Three-dimensional Structure of YaaE from Bacillus subtilis, a Glutaminase Implicated in Pyridoxal-5′-phosphate Biosynthesis*

The structure of YaaE from Bacillus subtilis was determined at 2.5-Å resolution. YaaE is a member of the triad glutamine aminotransferase family and functions in a recently identified alternate pathway for the biosynthesis of vitamin B6. Proposed active residues include conserved Cys-79, His-170, and Glu-172. YaaE shows similarity to HisH, a glutaminase involved in histidine biosynthesis. YaaD associates with YaaE. A homology model of this protein was constructed. YaaD is predicted to be a (β/α)8 barrel on the basis of sequence comparisons. The predicted active site includes highly conserved residues 211–216 and 233–235. Finally, a homology model of a putative YaaD-YaaE complex was prepared using the structure of HisH-F as a model. This model predicts that the ammonia molecule generated by YaaE is channeled through the center of the YaaD barrel to the putative YaaD active site.

Pyridoxal-5Ј-phosphate (PLP), 1 one of the B 6 vitamers, is an essential cofactor in all living systems. Its mechanistic function is the stabilization of carbanions adjacent to amino groups. This cofactor therefore plays a key role in amino acid metabolism; PLP-containing enzymes catalyze racemization, decarboxylation, transamination, and side-chain substitution reactions of amino acids. PLP is biosynthesized in bacteria, fungi, and plants, but it is an essential nutrient in animals (1).
It was clear from studies involving the incorporation of isotopically labeled PLP precursors that the PLP biosynthetic pathway in eukaryotes such as Saccharomyces cerevisiae is different from the well studied Escherichia coli pathway (Fig.  1). In E. coli, PLP is biosynthesized from D-erythrose-4-phosphate, deoxy-D-xylulose-5 phosphate, and glutamate (2), whereas in yeast, it is biosynthesized from an unidentified pentulose or pentose, glyceraldehyde, and glutamine (3)(4)(5). All of the genes required for PLP biosynthesis in E. coli have now been identified, the biosynthesis has been fully reconstituted in vitro, and the structures of three of the biosynthetic enzymes have been determined (6). In contrast, PLP biosynthesis in S. cerevisiae has not yet been reconstituted in a cell-free sys-tem, and the genes involved in the pyridine ring formation have only recently been identified (7)(8)(9)(10)(11)(12)(13)(14). These are PDX1 and PDX2 in S. cerevisiae. Extensive phylogenetic analysis revealed that PLP biosynthesis, by what was considered previously to be the eukaryotic pathway, is widespread, and this pathway has now been found in prokaryotes, eukaryotes, and in the archaebacteria (15).
YaaD and YaaE have been identified as the PLP biosynthetic enzymes in Bacillus subtilis (16). YaaD revealed no homology with any known enzyme class. However, YaaE shows high sequence similarity to glutamine aminotransferase, suggesting that this protein is involved in the hydrolysis of the amide of glutamine, releasing ammonia for incorporation into the pyridine ring in a reaction likely catalyzed by YaaD.
We report here the crystal structure of the B. subtilis YaaE at 2.5-Å resolution. A comparison of this structure with that of other glutaminases revealed a high structural similarity to the histidine biosynthetic protein HisH from Thermotoga maratima (17). We also report a homology model of YaaD and propose a model for the YaaD-YaaE complex.

EXPERIMENTAL PROCEDURES
Molecular Cloning-Standard methods were used for DNA restriction endonuclease digestion, ligation, and transformation of DNA (18). Plasmid DNA was purified with the Wizard™ Plus SV DNA miniprep kit (Promega). DNA fragments were separated by agarose gel electrophoresis, and then excised and purified with the QiaQuick gel extraction kit (Qiagen). E. coli strain DH5␣ was used as a recipient for transformations during plasmid construction and for plasmid propagation and storage. A PerkinElmer Life Sciences GeneAmp PCR System 2400 and Platinum Pfx DNA polymerase (Invitrogen) were used for PCR. The plasmid pET-28a was obtained from Novagen. Sequencing was performed at the Cornell BioResources Center. Genomic DNA from B. subtilis CU1065 was used as a template for the PCR amplification of yaaE using 5Ј-AGG AGC GCT GCT CAT ATG TTA ACA ATA GGT GTA CTA GG-3Ј (inserts an NdeI site (underlined) at the start codon) and 5Ј-CTA TCA ACG CTT CTC GAG CTT TAT TTG TGC TTA TAA TG-3Ј (inserts an XhoI site (underlined) after the stop codon) as the primer pair. The PCR product was digested with NdeI and XhoI and ligated into similarly digested pET-28a to give pCLK1501. All PCR-derived DNA was sequenced to ensure that no mutations had been introduced in the cloning process.
Expression and Purification of YaaE-For protein purification, the expression construct was transformed into E. coli B834(DE3) cells (Novagen), which are auxotrophic for methionine. One liter of Luria-Bertani medium containing 25 g/ml kanamycin was inoculated with 5 ml of saturated starter culture and incubated at 37°C. When the culture reached an A 595 of ϳ0.6, the cells were induced with 1 mM isopropyl-␤-D-thiogalactopyranoside for 6 h at 25°C. Cells were harvested by centrifugation and resuspended in 25 ml of cold extraction buffer (10 mM imidazole, 50 mM Tris, 500 mM NaCl, pH 8.0). All subsequent protein purification steps were carried out at 4°C. Cells were lysed by two passes through a French pressure cell at 15,000 p.s.i., and the insoluble cell debris was removed by high speed centrifugation. The clarified cell extract was mixed with a 3-ml slurry of nickel-nitrilotriacetic acid resin (Qiagen), which was pre-equilibrated in extraction buffer, and gently stirred for 1 h. The resin was centrifuged for 15 min at low speed, and the supernatant was decanted. The resin was resuspended in the ex-traction buffer and poured onto a nickel column. The column was thoroughly washed with the extraction buffer. The His-tagged protein was then eluted with 130 mM imidazole, 50 mM Tris, 500 mM NaCl, pH 8.0. The eluted enzyme was dialyzed against 10 mM Tris, pH 8.0, and was concentrated to 5 mg/ml.
Production of selenomethionine-labeled YaaE (SeMet-YaaE) follows the same protocol as above with the following modifications. The cells were grown in 1 liter of cultures containing M9 minimal medium supplemented with 40 g/ml L-amino acids (excluding methionine), 1% BME vitamin solution (Invitrogen), 0.4% (w/v) glucose, 2 mM MgSO 4 , 25 g/ml FeSO 4 ⅐7H 2 O, 0.1 mM CaCl 2 , 25 g/liter kanamycin, and 50 g/ liter L-selenomethionine. This medium was inoculated with a 1:20 dilution of cells from a 50-ml starter culture containing the above medium but with L-methionine in place of L-selenomethionine to encourage growth. Prior to inoculation, the starter cells were pelleted and washed with the induction culture to remove the L-methionine. Purification of the labeled enzyme then proceeded as for the native protein.
Crystallization of YaaE-The optimized conditions for both the native and the selenomethionine-labeled proteins were found to be 10 -12% (w/v) polyethylene glycol-8000, 100 mM MOPS, pH 7.0 -7.2, 6% (v/v) ethylene glycol, and a protein concentration of 2.5 mg/ml. Crystals of SeMet-YaaE also grew in the presence of 2 mM dithiothreitol. The crystals were grown at 18°C using the hanging-drop vapor diffusion technique. Drops (8 l) containing a 1:1 mixture of protein and reservoir solutions were optimal for crystal growth. Crystals of diffraction quality grew over a period of 1-2 weeks. YaaE crystallizes in the orthorhombic space group P2 1 2 1 2 1 with unit cell dimensions of a ϭ 45.50 Å, b ϭ 80.87 Å, and c ϭ 115.39 Å. Each asymmetric unit contains two monomers corresponding to a calculated solvent content of 50%.
Data Collection and Processing-Initially, a single-wavelength anomalous diffraction data set was collected on a single frozen SeMet-YaaE crystal at the Cornell High Energy Synchrotron Source (CHESS) beamline F2. A cryoprotectant solution of 18% ethylene glycol in the mother liquor was used to prevent damage during freezing. An x-ray absorption spectrum in the vicinity of the selenium absorption edge was determined for the SeMet-YaaE crystal by recording x-ray fluorescence as a function of wavelength. Diffraction data were then collected to 2.85 Å at the wavelength corresponding to the peak of this spectrum (0.979 Å). The data were measured in 0.5°oscillation steps with 2.5-min exposure times using a Quantum-4 CCD detector (San Diego Area Detector Systems Corp.) with a crystal-to-detector distance of 180 mm. A total of 90°of data was collected. These data were processed with DENZO and scaled using SCALEPACK (19). A three-wavelength MAD data set was collected on a single frozen SeMet-YaaE crystal on the 8-BM beamline at the Advanced Photon Source (APS). An x-ray fluorescence spectrum was again determined, and diffraction data were collected to 2.5-Å resolution at wavelengths corresponding to the inflection point of the spectrum (0.9796 Å), the peak of the spectrum (0.9794 Å), and a high energy remote (0.9642 Å). The data were measured in 1°o scillation steps with 15-s exposure times using a Quantum-315 CCD detector (San Diego Area Detector Systems Corp.) with a crystal-todetector distance of 350 mm. A total of 100°of data was collected for each wavelength. The data were processed with MOSFLM (20) and scaled using SCALA (21). Final data processing statistics are shown in Table I.
Structure Determination-The selenium atom positions were determined from the CHESS data using the Patterson search routine of CNS 1.0 (22). The MAD phasing component of CNS was used to calculate the initial phases and density modification (22). Symmetry averaging was attempted using a matrix generated by PROFESSS (23); the resulting electron density map was, unfortunately, uninterpretable. An interpretable map was calculated from the three-wavelength MAD data set collected at APS and from the selenium sites from the single wavelength CHESS data set using CNS. Electron density visualization and model building were done with O (24). Simulated annealing torsion angle refinement was followed by group B-factor refinement and energy minimization and performed using CNS. Several rounds of refinement were combined with model rebuilding in O after inspection of both 2F o Ϫ F c and F o Ϫ F c maps. The refinement was carried out using the peak wavelength of the MAD data set. In verifying the final model, a simulated annealing composite omit map was calculated, and the B-factors were refined individually. Water molecules were initially picked using CNS and then manually verified in O. Refinement statistics can be found in Table II, and a sample of the electron density is shown in Fig. 2. YaaD-YaaE Complex Modeling-Templates for a homology model of B. subtilis YaaD were identified using 3D-PSSM (27) and a sequence search of the Protein Data Bank (28). The initial list of nine candidate ␤-barrel structures was reduced to five (PDB codes 1GPW, 1JVN, 1DVJ, 1PII, and 2TPS) using the following biases: (a) favor structural diversity as indicated by root mean square deviation from other structures, (b) favor higher resolution structures, and (c) favor structures that bind ligands most similar to PLP. The highest sequence identity with YaaD was 20% for PDB code 1GPW. The ␤-barrel domains of the templates were structurally superimposed on PDB code 2TPS using CE (29), and a ClustalX (30) alignment with YaaD was manually adjusted based on the structural superposition. Predicted loop regions in YaaD were aligned only with loops of similar size in the template structures. Ten homology models were built using Modeler version 6 (31, 32) with two cycles of very slow MD (molecular dynamics) annealing (MD_LEVEL ϭ refine_4, LIBRARY_SCHEDULE ϭ 2). A DALI (33) search with the lowest energy model revealed no additional related structures.
The three lowest energy YaaD models and the YaaE crystal structure were superimposed on the HisH-F structure from yeast (PDB code 1JVN) and the HisH-HisF complex from T. maritima (PDB code 1GPW) using ProFit (34). Overlaps were resolved through a four-stage minimization using the OPLS-AA (35) force field and GB/SA solvation model (36) as implemented in version 7.2 of the program MacroModel (37). First, gross overlaps were corrected. Some residues were manually adjusted to remove knots, and the YaaE loop at residues 110 -113, YaaE side chains within 2.5 Å of YaaD, and YaaD residues within 6 Å of YaaE were energy-minimized with all other atoms frozen. Second, hydrogen atoms were energy-minimized with all other atoms frozen. Third, the initial minimization was repeated but with all YaaD atoms permitted to move subject to a 100 kJ/mol-Å 2 restraint. Due to software limitations, bond angle and torsion parameters were not applied to the restrained atoms in the initial minimizations, making it necessary to relieve some strained geometries through a final minimization. This fourth cycle included all force field parameters for restrained atoms, decreased the YaaD restraint to 10 kJ/mol-Å 2 , and allowed the previously frozen YaaE atoms to move subject to a 300 kJ/mol-Å 2 restraint.

RESULTS
Crystal Packing Analysis-The asymmetric unit contains two monomers (referred to hereafter as A and B). Superimposing the monomers reveals that residues 9 -14, 45-60, 87-96, and 106 -115 are in noticeably different conformations (Fig.  3a). These differences appear to be completely explainable in terms of crystal packing. An adjacent monomer packs against monomer A in such a way that the loop containing residues 9 -14 is forced back against the helix containing residues 45-60, forcing this helix into a somewhat distorted position; no  such packing constraint exists for monomer B, in which the loop containing residues 9 -14 is permitted to move freely. Residues 9 -14 were not visible in the initial MAD phasing map of monomer B and appeared only weakly in the later composite omit maps; residues 45-60 were visible in the initial map but were too disordered to be built until the later stages of refine-ment. In both monomer A and monomer B, the loop containing residues 87-96 is unconstrained and is somewhat disordered; even at the later stages of refinement, it was possible to build the main chain atoms only approximately. The majority of the residues in the generously allowed region of the Ramachandran plot are from these loops. These poorly ordered regions of FIG. 2. A stereodiagram showing the electron density surrounding a portion of the YaaE active site. The density is from a composite omit map calculated at 2.5-Å resolution using the final model. The map is contoured at 1.5 times the root mean square value of the map. The strand containing residues Ser-168 -Glu-172 lies to the right, and the turn containing Cys-79 and Ala-80 is shown at center, whereas Gly-46 and Gly-47 are at the upper left. Note the broken density for Gly-46 and Gly-47, which may form part of the oxyanion hole in the YaaD-YaaE complex.
FIG. 3. Structure of YaaE. a, stereo diagram showing the differences in conformation between monomer A (blue) and monomer B (green). Note that the ordering of residues 9 -14 in monomer A causes the helix containing residues 45-60 to adopt a noticeably different position. The loop containing residues 106 -115 in A makes a crystal contact with the C-terminal helix of an adjacent monomer and is not affected by the repositioning of helix 45-60. The loop containing residues 87-96 is somewhat disordered, the different conformations arising from flexibility rather than crystal contacts. b, a topology diagram showing the connections between the strands of the central mixed ␤-sheet and the flanking ␣-helices. The beginning and ending sequence numbers are shown for each secondary structural element. The orientation is the same as that for the monomer shown in panel a. c, a view of YaaE looking down the C-terminal end of the ␤-sheet. YaaE embodies the features of the triad aminotransferase fold (42): a twisted seven-stranded parallel ␤-sheet flanked by six ␣-helices. The active site cysteine is located on a tight turn between ␤4 and ␣5, which causes its phi, psi values to fall into the generously allowed region of the Ramachandran plot. the structure contain nearly 15% of the residues and are probably responsible for the slightly high crystallographic R-factor reported in Table II. The remaining residues in the generously allowed region are Cys-79 from both monomers; as detailed below, these residues serve as active site nucleophiles. Finally, the loop containing residues 106 -115 is pushed into a slightly different position in monomer A as a result of the C-terminal helix of an adjacent monomer of A packing against this loop; this also had the effect of making the C-terminal residues of monomer A somewhat more ordered than those of monomer B.
Overall Structure-YaaE is an ␣/␤ three-layer sandwich containing a seven-stranded twisted mixed parallel ␤-sheet flanked by six ␣-helices on the N-terminal stretch of the sheet, four on one side and two on the other. The first five strands in the sheet, ␤1-␤4, ␤12, and the last strand, ␤10, share the same orientation, whereas ␤11, the next-to-last strand in the sheet, lies antiparallel to the rest. This type of mixed ␤-sheet appears to be common to the triad aminotransferases. A pair of antiparallel ␤-strands is located along the exposed edge of the central ␤-sheet. A mixed, three-stranded region of short ␤-strands flanks the other side of the central sheet. The ␣-helices lie close to the N-terminal edge of the central sheet, whereas the flanking ␤-strands lie at the C-terminal edge of the central sheet. These connections are illustrated in the topology diagram shown in Fig. 3b.
The overall structure of YaaE is shown in Fig. 3c. The molecule folds as a single domain with approximate dimensions of 27 Å ϫ 22 Å ϫ 50 Å. YaaE adopts the form of a rough cylinder with a cleft between ␤4 and ␤12 and below a loop near the center of the upper face. At the bottom of this cleft lies a tight turn between ␤4 and ␣5, oriented in such a way that the side chain of Cys-79 protrudes into the center of the cavity. The side chain of His-170 lies 3.4 Å from the sulfur atom of Cys-79 and 2.7 Å from the carboxyl side chain of Glu-172 in an orientation suggesting a Glu-His-Cys catalytic triad. This, along with sequence and structural alignments of YaaE with known triad aminotransferases (38,39), led us to designate this cleft as the active site.
Active Site-The active site crevice lies below a two-strand stretch of antiparallel ␤-sheet. Cys-79 has (, ) values of (51.5, Ϫ132.2); average values for active site cysteines in triad aminotransferases are (50, Ϫ110). In keeping with the active sites of other aminotransferases, His-170 and Glu-172 are separated by Pro-171. Continuing with the analogy with other triad aminotransferases, we expected an oxyanion hole to be formed by the amide NH of Ala-80 and Gly-47; however, no water molecules were found bound in this location. Comparison with the structurally similar HisH from T. maritima leads us to suspect that in the absence of YaaD, the oxyanion hole is not well formed (17). By superimposing the active sites for the two monomers, it may be observed that three bound water molecules in the active site superimpose within 1 Å of their corresponding partner; two of these pairs are located in positions indicative of catalytic significance. One pair is located within hydrogen bonding distance of the histidine ring of His-170, the second is within hydrogen bonding distance of two backbone carbonyls (168 and 134) and one backbone amide (134 N) and appears to be positioned in such a way as to serve the function of nucleophile. When the active site residues of YaaE are superimposed with the active site residues of HisH from T. maritima (17) and S. cerevisiae (40), anthranilate synthase TrpG  1JVN) with inhibitor acivicin has been superimposed to illustrate possible functions of key residues. Active site residues of yeast HisH are shown in cyan; Gln-397 from the HisF subunit is indicated in magenta. By analogy, Glu-172, His-170, and Cys-79 of YaaE form the catalytic triad, whereas Gly-47 participates in the formation of the oxyanion hole. The active site water molecule was found in all other triad aminotransferases examined, within 1 Å of the position illustrated here. b, the structure of acivicin. from Serratia marcescens (41), and the aminotransferase domain of carbamoyl phosphate synthase from E. coli (39), analogues for this water molecule superimpose within 1 Å of each other. Furthermore, in carbamoyl phosphate synthetase and anthranilate synthase, this water molecule also lies within hydrogen bonding distance of the imidazole side chain of the catalytic histidine.

DISCUSSION
A comparison of the structure of YaaE with those of other known protein structures using the DALI server returned the highest scores (Table III) for free HisH from T. maritima (17) and the HisH subunit from the structure of HisH-F complex from S. cerevisae. HisH and the remaining entries in Table III are all triad aminotransferases or triad glutaminase domains of multifunctional proteins. A least squares superimposition of YaaE with each of these other proteins reveals that they all possess the same ␣/␤ structure with loops of various lengths and conformations; this motif has been designated as the triad aminotransferase fold (42). Salient characteristics of this fold include a seven-stranded parallel ␤-sheet flanked by ␣-helices with the active site located at the C-terminal end of the sheet and below an overhanging loop. The active site cystine residue is generally near the 80th residue in the domain and lies on a nucleophilic elbow: a tight turn between a ␤-strand and an ␣-helix at the bottom of the active site. A least squares superimposition of the active sites using the catalytic triad and oxyanion hole residues reveals that the active site of YaaE is largely similar to that of the other aminotransferases; however, there are a few minor, but notable, differences. YaaE lacks a FIG. 5. The reaction catalyzed by the HisHF complex. Notice that this reaction involves a nucleophilic attack by NH 3 followed by a ring closure. YaaD may catalyze related chemistry. PRFAR, N 1 -(5Ј-phosphoribulosylformimino)-5-aminoimidazole-4-carboxamide ribonucleotide; ImGP, imidazole glycerol phosphate; AICAR, 5Ј-(5-aminoimidazole-4-carboxamide)ribonucleotide.
FIG. 6. The complex model created using the crystal structure of YaaE and the lowest energy homology model of YaaD. In this model, YaaE binds to YaaD in much the same orientation as HisH binds to HisF. It can also bee seen that ␤-strands 7 and 8 appear to have some role in the association of YaaE and YaaD. YaaE binds to YaaD in much the same orientation as HisH binds to HisF. As with YaaE, HisH utilizes two ␤-strands in creating the protein-protein interface.
substrate-interacting glutamine present in the other aminotransferases. Additionally, a conserved histidine, which does not interact directly with the substrate, is absent in YaaE. In its place is a proline that serves to kink a loop into a somewhat different conformation in YaaE from the other glutaminases. Finally, YaaE has three arginine residues located above the active site, which the other glutaminases appear to lack. It is presently not possible to speculate on the functional significance, if any, of these differences. One additional point of similarity exists: a water molecule, located within 3.5 Å of the cysteine sulfur atom, is present in all glutaminases examined in Table III. When the active site residues are superimposed, these water molecules superimpose within 1 Å of each other. In every case, they are located in a position consistent with nucleophilic behavior and are anchored there by at least two hydrogen bonds; in two of the five cases examined, they are close enough to the catalytic histidine to hydrogen-bond to one of the nitrogen atoms of the imidazole ring.
The active site residues of YaaE are shown in Fig. 4 superimposed upon the active site residues of yeast HisH with bound (␣S,5S)-␣-amino-3-chloro-4,5-dihydro-5-isoxazoleacetic acid (acivicin) (40). This inhibitor functions by covalently modifying the nucleophilic sulfur of the active site cysteine. The superposition shows conservation of the catalytic triad (Cys-79-His-170-Glu-172); however, the only other conserved residue is Gly-47, which is believed to contribute to the oxyanion hole. In HisH, the ␣-amino group of the substrate is positioned to hydrogen-bond to Gln-87, whereas the carboxyl group interacts with the backbone nitrogen atoms of Ser-148 and Phe-149 and also forms a hydrogen bond to Gln-397. The superposition suggests that Arg-106, which is conserved in YaaE sequences, may play a role in substrate binding. If so, the ␣-amino and carboxyl groups of the substrate glutamine would be interchanged by rotation about the C␣-C␤ bond, keeping the terminal amide group near Cys-79. The water molecule shown is found in all triad glutaminase structures and is positioned for nucleophilic attack in the final step of product formation in which the glutamate-enzyme thioester intermediate is hydrolyzed to glutamate and free enzyme.
As seen in Table III, the protein with the closest structural similarity to YaaE is the histidine biosynthesis protein HisH. HisH complexes with HisF to catalyze the reaction shown in Fig. 5. Secondary and tertiary structural prediction of YaaD using the SCOP (structural classification of proteins) data base (43) indicated that YaaD most likely folds in a (␤/␣) 8 barrel motif. Since HisF is also a (␤/␣) 8 fold and it complexes with HisH, which had the closest structural match to YaaE, it seemed as though a modeling study would reveal useful information. The lowest energy complex model is shown in Fig. 6.
In this model, YaaE binds to YaaD in the same general orientation as HisH binds to HisF: the active site of YaaE faces up into the N-terminal end of the YaaD model. The closest approach of YaaE to YaaD is where strands ␤7 and ␤8 pack against the N-terminal end of the ␤-barrel of YaaD. These strands contain three invariant phenylalanine residues that face in the direction of YaaD. The orientation of ␤8 appears to be due to an invariant Pro-137, which is also responsible for distorting one of the strands involved in the YaaE active site. The three invariant arginine residues that line the opening to the active site of YaaE are close enough to YaaD to presumably play some role in the interaction of YaaE with YaaD.
One of the more interesting features of the HisH-F enzyme complex is the hydrophobic tunnel through the center of the ␤-barrel, finally delivering the ammonia to the HisF active site (17). In modeling YaaD, we looked for a corresponding channel. In the YaaD model, a number of invariant residues were located in the barrel interior and in two adjacent loops, one between residues 212 and 218 and the other between residues 234 and 236 (Fig. 7). The region between these two loops corresponds to one of the phosphate binding sites of HisF and is predicted to be a phosphate binding site in YaaD based on sequence analysis. The remaining loops consist largely of unconserved residues. A number of hydrophobic residues were located in the core of the barrel, but a number of invariant polar or charged residues and two aromatic residues are also present near the interface. Calculating the accessible surface area using SPOCK (44) does not unambiguously indicate either the presence or the absence of a channel large enough for NH 3 to pass through. During the model construction, no attempt was made to place conserved residues at the C-terminal end of the barrel; however, it is known that the interface of the barrel is very highly conserved for HisH-F (17,40), and it would seem reasonable to expect similar conservation of the interface residues of YaaD-YaaE.
In summary, the model of YaaD reported here allows us to identify possible active site residues, whereas the model of the YaaD-YaaE complex indicates that the two enzymes can bind in an orientation favorable toward channeling of the ammonia molecule. The existence of an ammonia channel through YaaD cannot be definitely inferred from the model because of ambiguities in side chain positions. However, the structural and functional similarities of YaaD-E and HisH-F strongly suggest ammonia channeling.