Portal de Periódicos da CAPES

Crystal Structure of Escherichia coli Glucose-1-Phosphate Thymidylyltransferase (RffH) Complexed with dTTP and Mg2+

2002; Elsevier BV; Volume: 277; Issue: 46 Linguagem: Inglês

10.1074/jbc.m206932200

1083-351X

J. Sivaraman, Véronique Sauvé, Allan Matte, Mirosław Cygler,

Enzyme Structure and Function

The enzyme glucose-1-phosphate thymidylyltransferase (RffH), the product of therffh gene, catalyzes one of the steps in the synthesis of enterobacterial common antigen (ECA), a cell surface glycolipid found in Gram-negative enteric bacteria. InEscherichia coli two gene products, RffH and RmlA, catalyze the same enzymatic reaction and are homologous in sequence; however, they are part of different operons and function in different pathways. We report the crystal structure of RffH bound to deoxythymidine triphosphate (dTTP), the phosphate donor, and Mg2+, refined at 2.6 Å to an R-factor of 22.3% (R free = 28.4%). The crystal structure of RffH shows a tetrameric enzyme best described as a dimer of dimers. Each monomer has an overall α/β fold and consists of two domains, a larger nucleotide binding domain (residues 1–115, 222–291) and a smaller sugar-binding domain (116–221), with the active site located at the domain interface. The Mg2+ ion is coordinated by two conserved aspartates and the α-phosphate of deoxythymidine triphosphate. Its location corresponds well to that in a structurally similar domain of N-acetylglucosamine-1-phosphate uridylyltransferase (GlmU). Analysis of the RffH, RmlA, and GlmU complexes with substrates and products provides an explanation for their different affinities for Mg2+ and leads to a proposal for the dynamics along the reaction pathway. The enzyme glucose-1-phosphate thymidylyltransferase (RffH), the product of therffh gene, catalyzes one of the steps in the synthesis of enterobacterial common antigen (ECA), a cell surface glycolipid found in Gram-negative enteric bacteria. InEscherichia coli two gene products, RffH and RmlA, catalyze the same enzymatic reaction and are homologous in sequence; however, they are part of different operons and function in different pathways. We report the crystal structure of RffH bound to deoxythymidine triphosphate (dTTP), the phosphate donor, and Mg2+, refined at 2.6 Å to an R-factor of 22.3% (R free = 28.4%). The crystal structure of RffH shows a tetrameric enzyme best described as a dimer of dimers. Each monomer has an overall α/β fold and consists of two domains, a larger nucleotide binding domain (residues 1–115, 222–291) and a smaller sugar-binding domain (116–221), with the active site located at the domain interface. The Mg2+ ion is coordinated by two conserved aspartates and the α-phosphate of deoxythymidine triphosphate. Its location corresponds well to that in a structurally similar domain of N-acetylglucosamine-1-phosphate uridylyltransferase (GlmU). Analysis of the RffH, RmlA, and GlmU complexes with substrates and products provides an explanation for their different affinities for Mg2+ and leads to a proposal for the dynamics along the reaction pathway. Many of the currently available antibiotics target enzymes involved in the synthesis of bacterial cell wall components. Lipopolysaccharides (LPS) are unique and complex glycolipids embedded in the outer membrane of Gram-negative bacteria. They are made of galactose, mannose, rhamnose, 4-acetamido-4, 6-dideoxyglucose, andN-acetylglucosamine. LPS consists of three structural domains, namely the hydrophobic lipid A, the core oligosaccharide, and the O-antigenic polysaccharide (O-PS) (1Godfroid F. Cloeckaert A. Taminiau B. Danese I. Tibor A. de Bolle X. Mertens P. Letesson J.J. Res. Microbiol. 2000; 151: 655-668Crossref PubMed Scopus (71) Google Scholar). Surface polymers have essential roles in the survival of bacteria with the enzymes involved in their formation often found critical to virulence (2Finlay B.B. Falkow S. Microbiol. Mol. Biol. Rev. 1997; 61: 136-169Crossref PubMed Scopus (1182) Google Scholar) and potentially a source of novel targets for therapeutic intervention. The enterobacterial common antigen (ECA) is a unique cell surface glycolipid that is present in all gram–negative enteric bacteria (3Kuhn H.M. Meier-Dieter U. Mayer H. FEMS Microbiol. Rev. 1988; 4: 195-222Crossref PubMed Google Scholar). The genes involved in ECA synthesis cluster near the rfflocus. The product of the rffh gene is a glucose-1-phosphate thymidylyltransferase (EC 2.7.7.24) that catalyzes the reaction that combines dTTP with α-d-glucose 1-phosphate (G-1-P) 1The abbreviations used are: G-1-P, glucose 1-phosphate; GlmU, N-acetylglucosamine-1-phosphate uridylyltransferase; NDP, nucleoside diphosphate; dTTP, deoxythymidine triphosphate; RffH, glucose-1-phosphate thymidylyltransferase; rms, root-mean-squares; dTDP, dioxythymidine diphosphate to yield pyrophosphate and dTDP-glucose. This reaction constitutes the first step in the synthesis of l-rhamnose, a component of the cell walls of both Gram-negative and Gram-positive bacteria (4Shibaev V.N. Adv. Carbohydr. Chem. Biochem. 1986; 44: 277-339Crossref PubMed Scopus (100) Google Scholar). In the Escherichia coli K12 genome, the rffh gene is paralogous to the rfba gene that encodes protein RmlA. The rfba gene is contained within the rfb gene cluster responsible for the synthesis of O-antigen (5Reeves P.R. Hobbs M. Valvano M.A. Skurnik M. Whitfield C. Coplin D. Kido N. Klena J. Maskell D. Raetz C.R. Rick P.D. Trends Microbiol. 1996; 4: 495-503Abstract Full Text PDF PubMed Scopus (418) Google Scholar), and its product RmlA catalyzes the same reaction as RffH. This duplication of function is reflected in a ∼68% amino acid sequence identity between these two enzymes. In addition, a second pair of closely related genes that encode two dTDP-d-glucose 4,6-dehydratases are also present in both clusters: rffg(rff cluster) and rfbb (rfb cluster). The presence of closely related genes in the rff andrfb clusters (both involved in the biosynthesis ofO-polysaccharides) is not unique for these two families (5Reeves P.R. Hobbs M. Valvano M.A. Skurnik M. Whitfield C. Coplin D. Kido N. Klena J. Maskell D. Raetz C.R. Rick P.D. Trends Microbiol. 1996; 4: 495-503Abstract Full Text PDF PubMed Scopus (418) Google Scholar,6Marolda C.L. Valvano M.A. J. Bacteriol. 1995; 177: 5539-5546Crossref PubMed Google Scholar). A similar duplication of functions has been reported for the GDP-mannose biosynthesis genes rfbM and rfbK with genes cpsB and cpsG from the cps cluster involved in the biosynthesis of colonic acid (7Stevenson G. Lee S.J. Romana L.K. Reeves P.R. Mol. Gen. Genet. 1991; 227: 173-180Crossref PubMed Scopus (37) Google Scholar). The rfb cluster also encodes three other enzymes involved in the l-rhamnose synthesis pathway, rmlB, rmlC, andrmlD (8Tsukioka Y. Yamashita Y. Oho T. Nakano Y. Koga T. J. Bacteriol. 1997; 179: 1126-1134Crossref PubMed Google Scholar). The l-rhamnose biosynthetic pathway is not found in mammals, which makes these enzymes potential targets for development of antibacterial drugs. The structure of RmlB has also been described recently (9Allard S.T. Giraud M.F. Whitfield C. Graninger M. Messner P. Naismith J.H. J. Mol. Biol. 2001; 307: 283-295Crossref PubMed Scopus (108) Google Scholar, 10Allard S.T. Beis K. Giraud M.F. Hegeman A.D. Gross J.W. Wilmouth R.C. Whitfield C. Graninger M. Messner P. Allen A.G. Maskell D.J. Naismith J.H. Structure. 2002; 10: 81-92Abstract Full Text Full Text PDF PubMed Scopus (90) Google Scholar). Recently, three-dimensional structures of three enzymes belonging to the glucose-1-phosphate thymidylyltransferase family and their complexes with substrate(s), products, and inhibitors have been reported. These include RmlA from Pseudomonas aeruginosa(11Blankenfeldt W. Asuncion M. Lam J.S. Naismith J.H. EMBO J. 2000; 19: 6652-6663Crossref PubMed Scopus (162) Google Scholar), RmlA from E. coli (12Zuccotti S. Zanardi D. Rosano C. Sturla L. Tonetti M. Bolognesi M. J. Mol. Biol. 2001; 313: 831-843Crossref PubMed Scopus (93) Google Scholar), and RmlA from Salmonella enterica LT2 (13Barton W.A. Lesniak J. Biggins J.B. Jeffrey P.D. Jiang J. Rajashankar K.R. Thorson J.S. Nikolov D.B. Nat. Struct. Biol. 2001; 8: 545-551Crossref PubMed Scopus (119) Google Scholar). These studies revealed high structural similarity among these enzymes, localized their substrate binding sites, defined the arrangement of the substrates and the product in the active site, and showed that dTTP binds to the enzyme in a strained conformation. While the kinetic mechanism of these enzymes is now well established to follow a sequential ordered bi-bi mechanism (12Zuccotti S. Zanardi D. Rosano C. Sturla L. Tonetti M. Bolognesi M. J. Mol. Biol. 2001; 313: 831-843Crossref PubMed Scopus (93) Google Scholar, 13Barton W.A. Lesniak J. Biggins J.B. Jeffrey P.D. Jiang J. Rajashankar K.R. Thorson J.S. Nikolov D.B. Nat. Struct. Biol. 2001; 8: 545-551Crossref PubMed Scopus (119) Google Scholar, 14Sheu K.F. Richard J.P. Frey P.A. Biochemistry. 1979; 18: 5548-5556Crossref PubMed Scopus (76) Google Scholar) and proceeds by a S n2 nucleophilic attack of the phosphoryl group of glucose 1-phosphate at the α-phosphate of dTTP (11Blankenfeldt W. Asuncion M. Lam J.S. Naismith J.H. EMBO J. 2000; 19: 6652-6663Crossref PubMed Scopus (162) Google Scholar), these studies have not clarified the role in catalysis of a strictly required Mg2+ ion (12Zuccotti S. Zanardi D. Rosano C. Sturla L. Tonetti M. Bolognesi M. J. Mol. Biol. 2001; 313: 831-843Crossref PubMed Scopus (93) Google Scholar, 15Bernstein R.L. Robbins P.W. J. Biol. Chem. 1965; 240: 391-397Abstract Full Text PDF PubMed Google Scholar). Neither Mg2+ nor Mn2+ ions were observed in the crystal structures of P. aeruginosa or E. colienzymes despite the co-crystallization efforts, which at the same time showed that this ion is not necessary for binding of the first substrate, dTTP. In the structure of the S. enterica enzyme (13Barton W.A. Lesniak J. Biggins J.B. Jeffrey P.D. Jiang J. Rajashankar K.R. Thorson J.S. Nikolov D.B. Nat. Struct. Biol. 2001; 8: 545-551Crossref PubMed Scopus (119) Google Scholar) a feature in the electron density map was interpreted as a Mg2+ ion. However, this ion contacts only the β-phosphate of dTTP and would make no contacts with the phosphate of the second substrate, glucose 1-P (13Barton W.A. Lesniak J. Biggins J.B. Jeffrey P.D. Jiang J. Rajashankar K.R. Thorson J.S. Nikolov D.B. Nat. Struct. Biol. 2001; 8: 545-551Crossref PubMed Scopus (119) Google Scholar). Since the other two structures showed that dTTP binding does not require Mg2+, the strict requirements for this ion in catalysis can not be explained by its observed position in the S. enterica structure. Therefore, the binding site of Mg2+ and its precise role in catalysis is still uncertain in this family of enzymes. RffH and RmlA belong to the superfamily of glucose-1-phosphate NTP-transferases, which also includes bacterial uridylyltransferases (e.g. GalF, GalU) and adenylyltransferases (e.g. GlgC). The enzymes from this superfamily in E. coli show 22–26% sequence identity within the ∼240-residue-long catalytic domain compared with RffH. RffH also shows low sequence identity (19% for 230 amino acids) withE. coli N-acetylglucosamine-1-phosphate uridylyltransferase (GlmU). Indeed, the catalytic domains of all these enzymes share the same fold and have conserved key catalytic residues. All of them require Mg2+ for catalysis, therefore indicating a common catalytic mechanism. Yet, whereas no metal ion has been identified in the structures of RmlA enzymes, the three-dimensional structures of GlmU from various sources complexed with the NDP-sugar product in the presence of either 10 mm Mg2+ (16Kostrewa D. D'Arcy A. Takacs B. Kamber M. J. Mol. Biol. 2001; 305: 279-289Crossref PubMed Scopus (85) Google Scholar) or 2–10 mm Co (17Olsen L.R. Roderick S.L. Biochemistry. 2001; 40: 1913-1921Crossref PubMed Scopus (126) Google Scholar) contain a well ordered metal ion. We have succeeded in obtaining crystals of RffH from E. coliin the presence of deoxythymidine triphosphate (dTTP) and Mg2+, with both ligands clearly visible in the electron density maps. The location of the Mg2+ ion in our structure differs from that described in S. enterica RmlA, and in combination with the previously determined structures of RmlA complexes from the three species, our data provide a clear explanation for the essential role of Mg2+ in the catalytic mechanism of these enzymes. This Mg2+-binding site coincides with that observed in GlmU enzymes, providing additional evidence for a common catalytic mechanism. Comparison of all available structures allows us to make hypotheses regarding conformational flexibility at the active site during catalysis. The rffH gene was cloned into a derivative of the pET-15b vector (Amersham Biosciences) containing a thrombin cleavage site to obtain an in-frame N-terminal fusion with His6. Plasmid DNA was transformed into E. coli DL41 (18Hendrickson W.A. Horton J.R. LeMaster D.M. EMBO J. 1990; 9: 1665-1672Crossref PubMed Scopus (1008) Google Scholar) for selenomethionine protein production. The transformed bacteria were grown at 37° to anA 600 of ∼0.8 in defined LeMaster medium supplemented with 25 mg/liter of l-selenomethionine. A 1 liter culture was induced with 100 μmisopropyl-1-thio-β-d-galactopyranoside (IPTG) and the culture continued at room temperature for an additional 15 h. Cells were harvested by centrifugation (4000 × g, 4 °C, 25 min) and re-suspended in 40 ml of lysis buffer (50 mm Tris-HCl, pH 7.5, 0.4 m NaCl, 5% (w/v) glycerol, 20 mm imidazole, 10 mm β mercaptoethanol) in which one tablet of complete protease inhibitors (Roche Diagnostics, Laval, Canada) was dissolved. Cells were lysed by sonication on ice for a total of five 30s cycles with 45s between cycles. The lysate was cleared by centrifugation (100,000 ×g, 4 °C, 30 min). The protein supernatant was first loaded on an equilibrated 5-ml DEAE-Sepharose (Amersham Biosciences) column, and the eluted proteins were loaded on a 5-ml bed volume nickel-nitrilotriacetic acid (Qiagen). The column was washed with 50 mm imidazole and the bound proteins eluted with buffer containing 150 mm imidazole. Purified RffH ran as a single band on both SDS-PAGE and native PAGE gels, and its behavior in solution was evaluated by dynamic light scattering (DLS) using a DynaPro-801 molecular sizing instrument (Protein Solutions, Charlottesville, VA). Crystals of RffH were obtained with the intact His-tag in the presence of dTTP/Mg2+. Selenomethionine-labeled protein was concentrated to 10 mg/ml by ultrafiltration and crystallized by hanging drop vapor diffusion. A volume of 2 μl of protein solution in buffer (20 mm Tris, pH 7.5, 0.2 m NaCl, 5 mm dithiothreitol, 5 mm dTTP) was mixed with 2 μl of reservoir solution containing 16% (w/v) polyethylene glycol (PEG) 8000, 0.1 m 2-(N-morpholino)ethanesulfonic acid (MES) buffer, pH 6.5, and 0.4 m MgCl2 and suspended over the reservoir solution. Crystals grew to the size of ∼0.3 × 0.15 × 0.1 mm3 in 7 days. Diffraction data to 2.6 Å resolution were collected at 295 K on an Raxis IiC area detector mounted on a RU-300 rotating anode generator (Molecular Structure Corporation, The Woodlands, TX). These crystals belong to space group P21212 with a = 59.3 Å, b = 71.7 Å, c = 144.3 Å, and Z = 8. Data integration and scaling was accomplished using the programs Denzo and Scalepack (19Otwinowski Z. Minor W. Methods Enzymol. 1997; 276: 307-326Crossref PubMed Scopus (38617) Google Scholar). The crystal structure was determined by the molecular replacement method with the program Amore (20Navaza J. Acta Crystallogr. 1994; A50: 157-163Crossref Scopus (5030) Google Scholar) using the coordinates of RmlA from P. aeruginosa (11Blankenfeldt W. Asuncion M. Lam J.S. Naismith J.H. EMBO J. 2000; 19: 6652-6663Crossref PubMed Scopus (162) Google Scholar) (Protein Data Bank code 1G2V) as the search model. Both independent molecules showed as strong peaks in the rotation function. The rigid body refinement after determining the translation components gave a correlation coefficient of 0.202 and R = 52.6%. Non-crystallographic averaging was carried out with the program package RAVE (21Kleywegt G.J. Jones T.A. Acta Crystallogr. Sect. D Biol. Crystallogr. 1999; 55: 941-944Crossref PubMed Scopus (156) Google Scholar) to improve the quality of the electron density map. Several cycles of map fitting with the program O (22Jones T.A. Zhou J.Y. Cowan S.W. Kjeldgaard M. Acta Crystallogr. 1991; A47: 110-119Crossref Scopus (13014) Google Scholar) and refinement using the program CNS (23Brünger A.T. Adams P.D. Clore G.M. DeLano W.L. Gros P. Grosse-Kunstleve R.W. Jiang J.S. Kuszewski J. Nilges M. Pannu N.S. Read R.J. Rice L.M. Simonson T. Warren G.L. Acta Crystallogr. Sec. D. 1998; 54: 905-921Crossref PubMed Scopus (16979) Google Scholar) led to convergence. The final R-factor is 0.223 (R free = 0.284) for all measured data (R = 0.203 and R free = 0.261 for data with I>ς) to 2.6 Å resolution. The model is of good quality as judged by the low root-mean-squares (rms) deviation for bond lengths and bond angles from target values and the Ramachandran plot (Table I). Only one residue, Tyr-29, is found in the disallowed region. This residue is located in a tight turn and is well defined in the electron density map. Coordinates of RffH have been deposited to the Protein Data Bank with accession code 1MC3.Table ICrystallographic statisticsData collectionResolution range (Å)20.0–2.6Wavelength (Å)1.5418Observedhkl47,293Unique hkl17,698Completeness (%)90.3OverallI/ςI9.5a Rsym(%)8.8Refinement and quality of the modelResolution range (Å)20.0–2.6b Rfree (%) no. reflections0.284 (1661)c Rwork (%) no. reflections0.223 (14870)Root mean square deviation bond lengths (Å)0.01Root mean square deviation bond angles (°)1.8Ramachandran plotFavored regions (%)85.0Allowed regions (%)12.3Generously allowed regions (%)2.3Disallowed regions (%)0.4Average B-factors (Å2)Main chain atoms (no. atoms)49.1 (2336)Side chain atoms (no. atoms)48.6 (2004)Overall protein atoms (no. atoms)48.8 (4340)Water (no. atoms)43.8 (165)Ligand (no. atoms)38.0 (56)Metals (no. atoms)34.9 (2)aRsym = Σ∣Iobs −Iav∣/ΣIavg.bRwork = Σ∣Fo −Fc∣/ΣFo.cRfree = Rwork, but for a random test set of 10% of reflections not included in the refinement. Open table in a new tab aRsym = Σ∣Iobs −Iav∣/ΣIavg. bRwork = Σ∣Fo −Fc∣/ΣFo. cRfree = Rwork, but for a random test set of 10% of reflections not included in the refinement. The structure of RffH was determined by molecular replacement using the RmlA structure fromP. aeruginosa (11Blankenfeldt W. Asuncion M. Lam J.S. Naismith J.H. EMBO J. 2000; 19: 6652-6663Crossref PubMed Scopus (162) Google Scholar) as a starting model. There are two molecules of RffH in the asymmetric unit. The histidine tag is present in the protein sample used for crystallization. Only one histidine from the construct is visible in the density and is included in the model. The last two C-terminal residues, Gln-292–Tyr-293 are disordered. The final model in the asymmetric unit contains two RffH molecules, each consisting of residues from His(−1)–Met1 to Pro-291 for each molecule, one Mg2+ ion, one dTTP molecule, and a total of 165 water molecules. Refinement statistics are summarized in Table I. The RffH molecule belongs to the α/β fold class (Fig.1). The overall size of the monomer is ∼35 × 41 × 55 Å3. As the RffH structure is similar to that of RmlA (11Blankenfeldt W. Asuncion M. Lam J.S. Naismith J.H. EMBO J. 2000; 19: 6652-6663Crossref PubMed Scopus (162) Google Scholar, 12Zuccotti S. Zanardi D. Rosano C. Sturla L. Tonetti M. Bolognesi M. J. Mol. Biol. 2001; 313: 831-843Crossref PubMed Scopus (93) Google Scholar, 13Barton W.A. Lesniak J. Biggins J.B. Jeffrey P.D. Jiang J. Rajashankar K.R. Thorson J.S. Nikolov D.B. Nat. Struct. Biol. 2001; 8: 545-551Crossref PubMed Scopus (119) Google Scholar) we will not describe the structure in detail but provide only a general overview. The molecule is built from several β-strand/α-helix units that assemble into a central seven-stranded β-sheet, with six parallel and one antiparallel strands and helices lining both sides of the sheet. The C terminus is capped with an antiparallel bundle of three α-helices, each being two turns long. Two additional two-stranded β-sheets are formed, one along the edge of the central sheet, and the other on top, near the C-terminal ends of the central sheet. Dynamic light scattering (DLS) data of purified RffH measured at a protein concentration of ∼7 mg/ml indicates the presence in solution of a single species with a molecular weight of ∼120 KDa. This molecular weight corresponds to a tetrameric form of the enzyme, similar to that of RmlA proteins from E. coli, P. aeruginosa, and S. enterica. The arrangement of molecules in the crystal lattice clearly shows the presence of tetramers. Two independent molecules in the asymmetric unit form a dimer with the pseudo 2-fold axis nearly parallel to the crystallographic y axis and pack very closely with a dyad symmetry-related dimer (rotation along z axis) into a tetramer. Therefore, the tetramer has an approximate 222 symmetry. The size of the tetramer is ∼55 × 80 × 80 Å3. The surface area of monomer A is ∼11,650 Å2 of which ∼1,380 Å2 is in contact with monomer B and ∼625 Å2 contacts the symmetry-related monomer A'. In total, 17% of the monomer surface area is utilized for tetramer formation. Superposition of a monomer of RffH and RmlA gives an rms deviation of 1.25 Å (283 Cα atom pairs), 1.20 Å (293 Cα atom pairs), and 1.10 Å (281 Cα atom pairs) for E. coli, P. aeruginosa, and S. enterica, respectively. Comparison of the RffH tetramer with tetramers of RmlA shows that a better description of these tetramers is that of dimers of dimers (Fig.2). Molecules A and B have a larger interface then A and A' (symmetry related to A), and this A/B dimer overlaps well with the equivalent dimers of each of the three RmlA structures (rms of 1.45 Å). There is, however, a difference of ∼15o in the relative orientation and ∼5 Å translation of the dimers within a tetramer when comparing RffH with RmlA. This modification of the quaternary arrangement has important consequences. Namely, the second, presumably allosteric dTTP binding site found in RmlA tetramers that is located at the interface between the two dimers is no longer available for binding dTTP in the RffH tetramer. The sidechain of an arginine, which in RmlA forms a salt bridge with a glutamate (Arg-219–Glu-255) and provides a stacking surface for the thymidine base, is shifted in RffH (Arg-217) into the position occupied by dTTP in the former, while maintaining interactions with the equivalent Glu-253. In the RffH-dTTP complex the dTTP binds at the C-terminal ends of the middle strands β1 and β4 of the central β-sheet. The triphosphate backbone assumes a conformation nearly perpendicular to the plane of the ribose ring, with the α-phosphate (Pα) and β-phosphate (Pβ) oxygen atoms in an eclipsed conformation, indicating some conformational strain. The phosphate backbone makes numerous hydrogen bonds with the protein, compensating for its conformational strain. The γ-phosphate (Pγ) is particularly well anchored and binds in a pocket formed by the segment Gly-8–Arg-13, which encompasses the highly conserved motif (G)GXGXR(L) (Fig.3). This sequence is also conserved in the related NTP transferases GlmU and GalF and is similar to that found in the GlgC family. Pγ forms two hydrogen bonds to the guanidine group of Arg-13, one to Thr-12OG and through a bridging water molecule to an oxygen atom of Pα and weakly to Glu-194. A H-bond is formed between the Pβ-O-Pγ bridging oxygen and NH of Gly-11, while the Pα-O-Pβ bridging oxygen forms a H-bond to Lys-23NZ, a residue potentially involved in catalysis (11Blankenfeldt W. Asuncion M. Lam J.S. Naismith J.H. EMBO J. 2000; 19: 6652-6663Crossref PubMed Scopus (162) Google Scholar). This conformation of dTTP is very similar to that observed in the previous RmlA structures. In the E. coli RmlA complex with the reaction product, dTDP-glucose, a sulfate ion is located in the same place as Pγ of dTTP, indicating high specificity of this binding pocket for such ions. The recognition of a pyrimidine base is primarily achieved by interactions with the loop Gln-80–Gly-85 (Fig. 3). The residue Gln-80 makes one hydrogen bond with N3 of the base. The O4 atom of the base is also hydrogen bonded to the backbone NH groups of Asp-84 and Gly-85. This interaction may explain the absence of activity with deoxycytidine triphosphate (dCTP), where the amino group replaces the O4 carbonyl group, resulting in a loss of a hydrogen bond to the mainchain NH group. The O2atom forms a hydrogen bond to the NH of Gly-8 from the same segment that tightly binds the triphosphate. There is also a hydrogen bond between the O3′ of deoxyribose and the sidechain of Gln-24. Superposition of the various RmlA complexes with bound dTTP clearly shows that the binding of this substrate is the same in all of these enzymes. The electron density map for RffH shows significant density near the Pα of dTTP in each monomer located ∼2.2 Å from several neighboring oxygen atoms, and the coordination environment suggests that this is a bound Mg2+ ion. Our observation of bound Mg2+is unique for structures of RmlA enzymes and is likely caused by the high (400 mm) concentration of MgCl2 in the protein crystallization solution. Five ligands in the Mg2+ion coordination sphere are: an oxygen atom from Pα of dTTP, two oxygen atoms from the Asp-223 sidechain, one oxygen atom from the Asp-108 sidechain, and one water molecule (Fig. 3). The sixth ligand, on the solvent-exposed side of the ion, is most likely another water molecule. However, it was not clearly visible in the electron density at this resolution. The two aspartates coordinating the Mg2+ ion, distant in the linear sequence, are strictly conserved in the proteins from the RffH/RmlA family (PFAM entry PF00483, www.sanger.ac.uk/Software/Pfam/). The position of the Mg2+ ion corresponds almost exactly to the position of a metal ion in the structures of GlmU complexed with the product, strongly supporting the notion that this is the catalytically important ion. At the same time, the requirement of a much higher concentration of Mg2+ for binding as compared with GlmU indicates most likely a significantly lower K D for Mg2+in the former. The position of the Mg2+ ion observed here is different from that in the structure of S. enterica RmlA where this ion is bound near Pβ of dTTP, on the side opposite to the binding site of G-1-P. This ion is coordinated by the mainchain atoms and by the oxygen of a glutamine sidechain, not conserved in other RffH/RmlA sequences. The location of this putative Mg2+ ion far from the G-1-P binding site, the lack of conservation of the sidechain providing coordination, and a rather atypical ligand environment makes this binding site unlikely to correspond to that for the catalytically essential Mg2+ ion. It is well established that catalysis in the RmlA family involves S n2 nucleophilic attack of a phosphoryl oxygen atom of glucose-1-P at Pα of dTTP and requires the presence of Mg2+ ions (11Blankenfeldt W. Asuncion M. Lam J.S. Naismith J.H. EMBO J. 2000; 19: 6652-6663Crossref PubMed Scopus (162) Google Scholar, 12Zuccotti S. Zanardi D. Rosano C. Sturla L. Tonetti M. Bolognesi M. J. Mol. Biol. 2001; 313: 831-843Crossref PubMed Scopus (93) Google Scholar, 13Barton W.A. Lesniak J. Biggins J.B. Jeffrey P.D. Jiang J. Rajashankar K.R. Thorson J.S. Nikolov D.B. Nat. Struct. Biol. 2001; 8: 545-551Crossref PubMed Scopus (119) Google Scholar,15Bernstein R.L. Robbins P.W. J. Biol. Chem. 1965; 240: 391-397Abstract Full Text PDF PubMed Google Scholar). The presence of two conserved aspartate residues that act as ligands to the Mg2+ ion in our structure, reveals the catalytically essential Mg2+ binding site for this family of enzymes. Superposition of RmlA with bound d-glucose 1-phosphate onto the structure of RffH with bound dTTP and Mg2+ shows that the phosphate group ofd-glucose 1-phosphate is in proximity to the Mg2+ ion (∼3.2 Å) and to the α-phosphate (<3 Å). The presence of the Mg2+ ion at the position observed in our structure would provide bridging interactions between the two phosphate groups upon binding of G-1-P, with oxygens of both phosphates participating in the coordination of this ion as equatorial ligands. Formation of a coordination bond to the Mg2+ ion by one of the oxygens of the G-1-P would affect the orientation of this phosphate. One of the remaining P-O bonds of G-1-P would point directly inline toward the α-phosphate of dTTP between the three oxygens, well positioned for nucleophilic attack. Consistent with this observation, a metal ion, either Co2+ (17Olsen L.R. Roderick S.L. Biochemistry. 2001; 40: 1913-1921Crossref PubMed Scopus (126) Google Scholar) or Mg2+ (16Kostrewa D. D'Arcy A. Takacs B. Kamber M. J. Mol. Biol. 2001; 305: 279-289Crossref PubMed Scopus (85) Google Scholar), has been observed to bridge the α and β phosphoryl groups of the UDP-GlcNAc product complexed with GlmU. In the structures of complexes of RmlA with either G-1-P or dTTP in the absence of bound Mg2+ the position of Asp-108 (or its equivalent) is maintained, whereas that of Asp-223 is found in two different conformations, forming either a salt bridge with an Arg residue (equivalent to Arg-142 of RffH) or a hydrogen bond to Trp (equivalent to Trp-221 of RffH). Superposition of GlmU and RffH fromE. coli reveals that the position of the first metal-coordinating residue, Asp-108, is conserved, whereas the second ligand is an asparagine originating in a different location than Asp-223, the second ligand in RffH. This Asn coordinates the metal at a different position within the coordination sphere. Importantly, in GlmU these two residues are found in approximately the same position in all structures, independent of the presence or absence of substrate or product. The preformed nature of the Mg2+ binding site in GlmU contrasts with the more flexible site in RffH/RmlA and is reflected by differences in affinity for Mg2+ in these two enzymes. Comparison of the structures of RmlA and GlmU NTP-transferases complexed with the product (NDP-glucose or NDP-glucosamine) shows a rather small spread in the positions of the base and the sugar, with two populations of conformations differing in the position of the phosphoryl group attached to the sugar (Fig.4). Its position is proximal to either the metal ion on one side or a water molecule on the other side. In each of these two conformations there is one close contact between the Pα and Pβ oxygen atoms but they occur on the opposite sides of the phosphate backbone (Figs. 3 and5 a). In the RmlA enzymes these two oxygen atoms are bridged by a water molecule that is also hydrogen bonded to an aspartate. In the GlmU enzymes, the two oxygen atoms on the opposite side are ligands of the metal ion (Fig.4). Comparison of the complexes containing sugar-1-P shows a similar same distribution of the 1-phosphate between two conformations (Fig.5 b). This distribution of the position of the phosphate group attached to the sugar suggests that this group may reorient during catalysis.Figure 5Stereo views of the superposition of reaction products. a, NDP-glucose from a complex with RmlA (1G1L,gray carbon-carbon bonds) and NDP-glucosamine from the complex with GlmU (1G97, green carbon-carbon bonds) and neighboring residues. The superposition is based on the nucleotide portion of the molecules. The molecules are shown inball-and-stick representation; labels initalics refers to RmlA. b, reaction substrate glucose 1-phosphate complexed with E. coli RmlA (1H5R,gray carbon-carbon bonds) and P. aeruginosa(1GOR, green carbon-carbon bonds). Labels initalics refer to RmlA.View Large Image Figure ViewerDownload (PPT) The kinetic data for NTP transferases indicate an ordered bi-bi kinetic mechanism. The NTP binds first in the active site. This event is neither associated with nor requires the presence of Mg2+, but is guided by interactions of the γ-phosphate with the highly conserved GGXGXRL sequence (Gly-8–Leu-14 loop in RffH) and by specific contacts made by the nucleotide base. In this conformation at least three bonds along the phosphate backbone are eclipsed, introducing steric strain into the NTP molecule. Binding of the phosphosugar occurs next. On its own this molecule can bind with two orientations of the phosphate, both stabilized by a network of hydrogen bonds through bridging water molecules (Fig. 5 b). We speculate that in the presence of bound NTP the phosphosugar binds with its phosphate facing away from the α-phosphate of NTP (Fig.5 b). Binding of the second substrate completes the Mg2+ binding site. The approach of the Mg2+ ion induces a rearrangement of Asp-223 (or its equivalent in RmlA enzymes) and a reorientation of the phosphoryl group of the phosphosugar to coordinate the ion. The orientation of the phosphate oxygen necessary to complete the octahedral coordination around Mg2+correctly orients the second phosphate oxygen for an inline attack on the NTP α-phosphate (on the side opposite to β-phosphate). In addition to these orientation effects, the Mg2+ would serve to lower the activation energy required for formation of the transition state, by neutralizing the doubly-negatively charged phosphorane intermediate as proposed for GlmU (16Kostrewa D. D'Arcy A. Takacs B. Kamber M. J. Mol. Biol. 2001; 305: 279-289Crossref PubMed Scopus (85) Google Scholar). Release of pyrophosphate concurrent with the formation of the bond to the phosphosugar occurs with an inversion of configuration on the NTP α-phosphate. The change of charge distribution on the diphosphate of the product likely destabilizes Mg2+ coordination and leads to its release with a concurrent shift of the phosphate bound to the sugar into an alternate conformation with a new hydrogen bond network (Fig. 4). Release of the product might be driven by the approach of another NTP molecule, whereby the γ-phosphate could bind to the now empty site formed by the GGXGXRL sequence. The proposal that Asp-110 of P. aeruginosa RmlA plays a key factor in enhancing catalysis by hydrogen bonding to the α-phosphoryl group of dTTP (11Blankenfeldt W. Asuncion M. Lam J.S. Naismith J.H. EMBO J. 2000; 19: 6652-6663Crossref PubMed Scopus (162) Google Scholar) requires re-evaluation in light of our structural results for RffH. Based on our model, the primary function of Asp-108 and equivalent residues in related enzymes is to coordinate Mg2+, which in turn would enhance catalysis by a combination of charge neutralization of the substrates, decreasing the electronegativity of the α-phosphoryl group and correctly positioning the 1-phosphate of the second substrate poised for nucleophilic attack. We thank Robert Larocque for cloning of the E. coli rffh gene, Dr. Enrico Stura for assistance with data collection, Dr. Stephane Raymond for maintaining the computing environment, and Dr. Joseph D. Schrag for helpful advice.