Crystal structure of the complex of subtilisin BPN' with its protein inhibitor Streptomyces subtilisin inhibitor. The structure at 4.3 Angstroms resolution

The crystal structure of the complex of subtilisin BPN′ (EC 3.4.21.14) with its protein inhibitor (Streptomyces subtilisin inhibitor) was solved at 4.3 Å resolution, thus establishing the following. (1) Two subtilisin BPN′ molecules (2E) associate with one dimeric inhibitor molecule (I₂) to form the complex molecule E₂I₂. (2) The conformation of neither the inhibitor nor subtilisin BPN′ undergoes any detectable change at this resolution upon complex formation. (3) The inhibitor binds to subtilisin to form an antiparallel β-sheet, as in the case of trypsin/ trypsin inhibitor complexes. (4) The scissible bond of the inhibitor is between Met73′ and Val74′, as proposed earlier (Ikenaka et al., 1974). (5) The protein inhibitor and the substrates bind to subtilisin BPN′ in essentially the same way.     

Crystal structure at 2.6 A resolution of the complex of subtilisin BPN' with streptomyces subtilisin inhibitor

The crystal structure of the complex of a bacterial alkaline serine proteinase, subtilisin BPN', with its proteinaceous inhibitor SSI (Streptomyces subtilisin inhibitor) was solved at 2.6 A resolution. Compared with other similar complexes involving serine proteinases of the trypsin family, the present structure is unique in several respects. (1) In addition to the usual antiparallel beta-sheet involving the P1, P2 and P3 residues of the inhibitor, the P4, P5 and P6 residues form an antiparallel beta-sheet with a previously unnoticed chain segment (residues 102 through 104, which was named the S4-6 site) of subtilisin BPN'. (2) The S4-6 site does not exist in serine proteinases of the trypsin family, whether of mammalian or microbial origin. (3) Global induced-fit movement seems to occur on SSI: a channel-like structure in SSI where hydrophobic side-chains are sandwiched between two lobes becomes about 2 A wider upon complexing with subtilisin. (4) The complex is most probably a Michaelis complex, as in most of the other complexes. (5) The main role of the "secondary contact region" of SSI seems to be to support the reactive site loop ("primary contact region"). Steric homology of the two contact regions between the inhibitors of the SSI family and the pancreatic secretory trypsin inhibitor-ovomucoid inhibitor family is so high that it seems to indicate divergent evolutionary processes and to support the general notion as to the relationship of prokaryotic and eukaryotic genes put forward by Doolittle (1978).     

Refined crystal structure of Streptomyces griseus trypsin at 1.7 A resolution

Streptomyces griseus trypsin (SGT) is a bacterial serine proteinase that is more homologous to mammalian than to other bacterial enzymes. The structure of SGT has been solved primarily by molecular replacement, though some low-resolution phase information was supplied by heavy-atom derivatives. The mammalian pancreatic serine proteinases bovine trypsin (BT) and alpha-chymotrypsin (CHT) were used as molecular replacement models. Because these proteins have low homology with SGT compared to the majority of other successful replacement models, new strategies were required for molecular replacement to succeed. The model of SGT has been refined at 1.7 A resolution to a final R-factor of 0.161 (1 A = 0.1 nm); the correlation coefficient between all observed and calculated structure factor amplitudes is 0.908. Solvent molecules located in the crystal structure play an important role in stabilizing buried charged and polar groups. An additional contribution to stability can be seen in the fact that the majority of the charged side-chains are involved in ionic interactions, sometimes linking the two domains of SGT. A comparison of SGT with BT shows that the greatest similarities are in the active-site and substrate-binding regions, consistent with their similar substrate specificities. The modeling of complexes of SGT with two inhibitors of BT, pancreatic trypsin inhibitor (PTI) and the third domain of Japanese quail ovomucoid (OMJPQ3), helps to explain why PTI inhibits SGT but OMJPQ3 does not. Like BT, but unlike other bacterial serine proteinases of known structure, SGT has a buried N terminus. SGT has also a well-defined Ca2+-binding site, but this site differs in location from that of BT.     

Refined structure of porcine pepsinogen at 1.8 A resolution

The molecular structure of porcine pepsinogen at 1.8 A resolution has been determined by a combination of molecular replacement and multiple isomorphous phasing techniques. The resulting structure was refined by restrained-parameter least-squares methods. The final R factor [formula: see text] is 0.164 for 32,264 reflections with I greater than or equal to sigma (I) in the resolution range of 8.0 to 1.8 A. The model consists of 2785 protein atoms in 370 residues, a phosphoryl group on Ser68 and 238 ordered water molecules. The resulting molecular stereochemistry is consistent with a well-refined crystal structure with co-ordinate accuracy in the range of 0.10 to 0.15 A for the well-ordered regions of the molecule (B less than 15 A2). For the enzyme portion of the zymogen, the root-mean-square difference in C alpha atom co-ordinates with the refined porcine pepsin structure is 0.90 A (284 common atoms) and with the C alpha atoms of penicillopepsin it is 1.63 A (275 common atoms). The additional 44 N-terminal amino acids of the prosegment (Leu1p to Leu44p, using the letter p after the residue number to distinguish the residues of the prosegment) adopt a relatively compact structure consisting of a long beta-strand followed by two approximately orthogonal alpha-helices and a short 3(10)-helix. Intimate contacts, both electrostatic and hydrophobic interactions, are made with residues in the pepsin active site. The N-terminal beta-strand, Leu1p to Leu6p, forms part of the six-stranded beta-sheet common to the aspartic proteinases. In the zymogen the first 13 residues of pepsin, Ile1 to Glu13, adopt a completely different conformation from that of the mature enzyme. The C alpha atom of Ile1 must move approximately 44 A in going from its position in the inactive zymogen to its observed position in active pepsin. Electrostatic interactions of Lys36pN and hydrogen-bonding interactions of Tyr37pOH, and Tyr90H with the two catalytic aspartate groups, Asp32 and Asp215, prevent substrate access to the active site of the zymogen. We have made a detailed comparison of the mammalian pepsinogen fold with the fungal aspartic proteinase fold of penicillopepsin, used for the molecular replacement solution. A structurally derived alignment of the two sequences is presented.     

Refined 1.8 A crystal structure of the lambda repressor-operator complex.

The crystal structure of the lambda repressor-operator complex has been refined to an R-factor of 18.9% at 1.8 A resolution. This refinement, using data collected at low temperature, has revealed the structure of the N-terminal arm and shows that the interactions of repressor with the two halves of the pseudo-symmetric operator site are significantly different. The two halves of the complex are most similar near the outer edge of the operator site (in a region where the lambda and 434 repressors make similar contacts), but they become increasingly different toward the center of the operator. There are striking differences near the center of the site where it appears that the arm makes significant contacts to only one half of the DNA site. This suggested a new way of aligning the operator sites in phage lambda. The high resolution structure confirms many of the previously noted features of the complex, but also reveals a number of new protein-DNA contacts. It also gives a better view of the extensive H-bonding networks that couple contacts made by different residues and different regions of the protein, and reveals important new details about the helix-turn-helix (HTH) region, and the positions of many water molecules in the complex.     

Refined structure of cytochrome c3 at 1.8 A resolution

The structure of cytochrome c3 from the sulfate-reducing bacterium Desulfovibrio vulgaris Miyazaki has been successfully refined at 1.8 A resolution. The crystallographic R factor is 0.176 for 9907 significant reflections. The isotropic temperature factors of individual atoms were refined and a total of 47 water molecules located on the difference map were incorporated in the refinement. The four heme groups are closely packed, with adjacent pairs of heme planes being nearly perpendicular to each other. The fifth and the sixth ligands of the heme iron atoms are histidine residues with N epsilon 2-Fe distances ranging from 1.88 A to 2.12 A. The histidine co-ordination to the heme iron is different for each heme group. The heme groups are all highly exposed to solvent, although the actual regions exposed differ among the hemes. The four heme groups are located in different environments, and the heme planes are deformed from planarity. The differences in the heme structures and their environments indicate that the four heme groups are non-equivalent. The chemical as well as the physical properties of cytochrome c3 should be interpreted in terms of the structural non-equivalence of the heme groups. The characteristic secondary structural non-equivalence of the heme groups. The characteristic secondary structures of the polypeptide chain of this molecule are three short alpha-helices, two short beta-strands and ten reverse turns.     

Inhibition of subtilisin BPN' by reaction site P1 mutants of Streptomyces subtilisin inhibitor

It has been shown that the P1 site (the center of the reactive site) of protease inhibitors corresponds to the specificity of the cognate protease, and consequently specificity of Streptomyces subtilisin inhibitor (SSI) can be altered by substitution of a single amino acid at the P1 site. In this paper, to investigate whether similar correlation between inhibitory activity of mutated SSI and substrate preference of protease is observed for subtilisin BPN', which has broad substrate specificity, a complete set of mutants of SSI at the reaction site P1 (position 73) was constructed by cassette and site-directed mutagenesis and their inhibitory activities toward subtilisin BPN' were measured. Mutated SSIs which have a polar (Ser, Thr, Gln, Asn), basic (Lys, Arg), or aromatic amino acid (Tyr, Phe, Trp, His), or Ala or Leu, at the P1 site showed almost the same strong inhibitory activity toward subtilisin as the wild type (Met) SSI. However, the inhibitory activity of SSI variants with an acidic (Glu, Asp), or a beta-branched aliphatic amino acid (Val, Ile), or Gly or Pro, at P1 was decreased. The values of the inhibitor constant (Ki) of mutated SSIs toward subtilisin BPN' were consistent with the substrate preference of subtilisin BPN'. A linear correlation was observed between log(1/Ki) of mutated SSIs and log(1/Km) of synthetic substrates. These results demonstrate that the inhibitory activities of P1 site mutants of SSI are linearly related to the substrate preference of subtilisin BPN', and indicate that the binding mode of the inhibitors with the protease may be similar to that of substrates, as in the case of trypsin and chymotrypsin.(ABSTRACT TRUNCATED AT 250 WORDS)     

Crystal structure of a bacterial protein proteinase inhibitor (Streptomyces subtilisin inhibitor) at 2.6 A resolution

The crystal structure of a bacterial protein proteinase inhibitor (Streptomyces subtilisin inhibitor) was solved at 2·6 Å resolution. Each subunit of the dimeric inhibitor has a five-stranded antiparallel β-sheet and two short α-helices. The subunit-subunit interface formed by a stack of two β-sheets provided by the two subunits resembles the dimer-dimer interface of concanavalin A. Conformation of the reactive site around the scissible bond Met73-Val74 seems very rigid. Between bovine pancreatic trypsin inhibitor (Kunitz) and the Streptomyces inhibitor, the reactive site conformations are almost identical with each other from the P₂ to P₂′ residues, while between the soybean trypsin inhibitor (Kunitz) and the Streptomyces inhibitor they are similar from the P₂ to P₁′ residues. There are overall similarities in conformation extending from the P₃ to P₂′ residues between the Streptomyces inhibitor and a hypothetical substrate presumed (Robertus et al., 1972b) to be bound to subtilisin BPN′ in a productive binding mode. Apart from the reactive site, there seems to be no structural relationship among the Streptomyces, bovine pancreatic and soybean inhibitors, suggesting their convergent evolution from separate ancestral proteins.     

Refined crystal structure of carboxypeptidase A at 1.54 A resolution

The crystal structure of bovine carboxypeptidase A (Cox) has been refined at 1.54 A resolution using the restrained least-squares algorithm of Hendrickson & Konnert (1981). The crystallographic R factor (formula; see text) for structure factors calculated from the final model is 0.190. Bond lengths and bond angles in the carboxypeptidase A model have root-mean-square deviations from ideal values of 0.025 A and 3.6 degrees, respectively. Four examples of a reverse turn like structure (the "Asx" turn) requiring an aspartic acid or asparagine residue are observed in this structure. The Asx turn has the same number of atoms as a reverse turn, but only one peptide bond, and the hydrogen bond that closes the turn is between the Asx side-chain CO group and a main-chain NH group. The distributions of CO-N and NH-O hydrogen bond angles in the alpha-helices and beta-sheet structures of carboxypeptidase A are centered about 156 degrees. A total of 192 water molecules per molecule of enzyme are included in the final model. Unlike the hydrogen bonding geometry observed in the secondary structure of the enzyme, the CO-O(wat) hydrogen bond angle is distributed about 131 degrees, indicating the role of the lone pair electrons of the carbonyl oxygen in the hydrogen bond interaction. Twenty four solvent molecules are observed buried within the protein. Several of these waters are organized into hydrogen-bonded chains containing up to five waters. The average temperature factor for atoms in carboxypeptidase A is 8 A2, and varies from 5 A2 in the center of the protein, to over 30 A2 at the surface.     

Calcium-mediated thermostability in the subtilisin superfamily: the crystal structure of Bacillus Ak.1 protease at 1.8 A resolution

Proteins of the subtilisin superfamily (subtilases) are widely distributed through many living species, where they perform a variety of processing functions. They are also used extensively in industry. In many of these enzymes, bound calcium ions play a key role in protecting against autolysis and thermal denaturation. We have determined the crystal structure of a highly thermostable protease from Bacillus sp. Ak.1 that is strongly stabilized by calcium. The crystal structure, determined at 1.8 A resolution (R=0. 182, Rfree=0.247), reveals the presence of four bound cations, three Ca(2+) and one Na(+). Two of the Ca(2+) binding sites, Ca-1 and Ca-2, correspond to sites also found in thermitase and the mesophilic subtilisins. The third calcium ion, however, is at a novel site that is created by two key amino acid substitutions near Ca-1, and has not been observed in any other subtilase. This site, acting cooperatively with Ca-1, appears to give substantially enhanced thermostability, compared with thermitase. Comparisons with the mesophilic subtilisins also point to the importance of aromatic clusters, reduced hydrophobic surface and constrained N and C termini in enhancing the thermostability of thermitase and Ak.1 protease. The Ak.1 protease also contains an unusual Cys-X-Cys disulfide bridge that modifies the active site cleft geometry.     

The refined crystal structure of subtilisin Carlsberg at 2.5 A resolution

We report here the X-ray crystal structure of native subtilisin Carlsberg, solved at 2.5 A resolution by molecular replacement and refined by restrained least squares to a crystallographic residual (Formula see text): of 0.206. we compare this structure to the crystal structure of subtilisin BPN'. We find that, despite 82 amino acid substitutions and one deletion in subtilisin Carlsberg relative to subtilisin BPN', the structures of these enzymes are remarkably similar. We calculate an r.m.s. difference between equivalent alpha-carbon positions in subtilisin Carlsberg and subtilisin BPN' of only 0.55 A. This confirms previous reports of extensive structural homology between these two subtilisins based on X-ray crystal structures of the complex of eglin-c with subtilisin Carlsberg [McPhalen, C.A., Schnebli, H.P. and James, M.N.G. (1985) FEBS Lett., 188, 55; Bode, W., Papamokos, E. and Musil, D. (1987) Eur. J. Biochem., 166, 673-692]. In addition, we find that the native active sites of subtilisins Carlsberg and BPN' are virtually identical. While conservative substitutions at residues 217 and 156 may have subtle effects on the environments of substrate-binding sites S1' and S1 respectively, we find no obvious structural correlate for reports that subtilisins Carlsberg and BPN' differ in their recognition of model substrates. In particular, we find no evidence that the hydrophobic binding pocket S1 in subtilisin Carlsberg is 'deeper', 'narrower' or 'less polar' than the corresponding binding site in subtilisin BPN'.     

Refined 1.2 A crystal structure of the complex formed between subtilisin Carlsberg and the inhibitor eglin c. Molecular structure of eglin and its detailed interaction with subtilisin.

The crystal structure of the complex formed between eglin c, an elastase inhibitor from the medical leech, and subtilisin Carlsberg has been determined at 1.2 A resolution by a combination of Patterson search methods and isomorphous replacement techniques. The structure has been refined to a crystallographic R-value of 0.18 (8-1.2 A). Eglin consists of a four-stranded beta-sheet with an alpha-helical segment and the protease-binding loop fixed on opposite sides. This loop, which contains the reactive site Leu45I--Asp46I, is mainly held in its conformation by unique electrostatic/hydrogen bond interactions of Thr44I and Asp46I with the side chains of Arg53I and Arg51I which protrude from the hydrophobic core of the molecule. The conformation around the reactive site is similar to that found in other proteinase inhibitors. The nine residues of the binding loop Gly40I--Arg48I are involved in direct contacts with subtilisin. In this interaction, eglin segment Pro42I--Thr44I forms a three-stranded anti-parallel beta-sheet with subtilisin segments Gly100--Gly102 and Ser125--Gly127. The reactive site peptide bond of eglin is intact, and Ser221 OG of the enzyme is 2.81 A apart from the carbonyl carbon.     

Complex of subtilisin BPN' with Streptomyces subtilisin inhibitor. Complex formation concomitant with change in reducibility of disulfide bonds in the inhibitor

Structure of the complex of Streptomyces subtilisin inhibitor (SSI) with subtilisin BPN' was studied by examining the thermal denaturation and reducibility of disulfide bonds. The denaturation temperature of the complex was significantly higher than that of the enzyme. Two disulfide bonds localized in the inhibitor side were completely reduced in the complex, whereas only one of them was reduced in the free SSI. Gel filtration of the reduced complex solution showed clearly that the main products of reduction of the complex were two peptide fragments of SSI divided at the active site. The resistive disulfide bond in the complexed inhibitor became accessible as a result of a large conformational change due to splitting of the half-reduced inhibitor.     

Refined crystal structure of ascorbate oxidase at 1.9 A resolution.

The crystal structure of the fully oxidized form of ascorbate oxidase (EC 1.10.3.3) from Zucchini has been refined at 1.90 A (1 A = 0.1 nm) resolution, using an energy-restrained least-squares refinement procedure. The refined model, which includes 8764 protein atoms, 9 copper atoms and 970 solvent molecules, has a crystallographic R-factor of 20.3% for 85,252 reflections between 8 and 1.90 A resolution. The root-mean-square deviation in bond lengths and bond angles from ideal values is 0.011 A and 2.99 degrees, respectively. The subunits of 552 residues (70,000 Mr) are arranged as tetramers with D2 symmetry. One of the dyads is realized by the crystallographic axis parallel to the c-axis giving one dimer in the asymmetric unit. The dimer related about this crystallographic axis is suggested as the dimer present in solution. Asn92 is the attachment site for one of the two N-linked sugar moieties, which has defined electron density for the N-linked N-acetyl-glucosamine ring. Each subunit is built up by three domains arranged sequentially on the polypeptide chain and tightly associated in space. The folding of all three domains is of a similar beta-barrel type and related to plastocyanin and azurin. An analysis of intra- and intertetramer hydrogen bond and van der Waals interactions is presented. Each subunit has four copper atoms bound as mononuclear and trinuclear species. The mononuclear copper has two histidine, a cysteine and a methionine ligand and represents the type-1 copper. It is located in domain 3. The bond lengths of the type-1 copper centre are comparable to the values for oxidized plastocyanin. The trinuclear cluster has eight histidine ligands symmetrically supplied from domain 1 and 3. It may be subdivided into a pair of copper atoms with histidine ligands whose ligating N-atoms (5 NE2 atoms and one ND1 atom) are arranged trigonal prismatic. The pair is the putative type-3 copper. The remaining copper has two histidine ligands and is the putative spectroscopic type-2 copper. Two oxygen atoms are bound to the trinuclear species as OH- or O2- and bridging the putative type-3 copper pair and as OH- or H2O bound to the putative type-2 copper trans to the copper pair. The bond lengths within the trinuclear copper site are similar to comparable binuclear model compounds. The putative binding site for the reducing substrate is close to the type-1 copper.(ABSTRACT TRUNCATED AT 400 WORDS)     

Refined structure of dienelactone hydrolase at 1.8 A

The structure of dienelactone hydrolase (DLH) from Pseudomonus sp. B13, after stereochemically restrained least-squares refinement at 1.8 A resolution, is described. The final molecular model of DLH has a conventional R value of 0.150 and includes all but the carboxyl-terminal three residues that are crystallographically disordered. The positions of 279 water molecules are included in the final model. The root-mean-square deviation from ideal bond distances for the model is 0.014 A and the error in atomic co-ordinates is estimated to be 0.15 A. DLH is a monomeric enzyme containing 236 amino acid residues and is a member of the beta-ketoadipate pathway found in bacteria and fungi. DLH is an alpha/beta protein containing seven helices and eight strands of beta-pleated sheet. A single 4-turn 3(10)-helix is seen. The active-site Cys123 residues at the N-terminal end of an alpha-helix that is peculiar in its consisting entirely of hydrophobic residues (except for a C-terminal lysine). The beta-sheet is composed of parallel strands except for strand 2, which gives rise to a short antiparallel region at the N-terminal end of the central beta-sheet. The active-site cysteine residue is part of a triad of residues consisting of Cys123, His202 and Asp171, and is reminiscent of the serine/cysteine proteases. As in papain and actinidin, the active thiol is partially oxidized during X-ray data collection. The positions of both the reduced and the oxidized sulphur are described. The active site geometry suggests that a change in the conformation of the native thiol occurs upon diffusion of substrate into the active site cleft of DLH. This enables nucleophilic attack by the gamma-sulphur to occur on the cyclic ester substrate through a ring-opening reaction.     

Protein structure refinement: Streptomyces griseus serine protease A at 1.8 A resolution.

The crystal structure of the bacterial serine protease from Streptomyces griseus (SGPA) has been refined at 1.8 Å resolution by a restrained parameter least-squares procedure (Konnert, 1976) to a conventional R factor of 0.139 for 12662 statistically significant reflections [I > 3σ(I)]. The number of variable parameters in the final model was 5912 which included positional and individual thermal parameters of the enzyme, and positions, B factors and occupancies of 175 solvent molecules. The algorithm used for this refinement allows for the simultaneous restraint on bond distances and distances related to interbond angles, the coplanarity of atoms in planar groups, the conservation of chirality of asymmetric centres, non-bonded contact distances, conformational torsional angles and individual isotropic temperature factors.The refined structure of SGPA differs from ideal bond lengths by an overall root-mean-square deviation of 0.02 Å; the corresponding value for angle distances is 0.038 Å. Comparison of the phase angles for the shell of data, 8.0 to 2.8 Å, between the multiple isomorphous replacement phases (Brayer et al., 1978a) and the refined phases, indicates an overall difference (r.m.s.) of 56.6 °. The average conformational angle of the peptide bond (ω) is 179.7 ° (root-mean-square deviation ± 2.5 °) for the 180 peptide bonds of SGPA. Of the 175 solvent molecules included during the course of the refinement, 22 with occupancies ranging from 1.00 to 0.38 are located in the active site and the substrate binding region. It was not until these water molecules were included in the refinement process that the active Ser195 adopted its final conformation (χ¹ = ?77 °). The resulting distance from O^γ of Ser195 to N^ε2 of His57 is 3.1 Å, which, when taken with the observed distortion from linearity (50 °), indicates a rather weak interaction.     

Refined crystal structure of Cd, Zn metallothionein at 2.0 A resolution

The crystal structure of Cd5,Zn2-metallothionein from rat liver has been refined at 2.0 A resolution of a R-value of 0.176 for all observed data. The five Cd positions in the asymmetric unit of the crystal create a pseudo-centrosymmetric constellation about a crystallographic 2-fold axis. Consequently, the distribution of anomalous differences is almost ideally centrosymmetric. Therefore, the previously reported metal positions and the protein model derived therefrom are incorrect. Direct methods were applied to the protein amplitudes to locate the Cd positions. The new positions were used to calculate a new electron density map based on the Cd anomalous scattering and partial structure to model the metal clusters and the protein. Phases calculated from this model predict the positions of three sites in a (NH4)2WS4 derivative. Single isomorphous replacement phases calculated with these tungsten sites confirm the positions of the Cd sites from the new direct methods calculations. The refined metallothionein structure has a root-mean-square deviation of 0.016 A from ideality of bonds and normal stereochemistry of phi, phi and chi torsion angles. The metallothionein crystal structure is in agreement with the structures for the alpha and beta domains in solution derived by nuclear magnetic resonance methods. The overall chain folds and all metal to cysteine bonds are the same in the two structure determinations. The handedness of a short helix in the alpha-domain (residues 41 to 45) is the same in both structures. The crystal structure provides information concerning the metal cluster geometry and cysteine solvent accessibility and side-chain stereochemistry. Short cysteine peptide sequences repeated in the structure adopt restricted conformations which favor the formation of amide to sulfur hydrogen bonds. The crystal packing reveals intimate association of molecules about the diagonal 2-fold axes and trapped ions of crystallization (modeled as phosphate and sodium). Variation in the chemical and structural environments of the metal sites is in accord with data for metal exchange reactions in metallothioneins.     

Refined crystal structure of type III chloramphenicol acetyltransferase at 1.75 A resolution

High level bacterial resistance to chloramphenicol is generally due to O-acetylation of the antibiotic in a reaction catalysed by chloramphenicol acetyltransferase (CAT, EC 2.3.1.28) in which acetyl-coenzyme A is the acyl donor. The crystal structure of the type III enzyme from Escherichia coli with chloramphenicol bound has been determined and refined at 1.75 A resolution, using a restrained parameter reciprocal space least squares procedure. The refined model, which includes chloramphenicol, 204 solvent molecules and two cobalt ions has a crystallographic R-factor of 18.3% for 27,300 reflections between 6 and 1.75 A resolution. The root-mean-square deviation in bond lengths from ideal values is 0.02 A. The cobalt ions play a crucial role in stabilizing the packing of the molecule in the crystal lattice. CAT is a trimer of identical subunits (monomer Mr 25,000) and the trimeric structure is stabilized by a number of hydrogen bonds, some of which result in the extension of a beta-sheet across the subunit interface. Chloramphenicol binds in a deep pocket located at the boundary between adjacent subunits of the trimer, such that the majority of residues forming the binding pocket belong to one subunit while the catalytically essential histidine belongs to the adjacent subunit. His195 is appropriately positioned to act as a general base catalyst in the reaction, and the required tautomeric stabilization is provided by an unusual interaction with a main-chain carbonyl oxygen.