Mechanistically significant diastereoselection in the sulfoximine inhibition of carboxypeptidase A.

The diastereomers of L-alpha-[ [S-(2-phenylethyl)sulfonimidoyl]methyl]benzenepropanoic acid bind differentially to carboxypeptidase A. These putative transition state-analogue inhibitors show unique and interpretationally significant pH dependences for Ki, as well as for the visible absorption spectra of their E.I complexes in the case of the cobalt-substituted enzyme. From the geometry of the enzymically preferred isomer, it may be concluded that the mechanism of peptide scission by the enzyme entails addition of a nucleophile to the si face of the bound-substrate prochiral carboxamide linkage. New interpretational constraints on the mode of action of the enzyme are thereby imposed.

The diastereomers of ~-a-[[S-(2-phenylethyl)sulfonimidoyl]methyl] benzenepropanoic acid bind differentially to carboxypeptidase A. These putative transition state-analogue inhibitors show unique and interpretationally significant pH dependences for Ki, as well as for the visible absorption spectra of their E-I complexes in the case of the cobalt-substituted enzyme. From the geometry of the enzymically preferred isomer, it may be concluded that the mechanism of peptide scission by the enzyme entails addition of a nucleophile to the si face of the bound-substrate prochiral carboxamide linkage. New interpretational constraints on the mode of action of the enzyme are thereby imposed.
The carboxamide functional group, as found in peptides, is a prochiral unit. Each such residue possesses a re and a si face, which in principle may have a different susceptibility to nucleophilic carbonyl addition, as occurs in enzymic proteolysis for example.
Indeed, because of the stereospecificity which is the hallmark of enzymic processes, it can be asserted that one or the other of these two faces in a substrate must be attacked exclusively in the catalytic process at the active site of any given proteolytic enzyme. However, all trace of this stereoselection is dissipated in the final products of enzymic peptide scission, so that this aspect of enzyme mechanism has been largely ignored in studies directed a t explaining the mode of action of proteases. Of course this information is known in the case of certain enzymes as a consequence of crystallographic evidence and an established mechanism, notably for the serine proteases, in which an enzyme-attached nucleophile has restricted scope for attacking the bound substrate. However, for other enzymes of uncertain mode of action, in particular carboxypeptidase A, an independent answer to this question of nucleophile stereopreference could yield a definitive indication of mechanism. The challenge is to provide a suitable experimental approach, so as to discern the geometry and the energetics of the essential enzymic interactions involving the tetrahedral adduct that is created, along with its associated transition states, in the course of peptide scission. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
$. TO whom correspondence should be addressed. itors for carboxypeptidase A and for the similar metalloprotease thermolysin, with the intention of approximating a tetrahedral adduct intermediate (1)(2)(3)(4)(5). However, crystallographic examination of enzyme complexes of such species has not resolved the mechanistic question. For several of these probes the tetrahedral moiety occupying the locus of a scissile linkage of a typical substrate (e.g.  seems to coordinate variably in a mono-or bidentate fashion with the active site zinc ion, in what appears generally to be mixed inner and outer sphere complexation (6,7). In this article we describe the inhibitor characteristics with regard to carboxypeptidase A for a diastereomeric pair of sulfoximine-containing substrate analogues, 3 and 4, that are found to be active in the micromolar concentration range. The sulfoximine group nitrogen atom (but not the oxygen) is sufficiently basic to coordinate strongly with a metal ion. The sulfoximine functionality also possesses a stable tetrahedral configuration, which in the cases shown presents a chiral ligand to the zinc ion within the enzyme. Consequently, the preferred geometry for a tetrahedral adduct analogue interacting with the active site may be revealed by a comparison of these two species as enzyme inhibitors.

6393
Sulfoximine Inhibition of Carboxypeptidase A nine (FAPP, K,,, = M ) (11) from Sigma. The synthesis, resolution, and absolute configuration of inhibitors 3 and 4 have been described (12); sulfone 6 was similarly prepared by peracid oxidation of the corresponding thioether intermediate from the preparation of 3 and 4, IR 1120 and 1300 cm".
Kinetic Analysis-The pH dependence of K; for inhibitors with both ZnCpA and CoCpA was determined a t 25.0 ( f O . l ) "C in buffers previously listed with FAPP as assay substrate at an initial concentration of less then K,, with spectrophotometric (328 nm, 1-cm path length) analysis and the method of initial rates. Enzyme concentration was maintained well below inhibitor concentration in Kj determinations (100-fold). Values of K, were obtained a t each pH by a nonlinear least-squares fit of data from perturbation of kJKm to the appropriate inhibition equation.
All pH values in this article are calibrated pH meter readings uncorrected for ionic strength effects. Tolerances listed are standard errors from least-squares analysis.
Spectral Titrations-Visible absorption spectra for concentrated solutionsof CoCpA (0.17-0.69 mM, 1-5-cmpath length) were recorded conventionally at various values of pH in the range 6-10 in the presence of the inhibitors 3 or 4. It was verified that [I] > Ki a t all p H values so examined. The intrinsic pK, , for the sulfoximine group has not been reported hitherto. We have secured a value for dimethyl sulfoximine by fitting its "C NMR chemical shift ( 8 relative to CH:SOCH:, or (CH,),SO,, internal reference) against the previously devised Hacidity function scale (13) in the alkalinity range of 0.1-15 N NaOH (an appropriately sigmoidal +4 ppm shift of signal was noted, corresponding to formation of the N-anion). A pK, , value of 15.8 ( f O . l ) was thereby obtained.

RESULTS
Design of Inhibitors-The primary catalytic specificity of carboxypeptidase A is for cleavage from peptides of individual C-terminal L-amino acid residues having a hydrophobic side chain. There also seems to be a secondary lipophilic interaction of the enzyme with the penultimate amino acid residue, for among the more susceptible peptides are acyl-PhePhe species such as 5 , the assay substrate herein employed (in which the free L configuration carboxyl substituent is also an essential specificity feature). With this knowledge, 3 and 4 were constructed in which the benzyl groups (Bzl) correspond spatially to those in 5 , and in which the sulfoximine group resides in the scissile (*IhcEJc ) carbonyl position. Furthermore, 3 and 4 have been resolved into their respective enantiomers, with relative configurations unambiguously established by crystallography (12). As a control, the substrate analogue 6 has also been prepared. The sulfone oxygens are noncoordinating, so that an inhibition comparison with 3 and 4 should reveal the significance of any sulfoximine-metal ion interaction at the enzyme active site. Due to extensive previous structural investigations of carboxypeptidase A complexed with similar substrate analogues, the potential mode of ligation of the inhibitors to the enzyme can be identified with some confidence. Inhibition Characteristics-In fact, absolute configurational assignments for 3 and 4 correlate with the enzyme's ability to recognize only an L configuration for the C-terminal amino acid residue in a peptide substrate. The more tightly binding enantiomers of 3 and 4, which by their synthesis possess an identical chirality at the methine carbon center, also have the same spatial arrangement next to the carboxyl group as has L-phenylalanine (12). Hence, they are given the L designation. The enzyme affinity of D-3 as measured by the competitive K , is at least 47-fold poorer than for L-3 at a pH of 7. The observed K; for D-4 is likewise greater than for L-4 (factor of 12). However, our synthesis of D-4 apparently yields some contamination with the extremely tight binding L-3, so that the true affinity of D-4 is probably masked. Subsequently described inhibition studies employ only the L-enantiomers, which are hereafter simply designated 3 and 4, and which have the absolute configurations depicted.
Both 3 and 4 have been shown to be strictly competitive, rapid equilibrium inhibitors in the enzymic hydrolysis of peptide substrate FAPP ( 5 ) by carboxypeptidase A. With 3 at concentrations of 0.58 or 1.78 pM at a pH of 7.6, there was a substantial retardation of k,,,/K,,,, but no perturbation of k,,,. Similar results were obtained with 4. Illustrative experimental competitive inhibition constants (pH 6.0-6. The much weaker affinity of 6 for carboxypeptidase A tends to confirm that the mildly basic sulfoximine nitrogen of 3 or 4 (pK, of -3 for conjugate acid, 14) does interact with the active site metal ion. The sulfone oxygens are so weakly basic (estimated pKo of -11 for conjugate acid, 15) that no coordination to a metal ion in competition with solvent water seems possible; indeed, 6 binds to the enzyme scarcely more strongly than does the "specificity fragment" 3-phenylpropanoic acid. Consequently, an explicit enzymic interaction involving the nitrogen (but not the sulfoximine oxygen) of 3 or 4 seems indicated.
p H Dependence for K;-The inhibitors 3 and 4 exhibit rather different dependence upon pH in their binding to carboxypeptidase A, and allowance must be made for this in comparing them. As shown in the Dixon plots presented in Fig. 1, the K, value for the more tightly bound inhibitor 3 appears invariant as pH increases up to a value of 9, above which enzyme affinity for 3 diminishes (pKi decreases). In contrast, 4 exhibits a gradually increasing K; with increasing pH. Greatest affinity of 4 for the enzyme is seen at a pH of -6 (pK; maximum), and at a pH of 9 the values of K, for 3 and 4 differ by 1,000-fold. Similar data have has been collected for cobalt carboxypeptidase A (CoCpA), the corresponding enzyme in which Co2+ has been substituted for Zn2+ in the active site (Fig. 2). This disparity in behavior for Breaks in asymptotic straight lines (slopes -1, 0, and +1) denote pK, values obtained by least-squares fitting of data (reaction velocity uers'sus concentration of H+) to a simple bell-shaped curve, with resulting kinetic parameters (kcat/Km)lim = 3.13 (f0.07) X lo6 M" s-', pK. = 5.94 (k0.06) and 9.23 (f0.06). Bottom: pKi uersus pH for 3 (filled circles) and for 4 (filled squares); breaks in asymptotic lines indicate fitted pK, values, as given in Table I   Apparently pK2 = pK1; therefore, no measurable inflection in pH profile for either. Inserted as a fixed value (corresponding to the lower pK, in kc,,/ K,) in fitting data to Scheme 1, in order to limit the number of adjustable parameters.
Inserted as a fixed value (corresponding to the higher pK, in kc,,/ K , and in K, for 3) in fitting data to Scheme 1 in order to limit the number of adjustable parameters.
Single point determination at pH of 5.98 K, increases with pH (14), a value of 2.3 mM was measured for 6 a t a pH of 7.6.
diastereomeric inhibitors differing only in configuration a t sulfur is potentially very significant mechanistically. Consequently, an endeavor at rationalizing the pH dependences is justified.
A manner in which this is to be done for the apparently complex pH dependence exhibited by 4 has been suggested previously (14). There are well-known enzymic deprotonations characterized by pK, values of -6.1 and -9.2, which affect catalytic behavior of the enzyme. The log(kcat/K,) uersus pH profile depicted at the top of Figs. 1 and 2 indicates a quenching of enzymic activity in acidic as well as in alkaline medium, as has been repeatedly observed elsewhere. The acidic limb pK, is known to be attributable to deprotonation of a water molecule which is coordinated to the active site metal ion in the absence of substrate (16). The alkaline limb pK,, is likewise associated with deprotonation of an active site metal ion ligand, most certainly the imidazole ring of His-196 in the free enzyme (17). The former assignment follows from the unambiguous reverse protonation inhibition characteristics of an analogue of 3 or 4 that presents a phenolate ligand to the active site metal ion; the latter assignment derives from NMR spectral evidence with CoCpA. Because these prototopic changes are so intimately involved in the processing of substrates by the enzyme, similar involvement in the binding of 3 and 4 should be anticipated. Indeed, analogous breaks in the pKi uersus pH profiles for 3 (alkaline limb) and for 4 (acidic limb) are cleanly observed. However, the gradual falloff in apparent pK; exhibited by the curves for 4 indicates an additional deprotonation occurring specifically in the E . I complex, with an intermediate pK, value. The fitted curve for 4 postulates a double inflection on the alkaline limb, with the same downward tilt at a pH of -9.2 as seen for the other pH profiles in the figures but with a compensatory upward inflection at slightly lower pH (specifically that of an E . I complex, 18). Scheme 1 provides a diagrammatic summary of our interpretation, while Table I lists parameters ((KJli,,,, pK1, pK,, pKJ fitting the pH dependence of Ki to the appropriate kinetic equation, also given in Scheme 1. The pKi uersus pH profile for 3 with ZnCpA is seemingly more plain, with only pK3, the alkaline limb pK,, discernible ( Fig. 1). However, we shall suggest that the curves for 3 and 4 actually have the same form, and that pKl and pK2 are merely hidden for 3 because they coincidentally have nearly the same value. If the intermediate pK2 inflection noted in Fig. 1 for the E . 4 complex were in the case of E . 3 shifted upward and to more acidic pH values, specifically to the vicinity of pKl, a cancellation would occur and the overall curve should be leveled as is observed. By this hypothesis both curves may be reconciled with a common explanation (Scheme 1). The pKi uersus p H profile for 3 with CoCpA is similar (Fig. Z), and in this instance we are able to show a small double inflection on the acidic limb in order to demonstrate the point. (Although the data here are ambiguous as regards the small inflection, pK, as obtained from the kcat/K,,, profile is known to have a lower value for CoCpA.) Fortunately, additional evidence exists that the E . I deprotonation detected by kinetic means for 4 (namely, pK2) does in fact take place also in the E . I complexes of 3 in more acidic solution.
Visible Spectra of CoCpA Complexes-Direct examination of E . I complexes is feasible in the case of cobalt-substituted carboxypeptidase A because the enzyme with Co2+ chelated at the active site has a characteristic visible spectrum due to the metal ion (9). The cobalt enzyme by itself in saline solution has a broad absorption maximum centered on 550 nm ( e -150 M" cm") with shoulders near 500 and 570 nm, as depicted in Figs. 3 and 4 (dotted line). The spectrum is noticeably perturbed upon addition of 1 molar equivalent of 3 or 4 but not further with extra increments of inhibitor. Moreover, the spectra of the E . I complexes are conspicuously pH-dependent. In the case of 3, the overall spectral intensity is enhanced upon addition of inhibitor, as may be seen in Fig.  3, and the longer wavelength shoulder becomes a distinct maximum at 567 nm in neutral solution. However, the latter peak clearly shrinks (absolutely as well as relative to the shorter wavelength maximum) as the pH is lowered, with an apparent isosbestic near 587 nm (change reversible on pH restoration). The absorption uersus p H profile has been sat- isfactorily fitted to a sigmoidal curve, yielding a pK, of 6.55 (f0.07) characterizing the transition. The spectral perturbation is even more marked in the case of 4; as seen in Fig. 4, a pronounced maximum at 571 nm is induced upon complexation with the enzyme in neutral saline solution. With increasing pH this peak reversibly diminishes and is shifted to a longer wavelength, and an additional shoulder appears to grow near 625 nm (isosbestics at 545, 596, and 647 nm). A fit of the absorption uersus pH profile gives a pKo of 8.54 (f0.09) in this instance. While it is evident that 3 and 4 create rather different ligand field environments for the active site metal ion in their complexes with CoCpA, meaning cannot presently be attached to the shapes of the individual curves in Figs. 3  and 4. However, the pH dependences collectively provide autonomous evidence of critical significance. Apparently within each E. I complex a functionality bearing a dissociable proton exists in intimate association with the metal ion. That group is 100-fold more acidic in E. 3 than in E . 4, resulting in a 2 pK-unit difference in the onset of spectral perturbation, as seen in Figs. 3 and 4. Because inhibitors 3 and 4 differ only in sulfoximine stereochemistry, and because the sulfoximine residue is a plausible mimic of the transient tetrahedral adduct in peptide hydrolysis, this observation has major ramifications mechanistically.
Reconciliation of Spectra with Kinetic Data-What is the identity of the functionality yielding the pH-induced spectral perturbation? That residue clearly regulates inhibitor binding, for in each case an inflection in the pKi uersus pH profiles (pK, values for 3 .CoCpA, 6.24, and for 4.CoCpA, 8.65) correlates with those obtained spectroscopically (pK, values for 3. CoCpA, 6.55, and for 4 . CoCpA, 8.54). The most reasonable explanation is that the sulfoximine NH of 3 or 4 is strongly acidified in consequence of coordination to the active site metal ion; i.e. the pertinent proton dissociation may be formulated as (Enz)M2+t:NH = S(0)R2 (Enz)M+-

N=S(0)R2 + H'. Such an ionization should perturb the Co2+
visible spectrum and could be expected to show an influence from inhibitor configuration. Moreover, it rationalizes the pH profiles for Ki with the inhibitors. As seen most clearly with 4, pKi diminishes above a pH of 6 as a consequence of the fact that hydroxide is the external ligand completing the coordination sphere of the active site metal ion above that pH in the absence of inhibitor or substrate. (As previously noted, the pK, of -6 seen in the k,,,/K,,, profiles of Figs. 1 and 2 has been shown to be due to (Enz)M2++:OH2 (16,17)). The comparatively weakly basic sulfoximine nitrogen of 4 cannot itself compete effectively with OH-as a ligand, but the conjugate base R2S(0)=Nis able to do so. Therefore, the apparent value of pKi decreases until the medium is sufficiently alkaline to deprotonate the metal ion-activated sulfoximine group within the E.1 complex, whereupon the pH profile levels off. This occurs near a pH of 8.5 in the case of 4, but closer to a pH of 6 with 3. While such strong acidification of an NH group might seem questionable, it is entirely consistent with the potent Lewis acid character of the active site metal ion. The pK, of -6 observed for (Enz)M2+c:OHz in the absence of inhibitor means that a solvent water molecule (intrinsic pK, of -15.7) has been acidified >109-fold by coordination to the metal ion. Independent determination of the acid dissociation constant for dimethyl sulfoximine (by 13C NMR spectral perturbation, employing a basicity function extrapolation based on overlapping indicators) yields an almost identical intrinsic pKa of 15.8. It is not at all unreasonable that the enzyme should similarly augment the acidity of a bound R2S(0)=NH group.

DISCUSSION
Although several tetrahedral adduct-analogue inhibitors for carboxypeptidase A have been described, a significant differential in enzyme affinity corresponding to the stereochemistry of the metal ion ligating moiety has not been reported previously.2 In the present instance, we find diastereomer 3 to bind from 10-to 1,000-fold more tightly than 4, depending on pH. Particularly revealing is the 100-fold greater acidification of the sulfoximine NH proton for 3 in consequence of nitrogen coordination to the metal ion. It appears that the active site metal ion exhibits a considerably greater Lewis acidity toward one diastereomer, for the ease of N-deprotonation ought to correlate directly with the dative electron-withdrawing capacity of the metal ion engaged with the sulfoximine nitrogen. This apparent Lewis acid anisotropy within the active site has ramifications for the catalytic mechanism, for it seems as if the acidity of the metal ion becomes focused on a single apex of the pyramidal species that is presented to it.
Independent kinetic evidence suggesting that the active site metal ion functions as a Lewis acid in the course of peptide scission exists in the form of the synergism test (20). We refer to the finding that the native enzyme (ZnCpA) is a poor * There is a single report involving diastereomeric phosphonamidothioate-containing inhibitors (19), which is inconclusive, however. catalyst for the cleavage of a thiopeptide substrate analogue (PhCONHCH2CONHCH2C(=S)NHCHBzlC02H), but that the thioamide linkage is cleaved efficiently by cadmium carboxypeptidase A (CdCpA, the enzymic species in which the active site Zn2+ has been substituted by Cd"). Furthermore, CdCpA is a poorer enzyme than ZnCpA for normal, carboxamide-containing peptide substrates (kJK,,,). This behavior is explained by polarizability considerations in the dative interaction between substrate and enzymic metal ion (hardsoft acid-base theory; "soft" acid Cd2+ beneficially interacts with soft base S, but " h a r d Zn2+ prefers hard 0 in substrate carbonyl). However, this behavior only makes sense if the metal ion functions in Lewis acid fashion to polarize the substrate carbonyl in the course of peptide hydr~lysis,~ as is also suggested by enzyme crystallographic investigations.
When these latter observations are combined with the results of the present investigation, a hypothesis regarding the stereochemical course of peptide scission at the active site of carboxypeptidase A emerges. The diastereomer yielding the stronger interaction with the enzyme (namely, 3) will be the one in which the sulfoximine nitrogen occupies the place of substrate-carbonyl oxygen in the tetrahedral adduct intervening in normal substrate hydrolysis. This defines the direction of pyramidalization of substrate carboxamide; the nucleophile attacking to form a tetrahedral species must approach toward the si face of the peptide linkage, in order to end up in the position of the sulfoximine oxygen atom in 3. This avenue in fact comprises the more solvent-exposed region of an E . S complex. Presented in Fig. 5 is a stereoview of 3 complexed with the active site region of carboxypeptidase A; congruity with the reaction intermediate is inferred.
What are the mechanistic implications of this conclusion? Foremost is a disavowal of an active role for the carboxylate side chain moiety of glutamic acid residue 270. This functional group has been repeatedly invoked in speculative mechanisms for carboxypeptidase A, either as a nucleophile or as a general base facilitating the entry of a water molecule into the scissile linkage. However, convincing evidence for either role is lack-'' Other investigators have apparently arrived independently at this same conclusion (21).
ing. In particular, the acidic limb pK, of -6.1, which is invariably found in the enzyme kinetics (kcat/Km and competitive Ki) and which was formerly attributed to that residue, is now known to be due to deprotonation of a water molecule which in the free enzyme binds to the active site metal ion and becomes replaced by substrate as previously indicated, rather than to the carboxyl side chain of Glu-270 (16, 17). Much additional evidence purportedly indicating proton exchange between this carboxylate residue and bound substrates is dubiously inte~preted.~ Although presence of the carboxylate may be essential for a functional enzyme (26), it could be simply providing a negative charge for electrostatic stabilization of electron-deficient intermediates in the catalytic cycle. Our reason for relegating Glu-270 to such a passive (although not necessarily insignificant) role is that its carboxylate clearly is impossibly placed to facilitate nucleophilic addition by deprotonation of an attacking water molecule, according to Fig. 5 . In widely circulated crystallographic depictions of carboxypeptidase A, Glu-270 is squarely placed over the re face of the bound (and metal ion-activated) substrate carboxamide linkage, whereas the favored binding of 3 indicates that addition takes place on the si face, which is diametrically For example, it has been suggested that Glu-270-CO; functions as a base in the enzyme-catalyzed, stereospecific enolization-H,Dexchange of ketone-containing inhibitor BMBP (24).
However, a revised crystal structure for the BMBP-carboxypeptidase A complex (25) reveals noncoordination of substrate ketone to metal ion, and that the more strongly basic metal-bound hydroxide is closer to the exchangeable CH, group of BMBP than is Glu-270-COT. The observed pH dependence of enzyme-induced proton exchange within substrate also suggests that (Enz)Zn+OH is the operative base, when the pH profile is properly interpreted (14,16). Finally, our hypothesized process (Zn'+. OH-+ ArCOCD2R % Zn'+. OHD + ArCOCDR-+ Zn". OD-+ ArCOCDHR) constitutes a valid intramolecular mechanism for proton exchange completely within the E . S complex. The original proposal that Glu-270-CO; abstracts the substrate deuterium explains nothing by way of catalysis, since in that case a second, intermolecular exchange step of the carboxyl with external solvent would be required prior to reintroduction of a proton into substrate, which step would not be expected to compete with intramolecular reprotonation of substrate enolate by Glu-270-C02D; i.e. it is only a diprotic acid intermediate such as Zn'+. OHD which allows rapid label exchange entirely within an E . S complex. Conjectured mechanism for peptide hydrolysis by carboxypeptidase A. Optimally oriented substrate carboxylate (held by Arg-145 and Asn-144) provides general base catalysis for metal ion-induced addition of water to si face of scissile carboxyamide linkage, and subsequently donates proton to nitrogen during breakdown of tetrahedral adduct. Role of Glu-270 carboxylate is passive in this scheme; it electrostatically stabilizes adjacent electron-deficient centers (metal ion and carbonyl) in early stages, and in conjunction with Arg-71 and -127 may provide catalytically beneficial electric field gradient across substrate-binding region (polarizing reaction center).
opposed and inaccessible to Glu-270.
How then is a nucleophile delivered to substrate? One proposal finding recent advocacy is a four-center mechanism, in which the substrate carbonyl supposedly enters the coordination sphere of the metal ion while a metal-bound hydroxide is concurrently transposed to the carbonyl (27)(28)(29). This scheme has been advanced by crystallographers on the finding that certain carbonyl hydrate containing inhibitors (e.g. 1) bind to the enzyme with both oxygens more or less equivalently coordinated to the active site metal ion. Such a process is compatible with the synergism test previously alluded to, and may be given stereochemical definition by the sulfoximine-binding results with 3 and 4. However, we have a problem with the chemical plausibility of such a scheme. It seems impracticable to ask a metal ion simultaneously to activate a nucleophilic water molecule (by facilitating deprotonation, 30) while also providing significant Lewis acid catalysis to the substrate carbonyl. Ligand hydroxide should so diminish the electron deficiency of the metal ion as to provide little residual acidity for substrate activation. Nor is it obvious that this proscription might be relaxed by specifying that the ligand hydroxide is only incipient (28). In other words, this proposal fails to explain the central problem of enzyme mechanism, what is the origin or the kinetic acceleration?
q p : -We should like to introduce an alternative idea, which appears not to have previously been advanced. Examination of Fig. 5 reveals that the sulfoximine oxygen in the bound inhibitor (which we contend approximates the position of the introduced nucleophile in the tetrahedral adduct for substrate hydrolysis) resides in close proximity to one of the substrate carboxylate oxygens. We shall explore the proposition that it is the substrate carboxylate that serves to deprotonate a water molecule, promoting nucleophilic addition to the adjacent carboxamide, which is activated by conventional coordination to the metal ion. It is well known that this carboxylate group is an essential specificity feature for recognition of substrates by carboxypeptidase A, and that for productive binding it forms a salt link with the side chain of Arg-145. However, this by no means precludes an active role as general base in the catalytic mechanism as well (precedent 31). We suggest that after 20 years of largely unsupported speculation that the carboxylate of Glu-270 plays such a role, some consideration be given to the possibility that the equally puissant carboxylate of substrate may be the actual proton transfer agent in the hydrolytic cycle. Our mechanistic suggestion is summarized in Fig. 6. We offer the following arguments in favor of this hypothesis: (i) As previously designated, addition to the scissile linkage is from the proper (si) face as indicated by the preferential binding and acidification of 3. (ii) The acidic (Zn") and basic (COT) catalytic groups are separate entities, unlike the four-center mechanism wherein the metal ion must fulfill a dual role. The enzymic catalysis may consequently be partly attributed to the conventional counterentropic effect of gathering and precisely orienting reactive functionality. (iii) In the latter regard, it is worth noting that for the carboxylate-containing side chain of Glu-270, nominally free rotations about no less than three carbon-carbon single bonds would have to be frozen (>CH-CH2-CH2-COT), were it to be brought into a fixed position to serve as a general base proton acceptor as in the conventional mechanism. In contrast, the substrate carboxylate is rigidly oriented. It is of course directly connected to the reaction center via the substrate backbone, with the conformation of the latter locked by virtue of interactions with the enzyme. Of the two oxygen atoms in the substrate carboxylate, the one distal to the reaction center is hydrogen bonded to an NH of the side chain carboxamide of Asn-144; this should tend to freeze rotation about the substrate >CH-COT bond, and to hold the proximal (catalytic) oxyanion in a fixed position.
When the Burgi-Dunitz (32) reaction coordinate for addition to carbonyl groups is projected onto the scissile carbonyl of the substrate (on the si face), using crystallographic coordinates for the emyome-GlyTyr complex (22), the nucleophile passes within 2.7 A (H-bonding distance) of the proximal oxygen atom of the substrate carboxylate precisely at the point where pyramidalization of the carbonyl group commences. Hence, there can be little question that the substrate carboxylate is suitable to serve as a general base assisting HzO addition. In contrast, there are no obvious conformational constraints on the side chain of Glu-270, such as would be needed to orient it rigidly in the manner that seems to be a prerequisite for effective intramolecular general base catalysis (33). However, were the Glu-270 carboxylate to have only a charge-neutralization role, electrostatically satisfying electron deficiency alternately in the metal ion or in the substrate carbonyl group at various stages of the catalytic cycle, this conformational flexibility could be of benefit in allowing optimal interaction guided by electronic attraction. (iv) A potential objection is that the geometry of the substrate carboxylate requires that the purportedly less basic anti lone pair of electrons on the carboxylate oxygen be a proton acceptor (34). However, it is not known how great the antilsyn basicity differential is in aqueous solution, and in any event the conjugate acid subsequently would serve to protonate the nitrogen of the scissile linkage prior to dissociation of the tetrahedral adduct (Fig.  6). This latter step can be rate limiting in model amide hydrolyses (29), and so a stronger acid here might be advantageous overall. (v) There is additional suggestive evidence that the substrate carboxylate plays more than the merely passive role of binding the reactant to the enzyme. A heterocyclic tetrazole 5-membered ring has been shown to be a useful surrogate for the carboxyl group in various biochemical analogues. It has a similar pK,, so that it provides a congeneric anion when incorporated via its carbon atom into a synthetic pseudosubstrate in place of a carboxylate. In many cases the analogue is accepted normally by its biochemical receptor. Such a substrate analogue for carboxypeptidase A has been examined (35). In contrast to a rapidly cleaved carboxylatecontaining model substrate, the corresponding peptide linkage was not hydrolyzed at all within the tetrazolate. This may indicate that the substrate-CO; normally does more than merely provide far recognition by Arg-145. We note that the anionic charge on a tetrazole ring is distributed over four nitrogen atoms rather than the two oxygens of a carboxylate, and judging by the product orientations obtained upon Nalkylation of tetrazole anions, electron density on the 2-and 3-position nitrogens is greater than for the 1-and 4-positions (which would be the obligatory catalytic proton acceptor at the active site) (36). Therefore, it may be that the heterocyclic ring is an incompetent general base for the catalytic mechanism.s (vi) Finally, a potentially serious objection to the proposed essential role of the substrate carboxylate must be met. The metalloenzyme thermolysin is homologous in many respects to carboxypeptidase A, but it is an endopeptidase which does not require a free C-terminal carboxylate in its substrates. Were the mechanisms of these enzymes identical (which ought not to be a foregone conclusion), the substrate carboxylate could not have the pivotal role prescribed for carboxypeptidase A. However, the active site of thermolysin contains an additional histidine residue (His-231), and its imidazole ring appears suitably positioned for proton acceptance on Ne from a nucleophile attacking the appropriate face of a bound substrate. Furthermore, the imidazole is "backedup" by a carboxylate (Asp-226) which is H-bonded to N6 in the fashion of the serine proteases (37). An imidazole moiety is the more suitable general base for an enzyme operating a t neutral pH, as demonstrated by its prevalence in enzymology (38): Our suggestion is that carboxypeptidase A makes do 'While it appears that the tetrazole-containing analogue exhibits a nonproductive binding mode (as does its carboxylate-containing parent substrate), this can have no adverse effect on k,,,/K, (18), which was found to be nil. It might be noted that the dual tight Hbonds between substrate carboxylate and Arg-145 would forestall anti-syn tautomerism of the conjugate acid intermediate in Fig. 6, but the poorer match with Arg-145 in the case of a tetrazole might not prevent N,+N, migration of a proton, retarding breakdown of a tetrahedral adduct.
'Involvement of His-231 in the mechanism of thermolysin is supported by enzyme inactivation with ethoxyformic anhydride and by a bell-shaped pH uersus catalytic rate profile, which fits a reverseprotonation (16) mechanism (acidic limb, M"cOH,, pKa of -5; alkaline limb, imidazolium ion, pK,, of -8.25, the latter correlating with the kinetics of carbethoxylation, 39). Furthermore, this partic-with a carboxylate base only because its proximity within the substrate never allowed an evolutionary replacement. The side chain of Glu-270 has similarly been conserved because it functions as an obligatory anion, but not necessarily as a proton acceptor.
In conclusion, we have attempted to initiate a case for a hitherto ignored mechanism for carboxypeptidase A. Our purpose is not to disenthrone completely Glu-270 from its place of prominence in conventional mechanistic speculations. Rather, we merely observe that another carboxylate is invariably present within the active site when substrates are productively bound. Because kinetic evidence fails to implicate Glu-270 in a prototopic capacity (i.e. the pK, , of -6.1 observed in enzyme kinetics is not attributable to that residue and such a pK, need not represent an enzymic general base in any event), we suggest that an open mind should embrace all possibilities and give consideration equally to proposals involving either carboxylate in a catalytic capacity. Our introduction of the sulfoximine group as an active site stereochemical probe is only the beginnning of a hoped-for discrimination between these alternatives.