The crystal structure of the ternary complex citrate synthase-oxaloacetate-carboxymethyl coenzyme A has been solved to a resolution of 1.9 A and refined to a conventional crystallographic R factor of 0.185. The structure resembles a proposed transition state of the condensation reaction and suggests that the condensation reaction proceeds through a neutral enol rather than an enolate intermediate. A mechanism for the condensation reaction is proposed which involves the participation of three key catalytic groups (Asp 375, His 274, and His 320) in two distinct steps. The proposed mechanism invokes concerted general acid-base catalysis twice to explain both the energetics of the reaction and the experimentally observed inversion of stereochemistry at the attacking carbon atom.

To probe the nature of the hydrophobic cores of proteins and to test potential ways of increasing protein thermostability, an attempt was made to improve the packing within T4 bacteriophage lysozyme by engineered amino acid replacements. Two mutations, Vat, which were designed to fil the largest cavities that exist in the folded structure of the native protein, were constructed. The mutant proteins have normal activities and their thermal stabilities are marginally lower than that ofwild-type lysozyme. Crystal structure analysis of the mutant proteins shows that the introduced amino acids are accommodated with very little perturbation of the three-dimensional structure. Incorporation of the more bulky hydrophobic residues within the core of the protein is expected to provide an increase in hydrophobic stabilization, but this is seen to be offset by the introduction of strain. Inspection of the mutant structures shows that in each case the introduced amino acid side chain is forced to adopt a non-optimal dihedral angle yX. Strain is also observed in the form ofbond angle distortion and in unfavorable van der Waals contacts. The results illustrate how the observed core structures of proteins represent a compromise between the hydrophobic effect, which will tend to maximize the core packing density, and the strain energy that would be incurred in eliminating all packing defects. The results also suggest that mutations designed to increase protein stability by filling existing cavities may be effective in some cases but are unlikely to provide a general method for increasing protein stability.It has been anticipated for some time that the folding of proteins is probably dominated by the hydrophobic effect, i.e., by the sequestering of nonpolar side chains away from solvent. Recent studies suggest that alterations in many solvent-exposed residues on the surfaces of proteins have little if any effect on folding and stability (1-4). These results lead one to consider protein folding from a somewhat different perspective. Is it possible that folding is determined not by the whole amino acid sequence but by a small subset of key residues? If so, which are the critical residues and how do they determine three-dimensional structure? In an attempt to address questions such as these, increasing attention, both theoretical (5) and experimental (6)(7)(8), is being directed to the hydrophobic cores of proteins. Such cores have densities comparable to crystals of simple organic molecules (9, 10) but the packing is not perfect and some cavities remain (11). What determines the size of these cavities?In the present study we have used site-directed mutagenesis to probe the nature of the core of T4 bacteriophage lysozyme. Specifically, we have asked whether the hydrophobic packing within the molecule can be improved by appropriate amino acid replacements. The hydrophobic surface area that is buried within a folded protein contributes directly to its free energy of stabilization (12, 13). Therefore, amino acid substitutions t...

The structures of four isomorphous crystals of ternary complexes of chicken heart citrate synthase with D- or L-malate and acetyl coenzyme A or carboxymethyl coenzyme A have been determined by X-ray crystallography and fully refined at 1.9-A resolution. The structures show that both L-malate and D-malate bind in a very similar way in the presence of acetylCoA and that the enzyme conformation is "closed". Hydrogen bond geometry is suggested to account for the difference in binding constants of the two stereoisomers. The structures suggest that steric hindrance can account for the observation that proton exchange of acetyl coenzyme A with solvent is catalyzed by citrate synthase in the presence of L-malate but not D-malate. The ternary complexes with malate reveal the mode of binding of the substrate acetylCoA in the ground state. The carbonyl oxygen of the acetyl group is hydrogen bonded to a water molecule and to histidine 274, allowing unambiguous identification of the orientation of this group. The structures support the hypothesis that carboxymethyl coenzyme A is a transition-state analogue for the enolization step of the reaction (Bayer et al., 1981) and additionally support proposed mechanisms for the condensation reaction (Karpusas et al., 1990; Alter et al., 1990).

The X-ray structure of a new crystal form of chicken heart muscle citrate synthase, grown from solutions containing citrate and coenzyme A or L-malate and acetyl coenzyme A, has been determined by molecular replacement at 2.8-A resolution. The space group is P4(3) with a = 58.9 A and c = 259.2 A and contains an entire dimer of molecular weight 100,000 in the asymmetric unit. Both "closed" conformation chicken heart and "open" conformation pig heart citrate synthase models (Brookhaven Protein Data Bank designations 3CTS and 1CTS) were used in the molecular replacement solution, with crystallographic refinement being initiated with the latter. The conventional crystallographic R factor of the final refined model is 19.6% for the data between 6- and 2.8-A resolution. The model has an rms deviation from ideal values of 0.034 A for bond lengths and of 3.6 degrees for bond angles. The conformation of the enzyme is essentially identical with that of a previously determined "open" form of pig heart muscle citrate synthase which crystallizes in a different space group, with one monomer in the asymmetric unit, from either phosphate or citrate solution. The crystalline environment of each subunit of the chicken enzyme is different, yet the conformation is the same in each. The open conformation is therefore not an artifact of crystal packing or crystallization conditions and is not species dependent. Both "open" and "closed" crystal forms of the chicken heart enzyme grow from the same drop, showing that both conformations of the enzyme are present at equilibrium in solution containing reaction products or substrate analogues.

The crystal structure of the functional amino-terminal two-domain fragment of human vascular cell adhesion molecule 1 (VCAM-1) has been determined at 1.9 A resolution. The crystals contain two copies of the molecule in the asymmetric unit. The structure was solved by multiple isomorphous replacement, using lead and selenium derivatives. Anomalous scattering had to be used to resolve the phase ambiguity of a lead derivative. Since the selenium derivative has very small isomorphous differences, the local scaling algorithm had to be used to obtain an interpretable difference Patterson map. The initial phases were improved by non-crystallographic averaging, solvent flattening and histogram matching. The structure has been refined to a crystallographic R factor of 20.4% (15-1.9 A, F>/= 3sigma) and consists of two Ig domains (D1 and D2). The angle between these domains differs by 12 degrees between the two copies of the molecule in the crystallographic asymmetric unit, demonstrating that some movement is possible at the interface. In the amino-terminal domain D1 there is an 'extra' disulfide bond, in addition to the conserved cross-sheet disulfide bond, at the top of the molecule. This bond, a hallmark of the integrin-binding subclass of Ig superfamily proteins, makes the top of this domain very compact. The feature that projects most prominently from D1 is the CD loop, near the base of the domain. The key residue for integrin binding, Asp40, is located in this loop and is easily accessible.