2.1. Protein production and crystallization

Clones were generated using the Polymerase Incomplete Primer Extension (PIPE) cloning method (Klock et al., 2008). The gene encoding HinfPDH (GenBank AAC22939, gi|1574749; Swiss-Prot TYRA) was amplified by polymerase chain reaction (PCR) from H. influenzae Rd KW20 genomic DNA using PfuTurbo DNA polymerase (Stratagene) and I-PIPE (Insert) primers (forward primer, 5′-­ctgtacttccagggcATGCGTGAATCCTATGCCA­ATGAAAACC-3′; reverse primer, 5′-aattaagtcgcgttaGCATAAAACGGCGTAGAA­CATCTTCAAT-3′; target sequence in upper case) that included sequences for the predicted 5′ and 3′ ends. The expression vector pSpeedET, which encodes an amino-terminal tobacco etch virus (TEV) protease-cleavable expression and purification tag (MG­SDKIHHHHHHENLYFQ/G), was PCR-amplified with V-PIPE (Vector) primers. The V-PIPE and I-PIPE PCR products were mixed to anneal the amplified DNA fragments together. E. coli GeneHogs (Invitrogen) competent cells were transformed with the V-PIPE/I-­PIPE mixture and dispensed onto selective LB–agar plates. The cloning junctions were confirmed by DNA sequencing. Using the PIPE method, the part of the gene encoding residues Met1–Phe80 was excluded from the final construct. Expression was performed in selenomethionine-containing medium with suppression of normal methionine synthesis (Van Duyne et al., 1993). At the end of fermentation, lysozyme was added to the culture to a final concentration of 250 µg ml⁻¹ and the cells were harvested and frozen. After one freeze–thaw cycle, the cells were sonicated in lysis buffer [50 mM HEPES pH 8.0, 50 mM NaCl, 10 mM imidazole, 1 mM tris(2-­carboxyethyl)phosphine–HCl (TCEP)] and the lysate was clarified by centrifugation at 32 500g for 30 min. The soluble fraction was passed over nickel-chelating resin (GE Healthcare) pre-equilibrated with lysis buffer, the resin was washed with wash buffer [50 mM HEPES pH 8.0, 300 mM NaCl, 40 mM imidazole, 10%(v/v) glycerol, 1 mM TCEP] and the protein was eluted with elution buffer [20 mM HEPES pH 8.0, 300 mM imidazole, 10%(v/v) glycerol, 1 mM TCEP]. The eluate was buffer-exchanged with TEV buffer (20 mM HEPES pH 8.0, 200 mM NaCl, 40 mM imidazole, 1 mM TCEP) using a PD-10 column (GE Healthcare) and incubated with 1 mg TEV protease per 15 mg of eluted protein. The protease-treated eluate was run over nickel-chelating resin (GE Healthcare) pre-equilibrated with HEPES crystallization buffer (20 mM HEPES pH 8.0, 200 mM NaCl, 40 mM imidazole, 1 mM TCEP) and the resin was washed with the same buffer. The flowthrough and wash fractions were combined and concentrated to 19.6 mg ml⁻¹ by centrifugal ultrafiltration (Millipore) for crystallization trials. HinfPDH was crystallized using the nanodroplet vapor-diffusion method (Santarsiero et al., 2002) with standard JCSG crystallization protocols (Lesley et al., 2002). Sitting drops composed of 200 nl protein mixed with 200 nl crystallization solution were equilibrated against a 50 µl reservoir at 293 K for 28 d prior to harvest. The crystallization reagent was composed of 0.04 M potassium dihydrogen phosphate, 20.0%(v/v) glycerol and 16.0%(w/v) PEG 8000. A rod-shaped crystal of approximate dimensions 0.1 × 0.05 × 0.05 mm was harvested for data collection. No additional cryoprotectant was added to the crystal. Initial diffraction screening was carried out using the Stanford Automated Mounting system (SAM; Cohen et al., 2002) at the Stanford Synchrotron Radiation Lightsource (SSRL, Menlo Park, California, USA). The crystal was indexed in the tetragonal space group P4₁2₁2. The oligomeric state of HinfPDH was determined to be a dimer by gel filtration using a 0.8 × 30 cm Shodex Protein KW-803 column (Thomson Instruments) equilibrated in 20 mM Tris, 200 mM NaCl, 0.5 mM TCEP pH 7.5 and pre-calibrated with gel-filtration standards (Bio-Rad). Protein con­centrations were determined using the Coomassie Plus assay (Pierce).

2.2. Data collection, structure solution and refinement

Multiple-wavelength anomalous diffraction (MAD) data were collected on beamline 11-1 at the SSRL using a 0.1 × 0.1 mm X-ray beam at wavelengths corresponding to the inflection (λ₁) and remote (λ₂) wavelengths of a selenium MAD experiment. The data sets were collected at 100 K using a Rayonix MAR Mosaic MX-325 CCD detector. The MAD data were integrated and reduced using XDS and scaled with the program XSCALE (Kabsch, 1993, 2010a,b). An initial substructure solution was obtained with SHELXD (Sheldrick, 2008) and the phases were refined using autoSHARP (Vonrhein et al., 2007), which gave a mean figure of merit of 0.48 with 14 selenium sites. Automated model building was performed with ARP/wARP (Cohen et al., 2004). Model completion and refinement were per­formed with Coot (Emsley & Cowtan, 2004) and REFMAC 5.2 (Winn et al., 2003) using the high-energy remote (λ₂) data set. The refinement included phase restraints from SHARP and TLS refinement with three TLS groups per chain. CCP4 programs were used for data conversion and other calculations (Collaborative Computational Project, Number 4, 1994).

During structure refinement, additional electron density was found at the active site. The density was well defined and could be unambiguously assigned to a tyrosine molecule and an NAD⁺ cofactor. As these molecules were not added during the crystallization experiment, they must have been acquired during protein expression and have remained bound during purification and crystallization. Data-collection and refinement statistics are summarized in Table 1.

2.3. Validation and deposition

The quality of the crystal structure was analyzed using the JCSG Quality Control server (https://smb.slac.stanford.edu/jcsg/QC ). This server verifies the stereochemical quality of the model using AutoDepInputTool (Yang et al., 2004), MolProbity (Chen et al., 2010) and WHATIF v.5.0 (Vriend, 1990), the agreement between the atomic model and the data using SFCHECK v.4.0 (Vaguine et al., 1999) and RESOLVE (Terwilliger, 2004), the protein sequence using ClustalW (Thompson et al., 1994), the atom occupancies using MOLEMAN2 (Kleywegt, 2000) and the consistency of NCS pairs. Protein quaternary-structure analysis was conducted using the PISA server (https://www.ebi.ac.uk/msd-srv/prot_int/pistart.html ; Krissinel & Henrick, 2005). Fig. 2(d) was adapted from PDBsum (Laskowski, 2009); all other figures were prepared with PyMOL (DeLano, 2002). Atomic coordinates and experimental structure factors have been deposited in the PDB under accession code 2pv7 .

3.1. Overall structure

The final model consists of a dimer of the PDH domain of TyrA (residues 92–371 for chains A and B), two nicotinamide adenine dinucleotides (NAD⁺), two tyrosines and 393 water molecules in the asymmetric unit (Figs. 2a, 2b and 2c). No electron density was observed for Gly80–Phe91, Asp312–Glu315 and Asn372–Gly377 in chain A and Gly80–Val91 and Asn372–Gly377 in chain B. The side chains of Lys212 and Lys348 from chain A and of Arg132, Lys212, Lys239, Gln325 and Ala371 from chain B were omitted owing to poor electron density. The Matthews coefficient (V_M; Matthews, 1968) was 2.98 ų Da⁻¹ and the estimated solvent content was 58.7%. The Ramachandran plot produced by MolProbity showed that 98% of the residues are in favored regions, with no outliers.

Each monomer consists of an N-terminal α/β dinucleotide-binding domain (residues 92–243) and a C-terminal α-helical dimerization domain (residues 244–371) (Fig. 2a). The active site is located at the domain interface. The N-terminal domain adopts a modified Rossmann fold, which consists of a parallel seven-stranded β-sheet (strand order β2-β1-β3-β4-β5-β6-β7) with the α1 helix on one face of the β-­sheet and the α2, α3, α4 and α5 helices on the other. Structural comparisons of HinfPDH and other nucleotide-binding proteins, including other TyrA enzymes, show that HinfPDH lacks the α–β structural motif between α2 and β2 that is present and is part of the β-­sheet in other nucleotide-binding proteins. The C-terminal domain consists of seven helices (α6–α12) that form the dimer interface; the helices from each monomer are intertwined into a tightly packed helical bundle with a buried surface of 11 000 Ų (Figs. 2b and 2c). The two monomers are structurally similar to each other, with an r.m.s.d. of 0.2 Å for 268 equivalent C^α atoms.

3.2. The active site

The active site is located in the cleft between the N- and C-terminal domains (Figs. 2a, 2b and 2c). One tyrosine and one NAD⁺ are bound in each monomer (Fig. 3a). His200, Ser179 and Arg297 (Fig. 3b) are among the important residues for enzyme catalysis and/or ligand binding and are conserved in TyrA enzymes across all kingdoms of life. The His200 imidazole is hydrogen bonded to the 4′-OH of the bound tyrosine. Ser179 hydrogen bonds to both the ribose O atom of nicotinamide nucleoside and the 4′-OH of the bound tyrosine and is important for orientating prephenate and NAD⁺ for catalysis. The Arg297 guanadinium forms a pair of electrostatic interactions with the tyrosine carboxyl, which also interacts with Gln301 from the α8 helix. The tyrosine amino group hydrogen bonds to Tyr306 from α9 and Tyr288′ from α8 of the adjacent monomer. His260 is located close to the tyrosine amino group and could be involved in regulation of tyrosine inhibition in a similar way to His217 in A. aeolicus.

The aromatic ring of the bound tyrosine packs against the nico­tinamide ring of NAD⁺ such that the 4′-OH of tyrosine is approximately 4 Å away from C4 of the nicotinamide ring. Assuming that prephenate adopts the same binding mode, the structure suggests that the hydride is transferred from prephenate to the si face of NAD⁺, which is consistent with a previous proton NMR study on the E. coli TyrA enzyme using isotope-labeled NAD-4-d (Hermes et al., 1984).

The binding mode of NAD⁺ is similar to those of arogenate de­hydrogenase from Synechocystis sp. and prephenate dehydrogenase from A. aeolicus (Legrand et al., 2006; Sun et al., 2006). The pyrophosphate of NAD⁺ interacts with the P-loop (Gly108–Gly113) between α1 and β1, forming hydrogen bonds to Lys111 and the main-chain amides of Lys111 and Leu112. The diol of the adenylyl ribose hydrogen bonds to Asp131 and Lys132 from the loop located between α2 and β2. The adenine ring is sandwiched between Val152 and Arg132 from helices α2 and α3, with its N1 hydrogen bonded to Trp135 from α2. The diol of the nicotinamide ribose is hydrogen bonded to Ser179 and Val152. The nicotinamide ring interacts with the protein mainly through hydrophobic interactions.

3.3. Structural comparison with other TyrA enzymes

A search with FATCAT (Ye & Godzik, 2004) using the HinfPDH coordinates identified the closest structural homologues of HinfPDH in the PDB as the prephenate dehydrogenases from A. aeolicus (AaeoPDH; PDB code 2g5c ; Sun et al., 2006) and Streptococcus thermophilus (SthePDH; PDB code 3dzb ; Z. Zhang, S. Eswara­moorthy, S. K. Burley & S. Swaminathan, unpublished work) and the arogenate dehydrogenase from Synechocystis sp. (SyneADH; PDB code 2f1k ; Legrand et al., 2006). HinfPDH is bifunctional, whereas the other three enzymes are monofunctional. The pairwise sequence identities between HinfPDH and AaeoPDH, SthePDH and SyneADH are 20, 27 and 25%, respectively. Despite the low sequence identity, the overall structures of these enzymes are very similar. The structures of AaeoPDH complexed with the ligands NAD⁺ (PDB code 2g5c ), NAD⁺ and L-tyrosine (PDB code 3ggg ), NADH and 4-hydroxylphenylpyruvate (HPP; PDB code 3ggo ) and NAD⁺ and 4-hydroxyphenylpropionate (PDB code 3ggp ) are available (Sun et al., 2009), but only the first two structures were used for comparison because they are sufficient to represent the two unique enzyme states; they are denoted AaeoPDH and AaeoPDH–Tyr–NAD⁺, respectively. SyneADH has an NADP⁺ bound at the active site and SthePDH has no ligand bound. The dehydrogenase activity of SyneADH is strictly dependent on arogenate and NADP⁺ (Legrand et al., 2006). Conversely, NAD⁺ and prephenate are the preferred cofactor and substrate for AaeoPDH, although a very low level of dehydrogenase activity is detected when NADP⁺ with pre­phenate or NAD⁺ with arogenate are used (Bonvin et al., 2006).

Although the overall fold is similar, structural comparisons revealed important differences around the active site. Compared with other TyrA enzymes, HinfPDH lacks an α–β structural motif between α2 and β2, the loop between β5 and β6 (L_β5–β6) is much shorter, and an extra helix α12 is found at the C-terminus (Fig. 4). Multiple sequence alignment of many representative TyrA proteins clearly shows that the α–β motif and L_β5–β6 represent unique structural differences between bifunctional and monofunctional TyrA enzymes (Fig. 5), but the extra C-terminal helix is less obviously discernable from the sequence comparisons, presumably because the exact end point of the prephenate dehydrogenase domain is difficult to determine for cases such as TyrA-aroF or TyrA-ACT fusions in which aroF and ACT are fused at the C-terminal end.

In HinfPDH, L_β5–β6 is eight residues shorter than in AaeoPDH, SthePDH and SyneADH. This loop is well ordered in the HinfPDH, AaeoPDH and SyneADH structures (Fig. 4c) but is disordered in the SthePDH structure owing to the absence of bound cofactor. In AaeoPDH and SyneADH, this loop extends along the cofactor-binding site. Ser155 in AaeoPDH and Gln120 in SyneADH are in structurally equivalent positions (Fig. 4c) and both form hydrogen bonds to the pyrophosphate O atom of the bound cofactor. However, in HinfPDH no residue is structurally equivalent because of the shorter loop and Lys111 from α1 instead provides the equivalent interaction with the pyrophosphate O atom (Fig. 4c). The equivalent residues to Lys111 in AaeoPDH and SyneADH are Phe40 and Leu10, respectively, and neither side chain can form hydrogen bonds to the pyrophosphate. This suggests that although a shorter L_β5–β6 loop has evolved in HinfPDH, the ability to bind cofactor is not affected.

In SyneADH, the phosphate group of the adenosine ribose of NADP⁺ is recognized by helical residues in the α–β motif. The phosphate group is stabilized by electrostatic interaction with Arg31 and hydrogen-bonding interactions with Gln32 and Thr35, as well as with the main chain of Arg31 and Gln32. The equivalent residues in AaeoPDH are Ile63, Asn64 and Ser67. A loss of electrostatic interaction caused by the substitution of Arg by Ile may explain why AaeoPDH prefers NAD⁺ over NADP⁺ as cofactor. In SthePDH, the equivalent residues are Arg36, Ser37 and Ser40. Thus, the ability to form electrostatic and hydrogen-bonding interactions is similar to that of SyneADH, suggesting that SthePDH is capable of binding NADP⁺. In HinfPDH, Arg132 is structurally equivalent to SyneADH Arg31, but residues equivalent to Gln32 and Thr35 are absent owing to the lack of the α–β structural motif. It is possible that HinfPDH also prefers NAD⁺ over NADP⁺ in a manner similar to bifunctional TyrA enzymes from E. coli and K. pneumonia.

A smaller local difference among these TyrA structures is the loop joining α7 and α8 and the adjacent residues (Fig. 4d). This loop takes a wider turn in HinfPDH compared with that in AaeoPDH, SthePDH and SyneADH and is positioned next to α12. Ser284–Leu290 in this region are highly conserved in TyrAc_Δ sequences, suggesting that they play important roles.

3.4. Global conformational change

Another important difference between HinfPDH and other TyrA structures is the relative orientation of the respective N- and C-­terminal domains. Pairwise structural alignment of HinfPDH with AaeoPDH, AaeoPDH–Tyr–NAD⁺, SthePDH and SyneADH using only the N-terminal domain gives r.m.s.d.s of 1.8, 1.7, 1.9 and 1.5 Å, respectively, for 128–132 superimposed C^α atoms (0.8–0.9 Å for core β-sheet residues). Upon structural superimposition, it is immediately noticeable that differences in the relative orientation of the N- and C-­terminal domains are present in these TyrA structures (Figs. 6a and 6b). A hinge region around Glu242–Asn244 connects the N- and C-­terminal domains at the domain interface opposite to the substrate-binding site. Superimposition using the C-terminal domain gives similar results, although the results are less obvious owing to internal structural differences within the C-terminal domains of these TyrA structures. Therefore, the discussion below is based on superimpositions using the N-terminal domain.

Using the C-terminal domain of HinfPDH as a reference, the C-­terminal domains of AaeoPDH, SthePDH and SyneADH are farther away from the substrate-binding site; therefore, HinfPDH represents the most closed form, SyneADH is the most open form and AaeoPDH and SthePDH are in intermediate states (Figs. 6a and 6b). Monomer B of AaeoPDH–Tyr–NAD⁺ has both NAD⁺ and tyrosine bound and the conformation is closed, similar to HinfPDH. Monomer A of AaeoPDH–Tyr–NAD⁺ only has NAD⁺ bound and the conformation is similar to AaeoPDH. Comparison of the AaeoPDH and AaeoPDH–Tyr–NAD⁺ structures suggest that tyrosine induces a conformational change upon binding. Hence, it is possible that the binding of tyrosine to HinfPDH also induces a conformational change from an open to a closed form and the closed form is captured in the current HinfPDH structure. A crystal structure of apo HinfPDH could provide direct evidence for this proposal. In the closed conformation, α6, α8, α9 and α8′ (α8 from monomer B) are close to the bound tyrosine and α8 and α12′ (α12 from monomer B) are near L_β5–β6, facilitating closure of the active site and the proper alignment of active-site residues for catalysis. Key active-site residues in this region include His260 (α6), Arg297, Gln301 (α8), Tyr288′ (α8′), Tyr306 (α9), Arg365′ (α12′) and Asp206 (L_β5–β6). If a tyrosine molecule is modeled adjacent to the cofactor in AaeoPDH and SthePDH, the residue equivalent to Arg297 of HinfPHDH is too distant to interact with the tyrosine. The conformational change is independent of cofactor binding as the structures of SyneADH with NADP⁺ bound, AaeoPDH with NAD⁺ bound and SthePDH without any cofactor are all in similar open conformations.

In the closed conformation of HinfPDH an ionic network con­sisting of Arg297, a bridging water molecule, Asp206 from L_β5–β6 and Arg365′ from α12′ is observed that may be involved in gating the active site (Fig. 6c). The bridging water is present in both monomers. The Arg365′ side chain adopts dual conformations, with one con­formation participating in the ionic network and the other pointing away from the active site. This dual conformation may be part of the gating mechanism, in which an ionic network forms on closure of the active site after substrate is bound and is broken when the product is released. It is worth noting that Asp206 is absolutely conserved and Arg365 is highly conserved in chorismate mutase/prephenate dehydrogenase sequences but not in monofunctional TyrA enzymes. Thus, the active-site gating mechanism might be different in bifunctional and monofunctional TyrA enzymes.

3.5. Insights into the catalytic mechanism

Previous studies on the pH-dependence of the E. coli TyrA enzyme showed that a catalytic group with a pK_a value of about 6.5 is de­protonated for dehydrogenase activity (Turnbull, Cleland et al., 1991). Subsequent site-directed mutagenesis experiments revealed that this critical catalytic residue is His197 (Christendat et al., 1998). The catalytic mechanism of the oxidative decarboxylation of E. coli TyrA was investigated and suggested a concerted mechanism in which hydride transfer and decarboxylation occur in a concerted manner. It was proposed that His197 provides the driving force for the dehydrogenase reaction by polarizing the 4′-hydroxyl group of prephenate (Christendat et al., 1998). It was also postulated that since the end-product of the reaction is aromatic, polarization of the 4′-OH group is sufficient to lower the energy barrier for the reaction, rather than deprotonation of the 4′-OH group to form a β-keto acid intermediate in a stepwise mechanism. The crystal structure of HinfPDH shows that the N^∊2 atom of His200 (equivalent to His197 in E. coli TyrA) is hydrogen bonded to the 4′-hydroxyl group of tyrosine at a distance of ∼2.6 Å. A hydrogen-bonding network between His200, His248 and Asp249 is observed in which His200 N^δ1 hydrogen bonds to His248 N(H)^∊2 and His248 N(H)^δ1 hydrogen bonds to Asp249 O^δ1. Asp249 is located near the protein surface. As for His200, His248 and Asp249 are highly conserved in TyrA sequences. This hydrogen-bonding network can help to maintain His200 N^∊2 in a deprotonated state. The Ser179 hydrogen bond to the 4′-OH of the bound tyrosine can provide an additional driving force for the reaction by polarizing the 4′-OH group since the equivalent Ser126 in AaeoPDH is critical for catalysis. In addition, HinfPDH and E. coli TyrA share 57% sequence identity in their prephenate dehydro­genase domains and all key active-site residues are conserved. Hence, HinfPDH is likely to adopt a concerted mechanism for dehydro­genase reaction as found for E. coli TyrA. In HinfPDH, tyrosine is bound directly at the catalytic site, suggesting that it acts as a competitive inhibitor.

How TyrA enzymes evolved to be specific for prephenate or arogenate is intriguing because prephenate and arogenate have very similar structures. In HinfPDH, Gln301, Tyr306, Tyr288 and His260 are positioned close to the amino and carboxyl groups of the bound tyrosine and could be involved in substrate specificity (Fig. 6c). Gln301, Tyr306 and Tyr288 are highly conserved in bifunctional TyrA sequences. In other TyrA sequences, Gln301 is replaced by Gly, Ser or Thr. Tyr306 is relatively conserved as Trp in TyrAp and TyrAc, but is Trp, Gly or Val in TyrAa. Tyr288 is located at the N-terminus of α8, where some local structural differences are found between HinfPDH and other TyrA structures. Alignment of TyrA sequences shows a gap of approximately four residues around Tyr288 in the TyrAp, TyrAa and TyrAc sequences. His260 is conserved in TyrA, except for some TyrAas, where it is Val or Gln. Given the structural resemblance between prephenate and arogenate, further experiments to elucidate the exact prephenate-binding mode will advance our understanding of substrate specificity in TyrA enzymes.