Two-step Ligand Binding in a (βα)8 Barrel Enzyme

Background: HisA catalyzes a ring-opening isomerization reaction in histidine biosynthesis. Results: Catalytic residues and conformational changes upon substrate binding are clarified by structures, kinetics, and mutagenesis. Conclusion: Closing of active site loops in HisA brings the substrate into a product-like conformation before catalysis. Significance: This exemplifies coupled conformational changes in a (βα)8 barrel enzyme and its substrate and clarifies the mechanistic cycle of HisA. HisA is a (βα)8 barrel enzyme that catalyzes the Amadori rearrangement of N′-[(5′-phosphoribosyl)formimino]-5-aminoimidazole-4-carboxamide ribonucleotide (ProFAR) to N′-((5′-phosphoribulosyl) formimino)-5-aminoimidazole-4-carboxamide-ribonucleotide (PRFAR) in the histidine biosynthesis pathway, and it is a paradigm for the study of enzyme evolution. Still, its exact catalytic mechanism has remained unclear. Here, we present crystal structures of wild type Salmonella enterica HisA (SeHisA) in its apo-state and of mutants D7N and D7N/D176A in complex with two different conformations of the labile substrate ProFAR, which was structurally visualized for the first time. Site-directed mutagenesis and kinetics demonstrated that Asp-7 acts as the catalytic base, and Asp-176 acts as the catalytic acid. The SeHisA structures with ProFAR display two different states of the long loops on the catalytic face of the structure and demonstrate that initial binding of ProFAR to the active site is independent of loop interactions. When the long loops enclose the substrate, ProFAR adopts an extended conformation where its non-reacting half is in a product-like conformation. This change is associated with shifts in a hydrogen bond network including His-47, Asp-129, Thr-171, and Ser-202, all shown to be functionally important. The closed conformation structure is highly similar to the bifunctional HisA homologue PriA in complex with PRFAR, thus proving that structure and mechanism are conserved between HisA and PriA. This study clarifies the mechanistic cycle of HisA and provides a striking example of how an enzyme and its substrate can undergo coordinated conformational changes before catalysis.


barrel enzyme HisA, or NЈ-[(5Ј-phosphoribosyl)-
formimino]-5-aminoimidazole-4-carboxamide ribonucleotide (ProFAR) 5 isomerase, catalyzes the fourth step in the histidine biosynthesis pathway. The reaction is an Amadori rearrangement, in which the aminoaldose ProFAR is converted to the corresponding ketose PRFAR (Fig. 1). HisA has attracted considerable attention as a model both for the evolution of catalytic function and for evolution of the (␤␣) 8 barrel architecture itself. Another (␤␣) 8 barrel in tryptophan biosynthesis, TrpF, catalyzes an equivalent Amadori rearrangement on a chemically related substrate (5Ј-phosphoribosylanthranilate (PRA)). TrpF activity has been imparted on HisA by directed evolution (1) and by serial passaging a Salmonella enterica strain that lacked the trpF gene (2). Further, some Actinobacteria possess a HisAlike enzyme named PriA (phosphoribosyl isomerase), which catalyzes both the HisA and TrpF reactions (3). In some members of the genus Corynebacterium, PriA has evolved to become a newly respecialized HisA enzyme, termed subHisA, when trpF was regained through horizontal gene transfer (4).
The most well characterized HisA is the enzyme from Thermotoga maritima (TmHisA) (5)(6)(7). It possesses striking 2-fold symmetry and a significant level of sequence identity between its N-and C-terminal half-barrels, providing evidence that HisA evolved by duplication and fusion of an ancestral half-barrel (5). This 2-fold symmetry has also been observed in the solved structures of the PriA enzymes from Streptomyces coelicolor (ScPriA) and Mycobacterium tuberculosis (MtPriA), the C. efficiens subHisA, and also TmHisF, the homologous (␤␣) 8 barrel that catalyzes the subsequent step in histidine biosynthesis (4, 5, 8 -10). However, TrpF does not possess the same symmetrical architecture (11,12).
To date, there has been no structure available of any HisA in complex with its substrate or product. For this reason, mechanistically important residues have been inferred through comparisons with related enzymes. The structure of T. maritima TrpF (TmTrpF) in complex with its product analogue, reduced 1-[(2-carboxyphenyl)amino]-1-deoxyribulose 5-phosphate (rCdRP), was used to infer a common catalytic mechanism for TmHisA and TmTrpF (7). In this mechanism, isomerization of each substrate (ProFAR and PRA) involves protonation of the furanose ring oxygen by a general acid, a Schiff base intermediate, proton abstraction at the C2Ј position by a general base, and a spontaneous enol-keto tautomerization (Fig. 1). Based on mutagenesis of structurally equivalent residues in TmHisA and TmTrpF, Asp-8 (TmHisA numbering) was proven to be essential for activity and subsequently suggested to be the general base (7). In the same study, mutagenesis of Asp-127 suggested that it was the most likely candidate for the catalytic acid in TmHisA.
In contrast to the results with TmHisA, structures of MtPriA in complex with both rCdRP (the TrpF product analogue) and PRFAR (the product of the HisA reaction) showed that Asp-130 in MtPriA, the equivalent of Asp-127 TmHisA, was too distant from the furanose ring oxygen to be responsible for catalysis. Instead, Asp-175 was proposed to be the catalytic acid, and this was supported by mutagenesis showing that the MtPriA(D175A) mutant had lost activity toward both of its substrates, ProFAR and PRA (10).
The discrepancies between the identities of the proposed catalytic residues in HisA and PriA and the lack of ligandbound HisA structures inspired us to study this enzyme further. Moreover, T. maritima is a deep-branching anaerobic hyperthermophile that is predicted to resemble the last common ancestor of all bacteria (13) and has acquired ϳ24% of its genetic material from archaeal species through horizontal gene transfer (14). Given this unusual evolutionary history of T. maritima, we sought a new representative of "archetypal" HisA enzymes. S. enterica HisA (SeHisA) is both a canonical example (15) and a model for enzyme evolution (2). Here we present structures of SeHisA in the apo-state and in complex with its substrate, ProFAR, at two distinct stages of the catalytic cycle and clarify the mechanism for ligand binding and catalysis of HisA based on structures, kinetics, and mutational data.

Experimental Procedures
Cloning-The hisA coding sequence from S. enterica was amplified from a colony suspension using PCR with Phusion High-Fidelity DNA polymerase (Thermo Scientific, Waltham, MA) and the primers SeHisA.for and SeHisA.rev (Table 1) before being cloned into pEXP5-CT/TOPO (Life Technologies, Inc.) according to the manufacturer's instructions. This vector encoded a C-terminal hexahistidine (His 6 ) tag for hisA. DNA sequencing confirmed the correctness of the resulting plasmid, pEXP5-CT-hisA.
Enzymatic synthesis of the HisA substrate, ProFAR, required phosphoribosyl pyrophosphate (PRPP) as the starting material. To produce this from ribose 5-phosphate, we cloned the S. enterica prsA gene, which encodes PRPP synthetase. The prsA gene was amplified directly from S. enterica cells with Phusion polymerase and the primers SePRPPS.for and SePRPPS.rev (Table 1). The primers introduced restriction sites for BamHI (5Ј end) and EcoRI (3Ј end), allowing the PCR product to be cloned into the expression vector pJEX401 (DNA2.0 Inc., Menlo Park, CA). The vector encoded an N-terminal His 6 tag for prsA. The construction of pJEX401-prsA was confirmed by DNA sequencing.
HisF and HisH, which form the heterodimer imidazole glycerol phosphate synthase, were required for the HisA-coupled enzyme assay. Vectors for expressing the Escherichia coli genes (pCA24N-hisF and pCA24N-hisH) were taken from the ASKA FIGURE 1. Mechanism for the isomerization of ProFAR to PRFAR catalyzed by HisA (7). AH/A Ϫ , catalytic acid; B Ϫ /HB, catalytic base. A, the ring oxygen of the reacting ribose in ProFAR is protonated by the general acid. The free electron pair of the neighboring amine nitrogen forms a double bond with the 1Ј carbon. B, the resulting Schiff base intermediate acts as an electron sink allowing deprotonation of the 2Ј carbon by a general base. C, the resulting enolamine form of PRFAR spontaneously tautomerizes to the corresponding keto-form (D).  (16). The sequences encoding C-terminal green fluorescent protein tags were removed from each vector by digestion with NotI (New England Biolabs, Ipswich, MA), followed by religation of the plasmid.
Site-directed Mutagenesis-Site-directed mutagenesis was carried out according to the QuikChange II protocol. The primers for introducing each hisA mutation (encoding the D7N, D129N, D176N, D176A, and S202A amino acid substitutions) into pEXP5-CT-hisA are listed in Table 1. A double mutant D7N/D176A was made by introduction of the D176A mutation into pEXP5-CT-hisA(D7N). Each mutated plasmid was used to transform either E. coli XL10-Gold or E. coli BL21-Gold(DE3) cells. Transformed cells were spread on LB agar plates containing 100 g/ml ampicillin. Single colonies were used to inoculate cultures, from which plasmid DNA was prepared using the QIAprep Miniprep kit (Qiagen, Hilden, Germany). The presence of each desired mutation was confirmed by sequencing.
Protein Expression and Purification-All proteins were expressed in E. coli BL21(DE3) or E. coli BL21-Gold(DE3) cells, apart from PRPP synthetase, which was expressed in E. coli MC1061. Single colonies were used to inoculate 10-ml aliquots of LB medium containing the appropriate antibiotic: ampicillin (50 or 100 g/ml) for pEXP5-CT-hisA; kanamycin (30 g/ml) for pJEX401-prsA; and chloramphenicol (34 g/ml) for pCA24N-hisF and pCA24N-hisH. After growth to saturation (overnight at 37°C), each culture was used to inoculate 1 L LB medium in a shake flask, which was incubated at 37°C until the A 600 reached ϳ0.5. Cultures were moved to room temperature to cool, and after 30 min, expression was induced by the addition of 0.5 mM IPTG. IPTG-induced protein expression was carried out at 22-28°C for 16 -20 h. Cells were harvested by centrifugation, and the pellets were stored at Ϫ80°C until use.
For purification of PRPP synthetase, cell pellets were resuspended in a lysis buffer comprising 50 mM Tris-HCl and 300 mM KCl, pH 7.0. For all other proteins, the lysis buffer comprised 50 mM Tris-HCl and 300 mM NaCl, pH 7.5 or 8.0. A nuclease (either DNase I or Benzonase; both from Sigma) and protease inhibitors (Complete Protease Inhibitor from Roche Applied Sciences (Basel, Switzerland) or Protease Inhibitor Mixture from Sigma) were added to each cell suspension. Cells were lysed by using either a cell disruptor (Constant Systems, Daventry, UK) or a Vibra-Cell sonicator (Sonics & Materials, Newtown, CT) and centrifuged at 30,000 ϫ g for 20 min. Each lysate was clarified using a 0.45-m syringe filter and added to an Ni 2ϩ -Sepharose gravity column equilibrated with lysis buffer. The column was incubated under slow rotation at 4°C for 20 min, before extensive washing with lysis buffer supplemented with 25 mM imidazole. His 6 -tagged proteins were eluted with lysis buffer supplemented with 500 mM imidazole. Protein-containing fractions were pooled. For kinetics, the pooled fractions were exchanged into lysis buffer supplemented with 5 mM 2-mercaptoethanol (without imidazole). For crystallization, the pooled fractions were loaded onto a HiLoad 16/60 Superdex 75 column equilibrated with 50 mM Tris-HCl, 300 mM Na 2 SO 4 , and 5 mM 2-mercaptoethanol, pH 8.0. All proteins were concentrated to 20 -30 mg/ml using a Vivaspin concentrator, aliquoted, flash-frozen in liquid nitrogen, and stored at Ϫ80°C.
Preparation of ProFAR-The HisA substrate, ProFAR, was prepared according to methods modified from Ref. 17. E. coli strain FB1, which lacks the his operon (18), was transformed with phisGIE-tac (for IPTG-inducible overexpression of HisG and HisI). A saturated 10-ml culture of the transformed strain was used to inoculate LB (1 liter) containing 100 g/ml ampicillin and was grown with shaking at 37°C. At A 600 ϳ0.6, protein expression was induced with IPTG (1 mM), and the culture was incubated for a further 16 h at 28°C. The cells were harvested by centrifugation (3,000 ϫ g, 15 min) and washed in 100 mM Tris-HCl, pH 7.5, before being stored at Ϫ80°C.
At the time of ProFAR synthesis, cell pellets were thawed, resuspended, and lysed in 50 mM Tris-HCl and 300 mM KCl, pH 7.5, with lysozyme (0.5 mg/ml) and Benzonase nuclease (50 units) added, at room temperature for 20 min. Debris was removed by centrifugation (17,000 ϫ g for 1 min), and the supernatant was used for ProFAR synthesis, as described previously (17). ProFAR was purified from the lysate by anion exchange chromatography with a HiPrep Q FF 6/10 column (GE Healthcare, Little Chalfont, UK). The column was equilibrated with 60 mM ammonium bicarbonate, and ProFAR was eluted with a gradient of 60 -250 mM ammonium bicarbonate. The presence of ProFAR in peak fractions was tested in HisA activity assays (see below) and confirmed with liquid chromatography mass spectrometry (LC-MS), using a Poroshell 120 EC-C18 3 ϫ 50-mm column. Pooled fractions were lyophilized to remove residual ammonium bicarbonate and stored at Ϫ80°C. The yield and purity of ProFAR were quantified using the HisA assay; each preparation was typically 15-25% pure.
Crystallization, Data Collection, and Refinement-Crystallization was done in sitting drop vapor diffusion experiments. For wild type SeHisA, crystals were obtained with 1.5 l of 15 mg/ml HisA protein and 1.5 l of reservoir solution containing 0.1 M HEPES, pH 7.5, 0.8 M NaH 2 PO 4 , and 0.8 M KH 2 PO 4 (JCSGϩ screen, Hampton Research, Aliso Viejo, CA) at 20°C. Diamond-shaped crystals grew to a size of ϳ0.2 ϫ 0.2 ϫ 0.2 mm within 24 h. Crystals were cryoprotected in reservoir solution supplemented with 20% glycerol. For SeHisA(D7N), similar crystals were obtained with 20 mg/ml protein and a reservoir solution containing 0.2 M ammonium acetate, 0.1 M sodium acetate, and 20% PEG 4000 (pH set to 5.15). These crystals were soaked for 2 min in a cryosolution containing 50 mM sodium HEPES, pH 7, 150 mM NaCl, 30% PEG 4000, and 15% glycerol, to which ProFAR had been added as a lyophilized powder. SeHisA(D7N/D176A) cocrystallized with ProFAR at 8°C in a drop containing 0.075 l of 23 mg/ml protein solution, to which ProFAR had been added as a lyophilized powder, and 0.125 l of reservoir solution containing 0.1 M Bicine, pH 9.0, and 20% PEG6K (JCSGϩ screen, Hampton Research). One rod-shaped crystal with the approximate dimensions 0.3 ϫ 0.04 ϫ 0.04 mm appeared after 2 days. The crystal was cryoprotected in reservoir solution supplemented with 15% glycerol. All crystals were vitrified in liquid nitrogen.
Data were collected at ESRF (Grenoble, France) and Diamond Light Source (Didcot, UK) and processed using XDS (19). Data statistics are summarized in Table 2. The wild type SeHisA structure was solved by molecular replacement with the pro-Structure and Mechanism of HisA from S. enterica OCTOBER 9, 2015 • VOLUME 290 • NUMBER 41 gram Phaser (20) using a mutant SeHisA 6 as the search model, which in turn had been solved using the Campylobacter jejuni HisA (PDB entry 4GJ1) as the search model. The SeHisA structure was used as the search model for SeHisA(D7N) and SeHisA(D7N/D176A). The structures were rebuilt using Coot (21) and refined in phenix.refine (22). Refinement statistics are presented in Table 2. Structure figures were prepared using PyMOL version 1.7 (Schrödinger, LLC). Detailed structure comparisons were done using the LSQ commands in O (23,24).
Enzyme Kinetics-The HisA activity assay was adapted from one described previously (25). Assay mixtures contained 50 mM Tris-HCl (pH 7.5), 5 mM 2-mercaptoethanol, 25 mM L-glutamine, 2 M purified HisF, and 2 M purified HisH. The ProFAR concentration was varied from 1 M to 1 mM, and each reaction was initiated by the addition of a HisA protein (either SeHisA or one of the mutated variants) to a final concentration of 0.1 M (increased to 10 M when activity was poor). Reactions were performed at 25°C, and a decrease in absorbance at 300 nm was detected with a Cary 100 spectrophotometer (Agilent Technologies). The extinction coefficient difference for the substrate and product of the coupled reaction (⑀ ProFAR-AICAR ϭ 5,637 M Ϫ1 cm Ϫ1 ) was determined previously (25). The initial reaction rates were plotted and fitted to the Michaelis-Menten model using GraphPad Prism. Each enzyme variant was assayed in biological duplicate and technical triplicate.

Results
Structure Determination-SeHisA, SeHisA(D7N) and SeHisA(D7N/D176A) were expressed in E. coli with C-terminal His 6 tags. The mutation of Asp-7 was assumed (and then shown; see below) to make HisA inactive. Apo-crystals were obtained under several conditions with commercial crystallization screens, most of which contained phosphate or sulfate. Both SeHisA and SeHisA(D7N) crystallized in space group P6 1 22. A 1.7 Å SeHisA structure was solved by molecular replacement and showed a single protein molecule in the asymmetric unit. The enzyme has two phosphate ions from the crystallization solution bound to the substrate-binding site ( Fig.  2A), and two loops are disordered.
Structure Determination of SeHisA(D7N) and SeHisA(D7N/ D176A) in Complex with ProFAR-To gain further insight into the mechanism of substrate binding and catalysis, we set out to determine substrate-bound structures of SeHisA. The HisA substrate, ProFAR ( Fig. 1) was synthesized according to previously published methods (17). After many fruitless attempts, a 1.6 Å complex structure in space group P3 1 2 1 with clear elec- 6 A. Söderholm, X. Guo, and M. Selmer, unpublished data.  (Fig. 2, B and C).
Overall Structure-SeHisA adopts a classical (␤␣) 8 barrel fold (Fig. 2, A-C). As expected for enzymes with this fold (26), the ␣␤ loops (connecting each ␣-helix to the subsequent ␤-strand), on the "back side" of the enzyme, are short; in contrast, the catalytic face of the protein exposes the longer and flexible ␤␣ loops 1-8. ␤␣ loops 1 and 6 are partially unstructured in the apo-SeHisA and SeHisA(D7N)-ProFAR structures. To test whether the extensive 2-fold symmetry observed in TmHisA (5) was also present in the S. enterica enzyme, the N-terminal and C-terminal halves (residues 1-123 and 124 -240) of the fully ordered SeHisA(D7N/D176A)-ProFAR structure were superimposed. The r.m.s. deviation between the halfbarrels was only 1.7 Å for 100 C ␣ atoms, and the sequence identity was 19%, similar to the 2.1 Å r.m.s. deviation and 23% sequence identity observed for the two halves of TmHisA. The loop lengths follow this 2-fold symmetric pattern, and the longest loops, 1 and 5, form ␤-hairpins stabilized by interactions with the end of the loop and the preceding ␤-strand.
Structure of the SeHisA(D7N/D176A)-ProFAR Complex-ProFAR binds to SeHisA(D7N/D176A) in an extended conformation across the binding cleft, with both ribose entities in the C2Ј endo-conformation (Fig. 3A). The complete ligand makes extensive interactions with surrounding residues in the C-terminal parts of the ␤ strands and in the eight ␤␣ loops that shape and cover the active site (Fig. 3). Loops 1, 2, 5, and 6 are covering the ligand and sealing the active site, for which reason we will refer to this as the closed structure (Fig. 2C). The closed loop structures are stabilized by interactions with ProFAR as well as by direct and water-mediated interloop hydrogen bonds. Following the 2-fold symmetry of both the enzyme and the substrate, the phosphate groups at the two ends of ProFAR are coordinated by the N termini of loops 4 and 8, together with the backbones of loops 3 and 7. At the reacting end of the ligand, the phosphate (phosphate 1) of ProFAR interacts with the backbone amide hydrogens of Gly-177 in loop 6, Gly-204, Gly-225, and Arg-226 and with four structured waters. The phosphate adjoining the non-reacting ribose (phosphate 2) forms hydrogen bonds with the backbone amide hydrogens of Gly-81, Gly-102, and Ser-103 and with the side chains of Ser-103 and Arg-83. Four ordered water molecules surround this phosphate, two of which are strongly coordinated by backbone atoms of residues Val-82, Thr-104, Gly-80, and Val-100 and two that are coordinated by loops 5 and 2.
The reacting ribose of ProFAR is kept in place by hydrogen bond interactions between both the 2Ј hydroxyl and its neighboring secondary amine to Asn-7 (replacing Asp-7) and between the 3Ј-hydroxyl and Ser-202. The non-reacting ribose is positioned by a hydrogen bond between the 2Ј-hydroxyl group and Asp-129. The identical sugar puckers of the two riboses are stabilized by a bridging water molecule coordinated by His-47 and from the opposite side by a hydrogen bond network between Ser-202, Thr-171, and Asp-129 (Fig. 3, A and B). Because Asp-176 is mutated to Ala in this structure, it cannot form a hydrogen bond to the ligand; however, the reacting ring oxygen is only 4.1 Å away from the Ala side chain, suggesting that it would be at appropriate hydrogen bonding distance to Asp-176 in the wild type enzyme (see below).
The carboxamide group of ProFAR forms direct hydrogen bonds with the carbonyl oxygens of Gly-19 on loop 1 and Gly-144 on loop 5. The carboxamide aminoimidazole of ProFAR is sandwiched between a hydrophobic surface formed by Val-49 and Leu-51 and Trp-145 in loop 5, which makes a -stacking interaction with its opposite side (Fig. 3A).  (Fig. 2). We therefore define SeHisA(D7N)-ProFAR as the open liganded structure. All ProFAR interactions with loops 1, 2, 5, and 6 are absent in this open structure (Fig. 4), demonstrating that loop ordering and loop closure are not required for substrate binding.
Comparison of ProFAR in the open and closed structures shows that phosphate 1 and the reacting ribose superpose well (Fig. 5A). Apart from loop interactions, they form identical  OCTOBER 9, 2015 • VOLUME 290 • NUMBER 41 interactions with the enzyme (Figs. 3 and 4). The rest of the ligand adopts two different conformations (Fig. 5A). In the open structure, the distance between the two phosphate groups is 1.5 Å shorter, and the part of ProFAR comprising the nonreacting ribose plus carboxamide aminoimidazole is rotated relative to the fixed part of the ligand. This results in a shift of 2.5 Å of the 3Ј-hydroxyl of this ribose toward the C-terminal half of the barrel and of 1.5 Å of the carboxamide toward the N-terminal half of the barrel (Fig. 5A).

Structure and Mechanism of HisA from S. enterica
These differences can be explained by conformational changes in the surrounding region going from the open to the closed structure. In the open structure, the side chain carboxylate of Asp-129 is 3.5 Å from the 2Ј-hydroxyl of the non-reacting ribose, making water-mediated interactions with the 3Ј-hydroxyl as well as with Thr-171 (Figs. 4 and 5B). Upon loop closure, it moves 4 Å to form the interactions observed in the closed structure (Fig. 5C). In addition, crystal packing interactions involving loop 5 in the open structure position Trp-145 far from the carboxamide aminoimidazole moiety, with which it stacks in the closed structure. The shorter distance between the phosphate groups in the open structure results in phosphate 2 being in a position closer to the center of the enzyme, hence lacking the interactions with the side chains of Arg-83 and Ser-103 that are observed in the closed structure (Fig. 5D).
The phosphate ions in the apo-SeHisA structure bind in positions that are similar to the phosphates of the substrate (Fig. 2). Phosphate 1 in the apo-structure resides in two alternative positions about 2 Å apart, interacting in one position with loop 8 and in the other with both loops 7 and 8 (Fig. 5E). The position of the corresponding phosphate in the closed ligand structure is between these positions, and in the open structure it is close to loop 8, lacking the amide interactions that occur upon closure of loop 6 (Fig. 5D). Phosphate 2 in the apo-structure is positioned at a similar distance from the center as in the closed structure and stabilized by the same interactions with Arg-83 and Ser-103 (Fig. 5E). Thus, in the closed structure, both phosphates form additional interactions with the enzyme.
Comparison with Previous Structures-The Dali server (27) was used to search the PDB for structures most similar to SeHisA(D7N/D176A)-ProFAR because this was the only one of our structures in which all of the loops were ordered. The closest structural homologue was the MtPriA(D11N)-PRFAR complex (PDB code 2Y88, Z-score 37.0). This was followed by the other PriA complex structures (Streptomyces sviceus HisAp with degraded ProFAR (PDB code 4TX9, Z-score 36.7) and MtPriA-PRFAR (PDB code 3ZS4, Z-score 36.6)) and then by several apo-PriA structures. The closest structural homologue annotated as a HisA was the enzyme from C. jejuni (PDB code 4GJ1, Z-score 33.6), which was the top hit when performing the Dali search using apo-SeHisA. Despite its lower structural similarity with SeHisA(D7N/D176A)-ProFAR, the C. jejuni enzyme (CjHisA) has a substantially higher degree of sequence identity (51%) than MtPriA (33%). This result emphasizes that with the significant structural rearrangements upon ligand binding, the conformational state has a larger impact on overall structural similarity than sequence identity. Notably, the SeHisA structure was less similar to TmHisA (PDB code 1QO2, Z-score 27.8) than it was to many HisF structures.
SeHisA(D7N/D176A) was carefully compared with TmHisA (5) and with MtPriA in complex with the reaction product PRFAR (10) because both proteins have been biochemically characterized, and the structures have fully ordered loops. The apo-TmHisA structure superimposes on SeHisA with an r.m.s. deviation of 1.84 Å over 208 C ␣ atoms. SeHisA and MtPriA superimpose with an r.m.s. deviation of 1.39 Å over 242 C ␣ atoms, and the structures are highly similar; notably, the conformations of all loops are almost identical.
A previously overlooked aspect of HisA catalysis is that ring opening (Fig. 1) should increase the overall length of the product, PRFAR, relative to the substrate, ProFAR. Comparison of the product PRFAR in the MtPriA structure with ProFAR in the two structures of SeHisA shows that PRFAR adopts a conformation that, besides the reacting ribose, is very similar to the extended conformation of ProFAR observed in the closed SeHisA(D7N/D176A) structure (Fig. 6). The protein-ligand interactions are virtually identical in the SeHisA(D7N/D176A)-ProFAR and MtPriA-PRFAR structures, including the hydrogen bond network of Asp-129 around the non-reacting ribose and the stacking of Trp-145 with the carboxamide aminoimidazole moiety.
ConSurf analysis (28) shows that 39 residues display more than 96% conservation among sequences with Ͼ25% sequence identity to SeHisA. This conservation was mapped on a structure-based sequence alignment of sequences from previously  OCTOBER 9, 2015 • VOLUME 290 • NUMBER 41 characterized HisA and PriA enzymes (Fig. 7). Mapping the surface conservation on SeHisA confirmed that the entire substrate-binding pocket displays high conservation (data not shown). The suggested catalytic residues in TmHisA and

Structure and Mechanism of HisA from S. enterica
MtPriA, corresponding to Asp-7, Asp-129, and Asp-176 in SeHisA, were all completely conserved. Of these, Asp-7 in ␤1 and Asp-176 in loop 6 are embedded in conserved regions. The phosphate-binding sites and a motif spanning ␤2 and loop 2 also display high conservation. The alignment showed that all side chains involved in direct or water-mediated hydrogen bonds with ProFAR are conserved except for Ser-103 (sometimes conservatively replaced with threonine). Of the 39 highly conserved residues, 10 are not conserved in TmHisA, demonstrating that TmHisA is an outlier. Most interestingly, the substrate-binding residue Ser-202 is 100% conserved according to ConSurf analysis but is replaced by alanine (Ala-194) in TmHisA. Only two of the highly conserved residues are different in the PriA sequences. One is Ser-81 in MtPriA, corresponding to Gly-79 in SeHisA, which has been shown to be critical for the TrpF activity of PriA (9).
Enzyme Kinetics-The steady state kinetic parameters of SeHisA were determined using a coupled assay (25). In brief, the conversion of ProFAR to PRFAR by SeHisA (Fig. 1) was coupled to the conversion of PRFAR to AICAR, which in turn is catalyzed by the HisF-HisH heterodimer, imidazole  glycerol phosphate synthase. Using this assay, we determined that SeHisA had a k cat of 7.8 Ϯ 2.4 s Ϫ1 and a K m of 17.0 Ϯ 0.1 M; therefore, the catalytic efficiency (k cat /K m ) was 4.5 ϫ 10 5 s Ϫ1 M Ϫ1 ( Table 3).
Identification of Catalytic Residues-SeHisA isomerizes Pro-FAR using acid-base catalysis (Fig. 1) (7, 29). Our structures showed that Asp-7 is well positioned to act as the catalytic base (Figs. 3 and 4), consistent with the data for TmHisA and MtPriA (7,10). The SeHisA(D7N/D176A)-ProFAR structure shows that closure of loop 6 places Asp-176 in position to be the catalytic acid (Fig. 3A), in analogy with MtPriA (10). Asp-129, the equivalent of the suggested catalytic acid in TmHisA (7), is coordinating the non-reacting ribose, 6.8 Å away from the oxygen to be protonated (Figs. 3 and 4). Our attention was also drawn to Ser-202, which coordinates the 3Ј-hydroxyl of the reacting ribose (Figs. 3 and 4) and is replaced by Ala-194 in the T. maritima enzyme. In order to probe the roles of these residues as proton donors and acceptors, while introducing minimally disruptive substitutions, we constructed the SeHisA (D7N), SeHisA(D129N), SeHisA(D176N), and SeHisA(S202A) variants. Steady state kinetics data for each mutant are shown in Table 3.
Asp-7 was confirmed to be essential for function because no activity was detected for the SeHisA(D7N) mutant. These data, along with the position of the side chain in the active site and its conservation between HisA and PriA enzymes, confirmed its role as the general base. The role of general acid appears to be played by Asp-176 because the D176N substitution led to a reduction in k cat of greater than 400-fold (Table 3). In contrast, the D129N substitution only reduced k cat by 15-fold. This was accompanied by a 3-fold reduction in the K m for ProFAR, resulting in a variant that retained almost 20% of overall catalytic efficiency ( Table 3). The S202A substitution led to an increase in K m from 17 to 40 M, supporting the role in ligand binding that was predicted from our structures and ruling out any role in catalysis.

Discussion
Here we have presented the structure of SeHisA, in addition to kinetic parameters and an in depth examination of the active site. Our two substrate-bound structures (SeHisA(D7N)-ProFAR, which represents initial substrate binding, and SeHisA(D7N/D176A)-ProFAR, where the substrate adopts a product-like conformation) provide greater insight into the HisA active site and reaction mechanism than was previously possible. These are the first ever protein structures to be co-crystallized with the labile metabolite ProFAR, which has a short half-life compared with crystallization time scales but is slightly more stable at neutral to alkaline pH (17). With this in mind, a closed ProFAR complex structure where the ordered loops cover the ligand was achieved with the doublemutant SeHisA(D7N/D176A) crystallized in an alkaline condition (pH 9) and at low temperature (8°C). A second, open, ProFAR complex structure with disordered loops was obtained by brief soaking of a SeHisA(D7N) crystal (obtained at pH 5.15) in ligand solution at physiological pH. In this crystal form, loop 5 is involved in crystal packing, preventing loop closure.
The (␤␣) 8 barrels of SeHisA and TmHisA both show strong 2-fold symmetry, confirming that this is a general feature of HisA enzymes. The symmetry of HisA and its bisphosphorylated substrate, ProFAR (Fig. 1), raises the question of how the enzyme recognizes the correct orientation of the substrate. The open SeHisA(D7N)-ProFAR structure shows that ligand-loop interactions are not involved in the initial direction-specific recognition of the substrate. Conserved but subtle deviations from perfect 2-fold symmetry seem sufficient to dictate the correct binding of ProFAR to the open state of HisA. At the end of ␤2, the conserved residues Val-49 and Leu-51 make hydrophobic interactions with the aminoimidazole carboxamide moiety (Fig. 4A). In contrast, at the end of symmetry-equivalent strand 6, the conserved residue Thr-171 (symmetry-related to Val-49) is engaged in a hydrogen bond network and water-mediated ligand binding (Figs. 4 and 5B). Docking of ProFAR in the  Structure and Mechanism of HisA from S. enterica OCTOBER 9, 2015 • VOLUME 290 • NUMBER 41 opposite orientation suggests that the aminoimidazole carboxamide moiety may also clash slightly with loop 2. Upon loop ordering, additional interactions are formed with the asymmetric central part of the ligand (Fig. 3). Our ProFAR-bound structures illuminate key details of the acid-base catalysis that is carried out by HisA. The general base is Asp-7, which is essential for activity (Table 3) and well positioned to abstract a proton from C2Ј according to the structures of SeHisA(D7N/D176A) and SeHisA(D7N) with ProFAR (Fig.  3). In the MtPriA-PRFAR structure, showing a state after ring opening, the equivalent Asn-11 forms hydrogen bonds to the C2Ј and C3Ј hydroxyls of the product. The structural details of the intermediate on which the general base acts (Fig. 1) are unknown. However, our structures suggest that Asp-7 could stabilize the Schiff base through a salt bridge, potentially taking on an additional function in catalysis. Asp-7 occupies a nearly identical position in the wild type apo-structure, and PROPKA3 (30) predicts that the pK a of its carboxylate is elevated to ϳ5.5 due to its relatively hydrophobic chemical environment. This indicates that it will be deprotonated at physiological pH, albeit with increased basicity as required to accept the C2Ј proton.
Our data confirm unequivocally that Asp-176 has the role of general acid. This residue is located on the flexible loop 6, which is only ordered in the closed SeHisA(D7N/D176A)-ProFAR structure. Despite being mutated to alanine in this structure, the amino acid overlays perfectly with the equivalent residue (Asp-175) in TmPriA-PRFAR (Fig. 6). In silico mutagenesis of Ala back to Asp showed that the Asp side chain is positioned to donate a proton to the furanose ring oxygen. During this work, we also solved a SeHisA(D7N) structure with partial electron density for ProFAR and loop 6, where the Asp-176 side chain was positioned for catalysis (data not shown). Further evidence for the role of Asp-176 was provided by the introduction of the D176N mutation, which reduced k cat by 400-fold but did not alter K m at all ( Table 3). The equivalent residue in TmHisA, Asp-169, would need to move 4 Å upon substrate binding in order to act in the same fashion. Such a movement is entirely consistent with the pattern of loop closures we have observed for SeHisA. Thus, the catalytic residues, Asp-7 and Asp-176 in SeHisA, are strictly conserved in SeHisA, TmHisA, and MtPriA (10).
In TmHisA, the equivalent of the D129N mutation was shown to reduce turnover ϳ2,500-fold (7), albeit in assays conducted 55°C below the temperature optimum of the enzyme. This suggested to the authors that Asp-129 might be the catalytic acid. Our ProFAR complex structures show that Asp-129 is too far from the furanose ring oxygen to play such a direct role in catalysis. Nevertheless, the D129N mutation in SeHisA did reduce k cat (by 15-fold) while also lowering K m (Table 3). This indicates a role for this aspartate in catalysis rather than binding (Table 3). Upon loop closure, Asp-129 moves 4 Å, making changes in its hydrogen bonding network to stabilize the extended product-like conformation of ProFAR together with Thr-171, Ser-202, and His-47 (Fig. 5C). In the open structure, only one of the side chain oxygens of Asp-129 is involved in a hydrogen bond (Fig. 4B), whereas in the closed structure, both side chain oxygens should function as hydrogen bond acceptors (Fig. 3B). Thus, the introduction of Asn in place of Asp-129 appears to be favorable for substrate binding to the open enzyme conformation, but the inability of Asn to accept a second hydrogen bond disfavors the transition to the product-like ligand conformation in the closed enzyme. These observations explain the effects of the D129N mutation on K m and k cat . The same logic is also likely to explain why mutation of the equivalent residue in MtPriA (D130A) leads to a 20-fold decrease in k cat while not affecting K m (10).
The functional importance of His-47 and Thr-171 has previously been demonstrated in HisA and/or PriA (7,9,10); here we have demonstrated that Ser-202 also has an important role. Ser-202 makes hydrogen bond interactions with the 3Ј hydroxyl of the reacting ribose in both the open and closed structure (Figs. 3, 4, and 5 (B and C)). Therefore, the negative impact of the S202A mutation on both k cat and K m (Table 3) is consistent with a role in initial binding of the substrate as well as in stabilization of the product-like conformation. The absence of an equivalent to Ser-202 in TmHisA (in which it is replaced by Ala) might explain why the D129N mutation has a more detrimental effect on this enzyme.
The phosphates in the apo-SeHisA structure most likely bind to the enzyme's preferred phosphate binding sites. Thus, electrostatic attraction by the conserved Arg-83 on the second phosphate, bringing it into the preferred binding site, may contribute to extending the substrate while phosphate 1 is kept tightly in place (Fig. 5, D and E). Preliminary density functional theory calculations indicate that there is FIGURE 8. The catalytic cycle of SeHisA. A, in the uncharged enzyme, loops 1 and 6 are flexible, and loops 2 and 5 are distant from each other. B, ProFAR docks into the fully exposed active site. C, binding of the substrate induces conformational changes in the enzyme. Loops 1, 2, 5, and 6 form interactions with the substrate, resulting in loop closure and burying of the ligand. Coupled to loop ordering, the non-reacting half of ProFAR moves toward the edge of the barrel and adopts a product-like extended conformation stabilized by additional enzyme interactions. D, the conformational changes of substrate and enzyme prepare the complex for catalysis. After the reaction, the loops open, and PRFAR is released. no significant energy difference between the two observed conformations of ProFAR in gas phase (data not shown). The movement of Gly-204 to make an additional interaction with phosphate 1 in the closed structure (Fig. 5D) is necessary sterically to allow closure of loop 6. In the MtPriA(D11N)-PRFAR structure (10), a salt bridge between Arg-19 and Asp-175 was suggested to be essential for recruiting loop 6. In contrast, the equivalent salt bridge (between Arg-15 and Asp-176) is not needed in SeHisA because the closed structure was obtained with a D176A mutant.
Sequence analysis showed that SeHisA, rather than TmHisA, is an archetypal representative of the HisA family (Fig. 7). Further, SeHisA is more similar to the PriA enzymes in sequence and structure than it is to TmHisA. The structural similarity is particularly close between SeHisA(D7N/D176A)-ProFAR and MtPriA(D11N)-PrFAR, in which all ligand-binding interactions are conserved. Thus, we have shown that neither structure nor the mechanism of ProFAR isomerization has diverged in HisA and PriA, despite selection for bifunctionality in the latter enzyme. When the sequence/structure distance from SeHisA to TmHisA and the uneven phylogenetic distribution of PriA enzymes are both considered, the most parsimonious explanation is that PriA enzymes have evolved from HisA ancestors in a handful of lineages (31).
Based on our crystal structures and mutagenesis of SeHisA, we can propose a general mechanism of substrate binding and catalysis in HisA (Fig. 8). The apo-SeHisA structure and the SeHisA(D7N)-ProFAR structure show that active site loops 1 and 6 are inherently flexible and that the catalytic cycle begins with a discrete, loop-independent, substrate-binding step (Fig. 8, A and B). Hydrophobic residues at the end of ␤2 (Val-49 and Leu-51) and the symmetryrelated hydrophilic ␤6 (Thr-171) guide the substrate to bind in the correct orientation. Next, as observed in the SeHisA(D7N/D176A)-ProFAR structure, loops 1 and 6 become ordered, and loops 2 and 5 adopt a closed conformation resulting in complete sequestration of the substrate. Coupled with loop closure, the substrate adopts its extended conformation. The non-reacting half of the substrate adopts a product-like conformation, as observed in MtPriA(D11N)-PrFAR, with phosphate 2 located in a similar position as it is in the apo-SeHisA structure ( Fig. 8C and supplemental Movie 1). The interactions stabilizing this conformation of ProFAR involve Asp-129 and the hydrogen-bonding network to Thr-171 and Ser-202, an electrostatic attraction on phosphate 2 by Arg-83, and additional interactions with loops 1, 2, 5, and 6. The functional importance of this transition is supported by the effect of the D129N mutation. In the closed state, the substrate is buried from the surrounding water, and the catalytic residues Asp-7 and Asp-176 are positioned for acidbase catalysis (Fig. 8D). The interactions of the enzyme with the product stay the same apart from around the reacting ribose, as shown in the MtPriA(D11N)-PRFAR structure (10). The catalytic cycle is completed when the loops open to release the product PRFAR. In conclusion, our two substrate-bound structures illuminate a two-step mechanism for ProFAR binding to HisA and provide an example of cou-pled conformational changes of enzyme and substrate in enzymatic catalysis.