The enzymatic inactivation of the antibiotic chloramphenicol (Cm) is a well-known mechanism of bacterial resistance. The chloramphenicol acetyltransferases (CATs) are the enzymes responsible for the resistance to Cm. CATs inactivate Cm by converting it to 1-acetoxy Cm, 3-acetoxy Cm, or 1,3-diacetoxy Cm. These Cm derivatives are no longer able to interrupt the function of the ribosomal peptidyl-transferase, and are thus unable to inhibit bacterial growth and survival.1, 2
The CAT enzymes can be divided into two classes: CATA (the classical CAT) and CATB. The latter group of enzymes is encoded by Gram-negative bacteria.3–6 CATBs and paralogous acetyltransferases obtained from Gram-positive bacteria7–9 have been classified as a novel protein family, the xenobiotic acetyltransferase (XAT) family.10 In common with the CATAs, the proteins in the CATB family also form homo-trimers with similar molecular masses. Although the CATBs have no primary sequence similarities with the CATAs, both families mediate acetyl-transfer reactions by the use of a similar biochemical mechanism. Moreover, CATBs have a common motif of multiple repeats of a six-residue consensus sequence known as an “isoleucine patch”,11 which is an important structural determinant of an unusual left-handed β-helical domain and is characterized by the frequent occurrence of an aliphatic residue (Leu, Ile, or Val) at position I, a small residue (Ser, Thr, Ala, or Val) at position I+4 and Gly at position I+1.10–12 Proteins that contain tandem repeated copies of the above patch are called hexapeptide proteins, most of which are found in microorganisms and higher plants and display catalytic functions as acyltransferases.10, 13
In the CATB family, only one member's crystal structure has been reported (CATB7, also known as PaXAT).13 Structurally similar to other hexapeptide proteins,14–16 CATB7 is trimeric and its active sites are located at the interface between two monomers.
We report here the crystal structure at 2.5 Å resolution of CATB2, an enzyme encoded by the multiresistance transposon Tn2424 from Escherichia coli. This significant improvement of our CATB2 structure (2.5 Å) from the known CATB7 structure (3.2 Å) may provide more details on the active site of the enzyme.
Materials and methods.
Purification and crystallization:
The CATB2 was overexpressed and purified as previously described.17 The CATB2 apoenzyme was crystallized using vapor-diffusion techniques in hanging drops. The protein sample was concentrated to 8 mg/ml in a buffer containing 10 mM TEA (triethanolamine) (pH 7.7), 0.5 mM EDTA, and 0.1 mM PMSF. The crystals of CATB2 were obtained by mixing equal volumes of the protein sample and a reservoir solution containing 12 mM NiCl2, 6.5% PEG-8k, and 1 mM NaN3 in a buffer of 0.1 M MES at pH 6.7.
Data collection and processing:
Data collection was performed using synchrotron radiation on beamline X8C at Brookhaven National Laboratory, New York. Complete data were collected to 2.5 Å resolution. The diffraction data were processed using the HKL program suite.18
Initial phases were obtained by the molecular replacement method with the program CNS19 using the diffraction data between 15 and 4.0 Å. The structure of the hexapeptide xenobiotic acetyltransferase from Pseudomonas aeruginosa PA103 (PaXAT, PDB entry code: 2XAT), omitting solvent molecules and ligands, was used as a model in the rotation and translation search.
Model rebuilding and refinement.
Refinement was carried out using the program Refmac from CCP4,20 and model building was done with the O program.21 Noncrystallographic symmetry constraints were employed. The refinement statistics are listed in Table I.
Table I. Refinement Statistics
Numbers in parentheses are related to the highest resolution shell.
There are two subunits in the asymmetric unit and each subunit includes all residues except the first one (from residue 2 to 210). The model shows good stereochemistry with root mean square (RMS) deviations of bond lengths and bond angles of 0.006 Å and 1.25°, respectively. Out of all 354 non-glycine and non-proline residues, the Ramachandran plot positions 302 residues in the most favored regions, 44 residues in additional allowed regions and 8 residues in generously allowed regions.22 The solvent content is estimated to be 68.8%, which is considered high in general but lower than that reported for the PaXAT used as the model for molecular replacement.13 Because of the significant sequence identity (72.9%) between PaXAT and CATB2, the present CATB2 model displays high similarity to the PaXAT structure, as indicated by the RMS deviation of 0.56 Å between the paired alpha-carbon atoms for residues 5–205.
Left-handed β helix domain.
Figure 1(A) illustrates the crystal structure of trimeric CATB2. Each subunit contains five copies of a hexapeptide repeat sequence, which form an unusual left-handed β-helix domain (LβH). The LβH domain is interrupted by an extended loop from 73 to 110. In a closer view to each subunit in the trimer, five stacking triangular coils can be clearly defined. Each coil is composed of three β-strands and three tight turns.
In the PaXAT structure, the substrate and cofactor binding sites were reported to be located in a short tunnel formed by the extended loop from subunit B and residues from the LβH domains of both the A and B subunits.13 We also find such a short tunnel in the CATB2 structure. Since the active sites of PaXAT and CATB2 are very similar [Fig. 1(B)], it is reasonable to assume that the substrate (Cm) and the cofactor (acetyl-CoA) adopt similar binding modes in both enzymes. His-79, the critical residue in the catalysis as suggested by previous structural comparisons and structure–function studies,10–13 can be well superimposed in PaXAT and CATB2. There are relatively few shifts of the identical side chains between the overlapped structures, with the exception of the Phe-9 side chain, which is shifted by about 2.0 Å. This, of course, can be explained by the presence or absence of Cm, since this residue is near the aromatic group of the substrate. Nonetheless, a series of residues in both the substrate and cofactor binding sites have been substituted in CATB2 when compared to PaXAT.
At the Cm binding site of CATB2, four residues (Lys-10, Leu-48, Tyr-91, Phe-105) are different in CATB2 as compared to PaXAT. The side chain of residue 10 (Lys-10 in CATB2 and Arg-10 in PaXAT, respectively) is not well located due to the lack of clear electron density in both enzymes. Since the side chain of Arg-10 in PaXAT is quite flexible even in the presence of well-defined substrate, it is unlikely that this side chain interacts directly with Cm. A possible role of Lys-10 in CATB2 is to shield Cm from the solvent via its interaction with the side chain of the negatively charged residue Glu94 from the nearby subunit. Similarly to the case of residue 10, the substitution at residue 48 (Leu-48 in CATB2 and Met-48 in PaXAT) does not bring about significant differences considering the size and hydrophobicity of both side chains. It is of interest to note the substitutions of residues 91 and 105 (Tyr-91 and Phe-105 in CATB2 corresponding to Phe-91 and Tyr-105 in PaXAT). Inspection of the substrate binding region reveals that one more hydrogen bond between Cm and the enzyme could take place because of the presence of Tyr-91 in CATB2. However, the difference at residue 105 has no such effects because the side chain of this residue does not point toward the substrate, even in the structure of the PaXAT-Cm complex.
At the acetyl-CoA binding site there are five substitutions, three of which are located in the binding region of the 3′-phosphate ADP moiety of the cofactor. These three substitutions include residue 159 (Arg → Lys), residue 145 (Gly → Lys), and residue 144 (Thr → Ala) [Fig. 1(B)]. In PaXAT, Arg-159 contributes to the cofactor binding through its interaction with the negatively charged phosphate group on the adenine ribose of acetyl-CoA, which would be maintained in CATB2 since the replaced Lys-159 is also a positively charged residue. In contrast to the replacement of Arg-159 by Lys-159, the change of residue 145 is more significant in that the glycine residue in PaXAT has been substituted by a positively charged lysine in CATB2. Upon the superimposition of the active sites of both enzymes, Lys-145 is found to be located in the vicinity of the phosphate group of acetyl-CoA. Thus it is plausible that Lys-145 could also contribute to the binding of the cofactor. Therefore, two positively charged residues (Lys-159 and Lys-145) at this site could be important determining factors for the binding of acetyl-CoA in CATB2. In PaXAT, the side chain of Thr-144 interacts with the ribose 2′-hydroxyl of the cofactor, but such hydrogen-bonding interaction is not possible in CATB2 because the threonine residue has been substituted by alanine. The other two substitutions in the cofactor binding site are at residues 81 and 121, i.e., Ala-81 is substituted by tyrosine and Thr-121 is replaced by serine. The bulky side chain of Tyr-81 in CATB2 may stack on the pantetheine arm of the cofactor and thus stabilize cofactor binding. The change of residue 121 (Thr → Ser) may be negligible because the cofactor interacts with residue 121 through its main-chain nitrogen, not side-chain atoms. In summary, the substitutions of residue 145 and residue 81 could increase the affinity of the enzyme for the cofactor, while the substitution of residue 144 could impede the stable binding of cofactor. Accordingly, we could expect that the binding for acetyl-CoA would not undergo remarkable change.
Although we have already solved the structure of the enzyme-Cm complex form to 3.2 Å (data not shown), we were not able to define a clear position for the substrate due to insufficient electron density. To better understand the enzyme–ligand interactions and the catalytic mechanism, crystals of CATB2 complexes with Cm and/or acetyl-CoA at a higher resolution will be needed.
This study was supported by a Canadian Space Agency contract for protein crystallization and by the infrastructure grant 2003-ER-2481 from the ≪Fonds pour la Formation de Chercheurs et l'Aide à la Recherche du Québec ≫ (FCAR). Final data collection was carried out at the National Synchrotron Light Source, Brookhaven National Laboratory, supported by a joint grant from the National Science and Engineering Research Council and Medical Research Council of Canada for a “consortium for operation of a protein crystallographic synchrotron beam line.” The authors thank Dr. Peter H. Rehse for his useful suggestions.