Crystal Structure of N-Succinylarginine Dihydrolase AstB, Bound to Substrate and Product, an Enzyme from the Arginine Catabolic Pathway of Escherichia coli*

The ammonia-producing arginine succinyltransferase pathway is the major pathway in Escherichia coli and related bacteria for arginine catabolism as a sole nitrogen source. This pathway consists of five steps, each catalyzed by a distinct enzyme. Here we report the crystal structure of N-succinylarginine dihydrolase AstB, the second enzyme of the arginine succinyltransferase pathway, providing the first structural insight into enzymes from this pathway. The enzyme exhibits a pseudo 5-fold symmetric α/β propeller fold of circularly arranged ββαβ modules enclosing the active site. The crystal structure indicates clearly that this enzyme belongs to the amidinotransferase (AT) superfamily and that the active site contains a Cys–His-Glu triad characteristic of the AT superfamily. Structures of the complexes of AstB with the reaction product and a C365S mutant with bound the N-succinylarginine substrate suggest a catalytic mechanism that consists of two cycles of hydrolysis and ammonia release, with each cycle utilizing a mechanism similar to that proposed for arginine deiminases. Like other members of the AT superfamily of enzymes, AstB possesses a flexible loop that is disordered in the absence of substrate and assumes an ordered conformation upon substrate binding, shielding the ligand from the bulk solvent, thereby controlling substrate access and product release.

transferase (AST) pathway, the arginine transaminase, oxidase, and oxygenase pathways, the arginine decarboxylase pathway, as well as others (1,2). Pseudomonas aeruginosa has four of these pathways (1). These pathways often have distinctive functions. For example, the arginine deiminase pathway generates carbamoyl phosphate for substrate level phosphorylation when oxygen is limiting (1). The presence of a particular arginine catabolic pathway may increase the ability of an organism to inhabit a much broader ecological niche.
Escherichia coli and related bacteria have two such metabolic routes (EcoCyc; ecocyc.org/ (3)): the arginine decarboxylase and AST pathways (4). The AST pathway accounts for 97% of arginine catabolism, whereas the arginine decarboxylase pathway accounts for only 3% (4). The AST pathway converts the carbon skeleton of arginine into glutamate, with the concomitant production of ammonia and conversion of succinyl-CoA to succinate and CoA (Fig. 1). The AST pathway consists of five enzymes: arginine succinyltransferase (AstA, EC 2.3.1.109), succinylarginine dihydrolase (AstB, EC 3.-.-.-), succinylornithine transaminase (AstC, EC 2.6.1.-), succinylglutamic semialdehyde dehydrogenase (AstD, EC 1.2.1.-), and succinylglutamate desuccinylase (AstE, EC 3.5.1.-), all contained within the astCADBE operon (aruCFGDBE operon in P. aeruginosa (5)). Nitrogen limitation induces transcription of the operon; ast mutants cannot utilize arginine as a nitrogen source and are impaired in ornithine utilization (4). Therefore, one function of the AST pathway is to provide nitrogen during nitrogen restriction. The ammonia produced is assimilated into glutamate and glutamine, which in turn provides nitrogen for the synthesis of virtually all nitrogen-containing compounds. Entry into stationary phase also induces the ast operon, and an ast mutant strain survives only poorly under conditions of carbon starvation (6,7). The observed phenotype may result from diminished generation of citric acid cycle intermediates. Slow growth lowers polyamine pools, and nitrogen limitation induces a variety of polyamine catabolic operons (8). The AST pathway has also been proposed to contribute to polyamine homeostasis by controlling levels of intracellular arginine and ornithine, the substrates for putrescine synthesis (8). Because arginine catabolism and the AST pathway perform several important physiological functions in E. coli, analogs of AST intermediates have been suggested as potential antimicrobial agents against pathogenic E. coli (9).
Of the five E. coli AST enzymes two have homologs (at a level of ϳ35% sequence identity) with known three-dimensional structures, namely AstC (Protein Data Bank codes 1OAT (10) and 1SFF (11)) and AstD 2 (Protein Data Bank code 1UZB; www.rcsb.org/pdb). Succinylglutamate desuccinylase AstE was predicted to be a member of the zinc-dependent carboxypeptidase family (12). Recently, Shirai and Mizuguchi (9), using sophisticated sequence analysis and fold recognition tools, proposed assignment of AstA and AstB to the acyl-CoA N-acyltransferase and amidinotransferase (␤/␣-propeller) fold families, respectively. N-succinylarginine dihydrolase (AstB), the second enzyme in the AST pathway, converts N-succinylarginine into N-succinylornithine with the release of ammonia and carbon dioxide (Scheme 1). The residues involved in catalysis were proposed to be Asp 173 , His 248 , and Cys 365 , with the cysteine playing the role of a nucleophile (9). Here we report the crystal structure of AstB from E. coli, its C365S mutant, and their complexes with substrate and product. The protein does indeed have the ␤/␣-propeller fold and contains a Cys-His-Asp catalytic triad with similarity to other amidinotransferases, suggesting a similar catalytic mechanism.

Cloning, Expression, and Purification
The astB gene was cloned into a derivative of the pET-15b vector (Amersham Biosciences). The C365S point mutation was introduced using QuikChange TM mutagenesis according to the manufacturer's instructions (Stratagene) and verified by DNA sequencing. The BL21(DE3) strain was transformed by the plasmid DNA, and the cells were grown at 37°C to an A 600 of ϳ0.8 in Circle Grow medium (Bio101 Inc.). Recombinant protein expression was induced by addition of isopropyl-1-thio-␤-D-galactopyranoside to a final concentration of 100 M in a 1-liter culture that was maintained at room temperature for an additional 15 h. Selenomethionine-labeled protein was prepared by transforming the E. coli methionine auxotroph DL41(DE3) with the plasmid DNA, and the cells were grown in LeMaster medium supplemented with 25 mg/liter of L-selenomethionine for selenomethionine labeling (13).
The cells were harvested by centrifugation (4000 ϫ g, 4°C, 25 min) and were resuspended 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 ␤-mercaptoethanol) containing one dissolved tablet of Complete TM protease inhibitor mixture (Roche Applied Science). The cells were lysed by sonication on ice for a total of five 30-s pulses with 45 s between each pulse for cooling. The lysate was then cleared by centrifugation (100,000 ϫ g, 4°C, 30 min). The protein supernatant was loaded on a 5-ml DEAE-Sepharose (Amersham Biosciences) column equilibrated with lysis buffer, and the flow-through fraction was collected and applied to a 5-ml nickel-nitrilotriacetic acid column (Qiagen), pre-equilibrated with lysis buffer. The column was washed extensively with buffer (50 mM Tris-HCl, pH 7.5, 50 mM imidazole, 0.4 M NaCl), and bound protein was eluted with the same buffer containing 150 mM imidazole. The protein was subsequently concentrated for crystallization with a concomitant buffer exchange by ultrafiltration to 50 mM Tris-HCl, pH 7.5, 0.2 M NaCl, 5% (w/v) glycerol, 10 mM dithiothreitol.
Dynamic light scattering measurements were carried out at 22°C on a DynaPro Plate Reader (Protein Solutions, Inc., Charlottesville, VA) at a protein concentration of 4 mg/ml. Gel filtration chromatography was performed using a Superose-12 column equilibrated in buffer (50 mM Tris-HCl, pH 7.5, 0.2 M NaCl) connected to an Akta Express fast protein liquid chromatography system (Amersham Biosciences). A sample of purified AstB was loaded at a flow rate of 0.8 ml/min, and the elution volume, V e , was determined. The apparent molecular weight was calculated using a standard curve of protein markers from a gel filtration calibration kit (Sigma).

Crystallization and Data Collection
The initial crystallization conditions were identified by sparse matrix screening using Screen I and II (Hampton Research, Liguna Niguel, CA). A triclinic crystal form was obtained from the His-tagged, selenomethionine-labeled protein after 5 days at 20°C in hanging drops containing 2 l of protein (7.3 mg/ml) in buffer (50 mM Tris-HCl, pH 7.5, 0.2 M NaCl, 5% (w/v) glycerol, 10 mM dithiothreitol, 4 mM glutamate) and 2 l of reservoir solution (15% (w/v) polyethylene glycol 10,000, 100 mM HEPES buffer, pH 7.5). These P1 crystals diffract to 2.3 Å resolution and have unit cell dimensions a ϭ 55.6 Å, b ϭ 93.8 Å, c ϭ 139.4 Å, ␣ ϭ 104.7°, ␤ ϭ 101.5°, and ␥ ϭ 90.0°with six monomers/asymmetric unit. Three Multiwavelength Anomalous Diffraction data sets (peak, inflection, and remote) about the selenium K absorption edge were collected at Beamline X8C (National Synchrotron Light Source, Brookhaven National Laboratory) using a Quantum-4 CCD area detector (Area Detector Systems Corporation, San Diego, CA) ( Table I). An additional data set was collected on a Micromax 007 rotating anode equipped with Osmic mirrors and an HTC image plate detector (Rigaku/MSC, The Woodlands, TX) ( Table I).
Sparse matrix screening of AstB in the presence of 5 mM N-succinylarginine resulted in an orthorhombic crystal form of the complex. The best crystals grew in hanging drops containing 2 l of protein (6 mg/ml) in buffer (0.2 M NaCl, 5% (v/v) glycerol, 10 mM dithiothreitol, 5 mM N-succinylarginine) and 2 l of reservoir solution (8.5% (w/v) polyethylene glycol 10,000, 0.1 M cacodylate buffer, pH 6.4, 0.2 M calcium acetate). They belong to the space group P2 1 2 1 2 1 with a ϭ 54.9 Å, b ϭ 166.9 Å, c ϭ 186.0 Å and contain four monomers/asymmetric unit. Isomorphous crystals were obtained for the AstB C365S mutant complexed with the substrate. Complete data sets for the wild type enzyme and the C365S mutant co-crystallized with N-succinylarginine were collected to resolutions of 1.95 and 1.7 Å, respectively, at Beamline X25 (National Synchrotron Light Source, Brookhaven National Laboratory on a Quantum-315 CCD area detector (Area Detector Systems Corporation)). When native AstB was co-crystallized with the substrate, we observed only the product in the crystal structure.   All of the crystal forms were soaked in a cryoprotectant solution consisting of mother liquor supplemented with 20% (w/v) glycerol, picked up in a nylon loop, and flash cooled at 100 K in the N 2 (gas) cold stream (Oxford Cryosystems, Oxford, UK). The data sets were integrated and scaled using either HKL2000 (14) or d*trek (15).

AstB Activity Assay
N-Succinylarginine was synthesized as described previously (4). Following the method described by Schneider et al. (4), the N-succinylarginine dihydrolase activity was measured in a coupled assay with glutamate dehydrogenase. Purified AstB (32 g of native or 122 g of AstB C365S mutant) was incubated for 1 h at 30°C with 10 mM N-succinylarginine in 100 mM Tris-HCl, pH 7.5, in a total volume of 1 ml. Following incubation, 200 l of the reaction mixture was added to 800 l of 1 mM ADP (Fluka), 1.7 mM ␣-ketoglutarate (Sigma), and 2.9 mM NADH. Oxidation of NADH to NAD ϩ occurs in a reaction catalyzed by glutamate dehydrogenase.
This reaction was followed by measuring the decrease in absorbance at 340 nm after addition of 4.9 units (100 g) of glutamate dehydrogenase (Sigma). Control incubations without enzyme or substrate were performed in parallel. One unit is defined as the amount of enzyme required to form 1 mol of product/min at 30°C

Structure Solution and Refinement
P1 Crystal Form-This crystal form has six monomers/asymmetric unit. These crystals showed a 2-fold noncrystallographic axis slightly off the crystallographic axis, generating pseudo C2 symmetry that initially confused the structure solution. The R sym in space group C2 (ϳ0.08) was very similar to that for data processed in space group P1 (0.06). Multiwavelength Anomalous Diffraction phasing was initially performed in space group C2 where three independent monomers were expected per asymmetric unit based on the Matthews coefficient of 2.7 Å 3 /Da (16). A total of 26 selenium sites were located from a threewavelength Multiwavelength Anomalous Diffraction experiment calculation using the program SOLVE (17) and were used to calculate an electron density map. Density modification with the program RE-SOLVE (18) resulted in a figure of merit of 0.73. This electron density map was sufficiently clear to build a partial model that was ϳ60% complete. However, the quality of the electron density map varied significantly from one molecule in the asymmetric unit to the other and subsequent refinement using CNS (19) stalled at an R factor of 0.48. To eliminate the possibility of systematic errors in the diffraction data as the source of the difficulty, a second data set was collected on this crystal form using a rotating anode source. The diffraction limit for this data set was similar to that of the synchrotron data set, suggesting that this crystal was of better quality. The self-rotation function calculated using this data set merged in space group P1 suggested the presence of noncrystallographic rather than crystallographic 2-fold symmetry. The partial model of an AstB molecule built previously from the C2 electron density map was used as a search model for molecular replacement in the second, P2 1 crystal form described below. Despite the lower resolution of this crystal form, we were able to extend the model to encompass residues 2-440. This improved model was subsequently used to locate six independent molecules in the asymmetric unit of the P1 crystal form using the program MOLREP (20). From this point the refinement using CNS (19) decreased the R factor rapidly, confirming the choice of the P1 space group. These data are 92% complete to 2.25 Å resolution, with partial data extending to 1.9 Å resolution. The refinement using all available data converged to an R factor of 0.213 and R free of 0.251 (Table  I). The final model includes six independent monomers, each containing residues Asn 2 -Ala 19 and His 31 -Arg 441 . Residues Gly 20 -Arg 30 are disordered and were not modeled. Difference electron density maps showed a strong peak in each molecule with clear octahedral coordination by oxygen atoms from the surface loop Ala 340 -Ser 346 , suggesting a bound metal ion. The ion-oxygen distances were in the range 2.6 -3.1 Å with the majority between 2.7 and 2.9 Å. Based on these distances the site is most likely occupied by a sodium or potassium ion. This density was modeled as a potassium ion and yielded reasonable B factors during subsequent refinement. This ion likely plays a structural role because it is far from the active site region. P2 1 Crystal Form-This crystal form diffracts only to 3 Å resolution and has eight molecules in the asymmetric unit. Initially, molecular replacement using a partial model (see above) located seven monomers in the asymmetric unit. The electron density was sufficient to extend and partially refine the model. Upon completion of the refinement of the model in the P1 crystal form, this refined model was used to locate all eight monomers in the asymmetric unit by molecular replacement (20). This structure was refined using CNS (19) to an R factor of 0.26 and an R free of 0.28. The loop Ala 19 -Arg 32 was disordered in seven of the eight molecules. The eighth molecule showed this loop in a closed conformation. Because the resolution of these data is low and there are no substantial difference between these models and the higher resolution P1 form, these coordinates were not deposited. P2 1 2 1 2 1 Crystal Form-The structures of the AstB-substrate complexes were solved by molecular replacement and refined using the program CNS (19). Each of the four independent molecules contained residues 2-445. The loop Gly 20 -Arg 30 , poorly ordered in other structures, is well ordered in every molecule. A ligand molecule is bound to each monomer of AstB. In addition, ϳ600 solvent molecules were positioned in the electron density. The final R factor is 0.217, and R free is 0.245 for the complex of AstB with the N-succinylornithine reaction product and for the C365S mutant co-crystallized with the substrate the R factor is 0.202 and R free is 0.225 (Table I). Coordinates of wild type AstB in space group P1, AstB-succinylornithine complex, and the C365S mutant of AstB with bound succinylarginine (both in space group P2 1 2 1 2 1 ) have been deposited in the Protein Data Bank (21) with codes 1YNF, 1YNH, and 1YNI, respectively.

RESULTS AND DISCUSSION
Purification, Mutagenesis, and Characterization-Wild type E. coli AstB protein and the C365S mutant were purified to apparent homogeneity as assessed by SDS-PAGE and native PAGE. The protein forms dimers in solution as determined by dynamic light scattering and gel filtration chromatography. Purified wild type AstB was highly active and had a specific activity of 5.3 units/mg as determined by measuring the release of ammonia upon conversion of N-succinylarginine to N-succinylornithine. A crude extract of nitrogen-limited (i.e. fully induced) wild type E. coli contained 0.025 units/mg total protein activity (4). In contrast, the purified C365S mutant had a specific activity of only 0.065 units/mg, indicating a crucial role for this cysteine residue in catalysis.
Overall Structure of the Monomer-The AstB molecule consists of a single globular domain of 447 amino acids with an ␣/␤ topology. The domain forms a propeller composed of five repeats (modules) of a 2␤1␤␣1␤ motif arranged circularly around 5-fold pseudo symmetry axis (Fig. 2). This fold has been observed previously and is called the ␣/␤ five-stranded propeller in the CATH database classification (version 2.5.1) (22,23) and the pentein ␤/␣-propeller in the SCOP database classification (24). The three strands of each module form a mixed ␤-sheet with the first, N-terminal strand of the repeat lying near the central axis of the propeller. The two innermost strands of the module are anti-parallel, with the ␣-helix forming a cross-over connection to the third ␤-strand, which is parallel to the second. The ␣-helix is parallel to this last ␤-strand and is out of the plane of the ␤-sheet on the outside of the propeller (Fig. 2). The connections between the modules vary in length and contain either a short 3 10 -helix or ␣-helix. The first module, which begins with the N terminus of the protein, differs somewhat from the other modules. It starts at the middle ␤-strand, followed by an ␣-helix and a third ␤-strand, whereas the innermost ␤-strand comes from a Cterminal segment, which follows the fifth module. This organization results in the first (Ala 2 ) and the last (Thr 373 ) residue of the propeller being adjacent to one another. The last ϳ75 residues (Glu 374 -Arg 441 ) fold into a two-helix hairpin stacked against the edges of the first and second modules of the propeller. The loops connecting the secondary structure elements within each module on one end of the sheet are short, whereas the loops on the opposite end are much longer. The connections between the modules are on the latter side and are also comparatively long. Thus one side of the propeller is relatively flat, whereas the opposite side has a more complex topography.
The independent molecules within the asymmetric unit of each crystal form and across the different crystal forms are very similar overall. The root mean square deviation between the C␣ atoms of molecules across different crystal forms is 0.2-0.4 Å. The only significant difference between the molecules is the conformation of a loop, Ala 19 -Gln 34 , which is largely disordered in the apo protein but becomes well ordered in the presence of bound substrate or product.
Dimer Formation-Gel filtration and dynamic light scattering studies both indicate that AstB forms homodimers in solution. This is consistent with the presence of homodimers in each of the three crystal forms (Fig. 2c). Each dimer has approximate dimensions of 92 ϫ 62 ϫ 54 Å, with the two monomers being related by a noncrystallographic 2-fold symmetry axis. The dimer interface is formed by a bundle of three short helices. The residues contributing to dimerization are from the second module (residues Asn 133 -Ser 139 and Ala 164 -Leu 171 ), the helix of the third module (residues Glu 216 -Leu 224 ), and the long loop that connects the first and second modules (residues Arg 74 -Phe 78 and Trp 96 ). These residues include several isoleucines, phenylalanines, alanines, and a proline, giving this surface a partially hydrophobic character. In addition to numerous van der Waals' interactions, the dimer is further stabilized by hydrogen bonds, some of which are bridged by water molecules. The surface area buried upon dimer formation calculated using the method of Lee and Richards (25) with a 1.4-Å probe radius is 900 Å 2 , which corresponds to 6% of the total surface area of each monomer. The two substrate-binding sites within the dimer are positioned on the same side of the elongated dimer, but each involves residues from only one monomer, suggesting that the active sites within each monomer of AstB function independently.
Substrate-binding Site-Attempts to capture the substrate N-succinylarginine by soaking it into the P1 and P2 1 crystal forms were unsuccessful. Because this could have been a result of enzymatic hydrolysis of the substrate in the crystal, a C365S mutant, in which a serine replaced the cysteine nucleophile, was constructed and expressed to eliminate the catalytic activity of the enzyme. Nevertheless, using the mutant AstB we were still unable to detect the enzyme-bound substrate either by soaking the crystals or by co-crystallization under conditions producing the P1 or P2 1 crystal forms. Further screening for suitable crystallization conditions of this inactive AstB mutant in the presence of N-succinylarginine yielded new crystallization conditions that resulted in an orthorhombic crystal form belonging to the space group P2 1 2 1 2 1 . An electron density map calculated from diffraction data collected from this crystal revealed a well defined substrate molecule bound in the active site (Fig. 3a). As a result of substrate binding, all of the residues in the Ala 19 -Arg 32 loop, which were disordered in the native crystal structures, were now clearly defined in electron density.
The substrate binds to the C365S mutant enzyme on the propeller face that contains the long cross-over loops between the modules (see above). The binding site is shaped as a ϳ15-Å deep tunnel that leads from the surface toward the protein center and rests on the residues from the ends of the innermost strands of the five ␤-sheets (Fig. 2, a and b). The sides of the tunnel are made of residues from the various cross-over loops, Ser 102 -Trp 107 , His 137 -Arg 138 , and Asn 359 -Gly 361 , and the bottom is lined with Ala 109 , Asn 110 , Ala 177 , Val 251 , and Asn 306 . The entrance to the substrate-binding tunnel is shielded by the mobile Ala 19 -Arg 32 loop. The substrate is oriented with its guanidinium group at the bottom of the tunnel and the succinate carboxylate closest to the surface at the entrance to the tunnel (Figs. 2 and 3a). All of the nitrogen and oxygen atoms of the substrate are involved in direct hydrogen bonds to the FIG. 2. Ribbon representation of AstB. a, stereo view of AstB approximately along the 5-fold pseudo symmetry axis from the side opposite to the bound substrate (shown in van der Waals' representation and colored blue). Each module is colored in succession: red, cyan, magenta, green, and blue. The C-terminal ␣-helical hairpin extension is colored yellow. The positions of the N and C termini are marked. b, C␣ backbone of AstB with the residues colored by conservation level. Magenta, highly conserved; aquamarine, semi-conserved; gray, others. The ligand is shown in van der Waals' representation. c, ribbon drawing of the AstB dimer viewed along the pseudo 2-fold axis. This and subsequent figures were generated using the program PyMol (www.pymol. org/). enzyme (Figs. 3a and 4a). Hydrogen bonds between the succinyl carboxyl group and ordered water molecules provide additional bridging interactions to the protein. The carboxylate group of the arginine moiety, in the middle of the extended substrate molecule, forms two salt bridges, to Arg 212 (two Hbonds) and to Arg 138 (one H-bond), and also forms a hydrogen bond to Asn 25 . The guanidinium group of the substrate is oriented through a salt bridge between its NH1 and NH2 atoms to Asp 250 and is hydrogen-bonded through NE to OD1 of Asn 110 . Further, the amide NH (former N terminus of arginine) is hydrogen-bonded to the carbonyl oxygen of Asn 359 , and the neighboring carbonyl oxygen of the substrate forms a hydrogen bond to the side chain of His 137 . Finally, the succinyl carboxyl group interacts with residues from the mobile loop: its OD1 atom is hydrogen-bonded to the NH group of Leu 21 and the OH of Ser 28 , whereas the OD2 is hydrogen-bonded to the NH group of Ala 19 and a bridging water molecule. Of importance is also the interaction of the Trp 107 side chain stacked against the alkyl chain of the arginine moiety of the substrate. In the substrate-bound state the mobile loop forms a lid over the substrate, completely burying it within the protein, and contributes several hydrogen bonds to the substrate. In the C365S FIG. 3. The final 3F o ؊ 2F c a -weighted electron density map contoured at 1 . a, around N-succinylornithine product in the substrate-binding site of AstB. b, N-succinylarginine substrate complexed with the AstB C365S mutant. The ligand and the surrounding residues are drawn in a ball-and-stick representation. Nitrogen atoms are shown in blue, and oxygen atoms are in red. The hydrogen bonds between the ligand and protein atoms are marked by green dashed lines. mutant, in which the serine replaces the cysteine nucleophile, the Ser 365 side chain is directed away from the substrate, and the Ser 365 hydroxyl makes two hydrogen bonds to Gly 362 . Although the active site His 248 is 3.3 Å from the guanidinium moiety of the substrate, the imidazole ring is nearly perpendicular to the plane of the guanidinium group, indicating that this histidine forms no hydrogen bond to the bound substrate (Fig. 4a).
With the exception of the contacts involving the missing guanidinium moiety, all of the previously described enzymesubstrate interactions are also observed in the wild type enzyme-product complex (Fig. 3b). Here, the side chain of the Cys 365 nucleophile is directed toward the substrate as expected from its catalytic role.
Sequence and Structural Similarity-Sequence analysis using PSI-BLAST (26) identified AstB homologs in 23 bacterial species. No homologs in other kingdoms were found. The sequence identity between E. coli AstB and the other bacterial orthologs varies from 85% sequence identity for Salmonella typhimurium to 42% for Zymomonas mobilis, indicating a high degree of sequence conservation for AstB among these species.
Mapping the positions of conserved residues identified in the sequence alignments onto the three-dimensional structure of AstB shows that the strictly conserved residues cluster predominantly around the substrate-binding site. Almost all of the residues that are within a distance of 8 Å from the bound substrate are fully conserved in these sequences, and the remaining residues are highly conserved. Indeed, this sphere of high conservation extends to ϳ12 Å from the substrate (Fig. 2b).
Although sequence comparison using BLAST (26) with sequences of proteins of known three-dimensional structure contained within the Protein Data Bank showed no homologs, a search for structurally similar proteins using the program DALI (27) identified statistically significant matches for experimentally determined structures of N-dimethylarginine dimethylaminohydrolase (Protein Data Bank code 1H70 (28)), arginine glycine amidinotransferase (Protein Data Bank code 1JDW (29)), two arginine deiminases (Protein Data Bank codes 1LXY (30) and 1RXX (31)) and two ribosome anti-association factors eIF6 (Protein Data Bank codes 1G61 and 1G62 (32)). The first four proteins are enzymes belonging to the amidino- transferase superfamily (AT, as classified within SCOP database (24)) with the structural similarity extending nearly throughout the entire protein. The substrates in the reaction catalyzed by arginine deiminases (Scheme 2) are similar to that of AstB, and both enzymes utilize water molecule(s) to release ammonia (33). The eIF6 factors that are structurally similar to AstB are much smaller, with only ϳ230 residues, and contain five repeats of the basic ␤␤␣␤ motif forming a minimal pentein propeller structure.
The AT superfamily of enzymes (PF02274, PFAM data base (34)) presently contains over 130 members from various bacterial species. The enzymes with known activities include glycine amidinotransferases (EC 2.1.4.1) involved in creatine biosynthesis, inosamine amidinotransferases (EC 2.1.4.2) involved in streptomycin biosynthesis, and arginine deiminases (EC 3.5.3.6) that convert arginine to citrulline. All of these enzymes catalyze amidine group transfer or hydrolysis with the first step of the mechanism involving nucleophilic attack by a cysteine residue on the substrate. In addition to the cysteine, catalytic residues also include a histidine and an aspartate or a glutamate. Structure-based alignment of AstB with four other enzymes from the AT superfamily with known structures showed less than 10% sequence identity. Nevertheless, based on structural similarity, conservation of the catalytic residues, and the common type of reaction, it is now clear that the family of enzymes sharing sequence similarity with E. coli AstB also belongs to the AT superfamily.
Using the software FUGUE (35) Shirai and Mizuguchi (9) recently constructed a model for the structure of E. coli AstB and suggested that this protein is a member of the AT superfamily. They correctly predicted that the fold of AstB is a ␤␤␣␤ propeller with five modules. However, their more detailed predictions were only partially correct. Indeed, although their assignment of Cys 365 and His 248 as part of the active site agrees with our structure, the third catalytic residue is Glu 174 and not Asp 173 as they predicted. Similarly, their prediction that Asp 119 and Asp 122 form hydrogen bonds to the guanidinium group of the substrate is incorrect; these two residues are part of the surface loop and are more than 20 Å away from the substrate.
Catalytic Mechanism-The structure of the C365S mutant complexed with N-succinylarginine and the wild type enzyme complexed with the N-succinylornithine product identified the location of the substrate-binding site and the disposition of the substrate and the product relative to the catalytic residues. These side chains include Cys 365 , His 248 , and Glu 174 . The comparison of AstB with other members of the AT superfamily (see above) shows a similar disposition of their catalytic residues with respect to the guanidinium moiety of the substrate (Fig.  4). In agreement with its predicted role as a nucleophile attacking the carbon of the guanidinium moiety, mutation of Cys 365 to serine severely compromised the activity of the enzyme. These observations indicate that AstB uses a catalytic mechanism similar to those of amidinotransferases and deiminases (30,31,36).
The reactions catalyzed by ADI and succinylarginine dihy-drolase differ in that ADI removes one NH 2 from the guanidinium moiety of the arginyl chain and replaces it by a carbonyl oxygen atom derived from a water molecule, whereas AstB carries the reaction further by removing the second NH 3 group and releasing CO 2 , leaving an ornithine side chain as the product. Detailed catalytic mechanisms have been proposed for the arginine deaminases (30,31). Surprisingly, comparison of the side chains in the vicinity of the guanidinium moiety of ADI (Protein Data Bank code 1LXY (30)) and AstB shows nearly identical environments (Fig. 4), raising a question as to why the succinylarginine dihydrolase does not stop at converting N-succinylarginine to N-succinylcitrulline but carries the reaction further through a second hydrolysis event.
The key catalytic residues Cys-His-Glu are conserved, as are the Asp 250 (Asp 271 in ADI) and Arg 212 (Arg 232 in ADI) that form salt bridges to the guanidinium and carboxylate groups of the substrate, respectively. The only difference in the vicinity of the guanidinium group is the side chain of residue 110, which in AstB is an asparagine, whereas in ADI it is an aspartate (Asp 161 in 1LXY) (Fig. 4). This side chain in AstB forms one hydrogen bond to the NE atom of the guanidinium moiety, whereas the equivalent residue, Asp 161 , in ADI forms two hydrogen bonds to NE and NH2. A review of available structures, reinforced by sequence alignment within the AT, ADI, and AstB families, shows that in other enzymes that substitute a carbonyl oxygen for NH 2 this side chain is always an aspartate, whereas in the AstB dihydrolase family it is always an asparagine. We speculate therefore that the residue that is hydrogen-bonded to the NE atom of the arginyl moiety of the substrate determines the outcome of the reaction. When this residue is an aspartate, as in ADI (Asp 161 ), this side chain forms two hydrogen bonds to the guanidinium moiety in the substrate and to the corresponding citrulline atoms in the product. A detailed reaction mechanism for ADI enzymes has been previously proposed (30,31). We propose that AstB employs a mechanism similar to that of ADI but with two ADI-like hydrolytic reaction cycles to replace NH1 and NH2 by oxygens. In AstB the side chain of Asn 110 is hydrogen-bonded to the arginyl NE (but not NH2). The first reaction cycle would convert N-succinylarginine to N-succinylcitrulline, replacing the NH1 atom with oxygen. This carbonyl oxygen would be in close proximity to the negatively charged acidic group of Asp 250 . A 180 o rotation around the N⌭-C⌮ bond would relieve the unfavorable CϭO . . . Asp 250 contact, would bring this carbonyl oxygen atom into hydrogen bonding distance of the ND2 atom of Asn 110 , and would place the NH2 atom into the position previously occupied by NH1, thereby preparing the stage for the second hydrolysis cycle. The presence of an aspartate rather than asparagine at this position in ADI enzymes would prevent such a rotation and lead to a release of the citrulline product.
The previously proposed mechanism for ADI suggests that NH1 of the substrate forms a hydrogen bond to the histidine (His 269 in ADI). The structure of the C365S mutant complexed with the N-succinylarginine substrate indicates that His 248 does not form a hydrogen bond to the guanidinium group of the bound substrate because it is nearly perpendicular to the plane of the guanidinium. Rather, such a hydrogen bond would form only after the transfer of a proton to the arginyl CZ, when this atom attains sp 3 hybridization. The resulting tetrahedral coordination would direct this NH1 toward the NE of His 248 , forming a transient hydrogen bond that would aid in the release of ammonia, NH 3 . Such a tetrahedral intermediate state has been observed in the structure of arginine deiminase (30) and shows the plausibility of a hydrogen bond between NH1 and the histidine (Fig. 4b). Based on the proposed model for the reaction mechanism, we expect that succinylcitrulline would also be a good substrate for AstB. According to the proposed model the replacement of Asn 110 by an aspartate should convert the dihydrolase activity of the wild type enzyme to a deiminase activity in the mutant, leading to the formation of N-succinylcitrulline rather than N-succinylornithine. Similarly, the mutation of the corresponding Asp in arginine deiminase to an asparagine should convert the enzyme into an arginine dihydrolase. We are now testing these predictions experimentally.
Conclusions-The crystal structure revealed that AstB has the ␣/␤ propeller fold and belongs to the AT protein superfamily. The catalytic center is comprised of residues Cys 365 , His 248 , and Glu 174 positioned near the bottom of a long cavity extending from one side of the protein near the propeller axis. Although the triad is superficially reminiscent of the catalytic triad of cysteine proteases, the disposition of the cysteine and histidine side chains in these two classes of enzymes is quite different. In cysteine proteases Cys . . . His are within a hydrogen-bonding distance, and the role of the histidine is to deprotonate the nucleophilic cysteine directly. In the AT superfamily cysteine and histidine are separated by more than 5 Å and are positioned on opposite sides of the substrate with the histidine acting on the substrate or the transition state.
The substrate and product bind in a polar cleft of AstB shielded from the solvent by a 13-residue-long (Ala 19 -Arg 32 ) loop unique to this family. This loop is disordered in the apo form of the enzyme, suggesting that AstB exists in an open conformation in the absence of a bound ligand. Comparison of the substrate-free and substrate-bound structures shows that the flap closes over the entrance to the substrate-binding tunnel and buries the N-succinylarginine. Subsequently, the flap must open to allow N-succinylornithine to depart. Further studies of the enzymatic mechanism and conformational mobility of AstB, as well as other enzymes of the AST pathway, may lead to the design of small molecule therapeutics that inhibit these enzymes.