The Proximal Residue Largely Determines the CO Distortion in

An ab initio investigation of a model heme prosthetic group based on the carbon monoxy myoglobin (MbCO). lMBC X-ray structure reproduces the large ...
0 downloads 0 Views 2MB Size
J. Phys. Chem. 1995, 99, 12677-12685

12677

The Proximal Residue Largely Determines the CO Distortion in Carbon Monoxy Globin Proteins. An ab Initiu Study of a Heme Prosthetic Unit? Philip Jewsbury,***** Shigeyoshi Y amamoto," Tsutomu Minato,l Minoru Saito,s and Teizo Kitagawas Institute for Molecular Science, Myodaiji, Okazaki 444, Japan, Protein Engineering Research Institute, 6-2-3 Furuedai, Suita, Osaka 565, Japan, Chukyo University, IO1 Tokodachi, Kaizu, Toyota 470-03, Japan, and Institute for Natural Science, Nara University, I500 Misasagi, Nara 631, Japan Received: February 13, 1995; In Final Form: June 15, 1995@

An ab initio investigation of a model heme prosthetic group based on the carbon monoxy myoglobin (MbCO) lMBC X-ray structure reproduces the large off-perpendicular distortions of the Fe-C-0 unit reported for the protein. The distortion is mainly caused by the nonequilibrium orientation of the proximal residue and not by the distal residue: inclusion of the distal residue in a supermolecule calculation has a smaller effect on the Fe-C-0 geometry. If such a mechanism primarily determines the Fe-C-0 distortion in the protein itself, then the large strain energies implied by the Fe-C-0 geometries in the X-ray structures are delivered by the protein tertiary structure, via the proximal residue, and not by the mobile distal side chain, as had been previously proposed. The structure-function relationship, as revealed by the X-ray structure, would then be clarified. Distortion of the Fe-C-0 geometry is largely determined by the proximal residue, and so FeC - 0 is nonperpendicular even in the His64Gly mutant. The distal residue is not subject to a large repulsive interaction with the carbonyl ligand; thus, its orientation in the solvated protein can be determined by weaker attractive electrostatic interactions, as inferred from recent experimental studies of distal residue mutant myoglobins. This result removes the need to invoke a large stabilization of the distal side chain orientation by a rigid hydrogen-bonding network, an interpretation of the physiological structure-function relationship that was at odds with the X-ray B factors and the mobility of surface residue side chains expected under physiological conditions.

Introduction How the globin proteins execute their physiological role of oxygen transport and storage in the presence of significant CO concentrations has been the focus of considerable research interest.' The physiological ligand, 02, binds to free porphyrin in a bent conformation,2 which, in the globin protein, is further stabilized by a hydrogen bond to the terminal oxygen atome3 However, the poisoning ligand, CO, binds to free porphyrin perpendicular to the heme plane,2 up to 25 000 times more strongly than ~ x y g e n .This ~ relative affinity is reduced to only 30 in myoglobin' (Mb), and how the heme pocket discriminates in favor of oxygen has become an important example of the control exerted by a protein over its binding site (see, for example, Strye@). The first high-resolution X-ray structure of MbCO (Protein D a t a b a ~ k(PDB) ~ . ~ structure 1MBC') showed the Fe-C-0 unit bent and tilted away from the distal side chain, the nearest residue to the binding site (shown in Figure 1). These conformations were originally interpreted in terms of a steric repulsion between the distal residue and the CO ligand,7-9 forcing the carbonyl into a nonperpendicular binding geometry and thereby reducing its affinity for the protein binding site. However, the bend angles observed in the crystal structures are large, and it has been difficult to account for such large strain

* Author for correspondence. Present address: Zeneca Pharmaceuticals, Mereside, Alderley Park, Macclesfield, Cheshire SKI0 4TG, U.K. Email: [email protected]. 'The preliminary results of this study have previously been communicated (J. Am. Chem. SOC. 1994, 116, 11586-11587). L Institute for Molecular Science. e Protein Engineering Research Institute. l 1 Chukyo University. Nara University. Abstract published in Advance ACS Absrracts, July 15, 1995. @

energies being delivered by the distal side chain? which, being exposed to the solvent, is quite mobile, as confirmed by the X-ray B factors7 and in molecular dynamics simulations.'0,'' More recently, the results of a series of experimental studies of site specific mutant MbCO proteins have suggested that the key distal residue-CO interaction is electrostatic in nature: only those distal mutants with a polar side chain show any significant change in their infrared (IR) spectra from that of the His64Gly mutant.I2-l4 Little difference was found between the IR,I2-l4 and resonance Raman,I5 spectra of distal mutants with aliphatic side chains of various steric bulk; nor did their X-ray structures differ significantly.I6 It has therefore been proposed that the main influence over the Fe-C-0 bonding is the extent of electrostatic polarization by the nearby distal r e s i d ~ e . ' ~ - ' ' - ' ~ While based on compelling experimental evidence, this reinterpretation of the Mb distal-CO interaction does not fully account for the functional role of the heme pocket. In particular, the wild type vco frequencies are lower, and the V F ~ Cfrequencies higher, than the aliphatic distal mutants, indicating that the distal-CO interaction in the wild type leads to increased FeCO back-bonding, that is, the distal histidine is stabilizing the CO ligand. In the electrostatic interpretation the distal-CO interaction is similar in nature, though weaker, to the distal0 2 interaction. It seems unlikely that this relatively small difference in the distal-CO and distal-02 interactions could lead to the very large change in the relative affinities of these ligands for the heme on placing it in the protein matrix. That, however, is the key functional achievement of the globin heme pocket. Furthermore, in this electrostatic interpretation, the shift to lower vco frequencies for the wild type relative to the His64Gly mutant MbCO implies that the carbonyl ligand is polarized by a positive electrostatic potential. The only candidate is the nearby NH hydrogen of the distal side chain, but the neutron diffraction

0022-3654/95/2099-12677$09.00/0 0 1995 American Chemical Society

12678 J. Phys. Chem., Vol. 99, No. 33, 1995

I

Jewsbury et al.

haem

proximal Histidine

Figure 1. Stereo representation of certain key residues in the carbon monoxy myoglobin heme pocket of PDB structure 2MB5. Two carbonyl 0 positions were resolved.

structure of wild type MbCO (PDB file 2MB5I9),the only direct experimental measure of the distal-CO conformation, shows the protonated Nd nitrogen pointing out of the heme pocket; the lone pair of the unprotonatedN, nitrogen is orientated toward the carbonyl group, indicating that the ligand lies in a negative electrostatic potential. A recent molecular dynamics (MD) study of wild type myoglobin1’ has, however, suggested that the solution phase orientation of the distal side chain may differ from that of the crystal: both the His64 tautomeric states had the side chain orientated with the polar hydrogen inside the heme pocket during the 90 ps trajectories. The 64N,H tautomer forms a stable electrostatic interaction with the carbonyl 0, while the 64NaH tautomer forms only a weak interaction with the ligand but is held inside the heme pocket by interactions with the 67Thr residue. The two peaks in the room temperature solution phase IR spectrum were assigned one to each distal tautomeric state on account of their significantly different distal-CO interactions. A third peak, observed at low pH andor low temperature, at a frequency similar to the His64Gly mutant, has been interpreted as representing an “open” pocket structure20 where the distal side chain has swung out of the heme pocket and is hydrogen bound to the Asp60 main chain oxygen.‘1,2’,22 This structural interpretation of the wild type IR spectrum is consistent with studies of the His64Gln mutant, which has an IR s p e ~ t r u m ’ similar ~ . ’ ~ to that assigned to the wild type 64NdH tautomer, but a connectivity resembling the 64N,H tautomer: MD simulations show that the His64Gln-CO interaction is “64NbH-like” since repulsions between the glutamine 0, and the carbonyl oxygen prevent the close interaction found for the wild type 64N,H tautomer.23 MD studies of the His64Leu and His64Gly mutants also suggest how steric interactions and the presence of solvent molecules in the heme pocket, respectively, may influence the ligand e n v i r ~ n m e n t .All ~ ~ these simulations, and an earlier study,24 have the carbonyl ligand distributed around an axis close to the heme normal, revealing that there is no significant steric influence of the distal residue over the CO orientation, and suggest that the key determinant of the vco frequency is electrostatic and not steric in nature, consistent with the recent spectroscopic results. The solution phase structure of wild type MbCO has recently been determined by Osapay et al. Contrary to what would be expected from the results discussed above, their heme pocket geometry is similar to that of the crystal phase X-ray structure, although the distal residue is displaced further out into the solvent.25 Their determination of the distal histidine Orientation, however, allows for only one distal tautomer and is subject to

uncertainties in the Fe-C-0 geometry. Since the solution phase IR spectrum under similar conditions shows two peaks, which have been interpreted as representing two distal side chain orientationdtautomers,or two Fe-C-0 conformations,it would be interesting to investigate what effect including these possibilities would have on the distal geometry determined from the NMR data. It is diffic~lt~.”,’~.”,~~~~~,~~ to reconcile the MbCO IR spectra, which have multiple peaks suggesting multiple heme pocket-ligand interactions, with the single heme pocket geometries resolved in recent X-ray and NMR studies. On the basis of the current spectroscopic evidence and supported by theoretical studies, it therefore appears that the distal-CO interaction in wild type MbCO is largely electrostatic, and stabilizing, in nature. Although the mutation of other residues, in particular Leu29 and Va168, has a significant effect on the CO stretching frequencyl2-I4 and binding affinity,28 it is unclear how the dramatic change in the relative affinity of CO and 0 2 for the protein heme, over that for a free heme, is achieved. In this paper we investigate the origin of the CO destabilization in the protein by determining the preferred ab initio orientation of a CO ligand in a model globin prosthetic group and comparing the results with those for a model free heme group. A secondary aim has been to assess the validity of the Fe-C-0 distortions in the X-ray structures, which have been recently criticized as unreasonable on the basis of normal mode analyses of an isolated Fe-C-0 unit.’* We find that, although the calculations are accurate to only a semiquantitative level, the distortions of the Fe-C-0 conformation in our model globin prosthetic group are similar to those found in the X-ray structures, but are primarily caused by the distorted orientation of the proximal residue: the Fe-C-0 geometry remains significantly distorted even in the absence of a distal side chain. However, when the proximal residue was allowed to relax from its orientation in the protein, a perpendicular Fe-C-0 conformation, Le. the orientation found experimentally in the free heme, resulted. If the same mechanism is determining the CO distortion in the protein itself, then these results have profound implications for the structure-function relationship in globin proteins: they reconcile apparently conflicting experimental data and suggest the “destabilizing factor” missing from the recent electrostatic reinterpretation of the structure-function relationship in MbCO.

Method We have based our geometry for a model globin prosthetic group on the highest resolution MbCO structure (PDB file 1MBC7), with the protonation site of the histidine side chains

Role of Proximal Residue in Globin Proteins

0

proximal Histidine

Figure 2. Model heme prosthetic group based on the PDB structure 1MBC: = 2.19 A, 4 = 86.3', and 8 = 130.3'. This represents the optimized "proximal 8z distal" geometry of Table 1.

assigned according to the neutron diffraction structure (PDB file: 2MB5I9). The porphyrin macrocycle was simplified to two amidinato ligand^,^^.^^ and the distal and proximal residues to their side chains, rotated to maintain C.,symmetry. A Hay and Wadt 16-electron p~eudopotential~' was used at the Fe center, with Goddard's triple-< d orbital c~ntraction.~~ The axial ligands were described by Dunning's ( 9 ~ 1 5bases,33 ~) contracted to [6111/411/1] with additional polarization (exponents 0.800) and diffuse (exponents 0.030, 0.040, 0.050 for C, N, 0, respectively) functions. The equatorial ligands were described by minimal STO-3G bases;34 the distal residue and proximal hydrogens, by 3-21G bases.35 All geometry optimizations were constrained to C, symmetry and calculated at the MP2 level using G a u ~ s i a n 9 2 . The ~ ~ bend @), tilt (z), and Fe-C bond length (RFe-C) have been optimized, in both the presence and absence of the distal residue, with the positions of all the other atoms determined from the X-ray structure (see Figure 2). The C-0 bond length (Rc-0) was kept frozen at 1.13 A, consistent with the free carbon monoxide bond length37and the bond length found in the model heme compound, Fe[1(TPP)(pyr)(CO),38 to prevent the overestimate of Fe-C back-bonding found in test calculations. The model free heme geometry was developed consistent with the model globin prosthetic unit, except that no distal group is included. In the free heme, the orientation of the proximal side chain is not influenced by the protein tertiary structure, but is free to adopt its equilibrium position. Thus, in this MP2

J. Phys. Chem., Vol. 99, No. 33, 1995 12679 geometry optimization of the Fe-C-0 geometry, the orientation of the proximal ligand was also allowed to relax from that in the protein; that is, the proximal parameters, RFe-N, 8, and 4, as well as the bend, tilt, and Fe-C bond length of the FeC - 0 unit, were included in the variable list. The same basis sets as described above were used. As well as these MP2 optimizations, a number of further calculations were made to probe the structural determinants of certain model heme geometries. To achieve a reasonable range of data within acceptable CPU limits, these calculations were made at the SCF level with all bond lengths constrained (RFe-C = 1.77 A, R F ~ - N= 2.13 A). Since SCF optimizations reproduced the geometrical trends of the MP2 calculations described above, we assume that these further SCF investigations will likewise be useful in determining the geometrical trends of other, closely related, model heme geometries. A number of simplifications of the two systems were necessary to allow an MP2 optimization of such a large system. The most significant are the rotation of the histidine side chains to maintain C.,symmetry, the simplification of the porphyrin macrocycle, and the minimal basis set description used for the equatorial ligand. Furthermore, these calculations have been performed in vucuo, with the majority of the protein neglected; here we consider the likely effect that these simplifications have on the CO orientation in these model systems. Orientation of the Histidine Side Chains. It is difficult to assess to what extent the rotation of the histidine side chains from their X-ray positions will affect the orientation adopted by the CO ligand. The proximal residue was rotated around its Fe-N bond and placed in the C, plane while maintaining the X-ray orientation of the imidazole ring relative to the heme plane; this probably has only a minor effect on the CO orientation, as the changes from the X-ray structure are small. The distal side chain was moved into the C.,plane around an arc parallel to the heme plane; a small rotation of the imidazole plane around the Fe-N, bond was also necessary to bring it parallel to the heme perpendicular. Its separation from the heme normal through the Fe atom was maintained, as was the approximate orientation of the imidazole bond vectors relative to the heme perpendicular. This operation should keep any steric distal-CO interaction constant, since the separation between the distal residue and the heme normal has been maintained; however, it will increase electrostatic intermolecular interactions since the distal imidazole plane is now orientated parallel to the heme normal, rather than the oblique orientation found in the crystal structure. It is important to consider the mobility of the side chains, as revealed by the X-ray B factors, in assessing the significance of these deviations from the X-ray geometry. The proximal side chain has B factors similar to those of the backbone atoms (3.56-4.73 A2), indicating a restrictive local environment; the distal side chain, however, is considerably more mobile, with B factors typical of the other side chains located on the protein surface (11.02-12.54 A2).7 It is thus important that the orientation of the proximal side chain is hardly altered in the model globin prosthetic unit; the changes in the orientation of the distal side chain, though larger, may be less functionally significant due to its inherently more mobile nature. The Fe-C-0 group is constrained to relax in the C.yplane, which is expected to have a softer potential than other distortion coordinates since they would involve interactions with the macrocycle n electron density at the nitrogen atoms. One conformer in the lMBC X-ray structure, the D conformer, is repelled from the distal side chain in a direction similar to the C, plane in our model globin prosthetic group. It has been argued that the other conformer in that structure, which has a

Jewsbury et al.

12680 J. Phys. Chem., Vol. 99, No. 33, 1995 7 6

2 5

-

. 1 4 f 3 w 2 1

0 150

155

180

165 Bend angle, B /

170

175

150

180

155

160

165 Bend angle, P /

170

175

180

w

60

82

84 86 Tilt angle, 1 / O

88

90

Figure 3. SCF (a) bending and (b) tilting potentials of the Fe-C-0 unit in the model heme prosthetic unit (solid circles) and an equivalent structure with the full porphyrin macrocycle (open circles).

distal-CO separation less than the distance from the distal side chain to the heme normal, arises from an attraction to the N,H distal tautomer1 and is therefore inconsistent with the assignment of the distal histidine protonation state made in the neutron diffraction structure (2MB5). Simplification of the Porphyrin Macrocycle. The amidinato structure we have used in this study is the largest of the model porphyrin ligands considered by Newton and Hall in their earlier study of the binding of 0 2 to metall~porphyrin.~~ This ligand has the advantage that its heavy atoms are in the same positions as their equivalents in the full porphyrin macrocycle; it is also expected to more faithfully represent the n electron delocalization of the macrocycle than the other model porphyrin ligands considered in previous studies. However, this simplification of the porphyrin ligand may have a significant bearing on the orientation of the CO ligand since the porphyrin macrocycle competes with the carbonyl for the Fe d electrons. We have investigated the effect of this simplification by comparing the SCF potentials for bending and tilting the FeC - 0 unit in the model heme prosthetic unit with those of an equivalent structure which includes a full porphyrin macrdcycle. The results are shown in Figure 3, where we have varied the bend and tilt angles independently over the ranges 80-90" and 150-180°, respectively; the same basis sets were used as have been described above. Basis Set Quality. It was necessary to use minimal basis sets in describing the model porphyrin atomic orbitals to allow for a sufficiently detailed description of the axial ligands at the MP2 level. Since the porphyrin ligand itself is incomplete in our model globin prosthetic unit, the effect of this choice of bases cannot be adequately investigated for the geometry considered above; however, a related set of calculations for Fe(C0)s can suggest the quality of the basis set description used here. In Figure 4 we show the SCF potentials for bending and tilting one axial ligand in Fe(CO)5 over the same range of angles considered in the previous section; all bond lengths were fixed to their experimental values.39 Three basis set descriptions of the CO ligands have been considered: the 6-311G** bases,40

80

82

64

86

08

90

Till angle, T /

Figure 4. The SCF (a) bending and (b) tilting potentials of one axial CO ligand in Fe(C0)S as a function of the ligand basis set: 6-311G** (open triangles); a [6111/411/1]contraction of Dunning's (9d5p)bases (open diamonds); and the description used in this study, i.e. a [6111/ 41 1/11 contractions of Dunning's (9s/5p) bases for the axial ligands, and STO-3G bases for the equatorial ligands (solid circles).

the [6111/411/1] contraction of Dunning's (9s/5p) bases used above, and the combination of bases used in the optimizations above (i.e. the [6111/411/1] contraction of Dunnin's (9s/5p) bases for the axial ligands, and a STO-3G description for the equatorial ligands). The basis set used at the Fe center was, in each case, the same as that described above. Although the tilting coordinate is found to be significantly softer when minimal STO-3G bases are used for the equatorial centers, the effect of this choice of basis set will, to some extent, be offset by the overestimated tilting potential of the model porphyrin macrocycle (see Figure 3); improvement of the equatorial basis description is not feasible in the large MP2 optimizations reported below. It is noteworthy that the major distortion coordinate of the Fe-C-0 unit in the MP2 optimizations is the bend angle (see Table 1). Neglect of Solvent and Remaining Protein Residues. Although desirable, it is clearly beyond current capabilities to include solvent molecules, or even the rest of the protein, in these MP2 calculations. Our simplification to model prosthetic units in vacuo is drastic, however: it is well-known from molecular dynamics simulations that the Coulombic interaction is significant even for atom pairs separated by up to 15 A; the side chains of other residues on the distal side of the protein pocket could also be involved in interacting with a carbonyl ligand; and the polar solvent molecules will screen electrostatic forces inside a low-dielectric protein.41-43 The combined effects of these omissions are hard to quantify. The mutation of two other residues on the distal side of the heme pocket, Leu29 and Val68, is known to significantly alter the MbCO IR spectrum. In the wild type, both these residues are aliphatic and therefore will interact with the ligand through dispersive forces and sterically limit the volume of the distal pocket available to the ligand. The neglect of these residues is likely to lead to an overestimate of the distortion induced by the proximal and distal histidines. The omission of the polar solvent water molecules neglects their screening of the elec-

J. Phys. Chem., Vol. 99, No. 33, 1995 12681

Role of Proximal Residue in Globin Proteins

TABLE 1: MP2 Optimization of the CO Orientation in a Model Heme Prosthetic Group” structural parameters co RFe-CIA tilt,bdo bend, /3/’ orientationC Ab Initio Optimizations free heme 1.79 88.6 179.6 1.8 globin heme 34.7 75.9 159.44 proximal only 1.79 distal & proximal 1.91 72.4 151.7 45.9 PDB Structure lMBCd C D

1.92 1.92

89.2 89.2

141.4 119.9

39.4 59.6

a The perpendicular orientation in the free heme optimization with no proximal strain and no distal residue (upper); the distorted orientations in the globin heme optimizations based on the PDB structure IMBC, both with and without the distal residue (middle); the orientation determined experimentally in the PDB structure (lower). Orientation of the Fe-C bond relative to the mean plane through the four porphyrin nitrogen atoms. An angle of 90’ lies on the heme normal. Orientation p of the CO bond relative to the normal to the mean plane through the four porphyrin nitrogens. An angle of 0” lies on the normal. d T ~ carbonyl o 0 positions were resolved: C, with occupancy 78%, and D, with occupancy 22%.

trostatic interactions between atoms in the protein; in particular, this is likely to overestimate the electrostatic influence of the distal side chain over the Fe-C-0 geometry. The neglect of protein residues further from the binding site cannot be readily assessed.

Results Table 1 shows the orientations of the CO ligand in the various systems considered. The extent of distortion from the heme normal in the “full” model prosthetic group is similar in magnitude to that of the X-ray structures; however, it is primarily determined by the distorted orientation of the proximal residue and not by interactions with the distal residue (compare the “proximal & distal” entry with the “proximal only” entry). This distorted orientation of the proximal ligand causes a rotation of the Fe d orbitals away from the axes defined by the heme plane toward those defined by the Fe-NHis bond, thus leading to an off-perpendicularbinding of the CO ligand. When the proximal orientation was allowed to relax from the protein geometry, a perpendicular Fe-C-0 orientation, similar to that in the X-ray crystal structure of Fe11(TPP)(pyr)(CO),38 was found. In this optimization, the equilibrium orientation of the proximal residue was 4 = 90.4”, 8 = 126.7”, and R F ~ - N =1.92 A. A further series of calculations were made at the SCF level to investigate which of the proximal parameters was the key determinant of the CO orientation. In these supplementary calculations, the proximal angles 8 and 4 were independently varied over the range of angles observed in the PDB structures of MbCO and carbon monoxy hemoglobin (HbCO). The orientation of the CO group (i.e. 1,9 and ‘G)was then optimized at the SCF level, with all other proximal parameters fixed to their average values in the X-ray structures. The results are shown in Table 2. Over these ranges, the Fe-C-0 geometry was left almost unaltered by the changes in 4 (changes of less than 1.5” in the CO orientation relative to the heme normal), whereas with 8 greater than 125’, or less than 115O, a large change in the CO orientation of up to 30” was found. Although the two coordinates 8 and q5 couple in determining the FeC - 0 conformation, it seems that the primary determinant of the distorted CO orientation in this model globin prosthetic unit is the nonequilibrium value of the angle 8. It is difficult to assess the influence of the proximal residue over the CO geometry from the X-ray structures available in the PDB

TABLE 2: SCF Geometry of the Carbonyl Ligand as a Function of the Proximal Orientation in a Distal-Free Model Globin Prosthetic Group CO geometry proximal geometry co 0 deg q5 deg tiltb,ddeg bend, Pldeg orientation‘ deg 123.1 80.0 76.1 162.6 31.3 123.1 82.5 76.2 31.9 161.9 123.1 84.9 76.5 161.7 31.9 123.1 87.5 77.1 161.6 31.3 90.0 77.9 123.1 161.7 30.5 110.0 84.9 89.9 177.0 -2.9d 84.9 77.5 115.0 162.6 29.9 120.0 84.9 76.6 161.7 31.7 123.1 84.9 76.5 161.7 31.9 125.0 84.9 84.6 161.6 31.9 130.0 84.9 84.7 126.2 59.0 135.0 84.9 84.7 126.2 59.1 The CO orientation was determined as the proximal angles, 8 and q5, were varied independently over the range of values found in the MbCO and HbCO PDB structures (see Table 3). All other proximal earameters were set to an average geometry which has R F ~ - N = 2.12 A, 8 = 123. lo, and $J = 84.9”. Orientation of the Fe-C bond relative to the mean plane through the four porphyrin nitrogen atoms. An angle of 90” lies on the heme normal. Orientation p of the CO bond relative to the normal to the mean plane through the four porphyrin nitrogens. An angle of 0” lies on the normal. Bend of the Fe-C-0 unit is in the opposite sense to the other geometries considered. ~

database: differences in the analyses of the electron density, in particular in the interpretation of the Fe-C-0 conformation, make comparisons between the studies of different research groups complicated. The strain energy delivered by the distorted orientation of the proximal residue in the protein (i.e. the difference in the energies of the “proximal only” and “free heme” entries in Table 1) is 7.2 kcal/mol. This is similar to that estimated by Case and Karplus in a molecular mechanics analysis of the X-ray structure, which they considered to be at the limit of the strain energy likely to be delivered by a protein to its binding sites9

Discussion Although these calculations are for a model heme prosthetic group and are accurate to only a semiquantitative degree due to limitations in the methodology, they reproduce the distorted Fe-C-0 geometry of the MbCO X-ray structure on which they were based. If the conclusions we can draw from these results are also applicable to the protein, then they have major implications for the structilre-function relationship in this important family of proteins. It is clear from Table 1 that the nonequilibrium orientation of the proximal residue is largely responsible for the nonperpendicular orientation of the CO ligand in this model heme prosthetic unit. This is important since it is unlikely that, in the MbCO protein under physiological conditions, the distal side chain alone could deliver the large strain energy implied by the distorted Fe-C-0 geometry in the X-ray structures; rather, being the side chain of a surface residue, it would move to relieve such a large steric repulsion with the carbonyl ligand. It has been previously proposed that a solvent hydrogen-bonding network, with a solvent molecule bridging the Arg45 and His64 side chains (see Figure l), stabilizes the distal orientation in the crystal, on the assumption that the distal side chain is in a sterically strained position. However, such a specific stabilization requires the appropriate orientation of all three groups and is not expected to persist under physiological conditions where surface residue side chains show increased mobility. Altematively, others have suggested that the CO ligand is not, in fact, significantly distorted,18.44i.e. that it lies close to the heme

12682 J. Phys. Chem., Vol. 99, No. 33, 1995

Jewsbury et al.

TABLE 3: Structural Parameters of the Proximal and Carbonyl Ligands in Selected HbCO and MbCO Protein Databank Files Together with the SCF Orientation of the CO Ligand in Model Heme Prosthetic Units Having Equivalent Proximal Residue Geometries but No Distal Side Chain crystal parameters SCF parameters resol$ion roximal geometry CO geometry CO geometry PDB file" a A Fe-Nli 81" $1" tbl" PI" Cpl" rbl" PI" @IO HbCO 1HCO 2.7 2.06 105.3 79.7 76.2 a 179.8 13.9 88.3 183.9 -2.2' 2.7 1.80 126.3 179.8 12.3 77.3 77.9 75.1 162.2 32.7 P 2.09 a 2HC0 2.7 125.2 179.5 11.4 77.8 161.7 89.2 79.1 30.5 2.08 125.8 179.9 88.9 79.5 2.7 10.6 77.8 161.8 30.4 P lBBB a 1.7 2.13 119.2 156.2 77.0 87.0 83.0 5.8 161.5 31.5 2.10 1.7 128.0 171.6 7.1 77.5 161.9 87.9 82.7 30.6 P a lCOH 2.9 2.25 116.9 153.0 29.2 76.5 85.2 86.9 161.6 31.9 1NIH 2.18 2.6 124.2 153.2 31.5 76.5 31.9 89.4 84.8 161.6 P 1SDH 2.4 2.20 127.9 142.4 a 35.7 76.6 162.0 31.4 84.8 88.0 2.4 2.16 120.8 155.0 25.1 88.2 89.4 77.2 161.6 31.2 P 2.14 a lPBX 2.5 129.9 152.5 14.2 76.8 85.3 74.5 162.0 31.2 2.15 2.5 124.9 125.3 32.7 77.4 161.8 30.8 88.2 79.1 P lECO 1.4 2.12 124.6 160.4 85.1 82.1 25.9 76.5 161.6 31.9 2.13 1HBG 1.5 127.1 168.9 77.6 88.5 81.2 7.9 162.0 30.4 MbCO 1MBC' C 2.19 130.3 141.4 1.6 86.3 89.2 39.4 17.2 162.4 30.4 D 119.9 59.6 A 2MBY 2.27 1.8 76.9 85.8 78.2 128.0 145.8 46.0 162.3 30.8 B 135.2 42.3 85.9 83.5 2MGK 2.27 126.6 168.7 2.0 76.8 17.8 162.2 31.0 2MGA 2.2 2.22 128.8 86.8 88.6 77.3 159.3 20.2 162.3 30.4 2.26 131.0 84.8 87.8 2MGC 1.9 76.8 30.7 155.8 24.8 162.5 2MGF 1.8 2.20 122.0 86.8 82.0 77.0 161.8 31.2 171.0 12.3 a Notes and references: lHCO and 2HC0 represent structures derived from the same human HbCO data set via two different refinement protocols;58 lBBB is a low-salt human HbCO lCOH has Co substituted for Fe in the &hemes of human HbC0;60 lNIH has Ni substituted for Fe in the a-hemes of human HbC0.6' lSDH is arcid clapn HbCO$* lHBG is marine bloodwarm HbC0;63 lPBX is antarctic fish HbCO;Mand lECO is erythrocruorin HbC0.b5 lHBG and lECO are both monomeric hemoglobins; lHBG has Leu as the distal residue. All MbCOs are sperm whale myoglobins: 1MBC7and 2MB5I9 are wild type, the latter a neutron diffraction structure. The structures 2MG? are recombinant mutants,I6each with their initiator Met residue intact. 2MGK has Asp122 mistakenly replaced by Asn122; 2MGA, 2MGC, and 2MGF each have an additional mutation at the distal site of His64 to Gly, Leu, and Gln, respectively. Orientation of the Fe-C bond relative to the mean plane through the four porphyrin nitrogen atoms. An angle 90" lies on the heme normal. Orientation 9 of the CO bond relative to the normal to the mean plane through the four porphyrin nitrogens. An angle of 0" lies on the normal. Bend of the Fe-C-0 unit in the opposite sense to the other geometries considered. e Two carbonyl 0 positions were resolved: C, with occupancy 78%, and D, with occupancy 22%. f Two carbonyl 0 positions were resolved: A, with occupancy 58.2%. and B, with occupancy 41.8%. normal, and thus the distal side chain can adopt its X-ray orientation because it is not sterically strained. In contrast, our results suggest that the X-ray Fe-C-0 conformation is, in fact, correct. That is, the CO ligand is significantly distorted; however the distal side chain is not in a strained position because the proximal residue causes the ligand to bind off-axis. The position of the distal side chain in solution can then be determined by a weak attractive electrostatic interaction with the carbonyl group. That is, our results support the recent electrostatic reinterpretation of the distal-CO interaction and reconcile the largely geometry-innocent influence of the distal residue over C - 0 binding, as inferred from studies of distal mutant M ~ C O S , ' ~with . ' ~ the distorted Fe-C-0 geometries observed in all X-ray structures to date (see Table 3). Differences between the orientation of the distal side chain observed in the neutron crystal structure and those implied by the IR spectra could then arise from differences in the solvent environment local to the distal residue in each phase. A hydrogenbonded solvent network, anchored by an ammonium counterion held between the exposed propionic side chains of the porphyrin IX macrocycle, is well resolved in the crystal structure; such a network is not expected under the solution phase conditions of the IR experiment and is not found in MD simulations.'I If the orientation of the proximal residue is responsible for the distorted CO orientation in the protein, then the strain energies implied by the large Fe-C-0 distortion found in the X-ray structure are mainly delivered by the protein tertiary structure via the proximal residue and not by the mobile distal side chain. This is intuitively reasonable since the proximal

side chain, unlike the distal side chain, has small B factors, similar to those of the backbone atoms, and a geometry relative to the porphyrin ring that is well conserved within experimental error among the various crystal structures available. If the orientation of the proximal side chain is the key structural factor in the functioning of the heme pocket, then we expect it to be faithfully reproduced across a wide range of globin proteins. The orientations of the proximal residue relative to the heme for the MbCO and HbCO structures available in the Protein Databank are shown in Table 3, together with the CO orientation in each structure. The CO orientations in equivalent model globin heme groups lacking a distal side chain have been determined at the SCF level and are also shown. Table 3 shows that the proximal orientation reported in all the PDB structures, with the exception of one, anomalous, proximal geometry (the a subunit of the lHCO HbCO structure), leads to a distorted CO orientation at the SCF level even in the absence ofa distal side chain. That is, the proximal residue, and not the distal residue, is causing the distortion of the Fe-C-0 group in all the HbCO and MbCO X-ray structures reported to date. Although including the distal residue in the "proximal & distal" calculation has only a limited effect on the distortion of the CO ligand from the heme normal, it does significantly change the Fe-C bond length by 0.12 A, giving a final value similar to the X-ray structure. It should be noted that the distances derived in the X-ray structures are dependent on the constraints used in the data analysis and that those found in the calculations here will be affected by the frozen Rc-0 value. However, this result suggests that the distal residue may also

Role of Proximal Residue in Globin Proteins

be responsible for weakening the binding of the CO ligand. This, however, is in contrast with the solution phase spectroscopic results which show that the interaction of the distal histidine side chain with the CO ligand leads to a shift to lower vco frequencies, that is, a stronger Fe-C bond. Thus, the neutron diffraction structure, on which these calculations have been based, may differ from that of the physiologically active protein. It is interesting to note that when the distal-CO interaction is broken by photodissociation of the CO ligand in the crystal phase, the initial motion of the distal side chain is away from the binding site.45 This is consistent with an attractive distalCO interaction in MbCO; if the distal side chain were sterically strained in the MbCO structure, its initial motion on the relief of this strain, i.e. after CO photodissociation, would be toward the binding site. If the distal-CO interaction in the physiologically active protein is taken to be attractive, then the final orientation adopted by the CO ligand is likely to be in the direction of the distal side chain; such an orientation is the major conformation found in the IMBC X-ray structure. The off-axis nature of the CO orientation, however, has been made possible by the influence of the proximal ligand; without this effect, the CO ligand would bind strongly, perpendicular to the heme plane. It appears that the proximal residue may be responsible for differentiating between the physiological and poisoning ligands by forcing the CO ligand to bind away from its preferred position on the heme normal; the role of the distal residue may be to stabilize the binding of a ligand at, and control access to, the binding site. Although intuitively pleasing, the implications of these results need to be discussed in the light of the wide range of studies that have been made concerning globin proteins and their model compounds. Distal Mutant Mbs. In their investigations of the distalCO interaction in sperm whale MbCO, Quillin et al. have determined the X-ray structures of three distal mutants, His64Gly (PDB file 2MGA), His64Leu (2MGC), and His64Gln (2MGF), as well as their recombinant wild type (2MGK).I6 They report that there is no experimentally significant difference in the FeC - 0 geometry of any of their structures; that is, the CO geometry is significantly distorted even in the His64Gly mutant (see Table 3). Thus, the distal residue is not the primary factor in determining the CO binding geometry, although it does have a significant effect on the vibrational frequencies of the FeC-0 nit.'*,'^,^^ They therefore suggested that the distal-CO interaction is largely electrostatic in nature,I4 although this implies that it is attractive and stabilizing in the wild type and thus not responsible for the significant changes in ligand affinity for the globin heme relative to the free heme. Our results suggest that the missing "destabilizing factor" in this electrostatic reinterpretation of the distal-CO interaction is the distorted orientation of the proximal side chain; this would also account for the distorted Fe-C-0 geometries they observe in all their mutant structures, since the binding of the proximal residue is distorted in all their mutants. Proximal Mutant Mbs. Barrick and co-workers have recently engineered a mutant Mb which has the proximal side chain separated from the protein tertiary ~ t r u c t u r e . ~They ~.~~ achieve this by site-directed mutation of the proximal residue His93 to Gly and then expression of the protein in Escherichia coli using a medium containing imidazole (Im). Their mutant then has a backbone structure similar to the wild type, but with the free ligand imidazole occupying the proximal binding site. They find that this alteration allows a significant shift of the imidazole ring toward the Fe atom and an increase in the rebinding rate of CO. By releasing the proximal side chain from the protein tertiary structure, Barrick et al. not only allow

J. Phys. Chem., Vol. 99, No. 33, 1995 12683 it to move toward the binding site but also inevitably change the electronic nature of the proximal ligand. However, this is not expected to account for the large geometry change adopted by the imidazole ligand upon liberation; it is more likely to arise from the removal of a restrictive linkage to the protein architecture. This mutant promises to be an excellent system to study the effect of the proximal ligand on the binding of a sixth ligand to the Mb binding site since the nature of the proximal ligand can be readily varied over a range of substituted imidazole rings with the distal environment maintained in situ. In particular, we would expect the X-ray structure of their His93Gly(Im) MbCO mutant to have a more nearly perpendicular Fe-C-0 geometry, although hydrogen bonding between residues on the proximal side of the heme pocket and the proximal imidazole ring may lead to some distortion of the axial ligand binding geometries. Spectroscopic Determination of the CO Orientation in MbCO. The average orientation of the CO dipole relative to the heme normal has been determined for wild type MbCO in solution by circular dichroism photoselection spectroscopy at room temperature26and at 10 K.27 Three average orientations of the dipole have been determined, one for each of the three vco peaks observed in the IR spectrum. The same technique has been applied to HbC0,26348various His64 and Val68 mutants of MbC0,'3,49and protoheme.50 These studies determine the orientation of the CO dipole relative to the heme normal as ranging from close to 20" in protoheme, HbCO, and the & state of wild type MbCO to around 30" for the A1.2 and A3 states of wild type MbCO; the MbCO mutants show a similar range of orientations. The results of this technique are not straightforward to interpret: they are not directly equivalent to the X-ray geometries but also reflect the dynamics of the prosthetic group. Consequently, even protoheme shows a 20" orientation of the CO dipole due to the mobility of the CO ligand, a distribution of ligand orientations that would not be determined in an X-ray structure. Henry has rationalized the apparent discrepancy between the results derived using this method and those expected from the X-ray Ivanov et al. have recently determined the orientation of the CO ligand in MbCO crystals by linear dichroism.44This method does not suffer the same sensitivity to a distribution of ligand orientations since the orientation of the CO dipole can be directly measured relative to the crystallographic axes; therefore, it was expected to be directly comparable with an X-ray structure determined in the same space group. They report that the three IR substates have CO orientations lying within 4" of each other, all less than 10" from the heme normal. This is in conflict with the orientations determined in the X-ray structures, which are confirmed as reasonable in this study. An implicit assumption in the analysis of the dichroism data is that the CO dipole is orientated parallel to the CO bond vector. While this is true for CO bound perpendicular to the heme plane, tilting and, in particular, bending of the Fe-C-0 unit will remove the cylindrical symmetry in the valence electron density that is responsible for the CO dipole. For example, in a bent FeC - 0 conformation, the n electron density will be enhanced in the "crook" of the Fe-C-0 ''arm''; this will result in an (Fe)C-0 dipole orientated closer to the heme normal than the CO bond vector. This may account for some of the discrepancy between the linear dichroism result and the orientation of CO in the X-ray structures, both of which were determined in the same crystalline phase and space group: the linear dichroism technique is sensitive to the valence Fe-C-0 electron density, whereas X-rays probe the total electron density. Model Compounds. Model heme compounds have been extensively studied in an effort to reproduce the properties of

12684 J. Phys. Chem., Vol. 99, No. 33, 1995 globin proteins and thereby understand how they achieve control over their binding sites. Much effort has concerned compounds with restraining straps on the distal side of the binding site to investigate the influence of steric constraints on the CO ligand. Although the original studies were supportive of a steric interpretation of the structure-function relationship in the protein,8 more recent work, including X-ray ~ t u d i e s , ~has ’,~~ suggested that the interaction between the strap and the ligand in these model compounds is electrostatic.I8 Our results suggest that the design of model compounds with strained proximal ligands is required to fully mimic the heme prosthetic unit; to date, few studies of such systems have been made. Traylor and co-workers have shown that straining the linkage between the proximal ligand and the macrocycle in “tailed” porphyrins reduces the binding affinities of both 0 2 and CO, each to differing extents;53however, the introduction of excessive strain causes a change in the CO dissociation mechanism, with the proximal ligand leaving before the carbonyl group, which possibly limits the usefulness of such model compounds. Yu and co-workers have shown that binding the proximal ligand 1,2-MezImto “picket fence” porphyrins leads to an increase in the Y F ~ - C frequency relative to that when N-MeIm is bound, presumably due to the increased steric bulk of the 2-methyl group “straining” the binding of the disubstituted ligand.54 Further studies are required before a functional understanding of the trends observed in these model compounds can be achieved. Ray et al. have recently investigated the conformations of the Fe-C-0 unit observed in globin proteins and their model compounds on the basis of energy analyses of an isolated FeC-0 group.’8 They concluded that the distortions in the X-ray structures were much too high in energy to be possible in the protein environment and suggested that the true orientation of the CO ligand should lie closer to the heme normal. Our results suggest that the distortions of the Fe-C-0 unit seen in the X-ray structure are, in fact, not as high in energy as expected and that the ligand geometries reported in the X-ray structures are reasonable. These distortions may become possible in the protein because the proximal ligand is not in its preferred binding orientation, in particular, the Fe-N, bond is stretched, which affects both the geometry and strength of CO binding to the heme binding site. Normal mode analyses of the Fe-CO unit in MbCO and its model compounds are complicated by uncertainty over what interactions are determining the frequencies observed. Both electrostatic and geometric factors can lead to shifts in frequency, and uncertainties in the assignment of the spectra and over what geometries and force constants to choose to represent the heme prosthetic unit make an unambiguous analysis of the trends observed in the experimental frequencies difficult to achieve.’8.55-56Our results suggest that it will be important to include the influence of a proximal ligand bound in a distorted orientation, and its consequent effect on the CO geometry and strength of the Fe-C-0 interaction, in the analysis. Molecular Dynamics Simulations. Henry has found that the CO orientations in a 2 ns molecular dynamics trajectory of wild type MbCO in vacuu are distributed around an axis close to the heme normal;24 Jewsbuxy and Kitagawa have found similar results for shorter simulations of wild type and distal mutant MbCOs with partially solvated heme pockets.‘’-23 Their results demonstrate that there is little systematic steric distortion of the Fe-C-0 geometry by the distal side chain; our results suggest that their failure to reproduce the Fe-C-0 geometries observed in the X-ray structures could be an artifact of their MD force fields not allowing for a direct coupling of the heme axial ligand orientations through the Fe d orbitals.

Jewsbury et al.

Functional Role of the Heme Pocket. If, as these results suggest, the distortion of the CO ligand in the heme pocket is caused by the orientation of the proximal ligand, it remains to explain why oxygen is not distorted to a similar extent. An ab initio study of an oxygen-bound model globin prosthetic unit is significantly more complicated than the CO-bound case considered here on account of the more complex nature of the Fe-02 i n t e r a ~ t i o n .It~ ~is worth noting, however, the different nature of the Fe-02 and Fe-CO interactions: electrons are donated by the oxygen ligand from an occupied n,rather than u, orbital and only one z* orbital is available for back-bonding. The Fe-02 interaction is therefore not expected to be as demandingly directional as that of Fe-CO, and its bending coordinate perpendicular to the plane of its back-bonding interaction is expected to be considerably softer than that of CO. It is therefore possible that CO, with its strong preference for a perpendicular geometry, is more affected by changes in the ligand binding geometry than 0 2 . Experimental studies have repeatedly shown that steric and electrostatic influences in the heme prosthetic group can affect the binding of CO and 02 in different ways. Summary Early interpretations of the structure-function relationship in globin proteins emphasized the role of the distal histidine residue in destabilizingthe binding of CO by sterically crowding its preferred binding position perpendicular to the heme plane. This interpretation was originally proposed since the other interactions expected in a protein environment would all be attractive and stabilizing and therefore could not account for the large change in the relative affinities of the two ligands for the heme located inside the protein. The large strain energies implied by the large distomons of the Fe-C-0 unit in the X-ray structures, which the distal side chain must deliver to the binding site in this interpretation, seem unlikely for the side chain of a surface residue. It has been suggested that a hydrogen-bonding network stabilizes the distal side chain orientation; such a stabilization is not expected under physiological conditions, however, and is inconsistent with the X-ray B factors and the results of MD simulations. Recent experimental and theoretical studies of wild type MbCO and some of its distal mutants have shown that the distal-CO interaction in myoglobin is largely electrostatic in nature. The large distortions of the Fe-C-0 unit observed in the X-ray structures have also been questioned, and various studies have proposed that the true CO orientation should be close to, or on, the heme normal. This reinterpretation of the structure-function relationship in the globin proteins is, however, incomplete since the shift in the vibrational frequencies of wild type MbCO relative to the His64Gly mutant implies that the wild type distal histidine is stabilizing the CO ligand. That is, there is no “destabilizing factor” in the current reinterpretation of the structure-function relationship in the globin proteins to account for the large change in the relative affinities of CO and 0 2 for the heme when it is located in the protein. Our ab initio results for the CO orientation in a series of model globin and free heme groups reproduce the Fe-C-0 geometries found in the X-ray structures on which the calculations were based and suggest that the distorted orientation of the proximal residue is responsible for delivering the strain to the CO binding conformation. In contrast, the distal side chain has only a small effect on the distortion of the carbonyl ligand from the heme normal. The orientation of the proximal residue relative to the heme plane is known to be an important reaction coordinate in the R T transition of hemoglobin, with the

-

Role of Proximal Residue in Globin Proteins proximal residue significantly more tilted in the T (deoxy, tense) state than the R (ligand bound, relaxed) state. Our results suggest that even in the R state the proximal residue is, in fact, still strained sufficiently to discourage the binding of CO. While the orientation of the proximal residue in the heme pocket allows for the distorted binding of CO, the final geometry adopted by the ligand is probably determined by weaker interactions with nearby residues in the distal side of the pocket. The results of this study suggest an important factor that is missing from the recent reinterpretation of the structurefunction relationship in the globin family of proteins, although further experimental and theoretical studies are required to test the validity of our proposal. In this respect, the spectroscopic and structural properties of the new proximal mutant Mb developed by Barrick and co-workers, His93Gly(Im), present an exciting opportunity to determine the influence of the proximal residue with the distal pocket structure maintained in situ.

Acknowledgment. These calculations were run at the Computer Centres of IMS, PERI, and Nara University. The authors thank Dr. H. Kashiwagi, of Kyushu Institute of Technology, Japan, for stimulating discussions. P.J. thanks the Japan Society for the Promotion of Science for a postdoctoral Fellowship under Grant 04NP0301 awarded to T.K. by the Japanese Ministry of Education, Science and Culture for the new research program “Intelligent Molecular Systems with Controlled Functionality”. References and Notes (1) Springer, B. A,; Silgar, S. G.; Olson, J. S.; Phillips, G. N., Jr. Chem. Rev. 1994, 94, 699-714. (2) Yu, N.-T.; Kerr, E. A. Vibrational Modes of Coordinated CO, CN-, 0 2 and NO. In Biological Applications of Raman Spectroscopy. Vol. 3: Resonance Raman Spectra of Heme and Metalloproteins; Spiro, T. G., Ed.; Wiley-Interscience: New York, 1988; pp 39-95. (3) Phillips, S. E. V.; Schoenbom, B. P. Nature 1981,292, 81-82. (4) Stryer, L. Biochemistry; W. H. Freeman and Co.: New York, 1988. (5) Abola, E. E.; Bemstein, F. C.; Bryant, S. H.; Koetzle, T. F.; Weng, J. Protein Data Bank In Crystallographic Databases- Information Content, Software Systems, Scientific Applications; Allen, F. H., Bergerhoff, G., Sievers, R., Eds.; Data Commission of the Int’l Union of Crystallography: BonnKambridgeKhester, 1987; pp 107-132. (6) Bemstein, F. C.; Koetzle, T. F.; Williams, G. J. B.; Meyer, E. F., Jr.; Brice, M. D.; Rodgers, J. R.; Kennard, 0.; Shimanouchi, T.; Tasumi, M. J. Mol. Biol. 1977, 112, 535-542. (7) Kuriyan, J.; Wilz, S.; Karplus, M.; Petsko, G. A. J. Mol. Biol. 1986, 192, 133-154. (8) Collman, J. P.; Brauman, J. I.; Halbert, T. R.; Suslick, K. S. Proc. Natl. Acad. Sci. U.S.A. 1976, 73, 3333-3337. (9) Case, D. A.; Karplus, M. J. Mol. Biol. 1978, 123, 697-701. (10) Kuczera, K.; Kuriyan, J.; Karplus, M. J. Mol. Biol. 1990,213, 35 1373. (11) Jewsbury, P.; Kitagawa, T. Biophys. J. 1994,67, 2236-2250. (12) Balasubramanian, S.; Lambright, D. G.; Boxer, S. G. Proc. Natl. Acad. Sci. U.S.A. 1993, 90, 4718-4722. (13) Braunstein, D. P.; Chu, K.; Egeberg, K. D.; Frauenfelder, H.; Mourant, J. D.; Nienhaus, G. U.; Ormos, P.; Silgar, S. G.; Springer, B. A,; Young, R. D. Biophys. J. 1993, 65, 2447-2454. (14) Li, T.; Quillin, M. L.; Phillips, G. N., Jr.; Olson, J. S. Biochemistry 1994, 33, 1433- 1446. (15) Sakan, Y.; Ogura, T.; Kitagawa, T.; Fraunfelter, F. A,; Mattera, R.; Ikeda-Saito, M. Biochemistry 1993,32, 5815-5824. (16) Quillin, M. L.; Arduni, R. M.; Olson, J. S.; Phillips, G.N., Jr. J. Mol. Biol. 1993, 234, 140-155. (17) Oldfield, E.; Guo, K.; Augspurger, J. D.; Dykstra, C. E. J. Am. Chem. SOC.1991, 113, 7357-7541. (18) Ray, G. B.; Li, X.-Y.; Ibers, J. A,; Sessler, J. L.; Spiro, T. G. J. Am. Chem. SOC.1994, 116, 162-176. (19) Cheng, X.; Schoenbom, B. P. J. Mol. Biol. 1991,220, 381-399. (20) Morikis, D.; Champion, P. M.; Springer, B. A.; Silgar, S. G. Biochemistry 1989, 28, 4791 -4800. (21) Ringe, D.; Petsko, G. A,; Ken, D.; Ortiz de Montellano, P. R. Biochemistry 1984, 23, 2-4. (22) Johnson, K. A.; Olson, J. S.; Phillips, G. N., Jr. J. Mol. Biol. 1989, 207, 459-463. (23) Jewsbury, P.; Kitagawa, T. Biophys. J. 1995, 68, 1283-1294. ~~

J. Phys. Chem., Vol. 99, No. 33, 1995 12685 (24) Henry, E. R. Biophys. J. 1993, 64, 869-885. (25) Osapay, K.; Theirault, Y.; Wright, P. E.; Case, D. A. J. Mol. Biol. 1994,244, 183-197. (26) Moore, J. H.; Hansen, P. A.; Hochstrasser, R. M. Proc. Natl Acad. Sci. U.S.A. 1988, 85, 5062-5066. (27) Ormos, P.; Brauenstein, D.; Frauenfelder, H.; Hong, M. K.; Lin, S.; Sauke, T. B.; Young, R. D. Proc. Natl. Acad. Sci. U.S.A. 1988, 85, 8492-8496. (28) Balasubramanian, S.; Lambright, D. G.; Simmons, J. H.; Gill, S. J.; Boxer, S. G. Biochemistry 1994, 33, 8355-8360. (29) Strich, A.; Veillard, A. Theor. Chim. Acta 1981, 60, 379-383. (30) Newton, J. E.; Hall, M. B. lnorg. Chem. 1984, 23, 4627-4632. (31) Hay, P. J.; Wadt, W. R. J. Chem. Phys. 1985, 82, 299-310. (32) Rappe, A. K.; Smedley, T. A.; Goddard, W. A,, 111. J. Phys. Chem. 1981,85, 2607-1611. (33) Dunning, T. H. J. Chem. Phys. 1970, 53, 2823-2834. (34) Hehre, W. J.; Stewart, R. F.; Pople, J. A. J. Chem. Phys. 1969, 51, 2657-2664. (35) Binkley, J. S.; Pople, J. A.; Hehre, W. J. J. Am. Chem. Soc. 1980, 102, 939-947. (36) Frisch, M. J.; Trucks, G. W.; Head-Gordon, M.; Gill, P. M. W.; Wong, M. W.; Foresman, J. B.; Johnson, B. G.; Schlegel, H. B.; Robb, M. A.; Replogle, E. S.; Gomperts, R.; Andres, J. L.; Raghavachari, K.; Binkley, J. S.; Gonzalez, C.; Martin, R. L.; Fox, D. J.; Defrees, D. J.; Baker, J.; Stewart, J. J. P.; Pople, J. A. Gaussian 92; Gaussian Inc.: Pittsburgh, PA, 1992. (37) Huber, K. P.; Herzberg, G. Molecular Spectra and Molecular Structure; Van Nostrand: New York, 1979. (38) Peng, S.-M.; Ibers, J. A. J. Am. Chem. SOC.1976,98, 8032-8036. (39) Beagley, B.; Schmidling, D. G. J. Mol. Struct. 1974, 22, 466468. (40) Krishnan, R.; Binkley, J. S.; Seeger, R.; Pople, J. A. J. Chem. Phys. 1980, 72, 650-654. (41) Sharp, K. A.; Honig, B. Annu. Rev. Biophys. Biophys. Chem. 1990, 19, 310-332. (42) Warshel, A.; Aqvist, J. Annu. Rev. Biophys. Biophys. Chem. 1991, 20, 267-298. (43) van Gunsteren, W. F.; Luque, F. J.; Timms, D.; Torda, A. E. Annu. Rev. Biophys. Biomol. Struct. 1994, 23, 847-863. (44) Ivanov, D.; Sage, J. T.; Keim, M.; Powell, J. R.; Asher, S. A.; Champion, P. M. J. Am. Chem. SOC. 1994, 116, 4139-4140. (45) Schlichting, I.; Berendzen, J.; Phillips, G. N., Jr.; Sweet, R. M. Nature 1994, 371, 808-812. (46) DePillis, G. D.; Decatur, S. M.; Bamck, D.; Boxer, S. G. J. Am. Chem. SOC.1994, 116, 6981-6982. (47) Bamck, D. Biochemistry 1994,33, 6546-6554. (48) Locke, B.; Lian, T.; Hochstrasser, R. M. Chem. Phys. 1991, 158, 409-419. (49) Braunstein, D.; Ansari, A.; Berendzen, J.; Cowen, B. R.; Egeberg, K. D.; Frauenfelder, H.; Hong, M. K.; Ormos, P.; Sauke, T. B.; Scholl, R.; Schulte, A.; Silgar, S. G.; Springer, B. A.; Steinbach, P. J.; Young, R. D. Proc. Natl. Acad. Sci. U.S.A. 1988, 85, 8497-8501. (50) Hansen, P. A.; Moore, J. N.; Hochstrasser, R. M. Chem. Phys. 1989, 131, 49-62. (51) Kim, K.; Ibers, J. A. J. Am. Chem. Soc. 1991, 113, 6077-6081. (52) Kim, K.; Fettinger, J.; Sessler, J. L.; Cyr, M.; Hugdahl, J.; Collman, J. P.; Ibers, J. A. J. Am. Chem. SOC.1989, 111, 403-405. (53) Geibel, J.; Cannon, J.; Campbell, D.; Traylor, T. G. J. Am. Chem. SOC. 1978, 100, 3575-3585. (54) Kerr, E. A.; Mackin, H. C.; Yu, N.-T. Biochemistry 1983,22,43734379. (55) Hirota, S.; Ogura, T.; Shinzawa-Itoh, K.; Yoshikawa, S.; Nagai, M.; Kitagawa, T. J. Phys. Chem. 1994, 98, 6652-6660. (56) Li, X.-Y.; Spiro, T. G. J. Am. Chem. SOC. 1988, 6024-6033. (57) Bytheway, I.; Hall, M. B. Chem. Rev. 1994,94, 639-658. (58) Baldwin, J. M. J. Mol. Biol. 1980, 136, 103-128. (59) Silva, M. M.; Rogers, P. H.; Amone, A. J. Biol. Chem. 1992,267, 17248-17257. (60) Luisi, B.; Shibayama, N. J. Mol. Biol. 1989, 206, 723-736. (61) Luisi, B.; Liddington, B.; Fermi, G.; Shibayama, N. J. Mol. Biol. 1990, 214, 7-14. (62) Royer, W. E., Jr.; Hendrickson, W. A.; Chiancone, E. J. Biol. Chem. 1989,264, 21052-21061. (63) Arents, G. A.; Love, W. E. J. Mol. Biol. 1989, 210, 149-161. (64) Camardella, L.; Caruso, C.; DAvino, R.; di Prisco, G.; Rutigliano, B.; Tambumni, M.; Fermi, G.; Perutz, M. F. J. Mol. Biol. 1992,224, 449460. (65) Steigemann, W.; Weber, E. J. Mol. Biol. 1979, 127, 309-338. JP950389Q