Differential Utilization of Enzyme-Substrate Interactions for Acylation but Not Deacylation during the Catalytic Cycle of Kex2 Protease*

Kex2 protease from Saccharomyces cerevisiae is the prototype for a family of eukaryotic proprotein processing proteases belonging to the subtilase superfamily of serine proteases. Kex2 can be distinguished from degradative subtilisins on the basis of stringent substrate specificity and distinct pre-steady-state behavior. To better understand these mechanistic differences, we have examined the effects of substrate residues at P1 and P4 on individual steps in the Kex2 catalytic cycle with a systematic series of isosteric peptidyl amide and ester substrates. The results demonstrate that substrates based on known, physiological cleavage sites exhibit high acylation rates (≥550 s−1) with Kex2. Substitution of Lys for the physiologically correct Arg at P1 resulted in a ≥200-fold drop in acylation rate with almost no apparent effect on binding or deacylation. In contrast, substitution of the physiologically incorrect Ala for Nle at P4 resulted in a much smaller defect in acylation and a modest but significant effect on binding with Lys at P1. This substitution also had no effect on deacylation. These results demonstrate that Kex2 utilizes enzyme-substrate interactions in different ways at different steps in the catalytic cycle, with the S1-P1 contact providing a key specificity determinant at the acylation step.

are active in intracellular compartments involved in sorting and secretion. As might be expected, the processing proteases are much more specific than the degradative subtilisins (13)(14)(15)(16)(17). Additionally, these enzymes can be distinguished on the basis of primary sequence and domain structure (18). Somewhat surprisingly, it has also been possible to distinguish these two families of serine proteases on the basis of their pre-steadystate behavior (9,14,16,19). This last finding suggests that the processing proteases may have adapted the classical serine protease mechanism (10,20,21) to their in vivo function during evolution in a manner not previously appreciated.
Serine proteases hydrolyze ester and amide substrates via a mechanism involving the generation and subsequent hydrolysis (or transamidation) of a covalent acylenzyme intermediate (10,(21)(22)(23). It had been thought classically that in cleavage of amide substrates, serine proteases were limited by the rate of acylation (commonly referred to as k 2 ), because this step involves the hydrolysis of a resonance-stabilized amide bond rather than the more labile ester cleaved in the deacylation step (whose rate is referred to as k 3 ). Early experiments with chromogenic and fluorogenic substrates analogous to single amino acids seemed to confirm this notion (9,21). Furthermore, substrates incorporating scissile esters rather than amides exhibited an initial burst of product formation caused by ratelimiting deacylation (21,24). For the subtilisin family, work with larger oligopeptide substrates also confirmed this expectation (9), in contrast to the evolutionarily unrelated serine proteases of the trypsin/chymotrypsin family (25).
However, the pre-steady-state behavior of Kex2 protease is at odds with this model, because the enzyme exhibits burst kinetics in cleavage of amide substrates based on known, physiological Kex2 cleavage sites (aliphatic/basic P 4 , basic P 2 , Arg at P 1 ; Refs. 14, 16, and 26). 1 Substitution of Lys for Arg at the P 1 position resulted in the loss of burst kinetics, indicating a change in rate-limiting step upon this substitution (16). Burst kinetics were also observed in cleavage of model amide substrates by the homologous enzyme furin (17,19). This work utilized tetrapeptide substrates with synthetic coumarin leaving groups (7-amino-4-methylcoumarin or 7-hydroxy-4-methylcoumarin; Refs. 14, 16, 17, 26, and 27). It was thus possible that this behavior was an artifact of the activated leaving group. However, otherwise identical substrates with either coumarin leaving groups or authentic peptide bonds are cleaved with essentially identical values of k cat /K M (15). Moreover, recent work has demonstrated that burst kinetics are also seen in Kex2 cleavage of an actual peptide bond, and this behavior requires the presence of the physiologically correct P 1 Arg (28). Thus, the acylation rate for Kex2 cleavage of peptide bonds must be substantially higher than some subsequent step, such as deacylation or product release. It has also recently been demonstrated that hydrolysis of the acylenzyme is rate-limiting for Kex2 protease in cleaving a substrate that exhibits burst hydrolysis (28).
This body of work suggested that catalysis by processing proteases such as Kex2 and furin involves an adaptation of the serine protease mechanism to allow rapid acylation in cleavage of sites resembling those in physiological substrates, e.g. with Arg at P 1 , but not in cleavage of nonphysiological sites. The much slower rate of deacylation relative to acylation in cleavage of physiologically correct sequences suggested that the energy of enzyme-substrate interactions might be used differentially during different points in the catalytic cycle. To test this model and better understand the means by which processing enzymes such as Kex2 and furin may distinguish authentic physiological substrates from other potential cleavage sites, we have measured or estimated substrate binding, acylation rate, and deacylation rate for sequences containing correct or incorrect P 4 and P 1 residues. The results are consistent with a model whereby correct enzyme-substrate contacts are used to drive acylation, with less effect on binding and essentially no effect on deacylation.
Steady-state Kinetics with Peptidyl-MCA Substrates-Kex2 protease was reacted with substrates in 0.2 M Bistris, pH 7.26, 1 mM CaCl 2 , 0.1% Triton X-100 at 21°C (standard conditions for this work). The reactions were quenched with 1 M acetic acid (600 l added to a reaction volume of 100 l), and product formation was measured by fluorescence as described (15). Some reactions were run in the above conditions but at 37°C, as reported previously (15,16). The difference in pH associated with the temperature change was negligible for these experiments (data not shown). For measurement of deacylation rates, some substrates were examined in 0.2 M sodium acetate, pH 6.0, 1 mM CaCl 2 , 0.1% Triton X-100 at 21°C. The reaction conditions are indicated in the tables and figure legends as appropriate. Active site titration with the substrate Ac-Nle-Tyr-Lys-Arg2MCA indicated that these different reaction conditions did not result in a change in the active enzyme concentration in the reaction (data not shown).
Steady-state Kinetics with Peptidyl-MCE Substrates-Substrate stocks were prepared as described (27). Kex2 cleavage of these substrates was examined in 0.2 M sodium acetate, pH 6.0, 1 mM CaCl 2 , 0.1% Triton X-100 at 21°C to minimize hydrolysis of the ester (27). The reactions were quenched with acetic acid, and fluorescence was measured exactly as described (27).
Rapid Quenched Flow Measurements-Z-Nle-Tyr-Lys-Lys2MCE and Z-Ala-Tyr-Lys-Lys2MCE were prepared in 0.1% aqueous trifluoroacetic acid and reacted with Kex2 in 0.2 M sodium acetate, pH 6.0, 1 mM CaCl 2 , 0.1% Triton X-100 in a Kintek RQF-3 rapid quenched flow mixer exactly as described (27). This protocol minimized nonspecific ester hydrolysis. Ac-Nle-Tyr-Lys-Arg2MCA was examined using either this procedure or the method previously reported (16), and no difference between these methods was observed.
Stopped Flow Fluorimetry-Z-Nle-Tyr-Lys-Arg2MCA, Ac-Nle-Tyr-Lys-Arg2MCA, and Z-Ala-Tyr-Lys-Arg2MCA (1-200 M) were reacted with Kex2 (260 -520 nM) in an Applied Photophysics stopped flow fluorimeter. The data were typically collected at photomultiplier tube voltages between 550 and 620 V with an offset of 4.99 V using an excitation wavelength of 380 or 385 nm with a 1-nm bandwidth. Emission was monitored with a 460-nm band pass interference filter (10 nm full width at half-maximum band pass; CVI Laser, Albuquerque, NM). Four data points/ms were collected for 100 -1000 ms, and the data were transferred to Kaleidagraph for analysis. For comparison, substrate base lines were prepared using mock reactions lacking enzyme. Base lines were stable, with no significant enzyme-independent change during the time course of the reaction. At least four trials were recorded at each substrate concentration.
Stopped Flow Data Analysis-The data from the individual time courses described above showed an initial drop in fluorescence attributed to completion of mixing in the cuvette, followed by an initial burst and a linear steady state (representative trials shown in Fig. 1). Time points after completion of the initial mixing phase (t Ͼ 1.5 ms) were fit to Equation 1 using nonlinear regression. Equation 1 describes an initial exponential transient followed by a linear steady state with an arbitrary intercept (to accommodate the arbitrary fluorescence units).
In this equation, F(t) is the fluorescence at time t, F 0 is the arbitrary intercept (in V), A is the apparent amplitude of the transient (also in V), k obs is the observed rate of the burst (in s Ϫ1 ), and V obs is the steady-state rate (in V/s). Residuals were calculated as (observed fluorescence Ϫ calculated fluorescence)/(calculated fluorescence) and showed no systematic deviation from the origin as a function of time (data not shown), indicating that the data were adequately described by this model. For Ac-Nle-Tyr-Lys-Arg2MCA it was occasionally necessary to constrain the fit by using the known base line as F 0 . The steady-state rate V obs should follow the Michaelis-Menten equation when followed as a function of substrate concentration (21), so V obs was plotted versus [S]. The resulting plots showed well behaved saturation behavior without apparent substrate inhibition or other complications ( Fig. 2 and data not shown). Kinetic parameters derived from these plots agreed reasonably well with values determined using other procedures (all values within 50%; data not shown).
To examine the acylation phase itself, we plotted the burst rate k obs as a function of substrate concentration [S]. This quantity should follow a modified Michaelis-Menten relationship (Equation 2), with the acylation rate k 2 as the maximal quantity (V max equivalent) and the enzyme-substrate affinity constant K S as the half-maximal binding constant (pre-steady-state K M ; Ref. 21).
The observed data did follow an apparent saturation relationship ( Fig.  3 and data not shown), but it did not seem well behaved. Observed burst rates at higher substrate concentrations deviated from predicted behavior, and data at lower substrate concentrations were equally well described by a linear model until a rate of ϳ550 s Ϫ1 was reached. This rate corresponds to a half-time of 1.3 ms, comparable with the apparent dead time for the instrument of 1.5 ms. These deviations from ideal behavior as predicted by both the saturation relationship and simulations suggested that the apparent saturation could be caused by a machine limit and not by true saturation of the initial burst phase, so we consider the apparent k 2 values to be lower bounds.

Pre-steady-state Rate Constants for the Nle-Tyr-Lys-Arg2
Cleavage Site-We wished to characterize Kex2 specificity throughout the catalytic cycle to expand upon our understanding of how this enzyme recognizes physiologically correct cleavage sites. We therefore focused on the Nle-Tyr-Lys-Arg2 sequence we had previously used to mimic the Met-Tyr-Lys-Arg2 cleavage site found in the Kex2 substrate pro-␣-factor (3,28). Deacylation is known to be rate-limiting for this sequence with amide substrates (28), so the deacylation rate (k 3 ) for this sequence is simply given by k cat . However, we also wished to examine earlier events in the catalytic cycle. We utilized Z-Nle-Tyr-Lys-Arg2MCA and Ac-Nle-Tyr-Lys-Arg2MCA for this purpose. Z-Nle-Tyr-Lys-Arg2MCA was chosen because we suspected that substitutions at P 1 resulted in rate-limiting acylation, necessitating the use of ester substrates to examine the deacylation step, and we had found that tetrapeptidyl methylcoumarin ester substrates (peptidyl MCEs; Ref. 27) and synthetic intermediates leading to them were easier to obtain as solid products with N-terminal Z groups than with N-terminal Ac groups, 3 whereas Ac-Nle-Tyr-Lys-Arg2MCA had previously been used to demonstrate rate-limiting deacylation (28). The use of peptidyl-MCE substrates also necessitated a lower pH to reduce nonspecific hydrolysis (27), but preliminary experiments using rapid quenched flow techniques at either pH 7.26 or pH 6.0 indicated that the pre-steady-state behavior of Kex2 cleavage of Ac-Nle-Tyr-Lys-Arg2MCA was not significantly altered at the lower pH (data not shown).
To examine early events in the Kex2 catalytic cycle (binding and acylation, referred to as K S and k 2 in the tables), we attempted to examine the saturation behavior of the burst phase of Kex2 cleavage of these substrates. When deacylation is rate-limiting, the initial approach to the steady state occurs as an exponential transient whose rate is a function of the substrate concentration, binding constant, and acylation rate (21). Therefore, examination of the apparent burst rate (k obs in Equation 1) as a function of substrate concentration should give a Michaelis-Menten relationship with K S as the presteady-state K M and k 2 as the pre-steady-state k cat . The steadystate portion of the reaction (represented as V obs in Equation 1) should in turn follow a Michaelis-Menten relationship with the known steady-state parameters. Initial experiments using rapid quench techniques suggested that the acylation rate was extremely rapid (data not shown), so we attempted to examine these processes by stopped flow fluorimetry.
Kex2 cleavage of Z-Nle-Tyr-Lys-Arg2MCA was examined at 21°C over a range of substrate concentrations from 1 to 100 M, and the resulting data were fit to Equation 1 (representative trials at 5 and 25 M substrate are shown in Fig. 1).
Analysis of the fitted data showed that the steady-state rate V obs did indeed follow the expected behavior (Fig. 2), and the apparent burst rates initially seemed to exhibit saturation. However, detailed examination of the pre-steady-state rates (k obs ) showed that there might be significant deviations from the predicted behavior at higher substrate concentrations, where the observed burst rate was ϳ550 s Ϫ1 (Fig. 3). The observed rate constants were approximately linear with substrate concentration until this point was reached and then diverged from predicted saturation behavior to remain at this apparent limiting rate (Fig. 3B). We were therefore concerned that the apparent saturation was due to a machine limit, in keeping with the observation that a rate constant of 550 s Ϫ1 corresponds to a half-time that is quite close to the dead time of the instrument (1.3 ms versus 1.5 ms; "Materials and Methods"). Therefore, we have chosen to report this value as a lower bound for the acylation rate constant (k 2 Ն 550 Ϫ1 ; Table I).
The lack of a reliable value for k 2 precludes reliable determination of K S using this approach, because the binding preequilibrium is measured as the pre-steady-state K M and hence as the substrate concentration giving half-maximal k obs . However, the linear relationship between k obs and substrate concentration below the machine limit implies that this range of substrate concentrations ([S] ϭ 25 M) is at or below K S for this substrate. Therefore, we are able to place a lower boundary on the binding constant for this substrate (Table I). Additionally, this suggests that the true acylation rate could be at least 2-fold higher than the apparent value, because a range of [S] at or below K S would imply that the observed rate constant k obs is at or below 1 ⁄2 of the maximal k obs (from the definition of K S and k obs ). However, we have chosen to report the experimental values throughout this paper.
Similar results were obtained in stopped flow characterization of Ac-Nle-Tyr-Lys-Arg2MCA (Table I). We again found that the acylation rate was equal to or greater than 550 s Ϫ1 . However, we were able to estimate a slightly lower boundary for K S with Ac-Nle-Tyr-Lys-Arg2MCA (Table I) 1% Triton X-100 at 21°C. Points recorded before mixing was complete (t Յ 1.5 ms) are shown as dashed lines; these data were omitted from fitting. The remaining data were fit to Equation 1 using nonlinear regression (smooth curves). As can be seen, both the linear steady-state rate (V obs in Equation 1) and the apparent burst rate k obs are increasing with increasing substrate concentration, as is the burst amplitude A. Four points/ms were recorded for 100 ms; only the first 50 ms are shown here to allow visualization of behavior at short times, but all data after 1.5 ms were used for fitting.  Fig. 1). At least four trials were conducted at each substrate concentration. Here, the steady-state rates (V obs ) determined by fitting Equation 1 to those data were plotted versus substrate concentration and fit to the Michaelis-Menten equation, giving a maximal V obs ϭ 39.4 V/s and K M ϭ 4 M. The observed burst amplitudes at high substrate concentration were averaged and used with the known enzyme concentration to normalize V obs , giving a k cat value of 19 s Ϫ1 under these conditions. These steady-state parameters are in accord with work using conventional saturation kinetics (data not shown). difference need not imply a difference in affinity (it could instead imply a difference in acylation rate).
We were able to exploit the rate-limiting deacylation exhibited by these substrates to conveniently measure k 3 as k cat . Once again, there is little difference between these two substrates, with both exhibiting deacylation rates between 15 and 20 s Ϫ1 (Table I). We also examined deacylation under alternative conditions, either a lower pH (6.0 versus 7.26) required for work with peptidyl-MCE substrates (27) or a higher temperature (37°C, as has been used in prior characterization of Kex2; Refs. 14 -16). The two substrates responded slightly differently to the changing conditions, with deacylation of Z-Nle-Tyr-Lys-Arg2MCA actually improving slightly at the lower pH, but all effects were small (Table I).
Even though we were unable to measure binding and acylation precisely, these experiments demonstrate several important points about cleavage of this sequence. First, the affinity for the substrate is rather low. Even with correct sequences, K S is greater than 10 M; for comparison, K M for this substrate is ϳ1 M. Moreover, these data show the extent to which the enzyme has evolved for fast acylation, because the attack on the amide bond is at least 30-fold faster than subsequent hydrolysis of the resulting ester. We therefore examined how substitutions in this substrate sequence altered the enzymesubstrate interactions during the catalytic cycle.

Pre-steady-state Rate Constants for the Nle-Tyr-Lys-Lys2 Cleavage Site: Effects of a P 1 Substitution-Previous work with
Kex2 protease had shown that substitution of Lys for Arg at P 1 results in significant decreases in k cat and k cat /K M (14,16). This substitution also results in a change in rate-limiting step, because substrates with Lys at P 1 no longer exhibit burst kinetics in formation of the acylation product (16,28). This lack of an initial burst with such sequences suggests that acylation is severely impaired for such sequences, such that it has become rate-limiting. To test this hypothesis, we compared Kex2 cleavage of a tetrapeptidyl methylcoumarin ester, Z-Nle-Tyr-Lys-Lys2MCE (27), to a similar methylcoumarinamide, Ac-Nle-Tyr-Lys-Lys2MCA (16). If the amide substrate is limited by acylation, the ester substrate should show an increase in k cat caused by its more rapid acylation, as is the case here (Table II and Ref. 27). This indicates that the substitution of Lys for Arg at P 1 does produce a dramatic reduction in acylation rate, such that acylation itself is now rate-limiting. This also implies that simple saturation kinetics with the amide substrate will measure K S and k 2 .
We therefore measured these quantities under conditions equivalent to those used in stopped flow characterization of equivalent substrates with P 1 Arg (Ac-Nle-Tyr-Lys-Lys2MCA versus Ac-Nle-Tyr-Lys-Arg2MCA; Table II). Although we were only able to place lower boundaries on these values for the Arg-containing cleavage sites, the comparison is nonetheless informative. Comparison of K S values shows that the physiologically correct (i.e. P 1 Arg) substrate is only bound at most 4-fold more tightly than the incorrect substrate. However, the acylation rates are dramatically different; in fact, the P 1 substitution of Lys for Arg results in a minimum defect of over 200-fold (over 3 kcal/mol in ⌬⌬G ‡ , calculated as in Ref. 21). Thus, the enzyme-substrate interaction at P 1 is utilized almost entirely for transition state stabilization at the first irreversible step, acylation. The data for substrate concentrations at or below 25 M were also fit by linear regression (dashed line). The equivalent results yielded by these two fitting models and the deviation from Michaelis-Menten behavior at higher substrate concentrations suggest that the apparent saturation could be due to a machine limit. B, the data from A are presented on a double-reciprocal plot and fit by linear regression. Inset, data with [S] ϭ 10 M are shown to aid in examination of the linear fit.
Measured or estimated by stopped flow at 21°C in 0.2 M Bistris, pH 7.26, 1 mM CaCl 2 , 0.1% Triton X-100. k 2 and K S were derived from plotting k obs versus [S] as shown in Fig. 3. k 3 was determined from plots of V obs versus [S] as shown in Fig. 2. c Measured or estimated by stopped flow at 21°C in 0.2 M Bistris, pH 7.26, 1 mM CaCl 2 , 0.1% Triton X-100. k 2 and K S were derived from plotting k obs versus [S] (Fig. 3).
d Measured by rapid quenched flow at multiple substrate concentrations in 0.2 M Bistris, pH 7.26, 1 mM CaCl 2 , 0.1% Triton X-100 at 21°C. The observed rate was insensitive to substrate concentration, indicating that the reaction was saturated. Analysis of stopped flow data and a previously published temperature dependence (28)  To measure deacylation, we examined the ester substrate Z-Nle-Tyr-Lys-Lys2MCE. Unlike the equivalent amide, this substrate does exhibit burst kinetics (27). Therefore, the steady-state k cat for this substrate provides a convenient approximation to the deacylation rate constant k 3 (25). Comparison of the deacylation rate for this sequence with the deacylation rate for the physiologically correct equivalent (Z-Nle-Tyr-Lys-Arg2MCA versus Z-Nle-Tyr-Lys-Lys2MCE; Table II) shows that these rate constants are within 50% of each other, in dramatic contrast to the results obtained at the acylation step. Thus, an enzyme-substrate interaction that is key for efficient acylation is essentially uninvolved in deacylation.
Pre-steady-state Rate Constants for the Ala-Tyr-Lys-Arg2 Cleavage Site: Effects of a P 4 Substitution-Unlike substitutions at P 1 , substitutions at P 4 do not result in a loss of burst kinetics for Kex2 cleavage of amide or ester substrates (data not shown). Therefore, stopped flow analysis was used to determine K S and k 2 in Kex2-dependent cleavage of Z-Ala-Tyr-Lys-Arg2MCA, and saturation kinetics were used to examine k 3 . Stopped flow fluorimetry confirmed that cleavage of this substrate proceeded with burst kinetics, but it once again proved impossible to measure K S and k 2 rigorously because of an possible machine limit (data not shown). We were able to estimate lower boundaries for these quantities (Table III), but these boundary values were essentially identical to those of the "correct" substrate, which has Nle at P 4 (Z-Nle-Tyr-Lys-Arg2MCA; Table III). Examination of the deacylation step with these substrates showed that the incorrect P 4 residue resulted in a slight increase in deacylation rate (Table III), but this effect was Ͻ3-fold.
To better understand the interplay between these two positions, we also examined the effect of a P 4 substitution on the individual steps in cleavage of sequences containing Lys at P 1 (Table III). In these substrates, acylation was rate-limiting, so K S and k 2 were measured by saturation kinetics with the amide substrates, and k 3 was measured by saturation kinetics with the equivalent esters. This work demonstrates that the P 4 side chain can impact binding, with K S for Ac-Nle-Tyr-Lys-Lys2MCA ϳ7-fold lower than that for Ac-Ala-Tyr-Lys-Lys2MCA (a 1.1 kcal/mol difference in ground-state binding). Additionally, in the context of the incorrect Lys at P 1 , the correct P 4 side chain did accelerate acylation (Table III), as we had previously proposed (16). This effect corresponds to ϳ1.8 kcal/mol utilized for transition state stabilization during the acylation reaction.
Examination of deacylation rates for Z-Nle-Tyr-Lys-Lys2MCE and Z-Ala-Tyr-Lys-Lys2MCE showed that they were essentially identical (Table III). Thus, Kex2 seems to utilize neither the enzyme-substrate contact at P 4 nor that at P 1 at the deacylation step, in marked contrast to acylation. DISCUSSION We have characterized the roles different enzyme-substrate interactions play at different steps in the Kex2 catalytic cycle. We found that the P 4 residue is used for both acylation (ϳ1.8 kcal/mol) and binding (ϳ1.1 kcal/mol) with little effect on deacylation (Table III). In contrast, the P 1 residue seems to primarily act at the acylation step (Ն3.1 kcal/mol), with little effect on either binding or deacylation (Tables II and III). Even the relatively conservative substitution of Lys for Arg at P 1 results in a large defect in acylation such that it becomes rate-determining. Thus, this enzyme seems to generate specificity primarily at the acylation step and not to utilize the same interactions for the chemically related deacylation step.
The approach we have taken to measure binding and acylation presupposes a one-step binding process with rapid preequilibrium. It is important to note that we have not experimentally validated this assumption, nor have we been able to satisfactorily rule out the related possibility that Kex2 cleavage of substrates with P 1 Arg is diffusion-controlled. However, the results of experiments with viscogens such as sucrose or glycerol are inconsistent with a diffusion-limited reaction. 3 Additionally, although two-step binding processes are not unknown for some serine proteases (21), we know of no evidence that requires a two-step process in Kex2 substrate binding. We therefore favor the minimal interpretation that substrate binding is a one-step process, followed by slower acylation with discrimination between Lys and Arg at P 1 occurring at the acylation step.
We have also made no attempt to measure product release, because we expect it to be rapid because of several lines of evidence. First, steady-state Kex2 cleavage of peptidyl-MCA substrates can be carried out with product concentrations approaching millimolar levels without detectable deviation from linearity as long as the remaining substrate is saturating, 3 suggesting that accumulation of substantial product does not significantly affect the kinetic behavior of this enzyme. Moreover, no product inhibition could be detected with the equivalent cleavage product Ac-Nle-Tyr-Lys-Arg at concentrations as high as 1 mM (28). Finally, the deacylation step for Ac-Nle-Tyr-Lys-Arg2MCA has been examined directly using a new mass spectrometric assay, and product release was found to be much faster than deacylation and therefore kinetically insignificant (28).
It is also important to consider to what extent the cleavage of b Obtained by saturation kinetics with Ac-Nle-Tyr-Lys-Lys2MCA in 0.2 M Bistris, pH 7.26, 1 mM CaCl 2 , 0.1% Triton X-100 at 21°C, using the Michaelis-Menten equation with k cat ϭ k 2 and K M ϭ K S . c Obtained by saturation kinetics at 21°C in 0.2 M sodium acetate, pH 6.0, 1 mM CaCl 2 , 0.1% Triton X-100, using Z-Nle-Tyr-Lys-Arg2MCA and Z-Nle-Tyr-Lys-Lys2MCE. For these substrates, k cat ϭ k 3 .  synthetic tetrapeptidyl substrates with coumarin leaving groups will accurately mimic the cleavage of authentic precursors. Although these substrates are more soluble than substrates in which cleavage occurs at a peptide bond and are better behaved kinetically (15), coumarin leaving groups display very low pK a values for substituents at the 7 position. For example, the 7-amino-4-methyl coumarin pK a is below 3, whereas the pK a for the 7-hydroxy group in the ester substrates is ϳ5.3. 3 Thus, these leaving groups are significantly more susceptible to nucleophilic attack by the active site serine than the peptide leaving groups found in authentic Kex2 substrates in vivo (21,22). However, peptidyl2MCA substrates are cleaved by Kex2 with k cat /K M values essentially identical to those obtained in cleavage of actual peptide bonds (15). For serine proteases, k cat /K M is determined by k 2 /K S , the ratio of the acylation rate to the enzyme-substrate binding constant (21). Therefore, for the values to be equivalent for these two types of substrates, any increase in acylation rate seen with coumarin leaving groups must be offset by an approximately equal decrease in affinity. Thus, the peptide leaving group may slow acylation but must simultaneously improve ground-state binding for this step. Such behavior is indicative of interactions that stabilize the ground state rather than the transition state (21,22), and therefore any as-yet unmeasured contact between Kex2 and authentic peptide leaving groups cannot be used to accelerate catalysis. Obviously, this argument does not preclude the possibility that such contacts could play an important role in initial substrate binding, but they cannot be used for driving either chemical step in the reaction.
Despite these caveats, this work nonetheless demonstrates that Kex2 protease carries out the acylation step extremely rapidly with substrates containing the correct P 1 residue (Arg), whereas neither the interaction with this residue nor the interaction with the P 4 residue are utilized to accelerate deacylation. Taken together with other work on P 2 substitutions, 3 these data suggest that side chain interactions are not utilized at all during the deacylation step. It is thus likely that interactions with the substrate backbone are sufficient to position the acylenzyme ester for hydrolysis, a hypothesis that could conceivably be examined by synthesis of substrates containing substitutions within the peptide backbone.
It is also intriguing to speculate as to why the deacylation rate constant is so low. Nonspecific Kex2 homologs in the subtilisin family also seem to exhibit little side chain specificity at the deacylation step (9), but the deacylation rates for these enzymes are nonetheless significantly higher than those seen with Kex2 (usually Ն300 s Ϫ1 for oligopeptide esters; Ref. 9). This raises the possibility that there may be some structural feature of the Kex2 acylenzyme that impedes hydrolysis. For instance, the precise binding of correct substrates to the Kex2 active site may simply favor attack at one face of the scissile carbonyl but not the other, because deacylation and acylation occur on opposite faces of the scissile carbonyl in serine proteases (29 -31). Alternatively, such behavior could result from increased conformational freedom of enzyme side chains in the vicinity of the scissile bond after release of the C-terminal cleavage product. If one or several side chains are now free to occupy new conformations that partially occlude the optimal trajectory for attack on the acylenzyme by water (31), the result could be a decrease in the efficiency of deacylation such as that we have observed.
This work clearly shows the extent to which the enzymesubstrate interactions of Kex2 are optimized for stabilizing the acylation transition state. Because this transition state is that for the first irreversible step, it is clear that Kex2 utilizes enzyme-substrate binding energy exceptionally well at this pivotal point in the catalytic cycle. It is hoped that a crystal structure for Kex2 or a close homolog will soon become available, shedding new insight on how this enzyme has utilized the structural chassis of the subtilisins to focus on a single step in the classical serine protease mechanism.