The electrostatic contribution to the free energy of folding was calculated for 21 salt bridges in 9 protein X-ray crystal structures using a continuum electrostatic approach with the DELPHI computer-program package. The majority (17) were found to be electrostatically destabilizing; the average free energy change, which is analogous to mutation of salt bridging side chains to hydrophobic isosteres, was calculated to be 3.5 kcal/mol. This is fundamentally different from stability measurements using pKa shifts, which effectively measure the strength of a salt bridge relative to 1 or more charged hydrogen bonds. The calculated effect was due to a large, unfavorable desolvation contribution that was not fully compensated by favorable interactions within the salt bridge and between salt-bridge partners and other polar and charged groups in the folded protein. Some of the salt bridges were studied in further detail to determine the effect of the choice of values for atomic radii, internal protein dielectric constant, and ionic strength used in the calculations. Increased ionic strength resulted in little or no change in calculated stability for 3 of 4 salt bridges over a range of 0.1-0.9 M. The results suggest that mutation of salt bridges, particularly those that are buried, to "hydrophobic bridges" (that pack at least as well as wild type) can result in proteins with increased stability. Due to the large penalty for burying uncompensated ionizable groups, salt bridges could help to limit the number of low free energy conformations of a molecule or complex and thus play a role in determining specificity (i.e., the uniqueness of a protein fold or protein-ligand binding geometry).

The GCN4 leucine zipper is a peptide homodimer that has been the subject of a number of experimental and theoretical investigations into the determinants of affinity and specificity. Here, we utilize this model system to investigate electrostatic effects in protein binding using continuum calculations. A particularly novel feature of the computations made here is that they provide an interaction-by-interaction breakdown of the electrostatic contributions to the free energy of docking that includes changes in the interaction of each functional group with solvent and changes in interactions between all pairs of functional groups on binding. The results show that~1! electrostatic effects disfavor binding by roughly 15 kcal0mol due to desolvation effects that are incompletely compensated in the bound state,~2! while no groups strongly stabilize binding, the groups that are most destabilizing are charged and polar side chains at the interface that have been implicated in determining binding specificity, and~3! attractive intramolecular interactions e.g., backbone hydrogen bonds! that are enhanced on binding due to reduced solvent screening in the bound state contribute significantly to affinity and are likely to be a general effect in other complexes. A comparison is made between the results obtained in an electrostatic analysis carried out calculationally and simulated results corresponding to idealized data from a scanning mutagenesis experiment. It is shown that scanning experiments provide incomplete information on interactions and, if overinterpreted, tend to overestimate the energetic effect of individual side chains that make attractive interactions. Finally, a comparison is made between the results available from a continuum electrostatic model and from a simpler surface-area dependent solvation model. In this case, although the simpler model neglects certain interactions, on average it performs rather well.

Continuum methods were used to calculate the electrostatic contributions of charged and polar side chains to the overall stability of a small 41-residue helical protein, the peripheral subunit-binding domain. The results of these calculations suggest several residues that are destabilizing, relative to hydrophobic isosteres. One position was chosen to test the results of these calculations. Arg8 is located on the surface of the protein in a region of positive electrostatic potential. The calculations suggest that Arg8 makes a significant, unfavorable electrostatic contribution to the overall stability. The experiments described in this paper represent the first direct experimental test of the theoretical methods, taking advantage of solid-phase peptide synthesis to incorporate approximately isosteric amino acid substitutions. Arg8 was replaced with norleucine (Nle), an amino acid that is hydrophobic and approximately isosteric, or with alpha-amino adipic acid (Aad), which is also approximately isosteric but oppositely charged. In this manner, it is possible to isolate electrostatic interactions from the effects of hydrophobic and van der Waals interactions. Both Arg8Nle and Arg8Aad are more thermostable than the wild-type sequence, testifying to the validity of the calculations. These replacements led to stability increases at 52.6 degrees C, the T(m) of the wild-type, of 0.86 and 1.08 kcal mol(-)(1), respectively. The stability of Arg8Nle is particularly interesting as a rare case in which replacement of a surface charge with a hydrophobic residue leads to an increase in the stability of the protein.

ABSTRACT:The therapeutic benefits of the antidepressant nefazodone have been hampered by several cases of acute hepatotoxicity/liver failure. Although the mechanism of hepatotoxicity remains unknown, it is possible that reactive metabolites of nefazodone play a causative role. Studies were initiated to determine whether nefazodone undergoes bioactivation in human liver microsomes to electrophilic intermediates. Following incubation of nefazodone with microsomes or recombinant P4503A4 in the presence of sulfydryl nucleophiles, conjugates derived from the addition of thiol to a monohydroxylated nefazodone metabolite were observed. Product ion spectra suggested that hydroxylation and sulfydryl conjugation occurred on the 3-chlorophenylpiperazine-ring, consistent with a bioactivation pathway involving initial formation of p-hydroxynefazodone, followed by its two-electron oxidation to the reactive quinone-imine intermediate. The formation of novel N-dearylated nefazodone metabolites was also discernible in these incubations, and 2-chloro-1,4-benzoquinone, a by-product of N-dearylation, was trapped with glutathione to afford the corresponding hydroquinone-sulfydryl adduct. Nefazodone also displayed NADPH-, time-, and concentration-dependent inactivation of P4503A4 activity, suggesting that reactive metabolites derived from nefazodone bioactivation are capable of covalently modifying P4503A4. A causative role for 2-chloro-1,4-benzoquinone and/or the quinone-imine intermediate(s) in nefazodone hepatotoxicity is speculated. Although the antianxiety agent buspirone, which contains a pyrimidine ring in place of the 3-chlorophenyl-ring, also generated phydroxybuspirone in liver microsomes, no sulfydryl conjugates of this metabolite were observed. This finding is consistent with the proposal that two-electron oxidation of p-hydroxybuspirone to the corresponding quinone-imine is less favorable due to differences in the protonation state at physiological pH and due to weaker resonance stabilization of the oxidation products as predicted from ab initio measurements.

Recent evidence suggests that the net effect of electrostatics is generally to destabilize protein binding due to large desolvation penalties. A novel method for computing ligand-charge distributions that optimize the tradeoff between ligand desolvation penalty and favorable interactions with a binding site has been applied to a model for bamase. The result is a ligand-charge distribution with a favorable electrostatic contribution to binding due, in part, to ligand point charges whose direct interaction with the binding site is unfavorable, but which make strong intra-molecular interactions that are uncloaked on binding and thus act to lessen the ligand desolvation penalty.