Biochemical Journal

Research article

Structural investigation of inhibitor designs targeting 3-dehydroquinate dehydratase from the shikimate pathway of Mycobacterium tuberculosis

Marcio V. B. Dias, William C. Snee, Karen M. Bromfield, Richard J. Payne, Satheesh K. Palaninathan, Alessio Ciulli, Nigel I. Howard, Chris Abell, James C. Sacchettini, Tom L. Blundell


The shikimate pathway is essential in Mycobacterium tuberculosis and its absence from humans makes the enzymes of this pathway potential drug targets. In the present paper, we provide structural insights into ligand and inhibitor binding to 3-dehydroquinate dehydratase (dehydroquinase) from M. tuberculosis (MtDHQase), the third enzyme of the shikimate pathway. The enzyme has been crystallized in complex with its reaction product, 3-dehydroshikimate, and with six different competitive inhibitors. The inhibitor 2,3-anhydroquinate mimics the flattened enol/enolate reaction intermediate and serves as an anchor molecule for four of the inhibitors investigated. MtDHQase also forms a complex with citrazinic acid, a planar analogue of the reaction product. The structure of MtDHQase in complex with a 2,3-anhydroquinate moiety attached to a biaryl group shows that this group extends to an active-site subpocket inducing significant structural rearrangement. The flexible extensions of inhibitors designed to form π-stacking interactions with the catalytic Tyr24 have been investigated. The high-resolution crystal structures of the MtDHQase complexes provide structural evidence for the role of the loop residues 19–24 in MtDHQase ligand binding and catalytic mechanism and provide a rationale for the design and efficacy of inhibitors.

  • 3-dehydroquinate dehydratase (dehydroquinase)
  • drug discovery
  • inhibitor design
  • Mycobacterium tuberculosis
  • shikimate pathway
  • transition state analogue


One strategy for the development of new therapies against pathogenic bacteria such as Mycobacterium tuberculosis is to target essential biosynthetic pathways of the micro-organism that are absent from humans, such as the shikimate pathway. The shikimate pathway produces chorismate, an important precursor of aromatic compounds in bacteria, fungi, plants and apicomplexan parasites [14]. This pathway comprises seven different enzymes, each of which catalyses a separate step of the pathway that converts erythrose-4-phosphate and phosphoenol pyruvate into chorismate [1,5]. Chorismate is the substrate for five branching pathways involved in the production of menaquinones, siderophores, aromatic amino acids (phenylalanine, tyrosine and tryptophan), vitamins E and K, p-aminobenzoic acid and other aromatic compounds [3]. Studies of gene disruption have shown that the shikimate pathway is essential for growth of M. tuberculosis [6].

The third enzyme of the shikimate pathway is DHQase [3-dehydroquinate dehydratase (dehydroquinase)], which is the product of the aroD gene. There are two forms of DHQase (type I and type II) with different structures and mechanisms, both of which catalyse the reversible conversion of 3-dehydroquinate into 3-dehydroshikimate [7]. DHQase functions in two metabolic pathways: the catabolic quinate pathway, enabling certain organisms to convert quinate into protocatechuate for use as a carbon source, and the biosynthetic shikimate pathway [810]. Type I DHQase exists as a heat-labile homodimer and is involved only in the biosynthetic pathway. Its mechanism involves a covalent iminium intermediate to catalyse the dehydration of 3-dehydroquinate through a syn elimination [11]. Type II DHQase exists as a heat-stable homododecamer that can function in both biosynthetic and catabolic pathways, probably involving dehydration through an enol/enolate intermediate [5] (Figure 1). The catalysis of type II DHQase involves an anti elimination of water processed through a stepwise E1CB mechanism. A base-catalysed abstraction of the axial proton at C2 is required for type II DHQase to form the enolate intermediate, and, in a second step, the hydroxy group of the C1 position is removed by acid catalysis to form the product. A conserved tyrosine residue of the active site is responsible for removing the pro-S proton in the first step of the reaction and also a conserved histidine residue finalizes the reaction acting as proton donor in the acid-catalysed reaction [7] (Figure 1).

Figure 1 Catalytic reaction mechanism for type II DHQase

In the first step of the reaction of Type II DHQase, a conserved tyrosine residue acts to remove the pro-S proton of the C2 to form the enolate intermediate. Simultaneously, an asparagine residue holds a conserved water molecule in the correct orientation to stabilize the intermediate, and, finally, a conserved histidine residue acts as a proton donor to catalyse the elimination of the C1 hydroxy group and liberate a water molecule producing the 3-dehydroshikimate.

M. tuberculosis uses the type II DHQase in a biosynthetic role. The crystallographic structures of type II DHQase from M. tuberculosis (MtDHQase) [7] and other bacteria [12,13] reveal a dodecamer formed from a tetramer of trimers with 23 symmetry and the trimer being the minimal active oligomeric form [7]. The parallel β-sheet of each subunit has strands in the following order: 2, 1, 3, 4 and 5, with two α-helices on each side of the sheet as in the flavodoxin type α/β fold [7,14]. The active site in type II DHQases is located in a cleft formed near the C-terminal regions of strands β1 and β3 of the parallel β-sheet, which is common among proteins with the flavodoxin-like fold [14]. Two of the key residues, Arg19 and Tyr24 in MtDHQase, are conserved in type II DHQases from other organisms [see Supplementary Figures S1 and S2 at; Arg23 and Tyr28 in Streptomyces coelicolor DHQase (ScDHQase), and Arg17 and Tyr22 in Helicobacter pylori DHQase (HpDHQase)] and have been identified by chemical modification and site-directed mutagenesis as being essential for DHQase enzyme activity [15]. Both of these residues are located on a flexible loop (residues 19–24 in MtDHQase) that is completely disordered in the apo-DHQase structure (PDB code 2DHQ) [7]. Structural studies of ScDHQase [12,1618] and HpDHQase [13,1920] were able to capture the full loop region in the presence of ligands and demonstrate that ligand binding causes the flexible-loop residues to form a lid that closes over the active site.

Structures of MtDHQase have been solved with and without ligands, including structures with the reaction intermediate analogues 2,3-anhydroquinate (2 in Figure 2; PDB code 1H0R) and 3-hydroxyimino-quinic acid (PDB code 1H0S), but only recently has the full closure of the flexible catalytic loop residues 19–24 been reported [20].

Figure 2 MtDQHase ligands

1, 3-dehydroshikimic acid (the cyclohexene carbon atom numbering progresses counter-clockwise from C1; C1, C3 and C5 of ligand 1 are labelled); 2, 2,3-anhydroquinate; 3, (1S,4R,5R)-3-(3-benzoylphenyl)-1,4,5-trihydroxycyclohex-2-enecarboxylic acid; 4, (1R,4R,5R)-1,4,5-trihydroxyl-3-(2-phenylcarbamoyl-vinyl)-cyclo-hex-2-enecarboxylic acid; 5, (1R,4R,5R)-1,4,5-trihydroxy-3-[3-phenoxy-prop-(E)-enyl]-cyclohex-2-enecarboxylic acid; 6, (1R,4R,5R)-3-(t-butylcarbamoyl)-1,4,5-trihydroxycyclohex-2-enecarboxylic acid; 7, citrazinic acid represented in both tautomeric forms.

This flexible-loop region is essential for substrate binding and catalysis, and has been an important element for DHQase inhibitor design. More importantly, this region plays a major role in the formation of a subpocket located adjacent to the enzyme active site of ScDHQase. The formation of this subpocket greatly increased the druggable space for ScDHQase and was the basis for the design of several nanomolar inhibitors [17,18,2127]. The inhibitors designed to target this subpocket of ScDHQase also showed potent inhibition of MtDHQase and many of them show ~10-fold or greater differences in inhibition between the two enzymes [18,21,22,24]. These studies strongly suggest a general conservation of features between the active sites of these two enzymes, but also that there are structural differences between them that need to be characterized.

In the present paper, we provide structures of MtDHQase that show all the residues in the active site, including the flexible-loop residues 19–24, in complex with the enzymatic reaction product 3-dehydroshikimate, and six competitive inhibitors. The in-hibitors include 2,3-anhydroquinate, an analogue based on the enol/enolate reaction intermediate, four compounds that extend from the 2,3-anhydroquinate template and citrazinic acid, whose planarity represents a novel class of MtDHQase inhibitor. The binary complex structures provide structural details for the lid-closure mechanism in MtDHQase, and also provide structural evidence and characterization of a subpocket located adjacent to the MtDHQase active site and its influence on MtDHQase inhibitor binding.


Cloning, overexpression and purification of MtDHQase

The aroD gene from M. tuberculosis [Rv2537c; UniProtKB/Swiss-Prot accession number P0A4Z6 (AROQ_MYCTU)] was cloned into either the pET28a with a thrombin cleavage site or the pET28b vector with a TEV (tobacco etch virus) cleavage site (modified from the original) after the N-terminal His6 tag. Both plasmids containing the aroD gene were transformed into BL21(DE3) competent Escherichia coli cells (Novagen) by heat shock. The cells were grown in LB (Luria–Bertani) or 2YT [1.6% (w/v) tryptone/1% (w/v) yeast extract/0.5% NaCl] medium at 37 °C until a D600 of 0.6 was reached and then induced with 1 mM IPTG (isopropyl β-D-thiogalactopyranoside) for 4–6 h at 37 °C. The cells were then harvested by centrifugation at 5010 g for 30 min. The pellet was suspended in approximately 30 ml of buffer composed of 50 mM Tris/HCl (pH 7.8) and 250 mM NaCl and was supplemented with one Roche Inhibitor Cocktail tablet (EDTA-free). The cells were then lysed by either sonication or French press and clarified by centrifugation at 21000 g for 40 min. The 0.2-μm-filtered supernatant was loaded on to a nickel column and the pure His6-tagged protein was eluted with a gradient concentration from 20 to 500 mM imidazole. MtDHQase structures with inhibitors 1, 6 and 7 (see Figure 2) were obtained from purified His6-tagged protein expressed using the pET28b vector. MtDHQase expressed using the pET28a vector was used for structures with inhibitors 25 (see Figure 2). The crystals for the complexes of MtDHQase with inhibitors 25 were obtained after cleavage of the His6 tag. To remove the His6 tag, MtDHQase was incubated overnight at 4 °C with thrombin (restriction grade; Invitrogen). The tag-free MtDHQase was purified further by size-exclusion chromatography using a Superdex 75 26/60 column (GE Healthcare) with a buffer containing 50 mM Tris/HCl (pH 7.8) and 250 mM NaCl.

MtDHQase ligands

The ligands (Figure 2) are 1, 3-dehydroshikimate; 2, 2,3 anhydroquinate; 3, (1S,4R,5R)-3-(3-benzoylphenyl)-1,4,5-trihydroxycyclohex-2-enecarboxylic acid; 4, (1R,4R,5R)-1,4,5-trihydroxyl-3-(2-phenylcarbamoyl-vinyl)-cyclo-hex-2-enecarboxylic acid); 5, (1R,4R,5R)-1,4,5-trihydroxy-3-[3-phenoxy-prop-(E)-enyl]cyclohex-2-enecarboxylic acid; 6, (1R,4R,5R)-3-(t-butylcarbamoyl)-1,4,5-trihydroxycyclohex-2-enecarboxylic acid; and 7, citrazinic acid represented in both tautomeric forms. The syntheses of inhibitors 26 were performed according to protocols published previously [18,21,22,26]. Inhibitor 7 was purchased from Sigma–Aldrich.

Crystallization with ligands, data collection and processing

MtDHQase expressed using the pET28a vector was concentrated to 10 or 40 mg/ml in a buffer composed of 50 mM Tris/HCl (pH 7.8) and 250 mM NaCl, and MtDHQase expressed using the pET28b vector was concentrated to 10 mg/ml in a buffer composed of 50 mM Tris/HCl (pH 7.5), 100 mM NaCl, 1 mM EDTA and 1 mM DTT (dithiothreitol) using Amicon Ultra 5 kDa molecular-mass cut-off protein concentrators. MtDHQase in complex with different compounds was crystallized using different conditions and techniques due to the difficulty of reproducing the crystals. The crystallization protocols and data collection/processing details for each crystal structure are described in Supplementary Table S1 at

Structure determination and refinement

The MtDHQase–1 structure was solved by molecular replacement using MOLREP [28] implemented in CCP4 [29] with the protein atomic co-ordinates for MtDHQase (PDB code 2DHQ). Non-protein atoms were removed from MtDHQase–1, which was then used as a molecular replacement model for structures MtDHQase–2, 3, 4, 6 and 7. The MtDHQase–5 structure was solved by molecular replacement using the MtDHQase–2 structure (PDB code 1H0R). Refinement was performed using REFMAC 5.2 from the CCP4 suite [29,30]. TLS (translation, liberation, screw-rotation) refinement was applied to structures of MtDHQase–1 and 7 (files were generated by TLSMD server) [31]. Visual inspection and manual rebuilding were performed using the programs XtalView/Xfit [32] and Coot 0.3.1 [33]. The water molecules were added manually and automatically using the program XtalView/Xfit [32] or with Coot 0.3.1 [33] and checked on the basis of B-factor values. The stereochemistry was checked using PROCHECK [34]. The Figures were prepared using UCSF Chimera [35], Raster3D [36], XtalView/Xfit [32] and PyMOL (

MtDHQase enzyme assay

The enzyme activity was assayed by monitoring the formation of the product 3-dehydroshikimate. The initial rate of the reaction was measured by the increase in absorbance at 234 nm, from the formation of the enone-carboxylate chromophore of 3-dehydroshikimate (ϵ=1.2 × 104 M−1·cm−1). The assays were performed in duplicate at 25 °C in 50 mM Tris/HCl (pH 7.0). A final enzyme concentration of 102 nM was used. The assay was initiated by the addition of the substrate (3-dehydroquinate), after incubating the buffer, inhibitor and enzyme at 25 °C for 2 min.

Kinetic parameters (Km and kcat) were obtained by measuring the initial rates of reaction over a range of substrate concentrations (typically 0.25–10 Km). The data were fitted to the Michaelis–Menten equation using least-squares fitting in GraFit (Erithacus Software). The values of Km and Vmax were determined using this software, and kcat was calculated from the latter value and the total enzyme concentration in the assay.

The kinetic data for inhibition studies were obtained by measuring the initial rates of reaction over a range of four to five inhibitor concentrations at four to five different substrate concentrations (0.5–5 Km). The inhibition constants (Ki) and the S.D. values associated with these values were determined using least-squares fitting using GraFit. GraFit was also used to carry out an F-test statistical analysis on the data, to confirm that the data satisfied a competitive inhibition model.

ITC (isothermal titration calorimetry)

ITC experiments were performed using a MicroCal VP-ITC instrument and all data were analysed with the software implemented in Origin (version 7). In all titrations, His6–MtDHQase was at a concentration of 80–160 μM and was buffered in 100 mM Tris/HCl (pH 8.0) and 50–100 mM NaCl, in the presence of 5% (v/v) DMSO. Titrations of inhibitor 7 (1.6 mM) and (3R,5R,6R)-3,5,6-trihydroxycyclohex-1-ene-1,3-dicarboxylic acid (780 μM) were performed at 30 and 15 °C respectively.

In a typical experiment, 19 injections of 15 μl were made at 4 min intervals from a 300 μl syringe rotating at 300 rev./min and loaded with inhibitor solution. In all titrations, an initial injection of 2 μl of inhibitor was made and the corresponding data discarded during data analysis. Control titrations of inhibitor to buffer were performed and subtracted from the inhibitor to protein titrations. The thermodynamic parameters were obtained by fitting the data to a single-site binding model.

STD (saturation transfer difference) NMR spectroscopy

1H-NMR spectroscopic experiments were performed at 278 K on a 700 MHz Bruker NMR spectrometer equipped with a 5 mm triple TXI cryoprobe with z gradients. STD experiments [37] employed a 40 ms selective Gaussian 180° shaped pulse at a frequency alternating between ‘on resonance’ (1.0 p.p.m.) and ‘off resonance’ (~80 p.p.m.) after every scan. Water suppression was achieved by using a W5 Watergate gradient spin-echo pulse sequence [38]. The resulting spectra were analysed with TopSpin. A sample of 0.35 mM citrazinic acid (7) was prepared in 25 mM Tris/HCl (pH 7.8), 20 mM NaCl and 10% (v/v) 2H2O, in the absence of protein, and the presence of 10 μM His6–DHQase to a total volume of 200 μl. The inhibitor (3R,5R,6R)-3,5,6-trihydroxycyclohex-1-ene-1,3-dicarboxylic acid was added to a final concentration of 400 μM to displace 7 and show specific binding in the active site. TSP {(trimethylsilyl)-[2H4]propionic acid} (20 μM) was present for calibration purposes.


The high-resolution crystal structures of MtDHQase in complex with the reaction product, 3-dehydroshikimate (1) and six competitive inhibitors (27) shown in Figure 3 define key interactions associated with the active-site residues including the flexible-loop residues 19–24 (Figures 3–7). The inhibitor 2,3-anhydroquinate (2) was designed to mimic the flattened enol/enolate intermediate in the reaction mechanism and inhibitors 36 were evolved from 2, extending to occupy a subpocket adjacent to the active site (Figures 3B–3E), based on the ScDHQase–2–glycerol structure (PDB code 1GU1) [12]. The inhibitor citrazinic acid (7) has similar substituents to the product, but lacks the ring puckering. The MtDHQase–1, 4, 6 and 7 and MtDHQase–2 and 3 structures were solved in space groups P21 and P1 respectively and both have two dodecamers in the asymmetric unit. No significant differences were observed in the three-dimensional structure and oligomeric state for MtDHQase in complex with these ligands compared with the previously reported apo-MtDHQase structure (PDB code 2DHQ) [7]. Table 1 provides the resolution and statistics of refinement of the structures (complete statistics of the structures can be seen in Supplementary Table S2 at

Figure 3 Surface representation of MtDHQase with close-up view of the active-site binding of 3-dehydroshikimate and six inhibitors

(A) 3-Dehydroshikimate (1); (B) inhibitor 2 and glycerol; (C) inhibitor 3; (D) inhibitor 4; (E) inhibitor 5; (F) inhibitor 6; (G) inhibitor 7. Electron density 2FoFc contour for the inhibitors are shown.

Figure 4 Ordering of the catalytic loop in the MtDHQase–1 structure

(A) Close-up comparison of MtDHQase–1 with the incomplete structure of apo-MtDHQase (PDB code 2DHQ) emphasizing the closure of the active site by flexible-loop residues 19–24 upon ligand binding and showing the location of the subpocket located adjacent to the active site observed in the MtDHQase–2 structure. Subunit G of MtDHQase–1 (residues 15–28 coloured yellow and ligand 1 coloured grey) is superimposed individually on the apo-MtDHQase subunit (residues 15–28 are coloured magenta) using UCSF Chimera. The location of the glycerol molecule (cyan) from MtDHQase–2 is shown based on the individual superposition of MtDHQase–2 subunit B on MtDHQase–1 subunit G; all other MtDHQase–2 atoms were removed for clarity. Inset: surface representation of the MtDHQase homododecamer with a black box encompassing one of the 12 enzyme active sites. Each MtDHQase subunit is coloured individually. The boxed active site is represented in the close-up view in Figure 3(A). (B) The 2.52 Å resolution electron-density map of MtDHQase–1 showing ligand 1 (green), loop residues 19–24 and His101 from subunit G, and Asp88* from subunit F (grey). The 2FoFc (1σ; blue) and FoFc (3σ; red) electron-density maps were calculated after omitting ligand 1 and residues 19–24, 101 and 88*. The blob feature in XtalView [32] has been applied to limit the electron-density display to within 1.5 Å of the subunit residues and the final Figure is rendered with Raster3D [36]. (C) Ligand-binding interactions of the MtDHQase–1 binary complex structure. 1 (green), loop residues 19–24 (yellow) and additional important MtDHQase residues (grey) involved in the binding of 1 are displayed as sticks; hydrogen-bond interactions between MtDHQase and 1 are indicated with blue lines and important direct and water-mediated intra-protein loop residue [1621] hydrogen bonds are indicated with black lines.

Figure 5 Residues forming hydrogen-bonding and hydrophobic interactions involved in the binding of the different ligands in the active site of MtDHQase

(A) 1; (B) 2; (C) 3; (D) 4; (E) 5. Different colours for residues are used for different subunits and the compounds are represented in green. The Figure for the complexes between MtDHQase and 6 and MtDHQase and 7 are not shown. The complex with MtDHQase and compound 6 has most of the flexible active-site loop disordered, and the complex with MtDHQase and 7 is very similar to the ones observed for the complexes either between MtDHQase and 1 or between MtDHQase and 2.

Figure 6 Glycerol-binding site and effect of different ligands on the glycerol pocket

(A) Glycerol-binding site and interactions of MtDHQase–2. Glycerol and inhibitor 2 (green) and important MtDHQase residues (yellow indicates MtDHQase subunit B, grey indicates neighbouring MtDHQase subunit) involved in the binding of glycerol are displayed as sticks; hydrogen-bond interactions are indicated as broken lines. Electron density 2FoFc contour for 2 and glycerol. The binding of glycerol to this pocket conserves most of the interactions formed by the guanidinium group of Arg19 and makes additional interactions with its three hydroxy groups. The glycerol molecule bound to the active-site flexible-loop pocket forms direct hydrogen-bonding interactions with the main-chain atoms of Asn12, Arg15, Arg18 and Arg19, and forms water-mediated hydrogen bonds with the side-chain atoms of Asn12 and Asp88* as well as with the main-chain atoms of Gly17, Arg18, Arg19 and Gly25. (B) Superposition of three different subunits of MtDHQase–3 (D, E and G) and the position of glycerol from MtDHQase–2 subunit B (yellow). The Figure shows the different ways inhibitor 5 binds in MtDHQase. (C) Individual superposition of MtDHQase–4 subunit A (yellow) on MtDHQase–5 subunit A (blue) with a close-up view of inhibitors and Tyr24. (D) Individual superposition of MtDHQase–6 subunit E (green) on MtDHQase–5 subunit E (blue) with a close-up view of inhibitors and key MtDHQase residues involved in inhibitor binding.

Figure 7 Binding of citrazinic acid (7) to MtDHQase

(A) Inhibitor-binding interactions of the MtDHQase–7 (green) and important MtDHQase residues (grey); hydrogen-bond interactions are indicated with blue lines. Loop residues 19–24 are coloured yellow. (B) Individual superposition of MtDHQase–7 subunit G (cyan) on MtDHQase–1 subunit G (grey) with a close-up view of the ligand and active-site residue positions/movements.

View this table:
Table 1 Crystallographic statistics of MtDHQase–1, 2, 3, 4, 5, 6 and 7 structures

Rcryst=Σ‖Fobs|−|Fcalc‖/Σ|Fobs|, Fobs and Fcalc are observed and calculated structure factor amplitudes. Rfree was calculated using a random subset of the data (5%) excluded from the refinement.

Ligand-induced stabilization of MtDHQase catalytic loop residues

The type II DHQase-catalysed conversion of 3-dehydroquinate into 3-dehydroshikimate requires the abstraction of the pro-S hydrogen from the C2 carbon atom by the negatively charged phenolate group of Tyr24 in M. tuberculosis and the overall anti-elimination of water by a stepwise E1CB mechanism resulting in the C1–C2 conjugation observed in 1 (Figure 1) [12]. The binding of 1 and substrate or analogues of substrates are assumed to stabilize the flexible active-site loop residues 19–24 into a catalytic conformation (Figures 3 and 4), even though this has not been observed with previous MtDHQase binary complex structures (PDB codes 1H0R and 1H0S). A recently reported MtDHQase structure (PDB code 2XB8) has captured the full loop region extended into a conformation similar to the crystal structures reported in the present paper (see below) [20].

The structure of MtDHQase in complex with 1 was solved to 2.5 Å (1 Å=0.1 nm) resolution and electron density (3σ) that is presumed to belong to the catalytic reaction product 3-dehydroshikimate (1) was observed in subunits C and G (Figure 4B) (the substrate was converted into product during crystallization). Electron density for the flexible active-site loop residues 19–24 in subunits C and G was also visible. However, in the other subunits, no discernible additional electron density for the ligand or the loop was observed.

The binding of ligand 1 is stabilized by hydrogen-bonding interactions with the backbone amides of Ile102 and Ser103, with the side-chain atoms of Arg19, Asn75, His81, Arg112 and Asp88* (* indicates residues from an adjacent subunit) and by hydrophobic interactions of its cyclohexene moiety with the side-chain atoms of Tyr24, His101, Leu13, Val105, Arg112 and the Cαs of Gly77 and Gly78 (Figures 4C and 5A and see Supplementary Table S3 at The binding of 1, as expected, seems to induce closure of the loop residues 19–24 through its interaction with the side-chain atoms of the loop residues Tyr24 and Arg19. The side-chain phenol of Tyr24 forms a hydrophobic interaction with the cyclohexene ring of 1, whereas the side chain of Arg19 is positioned to form hydrogen-bond interactions with the C3 carbonyl group of 1 (see Supplementary Table S3). In addition, the side-chain atoms of Arg19 appear to interact with Tyr24 through electrostatic interactions.

Loop residues 19–24 are involved in a network of water-mediated hydrogen bonds that co-ordinate the position of the catalytic Tyr24, orienting its phenolic oxygen atom 3.5 Å from the C2 atom of 1 (Figure 4C). The Tyr24 side chain is further positioned through hydrophobic interactions with the side-chain carbon atoms of Leu16 (3.8 Å), Val23 (4.3 Å) and Ile102 (4.3 Å) (not shown). Arg19 and Arg108 form electrostatic and hydrogen-bonding interactions respectively with Tyr24. These interactions are presumed to stabilize the electron-rich π-system of the Tyr24 side-chain. The Tyr24 phenol oxygen atom makes a hydrogen-bond interaction with Arg108 Nη1 (3.1 Å) and with a water molecule (WAT1) (2.8 Å) (Figure 4C). These groups are likely to be involved in proton abstraction from Tyr24 to generate the active phenolate.

The close proximity of the Tyr24 side chain to the C2 atom of 1 and its requirement for substrate conversion into 1 indicate that it may also play an important role in stabilizing reaction intermediates during catalysis. Inhibitors that are mimics of reaction intermediates can be strong binders and this was the basis for the design of 2,3-anhydroquinate (2), which has a structure with features assumed to be in the transition state, e.g. C2–C3 conjugation [39]. It acts as a competitive inhibitor with a Ki of 200 μM [26].

A new pocket for MtDHQase drug targeting

The structure of the MtDHQase–2–glycerol complex was solved to 2.0 Å resolution (Figures 3B and 6A). All subunits contained inhibitor 2 bound in the active site and a glycerol molecule bound in 13 of the 24 subunits. Inhibitor 2 is positioned in a similar mode as observed in previous structures determined in complex with MtDHQase and ScDHQase (PDB codes 1H0R and 1GU1 respectively); however, this structure successfully captured the full ordering of the flexible catalytic loop residues 19–24. Hydrogen-bonding interactions are observed with side-chain atoms of Arg112, His81, His101, Ser103 and Asp88* and with the main-chain atoms of Ile102 and Ser103 (Supplementary Figure S3A at The binding of inhibitor 2 is also stabilized by hydrophobic interactions similar to those observed for MtDHQase–1 (Figure 5B).

The structure also contains a glycerol molecule, presumed to be from the cryoprotectant solution, which is bound in a subpocket adjacent to the enzyme active site as observed previously for the structure of ScDHQase–2–glycerol (PDB code 1GU1). The glycerol molecule replaces the position of the side-chain atoms of Arg19 observed in the MtDHQase–1 structure (see Supplementary Figure S3B).

The side-chain atoms of Arg19 undergo significant structural rearrangement, rotating approximately 96° compared with the position of the Arg19 side chain in MtDHQase–1, resulting in the formation of several new intra-protein interactions and causing a small shift (1.4–1.1 Å) in the backbone atoms of residues 20–22 (see Supplementary Figure S3B). The new position of Arg19 forms a direct hydrogen bond between its side-chain Nη2 and Asp67* Oδ1 (3.0 Å) and forms several water-mediated hydrogen bonds with main-chain and side-chain atoms of Glu92* and the main-chain atoms of Arg15 and Arg18. The new position of Arg19's side chain also makes a hydrophobic interaction with the carbon side-chain atoms of Glu92*.

Insights into inhibitor design targeting the MtDHQase active-site subpocket

To investigate the importance of the MtDHQase subpocket on inhibitor binding, we have solved four structures of MtDHQase in complex with competitive type II DHQase inhibitors that were designed on the basis of the binding of 2 in the ScDHQase–glycerol structure [12,17,21,22].

We solved the 2.4 Å resolution crystal structure of MtDHQase in complex with inhibitor 3, a biaryl derivative of inhibitor 2 containing a phenyl ring directly attached to the C3 atom of the 2,3-anhydroquinate core that is sequentially connected to a terminal phenyl ring through a rigid carbonyl linker (Figures 2 and 3C). Ligand 3 inhibits both MtDHQase and ScDHQase with Ki values of 11 and 4.7 μM respectively [21]. Electron density was observed for inhibitor 3 in all subunits and for the flexible active-site loop residues 19–24 in most of the subunits of the asymmetric unit with the exception of the side chain of Arg19 in several subunits. The structure of MtDHQase–3 reveals that its biaryl extension binds in different conformations for several MtDHQase subunits and that the overall binding mode occupies only the edge position of the glycerol molecule in MtDHQase–2 (Figure 6B). Despite the multiple conformations observed for the terminal phenyl ring of 3, the binding mode of the anhydroquinate core and its primary phenyl ring are conserved in all subunits and the binding of inhibitor 3 does not cause significant alterations to loop residues 19–24 compared with MtDHQase–2–glycerol (Figure 6B).

The primary phenyl ring connected to the C3 atom of the 2,3-anhydroquinate core of 3 forms weak π–π stacking with Tyr24, and the terminal phenyl ring does not form additional stacking interactions (Figures 5C and 6B). The carbonyl linker of inhibitor 3 is positioned within hydrogen-bonding distance of the side-chain oxygen atom of Asp88* (Figure 5C). The biaryl extension of inhibitor 3 makes hydrophobic interactions with the side-chain and main-chain carbon atoms of Tyr24, Asn12, Leu13, Arg19, Glu20, Glu92*, Asp88* and Ala91* (Figure 5C). The terminal phenyl ring of 3 in several subunits is positioned ~4.5 Å from the glycerol molecule of MtDHQase–2 and occupies a hydrophobic cleft near the solvent-exposed region of the active site composed of the carbon atoms of Arg19, Glu20, Glu92*, Asp88* and Ala91* (not shown).

These interactions observed in the crystal structure of MtDHQase–3 might contribute to its increased inhibitor potency (~20-fold) compared with 2; however, the flexibility of the extension from the anhydroquinate anchor of 3 and its corresponding complementarity to the MtDHQase-binding pocket, is limited due to rigidity induced by the carbonyl linkage of the two aromatic rings. This is supported further by kinetic studies with other biaryl inhibitors from this series that show significant improvements in binding affinity using linkers with greater flexibility than 3 [21] (see Supplementary Table S4 at The rigidity of the biaryl extension prevents inhibitor 3 from reaching the position taken by the glycerol molecule in MtDHQase–2 and forming stacking interactions between the terminal phenyl ring of 3 and the side chain of Tyr24. To gain insight into the influence of the flexibility of different linkers in MtDHQase inhibitors, we have solved the structures for MtDHQase in complex with compounds 4 and 5 (Figures 3D and 3E).

Compounds 4 and 5 closely resemble one another, with the main difference being in the flexibility of the linker. Inhibitor 4 was designed to occupy the subpocket by extending the structure of 2 with a rigid olefinic amide linker attached to a terminal phenyl ring (Figure 2) and has Ki values of 2.3 and 2.1 μM against MtDHQase and ScDHQase respectively [18]. Alternatively, inhibitor 5 was designed to have greater flexibility of its side-chain phenyl moiety. Inhibitor 5 has a phenyl ring connected via a terminal ether linkage attached to the C3 atom of inhibitor 2, yielding one of the most potent DHQase inhibitors reported to date [18] (Figure 2 and see Supplementary Table S4). Inhibitor 5 inhibits both MtDHQase and ScDHQase with Ki values of 0.14 and 0.01 μM respectively [18].

The 1.9 Å resolution crystal structures of MtDHQase–4 and MtDHQase–5 reveal that both phenyl rings fully occupy the position of the MtDHQase–2 glycerol molecule (Figures 3C and 3D). Electron density was observed for inhibitor 4 and for all flexible active-site loop residues 19–24 in all subunits of the asymmetric unit. No significant differences in binding conformation were observed for inhibitor 4 in different subunits of the two dodecamers of the asymmetric unit. The complex of MtDHQase–5 has been crystallized in space group F23 and has only a monomer in the asymmetric unit. In this structure, electron density is observed for the active-site loop.

In MtDHQase–4, the amide linker of inhibitor 4 rigidifies its C3 side chain, hindering the π-stacking interaction between the Tyr24 side-chain phenol and the terminal phenyl ring of 4. The rotationally limited terminal phenyl ring of inhibitor 4 also makes hydrophobic interactions with the side-chain carbon atoms of Asn12, Arg15, Arg19, Glu20, Ala89* and Glu92* (Figure 5D) and causes the expulsion of a water molecule observed previously to link the glycerol molecule of MtDHQase–2 with Arg15 and Gly25 (not shown). The carbon linker atoms of 4 make hydrophobic interactions with the side-chain atoms of Asn12, Leu13, Tyr24, Asp88* and Gly77 (Figure 5D). In addition, the amide linker of 4 forms hydrogen-bond interactions with its carbonyl oxygen atom and the side-chain nitrogen atom of Asn12 and a water-mediated hydrogen bond with Gly78 and Pro11 probably potentiating its binding affinity (Figure 5D).

In MtDHQase–5, the more flexible terminal phenyl ring of 5 is shifted 1.2 Å from the terminal phenyl ring of 4, positioning it further into a hydrophobic portion of the subpocket (Figure 6C). The positioning of the terminal phenyl ring of 5 is 0.7 Å closer to the side-chain phenol of Tyr24 compared with the phenyl ring of inhibitor 4 and makes stronger edge-on π-stacking interactions with the side-chain phenol group. In addition, the terminal phenyl ring of 5 makes new hydrophobic interactions with the side-chain atoms of Leu13 and Leu16. The carbon linker atoms of 5 make similar interactions to those observed in MtDHQase–4, but does not make any hydrogen-bond interactions between its ether oxygen atom and the active-site residues (Figure 5E). The optimized positioning of the terminal phenyl ring of 5 into the hydrophobic patch comprising Leu13, Leu16 and Tyr24 is enabled through the increased flexibility of its ether linkage compared with the more rigid linker of 4 and results in a 16-fold increase in potency between these two inhibitors and a 1400fold increase from the template inhibitor 2 (Figures 5D and 5E and see Supplementary Table S4) [18].

The three crystal structures (MtDHQase–3, 4 and 5) indicate that the dramatic increases in potency for each of these compounds [18,21] correspond with the degree of complementarity for their aromatic extensions from the 2,3-anhydroquinate anchor with the side chain of Tyr24. The formation of stabilizing interactions between inhibitors and the flexible-loop region appears to be a significant factor in the increase of inhibitor potency which can be substantiated further by the crystal structure of MtDHQase–6.

We solved the 2.5 Å resolution crystal structure of MtDHQase in complex with 6 (Figure 3E), which contains a t-butyl amide extension from the C3 atom of 2 (Figure 2). Inhibitor 6 competitively inhibits MtDHQase and ScDHQase with Ki values of 27 and 29 μM respectively (Supplementary Table S4) [22].

Inhibitor 6 is bound in the active sites of 14 of 24 MtDHQase subunits per asymmetric unit. The active-site loop residues 19–24 for the MtDHQase–6 complex were predominantly disordered with only some subunits containing electron density for backbone atoms of this region. Inhibitor 6 binds in similar orientations in 11 of the subunits and takes up an alternative conformation for its t-butyl extension in the remaining three occupied subunits.

Analysis of the MtDHQase–6 structure reveals that the t-butyl amide group does not cause significant changes in the position of the anhydroquinate core of inhibitor 6 relative to its position in MtDHQase–5 or MtDHQase–2 (Figure 6D). The structure of MtDHQase–6 is unique in that the inhibitor only partly reaches the site of the glycerol molecule in MtDHQase–2. The t-butyl amide group of inhibitor 6 extends towards the MDHQase glycerol-binding site and makes hydrophobic interactions with the side-chain carbon atoms of Asn12 and Leu13, as well as Asp88* (not shown). The oxygen atom from the C3 amide linker group is within hydrogen-bonding distance of the side chain of Asp88*. The side-chain atoms of the active-site flexible-loop residues 19–24 are fully disordered, indicating a lack of stabilizing interactions between these atoms and the t-butyl moiety of inhibitor 6. The superposition of MtDHQase–6 and MtDHQase–5 shows that the t-butyl group is 1.6 Å from the position occupied by the side chain of Tyr24 in MtDHQase–1 and 2 and indicates that the flexible-loop residues remain disordered due to steric interference (Figure 6D). The disordering of the flexible loop in MtDHQase–6 appears to significantly reduce binding affinity approximately 200-fold compared with inhibitor 5, which binds with the greatest degree of complementarity to the active-site subpocket.

Discovery and characterization of a planar inhibitor of MtDHQase

The inhibitor series derived from inhibitor 2 are excellent probes for investigating the active-site subpocket of DHQase, but their relatively complex structures and non-trivial syntheses led us to search for more synthetically tractable chemical templates. We therefore screened a small collection of planar product analogues to identify new candidate anchor molecules. Kinetic studies identified compound 7 as a competitive inhibitor of MtDHQase with a Ki of 300 μM (see Supplementary Figure S4 at Inhibitor 7 (citrazinic acid), is an analogue of 1 that lacks the ring puckering due to its planar pyridine core (Figure 2) and contains two hydroxy groups, which can undergo keto–enol tautomerization (Figure 2), possibly enabling it to mimic the molecular conversion of the C3 keto group of substrate (3-dehydroquinate) into the C3 enol/enolate group of the reaction intermediate.

STD NMR spectroscopy was used to confirm the competitive binding of inhibitor 7 to MtDHQase followed by the determination of its ligand efficiency [40] using ITC (see Supplementary Figures S5 and S6 at The ligand efficiency is defined as a measure of the binding energy per non-hydrogen atom of a ligand [40]. Numerous studies have led to the establishment that successful lead molecules typically have ligand efficiency values ≥0.3 kcal·mol−1 (1 kcal=4.184 kJ) per non-hydrogen atom [41]. Inhibitor 7 has a very high ligand efficiency of 0.51 kcal·mol−1 per non-hydrogen atom, making it an attractive scaffold for the development of future series of MtDHQase inhibitors.

We solved the crystal structure of MtDHQase–7 to 2.25 Å resolution, providing structural details into its binding at the MtDHQase active site. Electron density for inhibitor 7 and for the active-site loop residues 19–24 is observed in 22 of the 24 subunits of the asymmetric unit (Figure 3F and see Supplementary Figure S7 at It is unclear which citrazinic acid tautomer is bound in the crystal structure of MTDHQase–7, because the resolution is insufficient to discern between 1.4 Å (C–OH) and 1.2 Å (C=O) bond lengths. However, on the basis of the interactions observed between MtDHQase and the tautomeric groups of inhibitor 7 (see below and Supplementary Table S3), we speculate that the keto-tautomer form is bound in the MtDHQase–7 crystal structure. The remainder of the present paper will refer to the keto-tautomer of 7 for purposes of describing and evaluating binding interactions.

MtDHQase–7 reveals that inhibitor 7 is predominantly bound using interactions similar to those observed in MtDHQase–1 and 2 except that the core of 7 is shifted towards the flexible-loop region by 0.8 Å possibly due to new hydrogen-bond interactions formed between the pyridine nitrogen atom of 7 and the subunit Asp88* (Figure 7 and see Supplementary Table S3). The flexible-loop region shows similar hydrogen-bond interactions with inhibitor 7 to those observed in MtDHQase–1 (see Supplementary Table S3); however, the hydrogen-bond interactions of the C3 carbonyl oxygen atom of 7 with Arg19 Nη2 (2.4 Å) and with a conserved water molecule (2.5 Å) (Figure 7A) appear stronger than those observed in the MtDHQase–1 structure.

Structural comparison of MtDHQase–1 and MtDHQase–7 reveals that binding of the more planar 2 does not cause significant movements of the flexible-loop region or the active-site residues upon binding of 7, with the exception of rearrangements for the side-chain atoms of His101 and Val105 (Figure 7B). In MtDHQase–7, His101 Cδ2 makes hydrophobic interactions with Val105 Cγ1, which has adopted an alternative side-chain conformation (not shown). His101 Nδ1 moves to 3.1 Å from Asn75 Oδ, and makes a new intra-protein hydrogen bond not observed in the MtDHQase–1 structure (Figure 7B). Furthermore, Glu99 Oϵ2 also moves 0.8 Å closer to His101 Nϵ2, making a stronger hydrogen-bonding interaction. In addition to these movements, the hydrogen bond formed by the pyridine nitrogen atom of 7 and the side-chain carboxy group of Asp88* of a neighbouring subunit causes subtle alteration of the salt-bridge interaction between Asp88* and Arg112 in comparison with MtDHQase–1 (Figure 7).

The structural movement of the catalytic His101 side chain results in the formation of new intra-protein interactions that are not observed in MtDHQase–1 and 2. This new orientation of His101 in MtDHQase–7 is presumed to be non-catalytic due to its increased intra-protein interactions (Figure 7) and appears to be a key feature in the ability of 7 to inhibit MtDHQase.


The MtDHQase crystallographic structures shown in the present paper reveal structural details for the flexible catalytic loop residues 19–24 in MtDHQase, and also provide structural evidence and characterization of a pocket adjacent to the MtDHQase active site and its influence on MtDHQase inhibitor binding. MtDHQase–1 reveals a series of intra-protein interactions involved in the co-ordination of the Tyr24 side chain into a presumed catalytically relevant position, which is co-ordinated further by the formation of electrostatic interactions with the side chain of Arg19. The requirement of these two residues for catalytic activity makes them attractive focal points for inhibitor development. Our investigation of MtDHQase in complex with inhibitors that occupy both the MtDHQase active site and the adjacent subpocket (36) and inhibitors that occupy only the MtDHQase active site (2 and 7) reveals that the most potent MtDHQase inhibitors replace the electrostatic interaction between the side-chain atoms of Arg19 and Tyr24 through the expulsion of the Arg19 side chain to the solvent as was observed by Peón et al. [20]. This structural perturbation of the MtDHQase loop residues 19–24 effectively expands the available area for creating protein–inhibitor binding interactions.

The lack of structural evidence and information for this active-site-adjacent subpocket in MtDHQase previously limited our understanding of differences in inhibitor complementarity between MtDHQase and ScDHQase. These structural investigations against MtDHQase and their comparison with crystals structures available for ScDHQase reveal key differences in the active-site subpockets of the enzymes from these two organisms. The flexibility of the terminal phenyl rings for 3 and 4 are both constrained by their rigid linkers and both display negligible differences in inhibition potency (Figure 1 and see Supplementary Table S4). However, inhibitor 5, which has a much more flexible linker, displays a significant (~14-fold) difference in potency against MtDHQase and ScDHQase (Figure 1 and see Supplementary Table S4).

Two structures of ScDHQase in complex with inhibitors containing terminal phenyl rings with linkers of similar length and flexibility to 5 have been deposited in the PDB (codes 2CJF and 2BT4) [17,18]. One of these structures is in complex with Biaryl-B, a biphenyl inhibitor that has a flexible thioether linkage and is a 5-fold more potent inhibitor of MtDHQase than ScDHQase (PDB code 2CJF) (see Supplementary Table S4) [19]. The other is a structure of ScDHQase in complex with CA2, a single ring extension inhibitor (PDB code 2BT4). The superposition of MtDHQase–5 and ScDHQase–Biaryl-B reveals several differences in the positioning of their respective terminal phenyl rings, providing insights into unique features of their subpockets. Although the terminal phenyl ring of Biaryl-B makes similar interactions to those of 5, the π-stacking interaction between it and the phenol group of ScDHQase Tyr28 appears to be stronger than is observed for MtDHQase–5. The terminal phenyl ring of Biaryl-B is rotated ~50° and is shifted 1.4 Å further into the pocket relative to 5 (see Supplementary Figure S8C at It is possible that the difference in the positioning of Biaryl-B is caused by its own biaryl system; however, this is probably not the case as the crystal structure of ScDHQase in complex with the single ring extension inhibitor CA2 also positions its terminal phenyl ring exactly as observed in ScDHQase–Biaryl-B (see Supplementary Figures S8A and S8B).

The superposition of MtDHQase–5 and ScDHQase–Biaryl-B shows that backbone atoms of the loop region are shifted by ~2.2 Å between these two structures and the main-chain carbonyl group of MtDHQsae Arg19 is positioned similarly to that of ScDHQase Gln22. The two peptide bonds are flipped 180° relative to each other, positioning the MtDHQase Arg19 backbone carbonyl group into the subpocket in a similar position to that observed in MtDHQase–2–glycerol (see Supplementary Figure S8C). Furthermore, two residues forming hydrophobic interactions with the terminal phenyl ring in ScDHQase–Biaryl-B are replaced with polar residues in MtDHQase. ScDHQase Leu19 is equivalently positioned to MtDHQase Arg15 and ScDHQase Thr96 is equivalently positioned to MtDHQase Glu92. The differences in polarity for these residues in addition to the flipped peptide bond of Arg19 presumably contribute to the differences in their inhibitor potencies.

The seven MtDHQase–ligand structures of the present study show the complete active-site structure of MtDHQase, including its full catalytic loop region, MtDHQase–1, 2 and 7 structures, and, in an extended conformation, MtDHQase–3, 4 and 5 structures, similar to the recent study by Peón et al. [20]. The present study clearly shows the formation of a stacking interaction between Tyr24 and an aromatic system such as the primary phenyl ring of inhibitor 3 or the terminal phenyl ring of inhibitors 4 and 5, is important in the gain of the affinity and that its engagement might be explored further using different linkers from the anhydroquinate core to increase complementarity. We also report the first structure of MtDHQase in complex with a planar molecule bound in the active site, representing a novel scaffold for MtDHQase inhibitor development. Very few studies have tried to optimize interactions of the anchor molecule to MtDHQase and no significant progress has been reported to the best of our knowledge [8,22,42]. Citrazinic acid (7) represents a stepping stone in this direction, and synthetic efforts are currently in progress to develop novel MtDHQase inhibitors on the basis of the planarity of citrazinic acid that target the MtDHQase subpocket as guided by the X-ray crystal structures of the present study.


Marcio Dias and William Snee planned the experiments, produced the protein, crystallized, solved the structures, analysed the results and wrote the paper; Karen Bromfield and Richard Payne synthesized the DHQase inhibitors; Satheesh Palaninathan helped in the data analysis and contributed to the writing of the paper; Alessio Ciulli and Nigel Howard performed the ITC and NMR studies and inhibition assays for citrazinic acid; Chris Abell, James Sacchettini and Tom Blundell directed the research and contributed to the writing of the paper.


This work was funded by a grant from the Bill and Melinda Gates Foundation (subcontract on Integrated Methods for Tuberculosis Drug Discovery grant to the Seattle Biomedical Research Institute). M.V.B.D. thanks the Conselho Nacional de Desenvolvimento Científico e Tecnológico (Brazil) for a postdoctoral fellowship.


We appreciate the support of staff scientists at beamline 19-ID and 23-ID of the Advanced Photon Source, Argonne National Laboratory, for help in data collection. We also thank Justin Roberts, Joel Freundlich and Hilary Baird for excellent technical assistance, and Siaska Castro and Tracey Musa for comments on the paper.


  • The structural co-ordinates reported for 3-dehydroquinate dehydratase from Mycobacterium tuberculosis in complex with various inhibitors have been deposited in the PDB under accession codes 3N59, 3N7A, 3N87, 3N86, 3N76, 3N8N and 3N8K.

Abbreviations: DHQase, 3-dehydroquinate dehydratase; HpDHQase, Helicobacter pylori DHQase; ITC, isothermal titration calorimetry; MtDHQase, Mycobacterium tuberculosis DHQase; ScDHQase, Streptocymes coelicolor DHQase; STD, saturation transfer difference


View Abstract