PROTEINS: Structure, Function, and Genetics 11:l-12 (1991)
The Electrostatic Potential of Escherichia coZi Dihydrofolate Reductase Jurgen Bajorath,' David H. Kitson,' Joseph Kraut: and Arnold T. Hagler' 'Biosym Technologies, Inc., 10065 Barnes Canyon Road, San Diego, California 92121, and 'Department of Chemistry, University of California at San Diego, La Jolla, California 92093
ABSTRACT Escherichia coli dihydrofolate reductase (DHFR) carries a net charge of -10 electrons yet it binds ligands with net charges of -4 (NADPH) and -2 (folate or dihydrofolate). Evaluation and analysis of the electrostatic potential of the enzyme give insight as to how this is accomplished. The results show that the enzyme is covered by an overall negative potential (as expected)except for the ligand binding sites, which are located inside "pockets" of positive potential that enable the enzyme to bind the negatively charged ligands. The electrostatic potential can be related to the asymmetric distribution of charged residues in the enzyme. The asymmetric charge distribution, along with the dielectric boundary that occurs at the solvent-protein interface, is analogous to the situation occurring in superoxide dismutase. Thus DHFR is another case where the shape of the active site focuses electric fields out into solution. The positive electrostatic potential at the entrance of the ligand binding site in E. coli DHFR is shown to be a direct consequence of the presence of three positively charged residues at positions 32,52, and 57-residues which have also been shown recently to contribute significantly to electronic polarization of the ligand folate. The latter has been postulated to be involved in the catalytic process. A similar structural motif of three positively charged amino acids that gives rise to a positive potential at the entrance to the active site is also found in DHFR from chicken liver, and is suggested to be a common feature in DHFRs from many species. It is noted that, although the net charges of DHFRs from different species vary from + 3 to -10, the enzymes are able to bind the same negatively charged ligands, and perform the same catalytic function. Key words: electrostatics,enzyme-substrateinteraction, solvent screening, active site potential, structure-function relationship 0 1991 WILEY-LISS, INC.
INTRODUCTION Dihydrofolate reductase (DHFR) catalyzes the NADPH- (reduced nicotinamide adenine dinucleotide phosphate) dependent reduction of folate to dihydrofolate and of dihydrofolate to tetrahydrofolate in bacterial and vertebrate species.' The enzyme is one of the most intensively studied proteins, and has been investigated, both experimentally and theoretically, from many different points of A variety of X-ray crystal structures of DHFRs, from different species and in complexes with various inhibitor^,"-'^ have been solved, and these provide the basis for much of the current research on the enzyme. In a recent study the changes in the electron density of the substrate folate when it binds to DHFR were investigated.16 The binding of folate, which carries a net charge of -2, and of NADPH, with a net charge of -4, to E . coli DHFR is somewhat intriguing since the enzyme also has a net negative charge, of -10. Binding of folate and NADPH to the like-charged DHFR is a more drastic example of the much studied effect seen in superoxide dismutase, where a 4-fold negatively charged protein binds a ligand with a net charge of -1 (17, and references therein). In addition, calculations of the effects of binding on the electronic structure of folate have shown that the ligand undergoes a significant polarization upon binding that leads to dramatic changes in the electronic structure. This electronic rearrangement may play a significant role in the enzymatic reaction.16 To gain some qualitative understanding of the apparently noncomplementary electrostatics of the ligand and of the protein to which it binds, and t o further describe the field postulated to polarize the substrate, it was of interest to study the electrostatic properties of the enzyme in a dielectric environment. The effect of the binding of the ligands on the electrostatic potential of the enzyme is also examined, and the origin of the active
Received August 13, 1990; revision accepted November 15, 1990. Address reprint requests to Arnold T. Hagler, Biosym Technologies, Inc., 10065 Barnes Canyon Road, San Diego, CA 92121. Jurgen Bajorath's current address: Bristol-Myers Squibb, Pharmaceutical Research InstituteSeattle, 3005 1st Avenue, Seattle, WA 98121.
2
J. BAJORATH ET AL.
Fig. 1. Stereo representation of the ternary complex of f.coli DHFR with folate and NADPH. The DHFR backbone is shown as a yellow ribbon; the substrate (folate) and the cofactor (NADPH) in
"ball-and-stick' representations, color coded according to atom type (C, green; N, blue; 0, red). This, and the other color figures, were produced with the Insight I1 software package.35
site potential is analyzed, based on the structural details of the protein around the ligand binding site. The long-range electrostatic potential of DHFR is investigated and implications for ligand binding are discussed. Finally, the electrostatics of DHFRs from different species are compared in order to evaluate the role of the net charge of the enzyme in ligand binding and reactivity. This is done by comparing the structures and electrostatic properties of E . coli and chicken liver DHFRs, together with the sequences of the enzymes from a variety of other bacterial and vertebrate species.
with only the waters allowed to move, was run to equilibrate the hydration shell. Steric overlaps and other sources of significant strain (arising from inaccuracies in the X-ray structure, or artifacts introduced by the hydration procedure) were then relaxed by energy minimization until the maximum derivative of the energy with respect to the atomic positions was less than 0.5 kcallmollA.21.22During the minimization, a tethering constant of 100 kcall mo1/A2 was applied to constrain the nonhydrogen atoms of the protein and the ligands to positions close to those of the X-ray structure.21 The resulting
METHODS The starting point for the study was the refined crystal structure of the ternary complex of E . coli DHFR with NADP' and fo1ate.l' This system was chosen as it is the only experimentally determined DHFR structure that includes a bound substrate. Thus, the structural basis for the calculation is well grounded in experiment. The NADP+ was converted to NADPH in order to mimic the active form of the enzyme. Figure 1 shows a representation of this ternary complex. Folate binds to the enzyme in a hydrophobic cleft with the glutamate moiety, which carries the two negatively charged carboxyl groups, located toward the protein surface. Before carrying out the electrostatics calculations, a molecular modeling protocol was carried out to prepare the starting structure. Hydrogen atoms were added to the protein and substrates in standard geometries and water molecules included to give a complete 5 A hydration shell around the ligand binding sites and a 3 A shell around the remainder of the protein. As suggested by recent enzyme kinetic studies the side chain of Asp-27 in the active site was modeled as uncharged (i.e., protonated). A short molecular dynamics simulation, 19720
Fig. 2. The electrostatic potential of E. coli DHFR in solution. The potential was calculated for the protein alone, without the bound ligands, which are, however, shown in these figures to indicate their positions relative to the potential. (a) The ternary complex of DHFR, folate, and NADPH, shown as a reference for b, in which the orientation and scale of the molecule are the same as is used in a. The enzyme is represented by a green ribbon, while folate and NADPH are represented by "ball-and-stick' models. (b) A close-up view of the potential of DHFR. Contours of negative potential around the protein are shown at a level of -0.6 kcalimol (red); the positive contour level is +0.6 kcalimol (blue) [a contour level of -0.6 kcalimol means that an atom of one unit of positive charge, sitting at a point on the contour, would have an interaction energy of -0.6 kcalimol (a favorable interaction) with the potential at that point]. Both the substrate folate (with a net charge of -2) and the cofactor NADPH (with a net charge of -4) are seen to bind in "pockets" of positive potential that are embedded within the largely negative potential around the 10-fold negatively charged protein. (c) The long-range electrostatic potential around E. coli DHFR. The positive potentials around the ligand binding sites extend far out (-30 A from the active site) into the solvent environment. DHFR is able to bind dihydrofolate in the polyglutamate form (as a polyanion), and the reason for this may be appreciated from this figure, since the polyglutamate tail can be accommodated within the large positive potential region around the entrance to the active site. The view in this figure is rotated slightly from that in a and b. The potentials in b and c were calculated using the Poisson-Boltzmann meth~d,'~-'' assuming a dielectric constant of 2 for the interior of the protein, and 80 for the exterior, an ionic strength of 20 mM (the strength used to prepare the complex for crystallization2') and a temperature of 300 K.
ELECTROSTATICS OF DHFR
Fig. 2. Legend appears on page 2.
3
4
J. BAJORATH ET AL.
structure was then used for the electrostatic potential calculations. The electrostatic potential calculations were carried out using the Finite Difference Poisson-Boltzmann m e t h ~ d ? ~ -and ’ ~ for this purpose the explicit water molecules in the hydration shell were omitted. The Poisson-Boltzmann method takes the different dielectric constants of the protein interior and the surrounding solvent into account (without the need to treat solvent molecules explicitly), thereby allowing for the shielding effects of the solvent. (It is, therefore, a more realistic technique for studying the electrostatic potential of proteins than the simple Coulombic formulation of charge-charge interactions applied to a protein in vacuo.) The dielectric constant was set to 2 for the protein interior and to 80 for the surrounding ~ o l v e n t . ~An ~ ~ ionic ’~ strength of 20 mM was used (the strength used experimentally to prepare the complex for crystallization”) and the temperature was set to 300 K. In order to analyze the electrostatics of the enzyme, several different calculations were carried out on the wild-type enzyme, the ternary complex, and proposed enzyme mutants. RESULTS AND DISCUSSION The Electrostatic Potential of E. coli DHFR in Solution As would be expected for a protein with a net charge of -10, DHFR is largely covered by a negative potential (Fig. 2-the ligands were not included in the calculation of the potential but are shown in the figures to indicate their positions relative to the potential). Despite the overall negative potential, the ligand binding sites exhibit “pockets”of positive potential that extend into the solvent. As Figure 2b shows, folate binds in the center of a region of positive potential a t the entrance to the active site. The adenosine phosphate group of NADPH also binds directly into the center of an area of positive potential. This is consistent with the fact that DHFR binds NADPH, with its adenosine phosphate group, as a cofactor, and not NADH, in which the phosphate group is replaced by a hydroxyl. Thus, the regions of positive potential appear to enable the enzyme to bind the negatively charged ligands. The LongRange Effect of the Active Site Potential The positive ligand binding site potentials of DHFR extend out significantly into the solvent environment (Fig. 2c). Only a t about 30 A from the protein surface is the enzyme completely surrounded by a (very weak) negative potential. A potential gradient (from negative to positive potential) exists 30 A away from the protein. To compare the potential of DHFR in vacuo (the state in which many protein simulations are carried out) with that of the enzyme in solution, we have
recalculated the potential with the dielectric constant of both protein and solvent set to 1 and with the ionic strength reduced to 0.0 M. That is, there is no dielectric boundary between the protein and the “solvent,” and the calculation is now equivalent to a Coulomb’s Law calculation in vacuo. This is found to have a significant influence on both the magnitude and the shape of the potential. With a dielectric constant for the solvent of 80, the most negative potential found at a distance of 5 A from the protein is approximately -3 kcal/mol (Fig. 3a). This means that a single positive charge, for example a sodium ion, would have an energy of only -3 kcal/mol a t this position. In contrast, when the dielectric of the solvent is 1, the most negative potential 5 A from the protein is approximately -240 kcal/mol (Fig. 3b), so that the sodium ion would have an energy of -240 kcal/mol at this position. The 80-fold decrease in energy in the presence of solvent reflects the very large dielectric screening effect of water and the concomitant diminution of the effect of the charge of the protein. At a larger distance from the protein the screening effect of the solvent is even more dramatic. The maximum potential found 30 A from the protein is approximately -0.15 kcal/mol in the presence of solvent (Fig. 2c), whereas, in vacuo, the maximum potential would be more than -60 kcal/mol (Fig. 3c). Furthermore, as can be seen by comparing Figures 2c and 3c, the presence of the dielectric boundary in the case where the solvent dielectric is 80 changes the shape of the potential around the enzyme. In the absence of this dielectric boundary the net total charge of the protein becomes more significant in determining the electrostatic environ-
Fig. 3. The effect of solvent screening and the presence of the protein/solvent dielectric boundary on the electrostatic potential around E. coli DHFR. (a) With a dielectric constant for the solvent of 80 (and 2 for the protein interior), the most negative potential found at a distance of 5 A from the protein is approximately -3 kcallmol [contours are shown at levels of -3.0 kcal/mol (red) and +3.0kcal/mol (blue)]. (Note, this figure was calculated using the same conditions as Fig. 2; the figures differ in the contour levels shown.) In this stereo figure, the enzyme is represented by a yellow ribbon, and folate and NADPH by color-coded line representations. Atoms of negatively charged protein residues are indicated by red balls and those of positively charged residues by blue balls. The figure shows that the charged residues in the protein are asymmetrically arranged, with most of the positive residues being located close to the ligand binding sites, while the negative residues are more widely distributed. (b) As seen by comparison with Figure 2, this asymmetric arrangement of charged residues leads to a similar asymmetry in the potential of the enzyme. When the dielectric of the solvent (as well as the protein interior) is 1, the most negative potential 5 A from the protein becomes approximately -240 kcal/mol. In this figure, the negative contour is shown at the -240 kcal/mol level. (c) As expected, the net negative charge of the protein totally dominates the electrostatic environment of the protein in vacuo. Although there are still pockets of positive potential around the ligand binding sites, they are very much smaller than those present when the solvent dielectric is 80 (Fig. 2c). Note that the contours in this figure are at levels approximately 200 times higher than those of Figure 2c.
ELECTROSTATICS OF DHFR
Fig. 3. Legend appears on page 4.
5
6
J. BAJORATH ET AL.
ment around the protein. The positive pockets around the ligand binding sites shrink considerably in size (Fig. 3c), and the negative potential dominates and completely encloses the protein at a closer distance than in the presence of solvent (Fig. 2c). In Figure 3c it is seen that the active site of DHFR is surrounded by a negative potential contour while, in contrast, the active site potential changes from positive to (very weakly) negative only at about 30 A from the protein in the presence of solvent (Fig. 2c). The comparison illustrates the effects of the dielectric boundary under conditions where the solvent is treated properly (i.e.) as a high dielectric medium and with correct ionic strength). Thus, the effect of the proteidsolvent dielectric boundary, together with the spatial arrangement of protein charges, focuses electric fields of the protein out into solution. This is analogous to the effect seen for superoxide dismutase.17 It is also of interest to compare the electrostatic potential for the situation where the protein is surrounded by the aqueous medium with the approximation of uniform (protein and solvent) dielectric 80, omitting the effects of ionic strength. This is another approximation that would be attractive to invoke in molecular mechanics/dynamics simulations for its simplicity. Since Coulomb’s law is valid in a uniform medium this is equivalent to dividing the potentials in Figure 3c by 80. In this case, the active site would still be surrounded by a negative potential contour (see above, comparison of Figs. 2c and 3c). This again emphasizes the importance of the dielectric boundary.
Asymmetric Charge Distribution in E. coli DHFR Figure 3 illustrates that positively and negatively charged residues are asymmetrically distributed in the enzyme. Positively charged residues are concentrated in the ligand binding site regions (on the “left” side in this figure) whereas negatively charged residues dominate the “right” side of the enzyme. The charge distribution appears to bisect the enzyme. This anisotropy of protein charges correlates directly with the electrostatic potential of the enzyme shown in Figure 2. The Origin of the Positive Electrostatic Potential at the Active Site An understanding of the origin of the pockets of positive potential into which the negative moieties of the ligands bind can be gained by examining the structural details of the ligand binding sites. The glutamate moiety of folate is surrounded by three positively charged protein residues, Lys-32, Arg-52, and Arg-57, the location of which, a t the boundary of the hydrophobic protein cleft, would be expected to induce a positive electrostatic field in the region of the protein where the glutamate moiety binds. Ini-
tially, the possible role of these three positively charged residues in giving rise to a positive potential around the glutamate moiety of folate was investigated based on calculations using a simple, Coulombic model.16 To confirm this role, and to study the field induced by these residues, a PoissonBoltzmann calculation of the potential of the enzyme was carried out in which the charges on these three residues were neutralized. As Figure 4 shows, the pocket of positive potential around the glutamate moiety of folate disappears completely when these charges are removed, indicating that they are directly responsible for this positive region of the electrostatic potential and supporting the qualitative conclusions drawn from the simpler Coulombic model. Examination of the protein structure also suggested that the positive potential pocket around the adenosine phosphate group of NADPH might be due to the presence of three positively charged residues, Arg-44, Lys-76, and Arg-98, in the direct vicinity of the phosphate group. The charges on these residues were also removed in the calculation, and, as Figure 4 shows, this leads to elimination of the positive potential that surrounds the phosphate group of the cofactor. Thus, we may rationalize the binding of these highly negatively charged ligands to the negatively charged protein in aqueous solution in terms of the pockets of positive potential at both the entrance to the active site cleft and the
~
~~
Fig. 4. The positive potential around the entrance to the folate binding site disappears when the potential is calculated with the charges turned off on the three (normally) positive protein residues (Lys-32, Arg-52, and Arg-57) located around the glutamate moiety of folate (and also calculated without the bound ligands). Thus, these three positively charged residues are largely responsible for this pocket of positive potential in E. coli DHFR. As in Figure 2, the potential is superimposed on a representation of the ternary complex. (a) A reference figure to show the orientation of the ternary complex in b. (b) A close-up view of the potential around this “mutant” enzyme. The positive potential into which the glutamate moiety of folate binds has disappeared. Only a small residual potential remains near this region. This potential is due to two positively charged residues, Arg-33 and Lys-58, which are located close to the entrance to the active site of the enzyme. The side chains of these two residues, however, point away from the folate binding site in the crystal structure and, therefore, as can be seen from this figure, do not contribute directly to the positive active site potential around the glutamate moiety in folate (calculation of the potential of the enzyme with the charges on residues 33 and 58 set to zero confirms that this is the case-the pocket of positive potential is largely unchanged). In the calculation of the potential displayed in this figure the charges were also removed on the three positive residues (Arg-44, Lys-76, and Arg-98) that are close to the adenosine phosphate group of NADPH. Comparison with Figure 2b indicates that these residues are largely responsible for the positive potential around this group. (c). The long-range positive potentials around the entrances to the substrate and cofactor binding sites have completely disappeared. (This figure is shown in the same orientation as Fig. 2c). At a distance of 30 A from the active site, an approaching ligand would normally (i.e., for the protein with all charged residues included) begin to “feel” a positive potential (Fig. 2c). In the case of this “mutant” protein (with the charges on Lys-32, Arg-52, Arg-57, Arg-44, Lys-76, and Arg-98 deleted) the ligand would “feel” a weak negative potential (of - -0.06 kcal/mol), which becomes more negative as the distance from the protein decreases.
ELECTROSTATICS OF DHFR
Fig. 4. Legend appears on page 6.
7
8
J. BAJORATH ET AL.
cofactor binding region. Furthermore, these pockets of positive potential are shown to be due, in both clefts, to three basic residues. In the active site cleft, these are residues Lys-32, Arg-52, and Arg-57, while in the cofactor binding site these are residues Arg44,Lys-76, and Arg-98.
The Electrostatic Potential of the Ternary Complex The charge distribution of the ligands and the ligand binding sites of the enzyme are complementary to one another, as can be seen in Figure 5, which shows the potential calculated for the ternary complex. The positive “pockets” around the binding sites are significantly reduced when the ligands are bound to the enzyme. The net charge of the ternary complex that was crystallized (DHFR-folateNADP’) is - 15 [the net charge of the complex studied here (with NADPH) is -161. That a crystal of this highly negatively charged complex can form indicates that there must be a large aqueous distribution of positive counter ions around the highly negatively charged protein complexes within the crystal, that is able to balance the electrostatic interactions between them and stabilize the packing arrangement. This is clear when one considers the effects of the negative potential, even after modulation by solvent, in attracting positive counterions in the solvent. A double charged cation such as Ca2 or Mg2+ would experience a gain of energy of approximately 7 kcal/mol if it came within -5 A of the complex. An analogous situation has been observed in molecular dynamics simulations of the crystal of Streptomyces griseus Protease A. (which has an overall charge of +5) in which the solvent of crystallization was included explicitly.28 In the crystal lattice, the counterions are predicted to be concentrated in positions where they are surrounded by protein residues of opposite charge, stabilizing the crystal in the observed packing arrangement.” It is interesting to speculate as to whether this large negative charge plays a role in protein aggregation in E . coli analogous to that which occurs in the cells of vertebrates, where it has been suggested that proteins exists as aggregates rather than isolated in ~olution.~’ As seen from the binding of the ligands, the net charge of the protein does not necessarily determine the charge of ligands that will bind. Rather, complementary local potentials seem to be required. These observations are consistent with conclusions from a recent study on different t r y p ~ i n s .Here ~ ~ too, local potential effects were shown to determine the active site potential and to be conserved, despite considerably different overall net charges of the two enzymes, rat and cow trypsin, that were studied. +
Structural and Functional Comparison With Chicken Liver DHFR In the E . coli enzyme, comparison of Figures 2 and 4 shows that the presence of the positive potential around the entrance to the active site is a direct consequence of the structural arrangement of the three positively charged amino acids at the entrance to the cleft. This structural feature has also been shown to play a significant role in the electronic polarization of folate by the enzyme.16 Thus, the more proper treatment of solvent by the PoissonBoltzmann approximation yields a similar potential in this local region to that found by a simple Coulombic treatment, further supporting the interpretation of the underlying basis for the polarization.16 Examination of the structure of chicken liver DHFR reveals that a similar arrangement of positive charges is found, in this case consisting of residues Arg-36, Lys-68, and Arg-70. Superposition of these structural motifs in the E . coli and chicken liver enzymes shows that the positively charged side chains occupy very similar positions within the ligand binding sites (Fig. 6). Furthermore, PoissonBoltzmann calculations for the chicken liver enzyme (equivalent to the calculations described for E . coli DHFR) reveal that these three positively charged amino acids are also directly responsible for a positive potential at the entrance to the active site in chicken liver DHFR. Thus, although chicken liver DHFR carries a net charge of + 1, under physiological conditions, and, therefore, presents a very different overall electrostatic environment compared to the E . coli enzyme, a common potential is observed around the active site entrance. These conclusions are similar to those found in a comparison of trypsins from different species, as described above, where a common active site potential was found despite very different overall net charges.31 Examination of the potential at the other end of the active site, in the region in which the pteridine moi-
Fig. 5. The potential of the ternary complex, i.e., with the bound ligands included in the calculation. (a) A reference figure to show the orientation of the ternary complex in b. (b) A close-up view of the potential around the ternary complex. This complex carries, at neutral pH, a net charge of -16. Comparison with Figure 2 shows that the positive potential around the ligand binding sites is significantly reduced due to the complementarity of protein and ligand charges. In the folate binding region, the two negatively charged carboxylate groups of the glutamate moiety of folate directly interact with the three positively charged residues at positions 32, 52, and 57. The net charge of 1 in this region is responsible for the small remaining fraction of positive potential around the glutamate moiety of folate. (c) The pockets of positive potential around the entrances to the ligand binding sites are very much reduced in size when the ligands are bound, in comparison with the protein without the bound ligands (Fig. 2c). Within the crystal of the ternary complex, adjacent complexes are packed closely together. This indicates that the crystal must contain a large aqueous distribution of positive counterions, which act both to balance the negative net charge of these protein complexes and to screen the unfavorable electrostatic interactions between them.
+
ELECTROSTATICS OF DHFR
Fig. 5. Legend appears on page 8
9
10
J . BAJORATH ET AL
Fig. 6. A stereo representation comparing the structural motifs that are responsible for the positive potential at the entrances to the active sites in E. coE DHFR (EC) and chicken liver DHFR (CL). The E. coli DHFR a-carbon trace is shown by a yellow ribbon and the chicken liver DHFR trace by a green ribbon. The enzymes are shown in equivalent orientations. In both enzymes,the three pos-
itively charged amino acids that induce the positive active site potential are shown with small spheres on the side chain atoms. Positions32,52, and 57 in E. coli DHFR correspond to 36,70, and 68 in the vertebrate enzyme, forming a similar structural arrangement with equivalent function.
TABLE I. Comparison of Net Charges of Bacterial and Vertebrate DHFRs
chicken liver DHFRs prompted us to raise a question as to the role of global charge-charge interactions. A comparison of the net charges of DHFRs from other bacterial and vertebrate species (derived from sequence data, ref. 32) is given in Table I. As seen here, the net charge on the various enzymes covers a range from - 10 for E . coli to 3 for porcine liver dihydrofolate reductase, with bacterial DHFRs seemingly having a tendency to be more negative than the enzymes from vertebrate species. That a wide range of net charges is observed for a series of enzymes with the same function further supports the hypothesis that, for some proteins, the overall electrostatic nature of the protein may be relatively unimportant for its function, and that it is arrangements of specific charges within a defined structural environment that give rise to important electrostatic interactions within protein^.^' Other cases of favorable interactions between like-charged species also support this hypothesis. For example, in addition to the E. coli DHFR ternary complex with NADPH and folate, superoxide dismutase17 is a well-studied example of a system where a negative ligand binds to a negative protein; the complex between trypsin and BPT133,34is a n example of a n
Bacterial Species Net charge E . coli -10 S. faecium -5 L. casei -4 N . gonorrhoeae -1
Vertebrate Species Net charge Mouse -2 Bovine liver -1 Human 0 Chicken liver +1 Porcine liver +3
ety of the substrate binds, reveals that here there is, however, a difference between the two enzymes. In the case of the E . coli enzyme, the potentials at the positions of the atoms of the pteridine ring are slightly negative-approximately 0.0 to -0.5 kcali mol. In the chicken liver enzyme the potentials a t the corresponding positions are positive, with values of + 2 kcal/mol.
-
Determinants of Protein-Protein and Protein-Ligand Binding As discussed in part above, the significant difference in the overall net charge of the E. coli and
+
ELECTROSTATICS OF DHFR
interaction between two positively charged proteins. The results obtained for these systems also show that local potential effects seem to determine ligand binding. In all of these cases, the global electrostatic potentials would be expected to inhibit binding.
CONCLUSIONS Electrostatic potential calculations have shown how E. coli DHFR, with its net charge of -10, is able to bind multiple negatively charged ligands despite its large negative net charge. It is shown that the dielectric boundary, along with the shape of the active site, focuses the electric field out into solution. In both bacterial and vertebrate enzymes, a structural motif of three positively charged amino acids in the ligand binding site is shown to give rise t o a region of positive potential around the entrance to the active site. This arrangement, and the positive active site potential that it induces, appears to be conserved for DHFRs from different species which show a wide distribution of overall net charges, supporting the role of this structural motif in both binding and catalysis. A similar motif is found in the cofactor binding site. The complementarity of the charges of the substrate and cofactor and the residues of the binding sites leads to a significant reduction of the positive binding site potentials when the ternary complex is formed. ACKNOWLEDGMENTS Financial support was provided by the National Institutes of Health (Grant GM30564). Computer calculations were carried out in part on the Cray X-MP/48 at the San Diego Supercomputer Center. The authors thank Dr. Richard Fine for helpful discussions. This work was carried out at Biosym Technologies, Inc. DEDICATION Jurgen Bajorath dedicates this work to Professor Wolfram Saenger, Berlin.
REFERENCES 1. Kraut, J., Matthews, D. A. Dihydrofolate reductase. In: “Biological Macromolecules and Assemblies,” Vol. 3. Jurnak, F. A,, McPherson, A. (eds.). New York: John Wiley, 1987:l-72. 2. Fierke, C. A,, Johnson, K. A,, Benkovic, S. J . Construction and evaluation of the kinetic scheme associated with dihydrofolate reductase from Escherichia coli. Biochemistry 26:4085-4092, 1987. 3. Villafranca, J. E., Howell, E. E., Voet, D. H., Strobel, M. S., Ogden, R. C., Abelson, J. N., Kraut, J . Directed mutagenesis of dihydrofolate reductase. Science 222:782-788, 1983. 4. Howell, E. E., Warren, M. S., Booth, C. L. J., Villafranca, J. E., Kraut, J. Construction of a n altered proton donation mechanism in Escherichia coli dihydrofolate reductase. Biochemistry 26:8591-8598, 1987. 5. Roberts, V. A,, Dauber-Osguthorpe, P., Osguthorpe, D. J., Wolff, J., Hagler, A. T. A comparison of the binding of the ligand trimethoprim to bacterial and vertebrate dihydrofolate reductases. Isr. J. Chem. 27:198-210, 1986. 6. Dauber-Osguthorpe, P., Roberts, V. A,, Osguthorpe, D. J., Wolff, J., Genest, M., Hagler, A. T. Structure and energet-
11
ics of ligand binding to proteins: Escherichia coli dihydrofolate reductase-trimethoprim, a drug-receptor system. Proteins 4:31-47, 1988. 7. Fleischmann, S. H., Brooks 111, C. L. Protein-drug interactions: Characterization of inhibitor binding in complexes of DHFR with trimethoprim and related derivatives. Proteins 7:52-61, 1990. 8. Singh, U. C., Benkovic, S. J . A free-energy perturbation study of the binding of methotrexate to mutants of dihydrofolate reductase. Proc. Natl. Acad. Sci. U.S.A. 85:95199523,1988. 9. Singh, U. C. Probing the salt bridge in the dihydrofolate reductase-methotrexate complex by using the coordinatecoupled free-energy perturbation method. Proc. Natl. Acad. Sci. U.S.A. 85:4280-4284, 1988. 10. Matthews, D. A., Alden, R. A,, Bolin, J. T., Freer, S. T., Hamlin, R., Xuong, N.-H., Kraut, J., Poe, M., Williams, M., Hoogsteen, K. Dihydrofolate reductase: X-ray structure of the binary complex with methotrexate. Science 197: 452-455, 1977. 11. Bolin, J . T., Filman, D. J., Matthews, D. A., Hamlin, R., Kraut, J. Crystal structures of Escherichia coli and Lactobacillus casei dihydrofolate reductase refined at 1.7 8, resolution. I. General features and binding of methotrexate. J. Biol. Chem. 257:13650-13662, 1982. 12. Filman, D. J., Bolin, J. T., Matthews, D. A., Kraut, J. Crystal structures of Escherichia coli and Lactobacillus casei dihydrofolate reductase a t 1.7 A resolution. 11. Environment of bound NADPH and implications for catalysis. J. Biol Chem. 257:13663-13672, 1982. 13. Matthews, D. A,, Bolin, J. T., Burridge, J . M., Filman, D. J., Volz, K. W., Kaufman, B. T., Beddell, C. R., Champness, J . N., Stammers, D. K., Kraut, J . Refined crystal structures of Escherichia coli and chicken liver dihydrofolate reductase containing bound trimethoprim. J . Biol. Chem. 260:381-391, 1985. 14. Champness, J. N., Stammers, D. K., Beddell, C. R. Crystallographic investigation of the cooperative interaction between trimethoprim, reduced cofactor and dihydrofolate reductase. FEBS Lett. 199:61-67, 1986. 15. Volz, K. W., Matthews, D. A,, Alden, R. A,, Freer, S. T., Hansch, C., Bernhard, T., Kaufman, B. T., Kraut, J. Crystal structure of avian dihydrofolate reductase containing phenyltriazine and NADPH. J . Biol. Chem. 25795282536, 1982. 16. Bajorath, J., Kitson, D. H., Fitzgerald, G., Andzelm, J., Kraut, J., Hagler, A. T. Electron redistribution on binding of a substrate to a n enzyme: Folate and dihydrofolate reductase. Proteins 9:217-224. 17. Klapper, I., Hagstrom, R., Fine, R., Sharp, K., Honig, B. Focusing of electric fields in the active site of Cu-Zn superoxide dismutase: Effects of ionic strength and aminoacid modification. Proteins 1:47-59, 1986. 18. Bystroff, C., Oatley, S. J., Kraut, J . Crystal structures of Escherichia coli dihydrofolate reductase: The NADP+ haloenzyme and the folate-NADP+ ternary complex. Substrate binding and a model for the transition state. Biochemistry 29:3263-3277, 1990. 19. Stone, S. R., Morrison, J. F. The pH-dependence of the binding of dihydrofolate and substrate analogues to dihydrofolate reductase from Escherichia coli. Biochim. Biophys. Acta 745:247-258, 1983. 20. Baccarani, D. P., Stone, D., Kuyper, L. J . Effect of single amino acid substitution on Escherichia coli dihydrofolate reductase catalysis and ligand binding. Biol. Chem. 256: 1738-1747, 1981. 21. Mackay, D. H. J., Cross, A. J., Hagler, A. T. The role of energy minimization in simulation strategies of biomolecular systems. In: “Prediction of Protein Structure and the Principles of Protein Conformation.” Fasman, G. D. (ed.). New York: Plenum Press, 1989:317-358. 22. DISCOVER, Vers. 2.5. Biosym Technologies, San Diego, 1989. 23. Gilson, M., Sharp, K., Honig, B. Calculating the electrostatic potential of molecules in solution: Method and error assessment. J. Comp. Chem. 9:327-335, 1987. 24. Gilson, M., Honig, B. Calculation of electrostatic potentials in a n enzyme active site. Nature (London)330:84-86, 1987. 25. Sharp, K. A,, Honig, B. Electrostatic interactions in mac-
12
26. 27. 28.
29. 30.
J . BAJORATH ET AL. romolecules: theory and applications. Annu. Rev. Biophys. Biophys. Chem. 19:301-332, 1990. DELPHI, Vers. 2.0. Columbia University, New York; Biosym Technologies, San Diego, 1990. Harvey, S. C. Treatment of electrostatic effects in molecular modeling. Proteins 5:78-92, 1989. Avbelj, F., Moult, J., Kitson, D. H., James, M. N. G., Hagler, A. T. Molecular dynamics study of the structure and dynamics of a protein molecule in a crystalline ionic environment, Streptomyces griseus Protease A. Biochemistry 2953658-8676,1990. Avbelj, F., Moult, J., Kitson, D. H., James, M. N. G., Hadzi, D., Hagler, A. T. To be submitted. Srere, P. A. Complexes of sequential metabolic enzymes. Annu. Rev. Biochem. 56:89-124, 1987.
31. Soman, K., Yang, A.-S., Honig, B., Fletterick, R. Electrical potentials in trypsin isoenzymes. Biochemistry 28:99189926,1989. 32. Matthews, D. A,, Kraut, J. Unpublished. 33. Struthers, R. S., Kitson, D. H., Hagler, A. T. Predicted three-dimensional structure of the protease inhibitor domain of the Alzheimer’s disease P-amyloid precursor. Proteins 9:1-11, 1991. 34. Marquart, M., Walter, J., Deisenhofer, J.,Bode, W., Huber, R. The geometry of the reactive site and of the peptide
groups in trypsin, trypsinogen and its complexes with inhibitors. Acta Crystallogr. B 39:480-490, 1983. 35. INSIGHT-11, Vers. 1.0., Biosym Technologies Inc., San Diego, 1990.
The Electrostatic Potential of Escherichia coZi Dihydrofolate Reductase Jurgen Bajorath,' David H. Kitson,' Joseph Kraut: and Arnold T. Hagler' 'Biosym Technologies, Inc., 10065 Barnes Canyon Road, San Diego, California 92121, and 'Department of Chemistry, University of California at San Diego, La Jolla, California 92093
ABSTRACT Escherichia coli dihydrofolate reductase (DHFR) carries a net charge of -10 electrons yet it binds ligands with net charges of -4 (NADPH) and -2 (folate or dihydrofolate). Evaluation and analysis of the electrostatic potential of the enzyme give insight as to how this is accomplished. The results show that the enzyme is covered by an overall negative potential (as expected)except for the ligand binding sites, which are located inside "pockets" of positive potential that enable the enzyme to bind the negatively charged ligands. The electrostatic potential can be related to the asymmetric distribution of charged residues in the enzyme. The asymmetric charge distribution, along with the dielectric boundary that occurs at the solvent-protein interface, is analogous to the situation occurring in superoxide dismutase. Thus DHFR is another case where the shape of the active site focuses electric fields out into solution. The positive electrostatic potential at the entrance of the ligand binding site in E. coli DHFR is shown to be a direct consequence of the presence of three positively charged residues at positions 32,52, and 57-residues which have also been shown recently to contribute significantly to electronic polarization of the ligand folate. The latter has been postulated to be involved in the catalytic process. A similar structural motif of three positively charged amino acids that gives rise to a positive potential at the entrance to the active site is also found in DHFR from chicken liver, and is suggested to be a common feature in DHFRs from many species. It is noted that, although the net charges of DHFRs from different species vary from + 3 to -10, the enzymes are able to bind the same negatively charged ligands, and perform the same catalytic function. Key words: electrostatics,enzyme-substrateinteraction, solvent screening, active site potential, structure-function relationship 0 1991 WILEY-LISS, INC.
INTRODUCTION Dihydrofolate reductase (DHFR) catalyzes the NADPH- (reduced nicotinamide adenine dinucleotide phosphate) dependent reduction of folate to dihydrofolate and of dihydrofolate to tetrahydrofolate in bacterial and vertebrate species.' The enzyme is one of the most intensively studied proteins, and has been investigated, both experimentally and theoretically, from many different points of A variety of X-ray crystal structures of DHFRs, from different species and in complexes with various inhibitor^,"-'^ have been solved, and these provide the basis for much of the current research on the enzyme. In a recent study the changes in the electron density of the substrate folate when it binds to DHFR were investigated.16 The binding of folate, which carries a net charge of -2, and of NADPH, with a net charge of -4, to E . coli DHFR is somewhat intriguing since the enzyme also has a net negative charge, of -10. Binding of folate and NADPH to the like-charged DHFR is a more drastic example of the much studied effect seen in superoxide dismutase, where a 4-fold negatively charged protein binds a ligand with a net charge of -1 (17, and references therein). In addition, calculations of the effects of binding on the electronic structure of folate have shown that the ligand undergoes a significant polarization upon binding that leads to dramatic changes in the electronic structure. This electronic rearrangement may play a significant role in the enzymatic reaction.16 To gain some qualitative understanding of the apparently noncomplementary electrostatics of the ligand and of the protein to which it binds, and t o further describe the field postulated to polarize the substrate, it was of interest to study the electrostatic properties of the enzyme in a dielectric environment. The effect of the binding of the ligands on the electrostatic potential of the enzyme is also examined, and the origin of the active
Received August 13, 1990; revision accepted November 15, 1990. Address reprint requests to Arnold T. Hagler, Biosym Technologies, Inc., 10065 Barnes Canyon Road, San Diego, CA 92121. Jurgen Bajorath's current address: Bristol-Myers Squibb, Pharmaceutical Research InstituteSeattle, 3005 1st Avenue, Seattle, WA 98121.
2
J. BAJORATH ET AL.
Fig. 1. Stereo representation of the ternary complex of f.coli DHFR with folate and NADPH. The DHFR backbone is shown as a yellow ribbon; the substrate (folate) and the cofactor (NADPH) in
"ball-and-stick' representations, color coded according to atom type (C, green; N, blue; 0, red). This, and the other color figures, were produced with the Insight I1 software package.35
site potential is analyzed, based on the structural details of the protein around the ligand binding site. The long-range electrostatic potential of DHFR is investigated and implications for ligand binding are discussed. Finally, the electrostatics of DHFRs from different species are compared in order to evaluate the role of the net charge of the enzyme in ligand binding and reactivity. This is done by comparing the structures and electrostatic properties of E . coli and chicken liver DHFRs, together with the sequences of the enzymes from a variety of other bacterial and vertebrate species.
with only the waters allowed to move, was run to equilibrate the hydration shell. Steric overlaps and other sources of significant strain (arising from inaccuracies in the X-ray structure, or artifacts introduced by the hydration procedure) were then relaxed by energy minimization until the maximum derivative of the energy with respect to the atomic positions was less than 0.5 kcallmollA.21.22During the minimization, a tethering constant of 100 kcall mo1/A2 was applied to constrain the nonhydrogen atoms of the protein and the ligands to positions close to those of the X-ray structure.21 The resulting
METHODS The starting point for the study was the refined crystal structure of the ternary complex of E . coli DHFR with NADP' and fo1ate.l' This system was chosen as it is the only experimentally determined DHFR structure that includes a bound substrate. Thus, the structural basis for the calculation is well grounded in experiment. The NADP+ was converted to NADPH in order to mimic the active form of the enzyme. Figure 1 shows a representation of this ternary complex. Folate binds to the enzyme in a hydrophobic cleft with the glutamate moiety, which carries the two negatively charged carboxyl groups, located toward the protein surface. Before carrying out the electrostatics calculations, a molecular modeling protocol was carried out to prepare the starting structure. Hydrogen atoms were added to the protein and substrates in standard geometries and water molecules included to give a complete 5 A hydration shell around the ligand binding sites and a 3 A shell around the remainder of the protein. As suggested by recent enzyme kinetic studies the side chain of Asp-27 in the active site was modeled as uncharged (i.e., protonated). A short molecular dynamics simulation, 19720
Fig. 2. The electrostatic potential of E. coli DHFR in solution. The potential was calculated for the protein alone, without the bound ligands, which are, however, shown in these figures to indicate their positions relative to the potential. (a) The ternary complex of DHFR, folate, and NADPH, shown as a reference for b, in which the orientation and scale of the molecule are the same as is used in a. The enzyme is represented by a green ribbon, while folate and NADPH are represented by "ball-and-stick' models. (b) A close-up view of the potential of DHFR. Contours of negative potential around the protein are shown at a level of -0.6 kcalimol (red); the positive contour level is +0.6 kcalimol (blue) [a contour level of -0.6 kcalimol means that an atom of one unit of positive charge, sitting at a point on the contour, would have an interaction energy of -0.6 kcalimol (a favorable interaction) with the potential at that point]. Both the substrate folate (with a net charge of -2) and the cofactor NADPH (with a net charge of -4) are seen to bind in "pockets" of positive potential that are embedded within the largely negative potential around the 10-fold negatively charged protein. (c) The long-range electrostatic potential around E. coli DHFR. The positive potentials around the ligand binding sites extend far out (-30 A from the active site) into the solvent environment. DHFR is able to bind dihydrofolate in the polyglutamate form (as a polyanion), and the reason for this may be appreciated from this figure, since the polyglutamate tail can be accommodated within the large positive potential region around the entrance to the active site. The view in this figure is rotated slightly from that in a and b. The potentials in b and c were calculated using the Poisson-Boltzmann meth~d,'~-'' assuming a dielectric constant of 2 for the interior of the protein, and 80 for the exterior, an ionic strength of 20 mM (the strength used to prepare the complex for crystallization2') and a temperature of 300 K.
ELECTROSTATICS OF DHFR
Fig. 2. Legend appears on page 2.
3
4
J. BAJORATH ET AL.
structure was then used for the electrostatic potential calculations. The electrostatic potential calculations were carried out using the Finite Difference Poisson-Boltzmann m e t h ~ d ? ~ -and ’ ~ for this purpose the explicit water molecules in the hydration shell were omitted. The Poisson-Boltzmann method takes the different dielectric constants of the protein interior and the surrounding solvent into account (without the need to treat solvent molecules explicitly), thereby allowing for the shielding effects of the solvent. (It is, therefore, a more realistic technique for studying the electrostatic potential of proteins than the simple Coulombic formulation of charge-charge interactions applied to a protein in vacuo.) The dielectric constant was set to 2 for the protein interior and to 80 for the surrounding ~ o l v e n t . ~An ~ ~ ionic ’~ strength of 20 mM was used (the strength used experimentally to prepare the complex for crystallization”) and the temperature was set to 300 K. In order to analyze the electrostatics of the enzyme, several different calculations were carried out on the wild-type enzyme, the ternary complex, and proposed enzyme mutants. RESULTS AND DISCUSSION The Electrostatic Potential of E. coli DHFR in Solution As would be expected for a protein with a net charge of -10, DHFR is largely covered by a negative potential (Fig. 2-the ligands were not included in the calculation of the potential but are shown in the figures to indicate their positions relative to the potential). Despite the overall negative potential, the ligand binding sites exhibit “pockets”of positive potential that extend into the solvent. As Figure 2b shows, folate binds in the center of a region of positive potential a t the entrance to the active site. The adenosine phosphate group of NADPH also binds directly into the center of an area of positive potential. This is consistent with the fact that DHFR binds NADPH, with its adenosine phosphate group, as a cofactor, and not NADH, in which the phosphate group is replaced by a hydroxyl. Thus, the regions of positive potential appear to enable the enzyme to bind the negatively charged ligands. The LongRange Effect of the Active Site Potential The positive ligand binding site potentials of DHFR extend out significantly into the solvent environment (Fig. 2c). Only a t about 30 A from the protein surface is the enzyme completely surrounded by a (very weak) negative potential. A potential gradient (from negative to positive potential) exists 30 A away from the protein. To compare the potential of DHFR in vacuo (the state in which many protein simulations are carried out) with that of the enzyme in solution, we have
recalculated the potential with the dielectric constant of both protein and solvent set to 1 and with the ionic strength reduced to 0.0 M. That is, there is no dielectric boundary between the protein and the “solvent,” and the calculation is now equivalent to a Coulomb’s Law calculation in vacuo. This is found to have a significant influence on both the magnitude and the shape of the potential. With a dielectric constant for the solvent of 80, the most negative potential found at a distance of 5 A from the protein is approximately -3 kcal/mol (Fig. 3a). This means that a single positive charge, for example a sodium ion, would have an energy of only -3 kcal/mol a t this position. In contrast, when the dielectric of the solvent is 1, the most negative potential 5 A from the protein is approximately -240 kcal/mol (Fig. 3b), so that the sodium ion would have an energy of -240 kcal/mol at this position. The 80-fold decrease in energy in the presence of solvent reflects the very large dielectric screening effect of water and the concomitant diminution of the effect of the charge of the protein. At a larger distance from the protein the screening effect of the solvent is even more dramatic. The maximum potential found 30 A from the protein is approximately -0.15 kcal/mol in the presence of solvent (Fig. 2c), whereas, in vacuo, the maximum potential would be more than -60 kcal/mol (Fig. 3c). Furthermore, as can be seen by comparing Figures 2c and 3c, the presence of the dielectric boundary in the case where the solvent dielectric is 80 changes the shape of the potential around the enzyme. In the absence of this dielectric boundary the net total charge of the protein becomes more significant in determining the electrostatic environ-
Fig. 3. The effect of solvent screening and the presence of the protein/solvent dielectric boundary on the electrostatic potential around E. coli DHFR. (a) With a dielectric constant for the solvent of 80 (and 2 for the protein interior), the most negative potential found at a distance of 5 A from the protein is approximately -3 kcallmol [contours are shown at levels of -3.0 kcal/mol (red) and +3.0kcal/mol (blue)]. (Note, this figure was calculated using the same conditions as Fig. 2; the figures differ in the contour levels shown.) In this stereo figure, the enzyme is represented by a yellow ribbon, and folate and NADPH by color-coded line representations. Atoms of negatively charged protein residues are indicated by red balls and those of positively charged residues by blue balls. The figure shows that the charged residues in the protein are asymmetrically arranged, with most of the positive residues being located close to the ligand binding sites, while the negative residues are more widely distributed. (b) As seen by comparison with Figure 2, this asymmetric arrangement of charged residues leads to a similar asymmetry in the potential of the enzyme. When the dielectric of the solvent (as well as the protein interior) is 1, the most negative potential 5 A from the protein becomes approximately -240 kcal/mol. In this figure, the negative contour is shown at the -240 kcal/mol level. (c) As expected, the net negative charge of the protein totally dominates the electrostatic environment of the protein in vacuo. Although there are still pockets of positive potential around the ligand binding sites, they are very much smaller than those present when the solvent dielectric is 80 (Fig. 2c). Note that the contours in this figure are at levels approximately 200 times higher than those of Figure 2c.
ELECTROSTATICS OF DHFR
Fig. 3. Legend appears on page 4.
5
6
J. BAJORATH ET AL.
ment around the protein. The positive pockets around the ligand binding sites shrink considerably in size (Fig. 3c), and the negative potential dominates and completely encloses the protein at a closer distance than in the presence of solvent (Fig. 2c). In Figure 3c it is seen that the active site of DHFR is surrounded by a negative potential contour while, in contrast, the active site potential changes from positive to (very weakly) negative only at about 30 A from the protein in the presence of solvent (Fig. 2c). The comparison illustrates the effects of the dielectric boundary under conditions where the solvent is treated properly (i.e.) as a high dielectric medium and with correct ionic strength). Thus, the effect of the proteidsolvent dielectric boundary, together with the spatial arrangement of protein charges, focuses electric fields of the protein out into solution. This is analogous to the effect seen for superoxide dismutase.17 It is also of interest to compare the electrostatic potential for the situation where the protein is surrounded by the aqueous medium with the approximation of uniform (protein and solvent) dielectric 80, omitting the effects of ionic strength. This is another approximation that would be attractive to invoke in molecular mechanics/dynamics simulations for its simplicity. Since Coulomb’s law is valid in a uniform medium this is equivalent to dividing the potentials in Figure 3c by 80. In this case, the active site would still be surrounded by a negative potential contour (see above, comparison of Figs. 2c and 3c). This again emphasizes the importance of the dielectric boundary.
Asymmetric Charge Distribution in E. coli DHFR Figure 3 illustrates that positively and negatively charged residues are asymmetrically distributed in the enzyme. Positively charged residues are concentrated in the ligand binding site regions (on the “left” side in this figure) whereas negatively charged residues dominate the “right” side of the enzyme. The charge distribution appears to bisect the enzyme. This anisotropy of protein charges correlates directly with the electrostatic potential of the enzyme shown in Figure 2. The Origin of the Positive Electrostatic Potential at the Active Site An understanding of the origin of the pockets of positive potential into which the negative moieties of the ligands bind can be gained by examining the structural details of the ligand binding sites. The glutamate moiety of folate is surrounded by three positively charged protein residues, Lys-32, Arg-52, and Arg-57, the location of which, a t the boundary of the hydrophobic protein cleft, would be expected to induce a positive electrostatic field in the region of the protein where the glutamate moiety binds. Ini-
tially, the possible role of these three positively charged residues in giving rise to a positive potential around the glutamate moiety of folate was investigated based on calculations using a simple, Coulombic model.16 To confirm this role, and to study the field induced by these residues, a PoissonBoltzmann calculation of the potential of the enzyme was carried out in which the charges on these three residues were neutralized. As Figure 4 shows, the pocket of positive potential around the glutamate moiety of folate disappears completely when these charges are removed, indicating that they are directly responsible for this positive region of the electrostatic potential and supporting the qualitative conclusions drawn from the simpler Coulombic model. Examination of the protein structure also suggested that the positive potential pocket around the adenosine phosphate group of NADPH might be due to the presence of three positively charged residues, Arg-44, Lys-76, and Arg-98, in the direct vicinity of the phosphate group. The charges on these residues were also removed in the calculation, and, as Figure 4 shows, this leads to elimination of the positive potential that surrounds the phosphate group of the cofactor. Thus, we may rationalize the binding of these highly negatively charged ligands to the negatively charged protein in aqueous solution in terms of the pockets of positive potential at both the entrance to the active site cleft and the
~
~~
Fig. 4. The positive potential around the entrance to the folate binding site disappears when the potential is calculated with the charges turned off on the three (normally) positive protein residues (Lys-32, Arg-52, and Arg-57) located around the glutamate moiety of folate (and also calculated without the bound ligands). Thus, these three positively charged residues are largely responsible for this pocket of positive potential in E. coli DHFR. As in Figure 2, the potential is superimposed on a representation of the ternary complex. (a) A reference figure to show the orientation of the ternary complex in b. (b) A close-up view of the potential around this “mutant” enzyme. The positive potential into which the glutamate moiety of folate binds has disappeared. Only a small residual potential remains near this region. This potential is due to two positively charged residues, Arg-33 and Lys-58, which are located close to the entrance to the active site of the enzyme. The side chains of these two residues, however, point away from the folate binding site in the crystal structure and, therefore, as can be seen from this figure, do not contribute directly to the positive active site potential around the glutamate moiety in folate (calculation of the potential of the enzyme with the charges on residues 33 and 58 set to zero confirms that this is the case-the pocket of positive potential is largely unchanged). In the calculation of the potential displayed in this figure the charges were also removed on the three positive residues (Arg-44, Lys-76, and Arg-98) that are close to the adenosine phosphate group of NADPH. Comparison with Figure 2b indicates that these residues are largely responsible for the positive potential around this group. (c). The long-range positive potentials around the entrances to the substrate and cofactor binding sites have completely disappeared. (This figure is shown in the same orientation as Fig. 2c). At a distance of 30 A from the active site, an approaching ligand would normally (i.e., for the protein with all charged residues included) begin to “feel” a positive potential (Fig. 2c). In the case of this “mutant” protein (with the charges on Lys-32, Arg-52, Arg-57, Arg-44, Lys-76, and Arg-98 deleted) the ligand would “feel” a weak negative potential (of - -0.06 kcal/mol), which becomes more negative as the distance from the protein decreases.
ELECTROSTATICS OF DHFR
Fig. 4. Legend appears on page 6.
7
8
J. BAJORATH ET AL.
cofactor binding region. Furthermore, these pockets of positive potential are shown to be due, in both clefts, to three basic residues. In the active site cleft, these are residues Lys-32, Arg-52, and Arg-57, while in the cofactor binding site these are residues Arg44,Lys-76, and Arg-98.
The Electrostatic Potential of the Ternary Complex The charge distribution of the ligands and the ligand binding sites of the enzyme are complementary to one another, as can be seen in Figure 5, which shows the potential calculated for the ternary complex. The positive “pockets” around the binding sites are significantly reduced when the ligands are bound to the enzyme. The net charge of the ternary complex that was crystallized (DHFR-folateNADP’) is - 15 [the net charge of the complex studied here (with NADPH) is -161. That a crystal of this highly negatively charged complex can form indicates that there must be a large aqueous distribution of positive counter ions around the highly negatively charged protein complexes within the crystal, that is able to balance the electrostatic interactions between them and stabilize the packing arrangement. This is clear when one considers the effects of the negative potential, even after modulation by solvent, in attracting positive counterions in the solvent. A double charged cation such as Ca2 or Mg2+ would experience a gain of energy of approximately 7 kcal/mol if it came within -5 A of the complex. An analogous situation has been observed in molecular dynamics simulations of the crystal of Streptomyces griseus Protease A. (which has an overall charge of +5) in which the solvent of crystallization was included explicitly.28 In the crystal lattice, the counterions are predicted to be concentrated in positions where they are surrounded by protein residues of opposite charge, stabilizing the crystal in the observed packing arrangement.” It is interesting to speculate as to whether this large negative charge plays a role in protein aggregation in E . coli analogous to that which occurs in the cells of vertebrates, where it has been suggested that proteins exists as aggregates rather than isolated in ~olution.~’ As seen from the binding of the ligands, the net charge of the protein does not necessarily determine the charge of ligands that will bind. Rather, complementary local potentials seem to be required. These observations are consistent with conclusions from a recent study on different t r y p ~ i n s .Here ~ ~ too, local potential effects were shown to determine the active site potential and to be conserved, despite considerably different overall net charges of the two enzymes, rat and cow trypsin, that were studied. +
Structural and Functional Comparison With Chicken Liver DHFR In the E . coli enzyme, comparison of Figures 2 and 4 shows that the presence of the positive potential around the entrance to the active site is a direct consequence of the structural arrangement of the three positively charged amino acids at the entrance to the cleft. This structural feature has also been shown to play a significant role in the electronic polarization of folate by the enzyme.16 Thus, the more proper treatment of solvent by the PoissonBoltzmann approximation yields a similar potential in this local region to that found by a simple Coulombic treatment, further supporting the interpretation of the underlying basis for the polarization.16 Examination of the structure of chicken liver DHFR reveals that a similar arrangement of positive charges is found, in this case consisting of residues Arg-36, Lys-68, and Arg-70. Superposition of these structural motifs in the E . coli and chicken liver enzymes shows that the positively charged side chains occupy very similar positions within the ligand binding sites (Fig. 6). Furthermore, PoissonBoltzmann calculations for the chicken liver enzyme (equivalent to the calculations described for E . coli DHFR) reveal that these three positively charged amino acids are also directly responsible for a positive potential at the entrance to the active site in chicken liver DHFR. Thus, although chicken liver DHFR carries a net charge of + 1, under physiological conditions, and, therefore, presents a very different overall electrostatic environment compared to the E . coli enzyme, a common potential is observed around the active site entrance. These conclusions are similar to those found in a comparison of trypsins from different species, as described above, where a common active site potential was found despite very different overall net charges.31 Examination of the potential at the other end of the active site, in the region in which the pteridine moi-
Fig. 5. The potential of the ternary complex, i.e., with the bound ligands included in the calculation. (a) A reference figure to show the orientation of the ternary complex in b. (b) A close-up view of the potential around the ternary complex. This complex carries, at neutral pH, a net charge of -16. Comparison with Figure 2 shows that the positive potential around the ligand binding sites is significantly reduced due to the complementarity of protein and ligand charges. In the folate binding region, the two negatively charged carboxylate groups of the glutamate moiety of folate directly interact with the three positively charged residues at positions 32, 52, and 57. The net charge of 1 in this region is responsible for the small remaining fraction of positive potential around the glutamate moiety of folate. (c) The pockets of positive potential around the entrances to the ligand binding sites are very much reduced in size when the ligands are bound, in comparison with the protein without the bound ligands (Fig. 2c). Within the crystal of the ternary complex, adjacent complexes are packed closely together. This indicates that the crystal must contain a large aqueous distribution of positive counterions, which act both to balance the negative net charge of these protein complexes and to screen the unfavorable electrostatic interactions between them.
+
ELECTROSTATICS OF DHFR
Fig. 5. Legend appears on page 8
9
10
J . BAJORATH ET AL
Fig. 6. A stereo representation comparing the structural motifs that are responsible for the positive potential at the entrances to the active sites in E. coE DHFR (EC) and chicken liver DHFR (CL). The E. coli DHFR a-carbon trace is shown by a yellow ribbon and the chicken liver DHFR trace by a green ribbon. The enzymes are shown in equivalent orientations. In both enzymes,the three pos-
itively charged amino acids that induce the positive active site potential are shown with small spheres on the side chain atoms. Positions32,52, and 57 in E. coli DHFR correspond to 36,70, and 68 in the vertebrate enzyme, forming a similar structural arrangement with equivalent function.
TABLE I. Comparison of Net Charges of Bacterial and Vertebrate DHFRs
chicken liver DHFRs prompted us to raise a question as to the role of global charge-charge interactions. A comparison of the net charges of DHFRs from other bacterial and vertebrate species (derived from sequence data, ref. 32) is given in Table I. As seen here, the net charge on the various enzymes covers a range from - 10 for E . coli to 3 for porcine liver dihydrofolate reductase, with bacterial DHFRs seemingly having a tendency to be more negative than the enzymes from vertebrate species. That a wide range of net charges is observed for a series of enzymes with the same function further supports the hypothesis that, for some proteins, the overall electrostatic nature of the protein may be relatively unimportant for its function, and that it is arrangements of specific charges within a defined structural environment that give rise to important electrostatic interactions within protein^.^' Other cases of favorable interactions between like-charged species also support this hypothesis. For example, in addition to the E. coli DHFR ternary complex with NADPH and folate, superoxide dismutase17 is a well-studied example of a system where a negative ligand binds to a negative protein; the complex between trypsin and BPT133,34is a n example of a n
Bacterial Species Net charge E . coli -10 S. faecium -5 L. casei -4 N . gonorrhoeae -1
Vertebrate Species Net charge Mouse -2 Bovine liver -1 Human 0 Chicken liver +1 Porcine liver +3
ety of the substrate binds, reveals that here there is, however, a difference between the two enzymes. In the case of the E . coli enzyme, the potentials at the positions of the atoms of the pteridine ring are slightly negative-approximately 0.0 to -0.5 kcali mol. In the chicken liver enzyme the potentials a t the corresponding positions are positive, with values of + 2 kcal/mol.
-
Determinants of Protein-Protein and Protein-Ligand Binding As discussed in part above, the significant difference in the overall net charge of the E. coli and
+
ELECTROSTATICS OF DHFR
interaction between two positively charged proteins. The results obtained for these systems also show that local potential effects seem to determine ligand binding. In all of these cases, the global electrostatic potentials would be expected to inhibit binding.
CONCLUSIONS Electrostatic potential calculations have shown how E. coli DHFR, with its net charge of -10, is able to bind multiple negatively charged ligands despite its large negative net charge. It is shown that the dielectric boundary, along with the shape of the active site, focuses the electric field out into solution. In both bacterial and vertebrate enzymes, a structural motif of three positively charged amino acids in the ligand binding site is shown to give rise t o a region of positive potential around the entrance to the active site. This arrangement, and the positive active site potential that it induces, appears to be conserved for DHFRs from different species which show a wide distribution of overall net charges, supporting the role of this structural motif in both binding and catalysis. A similar motif is found in the cofactor binding site. The complementarity of the charges of the substrate and cofactor and the residues of the binding sites leads to a significant reduction of the positive binding site potentials when the ternary complex is formed. ACKNOWLEDGMENTS Financial support was provided by the National Institutes of Health (Grant GM30564). Computer calculations were carried out in part on the Cray X-MP/48 at the San Diego Supercomputer Center. The authors thank Dr. Richard Fine for helpful discussions. This work was carried out at Biosym Technologies, Inc. DEDICATION Jurgen Bajorath dedicates this work to Professor Wolfram Saenger, Berlin.
REFERENCES 1. Kraut, J., Matthews, D. A. Dihydrofolate reductase. In: “Biological Macromolecules and Assemblies,” Vol. 3. Jurnak, F. A,, McPherson, A. (eds.). New York: John Wiley, 1987:l-72. 2. Fierke, C. A,, Johnson, K. A,, Benkovic, S. J . Construction and evaluation of the kinetic scheme associated with dihydrofolate reductase from Escherichia coli. Biochemistry 26:4085-4092, 1987. 3. Villafranca, J. E., Howell, E. E., Voet, D. H., Strobel, M. S., Ogden, R. C., Abelson, J. N., Kraut, J . Directed mutagenesis of dihydrofolate reductase. Science 222:782-788, 1983. 4. Howell, E. E., Warren, M. S., Booth, C. L. J., Villafranca, J. E., Kraut, J. Construction of a n altered proton donation mechanism in Escherichia coli dihydrofolate reductase. Biochemistry 26:8591-8598, 1987. 5. Roberts, V. A,, Dauber-Osguthorpe, P., Osguthorpe, D. J., Wolff, J., Hagler, A. T. A comparison of the binding of the ligand trimethoprim to bacterial and vertebrate dihydrofolate reductases. Isr. J. Chem. 27:198-210, 1986. 6. Dauber-Osguthorpe, P., Roberts, V. A,, Osguthorpe, D. J., Wolff, J., Genest, M., Hagler, A. T. Structure and energet-
11
ics of ligand binding to proteins: Escherichia coli dihydrofolate reductase-trimethoprim, a drug-receptor system. Proteins 4:31-47, 1988. 7. Fleischmann, S. H., Brooks 111, C. L. Protein-drug interactions: Characterization of inhibitor binding in complexes of DHFR with trimethoprim and related derivatives. Proteins 7:52-61, 1990. 8. Singh, U. C., Benkovic, S. J . A free-energy perturbation study of the binding of methotrexate to mutants of dihydrofolate reductase. Proc. Natl. Acad. Sci. U.S.A. 85:95199523,1988. 9. Singh, U. C. Probing the salt bridge in the dihydrofolate reductase-methotrexate complex by using the coordinatecoupled free-energy perturbation method. Proc. Natl. Acad. Sci. U.S.A. 85:4280-4284, 1988. 10. Matthews, D. A., Alden, R. A,, Bolin, J. T., Freer, S. T., Hamlin, R., Xuong, N.-H., Kraut, J., Poe, M., Williams, M., Hoogsteen, K. Dihydrofolate reductase: X-ray structure of the binary complex with methotrexate. Science 197: 452-455, 1977. 11. Bolin, J . T., Filman, D. J., Matthews, D. A., Hamlin, R., Kraut, J. Crystal structures of Escherichia coli and Lactobacillus casei dihydrofolate reductase refined at 1.7 8, resolution. I. General features and binding of methotrexate. J. Biol. Chem. 257:13650-13662, 1982. 12. Filman, D. J., Bolin, J. T., Matthews, D. A., Kraut, J. Crystal structures of Escherichia coli and Lactobacillus casei dihydrofolate reductase a t 1.7 A resolution. 11. Environment of bound NADPH and implications for catalysis. J. Biol Chem. 257:13663-13672, 1982. 13. Matthews, D. A,, Bolin, J. T., Burridge, J . M., Filman, D. J., Volz, K. W., Kaufman, B. T., Beddell, C. R., Champness, J . N., Stammers, D. K., Kraut, J . Refined crystal structures of Escherichia coli and chicken liver dihydrofolate reductase containing bound trimethoprim. J . Biol. Chem. 260:381-391, 1985. 14. Champness, J. N., Stammers, D. K., Beddell, C. R. Crystallographic investigation of the cooperative interaction between trimethoprim, reduced cofactor and dihydrofolate reductase. FEBS Lett. 199:61-67, 1986. 15. Volz, K. W., Matthews, D. A,, Alden, R. A,, Freer, S. T., Hansch, C., Bernhard, T., Kaufman, B. T., Kraut, J. Crystal structure of avian dihydrofolate reductase containing phenyltriazine and NADPH. J . Biol. Chem. 25795282536, 1982. 16. Bajorath, J., Kitson, D. H., Fitzgerald, G., Andzelm, J., Kraut, J., Hagler, A. T. Electron redistribution on binding of a substrate to a n enzyme: Folate and dihydrofolate reductase. Proteins 9:217-224. 17. Klapper, I., Hagstrom, R., Fine, R., Sharp, K., Honig, B. Focusing of electric fields in the active site of Cu-Zn superoxide dismutase: Effects of ionic strength and aminoacid modification. Proteins 1:47-59, 1986. 18. Bystroff, C., Oatley, S. J., Kraut, J . Crystal structures of Escherichia coli dihydrofolate reductase: The NADP+ haloenzyme and the folate-NADP+ ternary complex. Substrate binding and a model for the transition state. Biochemistry 29:3263-3277, 1990. 19. Stone, S. R., Morrison, J. F. The pH-dependence of the binding of dihydrofolate and substrate analogues to dihydrofolate reductase from Escherichia coli. Biochim. Biophys. Acta 745:247-258, 1983. 20. Baccarani, D. P., Stone, D., Kuyper, L. J . Effect of single amino acid substitution on Escherichia coli dihydrofolate reductase catalysis and ligand binding. Biol. Chem. 256: 1738-1747, 1981. 21. Mackay, D. H. J., Cross, A. J., Hagler, A. T. The role of energy minimization in simulation strategies of biomolecular systems. In: “Prediction of Protein Structure and the Principles of Protein Conformation.” Fasman, G. D. (ed.). New York: Plenum Press, 1989:317-358. 22. DISCOVER, Vers. 2.5. Biosym Technologies, San Diego, 1989. 23. Gilson, M., Sharp, K., Honig, B. Calculating the electrostatic potential of molecules in solution: Method and error assessment. J. Comp. Chem. 9:327-335, 1987. 24. Gilson, M., Honig, B. Calculation of electrostatic potentials in a n enzyme active site. Nature (London)330:84-86, 1987. 25. Sharp, K. A,, Honig, B. Electrostatic interactions in mac-
12
26. 27. 28.
29. 30.
J . BAJORATH ET AL. romolecules: theory and applications. Annu. Rev. Biophys. Biophys. Chem. 19:301-332, 1990. DELPHI, Vers. 2.0. Columbia University, New York; Biosym Technologies, San Diego, 1990. Harvey, S. C. Treatment of electrostatic effects in molecular modeling. Proteins 5:78-92, 1989. Avbelj, F., Moult, J., Kitson, D. H., James, M. N. G., Hagler, A. T. Molecular dynamics study of the structure and dynamics of a protein molecule in a crystalline ionic environment, Streptomyces griseus Protease A. Biochemistry 2953658-8676,1990. Avbelj, F., Moult, J., Kitson, D. H., James, M. N. G., Hadzi, D., Hagler, A. T. To be submitted. Srere, P. A. Complexes of sequential metabolic enzymes. Annu. Rev. Biochem. 56:89-124, 1987.
31. Soman, K., Yang, A.-S., Honig, B., Fletterick, R. Electrical potentials in trypsin isoenzymes. Biochemistry 28:99189926,1989. 32. Matthews, D. A,, Kraut, J. Unpublished. 33. Struthers, R. S., Kitson, D. H., Hagler, A. T. Predicted three-dimensional structure of the protease inhibitor domain of the Alzheimer’s disease P-amyloid precursor. Proteins 9:1-11, 1991. 34. Marquart, M., Walter, J., Deisenhofer, J.,Bode, W., Huber, R. The geometry of the reactive site and of the peptide
groups in trypsin, trypsinogen and its complexes with inhibitors. Acta Crystallogr. B 39:480-490, 1983. 35. INSIGHT-11, Vers. 1.0., Biosym Technologies Inc., San Diego, 1990.
