 (Molecular Simulations, Inc., San Diego, CA). The coordinates for the ACT-Chtr docked model proposed by Katz and Christianson (39) were used as a starting point. The fluorescent probes were constructed in the two-dimensional Sketcher application and then converted to molecular structure files. These were simultaneously imported into Quanta along with the published model, moved into proximity with their attachment sites, and covalently attached with the modeling palette. The Molecular Editor tool was used to add or remove hydrogens as necessary for proper valency. The initial orientation of the probes was varied for both probes by a rotation around the bond directly connected to the fluor in 90 o increments, generating a total of 16 structures. The probe aromatic rings were constrained to remain planar during the minimization and subsequent energy calculations. Each structure was then minimized first by the method of Steepest Descents, followed by the method of Adopted-Basis Newton-Raphson. These structures were then used to measure an average center-to-center distance between the probes in the encounter complex.

RESULTS
Site of DEACA Labeling of ␦-Chtr-The site or sites of labeling within ␦-Chtr depended on the overall stoichiometry of chymotrypsinogen labeling. Preparations having Ն1.0 DEACA groups per Chtr molecule had measurable labeling of the Ile 16 -Asn 245 polypeptide chain, rising to ϳ0.22 mol/mol for the sample containing 1.2 DEACA groups per Chtr. This most likely reflects partial labeling of some of the 14 free Lys residues present. On the other hand, preparations containing Յ0.75 DEACA groups per chymotrypsinogen molecule showed no (Ͻ0.01 mol/mol) such labeling of Ile 16 -Asn 245 . For these samples, all labeling was confined to the ␣-chain (residues 1-13) and, by inference, to the only amino group within this chain, the ␣-amino group of Cys 1 . Accordingly, preparations containing 0.75 DEACA per ␦-Chtr were used in all kinetics experiments.
Direct determination of the stoichiometry of ␣-chain labeling was provided by RP-HPLC analysis of the soluble fraction of a performic acid oxidized sample of DEACA-␦-Chtr (Fig. 1) as described under "Experimental Procedures." An analysis of the 75% labeled DEACA-␦-Chtr gave a labeled ␣-peptide/unlabeled ␣-peptide ratio of 3:1, measured by corrected peak area at A 215 . The assignment of peak 2 as the DEACA-labeled ␣-chain was indicated by its absorbance at 430 nm and confirmed by mass spectral analysis (1589, M ϩ 2 Na). The assignment of peak 1 as an underivatized ␣-chain was also confirmed by mass spectral analysis (1301, M ϩ H).
Functionality of Fluorescently Labeled Proteins-As shown by the results in Table I, MCM-S359C-rACT retains full activity toward Chtr inhibition, as measured by both the second- order rate constant for inhibition, k i , and the SI of ϳ1. Similarly, DEACA-␦-Chtr retains full activity toward rACT and toward hydrolysis of the standard substrate N-succinimidyl-AAPF-p-nitroanilide (data not shown). Finally, the reaction of MCM-S359C-rACT with DEACA-␦-Chtr has an SI value (ϳ1) and a second-order rate constant for inhibition ( Table I) that are quite similar to those found for rACT reaction with ␦-Chtr.
Fluorescence Spectra of DEACA-␦-Chtr, MCM-S359C-rACT, and the DEACA-␦-Chtr*MCM-S359C-rACT* Complex-Relative to the sum of the contributions of single-labeled complexes DEACA-␦-Chtr*wt-rACT* and ␦-Chtr*MCM-S359C-rACT*, excitation at 330 nm of the DEACA-␦-Chtr*MCM-S359C-rACT* complex formed at 25°C gives a decreased fluorescence at 400 nm, the donor emission maximum, and a comparable increased fluorescence at 475 nm, the acceptor emission maximum (Fig. 2). These changes provide clear evidence for FRET within the complex, with efficiencies, calculated as described under "Experimental Procedures," of 0.30 for donor and 0.33 for acceptor. Similar efficiencies of 0.32 and 0.36 for donor and acceptor were obtained for the DEACA-␦-Chtr*MCM-S359C-rACT* complex formed at 40°C. Incubations were performed in the presence of excess inhibitor to prevent proteolysis of the E*I* complex (9).
Rates of Serpin⅐Proteinase Complex Formation by Stoppedflow Fluorescence-On donor excitation at 330 nm, the fluorescence changes that follow mixing of MCM-S359C-rACT with DEACA-␦Ϫchymotrypsin at pH 7, 5°C are triphasic between 2-5000 ms, whether monitored at 400 nm (Fig. 3A, donor fluorescence) or at 475 nm (Fig. 4A, acceptor fluorescence). In the former case, intensity first decreases and then increases in two steps, whereas in the latter case, fluorescence intensity increases in all three steps. Three phases are also clearly evident on mixing MCM-S359C-rACT with ␦-Chtr (Fig. 3A, excitation at 330 nm, emission monitored at 400 nm). Shown in Fig.  3B is the calculated decrease in donor fluorescence due to FRET.
Rate constants for all three phases were calculated by fitting the observed fluorescence data to Scheme 1, in which the encounter complex, E⅐I, is separated from the final complex, E*I*, by two intermediates, EI a and EI b (Table II). The evident saturation in rate constant values as the [E]/[I] ratio is increased from 2 to 5 provides clear evidence for three first order processes following E⅐I formation.
Identical stopped-flow fluorescence experiments were also carried out with acceptor excitation at 430 nm and monitoring at 475 nm following addition of either MCM-S359C-rACT or rACT to DEACA-␦ϪChtr. In both cases, observed fluorescence increased, but the changes were much smaller than those seen in Figs. 3 and 4, as would be expected for a group at the amino terminus of a protein that is probably not directly involved in protein-protein contacts. As a result, although the data ob-tained could also be fit to Scheme 1 (Table II), the rate constant values obtained for k 3 were considerably less precise.

Rate of DEACA-␦-Chtr*MCM-S359C-rACT* (E*I*) Formation by Rapid Mixing Quenched Flow-In experiments exactly
paralleling the stopped-flow experiments described above, rapid quench kinetics was employed to determine the rates of E*I* formation between DEACA-␦ϪChtr and MCM-S359C-rACT using SDS-PAGE analysis of aliquots. The results (Fig.  4A), displayed alongside the acceptor fluorescence results, show that E*I* formation occurs only after most of the change in fluorescence has already taken place. Data were fit to Scheme 1, assuming that only E*I* is SDS-stable, accounting for the observed lag phase. Values of k 3 were obtained using values for k 1 and k 2 derived from stopped-flow analysis. SDS-PAGE analysis also showed no evidence for proteolysis of the E*I* complex (9) over the 5-s period required for full E*I* formation (Fig. 4B) despite proteinase being present in excess.
Calculation of Energy Transfer Efficiency and Distance-Apparent FRET efficiencies as a function of time, calculated from Equations 2 (acceptor) or 3 (donor) and stopped-flow traces as described under "Experimental Procedures" were fit to Scheme 1 using the rate constant values determined above, by direct fitting of observed fluorescence and quenched-flow results (Fig. 5), to yield efficiencies for E⅐I, EI a , EI b , and E*I* (Table III) as well as estimates of the distances between donors and acceptors in these species. The results show that efficiencies increase monotonically in the conversion of E⅐I to EI a , to EI b , and finally to E*I*, with correspondingly small decreases in the estimated distances separating the fluorophores. Also noteworthy is that the efficiency for E*I* of 0.33, measured from both donor and acceptor fluorescence at 5°C (Table III), is virtually identical to the values of 0.30 -0.32 and 0.33-0.36 determined above (see Fig. 2) at 25 and 40°C, respectively. This identity is strong evidence that incubation at 25 and 40°C for 15 min does not result in any major conformational change detectable by FRET in the structure of the E*I* complex formed within 5 s at 5°C (Fig. 5).
Structure Modeling-Although no crystal structure is available for the ACT⅐Chtr complex, the docked model for this complex proposed by Katz and Christianson (39) allows an estimation of the expected chromophore-to-chromophore distance on noncovalent complex formation, corresponding to one of the complexes, E⅐I, Ei a , or Ei b , shown in Scheme 1. In the docked model, the distance between the ACT Ser 359 side chain oxygen  and the amino-terminal ␣-amino group in ␣-Chtr is 33.6 Å. In our modeling, Ser 359 was first mutated to Cys 359 , an MCM group was attached to ACT-Cys 359 , and a DEACA group was attached to the amino-terminal ␣-amino group in Chtr. Energy minimization, performed using a large variety of starting orientations, gave an estimated MCM center-to-DEACA center distance of 35.8 Ϯ 2.0Å, in reasonable accord with the range of values estimated by FRET for the E⅐I, Ei a , or EI b complexes (26.4 -34.6 Å, Table III).

DISCUSSION
Kinetic Scheme-Recent pH-dependent studies of the reaction of a fluorescent derivative of a cysteine variant of rACT at the P13 position with Chtr enabled us to formulate a scheme for rACT interaction with Chtr (Scheme 2), similar to Scheme 1 but containing three intermediates between E⅐I and E*I* (23). The nature of the pH dependence of the fluorescent changes and the similarity of Scheme 2 to the interaction of substrates with Chtr (40,41) led us to propose that E⅐I conversion to EI a principally involves rearrangement of the RCL to the required canonical conformation. This rearrangement is followed by two conformational changes, corresponding to EI b and EI c formation, that result in a closer fitting of the RCL to the catalytic machinery of Chtr, as well as to insertion of a major portion of the RCL within the A ␤-sheet (EI c ); and also followed by conversion of EI c to E*I*, corresponding to covalent reaction between Chtr residue Ser 195 and the P1 residue of ACT to give acyl enzyme.
Our current results are consistent with Scheme 1. In Luo et al. (23), the time dependence of fluorescent change and of E*I* formation, measured at pH 7 and 10°C, could be adequately fit with three rate constants, and it was only in considering all of the rate data collected over the pH range of 5.0 to 8.0 that the need for four rate constants became clear. Thus, it is appropriate to compare rate constants k 1 (27.5 s Ϫ1 ), k 2 (2.5 s Ϫ1 ), and k 3 (0.6 s Ϫ1 ) measured in this work with rate constants k 1 (75 s Ϫ1 ), k 2 (5.6 s Ϫ1 ), and k 3, app (0.7 s Ϫ1 ), respectively, measured in Luo et al., where k 3, app , the overall rate constant for conversion of EI b to E*I*, is equal to k 3 k 4 /(k 3 ϩ k 4 ). This comparison shows each of the constants measured at 5°C in the present work to be lower within a factor of 2 Ϯ 1 than the values measured by Luo et al. at 10°C, a reasonable result for a conserved kinetic scheme.
A General Mechanism for E*I* Formation-The results col-  Table III indicate a small decrease in distance between the fluorophores attached to the amino terminus of Chtr and P1Ј on conversion of the E⅐I encounter complex to the SDS-stable E*I* complex. This is consistent with only smallscale movement of Chtr relative to ACT during this conversion, but it would not exclude larger movements that would leave the fluorophore-fluorophore distance little changed, such as those that could accompany partial insertion of the RCL in the A ␤-sheet. What is excludable is a large-scale movement of the enzyme across the length of the serpin molecule, as in the fully inserted (R conformation) model proposed by Wright and Scarsdale (12) and seen recently in the high resolution crystal structure of the trypsin*antitrypsin* complex (11). This conclusion is independent of the value of 2 in Equation 4. Above (see "Experimental Procedures"), we have provided the rationale for using 2 values of 0.667 or 0.476, showing that the resulting calculated distance in the E⅐I complex is consistent with that estimated from a docked model. However, even making the extremely unlikely assumption that 2 in the E*I* complex has its upper limit value of 4.0 (corresponding to the two fluorophores being held rigidly parallel to one another with aligned dipoles) leads to an upper limit fluorophore distance of only 42.6 Å, well below the value demanded by the fully inserted model.
Although our results appear incompatible with the crystal structure of the trypsin*antitrypsin* complex, both sets of results, as well as the apparently contradictory results of others (15,16), can be rationalized within the general mechanism for serpin⅐proteinase complex formation that we propose in Fig. 6, which is based in part on the recent results of Gooptu et al. (19) (see below). The important features of this mechanism are that: (a) partial insertion of the RCL, involving substantial structural change, occurs with the P1-P1Ј peptide linkage intact; (b) acyl enzyme formation occurs following partial insertion and is not concomitant with full insertion; and (c) the fully inserted form of the acyl enzyme, E*I* 2 is formed reversibly from the partially inserted form E*I* 1 , and differences in serpin structure are capable of affecting both the equilibrium position between these two states and the activation energy barrier for state-to-state interconversion. Here it should be noted that there is long-standing evidence for the reversibility of E*I* formation from E⅐I (42)(43)(44), although the overall equilibrium generally favors the acyl enzyme form.
Feature a is supported by both time-resolved and structural studies. In our own work, both current and previous (5,22,23), on the ACT/Chtr interaction, we have shown that large fluorescent changes, from probes placed at the P7, P13, and P1Ј positions in ACT precede E*I* formation. Similar conclusions from time-resolved studies have been reported for the interactions of antithrombin (AT) and thrombin (14) and antitrypsin and elastase (45). In addition, structural studies have indicated that the RCL of intact serpins can insert either partially (19,46) or fully into the A ␤-sheet (13,47), the latter yielding the latent serpin. In the proposed mechanism, complexation with enzyme would favor partial insertion of the intact RCL in forming EI c .
Feature b is a consequence of the present work, which unequivocally shows that acyl enzyme forms in the reaction of ACT and Chtr without full insertion of the RCL. This result differs from that of Stratikos and Gettins (15), who, employing FRET measurements from tryptophans within trypsin-to-dansyl groups placed at different specific positions within antitrypsin, demonstrated that enzyme was proximal to the bottom of the serpin as in E*I* 2 , in agreement with the crystal structure of the antitrypsin⅐trypsin complex (11). Moreover, earlier work by Stratikos and Gettins (48) showed that such insertion occurs reasonably rapidly. Similarly, Fa et al. (16), using donor-donor energy migration and several fluorescently labeled double Cys variants of PAI-1, showed that, within the PAI-1⅐u-PA complex, the P3 position (amino acid 344) was far removed from the P1Ј position (amino acid 347) at the top of the serpin but was proximal to amino acid 313 at the bottom of the serpin; in these studies, however, there was no clear indication of how quickly P3 separates from P1Ј.
Feature c rationalizes the accumulation of E*I* 1 in the interaction of ACT and Chtr, as opposed to the accumulation of E*I* 2 in the antitrypsin/trypsin and PAI-1/u-PA interactions, if it is assumed that the E*I* 1 complex has a higher relative stability for the ACT:Chtr pair. A structural basis for this assumption is provided by the crystal structure of the so-called ␦-conformation of the naturally occurring L55P variant of intact ACT, as recently determined by Gooptu et al. (19). In this structure, the space between the s3A and s5A strands is filled by partial insertion of residues from the RCL, which bends out of the A sheet at P12 and turns to join s1C, and by residues 164 -172, which are part of helix F-s3A turn in the native structure. We used the ␦-conformation as a model for the serpin portion of both EI c and E*I* 1 , although recognizing that the structures of these two species could well differ in detail.
In rationalizing the ␦-conformation, which has not yet been found for any other serpin, Gooptu et al. (19) point out the high homology between residues 164 -172 and residues P9-P1, which occupy the same positions in cleaved ACT (Table IV). Importantly, other serpins, in particular, antitrypsin and PAI-1, show considerably lower homology between these two sequences, consistent with the assumption of higher relative stability for ACT and, by extension, for Chtr*ACT* 1 . Whether such higher stability is sufficient to favor Chtr*ACT* 1 thermodynamically over Chtr*ACT* 2 or just increases the kinetic barrier for conversion of Chtr*ACT* 1 to a more stable Chtr*ACT* 2 is unknown. Physiologically it may not matter, because such conversion does not take place on heating Chtr*ACT* 1 at 25 or 40°C for 15 min, and serpin⅐proteinase complexes are cleared from the bloodstream rather rapidly, with a t1 ⁄2 of 12 min observed for the clearance of Chtr*ACT* (49,50).
Finally, feature c also rationalizes the apparent conflict in the results of Fa et al. (16), discussed above, and those of Bijnens et al. (17), working with an essentially identical serpinproteinase pair. The latter workers report that monoclonal antibodies binding to an epitope comprising residues 128 -131 in helix F and K 154 in the turn connecting helix F to s3A, at the bottom of the serpin, bind equally well to active PAI-1 and to the PAI-1⅐t-PA complex. They consider this result incompatible with full insertion of the RCL, as in E*I* 2 , in which enzyme at the bottom of the serpin would block access to epitope, but it would be compatible with E*I* 1 . A resolution of this apparent conflict is provided by the assumption that the monoclonal antibody binds only to E*I* 1 , because in this case antibody   a Calculated by fitting plot of acceptor efficiency versus time to Scheme I, using k 1 ϭ 27.5 s Ϫ1 , k 2 ϭ 2.5 s Ϫ1 , k 3 ϭ 0.6 s Ϫ1 .
b Calculated assuming 2 ϭ 2/3. c Calculated by fitting plot of donor efficiency versus time to Scheme I, using k 1 ϭ 27.5 s Ϫ1 , k 2 ϭ 2.5 s Ϫ1 , k 3 ϭ 0.6 s Ϫ1 . addition could shift an E*I* 1 /E*I* 2 equilibrium distribution from dominant E*I* 2 to dominant E*I* 1 . A similar explanation would account for the results of Picard et al. (18), who showed that a monoclonal antibody recognizing residues 366 -370 (P28-P24) in s5A in AT binds to the AT⅐thrombin and AT⅐Factor Xa complexes, as well as to the complex formed between AT and a FIG. 6. General mechanism for E*I* formation. Shown is a general mechanism for E*I* formation applied to Chtr*(gray)ACT*(yellow), which accounts for the results obtained for a variety of serpin-proteinase stable acyl enzyme complexes (see "Discussion"). In the structures shown, generated by QUANTA, the MCM group bound to the P1Ј residue of ACT is depicted as a blue triangle and the DEACA group bound to the amino terminus of ␦-Chtr as a red oval. The structure of the encounter complex is taken from the docked model of Katz and Christianson (39) in which the RCL (in red) shows no preinsertion into the A ␤-sheet. The carboxyl-terminal portion of helix F (in blue) is exposed to solvent. E⅐I is converted via three or more steps (23) to EI c , in which the RCL is partially inserted between the s3A and s5A strands, with full insertion blocked by insertion of the carboxyl terminus of helix F, and the P1-P1Ј bond is intact. The structure of the serpin portion of E⅐I c is based on the structure determined for the ␦-conformation of ACT (19). Residues 353-357 (P6-P2) are shown as part of a continuous RCL, although they are not visible in the determined structure. Conversion of E⅐I c to the acyl enzyme E*I* 1 is accomplished without major conformational change, with enzyme and P1Ј remaining in proximity (Table III). Conversion of E*I* 1 to E*I* 2 proceeds reversibly and results in translocation of the proteinase across the length of serpin, accompanied by a large separation between enzyme and P1Ј, with helix F displaced from ␤-sheet A by the RCL. The Chtr and ACT portions of E*I* 2 are based on the determined structures of ␣-Chtr (51) and cleaved ACT (52), respectively. The latter structure shows no observed density for the segment P1Ј to P6Ј. In the E*I* 2 structure shown, the probe is placed on the first observed residue in the cleaved structure (P7Ј). In the determined structure of cleaved antitrypsin (53), which clearly shows the P1Ј to P6Ј region, this portion of the RCL is extended even further away from the body of the serpin.

TABLE IV
Sequence comparisons A structure-based sequence comparison between the RCL from residues P9 to P1 and the carboxyl-terminal portion of helix F for four inhibitory serpins is shown. This portion of helix F is seen to insert into ␤-sheet A in L55P-ACT (19). a Identity is indicated by the one-letter code for the amino acid in common between the two secondary structure elements; similarity in terms of a non-negative value in a Blosum amino acid similarity matrix is indicated with an asterisk. hexapeptide corresponding to residues P14-P9, but does not bind to native, heparin-activated, latent, or cleaved AT.