Dynamical Structure of Carboxypeptidase A€¦ · 1. Introduction basis of X-rav studies to identifv how the extent, of The constituent amino acid residues of enzymes motional fl&tuations

J. Mol. Riol. (1989) 207. 201-216

Dynamical Structure of Carboxypeptidase A Marvin W. Makinent, John M. Troyer

Department of Biochemistry and Molecular Biology The University of Chicago, 920 East 58th Street

Chicago, IL 60637, U.S.A.

Harm van der Weti, Herman J. C. Berendsen and Wilfred F. van Gunsteren

Laboratory of Physical Chemistry, University of Groningen Nijenborgh 16, AG 9747 Groningen, The Netherlands

(Received 15 June 1988, and in revised form 30 November 1988)

Structural fluctuations of the apoenzyme form of carboxypeptidase A (EC 3.4.12.2) have been evaluated on the basis of molecular dynamics. The Konnert-Hendrickson refined co- ordinates of 2437 non-hydrogen atoms of the 307 amino acid residues derived from the X-ray structure of the holoenzyme served as the molecular model together with 548 calculated polar hydrogen atoms and 25 buried solvent molecules. Molecular dynamics simulations were carried out at 277 K, and the averaged structural properties of the protein were evaluated for the terminal 20 picosecond portion of a 48 picosecond trajectory. The average atomic displacement from the initial X-ray structure was 2.49 A for all atoms and 1.79 A for C” atoms. The average root-mean-square (r.m.s.) fluctuation of all atoms was 0.67 A as compared to 0.54 A evaluated from the X-ray-defined temperature factors. Corresponding r.m.s. fluctuations for backbone atoms were 0.56 A by molecular dynamics and 0.49 A by X-ray. On the basis of these molecular dynamics studies of the isolated molecule, it is shown that amino acid residues corresponding to intermolecular contact sites of the crystalline enzyme are associated with high amplitude motion. All eight segments of a-helix and eight regions of /?-strand were well preserved except for unwinding of the five C-terminal residues of the a-helix 112-122 that form part of an intermolecular contact in the crystal. Four regions of P-strand and one a-helix with residues adjacent to or in t’he active site constitute a core of constant secondary structure and are shown not to change in relative orientation to each other during the course of the trajectory. The absence of the zinc ion does not markedly influence the stereochemical relationships of active site residues in the dynamically averaged protein. The extent of motional fluctuations of each of the subsites of substrate recognition in the active site has been evaluated. Active site residues responsible for specificity of substrate binding or splitting of the scissile bond exhibit low simulated motion. In contrast, residues in more distal sites of substrate recognition exhibit markedly greater motional fluctuations. This differential extent of dynamical motion is related to structural requirements of substrate hydrolysis.

1. Introduction basis of X-rav studies to identifv how the extent, of

The constituent amino acid residues of enzymes motional fl&tuations in different structural

and proteins undergo motional fluctuations because domains of proteins may be related to function

of thermal energy. The amplitudes of these (Bennett & Huber, 1984). It has been suggested

fluctuations are routinely estimated through refine- that antigenic recognition of proteins is dependent

ment of X-ray diffraction data. However, until on structural mobility and flexibility (Westhof et

recently relatively little effort has been made on the al., 1984; Tainer et al., 1984). However, these analyses are restricted to the temperature factors of proteins in crystals and assume that they remain

t Aut,hor to whom all correspondence should be sent. representative of the molecule in solution. In 201

002-2836/89/090201 16 $03.00/O 0 1989 Academic Press Limited

202 M. W. Makinen et al.

crystals of proteins, the intermolecular contact regions can exhibit by necessity fluctuations of only low amplitude, and it has not been demonstrated by physical methods whether these regions exhibit motional fluctuations of comparably low amplitude in solution. A number of molecular dynamics investigations have been carried out to refine force field potentials, in part for the purpose of reproducing crystallographically refined protein structures (Levitt, 1983a,b; Whitlow & Teeter, 1986) even though the simulations are of the isolated protein in vacua. Such an approach pre- vents any inquiry into the question of whether the nature and amplitudes of the structural fluctuations may differ substantively in crystalline and solution environments. In this respect it is of some importance to point out that the results of Raman spectroscopic studies indicate that the low frequency motions of proteins in crystals and in solution are not totally identical (Genzel et al., 1976).

Here we present results of molecular dynamics calculations of carboxypeptidase A (EC 3.4.12.2), carried out to investigate the nature and extent of structural fluctuations in the protein. The crystallographically refined structure of this enzyme is accurately defined and a variety of inhibitor complexes have been characterized through difference Fourier syntheses (Lipscomb et al., 1968; Rees et al., 1980, 1981, 1983; Rees & Lipscomb, 1982; Christianson et al., 1987). Furthermore, the mechanism of catalytic action of this enzyme is well established (Makinen et al., 1984; Sander & Witsel, 1985, 1986). This enzyme with 307 amino acid residues constitutes the largest protein hitherto investigated by molecular dynamics methods. In comparison to studies carried out of smaller proteins such as bovine trypsin inhibitor (van Gunsteren et al., 1983; Levitt, 1983a,b), cytochrome c (Northrup et al., 1980, 1981), lysozyme (Post et al., 1986), and myoglobin (Levy et al., 1985), there is consequently a larger proportion of the amino acid residues in the protein interior inaccessible to direct contact with solvent molecules. Since molecular dynamics calculations of macromolecules are generally carried out for the molecule in vacua because of computational limitations, our results, reflecting the more extensive internal structure of the protein, should have proportionately lower errors in assessing conformational fluctuations. These factors, therefore, ensure a firm base on which to address critical questions concerning the nature and extent of structural fluctuations in the protein.

A particularly important objective in our molecular dynamics investigations has been to identify the relationships of protein structural fluctuations in the active site to catalytic action. Substrate hydrolysis catalyzed by carboxypeptidase A is a multi-point co-operative process that involves three separate regions of the enzyme (Makinen et al., 1984). These regions are schematically illustrated in Figure 1 and include: (1) the hydrophobic pocket

Figure 1. Schematic illustration of the binding interactions of an oligopeptide substrate in the active site of CPA. The active site comprises 5 subsites (S, to S4 and S,‘) located on both sides of the site of bond cleavage. Each subsite accommodates one amino acid residue of the substrate, and the positions P of the substrate are counted from the site of bond cleavage designated by the arrow. This diagram is taken from Kuo et al. (1983).

that binds the side-chain of the C-terminal residue together with Arg145 responsible for binding the C-terminal carboxylate group (S,‘); (2) the bond cleavage site containing the active site metal ion, its co-ordinating ligands, and the side-chain of Glu270 (S, ); and (3) amino acid residues known as the sites of secondary substrate recognition (S,-S,). Kinetic studies with synthetic oligopeptide substrates have shown that the interaction of the S,’ site with the C-terminal residue determines the value of K, and is dependent on the structure of the C-terminal side-chain. On the other hand, the steric interactions of the substrate with the S2 site determine the value of k,,, (Schechter, 1970). The relationships of the steric interactions in the S, site to the process of bond cleavage in the S1 site have not been elucidated. On the basis of magnetic resonance and cryoenzymologic studies, it has been suggested that these interactions are important to the catalytic event by inducing torsional alterations in the substrate during binding that “prepare” the tetrahedral intermediate for bond cleavage according to stereoelectronic requirements (Kuo et aZ., 1983; Makinen et al., 1984). On this basis, the requirements for specificity in substrate binding imply structural rigidity in the S1’ site while the torsional alterations imply flexibility in the S2 and S, sites. We have consequently evaluated the motional properties of the amino acid residues that constitute the various sites of substrate recognition. We find that there is a gradient in the extent of dynamical motion. Residues in subsites S,’ and S, exhibit low simulated motion while residues in the SZ and S3 subsites exhibit markedly greater motion. We relate the differential extent of dynamical fluctuations in carboxypeptidase A to stereochemical requirements of enzyme-substrate inter-

Dynamics of Carboxypeptidase A 203

actions for catalysis and indicate that comparable relationships probably obtain in other enzymes.

2. Methods

(a) The protein model

The co-ordinates of CPA? refined by the Konnert- Hendrickson method (Konnert, 1976; Hendrickson & Konnert, 1980) at 1.54 w resolution (1 a = 0.1 nm) to a crystallographic R-value of 0.190 (Rees et aE., 1983) served as the initial molecular model for MD calculations. This comprises 2437 non-hydrogen atoms for the 307 amino acid residues of the protein. Twenty-five solvent molecules buried within the protein were also included for which localized binding sites are well defined by X-ray studies. Since MD calculations were carried out only for simulation of the motion of the protein residues and solvent molecules in this investigation, the active site zinc ion was not included in the atomic listing (see below). For energy minimization and MD calculations, polar hydrogen atoms that may form hydrogen bonds were explicitly treated. The co-ordinates of these 548 hydrogen atoms incorporated into protein residues were calculated according to idealized valence angle and bond distance parameters.

In CPA there are 8 histidine residues. The assignment of the protot,ropir tautomer of each imidazole side-chain was made on the basis of the X-ray description of the protein (Quiocho & Lipscomb, 1971; Rees et al., 1983) and calculated nearest-neighbor interactions. The NE’-H tautomeric conformer is marginally favored energetically over the Ndl-H tautomer (Boschcov et al., 1983). All histidine residues proved to be best designated by a NC’-H conformation of the imidazole side-chain except for His166 and Hisl3. The former is best assigned to a Ndl-H conformation. Since inspection of the environment of the imidazole side-chain of His13 showed that it can simultaneously accommodate hydrogen bonds with the Oe2 of Glul7 and the carbonyl 0 of Asn8, the side-chain of His13 was consequently assigned as a protonated imidazolium group.

(b) Molecular dynamics calculations

The programs for carrying out MD calculations and analyzing results were taken from the Groningen Molecular Simulation System (GROMOS) library (van Gunsteren & Berendsen, 1987). The basic elements and philosophy underlying the use of this program package have been outlined (Berendsen & van Gunsteren, 1983; van Gunstcren et al., 1983; Aqvist et al., 1985; Kriiger et al., 1985), and parameters for solvent molecules have been described (Berendsen et al., 1981; Hermans et al., 1984). For the MD simulations carried out in this investigation. charged atom groups were neutralized to compensate for the missing shielding influence of bulk solvent. The partial charges associated with carboxylate, and protonated arginine, imidazole and lysine side-chains were those listed in the 37D4 set of the GROMOS manual (van Gunsteren & Berendsen. 1987). For MD calculations,

t Abbreviations used: CPA, carboxypeptidase A (peptidyl-L-amino acid hydrolase (EC 3.4.12.2)); correspondingly, ZnCPA is used to designate the native holoenzyme of the a-form as characterized by X-ray diffraction studies (Rees et aE., 1983); MD, molecular dynamics; r.m.s., root-mean-square.

the potential energy function describing the interactions between protein atoms was evaluated over the 2985 atoms in the apoenzyme for 3052 bonds, 4375 angles, 1567 improper dihedrals and 1626 dihedrals.

The MD simulations were carried out at constant temperature whereby the system is weakly coupled to a thermal bath by resealing all atomic velocities at each MD step (Berendsen et al., 1984). The strength of t,he coupling to the heat bath is determined by the temperature relaxation time T. Initially this constant was set to 0.01 ps for both solvent molecules and the protein atoms with subsequent resetting of r to 0,05 ps after temperature equilibration was achieved. Since the X-ray data for ZnCPA were collected on crystals at 4 “C (Rres et al., 1983), the MD simulations of the protein were correspondingly performed at 277 K. Bond lengths were constrained in the MD calculations by the SHAKE method (van Gunsteren & Berendsen. 1977: Ryckarrt, pt al., 1977).

The X-ray-defined structure without the zinc ion but with 548 explicitly treated hydrogen at’oms and 25 solvent molecules was first subjected to energy minimization by the method of steepest, descents (van Gunsteren & Karplus, 1980). After energy minimizat,ion the deviation of non-hydrogen atoms from their initial X-ray-defined positions, averaged over all protein residues, was 0.32 A, and there was no significant change in the conformation of protein residues. Tnitial velocities were derived from a Maxwellian distribution at 277 K. Non-bonded interactions were evaluated wit,hin a cut-off radius of 8.0 A with use of a pair list that was updated every 10 MD steps. For the protein the list) of non-bonded charged group pairs then totals approximat,ely 68,300, greatly reduced from the theoretical limit of I .779,441 for 2985 protein atoms and 25 water molecules. In general, a time step of 0.002 ps was used; occasionally this time step was too large for application of SHAKE and it was reduced to 0.001 ps. Almost invariably this problem arose wit,h respect to co-ordinate resetting by t,he SHAKE method for the hydrogen atoms of the guanidinium groups of arginine side-chains, due to thr circumstance that the SHAKE method is not well suited fi)r constraining large planar groups of atoms. Resetting of the time step in MD calculations in this manner results in somewhat incorrect initial velocities for the subsequent. run of MD calculations. However. the error was found to be no more than 5% of the pot,ential energy and it vanished within the first 5 to 10 MD steps subsequently calculated. Upon equilibration of the protein struct)urr, this problem did not arise further. Initial MD calculations were carried out with a VAX 11/750 computer; the final trajectory was calculated over a, 20 ps period on a, CYBER 205 supercomputer at the Academic Computation Center of Amsterda.m. During the MD simulations no special functions were applied t,o restrain the positions of solvent molecules or to maintain hydrogen bonds. For the force field potentials employed in GROMOS, such special constraints are not necessary, in contrast to MD simulations carried out h.y others (Levitt, 1983a,b; Post et al., 1986).

Preliminary MD calculations were carried out including the active site zinc ion in the atomic listing with use of interaction potentials in the GROMOS library for discrete charged amino acid residues (37C4 parameter set). Bulk solvent was excluded in MD simulations for computational reasons. The structure did not remain stable and the dipositive zinc ion migrated to occupy a position between the carboxylate side-chains of Glu72 and Glu270. In addition, discretely charged residues on the surfacr of


the molecule were found in distorted stereochemical relationships disrupting hydrogen bonding interactions in helices and causing global deviations from the structure of the protein as defined in X-ray studies. This illustrates that the use of fully charged amino acid residues (37C4 parameter set of GROMOS) requires the presence of bulk solvent with its dielectic screening influence. For these reasons, we have chosen to evaluate first the dynamical properties of the protein without the metal ion, using the interaction potentials for neutralized residues (37D4 parameter set of GROMOS) as described above.

(c) Molecular graphics

Analysis of the protein conformations generated through MD simulations was carried out with use of the programs FRODO (Jones, 1982, 1985) and HYDRA (R. E. Hubbard, personal communication) with an Evans

and Sutherland PS330 molecular graphics system inter- faced to a host VAX 111759 computer.

3. Results

(a) Time course

In Figure 2 we have compared parameters that monitor the approach of the protein to an equilibrium conformation. Over the time-course of the calculations, the radius of gyration monitoring the change in packing density of the protein residues has reached a stable plateau, showing no more than an approximate 1.4% contraction, in comparison to an average contraction of 2.9% reported for three different forms of lysozyme (Post et al., 1986), an enzyme of only 129 amino acid

3.0 - 0a b

, I

20 30 40 50

Time (ps)

Figure 2. Time-dependent evolution of parameters monitoring simulation of dynamical motion. (a) The change in the radius of gyration in Angstrom units; (b) the average deviation of all non-hydrogen atoms (a) and of all C” (A) atoms from their initial X-ray-defined positions; (c) the total potential energy of the configuration of the protein with 25 buried water molecules. The abscissa represents the time in picoseconds.


residues. In Figure 2 we have also compared the time-dependent averaged displacement of C” atoms and of all protein atoms from the initial X-ray structure as well as the change in potential energy of the configuration of the protein and 25 solvent molecules. Figure 2 shows that each parameter has reached its plateau value after approximately 25 to 30 picoseconds of MD steps. Therefore, the averaged structure of the protein with 25 buried solvent molecules was calculated from MD simulations over the 28 to 48 picosecond portion of t.he trajectory from co-ordinates stored at 0.05 picosecond intervals. The averaged structure and r.m.s. fluctuations calculated for the terminal 20 picosecond portion of the trajectory did not differ significantly from those calculated separately for each of four five-picosecond segments of this 20 picosecond portion, indicating that convergence of the trajectory had been achieved.

(b) Comparison of the time-averaged protein structure to X-ray results

(i) Atomic positions and jkctuations In Figure 3 we have compared the r.m.s. isotropic

fluctuations of polypeptide backbone atoms averaged per amino acid residue with corresponding values calculated from the X-ray-derived isotropic temperature factors of ZnCPA. In Figure 3(b) we plotted the displacement of C” atoms of the MD- averaged protein from their corresponding positions in ZnCPA. The average r.m.s. fluctuation of all atoms in the MD protein and in ZnCPA is 0.67 and 0.54 A, respectively. Corresponding values for the backbone atoms are 0.56 and 0.49 d. The average displacement of all atoms of the protein from the X-ray structure was 2.49 8, while the average displacement of C” atoms was 1.79 A. MD simulations of motion in other, comparably large proteins have not been carried out. However, the r.m.s. fluctuation of backbone atoms of sperm whale myoglobin averaged over the 100 to 300 picosecond interval of MD simulations is 0.78 A compared to that of 0.65 A derived from X-ray data (Levy et al., 1985). The average displacements for all atoms and C” atoms, respectively, calculated from MD data are l-82 and 1.07 A for the C-terminal fragment of the L7/L12 ribosomal protein having 68 amino acid residues (Aqvist et al., 1985), 2.2 and I.5 A for hen egg white lysozyme with 129 residues (Post et al., 1986), and 1.19 and 0.82 A for bovine trypsin inhibitor protein with 58 amino acid residues (van Gunsteren et al., 1983). The average r.m.s. fluctuation of the 25 buried water molecules, identified by Rees et aZ. (1983) and included in the co-ordinate listing, was 0.90 A compared to that of 0.72 a in ZnCPA. The average deviation of the water molecules from their initial X-ray-defined positions was 2.2 A. This is less than the average displacement of protein atoms. Essentially the same polar environment was preserved in the MD calculations for these 25 buried water molecules, as originally defined in X-ray

studies (Rees et al., 1983), without the need to apply special constraints of the type described by Levitt (1983a,b) or Post et al. (1986). We shall report the results for the 25 buried water molecules in greater detail elsewhere to compare results of inhibitor and substrate complexes.

In Figure 3(a) there is good agreement in the pattern and distribution of mobilities of residues of the MD protein with that derived from the temperature factors of ZnCPA. In particular, the high mobility of residues 52 to 60, 85 to 105, and 130 to 140, as observed through X-ray studies. is accurately simulated in MD calculations. Notable differences are seen in the N-terminal region where the simulated motion is greater and in the C-terminal region where motion according to X-ray results is increased. There is also increased simulated motion of residues 67 to 74. since this region contains residues His69 and Glu72, two of the three protein ligands to the metal ion, the increased motion is clearly due to the absence of the zinc ion in the MD calculations. All other regions in which there is greatly increased simulated motion compared to ZnCPA contain residues that are associated with intermolecular contacts in the P2, crystal of the enzyme (Rees et al., 1983). These differences are not unexpected since the corresponding steric restraints to motion are absent for MD simulations of the protein in vacua.

In Figure 3(b) several regions of the MD protein exhibit pronounced displacements from their corresponding positions in ZnCPA. It is of interest to note that of these regions residues 28 to 33, 129 to 137, and 149 to 154 are not associated with an increase in simulated motion compared to that observed in the crystal. On the other hand, residues 120 to 126 exhibit a large displacement in averaged position from that in ZnCPA and are associated with pronounced fluctuations. They also have steric contacts with residues near Phe279. as we shall describe later. Within the segment 270-281, only Gly275 undergoes a large displacement’ in averaged position from that in ZnCPA. The pronounced mobility of this latter region, therefore, must derive from its non-bonded contacts with the segment of highly mobile residues 120 to 126. It is of interest to note that the region 270 to 281 lines the active site and contains the catalytically required residue Glu270 and that residues Arg124, Arg127, and Phe279 form important’ interactions with extended oligopeptide substrates as sites of secondary substrate recognition (Quiocho & Lipscomb, 197 1; Kuo et al., 1983; Makinen et al., 1984).

(ii) Secondary structure A stereo view of the spatial distribution of the C”

atoms of ZnCPA and the MD-averaged protein is shown in Figure 4. In general, the secondary and tertiary structure of the MD-averaged protein remains similar to that of ZnCPA. There are localized regions in which the dihedral angles of the backbone atoms have changed markedly from their X-ray-defined values. In addition t,o the N and


6.0

(b)

I MD-averaged apacarboaypeptidoee A

11 28-48 ps

100 I50 200 250 300

Amino acid sequence number

Figure 3. Comparison of r.m.s. isotropic fluctuations and displacements of amino acid residues in ZnCPA and in the MD protein. (a) The r.m.s. isotropic fluctuations of backbone atoms (Cm, C, 0, N) averaged per amino acid residue calculated on the basis of X-ray refined temperature factors (Rees et al., 1983) and MD steps over the terminal 20 ps portion of a 48 ps trajectory. The X-ray-determined values for ZnCPA are represented in stick diagram form; the MD- based values are shown in histogram form. (b) The displacements of corresponding C” atoms in the MD-averaged structure from their initial X-ray-defined positions. In (a) the horizontal bars identify residues associated with the intermolecular contact sites of the crystalline enzyme.

C-terminal regions, the regions showing significant deviations are localized to residues 38-39, 41-45, 55-57, 119-140, 144, 151, 155, 156, 161, 162, 187-189, 198-199, 210-213, 252-254, 273-276 and 280. It is seen from Figure 3 that these regions also exhibit large motional fluctuations. Although there were large differences of the I$J and $ angles in these regions of ZnCPA and the MD-averaged protein, a pattern of positive and negative A4 and AJI values for the same or adjacent residues was observed.

Such correlated changes have been observed in MD calculations and have been shown to preserve the general direction of the backbone and the approximate positions of the side-chains (McCammon et al., 1977).

Rees et al. (1983) have shown that there are eight segments of a-helix and eight regions of b-strand in ZnCPA. In Table 1 we have listed the residue ranges for these 16 regions of secondary structure for which characteristic hydrogen bonds are found


Figure 4. Stereo view of the c” ribbon of the MD-averaged structure of CPA compared to that of X-ray-defined ZnCPA. The MD structure is represented by spheres and thick connector lines; the corresponding structure of Z&PA is represented by the narrow line trace.

by calculation in the MD-averaged protein that significant change was the loss of the hydrogen satisfy geometrical constraints of a donor-receptor bonds120HisN...116PheO ,..., 122GluN...118 (D. . .A) distance of I 3.40 b and a (D-H. . .A) Phe 0 formed with residues 116 to 122, resulting in angle 2 135”. Of the eight a-helices, the only a partial unfolding of the C-terminal portion of the

Table 1 Comparison of structural parameters for a-helical and P-strand regions of

carboxypeptidase A

Residue Ranget

MD ZnCPA

$1 (deg.)

Radius§ Rise/C,JI (4 (4

MD ZnCPA MD ZnCPA

A. a- He&es aHI ah a% a%

14-26 14-28 73-88 73-90 93-101 93-101

(112-122) 112-122 112-118 112-118 174-187 173-187 215231 215231 254-261 2.5-261 285-306 (285306) 285-294 285-294 292-299 292-299 297-306 297-304

32-39 32-39 47-53 47-52 6f-66 6966

104-108 104-108 191-196 192-196 200-204 201-204 239-241 239-241 265-27 1 267-271

5.3 2.25 2.26 1.54 1.54 2.9 2.25 2.23 1.56 1.53

11.0 2.31 2.32 1.47 1.46 35.5 2.32 2.29 1.73 1.69 11.3 2.04 2.04 1.74 1.52 3.4 2.31 2.30 1.47 1.48 6.3 2.24 2.22 1.46 1.49 2.6 2.21 2.20 1.54 1.55 2.4 2.33 2.28 1.48 I.43 5.4 2.26 2.32 1.52 1.54 3.6 2.18 2.16 1.53 1.61

16.6 2.24 2.29 1.56 1.56 ave.4 2.23f0.03 2.23f0.08 1.54+0.08 1.55+0.06

4.6 1.32 1.30 2.94 3.14 0.9 0.93 0.85 3.25 3.30 2.7 0.91 0.89 3.22 3.38 5.1 0.93 0.85 3.21 3.32 1.5 0.92 0.88 3.22 3.30 2.8 0.88 0.75 3.16 3.35 5.8 0.63 0.63 3.50 3.51 6.1 1.10 1.07 3.08 3.13

ave.8 0.95kO.18 090+0.19 3.2OkO.15 3.3OkO.12

t The residue range identified in X-ray studies for ZnCPA (Rees et al., 1983) and by calculation in the MD-protein. The helical parameters calculated for each segment in ZnCPA correspond to the residue range listed for the MD-protein in each case.

$ The angle between the helix axis in the MD-protein and that in ZnCPA when the MD-averaged protein is superpositioned onto ZnCPA by least-squares fitting of all corresponding @ atoms. The helix axis was calculated in each case as a line that is least-squares fitted to the positions of the designated (‘” atoms.

Fj Calculated for each region of secondary structure on the basis of the c” atoms. )I Averaged results are calculated including only residues 112 to 118 for the 4th segment of a-helix

and on the basis of 3 separate segments 285 to 294,292 to 299 and 297 to 306 for region 285 to 306. as discussed in the text.


a-helix 112-122. This change is seen in Figure 4. Structural alterations in this region of the protein are discussed in more detail below. Also, in Figure 4, the helical segment 285-306 in the MD- averaged protein is considerably more straight than in ZnCPA. In ZnCPA hydrogen bonds are not found by calculation for 295 Leu N. . .291 Glu 0, 300 Ile N. . ,296 Gly 0, and 306 Asn N. . .302 Glu 0 within the geometrical constraints employed although these hydrogen bonding interactions are maintained in the MD-averaged protein. These differences are accordingly enumerated in Table 1. On the other hand, all of the hydrogen bonding interactions defining regions of P-sheet remained intact in the MD-averaged protein.

In Table 1 we have compared the helical parameters that characterize each region of a-helix or b-strand. Comparison of corresponding sets of helical parameters of secondary structure shows that the radius and rise per residue of each region in the MD-averaged protein are essentially equivalent to those in ZnCPA. In addition, the average radius and rise per residue for the a-helical segments in the MD-averaged structure are close to those of 2.3 A and 1.5 A expected for an idealized a-helix, and the averaged values for /?-strand are very close to their expected values of 3.2 or 3.4 A for the radius and 1 .l and O-9 A for the rise per residue of parallel and antiparallel /?-sheet structures (cf.), Schulz & Schirmer, 1979). On this basis, the MD-averaged protein compares as well as the refined X-ray structure with respect to the values of these helical parameters for idealized structures.

In ZnCPA, the peptide bonds between residues Ser197-Tyr198, Pro205Tyr206, and Arg272- Asp273 are found in a cis conformation (Rees et al., 1981, 1983). In the MD-averaged protein, the cis conformation of residues 205-206 and 272-273 is preserved and differences in these regions involve only small shifts of both backbone and side-chain at’oms. The cis conformation of residues 197-198, however, is lost in the MD-averaged protein, and the peptide bond has assumed a tram structure. Tn the MD-averaged protein, the hydrogen bond between 197 Ser N and Glu 0”’ is preserved as in ZnCPA. Residues 197 and 198 exhibit considerable motion in comparison to residues 205-206 and 272-273. Inspection of the MD data showed that the loss of the cis conformation for residues 197-198 occurred within the first nine picoseconds of MD simulations. Inspection of our preliminary MD simulations of the Gly-Tyr inhibitor complex of apoCPA shows that the cis conformation of residues 197-198 is also lost. In view of the cis conformation preserved in the course of MD simulations for residues 205-206 and 272-273, it is possible that the cis conformation of residues 197-198 near the active site is intrinsically less stable.

In order to determine whether the relative orientations of the individual regions of a-helix and P-strand differ in the MD-averaged protein from that in ZnCPA, we compared the eight regions of cl-helix and of /?-strand listed in Table 1 in ZnCPA

to their counterparts in the MD-averaged protein by applying the analytical approach of Lesk & Chothia (1984). This analysis entails calculating the r.m.s. differences of the co-ordinates of backbone (Ca, C, 0, N) and C@ atoms in ZnCPA and in the MD-averaged protein for the 16 regions of secondary structure after least-squares superpositioning each individual homologous region in the MD protein onto its counterpart in ZnCPA. The common amino acid residues listed in Table 1 for each segment were employed as the basis of comparison. In addition to comparing the a-helix 285-306 in three segments as noted in Table 1, we employed only residues 112 to 118 of the fourth u-helix since the terminal residues of this a-helix have lost their helical conformation in the course of dynamics. The results are given in Table 2. For the 16 regions of secondary structure, equivalent sets of atoms from the MD protein and ZnCPA fitted onto each other with r.m.s. differences in atomic positions of 0.19 to 0.61 A, except for the p-strand segment PS,, which had an r.m.s. difference of 0.98 A, and these differences comprising the diagonal of Table 2 are comparable to the averaged r.m.s. fluctuation of backbone atoms. In contrast, when each of the regions of secondary structure is superpositioned onto itself and the differences between equivalent atoms in the other structural segments are calculated, the r.m.s. displacements are usually greater than the level attributable to noise in the co-ordinates. These results imply some structural rearrangement induced through dynamical motion. Therefore, to identify the regions of secondary structure that have remained the least changed in relative position and orientation to each other in the MD-averaged structure, we permuted the rows and columns to isolate an array that has the smallest Aij elements. For this purpose we applied a value of Aij < 1.6 A as an upper limit. This is just less than 2.5 times the value of the averaged r.m.s. fluctuation of atoms.

The results in Table 2 identify a core of four regions of P-strand and one a-helix that remain unchanged in position with respect to each other in both the MD-averaged protein and in ZnCPA. In Figure 5 we have illustrated the location of the core structure within the protein. We have also shown schematically the structural relationships that each of these segments of secondary structure have to each other within the core. These regions of secondary structure exhibit changes in relative orientation and position that are indistinguishable from the noise in the co-ordinates of their backbone atoms. Although the individual regions of u-helix and b-strand exhibit small changes in average position with respect to each other within the core structure, the changes do not indicate any sub- stantial cumulative shift. For example, superpositioning the helix aH, onto itself yields an r.m.s. difference in the backbone co-ordinates for the b-strands PS, and /?S, within the core of I.09 A and 0.87 A, respectively, while superpositioning flS, onto itself yields r.m.s. differences of 1.59 A and

Dynamics of Carboxypeptidaee A 209

Table 2 Fit of a-helices and /&strands of carboxypeptidase A

Aij(4S ~__--

Ati( & jX3, /X3, /X3, aH, aH,, aH,, aH,, /EC+ /?S, /?S, j3S, aH, aH, aH, aH, aH, aH,

[; iii!=] 0.68 0.49 0.48 0.77 0.58 130 044 1.59 1.32 1.06 0.69 1.21 1.11 1.22 1.63 2.11 0.99 1.20 1.35 1.21 0.93 1.13 1.31 1.49 1.73 1.43 1.84 1.57 1.28 1.45 0.79 2.56 2.39 1.37 1.85 0.67 2.31 1.61 1.60 1.62 0.91 2.47 1.01 1.75 1.76 0.97 0.59 2.00 1.74 1.01 2.36 1.29 1.16 1.26 1.24 1.15 1.62 I.16 1.03 1.66 2.16 2.64 2.26 1.42 1.65 3.78 3.15 2.45 2.66 1.77 0.89 0.75 2.21 2.10 1.76

ah, 0.32 3.63 3.33 3.12 2.66 1.48 0.74 2.52 4.30 3.44 3.02 2.45 3.50 2.31 2.36 3.96 1.88 4.10 ah, 0.46 3.78 3.28 3.13 2.93 1.99 0.88 1.82 4.69 3.28 2.63 2.73 3.76 2.68 3.23 3.52 1.24 4.03 ;::c 0.42 0.98 2.55 1.81 3.67 1.61 2.76 1.85 2.40 5.04 3.89 3.38 4.36 5.97 5.87 1.74 5.65 3.65 4.34 6.94 4.55 8.44 4.11 6.75 5.22 3GO 2.77 4.31 4.56 3.45 4.19 2.89 4.40 X.58 ;i.l(j 3.0#

/?S, 0.56 4.07 3.11 3.65 2.18 2.94 2.09 2.53 2.89 4.11 0.90 1.44 3.67 3.27 4.71 5.37 1.59 1.93

;:: 0.35 0.19 0.96 3.45 3.22 1.42 3.63 1.28 0.82 1.68 4.25 1.75 3.54 1.51 2.22 3.37 2.44 3.84 4.17 1.33 1.27 1.10 1.16 1.32 4.15 1.33 5.74 2.45 1.03 1.87 4.36 2.99 2.42 3.25 2.62 1.07 aHl 0.33 0.73 1.34 0.86 1.73 0.93 1.82 1.57 2.07 1.69 3.87 3.37 2.89 1.55 1.97 2.38 3.69 3.24 aH3 0.30 1.34 I.25 090 2.49 2.42 4.16 2.59 1.99 2.31 4.69 5.20 4.32 3.32 2.16 3.07 5.33 3.40 a% 0.61 1.53 1.99 1.99 1.63 2.77 2.22 3.26 3.29 2.18 1.35 2.33 1.03 2.36 3.87 2.48 3.54 1.75 aHs 0.56 1.52 1.62 1.44 2.34 3.88 4.28 4.12 2.95 2.63 3.23 4.33 2.56 3.51 3.59 2.27 3.88 2.63 a& 0.50 7.71 5.01 6.28 3.73 6.65 4.63 3.80 4.00 8.22 1.82 2.24 2.65 9.32 6.28 9.78 7.26 2.47 aH7 0.49 3.46 2.58 3.16 1.68 2.87 2.68 2.42 2.80 3.36 1.31 1.01 1.46 3.59 3,23 3.94 3.48 1.43

t r.m.s. differences (in A) in the co-ordinates of the main-chain and C@ atoms of the i a-helices and /?-strands upon least-squares superpositioning each segment in the MD-protein onto the corresponding segment in Z&PA.

$ r.m.8. differences in the atomic co-ordinates of the main-chain and C@ atoms of the j’* element of secondary structure after the molecules have been superpositioned by fitting the irh element (see the text).

5 The amino acid residues for each region of secondary structure are defined in Table 1.

n

Figure 5. Stereo view of the tertiary structure of the MD-averaged protein with schematical illustration of the segments of a-helix and /?-strand as cylinders and ribbons, respectively. The segments that constitute the constant core are lightly stippled. The position of the zinc ion in ZnCPA is indicated by the sphere to identify the active site of the enzyme. The projection is obtained by rotation of that in Fig. 4 by about 30” about the vertical axis and by 15” about the horizontal axis to illustrate more clearly the constant core of secondary structure. The lower diagram illustrates the relationships of the individual segments of secondary structure to each other that constitute the constant core. The crH, helix lies diagonally across the hydrogen-bonded strands of /?-sheet and forms non-bonded contacts with residues in each /?-strand.


0.58 A for the helix aH, and the P-strand PS,. Fitting the structures by superpositioning the P-sheet PS, outside of the core onto itself yields, on the other hand, values for Aij of 3.67, 5.04 and 3.38 A for /?S,, PS, and aH,, which are much higher than those for regions belonging to the constant core. The averaged r.m.s. fluctuation of the backbone atoms of these residues, furthermore, is 0*41( kO.05) A, less than that of 0*47( kO.08) A averaged over the residues of the remaining 11 segments of secondary structure. The average r.m.s. fluctuation of the core backbone atoms is also less than that of 0.56 A for all backbone atoms.

It is of interest to note in Figure 5 that the regions of a-helix and P-strand constituting the constant core are not only contiguous with each other but also are in the direct vicinity of residues important in catalysis and the specificity of substrate binding. The P-strand region /?S, contains the amino acid ligand to the metal ion His196, while the P-strand PS, and the a-helix aH, containing the residue Trp73 are adjacent to His69 and Glu72 that serve as the two other protein ligands to the metal ion. Furthermore, the p-sheet PC3 containing the catalytically required residue Glu270 exhibits relatively small values of Aij with respect to aH, and PS, that are symmetrical across the diagonal, indicating that its position remains relatively fixed with respect to these two regions of secondary structure.

(iii) Active site structure Figure 6 compares the conformations of active

site residues in the MD-averaged protein and in ZnCPA. There are, in general, only small shifts in the positions of backbone atoms. The averaged positions of the side-chains of His69, Glu72 and His196 that serve as donor ligands to the metal ion are nearly identical to those in ZnCPA despite their greater mobility and lack of contact with the zinc ion in MD simulations. Analysis of the calculated hydrogen bonding relationships showed that the hydrogen bond of the imidazole group of His196 with a solvent molecule remained intact, while the hydrogen bond of the imidazole side-chain of His69 with 142 Asp 06’ was broken. In Figure 6 it is seen that the orientation of the side-chain carboxylate group of Glu72 has changed from that in ZnCPA. However, this shift occurs with preservation of the hydrogen bond of 72 Glu Oe2 with 197 Ser N.

The position of Tyr248 has occupied a central role in discussions of CPA structure and function (cf. Makinen et al., 1984; Garde11 et aZ., 1985). Analysis of the electron density map of ZnCPA shows that the side-chain of Tyr248 is found in the “up” position (Rees et al., 1983), as seen in Figure 6, probably because of the lattice contacts with residues 120 to 122 of a neighboring molecule. On the other hand, difference Fourier maps of enzyme- inhibitor complexes indicate that the side-chain of Tyr248 is found in a variety of “down” conformations in which it may hydrogen bond to

the C-terminal carboxylate group (Rees et al., 1980; Rees & Lipscomb, 1982; Chistianson et al., 1985) or may lie adjacent to the penultimate peptide group (Rees & Lipscomb, 1982) of inhibitors. In these complexes the terminal carboxylate group of the inhibitor molecule also forms an attractive electro- static interaction with the side-chain of Arg145.

In Figure 6 the comparison of the conformation of active site residues and the environment of the side-chain of Tyr248 in the MD-averaged protein and in ZnCPA shows that this residue has shifted significantly in position. In the MD-averaged structure, the side-chain of Tyr248 has shifted markedly from its initial “up” conformation and is found partially buried in the active site so that the phenolic OH group serves as an acceptor for a hydrogen bond from the NC-H group of Argl45. Inspection of the time-dependent development of this structural change showed that it was completed essentially within the first four picoseconds of MD steps. Analysis of our (unpublished) preliminary MD calculations of the protein in di- and tripeptide complexes completed to 70 picoseconds of MD steps shows that the orientation of the side-chain of Tyr248 has similarly moved to a “down” position. Since the same force field potentials have been employed in both sets of MD calculations, the structural alteration of Tyr248 and of its immediate neighbors cannot be ascribed simply to computational artifacts. Chemical modification studies of Tyr248, particularly with an arsiniloazo derivative, show that this side-chain can occupy different conformations in crystalline and solution states (Vallee et al., 1971; Johansen et al., 1976). Since the conformation of Tyr248 in the MD- averaged protein is not unlike that in the dipeptide Gly-Tyr inhibitor complex, as simulated by MD and characterized by X-ray methods, it is likely that the position of Tyr248 in the MD-averaged structure resembles closely that found for the native enzyme in solution. Upon completion of our MD studies of oligopeptide complexes of CPA, we shall provide elsewhere a more detailed, comparative analysis of the structural alterations of t’his region of the protein.

(iv) Domain structure Leibman et al. (1985) have established that there

are three folding domains in ZnCPA, comprising residues 1 to 122, 123 to 196 and 197 to 307, respectively. These regions are designated as domains A, B and C, and give rise to a variety of A : B, B : C and A : C interdomain, non-bonded contacts through side-chain interactions. We have analyzed the interdomain contacts in the MD- averaged protein as well as over the entire 20 picosecond trajectory and have compared them to those described in ZnCPA by Leibman et al. (1985). This analysis was carried out by counting the number of interresidue contacts each side-chain made with a residue of a different domain at each interdomain interface over the calculated trajectory. An interresidue, interatomic distance


& PHEi79 4PHE2+9

Figure 6. Stereo view of the active site of the MD-averaged protein and of ZnCPA, shown in the upper and lower diagrams, respectively. The largest structural difference is seen in the shift in orientation of Tyr248 and Ile247 from an “up” position in crystalline ZnCPA to a “down” position in the MD-averaged protein so that the Tyr Oq atom is hydrogen-bonded to the side-chain of Arg145.

5 4.0 A between non-hydrogen atoms was used to count the number of interresidue contacts. Comparison of the total number of interdomain contacts for each residue averaged over the 406 co- ordinate sets comprising the trajectory to the number of contacts in the static structure of ZnCPA showed that amino acid residues experienced essentially equivalent numbers of interdomain, interresidue contacts in each MD step, as in ZnCPA. Residues showing a decrease in the average number of interresidue contacts in MD steps were found to maintain interactions, nonetheless, with the same residues as in ZnCPA. On the other hand, amino acid residues exhibiting the largest increase in the number of interdomain, interresidue contacts were

associated with an increase in motional fluctuations. None of the residues associated with an increase in contacts were associated with residues of the constant core of secondary structure identified in Table 2 except for the interactions of Trp73 involving Asn123, Phe279 and Leu280. These interactions are discussed below.

The most prominent alterations in the increase in interdomain contacts occurred for amino acid residues near the active site. Tyr198 changed from its cis conformation to a trans structure with its phenolic OV atom hydrogen-bonded with active site residues His69, Arg71 or Glu72. The unfolding of the C-terminal portion of the a-helix 112-122 caused a change in the position of Asn123 and


,124

P PHEllS P PHEll8

PHEl18

Figure 7. Stereo view of the environment of Trp73 and the C-terminal portion of the a-helix 112-122 of CPA. The upper diagram illustrates this region of the enzyme in the MD-averaged protein. The lower diagram belongs to ZnCPA. The unfolded terminal portion of the a-helix 112-122 in the MD protein is readily identified by comparing the relative positions of PhellS and Arg124 in the upper diagram to those in ZnCPA. This unfolding results in new interresidue contacts formed by Trp73 with Phe279, Ile280, Asn123 and other residues constituting subsites of secondary substrate recognition.

Arg124 to interact strongly with residues Trp73, His120 and Ser121. The conformational change of the side-chain of Tyr248 to a “down” position evoked as a consequence a variety of new contacts with residues in the active site as seen in Figure 6. While residues exhibiting a large increase in interdomain interactions were found not to belong to the constant core of secondary structure identified in Table 2, except for Trp73, a small number of residues from the core exhibited a significant change in their interdomain, interresidue contacts due to a change in nearest neighbors. These changes involved primarily subtle, structural alterations of nearby parts of the protein.

The larger structural changes associated with the unfolding of the C-terminal portion of the a-helix

112-122 evoke a particularly interesting series of interactions with Trp73, Phe279 and Leu280. These are illustrated in Figure 7. In ZnCPA, the side- chain of Trp73, as the N-terminal residue of helix uH2 within the constant core of secondary structure, has only very distant interactions (N 5 to 7 A) with atoms of His120, Asn123 and Arg124 and none with Phe279. In the course of dynamics, however, the conformation of the side-chain of Trp73 has changed so as to interact strongly with all four of these residues. Consequently Trp73, adjacent to residues His69 and Glu72 that ligate the metal ion, now experiences the motional fluctuations of Asn123, Arg124, Phe279 and Leu280 that constitute sites of secondary substrate recognition (Quiocho t Lipscomb, 1971). The sites


of secondary substrate recognition are probably responsible for evoking torsional changes in substrate conformation required for bond hydrolysis according to stereoelectronic requirements (Kuo et al., 1983). It is, therefore, of interest that these regions are associated with large motional fluctuations while the residues directly associated with catalysis and specificity of substrate binding exhibit only small fluctuations and belong to the most rigid parts of the enzyme.

4. Discussion

As described above, the present study of the dynamical structure of carboxypeptidase A was restricted to the apoenzyme form for computational reasons. It is likely, nonetheless, that the simulation of motion by MD has characterized accurately the nature and extent of structural fluctuations of the protein. A variety of physical studies indicate that the apoenzyme exhibits structural properties in solution similar to those of the native enzyme. The apoenzyme in solution binds peptide substrates as tightly as does the native enzyme (Auld & Holmquist, 1974). Zinc ions are bound rapidly by the apoenzyme with concomitant recovery of enzymatic activity, and the binding of zinc ions by the apoenzyme is inhibited by peptides (Coleman & Vallee, 1962; Davies et al., 1968; Auld & Vallee, 1970). Also, the overall conformation of the crystalline apoenzyme is similar to that of ZnCPA (Lipscomb et al., 1966). In addition, while the difference Fourier synthesis of the Gly-Tyr inhibitor complex of apoCPA shows conformational differences primarily affecting the ligands to the metal ion, the larger structural changes involving Tyr248 occur as in the binding of Gly-Tyr by ZnCPA (Rees & Lipscomb, 1983). Our results demonstrated that the MD-averaged protein structure remained close to that of native ZnCPA and that the averaged extent of motional fluctuations and the r.m.s. deviations of c” and other non- hydrogen atoms from their initial X-ray positions were not significantly different from those reported in MD studies of smaller proteins. Furthermore, in MD simulations the position of the side-chain of Tyr248 shifted to occupy a site comparable to that observed in enzyme-inhibitor complexes. These correlated observations of experimental studies of the apoenzyme with our MD results indicate that the MD-averaged structure represents a physically realistic structural model of the apoenzyme in solution.

A notable difference between our results and the description of the native enzyme is the loss of the cis peptide bond with residues Ser197-Tyr198. This cis peptide bond is apparently preserved in the structure of the Gly-Tyr inhibitor complex of apoCPA (Rees & Lipscomb, 1983). The region near Ser197-Tyr198 interacts through non-bonded contacts with residues near Phe279 and Glu122 that form intermolecular contacts in the crystal; also, the crystalline apoenzyme is formed for X-ray

studies by removing zinc from ZnCPA in crystals. It is, therefore, possible that these interactions help to preserve the cis peptide bond of residues 197 to 198 in the crystalline apoenzyme. In this respect, it is of interest to note that the apoenzyme in solution shows reduced affinity for ester substrates in contrast to peptide substrates (Auld & Holmquist, 1974) that has been ascribed to the greater torsional flexibility intrinsic to the C-O ester bond (Cleland, 1977). The difference Fourier synthesis of t’he inhibitor complex of ZnCPA with (-)-2-benzyl-3-p- methoxy-benzoyl-propionic acid, a ketonic inhibitor analog of non-specific esters, shows that the side- chain of Tyr198 interacts with the p-methoxy group (Rees et al., 1980). Alterations of the position of Tyr198 in the apoenzyme of the type observed through our MD studies may account for t’he decreased affinity of the apoenzyme for torsionally more flexible ester substrates.

A striking result shown in Figure 3 is the very large differences of MD-derived isotropic r.m.s. fluctuations for a number of regions of the protein with corresponding values calculated on the basis of the X-ray-refined temperature factors for ZnCPA in crystals. The regions displaying the greatest differences belong to the intermolecular contact sites of ZnCPA in crystals (Rees et al., 1983). These regions can be of only low amplitide motion in the crystal. The differential mobility of the intermolecular contact residues evaluated for the isolated molecule in MD simulations and for the crystalline enzyme by X-ray diffraction is not unexpected on this basis. Comparable differential mobility is observed also through MD studies of other proteins. For instance, the regions of hen egg white lysozyme exhibiting the largest main-chain and side-chain fluctuations, as evaluated through MD studies, are the coiled loop regions 16-23, 62-75, 83-87, 102-104 and 116-120, the P-bend 47-49, and the C-terminal region 126-129 (Post et al., 1986). A significant fraction of these residues, in particular, 16-23, 65-68, 73, 84, 102-106, 116-118, 126, 128 and 129 are involved in forming intermolecular contacts in either the triclinic (I-‘,) (Moult et al., 1976) or the tetragonal (P4,2,2) (Blake et al., 1967) crystal. Similarly, among the intermolecular contacts in the monoclinic crystal of sperm whale metmyoglobin (Takano, 1977). residues 52, 56, 57, 59, 117-118, 144 and 147-153 exhibit greater motional amplitude, as calculated on the basis of MD simulations for the isolated molecule (Levy et al., 1985), than in t,he crystal. Through crystallographic studies it has not been hitherto appreciated that the intermolecular contacts of proteins in crystals may represent some of the most mobile regions of the free molecule. It, is possible that regions of high motional fluctuations are critical as intermolecular recognition sites requiring subtle positional adjustments of residues for juxtaposed molecules to form a three- dimensional periodic array.

We have pointed out in the Introduction that an important objective in our investigations has been


to identify the relationships of structural fluctuations of active site residues to stereochemical requirements of substrate hydrolysis. In Table 3 we have enumerated the principal amino acid residues of each subsite of substrate recognition in the active site of the enzyme based on X-ray diffraction (Quiocho & Lipscomb, 1971), and magnetic resonance and molecular modeling studies (Kuo et al., 1983; M. W. Makinen & J. M. Troyer, unpublished observations). We have also listed the corresponding r.m.s. fluctuations of the backbone (C’, C, 0, N) atoms averaged per residue as an index of their relative amplitudes of motion. The salient

Table 3 Enumerated listing of amino acid residues

comprising sites of substrate-enzyme interactions in carboxypeptidase A with corresponding amplitudes of

simulated structural jluctuations of their C” atoms

Site of substrate Amino acid A%W,. interationt residue1 (4 ( fS.D.)

9’ L1

S2

S 3

Arg127 0.573 Asp142 0.434 Ala143 0.437 Am144 0.421 Arg145 0.382 Glu163 0.560 Thr164 0.551 Ile243 0.475 Ala250 0.537 Gly253 0.582 Ser254 0.463 Thr268 0.396

His69 0.469 Glu72 0.504 His196 0.485 Arg7 1 0.535 Trp73 0.441 Leu201 0.528 Glu270 0.507

Ser197 0.571 Tyr198 0.623 Ser199 0.636 Thr246 0.851 Ile247 0.921 Tyr248 0.651 Thr274 0.946 Gly275 1.098 Arg276 1.119 Tyr277 0.940 Glv278 0619 Phe279 0.888 Leu280 0.848

His120 0.700 Serl21 0.928 Glu122 1.228 Asn123 1.238 Arg124 1.062 Leu125 0.901

0.489 + 0.07 1

0.496f0.031

0.839 kO.169

t See Fig. 1 for schematic illustration of subsites of substrate recognition.

$ Assigned on the basis of X-ray crystallographic (Quiocho & Lipscomb, 1971; Rees & Lipscomb, 1982) and molecular modeling (Kuo et al., 1983; M. W. Makinen & J. M. Troyer, unpublished results) studies.

SD., standard deviation.

result that emerges from this comparison is that the residues directly responsible for the chemical step of bond hydrolysis and the specificity of binding exhibit the smallest motional amplitudes of all sites that, interact with substrates. On the other hand, residues in the more distal sites of secondary substrate recognition exhibit motional amplitudes that are significantly larger. This identification of residues involved directly in catalysis and in the specificity of binding as having only low amplitude motion coincides directly with expectation. The relatively quiet residues in the S,’ site present to the incoming substrate a binding cleft of considerable geometrical rigidity. If this region exhibited structural flexibility, the binding specificity would be decreased since a variety of amino acid side-chains could then be accommodated. Furthermore, a charged arginine side-chain for ion pairing with the terminal carboxylate group of the substrate is located in the S,’ region. If this region of the enzyme exhibited high amplitude motion, the binding of the side- chain of the C-terminal residue of the substrate in the hydrophobic pocket and the interaction of the side-chain of Arg145 with the terminal carboxylate group would be entropically unfavorable. In an analogous manner the catalytic action of residues in the S1 site responsible for bond cleavage would also be entropically unfavorable if they experienced large amplitude fluctuations. The requirements of residues in this subsite are to induce changes in the chemical bonding structure of the substrate by polarizing the carbonyl group for approach of the attacking nucleophilic side-chain of Glu270. These changes can be induced only by stereochemically specific, rigidly constrained juxtapositioning of atoms of catalytic groups on the enzyme with the atoms of the scissile bond of the substrate.

In contrast to the rigid geometrical requirements for specificity of substrate binding and cleavage of the scissile bond, the distal sites of secondary substrate recognition appear to be involved in inducing torsional changes in substrate structure that favor bond hydrolysis according to stereoelectronic principles (Kuo et al., 1983). Since the catalytic action of CPA requires multi-point, co- operative interactions between the substrate and active site residues, these torsional changes will enhance the efficiency of the bond cleavage step within the less flexible, more rigidly constrained region of the substrate. Since the binding specificity is determined through the S,’ site, torsional changes in substrate conformation concomitant with or after substrate binding are more likely to be induced through steric interactions with highly mobile residues than with rigidly positioned amino acid residues in the SZ and S3 subsites. On this basis, it is readily understood why tripeptides are more efficiently hydrolyzed by CPA than are dipeptides, and why the steric bulk of the residue of the substrate that interacts with the SZ subsite on the enzyme is directly correlated with catalytic efficiency (Schechter, 1970). While the detailed


structural interactions of substrates with the S, and S, subsites of the enzyme require further analysis through MD simulations, the present results identify for the first time their large amplitude fluctuations and likely catalytic significance. Since residues in the S2 and S, subsites form intermolecular contacts in the crystal, such fluctuations are necessarily damped in the crystalline enzyme.

Comparable relationships are observed on the basis of MD simulations of motion in sperm whale myoglobin (Levy et al., 1985) and hen egg white lysozyme (Post et al., 1986). In sperm whale myoglobin the amino acid residues in van der Waals’ contact with the heme group, to which molecular dioxygen binds, are associated with only relatively low amplitude motion, as simulated through MD calculations. Regions more distant from the heme group exhibit greater simulated motion. The MD simulations of lysozyme show in a parallel manner that residues exhibit low amplitude motion in the active site, particularly in subsites C and D where bond cleavage occurs and where the thermodynamically important binding forces result through substrate-enzyme interactions. Residues involved in substrate interactions in more distant regions of the a,ctive site cleft, such as Asp101 in subsite A and Argll4 in subsite F, are associated wit’h significantly greater motional fluctuations.

We thank Professor W. N. Lipscomb and Dr D. C. Rees for providing the atomic co-ordinates of ZnCPA. This work was supported by grants from the National Institutes of Health (GM 21900), the National Science Foundation (BBS 86128379 and BBS 8616566), the Collaborat,ivr Research Grants Program of NATO (RG 0104), and the Foundation for Chemical Research (SON) under the auspices of the Netherlands Organiza- tion for the Advancement of Pure Research (ZWO). M. W. M. also acknowledges senior fellowships from the tJohn E. Fogerty International Center of the National Institutes of Health and the European Molecular Biology Organization.

References

Ayvist, J., van Gunsteren, W. F., Leijonmarck, M. & Tapia, 0. (1985). J. Mol. Biol. 183, 461-477.

Auld, D. S. & Holmquist. B. (1974). Biochemistry, 13, 4355-4361.

Auld. D. S. & Vallee, B. L. (1970). Biochemistry, 9, 602-609.

Bennett, W. S. & Huber, R. (1984). CRC Crit. Rev. Biochem. 15, 291-384.

Berendsen. H. J. (:. & van Gunsteren, W. F. (1983). In Proc. Advan. Studies Inst. on Physics of Superionic Conductions and Electrode Materials (Perram, J. W., ed.), pp. 221-240, Plenum, New York.

Berendsen, H. J. C.. Postma, J. P. M., van Gunsteren, W. F. & Hermans, J. (1981). In Intermolecular Forces (Pullman. B., Rd.), pp. 331-342. D. Reidel, Dordecht, The Netherlands.

Berendsen, H. J. (‘., Postma. ?J. I’. M.. van Gunsteren, W. F.. DiNola, A. & Haak. .J. R. (1984). .I. Chem. Phys. 81. 3684 -3690.

Blake, C. C. F., Mair, G. A., North. A. (‘. T.. Phillips. D. C. & Sarma, V. R. (1967). Pror. Roy. ~qoc. (London), 167B, 365-377.

Boschcov, P., Seidel, W.. Muradian. ,J.. Tominaga. M.. Paiva. A. C. M. & Juliano, L. (1983). Bioorg. (‘hrm. 12, 34-44.

Christianson, D. W.. Kuo. L. C. & Lipscomb, LV. 9. (1985). J. Amer. Chem. Sot. 107, 8281-8283.

Christianson, D. W.. David, P. R. 8: Lipscomb. W. 8. (1987). Proc. Nat. Acad. Sri., U.S.A. 84, 1512-1515.

Cleland, W. W. (1977). Advan. Enzymol. 45. 273-386. Coleman. J. E. & Vallee. B. L. (1962). J. Biol. (‘hem. 237.

3430-3436. Davies. R. C., Riordan. J. F., Auld. D. S. & Vallee. K. I,.

(1968). Biochemistry, 7, 1090-1099. Gardell, S. J.. Craik, (1. S., Hilvert, D.. I’rdea. 11. S. &

Rutter, W. ,J. (1985). Nature (London), 317. 551-555.

Genzel, I,., Keilmann, F.. Martin. T. P.. Winterling, (i.. Yacoby, Y.. Frohlich, H. & Makinrn. 11. W. (1976). Biopolymers, 15, 219--225.

Hendrickson, W. A. & Konnert,. J. H. (1980). In Biomolecular Structure, Conformation, and Evolution (Srinivasan. R., ed.), pp. 43-57, Pergamon Press, Oxford University.

Hermans. ,J., Berendsen, H. ,J. (‘., van Gunstrren, W’. F. & Postma, J. P. M. (1984). Riopolymers. 23. 1.51:~- 1518.

Johansen, J. T., Klyosov, A. A. 8 Valler. 13. I,. ( 1976). Biochemistry. 15, 296-303.

Jones. T. A. (1982). In Computational (‘rystallography (Sayre, D.. ed.). pp. 303-317. (‘larrndon Press, Oxford.

Jones, T. A. (1985). In Methods in Enzymology (Wyrkoff. H. W.. Hirs. C. H. W. & Timasheff. S. S., eds), vol. 115, pp. 157-171, Academic Press, New York.

Konnert. J. H. (1976). Acta Crystallogr. SPCI. .-I, 32, 614-617.

Kriiger, P.. Strassburger. R’.. 12’ollmer. A. & \‘a,] Gunsteren, W. F. (1985). Eur. J. Riophys. 13. 77 -88.

Kuo. L. C., Fukuyama, J. M. & Makinen. M. FV. (1983). J. Mol. Biol. 163, 63-105.

Lesk, A. M. & Chothia. C. (1984). -1. Mol. Biol. 174, 175-191.

Levitt, M. (1983a). J. Mol. Biol. 168, 595-620. Levitt, M. (19836). J. Mol. Biol. 168. 621-657. Levy, R. M.. Sheridan, R. P., Keepers, .J. W.. Dubey.

G. S., Swaminathan. S. & Karplus, M. (1985). Biophys. J. 48, 509-518.

Liebman. M. S., Venanzi. (‘. A. & Wrinstrin. H. (1985). Biopolymers, 24, 1721-1758.

Lipscomb, W. N., Coppola, J. (“., Hartsuck. ,J. A., Ludwig, M. L., Muirhead, H., Sear]. ,J. (Ir Steitz. T. A. (1966). J. Mol. Biol. 19, 423-441.

Lipscomb. W. N.. Hartsuck, J. A.. Reeke, C. N., Jr. Quiocho. F. A., Bethge, P. H., Ludwig, M. L, Steitz, T. A., Muirhead, H. dt Coppola. ,J. (1968). Hrook?uwen STymp. Biol. 21, 24-90.

Makinen, M. W.. Wells. G. B. & Kang, S. 0. (1984). Advan. Inorg. Biochem. 6, l-69.

McCammon, J. A., Gelin, B. R. & Karplus, M. (1977). Nature (London), 267. 585-590.

Moult, J., Yonath. A., Traub. W.. Smilansky, A., Podjarny, A., Rabinovich. D. & Saya. A. (1976). J. Mol. Biol. 100, 179-195.

Northrup, S. H., Pear, M. R., McCammon. .J. A.. Karplus, M. & Takano. T. (1980). Nature (London), 287, 659-660.

Nort,hrup. S. H., Pear, M. R., Morgan, ,J. I).. MeCammon.


J. A. & Karplus, M. (1981). J. Mol. Biol. 153, 1087-1109.

Post, C. B., Brooks, B. R., Karplus, M., Dobson, C. M., Artymiuk, P. J., Cheetham, J. C. & Phillips, D. C. (1986). J. Mol. Biol. 190, 455-479.

Quiocho, F. A. 6 Lipscomb, W. N. (1971). Advan. Protein Chem. 25, l-78.

Rees, D. C. & Lipscomb, W. N. (1982). J. MOE. BioE. 160, 475498.

Rees, D. C. & Lipscomb, W. N. (1983). Proc. Nut. Acud. Sci., U.S.A. 80, 7151-7154.

Rees, D. C., Honzatko, R. B. t Lipscomb, W. N. (1980). Proc. Nat. Acad. Sk., U.S.A. 77, 3288-3291.

Rees, D. C., Lewis, M., Honzatko, R. B., Lipscomb, W. N. & Hardman, K. H. (1981). Proc. Nut. Acad. Sci., U.S.A. 78, 3408-3412.

Rees, D. C., Lewis, M. t Lipscomb, W. N. (1983). J. Mol. BioZ. 168, 367-387.

Ryckaert, J. P., Ciccotti, G. A. & Berendsen, H. J. C. (1977). J. Comput. Phys. 23, 327-341.

Sander, M. E. & Witzel, H. (1985). Biochem. Biophys. Res. Commun. 132, 681-687.

Sander, M. E. & Witzel, H. (1986). In Zinc Enzymes (Bertini, I., Luchinat, C., Maret, W. & Zeppezauer, M., eds), pp. 207-214, Birkhaueser, Basel.

Schechter, I. (1970). Eur. J. Biochem. 14, 516-520.

Schulz, G. E. & Schirmer, R. H. (1979). In Principlee of Protein Structure, chapt. 5, pp. 66-107, Springer- Verlag, New York.

Tainer, J. A., Getzoff, E. D., Alexander, H., Houghten, R. A., Olson, A. J., Lerner, R. A. & Hendrickson, W. A. (1984). Nature (London), 312, 127-133.

Takano, T. (1977). J. Mol. Biol. 110, 537-568. Vallee, B. L., Riordan, J. F., Johansen, J. T. &

Livingston, D. M. (1971). Cold Spring Harbor Symp. Quant. BioZ. 36, 517-531.

van Gunsteren, W. F. & Berendsen, H. J. C. (1977). Mol. Phys. 34, 1311-1327.

van Gunsteren, W. F. & Berendsen, H. J. C. (1987). Croningen Molecular Simulation (PROMOS) Library &znuaZ, p. 229, BIOMOS, Nijenborgh 16, 9747 AG Groningen, The Netherlands.

van Gunsteren, W. F. & Karplus, M. (1980). J. Comp. Chem. 1, 266-274.

van Gunsteren, W. F., Berendsen, H. J. C., Hermans, J., Hol, W. G. J. & Postma, J. P. M. (1983). Proc. Nut. Acud. Sk., U.S.A. 80, 4315-4319.

Westhof, E., Altschuh, D., Moras, D., Bloomer, A. C., Mondragon, A., Klug, A. & van Regenmortei, M. H. V. (1984). Nature (London), 311, 123-126.

Whitlow, M. & Teeter, M. M. (1986). J. Amer. Chem. Sot. 108, 7163-7172.

Edited by R. Huber

Documents

Dynamical Structure of Carboxypeptidase A€¦ · 1. Introduction basis of X-rav studies to identifv how the extent, of The constituent amino acid residues of enzymes motional fl&tuations