Binding of D-phenylalanine and D-tyrosine to carboxypeptidase A.

The structures of the complexes of carboxypeptidase A with the amino acids D-phenylalanine and D-tyrosine are reported as determined by x-ray crystallographic methods to a resolution of 2.0 A. In each individual study one molecule of amino acids binds to the enzyme in the COOH-terminal hydrophobic pocket: the carboxylate of the bound ligand salt links with Arg-145, and the alpha-amino group salt links with Glu-270. The carboxylate of Glu-270 must break its hydrogen bond with the native zinc-bound water molecule in order to exploit the latter interaction. This result is in accord with spectroscopic studies which indicate that the binding of D or L amino acids (or analogues thereof) allows for more facile displacement of the metal-bound water by anions (Bicknell, R., Schaffer, A., Bertini, I., Luchinat, C., Vallee, B. L., and Auld, D. S. (1988) Biochemistry 27, 1050-1057). Additionally, we observe a significant movement of the zinc-bound water molecule (approximately 1 A) upon the binding of D-ligands. We propose that this unanticipated movement also contributes to anion sensitivity. The structural results of the current x-ray study correct predictions made in an early model building study regarding the binding of D-phenylalanine (Lipscomb, W. N., Hartsuck, J. A., Reeke, G. N., Jr., Quiocho, F. A., Bethge, P. H., Ludwig, M. L., Steitz, T. A., Muirhead, H., and Coppola, J. C. (1968) Brookhaven Symp. Biol. 21, 24-90).

Bovine carboxypeptidase A is an exopeptidase of molecular weight 34,472 containing a divalent zinc ion bound to a polypeptide chain of 307 amino acids (Lipscomb, 1982(Lipscomb, , 1983Vallee et al., 1983;Vallee and Galdes, 1984). The complexes of carboxypeptidase A with various ligands have been investigated by x-ray crystallographic methods, and these structures have yielded important answers regarding the catalytic mechanism (Christianson and Lipscomb, l985,1986a, 1986b, 1986c, 1988, 1989Christianson et al., 1985Shoham et al., 1988). Of carboxypeptidase A substrates, D-amino acids do not comprise the biologically preferred COOH-terminal portion ( i e . P: portion; for a description of S,P notation, see Schecter and Berger, 1967). Schechter and Berger (1966) have shown that the enzyme hydrolyzes substrates with Pi D-amino acids only very slowly. Interestingly, D-Phe is a much more potent inhibitor of enzyme activity than is L-Phe: respective Kt values are 2 and 18 mM (Elkins-Kaufman and Neurath, * This work was supported in part by National Institutes of Health Grant GM 06920. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C.
Section 1734 solely to indicate this fact. Q Recipient of a visiting fellowship from the University of Florence.
Present address: Dept. of Chemistry, University of Pennsylvania, Philadelphia, P A 19104-6323. 1948Philadelphia, P A 19104-6323. , 1949Neurath and DeMaria, 1950). Studies of enzymeligand complexes involving these compounds may probe finer structural details of the enzyme-active site and particularly details pertaining to the zinc-bound water molecule. For example, recent spectroscopic studies have shown that the function of this particular water molecule is significantly affected upon the binding of zwitterionic L-or D-aminO acids and small, anionic carboxylate-containing inhibitors (Bicknell et al., 1988). The zinc-bound water molecule of carboxypeptidase A is more easily displaced by anions when the enzyme is first treated with single amino acids or carboxylate inhibitors. The postulated feature of anion sensitivity, the breaking of an enzyme hydrogen bond with the zinc-bound molecule, is revealed in the current x-ray crystallographic study. Additionally, the observed binding mode of D-Phe reported herein is remarkably different from that predicted in an early model building study (Lipscomb et al., 1968).
Important residues for binding and catalysis in the carboxypeptidase A active site include Glu-270, Arg-127, Arg-145, Arg-71, Tyr-248, the zinc ion (liganded by His-69, His-196, and the two carboxylate oxygens of Glu-72), and its bound water molecule. Of these components, the most functionally enigmatic remains the zinc-bound water molecule. X-ray studies of the native enzyme result in electron density, corresponding to this water molecule, which has been described as disordered (Rees et al., 1983) or reasonably ordered (Hardman and Lipscomb, 1984). Furthermore, the pK, of this zinc-bound water molecule continues to be the subject of much study and conjecture. The enzyme displays an acidic pKa at about 6 and a basic pK, at about 9 (Auld and Vallee, 1971;Auld et al., 1986;Hall et al., 1969). It is often though that the lower pK, reflects the ionization of the zinc-bound water during substrate turnover (Wooley, 1975;Markinen et al., 1979). However, the upper pK,, too, has been considered to reflect the ionization of the zinc-bound water molecule (Kaiser and Kaiser, 1972;Suh and Kaiser, 1976;Spratt et al., 1983), and this consideration has, in the free enzyme, some support from xray diffraction studies at basic pH values (Shoham et al., 1984). Studies of a mutant carboxypeptidase A from the rat are also interpretable to this end (Hilvert et al., 1986;Gardell et al., 1987).
A wealth of experimental results support the former assignment, including those results obtained with model complexes (Wooley, 1975;Groves and Olsen, 1985). Most notably, Groves and Olsen (1985) report a model (but non-catalytic) complex with a zinc-bound water of pK, 7.0. The pK, of the zinc-bound water in native carboxypeptidase A may be depressed further by a strong hydrogen bond with the y-carboxylate of residue Glu-270, the proposed intermediate proton donor in proteolysis Monzingo and Matthews, 1984). The strength of the association of this water molecule in carboxypeptidase A with the metal ion is remarkable; it is quite difficult to displace by other ligands such as chloride or 12849 azide unless carboxylate-bearing compounds, such as 3-phenylpropanoic acid or D-or L-phenylalanine, are present (Bick-ne11 et al., 1988;Bertini et al., 1988). Alternatively, chloride ion can displace zinc-bound water from carboxypeptidase A at low pH due to protonation of Glu-270 (Stephens et al., 1974)) but if Glu-270 is chemically modified the enzyme is anion sensitive at low as well as neutral pH values (Geoghegan et al., 1983). The current x-ray study shows that the following two factors are responsible for anion sensitivity of zinc-bound water: the breaking of its hydrogen bond with Glu-270, and the subsequent movement of this water molecule by as much Final data on D-Phe-and D-Tyr-soaked carboxypeptidase A crystals were collected to 2.0 A resolution on a Syntex (Nicolet; Madison, WI) automated four-circle x-ray diffractometer. Integrated intensities, estimated by using the Wyckoff step scan (Wyckoff et al., 1967), were corrected for Lorentz and polarization effects; additionally, a linear correction for crystal decay in the x-ray beam was applied based upon the average decay of four standard check reflections. After scaling and merging, R factors based on intensities were 0.049 and 0.078 for D-Phe and D-Tyr data sets, respectively. Model building was performed on an Evans and Sutherland PS300 interfaced with a VAX 11/780 with graphics software developed by Jones (1982). The enzyme-amino acid models were each refined by the reciprocal least squares method employing the stereochemically restrained least squares algorithm of Hendrickson and Konnert (1981). The crystallographic R factor for the carboxypeptidase A/D-Phe model is 0.168, and that for the carboxypeptidase A/D-Tyr model is 0.177. The highest peaks in final electron density maps, calculated for each model with Fourier coefficients lFol-lFcl and phases derived from each final model. were less than 4a.

RESULTS AND DISCUSSION
It is interesting to note that although the side chains of D-Phe and D-Tyr differ by a phenolic hydroxyl group, the two amino acids binds in identical fashion within the Si hydrophobic pocket. Distances of interactions between carboxypeptidase A and D-amino acids are recorded in Table I. The root mean square (rms) deviation between corresponding atomic coordinates of D-Phe and D-Tyr is 0.2 A; this deviation is the same as that resulting from the experimental rms error in each set of atomic coordinates (about 0.2 A). We propose that hydroxyl or other similarly sized para-substituents on a Pi benzyl side chain will perturb favorable enzyme-ligand association. A stereoview comparing the two binary complexes is presented in Fig. 1. The binding of the two amino acids differs in one respect, however, within the hydrophobic pocket: the phenolic hydroxyl of D-Tyr displaces a water molecule, whereas this water molecule is not displaced by the side chain of D-Phe. The phenolic hydroxyl of D-TY makes hydrogen bond contacts with two additional water molecules in the hydrophobic pocket which are not displaced by the ligand. Figs. 2 and 3, respectively, show difference electron density maps calculated for D-Phe and D-Tyr complexes with carboxypeptidase A. There are no other significant differences beyond the active site region in the structure of each enzymeamino acid complex.
In each complex, the side chain of residue Glu-270 is observed to move away from its position in the native enzyme, primarily by a rotation about the C,-CB bond. By virtue of this rotation, Glu-270 breaks a hydrogen bond with the zincbound water, Distances of relevant interactions in each complex are recorded in Table 11. The movement of Glu-270, probably driven by a stronger, hydrogen-bonded salt link between Glu-270 and the positively charged amino group of the bound D-amino acid, contributes to the anion sensitivity of the active site metal ion in the presence of amino acids (or their analogues) demonstrated in recent spectroscopic studies (Bicknell et aL, 1988). We judge this interaction to be strong and likely to involve an ion pair due to the short distance of the interactio?; the amino-carboxylate distance for D-Phe/ Glu-270 is 2.3 A, and that for ~-Tyr/Glu-270 is 2.4 A (although these distances appear anomalously short, carboxylate oxygens are known to engage in strong hydrogen bonds with nonhydrogen atom separations of 2.4 to 2.5 A; see Jeffrey and Maluszynska, 1982). The proposal of Bicknell and colleagues (1988) is supported in these crystal structures: by breaking the hydrogen bond between Glu-270 and zinc-bound water, the water molecule is more readily displaced by anions. We observe that the water molecule also moves from its position in the native enzyme, although it remains coordinated to zinc (the zinc-water distance in each complex is 2.3 A, compared with 2.1 in the native enzyme). The zincbound water moves 1.0 A in the carboxypeptidase A-D-Phe complex and 0.5 A in the carboxypeptidase A-D-Tyr complex.
We propose that the actual motion of zinc-bound water subsequent to ligand binding also contributes to its anion sensitivity.
The binding of single amino acids will provide a strong, favorable charge-charge interaction with Glu-270. We illustrate the nature of this interaction schematically in Fig. 4. The interaction of Glu-270 with the amino group of a single L-Phe molecule in the Si subsite would be less optimal than that observed for D-Phe. Crystallographic investigation of the binary carboxypeptidase A-L-Phe complex has yet been unsuccessful, so these conclusions regarding L-Phe remain partially speculative until a satisfactory structure determination is made. However, a carboxypeptidase A/L-Phe interaction was observed in a ternary enzyme-substrate-product complex  in which the amino nitrogen of L-Phe was 3.3 & 0.2 A away from the nearest carboxylate oxygen of Glu-270. This distance is longer than that observed for the interaction of D-Phe with Glu-270, and it is thus indicative of a weaker hydrogen bond interaction with L-Phe. Two binding sites for D-Phe (the "metal" and "non-metal" sites) were observed in the spectroscopic studies (Bicknell et al., 1988), but only one is observed occupied in each of the current crystal structures. We conclude that the proteinamino acid interaction we observe is that of the non-metal site, and the metal-carboxylate carbon distayce of 4.8 f 0.2 A is in accord with the distance of 4.2 & 0.4 A observed in a recent spectroscopic study . Since higher concentrations of amino acids caused severe crystal deterioration, we were unable to observe saturation of the metal site in the crystal. This lack of saturation is due solely to ligand concentration and not to crystal lattice obstructions, since the metal site is observed occupied by larger inhibitors spanning several enzyme subsites (Christianson and Lipscomb, 1989).
An interesting difference is noted when comparing the two electron density maps of Figs. 2 and 3. Although the zincbound water is well-ordered and its electron density spherically symmetric in the carboxypeptidase A/D-Phe structure (with thermal B = 13 A'), this water molecule is not similarly well behaved in the carboxypeptidase A-D-Tyr complex. The shape of the electron density above the zinc ion is rather elongated in the latter complex; this feature is reminiscent of the phenomenon described for the zinc-bound water of the native enzyme (Rees et al., 1983). In native carboxypeptidase A, the odd-shaped electron density above the zinc ion was interpeted as a disordered water molecule (with thermal B = 26 A'). Our interpretation is the same for the elongated density found in the carboxypeptidase A-D-Tyr complex (the peak above zinc displays a thermal B = 25 A'). This anomaly for zinc-bound water between the D-Phe and D-Tyr complexes could be artifactual in origin. The poorly shaped electron density may reflect the quality of diffraction data acquired for the D-Tyr complex relative to that acquired for the D-Phe complex. The higher R factor for scaling and merging replicate reflections in the carboxypeptidase A/D-Tyr data set (0.078) relative to that for the carboxypeptidase A/D-Phe data set (0.049), and the slightly higher crystallographic R factor for the carboxypeptidase A-D-Tyr complex (0.177) relative to that for the carboxypeptidase A-D-Phe complex (0.168) indicate to Glu-270, D-Tyr, and the two partially ordered water molecules (one of which is coordinated to zinc) in the region above the active site zinc ion. Positive electron density differences are contoured at a level of 20. The coordinates of the final model are superimposed, and D-Tyr is highlighted. Note the asymmetric and ambiguous nature of the solvent peak above the zinc ion. As in the native carboxypeptidase A structure, this density was fit by two mutually exclusive water molecules (separation = 1.9 A). This anomaly may be chemically relevant, or instead it may be artifactual due $0 the quality of the carboxypeptidase A/D-Tyr x-ray data. The thermal B factor of the zinc-bound water is 25 A'.

Zinc-bound water interactions
An asterisk denotes a possible hydrogen bond as judged from both distance and geometric criteria. The rms error in atomic coordinates is about k0.2 A. CPA. carboxwevtidase A. that the quality of data may affect, among other factors, the confidence of density corresponding to water molecules.
There is generally no great difference in the binding of the D-amino acids of this study (excluding amino groups) within the hydrophobic pocket of carboxypeptidase A when compared with the binding of L-amino acids either contained within longer molecules (e.g. glycyl-L-tyrosine (Christianson and Lipscomb, 1986)) or when bound in ternary complex such as the products of amidolysis (e.g. the enzyme-substrateproduct complex with benzoyl-L-phenylalanine plus L-phenylalanine ). Two essential features govern binding in the Si subsite of the enzyme: the hydrophobic pocket is complementary to a large aromatic side chain, and Arg-145 provides a strong salt link with the terminal carboxylate of substrate, product, or inhibitor. Hence, once anchored in the Si subsite by these two interactions, the configuration of a single amino acid governs the efficiency of an additional salt link between its amino group and Glu-270. Spectroscopic data are interpretable such that the binding of two molecules of L-Phe is less efficient than the binding of two molecules of D-Phe, perhaps due to steric interactions between the L-amino group and a zinc-bound carboxylate of a second molecule (Bicknell et d., 1988). However, such an interaction is not impossible; indeed, it is accommodated quite satisfactorily in a novel enzyme-substrate-product complex .
Interestingly, it has been found that simple carboxylate inhibitors such as P-phenylpropionate, phenylacetate, and acetate also activate the zinc-bound water toward displacement by anions (Bicknell et d., 1988). This activation, however, is not as strong as that measured for the amino acids. Nevertheless, the hypothesis which rationalizes the effects of D-and L-amino acids cannot hold for these simple carboxylate inhibitors which lack an a-amino group. An alternative hypothesis for the simple carboxylate inhibitors is that the aromatic side chain (if any) resides in the enzyme's Si hydrophobic pocket. The conformationally mobile carboxylate of the inhibitor may then engage in a carboxylate-carboxyl hydrogen bond interaction (Sawyer and James, 1982) with Glu- 270. An interaction of this type was observed for Pi product glycine in the complex of carboxypeptidase A with the cleaved potato inhibitor . Since glycine is almost isosteric with acetate, analogous interactions for acetate in the Si subsite would not be unexpected. A carboxylatecarboxyl interaction would disrupt the hydrogen bond between Glu-270 and zinc-bound water, facilitate some movement of this water, and resultantly confer anion sensitivity.

SUMMARY
The binding of D-amino acids to carboxypeptidase A confers anion sensitivity upon the function of the enzyme by breaking the hydrogen bond between the active site base, Glu-270, and the zinc-bound water molecule. An additional factor contributing to anion sensitivity is the substantial movement (-1 A) of the zinc-bound water molecule which accompanies ligand binding. To be sure, the surviving zinc-water interaction may be strong, but spectroscopic studies indicate that it is sufficiently weak to be displaced by anions such as chloride, azide, or the carboxylate of a second amino acid. Based on prior crystallographic experiments, the less potent anion sensitivity conferred by simple carboxylate inhibitors which lack an aamino group arises from a carboxyl-carboxylate interaction with Glu-270. The binding mode for D-Phe does not differ much from that observed for L-Phe as observed in the ternary complex of carboxypeptidase A with benzoyl-L-phenylalanine plus L-phenylalanine . However, the a-amino group of D-Phe makes a stronger salt link with Glu-270 than that observed for L-Phe. Finally, we note that the results of the current study correct the predictions of an early model building study regarding the carboxypeptidase A-D-Phe complex (Lipscomb et d., 1968).