Ribose phosphate isomerase (D-ribose-5-phosphate ketol-isomerase; EC 220.127.116.11; RPI), an enzyme involved in the first step of the non-oxidative branch of the pentose phosphate pathway, catalyzes the reversible conversion of ribose-5-phosphate to ribulose 5-phosphate. Two ribose phosphate isomerases, RpiA and RpiB, have been identified in Escherichia coli.1 RpiA, which is constitutively expressed, accounts for about 99% of the total RPI activity for strains grown in nutrient broth. E. coli strains defective in the rpiA gene are still able to use ribose as a carbon source due to the presence of the second RPI, a ribose-inducible RpiB.2 In contrast, the double mutant rpiA−: rpiB− shows severe inhibition of growth due to the accumulation of ribose-5-phosphate.3 RpiA, from E. coli, exists as a homodimer in solution with a subunit molecular weight of 22.8 kDa.4, 5 Despite the differences in the amino acid sequence and molecular weight, both RpiA and RpiB can efficiently catalyze the interconversion of ribose-5-phosphate and ribulose-5-phosphate.3 Sequence alignment of RPIs from various organisms shows a conservation of sequence around two regions (corresponding to RpiA)  GXG(T/S)GST  and  DGADE-(X)8-KGXG  throughout the family. In RPIs, the transfer of protons between the first and the second carbon atoms of the pentose-5′-phosphate substrate has been suggested to proceed through a single base mechanism involving a cis-enediol(ate) intermediate.6 Through site-directed mutagenesis, Asp87 of spinach RPI has been suggested to play the role of a general base in the interconversion of ribose-5-phosphate and ribulose-5-phosphate. A similar acid-base mechanism has been suggested to function in rabbit phosphoglucose isomerase7 and triosephosphate isomerase from Leishmania.8 Although sequence information is available for RPIs from various organisms, no structural information for this family of enzyme is available. To understand more about this family of proteins, we have determined the crystal structure of E. coli RpiA at 2.5 Å resolution.
Materials and methods.
The gene encoding E. coli RpiA was cloned and the protein, labeled with selenomethionine, was expressed and purified as a N-terminal GST fusion.9 The GST tag was removed by thrombin cleavage before crystallization trials. Two crystal forms were obtained by the hanging drop vapor diffusion by equilibrating drops containing 2 μL of protein (4.7 mg/mL) in buffer [50 mM Tris, pH 8.5, 400 mM NaCl, and 5 % (v/v) glycerol] and 2 μL reservoir solution consisting of either (a) 1.05 M trisodium citrate, 0.1 M HEPES buffer, pH 7.5 or (b) 1.2 M ammonium sulfate, 3% (v/v) isopropanol, respectively, suspended over 1 mL of reservoir solution. Crystals of the former condition belong to space group P1 with unit cell dimensions a = 42.6, b = 42.6, c = 60.8 Å, α = 101.3°, β = 90.5°, and γ = 98.8° (Z = 2), whereas the latter belong to orthorhombic space group C2221 with unit cell dimensions a = 70.3 Å, b = 71.8 Å, c = 193.1 Å (Z = 16). Further investigations were conducted by using the orthorhombic crystals. Before data collection, the crystals were soaked for 1 min in a cryoprotecting solution of 1.4 M ammonium sulfate, 3% (v/v) isopropanol, and 25% (v/v) glycerol and flash cooled to 100 K in a cold stream of N2 gas.
The structure of RpiA was determined by MAD phasing from a SeMet-labeled protein crystal (Table I). Data processing and scaling were performed with HKL2000.10 Twelve of the expected 14 Se sites in the asymmetric unit were found with the program SOLVE11 by using 2.5 Å resolution data-yielding phases with an overall figure of merit (FOM) of 0.42. Density modification with the program RESOLVE12 improved the quality of the map and the overall FOM increased to 0.60. Approximately 70% of the protein main-chain and 45% of the side-chain atoms were built automatically with RESOLVE. Further model building was performed by using program O.13 The complete model was refined against the high-energy remote wavelength data set using program CNS version 1.0.14 Noncrystallographic symmetry restraints were applied throughout the refinement. During refinement, 10% of the reflections were set aside to monitor Rfree. Water molecules were added in the last stage of refinement. The final model has an R-factor of 0.227 and Rfree of 0.270 for the data to 2.5 Å for all reflections. In the last round of refinement, test set reflections were also included (R-factor = 0.230). The final model contains all residues, from Met1 to Lys219 for each monomer, and 131 water molecules. The model has good stereochemistry and has no outliers in the Ramachandran plot as analyzed with the program PROCHECK.15 Coordinates of RpiA have been deposited in the Protein Data Bank, RCSB with the accession code 1LKZ.
Table I. Data Collection and Refinement Statistics
Number of reflections before merging.
Rsym = (Σ/Iobs − Iavg)ΣIavg.
Refinement performed by using data collected at the hard remote wavelength.
Rwork = (Σ/Fobs − Fcalc|)/ΣFobs.
Rfree = Rwork, but for a set of random set of 10% of the unique reflections.
There are two molecules of RpiA in the asymmetric unit. Each molecule has an overall α/β architecture with the size of ∼5 × 45 × 30 Å3 and is folded into two domains [Fig. 1(a)]. The larger domain consists of N-terminal residues Met1 to Phe129 and the C-terminal residues Ala199 to Lys219. This domain is composed of a six-stranded β-sheet, with five parallel β-strands and an antiparallel β-strand at the edge, flanked by three α-helices on one side and a single helix on the other side [Fig. 1(a)]. The smaller domain, containing residues Pro130 to Phe198, is made up of a four-stranded antiparallel β-sheet with two α-helices on one side [Fig. 1(a)] comprising a ferredoxin-like fold.
Dynamic Light Scattering (DLS) data of the purified RpiA, at a concentration of 4–5 mg/mL, indicated a species with a molecular weight of ∼50 kDa, which corresponds well to the dimeric form of the enzyme observed in the crystal and agrees with earlier reports.4, 5 The dimer has dimensions of 75 × 45 × 40 Å.3 The molecules within the dimer are related by twofold noncrystallographic symmetry. The surface of the monomer that is buried by dimer formation is ∼860 Å,2 which accounts for 10% of the total surface area. There are nine hydrogen bonds (<3.25 Å) between the subunits with most of them involving the small domain, which is largely involved in the subunit interactions.
RpiA is the first structure to be determined in the family of ribose phosphate isomerases. Both of its domains have structural homologues in other proteins as found by using the program DALI.16 The closest homologue to the large domain is glutaconate CoA-transferase (GCT)17 from Acidaminococcus fermentans with a root-mean-square deviation (RMSD) of 2.7 Å for 123 Cα pairs [Fig. 1(b)]. The residues 8–168 and 188–200 of the B-subunit of GCT overlap with the larger domain of RpiA. (PDB code 1poi). RpiA also shows similarity to 1-deoxy-D-xylulose-5-phosphate reductoisomerase18 from E. coli (RMSD = 3.4 Å for 94 Cα pairs; PDB code 1k5h), periplasmic D-ribose-binding protein19 of E. coli (RMSD = 3.7 Å for 103 Cα pairs; PDB code 2dri) and rabbit phosphoglucose isomerase7 (RMSD = 3.8 Å for 96 Cα pairs; PDB code 1dqr). In each case, the ordering of β-strands within the central β-sheet, 2-1-3-4-5-6, is maintained. The small domain shows fold similarity to ribosomal protein S620 from Thermus thermophilus with an RMSD of 2.2 Å for 65 Cα pairs (PDB code 1ris) and to bacterial elongation factor G (EF-G)21 from T. thermophilus with an RMSD of 2.3 Å for 68 Cα pairs (PDB code 1dar).
Potential catalytic residues have been probed by site-directed mutagenesis in the RPI from spinach leaves. Mutation of residues conserved in proteins from this family showed that Asp87 and Lys100 are crucial for enzymatic activity.6 Replacement of Asp87 by an alanine reduced activity by 105, affecting predominantly kcat, whereas the Lys100Ala replacement led to 104 reduction of activity affecting both kcat and KM.6 The side-chain of Asp90 is also important for catalysis, although its mutation to an alanine reduced activity only by a factor of 100. These residues correspond to Asp81, Asp84, and Lys94 of E. coli RpiA and are in close proximity to each other in the three-dimensional structure, delineating the putative location of the active site. These residues form a depression in the surface of the molecule that is easily accessible to the substrate molecule. The electrostatic potential surface calculated for RpiA indicates a basic character at the bottom of this depression (occupied by Lys94) and an acidic character along its sides.
Superposing the crystal structure of RpiA with that of phosphoglucose isomerase7 and periplasmic D-ribose binding protein from E. coli,19 with their ligands [Fig. 1(b)] showed that part of the substrate-binding pocket of these enzymes overlap with the proposed active site region of RpiA within the large domain. However, the binding site within RpiA appears to be shallower than in the other two proteins and is contained within the monomer, unlike in phosphoglucose isomerase where both subunits contribute to the active site. The locations of key catalytic residues of phosphoglucose isomerase are proximal, but not identical, to those proposed for RpiA. In GCT, the proposed substrate-binding site is located relatively distant from the expected binding site of RpiA, consistent with the different enzymatic activity of GCT.20 Overall, the sequence and structure comparisons, together with published mutagenesis data,6 indicate that the residues Asp81, Asp84, and Lys94 of E. coli RpiA form part of the active site and could facilitate a general acid-base catalytic mechanism for the isomerization step.
The authors thank Mr. R. Larocque for helping with cloning, Ms. Yunge Li for advice on crystallization, Dr. S. Raymond for database development, and Dr. J.D. Schrag for comments on the manuscript. We also thank Leonid Flaks (beamline X8C, NSLS) for assistance with data collection. This work was supported in part by CIHR grant No. 200103GSP-90094-GMX-CFAA-19924 to M.C.