Imidazole Glycerol Phosphate Synthase fromThermotoga maritima

Imidazole glycerol phosphate synthase, which links histidine and de novo purine biosynthesis, is a member of the glutamine amidotransferase family. In bacteria, imidazole glycerol phosphate synthase constitutes a bienzyme complex of the glutaminase subunit HisH and the synthase subunit HisF. Nascent ammonia produced by HisH reacts at the active site of HisF withN′-((5′-phosphoribulosyl)formimino)-5-aminoimidazole-4-carboxamide-ribonucleotide to yield the products imidazole glycerol phosphate and 5-aminoimidazole-4-carboxamide ribotide. In order to elucidate the interactions between HisH and HisF and the catalytic mechanism of the HisF reaction, the enzymes tHisH and tHisF from Thermotoga maritima were produced in Escherichia coli, purified, and characterized. Isolated tHisH showed no detectable glutaminase activity but was stimulated by complex formation with tHisF to which either the product imidazole glycerol phosphate or a substrate analogue were bound. Eight conserved amino acids at the putative active site of tHisF were exchanged by site-directed mutagenesis, and the purified variants were investigated by steady-state kinetics. Aspartate 11 appeared to be essential for the synthase activity both in vitro and in vivo, and aspartate 130 could be partially replaced only by glutamate. The carboxylate groups of these residues could provide general acid/base catalysis in the proposed catalytic mechanism of the synthase reaction.

During the synthesis of various biomolecules including amino acids, nucleotides, and coenzymes, the amido group of glutamine is transferred to a large variety of acceptor substrates by glutamine amidotransferases (GATases) 1 (1,2). GATases catalyze two separate reactions at two active sites that are either located on a single polypeptide chain or on different subunits. In the glutaminase reaction, glutamine is hydrolyzed to glutamate and ammonia, which in the synthase reaction is added to an acceptor substrate that is specific for each GATase.
There are two classes of GATases that can be discriminated by catalytically essential residues in their glutaminase domains (1,3). The key feature of class I GATases is the catalytic triad Cys-His-Glu. Recent x-ray structure determinations of three class I GATases, Escherichia coli carbamoyl phosphate synthase (4,5), E. coli GMP synthase (6), and S. solfataricus anthranilate synthase (7) indicate a common fold of their glutaminase domains, which is similar to the well known ␣/␤ hydrolase fold (8). Class II GATases belong to the large family of Ntn hydrolases (9), and their only catalytically essential amino acid is the conserved N-terminal cysteine. The corresponding synthase domains within each class are structurally, evolutionary, and functionally unrelated (2), supporting the hypothesis that glutamine-hydrolyzing enzymes were recruited independently by previously ammonium-dependent enzymes (3). Along these lines, most GATases can use ammonium salts as an alternative source of ammonia (1,2).
The imidazole glycerol phosphate (ImGP) synthase is a class I GATase, which in bacteria constitutes a bienzyme complex consisting of the glutaminase subunit HisH and the synthase subunit HisF (10,11). The ammonia produced by HisH reacts with the substrate of HisF, which is NЈ-((5Ј-phosphoribulosyl) formimino)-5-aminoimidazole-4-carboxamide-ribonucleotide (PRFAR). The products of this reaction, ImGP and 5aminoimidazole-4-carboxamide ribotide (AICAR), are further used in histidine and de novo purine biosynthesis, respectively. In yeast, the glutaminase and synthase activities are located on a single polypeptide chain, which is termed HIS7 (12,13). The mechanism of glutamine hydrolysis by HisH can be deduced from the class I GATase carbamoyl phosphate synthase (5). However, due to the lack of a high resolution structure for the bienzyme complex, the ammonia transfer from HisH to HisF and the mechanism of the HisF reaction are only poorly understood.
In order to address these questions, the thermostable variants tHisF and tHisH from Thermotoga maritima were produced in E. coli, purified and characterized by hydrodynamic and spectroscopic measurements, limited proteolysis, and steady-state enzyme kinetics. Moreover, the high resolution x-ray structure of isolated tHisF (14) was used to identify and probe amino acid residues that are potentially involved in catalysis of the synthase reaction. It was shown that tHisH is activated by complex formation with tHisF containing ImGP or a substrate analogue bound to its active site. A flexible loop region in tHisF appears to be important for these functional interactions with tHisH. Furthermore, two aspartate residues at the active site of tHisF were demonstrated to be essential for catalysis. One of them was partially replaceable by glutamate, as shown by saturation random mutagenesis and complementation in vivo of an E. coli hisF deletion strain. Based on these findings, a chemically plausible mechanism for the HisF synthase reaction was derived, which involves general acid/base catalysis.

EXPERIMENTAL PROCEDURES
DNA Manipulation and Sequence Analysis-Preparation of DNA, digestion with restriction endonucleases, and DNA ligation were performed as described (15). Oligonucleotides were purchased from Metabion. DNA was amplified by PCR using cloned Pfu polymerase (Stratagene). For PCR with mutagenic oligonucleotides, Taq polymerase (Roche Diagnostics) was used. DNA was extracted from agarose gels using the QIAquick gel extraction kit (Qiagen). DNA sequencing was performed by the "Göttingen Genomics Laboratory" and by the "Zentrum fü r Molekulare Medizin der Universität Köln," using standard methods. N-terminal protein sequencing was performed by Dr. Paul Jenö (Biozentrum, University of Basel) and the "Zentrum fü r Molekulare Medizin der Universität Köln," again using standard methods.
Subcloning of the hisF and hisH Genes from T. maritima-The hisF gene variants from Thermotoga maritima (thisF; Ref. 16) encoding the synthase subunit of imidazole glycerol phosphate synthase were cloned into the expression vector pET11c (Novagen) using the restriction enzymes NdeI and BamHI. The hisH gene of T. maritima (thisH; Ref. 16) was amplified by PCR, using purified chromosomal DNA (Qbiogen) as a template. The oligonucleotides 5Ј-GGT GTG ATA GCA TGC GTA TCG-3Ј with a SphI-site (in boldface type) and 5Ј-CTA CCA AGC TTC TGA AGA GAT CTA TCG-3Ј with a HindIII-site (in boldface type) were used as 5Ј-and 3Ј-primers, respectively. Using the two newly introduced restriction sites, the amplified DNA fragment was cloned into the vector pDS56/RBSII/SphI (17) to yield the plasmid pDS56/RBSIISphI-thisH. All inserts were entirely sequenced to exclude inadvertent PCR mutations.
Site-directed Mutagenesis of thisF-Point mutations were introduced into thisF by PCR-based methods using either mutagenic 5Ј-primer for construction of thisF_C9A, D11N, K19S, or the megaprimer method (18) for construction of thisF_D51N, N103A, D130N, D176N, and D183N. In both approaches, the plasmid pET11c-thisF (19) was used as the template. The following mutagenic 5Ј-primers were used (NdeI restriction sites are in boldface type, and base substitutions to introduce an amino acid exchange are underlined): 5Ј-TGA TGA AGA CAT ATG CTC GCT AAA AGA ATA ATC GCG GCT CTC GAT-3Ј for construction of C9A, 5Ј-TGA TGA AGA CAT ATG CTC GCT AAA AGA ATA ATC GCG TGC CTC AAT GTG AAA GAC-3Ј for construction of D11N, and 5Ј-TGA TGA AGA CAT ATG CTC GCT AAA AGA ATA ATC GCG TGT CTC GAT GTG AAA GAC GGT CGT GTG GTG AGC GGA ACG AAC TTC-3Ј for construction of K19S. In all PCRs, the oligonucleotide 5Ј-CCG GAT CCA GCG TCA TCA CAA-3Ј containing a BamHI restriction site (in boldface type) was used as the 3Ј-primer. For the production of megaprimers, the following mutagenic oligonucleotides were used (base substitutions to introduce restriction sites for the control of the reaction are in boldface type, and base substitutions to introduce an amino acid exchange are underlined): 5Ј-CGC GGT AAT ATT CAG AAA AAC GAG-3Ј with a new SspI restriction site for construction of D51N, 5Ј-CAC AGC CGC AGT GGC TAT GCT CAC CTT GTC-3Ј with a new TspRI site for construction of N103A, 5Ј-CAC TCT TTT TGC ATT AAT CGC CAC GAC-3Ј with a new VspI restriction site for construction of D130N, 5Ј-GAC AGA AAC GGC ACC AAA TCG G-3Ј with a new BshNI restriction site for construction of D176N, and 5Ј-GGC ACA AAA TCG GGT TAC AAC ACT GAG ATG ATA AGG-3Ј with a new TspRI restriction site for construction of D183N. For production of the megaprimers for thisF_D51N, N103A, and D130N, the corresponding mutagenic oligonucleotides listed above were used as 3Ј-primers, and the oligonucleotide 5Ј-TGA TGA AGA CAT ATG CTC GCT AAA AG-3Ј (NdeI restriction site in boldface type) was used as 5Ј-primer. The megaprimers were purified by agarose gel electrophoresis and used in the second PCR as 5Ј-primers, whereas the oligonucleotide 5Ј-CCG GAT CCA GCG TCA TCA CAA-3Ј (BamHI restriction site in boldface type) was used as 3Ј-primer. The construction of the variants thisF_D176N and D183N followed the same protocol except that in the first PCR the mutagenic oligonucleotides were used as 5Ј-primers and the oligonucleotide 5Ј-CCG GAT CCA GCG TCA TCA CAA-3Ј (BamHI restriction site in boldface type) was used as 3Ј-primer. The purified megaprimers were used as 3Ј-primers, and the oligonucleotide 5Ј-TGA TGA AGA CAT ATG CTC GCT AAA AG-3Ј (NdeI restriction site in boldface type) was used as 5Ј-primer. The resulting full-length products were digested with NdeI and BamHI and ligated into pET11c. In order to confirm the base substitutions and to exclude inadvertent additional ones, all thisF gene variants were entirely sequenced.
Randomization of thisF Codons-The thisF codons representing amino acids 11 and 130 were randomized in PCR-based approaches using degenerated primers and pET11c-thisF as template. For randomization of position 11, the oligonucleotide 5Ј-TAT ACG CAT GCT CGC TAA AAG AAT AAT CGC GTG CCT CNN SGT GAA GAC-3Ј with a SphI restriction site (in boldface type) was used, where N represents equal molar mixtures of all four bases, and S represents an equal molar mixture of G and C. In a PCR, the degenerate oligonucleotide was used as 5Ј-primer, and the oligonucleotide 5Ј-GTC GAC GGA TCC ACA ACC CCT CCA G-3Ј with a BamHI restriction site (in boldface type) was used as 3Ј-primer to yield an ensemble of thisF_D11NNS gene variants. For randomization of position 130, a megaprimer was produced using the degenerate oligonucleotide 5Ј-GTC GTG GCG ATT NNS GCA AAA AGA-3Ј as 3Ј-primer and the oligonucleotide 5Ј-TAT ACG CAT GCT CGC TAA AAG AAT AAT CGC-3Ј with a SphI restriction site (in boldface type) as 5Ј-primer. In a second PCR, the purified megaprimer was used as 5Ј-primer, and the oligonucleotide 5Ј-GTC GAC GGA TCC ACA ACC CCT CCA G-3Ј with a BamHI restriction site (in boldface type) was used as 3Ј-primer to yield an ensemble of thisF_D130NNS gene variants.
In Vivo Complementation-The PCR-amplified ensembles of thisF genes containing randomized codons at amino acid position 11 or 130 were ligated into a modified pDS56/RBSII vector (termed pTNA), which contains a truncated derivative of the tryptophanase operon promotor (20) that permits constitutive gene expression in E. coli. The auxotrophic E. coli strain UTH860 (⌬hisF), which carries a mutant ehisF gene that encodes an inactive HisF protein (21), was transformed separately with the two plasmid libraries and plated onto LB medium containing 150 g ml Ϫ1 ampicillin (15). The resulting lawns of at least 10 5 clones were rinsed off the plates, and plasmid DNA was prepared from the mixture of colonies. The ⌬hisF strain was retransformed with these plasmid libraries, and aliquots were streaked on LB medium plates (nonselective plates) or on minimal medium plates (22) without histidine (selective plates) and incubated at 37°C. From the nonselective plates, 16 colonies with randomized thisF codon 11 and 18 colonies with randomized thisF codon 130 were picked, and the corresponding thisF genes were entirely sequenced.
The mutational saturation is given by the following, where p is the probability that, for m randomized codons appearing with the relative frequency f i , practically all possible amino acid combinations are present in a library containing n independent clones (23). Under the assumption that all 32 permitted codons appear with equal frequency (f i ϭ 1 ⁄32), mutational saturation (p Ն 0.99) of one codon (m ϭ 1) is attained with a library of more than 145 independent clones. Although the bases within codons 11 and 130 were not randomly distributed, every allowed base was found at all three codon positions. None of the thisF genes contained additional mutations. Considering the size of the two libraries, each of which contained at least 10 5 independent clones, it can be concluded that all 20 amino acids were represented at both randomized positions. To select for functional amino acids at position 11 or 130 of tHisF, for each library one selective plate was incubated for various time periods. A number of clones appeared on both plates overnight and, after 48 h, additional colonies appeared on the selective plate with thisF genes that were randomized at position 130. Incubation was continued for 1 week, but no additional colonies appeared. A number of colonies that grew on selective medium after different periods of time were picked, and the thisF_D11NNS or thisF_D130NNS genes that encoded functional tHisF proteins were sequenced.
Purification of tHisF, tHisF Variants, and tHisH-Wild-type tHisF and its variants containing individual amino acid exchanges were purified as described (19). The yield was between 7 and 17 mg of purified enzyme per g of wet cell mass. Heterologous expression of thisH was conducted in E. coli W3110 ⌬trpEA2 cells containing pDS56/ RBSIISphI-thisH and the repressor plasmid pDMI, 1, as described for hisA from T. maritima (19). The cells were grown in 1 liter of LB medium supplemented with 0.15 mg/ml ampicillin and 0.075 mg/ml kanamycin. Overexpression of thisH was induced by adding 1 mM isopropyl-1-thio-␤-D-galactopyranoside at an optical density at 600 nm of about 0.6, and incubation was continued overnight. The cell suspension was washed with 100 mM potassium phosphate buffer at pH 7.5, containing 1 mM EDTA and 1 mM dithiothreitol, resuspended (5 ml of buffer per g wet cell mass), and lysed by sonification (Branson Sonifier W-250, 2 ϫ 2 min, 50% pulse, 0°C). According to SDS-PAGE, about 60% of tHisH were found in the soluble fraction of the cell extract. Benzonase (Merck) (50 units) was added to this fraction, which was then incubated for 1 h at 37°C and subsequently for 20 min at 75°C. The resulting suspension was centrifuged (Sorvall SS34, 12,000 rpm, 30 min, 4°C), and the pellet, which contained heat-labile host proteins, was discarded. The supernatant was dialyzed against 10 mM Tris/HCl buffer at pH 8.0, containing 1 mM EDTA and 1 mM dithiothreitol, and loaded on an anion exchange column (POROS HQ20; 1 ϫ 10 cm, PE Biosystems) that was equilibrated with the same buffer at room temperature. The column was washed with four volumes of equilibration buffer, and bound proteins were eluted with 1.5 liters of a linear gradient of 0 -1 M sodium chloride at pH 8.0. tHisH eluted between 130 and 150 mM sodium chloride, as judged from SDS-PAGE and conductivity measurements. Fractions containing tHisH were pooled, dialyzed against 10 mM potassium phosphate buffer at pH 7.5, containing 1 mM dithiothreitol, loaded on a hydroxylapatite column (3.6 ϫ 20 cm; Novartis) that was equilibrated with the same buffer, and eluted with 2 liters of a linear gradient of 10 -500 mM potassium phosphate. tHisH eluted between 100 and 150 mM potassium phosphate with a purity above 95%, as judged from SDS-PAGE. The purification yielded ϳ5 mg of tHisH per g of wet cells. Fractions containing pure tHisH were pooled, dialyzed against 50 mM potassium phosphate buffer at pH 7.5, containing 1 mM EDTA and 1 mM dithiothreitol, concentrated to 1.8 mg/ml using Centricon-10 concentration devices (Millipore), and shock-frozen in liquid nitrogen.
Analytical Methods-Purification of proteins was followed by electrophoresis on 12.5 or 15% SDS-polyacrylamide gels using the system of Laemmli (24) and staining with Coomassie Blue. During purification, the protein concentration was determined according to Bradford (25). The concentration of purified proteins was determined with molar extinction coefficients at 280 nm that were calculated from the amino acid sequence (26). Analytical gel filtration was performed at a flow rate of 0.5 ml/min on a Superdex 75 column (1 ϫ 30 cm; Amersham Pharmacia Biotech) that was equilibrated with 50 mM potassium phosphate at pH 7.5, containing 300 mM sodium chloride. Apparent molecular masses were determined from the corresponding elution volumes, using a calibration curve that was obtained with standard proteins. Sedimentation equilibrium runs were performed in a Beckman analytical ultracentrifuge (model Optima XLA), following the absorption at 278 nm. Runs with tHisF were performed as described (27). Runs with tHisH were performed at 24,000 rpm and protein concentrations of 12 and 23 M in 50 mM potassium phosphate, pH 7.5, at 20°C, containing 25 mM potassium chloride. Runs with tHisH-tHisF were performed at 18,000 and 24,000 rpm with 20 M protein in 50 mM potassium phosphate at pH 7.5 at 20°C, containing 300 mM sodium chloride. To determine apparent molecular masses, the runs were analyzed as described (27). Fluorescence spectra were recorded with a F-4500 spectrofluorimeter (Hitachi) or a Cary Eclipse spectrofluorimeter (Varian). Proteolytic The single tryptophan residue of tHisF has its emission maximum at 323 nm, indicating that it is shielded from solvent. The single tryptophan residue of tHisH has its emission maximum at 339 nm, indicating that it is partly accessible to solvent. In the tHisH-tHisF complex, the emission maximum lies at 326 nm, indicating that both tryptophan residues are shielded from solvent. b, formation of the stoichiometric tHisH-tHisF complex. 10 M tHisH was titrated with tHisF, and complex formation was followed by a shift of the fluorescence emission maximum (F em max ) from 339 nm to lower wavelengths. The titration curve is linear up to about equal molar concentrations of tH-isF and tHisH and remains constant at higher ratios, indicating that the thermodynamic dissociation constant for complex formation, K d , is Ͻ Ͻ10 M.
stability was tested at room temperature by incubating 10 nmol of substrate protein with 64 pmol of trypsin in 1 ml of 50 mM potassium phosphate, pH 7.5. The reaction was stopped after different time intervals by adding one volume of 2ϫ SDS-PAGE sample buffer and heating for 5 min at 95°C. The time course of proteolysis was followed on Tris-Tricine gels containing 20% acrylamide (28).
Steady-state Enzyme Kinetics-The ammonia-dependent activity of isolated tHisF was measured by recording entire progress curves in 50 mM Tris acetate buffer, pH 8.5, at 25°C. In order to determine K m PRFAR , the enzyme was saturated with ammonia by adding 100 mM ammonium acetate corresponding to 14.4 mM NH 3 at pH 8.5. PRFAR, 20 M, were synthesized in situ from ProFAR, using a molar excess of HisA from T. maritima (19) and completely converted by tHisF to ImGP and AICAR.
The reaction was quantified by the decrease in absorption at 300 nm, using ⌬⑀ 300 (PRFAR-AICAR) ϭ 5.64 mM Ϫ1 cm Ϫ1 (11). In order to determine K m NH3 , the reaction was performed in the presence of 50 M PRFAR at various concentrations of ammonium acetate between 0 and 200 mM, corresponding to 0 and 35 mM NH 3 at pH 8.5. The glutaminedependent activity of the tHisH-tHisF complex was measured in an analogous way as the ammonia-dependent reaction of tHisF, but the pH value was set to 8.0. In order to determine K m PRFAR , the reaction was performed with 20 M ProFAR and 5 mM L-glutamine. To determine K m Gln , the reaction was performed in the presence of 50 M PRFAR at various concentrations of glutamine between 0 and 7 mM. The progress curves were analyzed with the integrated form of the Michaelis-Menten equation (29), yielding values for K m and V max . The glutaminase activ-FIG. 2. The reactions catalyzed by the tHisH-tHisF complex and the individual enzymes were monitored spectroscopically using coupled enzymatic assays. a, physiological reaction of the imidazole glycerol phosphate synthase (tHisH-tHisF). b, tHisH-tHisF assay. PRFAR was synthesized in situ from ProFAR with a molar excess of HisA from T. maritima (tHisA); the glutaminedependent conversion of PRFAR into ImGP and AICAR was quantified using ⌬⑀ 300 (PRFAR-AICAR) ϭ 5640 M Ϫ1 cm Ϫ1 (11). c, tHisF assay. The ammoniadependent reaction was measured by replacing glutamine with ammonium acetate. d, tHisH assay. The glutamate produced by the glutaminase activity of tHisH was oxidized by a molar excess of glutamate dehydrogenase (GDH) to 2-oxoglutarate and ammonium; the reaction was quantified by the concomitant reduction of NAD ϩ to NADH ϩ H ϩ , using The ligand L bound to tHisF is either ImGP or ProFAR. The conditions for the different assays are listed in Table I (for  the synthase reactions) and Table II (for the glutaminase reaction). ity of tHisH in complex with liganded tHisF was measured in a coupled enzymatic assay with bovine liver glutamate dehydrogenase (Sigma). Glutamate that was produced by tHisH was oxidized by a molar excess of glutamate dehydrogenase in the presence of NAD ϩ , yielding 2-oxoglutarate and NADH ϩ H ϩ ϩ NH 4 ϩ . The reaction was quantified by the increase in absorption at 340 nm, using ⌬⑀ 340 (NADH-NAD ϩ ) ϭ 6300 M Ϫ1 cm Ϫ1 . The values for K m Gln and V max were deduced from initial velocity measurement at various glutamine concentrations.

RESULTS
Production and Purification of tHisF, tHisH, and the tHisH-tHisF Complex-tHisF was produced and purified as described previously (19). The thisH gene (16) was cloned into the expression vector pDS56/RBSII/SphI. Proteins were expressed in E. coli from this plasmid under the control of a lac operator system (17). tHisH was purified from the soluble fraction of the cell homogenate by first heat-precipitating E. coli host proteins, followed by anion exchange and hydroxylapatite chromatography. The final preparation was more than 98% pure as judged by SDS-PAGE and gel filtration chromatography on Superdex 75 (data not shown). The presence of the N-terminal methionine residues of both purified tHisF and tHisH were confirmed by N-terminal protein sequencing. The complex of tHisF and tHisH (tHisH-tHisF) was prepared by mixing equal molar amounts of both proteins, followed by gel filtration chromatography in order to remove any unintentional surplus of either of the components. The concentration of the purified proteins was determined by absorption spectroscopy, using calculated molar extinction coefficients at 280 nm (26) of 11,500 M Ϫ1 cm Ϫ1 for tHisF, 17,400 M Ϫ1 cm Ϫ1 for tHisH, and 28,900 M Ϫ1 cm Ϫ1 for the tHisH-tHisF complex.
Association States-The molecular masses and the association states of recombinant tHisF, tHisH, and tHisH-tHisF were determined by analytical gel filtration chromatography and equilibrium analytical ultracentrifugation. tHisF eluted from a calibrated Superdex 75 column with an apparent molecular mass of 26.4 kDa (27), which is similar to the molecular mass for the monomer as calculated from the amino acid sequence (27.7 kDa). tHisH eluted with a significantly lower apparent molecular mass (17.4 kDa) than the calculated molecular mass for the monomer (23.1 kDa). A similarly retarded elution has been observed previously for eHisH (11), for unknown reasons. Mixed equal molar amounts of tHisH and tHisF eluted as a single peak with an apparent molecular mass of 41.8 kDa, which is smaller than the calculated molecular mass for the 1:1 complex (50.8 kDa). In order to further clarify their association states, the molecular masses of tHisH, tHisF, and tHisH-tHisF were determined by analytical ultracentrifugation. The data from several independent sedimentation equilibrium runs yielded average molecular masses of 24.6 kDa for tHisH, 49.8 kDa for tHisH-tHisF, and, dependent on protein concentration, between 26.0 and 38.4 kDa for tHisF (27). These results confirm that isolated tHisH and tHisF are essentially monomeric proteins that assemble to a stable 1:1 heterodimeric bienzyme complex. These association states of the T. maritima proteins are identical to those of the E. coli homologs (11), which excludes the idea that a higher association state is responsible for the increased thermostability of the ImGP synthase from the hyperthermophilic compared with that of the mesophilic bacterium (30).
Fluorescence Spectroscopic Analysis-Fluorescence emission spectra were monitored to determine the relative solvent accessibility of the single tryptophan residues in tHisF (Trp 156 ) and tHisH (Trp 123 ). As judged from the emission maxima at 323 and 339 nm, Trp 156 in tHisF appears to be shielded from solvent, while Trp 123 in tHisH appears to be partly exposed to solvent (Fig. 1a). A solution containing equal molar amounts of tHisF and tHisH showed an emission maximum at 326 nm, indicating that complex formation leads to the burial of Trp 123 of tHisH but does not change the solvent accessibility of Trp 156 in tHisF. Along these lines, the titration of tHisH with tHisF leads to a linear shift of the emission maximum from 339 nm to lower wavelengths, until approximately equal molar amounts of tHisH and tHisF are present. An excess of tHisF does not lead to a further shift of the emission maximum (Fig. 1b). These data suggest that, at the applied protein concentrations, the binding of tHisF and tHisH is stoichiometric, indicating a high affinity of the two proteins. Although the thermodynamic dissociation constant K d cannot be determined by this method, it has to be much smaller than the equivalence concentration, which is 10 M.
Steady-state Enzyme Kinetics-The catalytic activities of isolated tHisF and tHisH and of the tHisH-tHisF complex were measured under steady-state conditions with coupled enzymatic assays, using absorption spectroscopy (Fig. 2). The synthase activities of the tHisH-tHisF complex and of isolated tHisF were measured, using glutamine and ammonium acetate as ammonia donors, respectively (Fig. 2, b and c). The glutaminase activity of tHisH was measured in the presence and absence of tHisF and the ligands ImGP or ProFAR (Fig. 2d), which were shown to activate the glutaminase activity of eHisH when bound to the active site of eHisF (11). Entire progress curves or initial velocities were measured and fitted to the integrated or the simple Michaelis-Menten equation, respectively. The resulting k cat and K m values were compared with those of the E. coli homologs eHisH, eHisF, and eHisH- eHisF (11) and with those of HIS7 from Saccharomyces cerevisiae, in which the glutaminase and the synthase domains are fused (13). Table I summarizes the results of the synthase activity measurements. In the ammonia-dependent reaction, at comparable temperatures and pH values, isolated tHisF exhibited 2-5-fold higher catalytic efficiencies (k cat /K m PRFAR and k cat / K m NH3 ) than isolated eHisF. The higher catalytic efficiency was caused by lower K m values of tHisF for both PRFAR and ammonia, while the turnover rate k cat was slightly lower for tHisF than for eHisF. In the glutamine-dependent reaction, the tHisH-tHisF complex exhibited catalytic activities comparable to tHisF in the ammonia-dependent reaction. In contrast, the catalytic efficiency k cat /K m PRFAR of the eHisH-eHisF complex in the glutamine-dependent reaction was about 25-fold higher than that of eHisF in the ammonia-dependent reaction, due to a much higher K m PRFAR value of eHisF. The decreased activity of eHisF in the ammonia-dependent reaction might be due to inhibition by high concentrations of ammonium chloride, the ammonia donor (11). tHisF was also inhibited by ammonium chloride, which was therefore replaced by ammonium acetate that showed no detectable inhibitory effect at the applied concentrations (data not shown). Similar catalytic efficiencies were observed for the glutamine-dependent synthase reactions of the bifunctional HIS7 enzyme and the tHisH-tHisF complex (Table I).
In the absence of tHisF, tHisH did not show any measurable glutaminase activity. However, as observed with the E. coli enzyme (11), the binding of either the product ImGP or the substrate analogue ProFAR to the active site of tHisF led to a strong stimulation of tHisH. In contrast, neither the T. maritima, nor the E. coli HisH glutaminase activities are stimulated by AICAR, the second product of the ImGP synthase reaction (Ref. 11 and data not shown). The steady-state parameters of the glutaminase activity of tHisH were therefore measured in the presence of tHisF and saturating concentrations of either ImGP or ProFAR (Table II). The k cat of the glutaminase reaction is 0.1 s Ϫ1 and lower by a factor of about 4 -8 than the k cat value of the glutamine-dependent synthase reaction (Table  I), where PRFAR (instead of ImGP or ProFAR) is bound to the active site of tHisF in the tHisH-tHisF complex. Similarly, in the E. coli enzyme, the k cat of the glutaminase reaction is lower by a factor of 3-4 than the k cat of the glutamine-dependent synthase reaction. In addition, the K m values for glutamine are increased about 10-fold (Tables I and II). Obviously, both in the T. maritima and the E. coli ImGP synthase, the activation of the glutaminase reaction by the synthase subunit is more efficient when the native substrate PRFAR is bound rather than ImGP or ProFAR. In contrast, in HIS7, where glutaminase and synthase activities are located on the same bifunctional polypeptide chain, ImGP and ProFAR are equally efficient glutaminase activators as PRFAR (Tables I and II). The dependence of the glutaminase activity of tHisH on the presence of a ligand bound to the active site of tHisF can be used to estimate the thermodynamic dissociation constants K d for the binding of these ligands. As shown in Table II, ImGP binds with similarly low affinity to tHisF and eHisF, whereas ProFAR binds almost 20-fold more strongly to tHisF than to eHisF.
Limited Proteolysis-The sensitivity of a protein toward proteolytic degradation can provide valuable information on its stability as well as on the flexibility of potential cleavage sites (31). For this reason, tHisF, tHisH and tHisH-tHisF were digested with trypsin, and the results were analyzed by SDS-PAGE (Fig. 3). Isolated tHisF (27.7 kDa) was cleaved once to yield fragments with apparent molecular masses of about 24 kDa (Fig. 3a) and about 4 kDa (not shown). These fragments, which were detectable already after 5 min of incubation, were enriched at the cost of intact tHisF in a time-dependent manner. After 120 min, about one-quarter of tHisF was degraded (Fig. 3a). N-terminal sequencing of the large fragment showed that cleavage occurred C-terminal of Arg 27 (Fig. 4). Arg 27 is located in a long and presumably flexible region at the Cterminal (active site) face of the ␤-barrel (cf. arrow in Fig. 4a). In contrast to tHisF, at the given conditions isolated tHisH was completely resistant to trypsin (data not shown). Incubation of the tHisH-tHisF complex with trypsin again resulted in the degradation of tHisF, albeit with significantly higher rate; after 120 min, almost all tHisF was degraded into fragments of about 24 kDa (Fig. 3b) and about 4 kDa (not shown). These results show that the single trypsin cleavage site at Arg 27 is more susceptible in complexed tHisF compared with isolated tHisF.
Mutational Analysis of the Active Site of tHisF-Glutamine hydrolysis reactions catalyzed by other members of the class I GATase family have been characterized in detail, for example in the case of carbamoyl phosphate synthase (5,32,33). In contrast, the complicated reaction mechanism that leads to the formation of the imidazole ring catalyzed by HisF (Fig. 2a) has not been investigated so far. The first step toward this goal is to identify the catalytic amino acid residues. These should (a) be invariant in the 25 currently available HisF sequences; (b) lie at the C-terminal face of the central ␤-barrel of tHisF, which is the location of the active sites of all known (␤␣) 8 -barrel enzymes (34); and (c) be able to reversibly provide and abstract a proton, since the HisF reaction probably involves general acid/ base catalysis. Eight residues that fulfill these criteria were replaced by site-directed mutagenesis with structurally similar amino acids lacking the putative functional group (Fig. 4a). The mutant thisF genes were expressed in E. coli, and the respec- 3. Isolated tHisF is less susceptible to proteolytic attack than complexed tHisF. 10 M isolated tHisF (a) or a 10 M concentration of the tHisH-tHisF complex (b) was digested with 64 nM trypsin for the indicated time intervals in 50 mM potassium phosphate, pH 7.5, at 25°C. tHisF (27.7 kDa) is degraded to one large fragment with M app of ϳ24 kDa and one small fragment with M app of ϳ4 kDa (not shown). This cleavage, which takes place C-terminal of Arg 27 (cf. arrow in Fig. 4), occurs more slowly in isolated compared with complexed tHisF, indicating conformational changes in the flexible loop region between strand ␤ 1 and helix ␣ 1 . tHisH (23.1 kDa) is resistant to trypsin, both in the tHisH-tHisF complex (b) and in the isolated form (not shown). Band 1, tHisF; band 2, large tryptic fragment of tHisF (a) or a mixture of tHisH and the large tryptic fragment of tHisF (b). M, standard proteins. 4. Conserved and catalytically important amino acid residues in the (␤␣) 8 -barrel structure of tHisF (14). a, structure-based sequence alignment. Secondary structural elements are identified by blue arrows (␤ strands) and red cylinders (␣ helices). Conserved, amino acids that are invariant in all known 25 HisF sequences and identical in at least 22 sequences are shown in uppercase and lowercase, respectively. Exchanged, eight conserved amino acid residues close to the proposed active site of tHisF were replaced individually with structurally similar residues (in boldface type). Amino acids involved in phosphate binding are underlined. The single trypsin cleavage site within tHisF, which is located C-terminal to Arg 27 , is marked by an arrow (cf. Fig. 3). b, ribbon diagrams showing a side view on the central ␤-barrel. P(N) and P(C) are the phosphate ions bound to the N-and C-terminal halves of tHisF. Asp 11 and Asp 130 (D11 and D130) are essential for catalysis of the tHisF reaction, Asp 176 (D176) appears to be important for catalysis, and the flexible region containing Lys 19 (K19) and Arg 27 (R27) is involved in interactions between tHisF and tHisH that take place upon the reaction of nascent ammonia with PRFAR (side chains in green). Side chains of conserved amino acids at the N-terminal face of the ␤-barrel, which might be involved in binding of tHisH, are in black (see "Results" and "Discussion" for details.) tive gene products were purified in the same way as described for the wild-type protein (19). In order to exclude structural perturbations (or denaturation) caused by the introduced amino acid replacements, all purified tHisF variants were analyzed by fluorescence spectroscopy and analytical gel filtration chromatography. No significant difference from wild-type tHisF was detected in any of the variants, confirming the assumption that residues at the C-terminal face of (␤␣) 8 -barrels do not contribute significantly to protein stability (35). Also, the association with tHisH was not impaired (data not shown).
The steady-state enzyme kinetic constants of the isolated tHisF variants and of the corresponding tHisH-tHisF complexes were determined and compared with the constants of the wild-type enzymes (Tables III and IV). The catalytic efficiencies k cat /K m of both isolated and complexed tHisF_C9A, tHisF_D51N, tHisF_N103A, and tHisF_D183N were not significantly different from wild-type tHisF, ruling out any central catalytic role for the replaced residues. Also, the ammonia-dependent reactions of isolated tHisF_K19S were similarly efficient as those of wild-type tHisF (Table III). In contrast, the efficiencies of the glutamine-dependent reactions of the tHisH-tHisF_K19S complex were significantly impaired (Table IV). The variant tHisF_D176N showed a 40 -50 fold decrease in k cat , both in isolated form and in complex with tHisH. The strongest effects, however, were found for tHisF_D11N and tHisF_D130N, the catalytic efficiencies k cat /K m PRFAR of which were decreased by approximately 5 orders of magnitude. While for tHisF_D11N only the k cat was affected, both k cat and K m were drastically impaired in tHisF_D130N. These results suggest that Asp 11 and Asp 130 play essential roles in the catalysis of the tHisF reaction.

Saturation Random Mutagenesis and Complementation in
Vivo-To further test the crucial role of Asp 11 and Asp 130 for catalysis, both residues were replaced by all 19 alternative amino acids, and their function in catalysis was probed by selection in vivo. To this end, the corresponding codons at amino acid positions 11 or 130 were subjected to saturation random mutagenesis using degenerate oligonucleotides. The randomization of the third codon position was limited to G and C. This restriction eliminates 32 out of 64 codons, but the remainder codons still represent all 20 amino acids.
Histidine auxotrophic E. coli cells lacking a functional hisF gene (⌬hisF cells) were transformed with a plasmid that allows constitutive expression in E. coli of the cloned thisF_D11NNS or thisF_D130NNS ensembles (20). Transformants were streaked onto selective medium without histidine and incubated at 37°C; a small aliquot was streaked onto nonselective LB medium. From the nonselective plates, colonies containing thisF_D11NNS or thisF_D130NNS were randomly picked, and the thisF genes were sequenced in order to confirm the random distribution of bases at each position of the codons. Sequencing of thisF from colonies grown on selective medium showed that at position 11 only the wild-type amino acid aspartate allowed functional complementation of the ⌬hisF cells. Similarly, all overnight grown colonies on the selective plate with the randomized codon at position 130 coded for the wild-type amino acid aspartate. However, all colonies that appeared on this plate after 48 h contained a codon for glutamate at position 130 (Table V).
In order to test its catalytic activity in vitro, tHisF_D130E was produced in E. coli, purified and characterized as described above for the other tHisF variants. tHisF_D130E was identical  to the wild-type enzyme with respect to fluorescence properties and complex formation with tHisH (data not shown). Its steady-state enzyme kinetic parameters, however, were between those of wild-type tHisF and tHisF_D130N; the k cat value was reduced by a factor of about 400 -500, and the K m PRFAR was increased almost 20-fold (Table III). DISCUSSION Catalytic Properties and Allosteric Interactions of T. maritima ImGP Synthase-Most investigated enzymes from the hyperthermophilic bacterium T. maritima are only marginally active at room temperature, probably due to conformational rigidity at their active sites (36). In contrast, at comparable temperatures of 25 and 30°C, the isolated synthase subunit tHisF shows a higher catalytic efficiency (k cat /K m PRFAR and k cat /K m NH3 ) than eHisF, due to much lower K m values (Table I). Similarly, both phosphoribosyl anthranilate isomerase and indoleglycerol phosphate synthase from T. maritima, which are involved in tryptophan biosynthesis, have lower K m values and higher catalytic efficiencies than their homologues from E. coli (37,38). It has been speculated that the high catalytic efficiency of T. maritima phosphoribosyl anthranilate isomerase is important for rapid processing of its thermolabile substrate, and a similar argument may hold for the procession by tHisF of the extremely labile PRFAR (39). However, the catalytic efficiencies of the overall tHisH-tHisF reaction (k cat /K m PRFAR and k cat /K m Gln ; Table I) and of the glutaminase reaction (k cat /K m Gln ; Table II) were significantly lower than those of the corresponding enzymes from E. coli and S. cerevisiae. Thus, the catalytic activity of tHisH seems to limit, at least at room temperature, the overall catalytic efficiency of T. maritima ImGP synthase. This result indicates conformational rigidity of tHisH, which is supported by its resistance to digestion by trypsin (Fig. 3).
The quaternary structure of the tHisH-tHisF complex has to provide the basis for (a) the activation of the glutaminase reaction at the active site of tHisH by the binding of PRFAR (or ImGP or ProFAR) to the active site of tHisF and (b) the transfer of nascent ammonia between tHisH and tHisF in a way that excludes contact with water, which inevitably would protonate ammonia to the nonreactive ammonium ion. The active site of tHisF is located at the C-terminal face of the central ␤-barrel (Fig. 4b), but several lines of evidence suggest that tHisH docks to the N-terminal face of the barrel. About half of the highly conserved and invariant residues are located close to the Nterminal face of the barrel of tHisF, indicating an important functional role of this region (Fig. 4). Two of these conserved residues, Arg 5 and Glu 46 , were shown to be important for the glutamine-dependent but not for the ammonia-dependent reaction of the eHisH-eHisF complex (40). Furthermore, the region between strand ␤ 1 and helix ␣ 1 , which is located at the C-terminal face of tHisF (Fig. 4), is not protected by tHisH against trypsinolysis (Fig. 3). Moreover, docking of tHisH to the C-terminal face of tHisF would impede access of the large substrate PRFAR and should therefore impair catalytic activity. However, both the K m PRFAR values and the catalytic efficiencies k cat /K m PRFAR are practically identical for isolated tHisF and tHisF in complex with tHisH (Table I).
How is the binding of PRFAR to tHisF at the C-terminal face of the ␤-barrel coupled with the glutaminase activity of tHisH, which is presumably bound to the N-terminal face of the barrel? Arg 27 is located in a flexible region between strand ␤ 1 and helix ␣ 1 at the C-terminal face of tHisF (Fig. 4). The cleavage of trypsin at Arg 27 is accelerated in the tHisH-tHisF complex compared with isolated tHisF (Fig. 3). This result indicates that the binding of tHisH to the N-terminal face of tHisF induces a long range conformational transition that makes Arg 27 more available to proteolytic cleavage. The amino acid exchange K19S, moreover, which lies in the same flexible region as Arg 27 (Fig. 4), causes larger effects of k cat and K m PRFAR in the glutamine-dependent reaction of the tHisH-tHisF complex (Table IV) than in the ammonia-dependent reaction of FIG. 5. Plausible mechanism of the tHisF reaction based on general acid/ base catalysis. R 1 , ribosephosphate, R 2 , 3-phosphoglycerol. The general acid H-A 1 and the general base A 2 Ϫ are probably identical with Asp 11 and Asp 130 , respectively (see "Discussion" for details.) isolated tHisF (Table III). Thus, the flexible region between strand ␤ 1 and helix ␣ 1 is involved in interactions between tHisF and tHisH that take place upon the reaction of nascent ammonia with PRFAR. Since uncharged ammonia is the nucleophile to attack PRFAR (Fig. 2), ammonia has to be transferred from tHisH to tHisF without contact with solvent. As a consequence, the transfer will probably occur through a hydrophobic channel connecting the two active sites, but a high resolution x-ray structure of the tHisH-tHisF complex is necessary to verify this hypothesis. Ammonia channels of this kind were recently identified in the GATases carbamoyl phosphate synthase (4,41), glutamine phosphoribosyl pyrophosphate amidotransferase (42), and asparagine synthase B (43).
Plausible Mechanism of the HisF Reaction-The mechanism of the cycloligase/lyase HisF reaction (Fig. 2a) is complicated and unique in primary metabolism (13). Two aspartate residues, Asp 11 and Asp 130 , are catalytically essential, since their replacement by asparagines causes an almost complete loss of tHisF activity (Table III). Saturation random mutagenesis of the corresponding codons and complementation studies showed that the function of Asp 11 could not be substituted by any other residue. In contrast, the variant tHisF_D130E complemented a ⌬hisF strain in vivo, albeit more slowly than wild-type tHisF (Table V). Thus, the restriction for aspartate at position 130 is not as exclusive as that at position 11, but a carboxylate side chain appears to be essential at both positions. These results support the hypothesis that the tHisF reaction depends on general acid/base catalysis, and a chemically plausible reaction mechanism was developed on the basis of this finding (Fig. 5). The proposed sequence of reactions starts with the substitution of the phosphoribulosyl carbonyl oxygen of PRFAR by ammonia, resulting in the release of water and the formation of Imine I. In the next step, the addition of water results in the generation of the first reaction product AICAR and of Imine II. In the subsequent step, the imidazole ring of ImGP is closed, and water is released in a reaction that is catalyzed by the general acid H-A 1 and the general base A 2 Ϫ . In support of this mechanism, ProFAR (and probably also PRFAR) are slowly hydrolyzed in the absence of enzyme to yield AICAR and other unidentified products, and the rate of this reaction increases with increasing proton concentration (44). In order to act as a general acid at physiological pH, the pK a value of an aspartate within an enzyme must be shifted by several units compared with free aspartate. Such a shift should be very sensitive to the immediate environment of the carboxylate function and therefore to the length of the amino acid side chain. As shown by complementation in vivo, Asp 11 is absolutely required for catalysis and cannot be functionally replaced by any other amino acid (Table V); it might therefore be identical with the general acid H-A 1 . Asp 130 can be functionally replaced by Glu (Table V) and therefore probably has a more "normal" pK a value; it might therefore be identical with the general base A 2 Ϫ . The reactions from PRFAR to Imine I and of Imine I to Imine II might also be accelerated by general acid/base catalysis, and Asp 11 and Asp 130 could play an essential role here as well. Asp 176 could also be involved in either of the reaction steps, since the k cat values of the tHisF_D176N variant are decreased about 50-fold compared with wild-type tHisF (Tables III and IV).