Structural Basis for Catalytic and Inhibitory Mechanisms of β-Hydroxyacyl-acyl Carrier Protein Dehydratase (FabZ)*

β-Hydroxyacyl-acyl carrier protein dehydratase (FabZ) is an important enzyme for the elongation cycles of both saturated and unsaturated fatty acids biosyntheses in the type II fatty acid biosynthesis system (FAS II) pathway. FabZ has been an essential target for the discovery of compounds effective against pathogenic microbes. In this work, to characterize the catalytic and inhibitory mechanisms of FabZ, the crystal structures of the FabZ of Helicobacter pylori (HpFabZ) and its complexes with two newly discovered inhibitors have been solved. Different from the structures of other bacterial FabZs, HpFabZ contains an extra short two-turn α-helix (α4) between α3 and β3, which plays an important role in shaping the substrate-binding tunnel. Residue Tyr-100 at the entrance of the tunnel adopts either an open or closed conformation in the crystal structure. The crystal structural characterization, the binding affinity determination, and the enzymatic activity assay of the HpFabZ mutant (Y100A) confirm the importance of Tyr-100 in catalytic activity and substrate binding. Residue Phe-83 at the exit tunnel was also refined in two alternative conformations, leading the tunnel to form an L-shape and U-shape. All these data thus contributed much to understanding the catalytic mechanism of HpFabZ. In addition, the co-crystal structures of HpFabZ with its inhibitors have suggested that the enzymatic activity of HpFabZ could be inhibited either by occupying the entrance of the tunnel or plugging the tunnel to prevent the substrate from accessing the active site. Our study has provided some insights into the catalytic and inhibitory mechanisms of FabZ, thus facilitating antibacterial agent development.

␤-Hydroxyacyl-acyl carrier protein dehydratase (FabZ) is an important enzyme for the elongation cycles of both saturated and unsaturated fatty acids biosyntheses in the type II fatty acid biosynthesis system (FAS II) pathway. FabZ has been an essential target for the discovery of compounds effective against pathogenic microbes. In this work, to characterize the catalytic and inhibitory mechanisms of FabZ, the crystal structures of the FabZ of Helicobacter pylori (HpFabZ) and its complexes with two newly discovered inhibitors have been solved. Different from the structures of other bacterial FabZs, HpFabZ contains an extra short two-turn ␣-helix (␣4) between ␣3 and ␤3, which plays an important role in shaping the substrate-binding tunnel. Residue Tyr-100 at the entrance of the tunnel adopts either an open or closed conformation in the crystal structure. The crystal structural characterization, the binding affinity determination, and the enzymatic activity assay of the HpFabZ mutant (Y100A) confirm the importance of Tyr-100 in catalytic activity and substrate binding. Residue Phe-83 at the exit tunnel was also refined in two alternative conformations, leading the tunnel to form an L-shape and U-shape. All these data thus contributed much to understanding the catalytic mechanism of HpFabZ. In addition, the co-crystal structures of HpFabZ with its inhibitors have suggested that the enzymatic activity of HpFabZ could be inhibited either by occupying the entrance of the tunnel or plugging the tunnel to prevent the substrate from accessing the active site. Our study has provided some insights into the catalytic and inhibitory mechanisms of FabZ, thus facilitating antibacterial agent development.
Recently, because of the major difference between the enzymes involved in the type II fatty acid biosynthesis system (FAS II) 4 found in bacteria and the counterpart in mammals and yeast, the enzymes involved in FAS II have been looked at as a promising antibacterial agent development target (1)(2)(3). The synthesis of fatty acid in vivo includes initiation and elongation phases. During the elongation cycle of FAS II, four consecutive reactions complete the extension of two carbons (4,5). In the third step of the elongation cycle, the dehydration of ␤-hydroxy-ACP to trans-2-acyl-ACP is catalyzed by FabA or FabZ. FabA is a bifunctional enzyme and only found in Gram-negative bacteria with its partner, FabB, to participate in the formation of unsaturated fatty acids (6). FabZ has ubiquitous distribution in the FAS II pathway, and it is a primary dehydratase that participates in the elongation cycles of both saturated and unsaturated fatty acid biosyntheses. Accordingly, FabZ presents itself as a suitable, yet unexplored, target for the discovery of compounds effective against pathogenic microbes (5,7).
So far, although the detailed enzymatic characterization has been performed for the FabZs from Enterococcus faecalis (EfFabZ) (8,9), Pseudomonas aeruginosa (PaFabZ) (7), and Plasmodium falciparum (PfFabZ) (2,5,10), and the crystal structures of FabZs from P. aeruginosa and P. falciparum have been determined, the structural basis underlying the catalytic mechanism of FabZ still remains unclear (3,7,10). However, because these structures were solved without the presence of inhibitor, they provide little insight into the mechanism of FabZ inhibition.
In our work, to uncover the structural basis underlying the catalytic mechanism of FabZ, the crystal structures of both FabZ from Helicobacter pylori (HpFabZ) and its Y100A mutant have been determined, and their catalytic activities and binding characteristics to ACP investigated. To illustrate the inhibitory mechanism of FabZ, two new inhibitors of FabZ were discovered, and their crystal structures in complex with HpFabZ were determined, which has provided the first structural model of the inhibitor-FabZ complex.

EXPERIMENTAL PROCEDURES
Expression and Purification of Recombinant HpFabZ-The expression, purification, and enzymatic assay of HpFabZ were performed as described previously (4 SPR Technology-based Binding Assay-The compounds for the FabZ inhibitor random screening were from the in-house library. The binding affinity assay of inhibitor against HpFabZ in vitro was carried out by using the SPR technology-based Biacore 3000 instrument (Biacore AB, Uppsala, Sweden) for the primary screening (11). Immobilization of protein to the hydrophilic carboxymethylated dextran matrix of the sensor chip CM5 (Biacore) was performed by the standard primary amine coupling reaction. HpFabZ to be covalently bound to the matrix was dissolved in 10 mM sodium acetate buffer (pH 4.3). Equilibration of the baseline was completed by a continuous flow of HBS-EP running buffer (10 mM HEPES, 150 mM NaCl, 3.4 mM EDTA, and 0.005% (v/v) surfactant P20, pH 7.4) through the chip for 1ϳ2 h. Biacore data were collected at 25°C with HBS-EP as the running buffer at a constant flow of 20 l/min. The equilibrium dissociation constant (K D ) for evaluation of the protein-ligand binding affinity was determined by the steady-state affinity fitting analysis of the Biacore data.
Enzymatic Inhibition Assay-Following the SPR technologybased binding assay, the compounds with high binding affinities to FabZ (K D Ͻ 10 M) were further validated by enzymatic assay. The enzymatic inhibition assay of HpFabZ was monitored by the spectrophotometric method (4,5). In brief, the activity of HpFabZ was measured by detection of the decrease in absorbance at 260 nm for the conversion of crotonoyl-CoA to ␤-hydroxybutyryl-CoA. The compound dissolved in 1% Me 2 SO was incubated with the enzyme for 1 h before the assay was started. The 50% inhibitory concentration (IC 50 ) of each inhibitor was estimated by fitting the inhibition data to a dosedependent curve using a logistic derivative equation (11). The inhibition mode assay for HpFabZ was also monitored using the spectrophotometric approach (4,5). The reaction mixture consisted of 10 g of HpFabZ in 20 mM Tris-HCl at pH 8.0, 500 mM NaCl, 1 mM EDTA, and 1% Me 2 SO in a total volume of 200 l. The decrease in absorption at 260 nm using crotonoyl-CoA as substrate was monitored for 5 min at 25°C. The inhibitor type was determined in the presence of various inhibitor concentrations (0 -50 M). After 2 h of incubation, the reaction was started by the addition of crotonoyl-CoA (10 -250 M). The K i values were obtained from Lineweaver-Burk double-reciprocal plots and subsequent secondary plots.
Site-directed Mutagenesis-Site-directed mutagenesis was performed using the QuikChange TM site-directed mutagenesis kit (Stragene) according to the manufacturer's instructions. pQE30-HpFabZ plasmid was used as template for construction of the HpFabZ (Y100A) mutant. The mutant was sequenced to confirm the desired mutation site. The expression and purification of HpFabZ (Y100A) were similar to those of HpFabZ as described above. The enzymatic characterization of HpFabZ (Y100A) mutant was determined according to the previously published method (4). The binding affinity of HpFabZ or its mutant to holo-HpACP was performed on the SPR technology-based Biacore 3000 instrument (12), in which the holo-HpACP protein was obtained and immobilized on the CM5 chip as described in our previous work (13).
Crystallization and Data Collection-The purified HpFabZ was dialyzed against 20 mM Tris-HCl at pH 9.0, 500 mM NaCl and concentrated to ϳ10 mg/ml. For crystallization, 1 l of protein was mixed with an equal volume of reservoir solution containing 2 M sodium formate, 0.1 M sodium acetate trihydrate at pH 3.6 -5.6 and benzamidine-HCl was added to a final concentration of 2% (w/v). The mixture was equilibrated against 500 l of the reservoir solution at 277 K by the hanging-drop vapor-diffusion method. Crystals of dimensions 0.5 ϫ 0.3 ϫ 0.3 mm 3 were obtained after 7 days. Inhibitor was added to the original drop to a final concentration of ϳ20 mM, and the crystals were soaked for 24 h. Similarly, crystals of HpFabZ (Y100A) mutant in dimensions of 0.3 ϫ 0.3 ϫ 0.1 mm 3 were obtained after 5 days in reservoir solution (2 M sodium formate, 0.1 M sodium acetate trihydrate, pH 4.0 -6.2).
Diffraction data were collected at 100 K using CuK␣ x-ray with a Rigaku R-AXIS IVϩϩ image plate. Before crystals were flash-frozen in liquid nitrogen, the drop was dehydrated against 500 l of reservoir solution containing 4.0 M sodium formate for 24 h. The data were processed using HKL2000 (14). Analysis of the diffraction data indicated that the native crystals belong to space group P2 1 2 1 2 1 and mutant crystals to C2. The crystallographic statistics are summarized in Table 1. The structures were solved by molecular replacement method with the crystal structure of PaFabZ as the search model and refined by using the program CNS (15). Electron density interpretation and model building were performed using the computer graphics program "coot" (16).

RESULTS
Structure of HpFabZ-The native HpFabZ protein used for structure determination was not truncated. The crystal structure was determined by molecular replacement (MR) to a resolution of 2.2 Å. The final structure of HpFabZ is well ordered, and the electron densities for all residues with the exception of the N terminus (residues 1-10) are clearly interpretable. The statistics of the diffraction data and structure refinement are listed in Table 1. Similar to the structures of PfFabZ (2) and PaFabZ (7), the x-ray crystal structure of the native HpFabZ is determined as a hexameric architecture, displaying a classic "trimer of dimmers" organization ( Fig. 1A). The six monomers arrange a ring-like (A-B-F-E-C-D-A) contact topology.
The overall structure of the HpFabZ dimer is similar to the dimers of PaFabZ (7), PfFabZ (2), and EcFabA (17). The HpFabZ dimer is formed mainly via the hydrogen bonds between residues 102-108 of the two ␤3 strands from two dif-ferent monomers, thus all the sheets constitute a 12-stranded curved anti-parallel ␤-sheet layer ( Fig. 2A). The dimer is further stabilized by the hydrophobic interactions within residues Val-68, Leu-69, and Val-71 in the last two turns of ␣3 helix, and interactions between residues Pro-22, His-23, and Phe-59Ј (the prime indicates the residue from the other subunit in the dimer). The dimer forms two substrate-binding tunnels with catalytic sites on the interface, which are ϳ20 Å away from each other ( Fig. 2A). Each catalytic site is formed by two highly conserved residues, His-58 and Glu-72Ј (His-49 and Glu-63Ј in PaFabZ; His-133 and Glu-147Ј in PfFabZ) that are located in the middle kink of the substrate-binding tunnel made up by the conserved residues nearby (Figs. 1D and 2, B and C).  (Fig. 1D). The entrance is near the ACP binding groove (7), which is composed of a floor (residues Tyr-100, Met-102, Thr-103, Ile-132, Met-154, and Ala-156) and an edge (residues Lys-129, Lys-108Ј, and Arg-110Ј).
Residue Tyr-100 of each monomer acts as a doorkeeper of the tunnel, residing between the entrance and the ACP groove (Fig. 2, B and C). Remarkably, Tyr-100 adopts two types of conformations in the whole structure of HpFabZ hexamer: Tyr-100 is recognized in a closed conformation in  (Fig. 2B), Tyr-100 flops ϳ120°around the C ␣ -C ␤ bond and is stabilized by the Van der Waals interactions with Met-154, Lys-62Ј, and Ile-64Ј, allowing the chains of substrates to enter the tunnel. In contrast to the high conservation of residues near the entrance, residues near the bottom of the tunnel are not conserved (Fig. 1D). Residues Pro-91, Ile-93, Ala-94, Lys-95, and Thr-96 in the unique ␣4 helix constitute half of the bottom of the tunnel. Residues Ile-20, Leu-21, Pro-22, Ala-94, Lys-95, and Lys-97 form a "back door (exit)" for the tunnel. Like Tyr-100, Phe-83 (Phe-74 in PaFabZ, Leu-158 in PfFabZ, and Gly-95 in EcFabA) acting as a doorkeeper from the exit also adopts two major conformations, closed and open (Fig. 2, B and C), In the open conformation, Phe-83 rotates ϳ120°around the C ␣ -C ␤ bond and its side chain points toward Ile-93, exposing the exit to the bulk solvent. Meanwhile, the two conformations of Phe-83 lead the tunnel to form two shapes, L-shape and U-shape, as will be discussed below.
Three dimers interact with each other through hydrogen bonds (H-bonds) and hydrophobic interactions to form a hexamer (Fig. 1A). On the dimer-dimer interface, Thr-49 at the loop between ␣2 and ␤2 forms H-bonds with Asp-31Ј at the N-terminal of ␤1Ј and Lys-46Ј at the C-terminal of ␤2Ј; and Phe-50 hydrophobically interacts with Ile-14Ј, Leu-18Ј, and Leu-28. Because of these strong H-bonds and hydrophobic interactions, HpFabZ still possesses enzymatic activity and high stability at extreme conditions, such as high temperature (Ͼ90°C) and extreme pH value as described in our previous report (4). In addition, the ␤3 and ␤6 strands of monomers B and C turned around toward the ␤5 strand with the largest displacement of 8.8 Å (Fig. 2D), causing the formation of two "grooves" that are ϳ4 Å wide and ϳ16 Å long on the surfaces of the dimer A/B and dimer C/D (Fig. 2, E and F), which have never been observed in the structures of PaFabZ, PfFabZ, and EcFabA. As will be discussed below, these grooves facilitate the binding of inhibitor.
The Role of Tyr-100 in the Catalytic Activity of HpFabZ-As mentioned above, Tyr-100 adopts two conformations in the crystal structure (Fig. 2, B and C). To investigate whether the flexibility of Tyr-100 is associated with the function of HpFabZ, site-directed mutagenesis of tyrosine 100 to alanine 100 (Y100A) was performed. The enzymatic assay showed that this mutant possessed less than 50% enzymatic activity of the wild type with K m at 197.8 Ϯ 12.6 M, k cat at 0.0039 s Ϫ1 , and k cat /K m   (Fig. 3A). The crystal structure of the mutant was also obtained ( Fig. 3 and supplemental Figs. S2 and S3). All the results have thereby suggested that residue Tyr-100 plays an important role in the fatty acid biosynthesis. The crystal structure of HpFabZ (Y100A) mutant was solved at resolution of 2.5 Å. The structure of the hexamer is very similar to that of the wild-type enzyme except for two minor differences. One difference is that the N terminus of ␤3 strand (residues Ile-98 and Val-99) in mutant FabZ Y100A becomes a loop in monomers A and E (supplemental Fig. S2). The other is the disappearance of the stereo specific blockade for the tunnel entrance (Fig. 3, C and D). As a result, a new tunnel entrance was formed by residues Ile-98, Ala-100, Phe-101, Met-102, Phe-59Ј, Ile-64Ј, Phe-109Ј, Arg-110Ј, and Pro-112Ј, and this entrance, ϳ15 Å in width, completely exposed the active site to the bulk solvent (Fig. 3D), indicating that the mutant may be more suitable for ACP binding, because the ACP binding groove of Y100A mutant is widened. To further investigate the role of Tyr-100 in ACP binding, the binding affinities of ACP with both HpFabZ and its Y100A mutant were determined by using the SPR technique-based assay (Fig. 3B). Indeed, the binding affinity of HpFabZ (Y100A) mutant to holo-HpACP was significantly enhanced compared with that of wild-type HpFabZ.

Structural Mechanisms of FabZ Catalysis and Inhibition
Discovery of HpFabZ Inhibitor-The compounds for FabZ inhibitor screening were from the in-house library. In the primary screening, the compound binding affinity assay against FabZ enzyme was carried out by SPR technology-based Biacore instrument. At this cycle, about 10 compounds with high binding affinities to FabZ (K D Ͻ 10 M) were obtained and further applied to enzymatic inhibition assay, among which 2 compounds (named compound 1 and compound 2 in the article) were discovered to exhibit inhibition against HpFabZ with an IC 50 value of 39.8 Ϯ 0.35 and 47.6 Ϯ 0.29 M, respectively. In addition, the competitive inhibition modes for both of these two inhibitors were also determined by K i values of 9.7 and 4.3 M, respectively (Fig. 4, C and F), comparable with their K D values of 8.61 and 7.59 M produced from the SPR assay (Fig. 4, B and E).
To test whether these two compounds could also inhibit FabZs of other species, the recombinant EcFabZ was cloned and purified (see supplemental data). The enzyme inhibition assay showed that these two HpFabZ inhibitors could also inhibit the enzymatic activity of EcFabZ by IC 50 values of 32.8 Ϯ 0.65 (compound 1) and 40.4 Ϯ 0.41 M (compound 2), thus suggesting that these two inhibitors might act as leads for developing antibiotics.
In addition, we have ever also performed the antibacterial assay for these two inhibitors in vivo. However, because of their too poor solubility in medium, no acceptable results could be obtained in the related assays. The surfaces of two active tunnels in the dimer are represented, respectively, and the entrance and exit of each tunnel are also marked by arrows. B, side view of HpFabZ active tunnel surface (the residues Phe-83 and Tyr-100 belong to monomer F). The picture was produced by Molcad of the program suite Sybyl. 5 Door residues (Phe-83, Tyr-100) and active residues (Glu-72, His-58Ј) are shown as sticks. The tunnel shows a conformation with entrance door opened and exit door closed; thus the substrate chain has to stretch into the bottom of the L-shaped tunnel due to the closing of the exit door. C, side-view of HpFabZ active tunnel surface (the residues Phe-83 and Tyr-100 belong to monomer A). The tunnel possesses another conformation with entrance door closed and exit door opened, then long substrate chain can stretch out from the exit door. D, schematic picture of the superposition between dimer A/B and dimer C/D. ␤3 and ␤6 in dimer D as well as ␤3Ј and ␤6Ј in dimer B are labeled. There is an obvious shift (the largest displacement is 8.8 Å) of these two strands (␤3 and ␤3Ј) between monomer B and D, and the shift is shown with arrows. E, top view of HpFabZ active tunnel surface (dimer A/B, the residue Tyr-100 belongs to monomer B), produced by PyMOL. The tunnel is not covered by the surface, exposing part of its hydrophobic inner wall to the bulk solvent. The door residue Tyr-100 points toward Pro-112Ј to display a closed conformation. These two residues are shown as sticks. F, top view of HpFabZ active tunnel surface (dimer A/B, the residue Tyr-100 belongs to monomer A). The tunnel is covered completely by the surface. Tyr-100 also displays a closed conformation.
More recently, by screening natural product library, Tasdemir et al. (19) discovered several flavonoids as PfFabZ inhibitors. We also discovered two favonoids PfFabZ inhibitors (compounds 3 and 4) that exhibit moderate binding affinities and inhibitory activities toward HpFabZ (supplemental Fig. S4), thereby indicating that HpFabZ might act as a potent target for discovering inhibitors against not only H. pylori but also other bacteria.
Structure of the Inhibitor-Hp-FabZ Complex-To investigate the potential inhibition mechanism of HpFabZ, the crystal structures of HpFabZ in complex with compounds 1 and 2 were determined (Fig. 5). The binding models of compounds 1 and 2 with HpFabZ are shown in Fig. 6. Compound 1 is mainly composed of a pyridine ring and a 2,4-dihydroxy-3,5-dibromo phenyl ring linked by a formohydrazide moiety. The whole structure of the compound is arranged in a trans configuration around the CϭC bond (Fig. 4A). Interestingly, two C, top view of the HpFabZ active tunnel surface (the residue Tyr-100 belongs to monomer F). The door residue Tyr-100 displays an opened conformation. D, top view of the active tunnel surface for Y100A mutant (the residue Ala-100 belongs to monomer F). The entrance is enlarged and exposed part of its hydrophobic inner wall to the bulk solvent. molecules of compound 1 bind to one hexamer of HpFabZ with two distinct interaction models: one molecule fits into the groove around the entrance (Fig. 6A) (designated as model A hereinafter); the other molecule lies at the active site of the tunnel (Fig. 6B) (called model B below). For model A, Tyr-100 and Phe-83 adopt open and closed conformations, respectively, thus the pyridine ring of compound 1 is located between the phenol ring of Tyr-100 and the pyrrolidine ring of Pro-112Ј, forming a sandwich structure; and the 2,4-dihydroxy-3,5-dibromo phenyl ring at the other end of compound 1 binds into a pocket formed by Arg-158, Glu-159, Phe-59Ј, Lys-62Ј through hydrophobic interactions. For model B, the whole molecule of compound 1 entered into the middle of the tunnel that is located near the active site of HpFabZ (His-58 and Glu-72Ј). The pyridine ring almost accessed the exit of the tunnel, contacting with Ile-20, Leu-21, Pro-22, Phe-83, Ala-94, Lys-97, and Val-99 via hydrophobic interactions. While the 2,4-dihydroxy-3,5-dibromo phenyl ring is sandwiched between Ile-98 and Phe-59Ј, and Tyr-100 adopted a closed conformation and Phe-83 adopted an open conformation.
Unlike compound 1, compound 2 can only bind to monomer B of HpFabZ with a binding model similar to model A and adopt an extended planar conformation to fully fill the groove of the entrance (Figs. 4D and 6C): the phenyl ring is sandwiched between Tyr-100 and Pro-112Ј through a hydrophobic interaction; the 2-chloro-benzoic acid moiety is stabilized by cationinteraction with the positive charged carbamidine group of Arg-158 and H-bond interaction with the backbone carbonyl oxygen of Pro-60Ј (20). The oxygen of the 2-methoxy-ethyl chain of compound 2 also forms H-bond with the backbone nitrogen of Phe-101.
The coordinates and structure factors of HpFabZ, HpFabZ complexed with compound 1, HpFabZ complexed with compound 2, and HpFabZ (Y100A) have been deposited in the Protein Data Bank (PDB accession codes: 2GLL, 2GLP, 2GLM, and 2GLV). The codes of the PDB Chemical Component Dictionary for compounds 1 and 2 are BDE and SCB in the crystal structures.

DISCUSSION
To date, only the crystal structures of PaFabZ and PfFabZ have been determined (2,7). In this work, the x-ray crystal structure of HpFabZ was determined. Although the overall structure of HpFabZ is similar to those of PaFabZ and PfFabZ (supplemental Fig. S1), the crystal structure of HpFabZ still has its unique features. For the monomer of HpFabZ, an additional ␣ helix (␣4) that does not exist in PaFabZ or PfFabZ could be observed between ␣3 and ␤3 (Fig. 1D). This extra short two-turn ␣-helix shapes half of the tunnel bottom (Fig. 2, B and C), which solidifies the tunnel in comparison with the tunnels of PaFabZ and PfFabZ. The crystal structure of HpFabZ also reveals the flexibility of Phe-83, a non-conserved residue at the bottom of the tunnel. When Phe-83 adopts the closed conformation, the exit (back door) of the tunnel is closed and the tunnel is shaped as an L-form (Fig. 2B); when Phe-83 turns out as in the open conformation, the exit is opened up and the tunnel turns into a U-form ( Fig. 2C). The existence of the two forms of the tunnel in HpFabZ clearly explained the paradox that the tunnel length of FabZ (ϳ18 Å) is not adequate to accommodate longer (Ն16carbon acyl chain) substrates (2), for the tails of long subtracts may thread the exit. While in EcFabA, this doorkeeper does not exist (the corresponding residue is Gly-95) (Fig. 1D), it can thus only catalyze short substrates (Յ10-carbon).
Another major difference between the hexamer structure of HpFabZ and those of PaFabZ and PfFabZ is that the former is unsymmetrical and the later are highly symmetrical. For the hexamer structure of HpFabZ, monomers A, D, E, and F fold into an identical conformation, and monomers B and C adopt a different conformation. Structural superposition indicates that the difference between these two conformations comes from the positions of ␤3 and ␤6 strands (Fig. 2D). The ␤3 and ␤6 strands of monomers B and C turn around toward the ␤5 strand, the largest displacement being up to 8.8 Å. The exceptional conformations of monomers B and C cause another structural specialty of HpFabZ, i.e. there is a ϳ4-Å wide and ϳ16-Å long groove on the surfaces of dimers A/B and C/D. This groove, never seen in the structures of PaFabZ and PfFabZ, is located around the entrance of the tunnel. As shown in Fig. 2, E and F, this groove is exactly where the inhibitors reside.
In this work, two different HpFabZ-inhibitor binding models have been proposed. As indicated in the complex structures, the hydrophobic active tunnel is exposed to bulk solvent due to the displacement of ␤3 and ␤6 strands in monomers B and C. Such displacement contributed a lot to the hydrophobic interactions between protein and inhibitors. While in the structures of PaFabZ and PfFabZ, six monomers are highly symmetric with a 3-fold axis and no displacement could be observed in ␤3 strand. The active tunnels of the hexamer are completely blocked by the surface, thus interfering with the inhibitors entry into the active tunnel to form co-crystal. However, such a ␤3 strand orientation difference between HpFabZ and PaFabZ might be related to crystal packing, although their ␤3 strands might exhibit much more flexible conformation in solution (2). Therefore, according to the above-mentioned indication, the determined inhibitors seem to show inhibition activities against HpFabZ, PfFabZ, and EcFabZ as indicated by enzyme assay but fail to form crystals in complex with PfFabZ and EcFabZ. Additionally, the poor solubility of the inhibitors and the weak hydrophobic inhibitor-enzyme interaction may lead to the relatively low occupancy of the inhibitors in the complex structure, thus resulting in a high B-factor. In fact, such a high B-factor issue seems to be common for soaking-approachbased protein/small molecule complex structures (21,22).
It is noticeable that there are above normal number of water molecules in the structures (over twice as many as in other similar structures), which, we think, could be also acceptable. During the structural analysis, the CNS program (water_pick. inp) was used for water molecule pick up. The sigma cutoff value was set to 3.0. Most of the water molecules had excellent 2Fo-Fc electron densities contoured at 1.25 , indicating the accuracy of the picked water molecules. The large number of the water molecules containing in the crystals might be due to the following dehydration approach. In considering that the crystals would be cracked when soaked in the cryoprotectant even with a low concentration of glycerol, the crystal-containing drop was dehydrated against 500 l of reservoir solution with 4.0 M sodium formate for 24 h before the crystals were flash-frozen in liquid nitrogen. Thus, it seems that our different dehydration approach has caused the difference in the number of water molecules in the crystals. The x-ray crystal structure of HpFabZ revealed the flexibility of Tyr-100 which adopted two conformations, open and closed conformations, at the entrance of the tunnel (Fig. 2, B and C). Although the flexibility of corresponding residues in PaFabZ (Tyr-88) (7) and PfFabZ (Leu-170) (2) were addressed by the x-ray crystal structure determinations, they have not been appreciated. Our site-directed mutagenesis revealed that the enzymatic activity of HpFabZ (Y100A) mutant decreases less than 50% (Fig. 3A). This highlights that Tyr-100 might facilitate the reaction of fatty acid biosynthesis. The binding data indicates that ACP binds to HpFabZ (Y100A) mutant much stronger than to the wild-type HpFabZ. In particular, the dissociation step of ACP from the HpFabZ mutant is extremely slow as shown in Fig. 3B. This may be the major reason for the enzy-matic activity of the mutant dropping less than 50%. After the dehydration of substrate (␤-hydroxyacyl-ACP) is completed, the product (trans-2-acyl-ACP) has to exit and vacate the tunnel for the next molecule of â-hydroxyacyl-ACP. Slow dissociation of ACP from FabZ thus slows down the catalytic rate of the enzyme. Indeed, the x-ray structure shows that the entrance of the mutant is enlarged due to the small size of the side chain of substituted Ala-100, which is beneficial to binding with ACP. From these experimental results, we can deduce that the side chain of Tyr-100 may switch frequently between the open and closed conformations (Fig. 2), and the waggling movement of Tyr-100 is an important driving force for products leaving FabZ. Based on the mutagenesis result and x-ray structures, a dynamic model for the catalytic process of FabZ during the biosynthesis of fatty acids can be devised. Firstly, ␤-hydroxyacyl-ACP binds to HpFabZ with Tyr-100 in an open conformation, and the substrate ␤-hydroxyacyl-ACP enters into the tunnel for dehydration. After the reaction is finished, Tyr-100 turns to the closed conformation and stimulates the dehydration product trans-2-acyl-ACP to leave FabZ, and the next molecule of ␤-hydroxyacyl-ACP binds to FabZ for dehydration.
Enzymes encoded in the FAS II pathway have recently been appreciated as promising targets for discovering anti-bacterial and anti-parasitemia (malarial) drugs (3,8,23). The currently available FabZ inhibitors are NAS derivatives originally found as EcFabA inhibitors and flavonoids against PaFabZ (5,19). So far, no inhibitors have been reported for FabZs of other microorganisms. We have discovered four HpFabZ inhibitors through random screening of our in-house chemical library (Fig. 4, A and D and supplemental Fig. S4), and for the first time, we obtained the structures of HpFabZ complexed with two of the inhibitors (compounds 1 and 2). In addition to residing in the grooves on the surface, one molecule of compound 1 has entered into the center of the tunnel. This indicates that the enzymatic activity of HpFabZ can be inhibited via two ways, either occupying the entrance of the tunnel or plugging the tunnel to prevent the substrate from accessing the active site.