The Crystal Structure of Shikimate Dehydrogenase (AroE) Reveals a Unique NADPH Binding Mode

Abstract

Shikimate dehydrogenase catalyzes the NADPH-dependent reversible reduction of 3-dehydroshikimate to shikimate. We report the first X-ray structure of shikimate dehydrogenase from Haemophilus influenzae to 2.4-Å resolution and its complex with NADPH to 1.95-Å resolution. The molecule contains two domains, a catalytic domain with a novel open twisted α/β motif and an NADPH binding domain with a typical Rossmann fold. The enzyme contains a unique glycine-rich P-loop with a conserved sequence motif, GAGGXX, that results in NADPH adopting a nonstandard binding mode with the nicotinamide and ribose moieties disordered in the binary complex. A deep pocket with a narrow entrance between the two domains, containing strictly conserved residues primarily contributed by the catalytic domain, is identified as a potential 3-dehydroshikimate binding pocket. The flexibility of the nicotinamide mononucleotide portion of NADPH may be necessary for the substrate 3-dehydroshikimate to enter the pocket and for the release of the product shikimate.

The shikimate pathway is responsible for the biosynthesis of the aromatic amino acids and other aromatic compounds (1). It is conserved in bacteria, fungi, algae, plants, and parasites (23) but absent in mammals. Deletion of genes involved in shikimic acid biosynthesis results in attenuation of microbial pathogenicity (11). Potent compounds have been identified that inhibit several enzymes of this pathway. Glyphosate, widely used as a herbicide, targets 5-enolpyruvylshikimate 3-phosphate synthase (AroA) (4). By a high-throughput screening approach, aryl bis-sulfonamide derivatives exhibiting good in vitro antimicrobial activity have recently been identified as potent inhibitors of 3-dehydroquinate synthase (AroB) (S. Cockerill, S. Chana, S. Chapman, I. Charles, M. Dolman, E. Goulding, A. Hawkins, D. Madge, P. Maunder, H. Lamb, D. McNamara, L. Rylance, K. Reynolds, G. Spacey, and J. Stables, Abstr. 41st Intersci. Conf. Antimicrob. Agents Chemother., abstr. F-1693, p. 242, 2001). Fluorinated analogues of shikimate were reported to inhibit the growth of Plasmodium falciparum in vitro, presumably through competitive inhibition of the shikimate pathway (19). Thus, components of aromatic acid biosynthesis represent valid targets for development of new broad-spectrum antimicrobial agents (8), herbicides (4), and antiparasitic drugs (19).

Seven enzymes of the shikimate pathway catalyze sequential conversion of erythrose 4-phosphate and phosphoenolpyruvate via shikimate to chorismate (1). Shikimate dehydrogenase (EC 1.1.1.25, AroE) is the fourth enzyme in the pathway. It catalyzes the NADPH-dependent reduction of 3-dehydroshikimate to shikimate. The enzyme exists as a component of the pentafunctional arom enzyme complex in fungi and yeasts (6) and is present as a bifunctional enzyme with 3-dehydroquinate dehydratase (AroD) in plants (5), whereas in bacteria it exists as a single monofunctional enzyme (3).

To provide structural information for rational drug discovery, expression and purification of AroE from Mycobacterium tuberculosis (18) and crystallization of AroE from Escherichia coli (17) have been reported. Here we present the crystal structures of AroE from Haemophilus influenzae both in its apo form and in complex with NADPH. These structures provide three-dimensional templates for future efforts towards structure-based inhibitor design.

MATERIALS AND METHODS

Overexpression and purification.

The open reading frame including aroE was amplified from H. influenzae genomic DNA obtained from the American Type Culture Collection (ATCC 51907D) with the PCR primers CATCTCCATGGATCTTTATGCTGTGTGGGGCA and TGAAGAGCTCTAACATCGCCTTTTTAAGCTGTTCA. The resultant PCR product was digested with NcoI and SacI restriction endonucleases (the sites for each were incorporated during amplification and are underlined in the oligonucleotide sequences above) and ligated into an NcoI-SacI digested kanamycin-resistant arabinose-inducible E. coli expression vector (pSX12). The final construct contained the open reading frame including H. influenzae AroE (identical to the SWISS-PROT reference sequence AROE_HAEIN [P43876]) fused to the sequence ELHHHHHH at the C terminus. To generate selenomethionine-substituted AroE, the expression plasmid was moved into the methionine-auxotrophic E. coli strain DL41 (10). The cells were grown in a 96-well fermentor at 24°C. Recombinant AroE was purified using ProBond nickel-chelating resin (Invitrogen, Carlsbad, Calif.) followed by diafiltration into 150 mM NaCl-25 mM Tris, pH 7.9.

Crystallization, data collection, and processing.

AroE (∼10 mg/ml) samples were screened using Syrrx's nanoliter crystallization, a highly parallel sitting drop submicroliter vapor diffusion crystallization technology (26). This technology represents a significant reduction in the sample requirements for structure determination. To generate AroE:NADPH complex, protein samples were incubated with 5 mM NADPH prior to crystallization. Crystals of selenomethionine-substituted apo-AroE were obtained at 4°C in 100-nl sitting drops containing a 1:1 mixture of protein solution with crystallization buffer [1.0 M sodium citrate, 0.1 M 2-(cyclohexylamino)ethanesulfonic acid (CHES), pH 8.8], whereas the crystals of the AroE:NADPH complex were obtained with a different crystallization buffer (12% polyethylene glycol 8000, 0.15 M calcium acetate, 0.1 M imidazole, pH 7.8). The crystals were cryoprotected and harvested in the presence of the same crystallization buffer containing up to 20% (vol/vol) ethylene glycol. A multiple-wavelength anomalous diffraction (MAD) data set of selenomethionine apo-AroE was collected on beamline 5.0.2 (selenium edge with λ₁ = 0.9786 Å, λ₂ = 0.9792 Å, and λ₃ = 0.9537 Å) on a single crystal, to a resolution of 2.4 Å, and a native data set of AroE:NADPH complex was collected on beamline 5.0.3 to 1.95 Å at Advanced Light Source (ALS) (Table 1), with an Area Detector Systems Corp. charge-coupled device detector. Apo-AroE crystals belong to space group P2₁2₁2₁ with unit cell dimensions of a = 78.2 Å, b = 86.6 Å, and c = 92.9 Å; two monomers per asymmetric unit; and a solvent content of 51.4%. The AroE:NADPH complex crystals belong to space group C222₁ with unit cell dimensions of a = 84.1 Å, b = 82.7 Å, and c = 92.4 Å; one monomer per asymmetric unit; and a solvent content of 46.6%. Reflection data were indexed, integrated, and scaled using MOSFLM/SCALA (7) or HKL2000 (22).

TABLE 1.

Crystallographic data, phasing, and refinement statistics

Type of value	λ1 (peak, max f")	λ2 (edge, min f′)	λ3 (high-energy remote)	AroE:NADPH complex
Crystallographic data
Wavelength (Å)	0.9786	0.9792	0.9537	1.00
Resolution (Å)	2.4	2.4	2.4	1.95
No. of measured reflections	168,448	167,921	172,256	76,173
No. of unique reflections	25,129	25,173	25,244	21,069
Redundancy	6.7	6.7	6.8	3.6
Completeness (%, highest shell)	99.4 (96.8)	99.4 (96.7)	99.8 (99.1)	99.1 (92.0)
Mean I/σI (highest shell)	9.7 (4.2)	9.0 (2.3)	9.5 (2.9)	9.8 (2.0)
Rsyma (%, highest shell)	5.6 (17.4)	6.2 (29.1)	5.9 (25.3)	13.3 (52.5)
Phasing
No. of Se sites	12	12	12
Figure of merit			0.69 (acentric), 0.59 (centric)
Refinement
Rcrystb/Rfreec (%)			22.8/27.6	20.3/24.4
Missing residues			Molecule A: 192-196; B: 191-196	Molecule A: 191-197
No. of protein atoms			4,124	2,055
No. of solvent sites			133	257
No. of cofactor atoms			0	31
cis peptide			Asn9-Pro10, Ser62-Pro63	Asn9-Pro10, Ser62-Pro63
rmsd of bonds (Å)			0.013	0.016
rmsd of angles (°)			1.5	2.0
Avg B factor (Å2)			25.5	10.7
Ramachandran plot (%)
Most favored regions			93.0	92.7
Additional allowed regions			7.0	7.3
Generously allowed regions			0	0
Disallowed regions			0	0

^aR_sym = Σ | I − <I> |/ΣI, where I is the observed intensity and <I> is the average intensity of multiple symmetry-related observations of that reflection.

^bR_cryst = Σ ‖ F_obs | − | F_calc ‖/Σ |F_obs|, where |F_obs| and |F_calc| are the observed and calculated structure factors, respectively.

^cR_free = Σ ‖ F_obs | − | F_calc ‖/Σ |F_obs| for 5% of the data not used at any stage of structural refinement.

Structure determination, model building, and refinement.

The crystal structure of apo-AroE was solved by MAD. Twelve out of 14 expected selenium sites in the asymmetric unit were found by using SHELXD (24). Final refinement of selenium parameters and calculation of phases were performed with the program SHARP (15), giving an overall figure of merit of 0.69 and a readily interpretable electron density map. The model building and substrate modeling were carried out with the program XtalView (20). The crystal structure of AroE:NADPH complex was solved by molecular replacement with the program AMoRe (7) and apo-AroE structure as the search model. Both models were refined using the program REFMAC (7). Poorly defined residues have been omitted from the models. The final model has good stereochemistry, as determined by the program PROCHECK (7). The figures were generated using MOLSCRIPT (14), BOBSCRIPT (9), GRASP (21), and GLR (L. Esser, personal communication) and rendered with POV-RAY (www.povray.org).

Coordinates.

Coordinates have been deposited in the Protein Data Bank (Protein Data Bank accession codes: AroE, 1P74; AroE:NADPH, 1P77).

RESULTS AND DISCUSSION

Structure determination and overall structure.

The crystal structure of H. influenzae apo-AroE was determined by the MAD phasing method and refined to 2.4-Å resolution to an R factor and R_free of 22.8 and 27.6%, respectively (Table 1). The final model of the apo enzyme includes 533 amino acids (residues 1 to 191 and 197 to 272 for molecule A and residues 1 to 190 and 197 to 272 for molecule B) and 133 water molecules. Crystals of the binary complex of AroE with its cofactor NADPH were obtained by cocrystallization under different conditions. The complex structure was solved by molecular replacement, and the final model, refined to 1.95-Å resolution with an R factor and R_free of 19.4 and 23.8%, respectively, contains 265 amino acids (residues 1 to 190 and 198 to 272), an NADPH, an acetate molecule from the crystallization solution, and 256 water molecules. The acetate molecule coordinates the molecules in the lattice via two salt bridges and a hydrogen bonding interaction with three symmetry-related AroE molecules. This may function to improve the diffraction quality of the crystals.

The three-dimensional structure of AroE is illustrated in Fig. 1A. The 272 residues in the AroE molecule form two structural domains: a catalytic domain and an NADPH binding domain. The NADPH binding domain is composed of a contiguous polypeptide chain (residues 119 to 238). This domain is symmetrically built from two halves with identical topology of β-α-β-α-β, together forming a single parallel β-sheet flanked by α-helices. The domain structure is very similar to those of NAD(P)H binding domains in other dehydrogenases (16), despite the absence of amino acid sequence homology.

FIG. 1.

(A) Ribbon diagram of the AroE:NADPH complex structure. The NADPH binding domain is shown in green at the top, the catalytic domain is shown in red at the bottom, and NADPH is shown in a ball-and-stick representation. (B) Alignment of 12 primary sequences of AroE from 10 bacteria and two archaea. Above the alignment is shown the secondary structure assignment based on the H. influenzae AroE crystal structure; the catalytic domain is shown in red, and the NADPH binding domain is shown in green. The boxes, arrows, and lines correspond to α-helices, β-strands, and loops, respectively. The two cis-peptide prolines are strictly conserved in all 12 sequences and are shown in green and marked with an asterisk. The residues related to NADPH binding are shown in blue. The residues of the glycine-rich P-loop are boxed. The residues forming the potential 3-dehydroshikimate binding pocket are shown in orange and are highly conserved, although not invariant. The sequences were drawn from SWISS-PROT and represent the organisms indicated: AroE_HAEIN, H. influenzae; AroE_HELPY, Helicobacter pylori; AroE_AQUAE, Aquifex aeolicus; AroE_BACSU, Bacillus subtilis; AroE_STAAM, Staphylococcus aureus; AroE_STRPN, Streptococcus pneumoniae; AroE_THEMA, Thermotoga maritima; AroE_ECOLI, E. coli; AroE_PSEAE, Pseudomonas aeruginosa; AroE_NEIGO, Neisseria gonorrhoeae; AroE_METJA, Methanococcus jannaschii; AroE_PYRFU, Pyrococcus furiosus. (C) Hydrophobic interactions between the catalytic domain and NADPH binding domain. The residues involved in interactions are shown in ball-and-stick format. (D) The two molecules in the asymmetric unit of the apo-AroE structure are superimposed on the molecule of AroE:NADPH complex structure. Red, AroE molecule in complex with NADPH; green, AroE molecule A in apo structure; blue, molecule B in apo structure.

The catalytic domain, which is comprised of the first 118 residues and last 34 residues of AroE, adopts a novel open twisted α/β structure with a six-strand central β-sheet packing against three α-helices (α2, α3, and α4) on one side and four (α1, α5, α10, and α11) on another side. The central β-sheet is arranged as 2-1-3-5-6-4, with all strands parallel except strand 5. The secondary structure motifs for catalytic domain are arranged as β1α1β2α2β3α3β4α4β5β6α5•α10α11. Two cis peptides, Asn9-Pro10 and Ser62-Pro63, are observed in the AroE structure. The two cis-peptide prolines are strictly conserved in AroE sequences (Fig. 1B), and both exist in a sharp turn linking a β-strand and a following α-helix, suggesting that they are structurally conserved.

The catalytic domain and the NADPH binding domain mainly pack through interdigitated hydrophobic side chains (Fig. 1C). Helix α5 interacts with the strands β11 and β10 of the central β-sheet and helix α6. Helix α4 interacts with helices α6 and α7. The hydrophobic core, formed between the two domains, locks the positions of the two domains relative to one another. Overall, the three independent AroE molecules whose structures were determined (two apo and one NADPH bound) are very similar (Fig. 1D). The main chain atoms superimpose with a root mean square deviation (rmsd) of 0.84, 1.22, and 1.04 Å between the two independent molecules of apo-AroE, molecule A of the apo enzyme and the NADPH complex, and molecule B of the apo enzyme and the NADPH complex, respectively. The conformation of the catalytic domain is more rigid than that of the NADPH binding domain. The rmsd's are 0.55, 0.75, and 0.68 Å between the catalytic domains and 0.98, 1.24, and 0.81 Å between the NADPH binding domains of AroE subunits. The major backbone differences between the apo and NADPH-bound AroE occur in the glycine-rich P-loop linking β7 and α6, the loop linking β10 and α8, the loop linking β11 and α9, helix α9, and helix α11 (Fig. 1D). The conformational changes occurring in the glycine-rich P-loop and in the loop linking β10 and α8 are a result of NADPH binding.

Unusual NADPH binding mode.

A systematic survey of NAD(P) protein complexes showed that NADP interacts with a variety of proteins more variably than does NAD (2). Nevertheless, typically each half of the NAD(P)H binding domain binds one nucleotide moiety of the dinucleotide NAD(P). In contrast, NADPH in the crystal structure of the AroE:NADPH complex was found to adopt an unusual binding mode. Only the adenine, ribose, 2′-phosphate, and 5′-diphosphate of NADPH were observed to interact with the protein. The nicotinamide and ribose moieties in the nicotinamide mononucleotide (NMN) portion of NADPH are disordered in the crystal structure. Lacking clear electron density, they were omitted from the model (Fig. 2A).

FIG.2.

(A) Electron density map of NADPH in the AroE:NADPH complex structure. The map was calculated in CCP4 with SIGMAA-weighted coefficients and model-derived phases for diffraction data between 20 and 1.95 Å, contoured at 1.0 σ. (B) Interactions between NADPH and AroE in the complex. (C) Overlay of NADPH modeled in a more canonical orientation (purple), as exemplified by NADPH-dependent alcohol dehydrogenase (13), onto the structure of the observed NADPH (gold) bound to AroE. Note the 90° turn in the phosphates relative to the more common orientation, radically repositioning the NMN moiety.

An extensive array of interactions is observed between AroE and NADPH (Fig. 2B). First, multiple interactions with the 2′-phosphate provide an anchor for the 2′-phosphate-AMP half of NADPH, defining the coenzyme specificity. The negatively charged 2′-phosphate moiety forms salt bridges with two positively charged side chains (R150 and K154) and hydrogen bonds with N149 and T151. In addition, the N149 side chain carbonyl also accepts a proton from the 3′-hydroxyl of the adenine ribose. The δ-guanido moiety of R150 stacks with adenine on one side, and A190 forms hydrophobic interactions with adenine on the other side. Among these residues, three of them (N149, R150, and T151) are strictly conserved among AroE sequences (Fig. 1B), suggesting that they may play important roles in coenzyme selectivity.

Second, AroE contains a unique glycine-rich P-loop with a conserved sequence motif GAGGXX (where X is any amino acid), in contrast to the typical GXGXXG motif observed in other dinucleotide binding proteins containing Rossmann folds, such as the short-chain dehydrogenase family (12). The glycine-rich P-loop motif forms a tight turn between the end of the first β-strand (β7) and the beginning of the so-called “dinucleotide binding helix” (α6) in the Rossmann fold. In the AroE:NADPH complex structure, the glycine-rich P-loop makes four hydrogen bonds with the NADPH. The main chain N-H of the P2 and P3 residues form two hydrogen bonds with the 3′-hydroxyl of the adenine ribose, and the main chain N-H of the P4 and P5 residues form two hydrogen bonds with the second phosphate (Fig. 2B). One consequence of the geometry formed by the unique glycine-rich P-loop in AroE is that the NMN phosphate makes a 90° bend, whereas in the typical NADPH binding geometry the phosphate is nearly linear (Fig. 2C). The substitution of glycine at position P4 in AroE allows the NMN half of NADPH to adopt this orientation, which is precluded in other dinucleotide binding proteins by the presence of a larger side chain at P4.

Potential 3-dehydroshikimate binding site.

The nature of the area around the NADPH indicates a potential site for 3-dehydroshikimate binding. An obvious pocket is observed with a narrow entrance between the two domains. The pocket is solvent rich, and a total of 30 water molecules are modeled inside and nearby the pocket in the 1.95-Å resolution AroE:NADPH complex structure. A groove between α helices 1, 5, and 10 and β-strands 1, 3, and 5 in the catalytic domain forms the bottom of the pocket. The side wall of the pocket is formed from two β-hairpin loops (between β1 and α1 and between β3 and α3) from the catalytic domain on one side and another β-hairpin loop between β11 and α9 from the NADPH binding domain on the other side (Fig. 1A and 3A). The pocket is narrow, with a width of 8 Å, and is approximately 10 Å deep. Sequence alignment analysis of AroE orthologs (Fig. 1B) shows that most of the residues exposed to the pocket (V6, S14, K15, S16, I19, N59, T61, K65, N86, N100, D102, M241, L242, and Q245) are strictly conserved, suggesting that it is a good candidate for the 3-dehydroshikimate binding pocket.

FIG. 3.

(A) The potential 3-dehydroshikimate binding pocket. (B) An approximate model of docking of 3-dehydroshikimate to AroE in the proposed pocket. (C) An approximate model of “ordered” NADPH and 3-dehydroshikimate. Protein surface charge distribution was calculated and displayed by the program GRASP; potentials less than −10 kT, neutral, and greater than 10 kT are displayed in red, white, and blue, respectively. The orientation of the molecule is related to that shown in Fig. 1A.

To help comprehend the unique binding mode of NADPH, we carried out a preliminary study of docking of a 3-dehydroshikimate onto the structure (Fig. 3B). Attempts to orient the substrate in the pocket were guided by three criteria. First, the 3-dehydroshikimate was docked to form as many hydrogen bonds as possible. Second, the 3-carbonyl group of 3-dehydroshikimate should be accessible to the nicotinamide moiety of NADPH to allow the dehydrogenation reaction. Third, the negatively charged carboxylate group of 3-dehydroshikimate would prefer to form a salt bridge with a positively charged amino acid side chain, like that of arginine or lysine. This has been observed in the crystal structure of 5-enolpyruvylshikimate 3-phosphate synthase complexed with shikimate 3-phosphate (25). Two conserved lysine residues in the potential 3-dehydroshikimate binding pocket, K15 and K65, were observed. However, only the side chain amino group of K65 faces inside the pocket. Thus, it was assumed that K65 stabilizes the carboxylate group of 3-dehydroshikimate.

In docking the 3-dehydroshikimate, it became apparent that, in the dehydrogenation reaction, the nicotinamide moiety of NADPH is located somewhere close to the entrance of the pocket, as shown in Fig. 3C. Given the narrow entrance of the pocket, this conformation may block, or at least interfere with, entry of the substrate 3-dehydroshikimate to the pocket and with release of the product shikimate. Thus, the flexibility of the NMN portion of NADPH may be necessary for the dehydrogenation reaction. This may explain why the NMN part of NADPH is disordered in the AroE:NADPH complex structure.

The structure of H. influenzae AroE provides the first three-dimensional view of this enzyme. It reveals that AroE is a monomer composed of two domains. The catalytic domain is arranged as a novel fold. The NADPH binding domain has a typical Rossmann fold and a unique glycine-rich P-loop with a conserved sequence motif of GAGGXX. The complex with NADPH allows the identification of residues involved in cofactor binding and specificity. NADPH adopts a unique binding mode with the nicotinamide and ribose moieties disordered in the binary complex. A deep pocket between the two domains lined with strictly conserved residues is identified as the probable 3-dehydroshikimate binding pocket.

All of the enzymes that make up this pathway are potentially targets for the design of novel drugs directed against pathogenic bacteria and the parasites. The structures of AroE are the first steps in efforts to use structural templates for the synthesis of effective inhibitors of the shikimate pathway for biosynthesis of aromatic compounds. Further experiments suggested by the structures may enable the development of new broad-spectrum antimicrobial agents, herbicides, and antiparasitic drugs.

ADDENDUM

While this paper was under review, two relevant papers were published online, ahead of print, by the Journal of Biological Chemistry: J. Benach, I. Lee, W. Edstrom, A. P. Kuzin, Y. Chiang, T. B. Acton, G. T. Montelione, and J. F. Hunt, The 2.3-Å crystal structure of the shikimate 5-dehydrogenase orthologue YdiB from Escherichia coli suggests a novel catalytic environment for an NAD-dependent dehydrogenase, J. Biol. Chem. 278:19176-19182, 2003, and G. Michel, A. W. Roszak, V. Sauve, J. McLean, A. Matte, J. R. Coggins, M. Cygler, and A. J. Lapthorn, Structures of shikimate dehydrogenase AroE and its paralog YdiB: a common structural framework for different activities, J. Biol. Chem. 278:19463-19472, 2003. The structure of E. coli shikimate dehydrogenase described by Michel et al. reveals a fully ordered NADPH molecule in the active site. The structures of E. coli YdiB in both papers reveal a fully ordered NADH. These results agree well with our model presented in Fig. 3C