Portal de Periódicos da CAPES

Crystal Structure of ATP Phosphoribosyltransferase fromMycobacterium tuberculosis

2003; Elsevier BV; Volume: 278; Issue: 10 Linguagem: Inglês

10.1074/jbc.m212124200

1083-351X

Yoonsang Cho, Vivek Sharma, James C. Sacchettini,

HIV/AIDS drug development and treatment

The N-1-(5′-phosphoribosyl)-ATP transferase catalyzes the first step of the histidine biosynthetic pathway and is regulated by a feedback mechanism by the product histidine. The crystal structures of theN-1-(5′-phosphoribosyl)-ATP transferase fromMycobacterium tuberculosis in complex with inhibitor histidine and AMP has been determined to 1.8 Å resolution and without ligands to 2.7 Å resolution. The active enzyme exists primarily as a dimer, and the histidine-inhibited form is a hexamer. The structure represents a new fold for a phosphoribosyltransferase, consisting of three continuous domains. The inhibitor AMP binds in the active site cavity formed between the two catalytic domains. A model for the mechanism of allosteric inhibition has been derived from conformational differences between the AMP:His-bound and apo structures. The N-1-(5′-phosphoribosyl)-ATP transferase catalyzes the first step of the histidine biosynthetic pathway and is regulated by a feedback mechanism by the product histidine. The crystal structures of theN-1-(5′-phosphoribosyl)-ATP transferase fromMycobacterium tuberculosis in complex with inhibitor histidine and AMP has been determined to 1.8 Å resolution and without ligands to 2.7 Å resolution. The active enzyme exists primarily as a dimer, and the histidine-inhibited form is a hexamer. The structure represents a new fold for a phosphoribosyltransferase, consisting of three continuous domains. The inhibitor AMP binds in the active site cavity formed between the two catalytic domains. A model for the mechanism of allosteric inhibition has been derived from conformational differences between the AMP:His-bound and apo structures. N-1-(5′-phosphoribosyl)-ATP transferase 5,5′-dithiobis(2-nitrobenzoic acid) 5′-phosphoribosyl 1′-pyrophosphate 4-morpholineethanesulfonic acid multiple anomalous dispersion thiobenzoate anion root mean square deviation ATP-PRTase fromM. tuberculosis The N-1-(5′-phosphoribosyl)-ATP transferase (ATP-PRTase)1encoded by the hisG locus catalyzes the condensation of ATP with PRPP, the first reaction in the histidine biosynthetic pathway. The reaction is a Mg2+-dependent transfer of the phosphoribosyl moiety from 5′-phosphoribosyl 1′-pyrophosphate (PRPP) to the N1 nitrogen of adenosine ring of ATP yielding phosphoribosyl-ATP and inorganic pyrophosphate (PPi) (Scheme FS1) (1Martin R.G. J. Biol. Chem. 1963; 238: 257-268Google Scholar). The activity and the expression of ATP-PRTase are regulated by feedback inhibition and by repression of the his operon in response to host iron, respectively (2Dall-Larsen T. Int. J. Biochem. 1988; 20: 231-235Google Scholar, 3Rodriguez G.M. Gold B. Gomez M. Dussurget O. Smith I. Tubercle Lung Dis. 1999; 79: 287-298Google Scholar). Given the high energetic costs associated with the synthesis of a histidine molecule and the direct connections of the histidine pathway with purine, pyrimidine, and tryptophan biosynthesis, a multilevel regulatory control has been selectively retained in all bacteria studied to date. Whereas the transcriptional regulation based on nutrient conditions controls the steady-state level of enzyme over several bacterial generations, the feedback inhibition of ATP-PRTase serves as a fine-tuning control that provides rapid regulation of biosynthetic activity as a function of the available histidine. The ATP-PRTase-catalyzed reaction has been studied for more than 4 decades and was originally believed to proceed via the formation of a 5′-phosphoribosyl enzyme covalent intermediate (4Bell R.M. Koshland Jr., D.E. Biochem. Biophys. Res. Commun. 1970; 38: 539-545Google Scholar, 5Koshland Jr., D.E. Biol. Rev. Camb. Philos. Soc. 1953; 28: 416-436Google Scholar). Detailed kinetic studies refuted the presence of such an intermediate (6Brashear W.T. Parsons S.M. J. Biol. Chem. 1975; 250: 6885-6890Google Scholar). Steady-state studies of the enzymatic reaction in both directions were consistent with a sequential mechanism (7Cleland W.W. Boyer P.D. The Enzymes. Academic Press, New York1970: 1-65Google Scholar) where ATP binding precedes binding of PRPP (8Musick W.D. CRC Crit. Rev. Biochem. 1981; 11: 1-34Google Scholar). The ATP-PRTase reaction has also been shown to be completely reversible as addition of pyrophosphate to phosphoribosyl-ATP yields ATP and PRPP (9Ames B.N. Martin R.G. Garry B.J. J. Biol. Chem. 1961; 236: 2019-2026Google Scholar). The synergistic inhibition of the enzyme was demonstrated to occur allosterically by histidine and competitively by AMP, ADP, or guanosine tetraphosphate (10Morton D.P. Parsons S.M. Arch. Biochem. Biophys. 1977; 181: 643-648Google Scholar). AMP and ADP are both competitive inhibitors with respect to PRPP and ATP (1Martin R.G. J. Biol. Chem. 1963; 238: 257-268Google Scholar). Histidine inhibition was first thought to be "noncompetitive" with PRPP and ATP (1Martin R.G. J. Biol. Chem. 1963; 238: 257-268Google Scholar). A single histidine was later proposed to interact with more than one molecule of the enzyme, in a site shown to be allosteric in nature (11Bell R.M. Parsons S.M. Dubravac S.A. Redfield A.G. Koshland Jr., D.E. J. Biol. Chem. 1974; 249: 4110-4118Google Scholar). A clear understanding of the molecular basis of ATP-PRTase activity and the mechanism of its regulation by histidine has been elusive due to the lack of structural information. Although structures of several PRTases are known (12Sinha S.C. Smith J.L. Curr. Opin. Struct. Biol. 2001; 11: 733-739Google Scholar), lack of sequence similarity precluded analyses based on homology modeling. Based on their structural folds, the PRTases have been subdivided into two groups (13Scapin G. Ozturk D.H. Grubmeyer C. Sacchettini J.C. Biochemistry. 1995; 34: 10744-10754Google Scholar, 14Eads J.C. Ozturk D. Wexler T.B. Grubmeyer C. Sacchettini J.C. Structure (Camb.). 1997; 5: 47-58Google Scholar). The type I PRTases have a central parallel five-stranded β-sheet surrounded by α-helices. Type II PRTases, such as quinolinic acid PRTase, have a modified α/β-barrel as the catalytic core. Association of alternate structural motifs with PRTases has suggested a convergent evolution of these enzymes. In this study we report the structure of ATP-PRTase fromMycobacterium tuberculosis (mtATP-PRTase) without bound ligands (apo) and in a ternary complex with the inhibitors AMP and histidine (AMP:His). These structures represent a new fold for a PRTase with a modular organization of the regulatory histidine binding domain and catalytic PRTase domains. Comparison of the inhibitor-bound structure with the apo form reveals the structural basis of the allosteric regulation by histidine. ATP, AMP, l-histidine, and 5,5′-dithiobis(2-nitrobenzoic acid) (DTNB) were purchased from Sigma. PRPP and lithium sulfate were purchased from Fluka. Standard proteins for calibrating gel filtration column were purchased from AmershamBiosciences. ThehisG gene, Rv2121c from M. tuberculosis H37Rv genome, was identified from the TubercuList sequence data base (15Cole S.T. Brosch R. Parkhill J. Garnier T. Churcher C. Harris D. Gordon S.V. Eiglmeier K. Gas S. Barry III, C.E. Tekaia F. Badcock K. Basham D. Brown D. Chillingworth T. Connor R. Davies R. Devlin K. Feltwell T. Gentles S. Hamlin N. Holroyd S. Hornsby T. Jagels K. Barrell B.G. Nature. 1998; 393: 537-544Google Scholar). The hisG gene was amplified using M. tuberculosisgenomic DNA as a template. The PCR product was cloned into a pET28a expression vector (Novagen) with N-terminal His tag and transformed into Escherichia coli overexpression strain, BL21(DE3). Cells were incubated at 37 °C until the optical density reached 0.6 and induced with 1 mmisopropyl-1-thio-β-d-galactopyranoside, and incubation was continued for an additional 4 h. The bacterial cells were harvested by centrifugation and resuspended in 20 mm sodium phosphate (pH 7.5) containing 0.5 m NaCl and 0.1m imidazole. The cells were lysed using a French press. The cell extract was applied onto a 5-ml nickel-nitrilotriacetic acid column (Amersham Biosciences), and the target protein was eluted using an imidazole gradient. The eluate was concentrated by Centriprep (Amicon) to 20 mg/ml and applied onto a Sephadex 200 gel-filtration column (Amersham Biosciences) equilibrated with 20 mm HEPES (1 mm EDTA and dithiothreitol (pH 7.5)) as a final step. The protein was more than 95% pure as observed on an SDS-PAGE gel. Selenomethionylated protein was prepared according to published methods (16Davies C. Heath R.J. White S.W. Rock C.O. Struct. Fold. Des. 2000; 8: 185-195Google Scholar). The pET28a-hisG plasmid was transformed into E. coli B834(DE3) (Novagen) Met auxotroph strain. Cells were grown in LB medium until an optical density of 0.6 was obtained. Cells were pelleted by centrifugation, washed with LB medium, and resuspended in M9 minimal medium lacking l-Met. SeMet was then added to a final concentration of 0.05 μg/ml along with 35 μg/ml kanamycin. Cultures were then induced with 1 mmisopropyl-1-thio-β-d-galactopyranoside followed by incubation for 4 h at 37 °C. The protein was purified using the same methods as for the apoprotein. Initial crystallization conditions were obtained using Crystal Screen 2 from Hampton Research. Crystals were grown using the hanging drop vapor diffusion method at 16 °C. The apocrystals were obtained by mixing equal volumes (2–3 μl) of 20 mg/ml protein with a buffer containing 0.1 m MES (pH 6.5) and 1.8 m magnesium sulfate as a precipitant. AMP:His-crystals were obtained in condition number 15 of Crystal Screen 2 from Hampton Research (0.1 m sodium citrate (pH 5.6), 0.5 m ammonium sulfate, and 1.0 m lithium sulfate) in the presence of 5 mm AMP and 100 μm histidine. A complete and redundant high resolution data set was collected at BioCARS beamline 14BMC at the Advanced Photon Source, Argonne National Laboratory. Multiple anomalous dispersion (MAD) data sets were collected for both the apocrystal (MAD1) and AMP:His crystal (MAD2) (Table I). All data sets were indexed and scaled using MOSFLM and SCALA of the CCP4 program suite (17Collaborative Computational Project 4 Acta Crystallogr. Sect. D Biol. Crystallogr. 1994; 50: 760-763Google Scholar). Unit cell dimensions for apocrystal were a = b = 132.5 Å, c = 110.5 Å, α = β = 90, and γ = 120. Space group was R32. The inhibitor complex crystallized also in the space group R32, but the cell dimension changed by about 14 and 11% in a, b (113.8 Å), and c (124.3 Å), respectively. Calculation of solvent content (18Matthews B.W. J. Mol. Biol. 1968; 33: 491-497Google Scholar) indicated that for both crystals the asymmetric unit contained one protomer of ATP-PRTase and 58 (apo) or 48% (AMP:His) solvent.Table IData collection and refinement statisticsData collectionaValues in parentheses are for the highest resolution shell: 3.0–3.17 Å for the MAD1 data sets; 2.6–2.75 Å for the MAD2 data sets; 2.7–2.97 Å for the apo data sets, and 1.8–1.93 Å for the AMP:His data sets.WavelengthResolution% completeR symbRsym=100×ΣhΣi|I(h)l−〈I(h)〉|ΣhΣi〈I(h)i〉where I is the observed intensity, and 〈I〉 is the average intensity of multiple observations of symmetry-related reflections.ÅÅMAD10.9642 (remote)3.099.9 (99.8)0.195 (0.569)0.9795 (peak)3.099.9 (99.8)0.109 (0.306)0.9798 (inflection)3.093.7 (99.6)0.171 (0.701)MAD20.9571 (remote)2.699.1 (99.7)0.088 (0.447)0.9798 (peak)2.699.1 (99.6)0.081 (0.185)0.9800 (inflection)2.699.1 (99.7)0.156 (0.207)Apo1.02.799.1 (99.9)0.067 (0.150)AMP:His1.01.897.5 (99.5)0.047 (0.134)Refinement statisticsApoAMP:HisResolution range (Å)28–2.720–1.8No. reflectionsWorking set985025,850Test set5471366No. atomsProtein20732097Solvent199251R crystcR=∑‖Fo|−|Fc‖/Σ|Fo|.Rcryst and Rfree were calculated using the working and the test reflection sets, respectively. 5% of the entire reflection was taken as test set.19.219.8R freecR=∑‖Fo|−|Fc‖/Σ|Fo|.Rcryst and Rfree were calculated using the working and the test reflection sets, respectively. 5% of the entire reflection was taken as test set.26.123.6Average B (Å2)18.317.9r.m.s.d. from idealBond length0.048 Å0.014 ÅBond angle3.6 °1.5 °a Values in parentheses are for the highest resolution shell: 3.0–3.17 Å for the MAD1 data sets; 2.6–2.75 Å for the MAD2 data sets; 2.7–2.97 Å for the apo data sets, and 1.8–1.93 Å for the AMP:His data sets.b Rsym=100×ΣhΣi|I(h)l−〈I(h)〉|ΣhΣi〈I(h)i〉where I is the observed intensity, and 〈I〉 is the average intensity of multiple observations of symmetry-related reflections.c R=∑‖Fo|−|Fc‖/Σ|Fo|.Rcryst and Rfree were calculated using the working and the test reflection sets, respectively. 5% of the entire reflection was taken as test set. Open table in a new tab Selenium sites were located using SOLVE (19Terwilliger T.C. Berendzen J. Acta Crystallogr. Sect. D Biol. Crystallogr. 1999; 55: 849-861Google Scholar) with three different wavelength MAD data. The sites were refined using MLPHARE (20Otwinowski Z. Wolf W. Evans P.R. Leslie A.G.W. Isomorphous Replacement and Anomalous Scattering: Daresbury Study Weekend Proceedings. Daresbury Laboratory, Daresbury, UK1991: 80-86Google Scholar), and protein phases were calculated with SHARP (21de La Fortelle E. Bricogne G. Methods Enzymol. 1997; 276: 472-494Google Scholar) (30–3.0 Å) and improved by density modification using CNS (crystallography and NMR system) (22Brünger 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. Sect. D Biol. Crystallogr. 1998; 54: 905-921Google Scholar). A polyalanine backbone model was built into the electron density using O (23Jones T.A. Zhow J.Y. Cowan S.W. Kjeldgaard M. Acta Crystallogr. Sect. A. 1991; 47: 110-119Google Scholar). Based on marker amino acids such as SeMet, Arg, and aromatic residues, polyalanines were converted to the original sequence. Initial refinement was performed by rigid body refinement, simulated annealing and individual B factor refinement. InitialR factor and R free were 35 and 42%, respectively. After an intensive series of manual rebuilding and refinement, the R factor andR free dropped down to 28 and 33%, respectively. Solvent molecules were picked using Xfit (24McRee D.E. J. Struct. Biol. 1999; 125: 156-165Google Scholar) and refined. As a final refinement step, the Restrained TLS refinement with Refmac5 (25Winn M.D. Isupov M.N. Murshudov G.N. Acta Crystallogr. Sect. D Biol. Crystallogr. 2000; 57: 122-133Google Scholar) was used, and the R factors were 19.2 and 26.1% (TableI, bottom). Molecular replacement of the AMP:His-bound data was attempted with the apo structure as a search model. However, any reasonable solution was not obtained from the whole molecule or separate domains. Therefore, another MAD experiment was performed. Four Se sites were determined by SOLVE, and phases were calculated with SHARP up to 2.6 Å. The MAD map was made after solvent flattening with DM (density modification) (26Cowtan K.D. Joint CCP4 and ESF-EACBM Newsletter on Protein Crystallography. 1994; 31: 34-38Google Scholar) of the CCP4 program suite. The apo structure was manually fitted into the electron density to make an initial model for the inhibitor-bound structure. Positional refinement and molecular dynamics were performed, and theR free was 30%. Shake & Warp (27Kantardjieff K.A. Höchtl P. Segelke B.W. Tao F.-M. Rupp B. Acta Crystallogr. Sect. D Biol. Crystallogr. 2002; 58: 735-743Google Scholar) was used to remove phase bias from the model. Solvent molecules were picked and the restrained TLS refinement with Refmac5 was performed. The refinement statistics are shown in Table I, bottom. We followed an experimental procedure described previously (28Riddles P.W. Blakeley R.L. Zerner B. Methods Enzymol. 1983; 91: 49-60Google Scholar) for characterizing the number of free cysteines per molecule of protein. Briefly, 0.1 ml of a protein solution was added to 3.1 ml of reaction buffer containing 0.3 mm DTNB to achieve a final concentration of 0.3 mg/ml (9.4 μm) of freshly prepared reaction mixture. The absorbance of 2-nitro-5-thiobenzoate anion (TNB2−) was measured at 412 nm until it reached a plateau. The numbers of free cysteines were calculated from the absorbance (0.18 and 0.35 absorbance units for the apo and AMP:His form, respectively) and molar absorption coefficient of TNB2− (14,150 m−1cm−1) covalently linked to free cysteines. The numbers of the free cysteines corresponding to the obtained absorbance were 2 and 1 (equivalent to 9.3 and 21.3 μm TNB2−). A Superdex 200 gel filtration column (24-ml bed volume, Amersham Biosciences) was used to estimate the molecular weight of ATP-PRTase and to observe the effect of different ligands on oligomerization. The column was calibrated using low and high molecular standard proteins (from Amersham Biosciences) in 20 mm HEPES (pH 7.5), 0.1 m NaCl, 1 mm EDTA, and dithiothreitol. 100 μm histidine and 1 mm AMP were added in the same buffer to observe the change of oligomeric status in the presence of the inhibitors. In the absence of histidine at 4 °C, more than 99% of the apoenzyme eluted as a dimer at a low protein concentration (less than 50 μg/ml). We were not able to detect the dimer when the protein was preincubated at 10 μm histidine; only hexamers and higher oligomers were detected. The x-ray structure of the recombinant mtATP-PRTase was solved from electron density maps calculated by MAD methods using crystals of selenomethionylated protein formed in the space group R32. Crystals were produced in the absence of any ligands or after incubation of protein with two inhibitors, adenosine monophosphate and histidine (AMP:His). The structures have been refined to R factors of 19.2 (apo) and 19.8% (AMP:His) at resolutions of 2.7 and 1.8 Å, respectively (Table I). In both cases, the refined structure contains 276 of the 284 residues present in mtATP-PRTase. The residues 186–193 were disordered and omitted from the final model. mtATP-PRTase is an elongated molecule consisting of 10 α-helices and 15 β-strands (Fig.1 a) composed in 3 domains. Domain I (residues 1–90, 175–184, and 194–211) contains a central β-sheet consisting of four parallel β-strands (β1, β3, β4, and β5) and two anti-parallel strands (β2 and β11). The β-sheet is surrounded by 3 α-helices, α1 on one side and α2 and α3 on the other side. Domain II (residues 91–174) is also an α/β-structure composed of four (β7–10) parallel β-strands and one (β6) anti-parallel β-strand with two α-helices on each side (α4 and α5 on one side and α6 and α7 on the other side). Domain III (residues 212–284) has one β-sheet consisting of four anti-parallel β-strands (β12–15) with two α-helices (α9 and α10) on one side of the β-sheet. Domains I and II form the catalytic core of ATP-PRTase. The competitive inhibitor AMP binds in a cleft located between the two domains (Fig.1 b) and makes the most of its bonding interactions with residues from domain II. The feedback inhibitor histidine was located far from the active site in domain III (Fig. 1). The electron density of both inhibitors is shown in Fig. 1, c and d.The catalytic core of ATP-PRTase (domains I and II) is similar to theE. coli glutamine-binding protein (Protein Data Bank code 1WDN; r.m.s.d. 3.4 for 172 Cα atoms) (29Sun Y.J. Rose J. Wang B.C. Hsiao C.D. J. Mol. Biol. 1998; 278: 219-229Google Scholar) (Fig. 2 a), an E. coli histidine-binding protein (Protein Data Bank code1HSL; r.m.s.d. 3.2 for 164 Cα atoms) (30Yao N. Trakhanov S. Quiocho F.A. Biochemistry. 1994; 33: 4769-4779Google Scholar) as well as the ligand binding core of a glutamate receptor fromSynechocystis sp. (Protein Data Bank code1IIW) (31Mayer M.L. Olson R. Gouaux E. J. Mol. Biol. 2001; 311: 815-836Google Scholar) and that of rat (Protein Data Bank code 1LB8) (32Sun Y. Olson R. Horning M. Armstrong N. Mayer M. Gouaux E. Nature. 2002; 417: 245-253Google Scholar, 33.Deleted in proof.Google Scholar). A VAST 2On-line address, www.ncbi.nlm.nih.gov/structure/VAST/vastsearch.html. structural similarity search using domain III found that the domain shares a high degree of similarity with the E. coli signal transducing protein PII (Protein Data Bank code 2PII; r.m.s.d. 1.4 for 63 Cα atoms) and the guanine nucleotide exchange factor domain from human elongation factor-1β (Protein Data Bank code 1B64; r.m.s.d. 2.0 for 57 Cα atoms). Whereas all the four β-strands and two α-helices are conserved between structures of PII and domain III of the ATP-PRTase, the two differ in the length of their connecting loops (i.e.7 residues longer in the case of PII) (Fig. 2 a). Interestingly, proteins of PII family (34Cheah E. Carr P.D. Suffolk P.M. Vasudevan S.G. Dixon N.E. Ollis D.L. Structure. 1994; 2: 981-990Google Scholar) and GlnK (35Xu Y. Cheah E. Carr P.D. van Heeswijk W.C. Westerhoff H.V. Vasudevan S.G. Ollis D.L. J. Mol. Biol. 1998; 282: 149-165Google Scholar) form a trimer similar to that observed for domain III of ATP-PRTase (35Xu Y. Cheah E. Carr P.D. van Heeswijk W.C. Westerhoff H.V. Vasudevan S.G. Ollis D.L. J. Mol. Biol. 1998; 282: 149-165Google Scholar, 36van Heeswijk W.C. Wen D. Clancy P. Jaggi R. Ollis D.L. Westerhoff H.V. Vasudevan S.G. Proc. Natl. Acad. Sci. U. S. A. 2000; 97: 3942-3947Google Scholar). Gel filtration, sedimentation velocity ultracentrifugation, and light scattering experiments on the E. coli enzyme have demonstrated that the ATP-PRTase exists in equilibrium between its active dimeric form (Fig. 2 c) and inactive higher oligomeric forms (37Klungsoyr L. Kryvi H. Biochim. Biophys. Acta. 1971; 227: 327-336Google Scholar, 38Tebar A.R. Fernandez V.M. Martin Del Rio R. Ballesteros A.O. Experientia (Basel). 1973; 29: 1477-1479Google Scholar, 39Lohkamp B. Coggins J.R. Lapthorn A.J. Acta Crystallogr. Sect. D Biol. Crystallogr. 2000; 56: 1488-1491Google Scholar). Gel filtration experiments showed similar behavior for the mt ATP-PRTase (see "Experimental Procedures"). In general, ATPase hexamers are more abundant at concentrations of enzyme higher than 1 mg/ml or in the presence of stoichiometric AMP, phosphoribosyl-ATP, and histidine and particularly in the combination of one of the nucleotides and histidine (37Klungsoyr L. Kryvi H. Biochim. Biophys. Acta. 1971; 227: 327-336Google Scholar). On the other hand, low enzyme concentrations (50 μg/ml) or the presence of the substrate PRPP seems to dissociate the hexamers, or higher oligomers, into active dimers (38Tebar A.R. Fernandez V.M. Martin Del Rio R. Ballesteros A.O. Experientia (Basel). 1973; 29: 1477-1479Google Scholar). Thus regulation of the oligomeric state of ATP-PRTase appears to be an efficient way of controlling the enzyme activity by sensing the intracellular concentrations of both enzyme and histidine. At low in vivointracellular histidine levels and enzyme concentrations, ATP-PRTase most likely exists as active dimers and constitutively replenishes the histidine pool. Under conditions of high histidine demand, such as active assimilation of nitrogen, transcriptional derepression of thehisG gene perhaps allows even higher intracellular concentration of ATP-PRTase that may be hexameric. However, once the histidine level exceeds the demand, the expression of hisGgene is reduced, and the existing ATP-PRTase is inhibited by histidine. Some proteobacteria have a shorter version of the ATP-PRTase, missing about 100 residues from the C terminus (domain III). In these bacteria, HisG can associate with another protein HisZ, a parahomolog of aminoacyl-tRNA synthetase that is functionally unknown (40Sissler M. Delorme C. Bond J. Ehrlich S.D. Renault P. Francklyn C. Proc. Natl. Acad. Sci. U. S. A. 1999; 96: 8985-8990Google Scholar). Recent equilibrium sedimentation studies on HisG and HisZ fromLactococcus lactis show that they individually form stable homodimers. However, together the two proteins form an octameric structure that can be destabilized by allosteric regulators AMP and histidine (41Bovee M.L. Champagne K.S. Demeler B. Francklyn C.S. Biochemistry. 2002; 41: 11838-11846Google Scholar). No homolog of HisZ is found in M. tuberculosis genome. However, given that HisZ is required for the activity and regulation of the truncated HisG, it is tempting to speculate that it may be compensating for some of the functions of the missing domain III. Whereas alternate roles and mechanisms for regulation of HisZ may not be ruled out, quaternary associations of HisG, both homo- or heteromeric, seem to have direct influence on the function and regulation of these enzymes. In both the apo and AMP:His structures of ATP-PRTase the packing in the crystal is consistent with a hexamer because of crystallographic 3- and 2-fold symmetry axes in the R32 space group that generates a "trimer of dimers" (Fig. 3 a). However, comparison of the intersubunit interactions in the two structures showed that the hexamers are different with the histidine-containing complex being much more compact than the apoprotein (Fig. 3, d and e). In the case of the AMP:His hexamer, the subunit-accessible surface area buried is 3078 Å2, and it is only 2417 Å2 in the apo form. The dimer interface buries 1203 and 965 Å2 of accessible surface of each subunit in apo or AMP:His forms, respectively. The interactions at the dimer interface are primarily from the catalytic core (domains I and II), whereas those involved at the hexamer interface are mainly from domain III. The most prominent structural feature of the AMP:His hexamer is the extended β-sheet for domain III formed by the C-terminal β-strand (residues 280–284) with the penultimate β-strand (β15, residues 273–276) of the adjacent subunit (Fig. 3 b). The catalytic site of ATP-PRTase is formed by a cleft located between domains I and II (Fig. 1,a and b). The substrate-binding sites could be identified from highly negative electrostatic potential of the protein, presumably involved in binding to the Mg2+ ions required for catalysis and by the presence of sulfate ions from the crystallization buffer, marking the probable binding sites of phosphate groups of the substrates. The inhibitor AMP (competitive with respect to ATP) was located in clear electron density from omit maps calculated from diffraction data collected from crystals of HisG that were incubated with AMP and His prior to crystallization. AMP bound to the expected ATP-phosphoribosyltransferase signature sequence region (Glu141–Leu162), which was identified from PROSITE (42Appel R.D. Bairoch A. Hochstrasser D.F. Trends Biochem. Sci. 1994; 19: 258-260Google Scholar) (documentation number PDOC01020). Residues from both domains I and II contributed to the binding of AMP (Fig.4 a). The phosphate of the AMP is coordinated by the P-loop motif (residues from Asp154 to Thr161) found in the domain II. One of the phosphate oxygens of the AMP, O1P, forms hydrogen bonds with backbone amides of Gly159 and Gly157. O2P makes a hydrogen bond with OG1 of Thr161, and O3P hydrogen bonds with backbone amides of Thr161 and Arg160 as well as with OG of Ser158. O5 of the AMP interacts with the backbone nitrogen of Ser158. N1 of the adenosine base forms hydrogen bonds with three ordered water molecules. One of them is a water-mediated interaction between the N1 of AMP and OD2 of Asp70. N1 also hydrogen bonds via an ordered water molecule, with OD1 of Asp70 and OG of Ser90. N6 of the adenosine ring forms a hydrogen bond with OH of Tyr116. Of these, only the interactions of residues 70 and 90 are from domain I and the others are from domain II. The 2-OH and 3-OH of the AMP are close to domain I of the neighboring subunit and interact with the side chain carboxyls of Asp30′ and Asp33′ of that subunit. As these interactions would contribute to the compactness of the hexamer, they may be responsible for the synergistic behavior of AMP toward inhibition with histidine. Analysis of the structure suggests that ATP would bind in a manner very similar to AMP binding. The presence of a tightly associated sulfate ion close to the AMP-binding site along with several basic residues (Arg49, Lys9, Lys32, and Arg160) indicate that PRPP may bind in a region adjacent to the AMP-binding site (Fig. 4 b). Moreover, consideration of the catalytic reaction would require that PRPP be oriented such that the C1 carbon of the ribosyl group of PRPP is in close proximity to N1 of the adenine ring of the ATP. The 5′-phosphate of PRPP is more likely to occupy the location occupied by the sulfate ion bound to residues Lys9 and Arg49. In this model, the leaving pyrophosphate group would interact with residue Lys32. The location of probable PRPP-binding site at the dimer interface suggests that the PRPP bound to one subunit of the dimer would be located close to the ATP bound to the other subunit of the dimer. In this model PRPP would bury the bound ATP that is consistent with the sequential mechanism observed in other PRTases where binding of base precedes binding of PRPP. The major conformational change observed in the histidine-bound form is a large twist of the domain III relative to the domain I and II (Fig.2 b). When domain I of the apoenzyme and that of the AMP:His enzyme were superimposed, the r.m.s.d. of the domain I was only 1.46 Å and that of domain II was 2.19 Å. The r.m.s.d. of the domain III, however, was 12.89 Å, due to a solid body movement of the β-sheet of the domain III induced by residues involved in binding to histidine (Fig. 2 b). Six histidines bind to the domain III clusters at both ends of three dimers, stabilizing the hexamer (Fig.3 a). These histidines are completely embedded in the domain III cluster (Fig. 3 c). Molecular surface representations of the hexamers show this conformational change from an "open" cluster (Fig. 3 d) to a "closed" cluster (Fig. 3 e). The residues involved in binding to each histidine are contributed by the two adjacent domain IIIs suggesting that direct interactions with histidine (Fig. 5 a) are responsible for bringing the three dimers together to form the hexamer. The interactions include a well ordered hydrogen bonding network with residues Asp218 and Ala273 from one subunit, residues Leu234, Gly235, Ser236, Thr238, and Leu253 of the second subunit, and an ordered water molecule. Inhibition resulting from hexamer formation is somewhat reminiscent of the allosteric mechanism observed in ribonucleotide reductase (43Kashlan O.B. Scott C.P. Lear J.D. Cooperman B.S. Biochemistry. 2002; 41: 462-474Google Scholar). Feedback inhibitor-based oligomerization, resulting in either altered topology or reduced access to the active site, is emerging as a way of regulating enzymes. In the case of ATP-PRTase, the allosteric inhibition by histidine can be synergistically favored by the competitive inhibitor AMP, thus adding yet another dimension to the regulation of activity. Maximal inhibition is observed when both inhibitors AMP:His are bound (11Bell R.M. Parsons S.M. Dubravac S.A. Redfield A.G. Koshland Jr., D.E. J. Biol. Chem. 1974; 249: 4110-4118Google Scholar). The structures suggest that the reason for the synergistic behavior is that binding of histidine reorients some key active site residues (Tyr116, Arg135, Arg137, Asp154, and Arg160) in the active site, and in return binding of AMP establishes additional inter-subunit interactions that stabilize the histidine-bound hexamer. These interactions are only possible with the global conformational change triggered by histidine. The presence of disulfide bonds in prokaryote intracellular enzymes has not been well documented, although crystallographic studies have shown the existence of disulfide bonds in a handful of prokaryotic enzymes (44Bourne Y. Redford S.M. Steinman H.M. Lepock J.R. Tainer J.A. Getzoff E.D. Proc. Natl. Acad. Sci. U. S. A. 1996; 93: 12774-12779Google Scholar,45Wells W.W. Yang Y. Deits T.L. Gan Z.R. Adv. Enzymol. Relat. Areas Mol. Biol. 1993; 66: 149-201Google Scholar). In the ATP-PRTase structure we not only observe a disulfide bond between Cys73 and Cys175 but also found that it was not present in the PRTase with AMP and histidine (Fig.5 b). In this structure the distance between the two Cαs of the cysteines was 8.6 Å, too far for disulfide bond formation. Two possible scenarios can explain this observation. First, the lack of a disulfide bond could be from the strain imposed by the conformational changes observed in the AMP:His structure possibly due to a closure between domains I and II (see Fig. 5 b). It could also be due to exposure of crystals to synchrotron radiation. Structurally and functionally significant disulfide bonds have been shown as broken in crystals exposed to synchrotron radiation (46Weik M. Ravelli R.B. Kryger G. McSweeney S. Raves M.L. Harel M. Gros P. Silman I. Kroon J. Sussman J.L. Proc. Natl. Acad. Sci. U. S. A. 2000; 97: 623-628Google Scholar). DTNB was used to determine the presence or absence of the disulfide bond between Cys73 and Cys175. In the absence of any ligands, the absorbance of TNB2− at 412 nm suggested that the disulfide was present in only two of the four cysteines. However, when the enzyme was incubated with 100 μm AMP or histidine, or 2 mm AMP with 100 μm of histidine, the molar ratio of cysteines modified per protomer was reduced to about 0.5 (see "Experimental Procedures"). We believe the reduction was due to the formation of hexamer that reduces exposure of all cysteines. When the same experiments were performed in the presence of 6 m guanidinium hydrochloride, all protein forms again showed only two free cysteines. These results suggest that the disulfide is present in both the apoprotein and AMP:His protein and that the observed broken disulfide was the result of the radiation, although we cannot rule out the possibility that in the inhibitor-bound protein the disulfide rapidly reforms upon denaturation. The structure described here provides an explanation of the molecular basis of feedback inhibition of the histidine biosynthetic pathway by allosteric regulation of ATP-PRTase by histidine. The binding of histidine seems to influence activity both by stabilizing the inactive hexameric form and by sterically hindering the binding of substrates to the catalytic site. Although the presence of an allosteric domain that binds the end product of the pathway has been observed in several enzymes, to our knowledge this represents the first example for the PRTases. ATP-PRTase also appears to be another example of the convergent evolution of the PRTases. We thank the scientists of BioCARS beamlines at Advanced Photon Source, Argonne National Laboratory, for help in data collection. The use of the Advanced Photon Source was supported by the United States Department of Energy, Basic Energy Sciences, Office of Science, under Contract W-31-109-Eng-38. The use of the BioCARS Sector 14 was supported by the National Institutes of Health, National Center for Research Resources, Grant RR07707. We thank Dr. Bernhard Rupp (Lawrence Livermore National Laboratory) for help in performing the Shake&Warp program and for comments on the manuscript.