Crystal Structure of the Mycobacterium tuberculosis Enoyl-ACP Reductase, InhA, in Complex with NAD+ and a C16 Fatty Acyl Substrate*

Enoyl-ACP reductases participate in fatty acid biosynthesis by utilizing NADH to reduce the transdouble bond between positions C2 and C3 of a fatty acyl chain linked to the acyl carrier protein. The enoyl-ACP reductase fromMycobacterium tuberculosis, known as InhA, is a member of an unusual FAS-II system that prefers longer chain fatty acyl substrates for the purpose of synthesizing mycolic acids, a major component of mycobacterial cell walls. The crystal structure of InhA in complex with NAD+ and a C16 fatty acyl substrate,trans-2-hexadecenoyl-(N-acetylcysteamine)-thioester, reveals that the substrate binds in a general “U-shaped” conformation, with the trans double bond positioned directly adjacent to the nicotinamide ring of NAD+. The side chain of Tyr158 directly interacts with the thioester carbonyl oxygen of the C16 fatty acyl substrate and therefore could help stabilize the enolate intermediate, proposed to form during substrate catalysis. Hydrophobic residues, primarily from the substrate binding loop (residues 196–219), engulf the fatty acyl chain portion of the substrate. The substrate binding loop of InhA is longer than that of other enoyl-ACP reductases and creates a deeper substrate binding crevice, consistent with the ability of InhA to recognize longer chain fatty acyl substrates.

Mammals, birds, and yeast produce fatty acids via a FAS-I system, a biosynthetic pathway in which each enzymatic activity resides on the single polypeptide chain of a very large multifunctional enzyme (1)(2)(3). In contrast, most plants and bacteria utilize a FAS-II system in which each of the enzymatic activities corresponds to individual polypeptides (4 -6). The best characterized FAS-II system is that of Escherichia coli, which includes ␤-ketoacyl-ACP synthases (FabB, FabF, and FabH), a ␤-ketoacyl-ACP reductase (FabG), ␤-hydroxyacyl-ACP dehydrases (FabA and FabZ), and an enoyl-ACP reductase (currently known as FabI and formerly known as EnvM). The E. coli FAS-II system initializes fatty acid biosynthesis by condensing acetyl and malonyl substrates, and then it continues the process by elongating the resulting product through the consecutive addition of C2 units, until the fatty acyl chain reaches a length of about 16 carbons.
Some bacteria, such as mycobacteria, possess both a FAS-I and a FAS-II system (7). The mycobacterial FAS-I system displays a bimodal distribution of products centered on C16 and C24 -C26 (7,8). This FAS-II system prefers C16 as a starting substrate (7) and can extend up to C56 (9), indicating that the mycobacterial FAS-II system utilizes the products of the FAS-I system as primers to extend fatty acyl chain lengths even further. The longer chain products of the FAS-II system are the precursors of mycolic acids. Mycolic acids are long chain ␣-alkyl-␤-hydroxy fatty acids, which are a major component of mycobacterial cell walls (10,11).
Isoniazid, a drug used as a first-line antibiotic for the treatment of Mycobacterium tuberculosis infections for the past 40 years, is known to inhibit mycolic acid biosynthesis (12)(13)(14)(15). More recently, isoniazid was shown to prevent radiolabel from being incorporated into the chain extension of C24 -C26 fatty acyl substrates (16 -20), suggesting that the target of isoniazid action is within the mycobacterial FAS-II system. Using a genetic approach to isolate the isoniazid target, a single open reading frame was identified (referred to as inhA) in which transformation of the wild-type gene carried on a multicopy plasmid or allelic exchange with a single amino acid substitution mutant (Ser 94 to Ala) was sufficient to confer isoniazid resistance in Mycobacterium smegmatis (21).
The protein encoded by the inhA gene, referred to as InhA, has a similar amino acid sequence to two previously characterized enoyl-ACP reductases, namely FabI from E. coli (28% identity) (22)(23)(24)(25), and ENR 1 from Brassica napus (oilseed rape) (23% identity) (26,27). Further analysis revealed that InhA catalyzes the NADH-dependent reduction of the trans double bond between positions C2 and C3 of fatty acyl substrates (28). In addition, InhA prefers fatty acyl substrates of C16 or greater, consistent with its being a member of the mycobacterial FAS-II system (28).
The connection between isoniazid action and InhA inhibition is somewhat complicated. Isoniazid is a pro-drug that must be converted, via a mycobacterial catalase-peroxidase (KatG), into an activated form of the drug (29 -32) (for recent reviews, see Refs. [33][34][35]. The activated form of isoniazid (suspected to be an isonicotinic-acyl radical) becomes covalently attached to the nicotinamide ring of the NADH bound within the active site of InhA, creating an NADH adduct that acts as a very tightly bound inhibitor (36). In contrast, the S94A mutation of InhA, which is resistant to the effects of isoniazid, has a decreased affinity for NADH (28,37).
The World Health Organization has estimated that one-third of the world's population, nearly 2 billion people, is infected with tuberculosis. Tuberculosis kills more adults now than all other infectious diseases combined, and it is the leading cause of death in HIV-positive individuals. Data from the World Health Organization-sponsored Global Project on Anti-tuberculosis Drug Resistance Surveillance indicates that 36% of tuberculosis patients are infected with a strain that is resistant to traditional antibiotics (isoniazid or rifampicin). Even though mutations within the inhA gene are known to facilitate isoniazid resistance (21, 38 -41), InhA remains a good candidate for drug design. Several reasons for this are as follows: (i) the vast majority of the mutations found in isoniazid-resistant clinical isolates are associated with the isoniazid activator (the KatG catalase-peroxidase) (38 -41); (ii) only one enoyl-ACP reductase is found in M. tuberculosis, unlike some of the other enzymes of bacterial FAS-II systems (42); and (iii) the longer substrate chain length specificity of InhA distinguishes it from the enoyl-ACP reductases from other sources, such as the enoyl-ACP reductase component of the human FAS-I system.
To facilitate the drug design process and increase our understanding of the mechanism of an enoyl-ACP reductase reaction, full characterization of the active site of InhA is now available. In an initial report, the crystal structure of InhA in complex with NADH was determined at 2.5-Å resolution (37), which revealed the putative location of the fatty acyl substrate binding pocket, lying adjacent to the bound NADH (28). We describe here the crystal structure of InhA in complex with NAD ϩ and a C16 fatty acyl substrate (trans-2-hexadecenoyl-(N-acetylcysteamine)-thioester). The substrate contains a fatty acyl chain of 16 carbons, with a trans double bond between positions C2 and C3 (trans-2-hexadecenoyl moiety), which is attached through a thioester linkage to a small portion of the phosphopantetheine arm of the acyl carrier (N-acetylcysteamine moiety). The chemical formula is CH 3 -(CH 2 ) 12 -CHϭCH-CO-S-CH 2 -CH 2 -NH-CO-CH 3 . This compound is not a natural biological substrate of InhA; however, it and long chain fatty acyl derivatives of coenzyme A are functional substrates in vitro.

EXPERIMENTAL PROCEDURES
Synthesis of C16 Fatty Acyl Substrate-Trans-2-hexadecenoyl-(Nacetylcysteamine)-thioester was prepared in two stages. In the first step, trans-2-hexadecenoic acid was activated by reaction with 1 equivalent of ethyl chloroformate and 1 equivalent of diisopropylethylamine in tert-butyl methyl ether. The reaction mixture was stirred at room temperature overnight and then filtered and concentrated, giving the mixed anhydride as an oil. In the second step, the thioester was formed by the coupling of N-acetylcysteamine with the mixed anhydride. To the oil was added a solution of N-acetylcysteamine in ethanol, ethyl acetate, and 0.3 M aqueous sodium bicarbonate (1:1:1 v/v/v). After the reaction mixture was stirred for 4 h at room temperature, the solvents were removed by rotary evaporation. The product was purified by flash chromatography on silica gel using hexane, ethyl acetate, and acetic acid (75:25:1 v/v/v) as the eluting agent. The product was obtained in 37% overall yield and the structure was confirmed by 1 H NMR spectroscopy.
Crystallization of the Ternary Complex-Expression of the M. tuberculosis inhA gene within an E. coli host, purification of the resulting soluble InhA protein, and production of hexagonal crystals (space group P6 2 22) containing InhA in complex with NADH have been reported previously (28,37). Monoclinic crystals (space group C2) of the ternary complex (InhA, NAD ϩ , and C16 fatty acyl substrate) were produced under new conditions, using the hanging drop vapor diffusion method. The drop was created by combining an equal amount of well solution (10% polyethylene glycol-4000, 6% Me 2 SO, 100 mM ammonium acetate, and 100 mM ADA, pH 6.8) with a concentrated protein mixture. The protein mixture was produced by first solubilizing the C16 fatty acyl substrate in Me 2 SO (600 M stock solution) and then slowly adding 1 ml of stock solution at room temperature to 100 ml of a dilute solution of InhA (0.1 mg/ml ϭ 3 M) in the presence of 600 M NAD ϩ . The final molar ratio of the mixture was 1:200:2 for InhA, NAD ϩ , and C16 fatty acyl substrates, respectively, and the final concentration of Me 2 SO was 1% (v/v). The mixture was concentrated at 4°C with an Amicon Centricon-10 until the concentration of InhA reached 10 mg/ml. X-ray Diffraction Data Collection-X-ray diffraction data were collected to 2.8-Å resolution at room temperature using a MacScience DIP2030 image plate detector with double-focusing mirrors coupled to a Rigaku x-ray generator utilizing a copper rotating anode (CuK␣ wavelength ϭ 1.54 Å). A 0.0075-mm nickel filter was placed in line before the mirrors, and the diameter of the x-ray beam collimator was 0.5 mm. The detector was placed 230 mm from the crystal with no offset in the 2 angle. Each frame of data was exposed for 15 min and consisted of a 1.5°rotation of the crystal. The data from two crystals (size 0.4 ϫ 0.4 ϫ 0.1 mm) were processed and merged using the DENZO/ Scalepack software package (43), and a 3 cutoff was applied to the final structure factors. Data collection statistics are listed in Table I.
Initial Phase Determination and Model Refinement-All calculations were performed with the XPLOR software package (44). A single subunit of InhA (protein coordinates only) derived from the hexagonal crystal form (Protein Data Bank entry 1eny) was used as a search model with the molecular replacement technique. Six subunits were found in the asymmetric unit of the monoclinic crystal form, corresponding to a solvent content of 53%. Four of the subunits form a homotetramer based on noncrystallographic (local) 2-fold symmetry axes, and the remaining two subunits form a second homotetramer based on a combination of both noncrystallographic (local) and crystallographic 2-fold symmetry axes.
Subsequent to rigid body refinement (R factor ϭ 30.2%, R free ϭ 38.3%), noncrystallographic (NCS) restraints were applied to both the atomic positions and individual temperature factors. After simulated annealing refinement (R factor ϭ 25.6%, R free ϭ 36.0%), the NAD ϩ and C16 fatty acyl substrate could be identified in a difference Fourier (F o -F c ) electron density map, confirming the correctness of the molecular replacement solution. Subsequent to conventional positional refinement in the presence of NAD ϩ and the C16 fatty acyl substrate (R factor ϭ 23.4%, R free ϭ 35.6%), several rounds of conventional positional refinement were performed during the addition of 682 ordered water molecules (R factor ϭ 21.7%, R free ϭ 34.4%). Model refinement statistics are listed in Table II, and an omit electron density map (XPLOR, Ref. 44) covering the NAD ϩ and C16 fatty acyl substrates is shown in Fig. 2A.
TABLE I X-ray diffraction data statistics Completeness ϭ number of F observed /number of F expected ) ϫ 100; R sym ϭ ⌺ԽI Ϫ ͗I͘Խ/⌺I) ϫ 100; ͗I/I͘ ϭ ⌺ (I/I)/number of I; R factor ϭ (⌺ (Խ F observed Ϫ F calculatedԽ)/⌺ ( F observed )) ϫ 100; R free ϭ R value of portion of data omitted from model refinement. I, intensity; F, amplitude. Analytical Size-exclusion Chromatography-The quaternary structure of InhA was investigated by determining the molecular weight of InhA in aqueous solution using analytical size-exclusion chromatography. A prepacked Pharmacia Superdex-200 HR-10/30 column (separation range 10 -600 kDa) was equilibrated in 150 mM NaCl and 50 mM HEPES, pH 7.5. The flow rate was set at 0.25 ml/min and controlled by a BioCAD SPRINT system from PerSeptive Biosystems (Cambridge, MA). The column was calibrated by individually applying a ϳ5 mg/ml solution of each standard protein obtained from a Pharmacia gel-filtration calibration kit to the column and then calculating a standard line by plotting the resulting elution volumes versus known molecular mass information. An ϳ5 mg/ml solution of InhA was applied to the column, and the molecular mass of InhA was estimated by comparing the value of its column elution volume with the standard line.

RESULTS AND DISCUSSION
Overall Tertiary and Quaternary Structure of InhA-InhA is a member of the short chain dehydrogenase/reductase (SDR) family of enzymes (45,46). The main characteristic of this family is a polypeptide backbone topology in which each subunit consists of a single domain with a central core that contains a Rossmann fold supporting an NADH binding site. Within InhA, several ␣-helices and ␤-strands of the Rossmann fold extend beyond the NADH binding site, creating a deep crevice for the fatty acyl substrate (28,37). Part of this extension, referred to here as the "substrate binding loop" (residues 196 -219), consists of two perpendicular ␣-helices that form one side of the fatty acyl substrate binding crevice (Fig. 1).
Analytical size-exclusion chromatography demonstrates that InhA is a homotetramer in aqueous solution. The crystal structure of InhA revealed that the homotetramer is a molecule possessing internal 222 symmetry. Three perpendicular twofold symmetry axes intersect in the center of the molecule, creating a molecule where each subunit is related to every other subunit by one of the perpendicular 2-fold symmetry axes (Fig. 1B). According to the DSSP program (47) (and the color code of Fig. 1B), the amount of buried surface area between subunit 1 (green) and subunit 2 (purple) is 1550 Å 2 (or ϳ12% of a single subunit), between subunit 1 and subunit 3 (cyan) is 700 Å 2 (or ϳ5% of a single subunit), and between subunit 1 and subunit 4 (magenta) is 1500 Å 2 (or ϳ12% of a single subunit). Therefore, the total buried surface area for any given subunit is 3750 Å 2 (or ϳ29%), indicating that a fairly large portion of each subunit of the homotetramer is involved in intramolecular contacts. Furthermore, the organization of the InhA homotetramer is identical to that of the homotetramers found in the crystal structures of two other enoyl-ACP reductases, namely E. coli FabI (1.6 Å r.m.s. deviation in ␣-carbon position over 848 residues) (48) and B. napus ENR (a 1.6-Å r.m.s. deviation in ␣-carbon position over 632 residues) (49).
Location and Conformation of the InhA-bound C16 Fatty Acyl Substrate-Within the orientation of Fig. 1A, the bound NAD ϩ sits on top of the shelf created by the Rossmann fold, and the C16 fatty acyl substrate sits on top of the NAD ϩ and is held in place by the substrate binding loop. The InhA-bound C16 fatty acyl substrate folds into a general "U-shaped" conformation (Fig. 2). This conformation is not unusual for protein-bound fatty acids. It has been seen before in the crystal structure of human muscle fatty acid binding protein in complex with several different long chain fatty acids (50).
The fatty acyl substrate binding crevice of InhA has an oval-shaped cavity with the approximate dimensions 16 ϫ 13 ϫ 7 Å. One side of the cavity is completely open and exposed to solvent (Fig. 2, B and C, major portal on right side), whereas the other side contains a small opening (Fig. 2, B and C, minor portal on upper left side). The N-acetylcysteamine portion of the substrate faces the major portal, which is consistent with the requirement that this region of the natural substrate should face outward, toward the solvent, when attached to the acyl carrier protein.
The substrate binding crevice of InhA must accommodate fatty acyl substrates that are longer than C16 in order to produce the precursors of mycolic acids (28). The major and minor portals may provide some assistance with substrate binding. For example, the -end of the fatty acyl substrate (position C16) faces the major portal, which could allow chain extensions beyond C16 (of natural substrates) to interact with the acyl carrier protein. Alternatively, the upper edge of the U-shaped turn (positions C8 -C10) faces the minor portal, which could permit the fatty acyl chain to thread out through the minor portal and then loop back in, so that the -end of the fatty acyl chain is still anchored inside the binding crevice.
The trans double bond, between positions C2 and C3 of the fatty acyl chain, faces the closed end of the substrate binding crevice (Fig. 2B, lower left) and is located directly adjacent to the nicotinamide ring of NAD ϩ . A comparison of the position of the nicotinamide ring between the binary complex (InhA and NADH) (37) and the ternary complex (InhA, NAD ϩ , and C16 fatty acyl substrate) shows no significant difference (only an 0.7-Å r.m.s. deviation over 9 atoms), indicating that NAD ϩ is an appropriate model for NADH at this resolution. Furthermore, the C3 position of the bound fatty acyl chain is closer (4.8Å) than the C2 position (5.1Å) to the C4 position of the nicotinamide ring of NAD ϩ . These distances are consistent with the previous proposal that hydride ion transfer is directed from the C4 position of the nicotinamide ring of NADH to the C3 position of the fatty acyl chain during substrate catalysis (28).
The C2 and C3 positions of the fatty acyl chain contain the flat (dihedral ϳ180°) trans double bond, whereas positions C4 -C8 form the turn of the U-shaped conformation ( Fig. 2A). Considering that the trans double bond is so close to the turn, it is tempting to speculate that during substrate catalysis, the active site may impose a slightly tighter U-shaped conformation upon the substrate by adding a twist to the substrate double bond. This requirement could facilitate the conversion from planar carbons to tetrahedral carbons through conformational strain of the trans double bond.
Specific Interactions between Bound Substrate and InhA Active Site-Using a 1.4-Å radius probe, the Connolly solvent accessible surface area (51) of the fatty acyl substrate binding crevice is approximately 670 Å 2 , with 72% of the area covered by nonpolar atoms and the remaining 28% of the area covered by polar atoms (Fig. 2B). This surface incorporates one buried water molecule, which is within hydrogen bonding distance of the fatty acyl substrate thioester sulfur. All hydrogen bonds and van der Waals contact distances between the C16 fatty acyl substrate and the InhA active site are listed in Table III.
The fatty acyl chain portion of the substrate is completely surrounded by the side chains of hydrophobic residues (Fig. 3). The majority of these hydrophobic side chains reside on the substrate binding loop (Ala 198 , Met 199 , Ala 201 , Ile 202 , Leu 207 , Ile 215 , Leu 218 ). Additional hydrophobic residues, having side There is only one direct hydrogen bond between the fatty acyl substrate and the InhA protein, namely between the thioester carbonyl oxygen and the side chain hydroxyl oxygen of Tyr 158 (Fig. 4A), as Tyr 158 is conserved in both the bacterial FabIs and plant ENRs, this interaction is likely to be a key feature of fatty acyl substrate binding common to other enoyl-ACP reductases.
Two additional hydrogen bonds occur between the fatty acyl substrate and NAD ϩ . These are between the amide nitrogen of the N-acetylcysteamine portion of the substrate and a phosphate oxygen of NAD ϩ and between the substrate thioester carbonyl oxygen and the 2Ј-hydroxyl oxygen of the nicotinamide ribose of NAD ϩ . Furthermore, there is an additional hydrogen bond between the substrate thioester sulfur and an ordered water molecule held by a phosphate oxygen of NAD ϩ and the side chain of Thr 196 (Fig. 4A).
Substrate Binding Loop Length Determines Substrate Binding Crevice Depth-An interesting correlation between substrate binding loop length and fatty acyl substrate chain length specificity can be inferred from amino acid sequence alignment of the known FAS-II enoyl-ACP reductases. Enoyl-ACP reductases that prefer shorter chain substrates (C2-C16) have shorter substrate binding loops (13 residues) (such as E. coli FabI and B. napus ENR), whereas longer substrate binding loops are seen in those enzymes that use longer chain substrates (C16 -C56). (23 residues) (such as M. smegmatis and M. tuberculosis InhA) (Fig. 5). This finding suggests that the size of the substrate binding loop is a primary determinant of the enzymes ability to distinguish between shorter versus longer chain substrates.
Superposition of the crystal structure of an E. coli FabI ternary complex (FabI, NAD ϩ , and a benzodiazaborine drug) (48) with that of the M. tuberculosis InhA ternary complex (InhA, NAD ϩ , and C16 fatty acyl substrate) demonstrates that there is a significant difference between these two enzymes with respect to the location of their substrate binding loops (Fig. 1C). The longer substrate binding loop of InhA creates a substrate binding crevice with more depth than that of FabI. The distance from the bottom of the C16 fatty acyl substrate to the top of the substrate binding loop is 18 Å for the InhA ternary complex, and it is only 10 Å for FabI superimposed onto the InhA ternary complex.
The crystal structure of the E. coli FabI binary complex (FabI and NAD ϩ ) indicates that the substrate binding loop is disordered in the absence of a fatty acyl substrate (48). Upon binding of a benzodiazaborine inhibitor, the shorter substrate binding loop is ordered, and the substrate binding loop makes direct contact with the bound inhibitor (48). The shape of the FabI substrate binding crevice, which results from the ordered substrate binding loop, is similar to InhA, in that both a major and minor portal are retained (Fig. 2C). Interestingly, even though a fatty acyl substrate is not present, the resulting binding crevice is continuous from one portal to the other, appearing as an open vertical tube, which lies adjacent to the nicotinamide ring of NAD ϩ . When the InhA ternary complex (InhA, NAD ϩ , and C16 fatty acyl substrate) is superimposed onto the FabI ternary complex (FabI, NAD ϩ , and a benzodiazaborine drug), the thioester and trans double bond portion of the C16 fatty acyl substrate overlap with the diazaborine drug, whereas the turn of the fatty acyl substrate (positions C4 -C8) fits into the open vertical tube (Fig. 2C).
The position of the shorter FabI substrate binding loop occludes the majority of the fatty acyl chain beyond C8. If this conformation of the substrate binding loop is indicative of how FabI interacts with fatty acyl substrates, then the substrates would not be accommodated in a U-shaped conformation, and chain extensions beyond C8 would most likely be directed out the minor portal. This situation would limit the ability of FabI to bind only shorter chain fatty acyl substrates.
In contrast, the geometry of the ␣-helices of the longer substrate binding loop of InhA would sterically hinder the loop from folding downward into the substrate binding cavity, leaving a larger opening for the fatty acyl substrate. The longer substrate binding loop of InhA would be prevented from making proper contact with shorter chain fatty acyl substrates and therefore would provide InhA with selection criteria against shorter chain substrates. The deeper substrate binding crevice of InhA would only be appropriate for binding of longer chain fatty acyl substrates, consistent with the substrate specificity of the enzyme and the role InhA plays in mycolic acid biosynthesis (28).
Conformational Changes within InhA upon Substrate Binding-A comparison of the structure of the binary complex (InhA and NADH) (37) with that of the ternary complex (InhA, NAD ϩ , and C16 fatty acyl substrate) clearly identifies two major conformational changes in the protein upon fatty acyl substrate binding. The substrate binding loop (residues 196 -219) was shifted 4.0 Å, and the side chain of Tyr 158 rotated by  different colored ribbon (green, purple, cyan, and magenta). The substrate binding loops, bound NAD ϩ , and bound C16 fatty acyl substrates are shown with the same coloring as in A. Three perpendicular 2-fold symmetry axes intersect in the center of the homotetramer, creating a molecule with internal 222 symmetry. C, superposition of the backbone of E. coli short chain substrate enoyl-ACP reductase (FabI) (Protein Data Bank entry 1dfg) (dark blue) with that of M. tuberculosis long chain substrate enoyl-ACP reductase (InhA) (green) demonstrates that the location of the substrate binding loop (yellow, InhA; cyan, FabI) differs between the two structures. Panels A and B were produced using the MOLSCRIPT program (55) coupled to the RASTER-3D program (56), and panel C was produced using the SPOCK program (57). 60° (Fig. 4B). Together these changes act to widen the substrate binding cavity to accommodate the bound C16 fatty acyl substrate.
The substrate binding loops are located on the outer surface of the homotetramer (Fig. 1B), and widening of the substrate binding cavity is accomplished when the loops push outward toward the solvent. In addition, no hydrogen bonds are broken during this conformational change. Furthermore, the loops are not involved in contacts between subunits of the homotetramer, and loop movement does not disrupt organization of the homotetramer (0.4-Å r.m.s. deviation in ␣-carbon position over 972 residues between the binary and ternary complexes when the substrate binding loops are omitted from consideration). This lack of major structural reorganization indicates that the FIG. 2. Conformation of bound substrate and shape of binding crevice. A, atomic coordinates of the InhA-bound NAD ϩ and C16 fatty acyl substrate, trans-2-hexadecenoyl-(N-acetylcysteamine)-thioester, superimposed onto the final NCS-averaged simulated annealing "omit" electron density (44) map contoured at 0.8 . The different atom types are represented by unique colors: carbon (gray), oxygen (red), nitrogen (blue), sulfur (yellow), and phosphorus (magenta). The C16 fatty acyl substrate folds into a general U-shaped conformation, with the trans double bond located directly over the nicotinamide ring of NAD ϩ . B, solvent-accessible surface of the M. tuberculosis long chain substrate enoyl-ACP reductase (InhA) binding crevice shown in black dots. A portion of the bound NAD ϩ (nicotinamide, nicotinamide ribose, and phosphates) is shown in green, and the bound C16 fatty acyl substrate is shown in red. C, solvent-accessible surface of the E. coli short chain substrate enoyl-ACP reductase (FabI) binding crevice shown in black dots. A portion of the bound NAD ϩ (nicotinamide, nicotinamide ribose, and phosphates) is shown in green, and the bound benzodiazaborine drug is shown in cyan (Protein Data Bank entry 1dfg). The C16 fatty acyl substrate from the InhA ternary complex is superimposed onto the FabI surface. This figure was produced using the SPOCK program (57). substrate binding loop is somewhat flexible and can move independently and therefore may provide the active site with additional adaptability to accommodate substrates of varying chain length during the fatty acyl chain elongation process. SDR Family Catalytic Triad and Mechanism of InhA Catalysis-In addition to FAS-II system enoyl-ACP reductases, numerous steroid dehydrogenases have been shown to be members of the SDR family (45,46). The 7␣-hydroxysteroid dehydrogenase (7␣-HSDH), catalyzes the NAD ϩ -dependent dehydrogenation of the hydroxyl group at position seven on the steroid skeleton of bile acids (52). Amino acid sequence alignment of M. tuberculosis InhA with E. coli 7␣-HSDH shows 22% identity, and superposition of their crystal structures shows that they share a very similar backbone topology (1.5-Å r.m.s. deviation in ␣-carbon position over 187 residues for a single subunit). In addition, the bile acid substrate binds to E. coli 7␣-HSDH in nearly the same location that the C16 fatty acyl substrate binds to InhA (53). The SDR family catalytic triad of E. coli 7␣-HSDH is Ser 146 -Tyr 159 -Lys 163 . The tyrosine and serine residues directly interact with the bile acid substrate hydroxyl group (produce oxo group), whereas the lysine residue interacts with the 3Ј-hydroxyl group of the nicotinamide ribose of the substrate NAD ϩ (product NADH) (Fig. 6A).
The tyrosine and lysine residues of the SDR family catalytic triad are strictly conserved and therefore mark the location of a consensus sequence (Tyr-X-X-X-Lys), where separation between the tyrosine and lysine can vary from three to seven residues (54). Among the various crystal structures of members of the SDR family, the location of the lysine residue is nearly identical, whereas the position of the tyrosine residues has a tendency to vary somewhat among different SDR family members (46). This trend suggests that the function of the lysine residue is to hold the NADH in place. The function of the tyrosine residue is to directly interact with the substrate, and that variation of the position of the tyrosine residue among different SDR family members allows the tyrosine residue to complement and accommodate different types of substrates. The structurally analogous SDR family catalytic triad of InhA is Phe 149 -Tyr 158 -Lys 165 (Fig. 6B). The fact that the side chain of Tyr 158 interacts with the C16 fatty acyl substrate thioester carbonyl oxygen and that the side chain of Lys 165 interacts with the 3Ј-hydroxyl oxygen of the nicotinamide ribose of NAD ϩ is consistent with interactions in other SDR family members.
The proposed mechanism of how InhA catalyzes the reduction of the trans double bond between positions C2 and C3 of the fatty acyl substrate consists of the formation of an enolate intermediate through the direct transfer of a hydride ion from NADH to position C3 of the substrate, which is followed by protonation of position C2 (28). In the past, it has been proposed that the side chain of Lys 165 stabilizes the enolate intermediate (28). However, in light of the above SDR family comparison and the fact that the side chain hydroxyl oxygen of Tyr 158 forms a direct hydrogen bond with the fatty acyl substrate thioester carbonyl oxygen, we propose that the side chain of Tyr 158 assumes this role. In addition, the crystal structure of the InhA ternary complex suggests that the 2Ј-hydroxyl oxygen of the nicotinamide ribose of NADH may assist in the stabilization of the enolate intermediate because it forms a hydrogen bond to the fatty acyl substrate thioester carbonyl oxygen as well (Fig. 4A).
The role of the remaining residue of the catalytic triad (Ser 146 in E. coli 7␣-HSDH and Phe 149 in M. tuberculosis InhA) is more ambiguous. In terms of the SDR family catalytic triad, this residue is the only component that distinguishes a dehydrogenase (7␣-HSDH) from a reductase (InhA). In terms of function, the main feature that distinguishes these two types of enzymes is the position on the substrate at which hydride transfer occurs (carbonyl carbon for the dehydrogenase and enoyl carbon for the reductase). Furthermore, the fatty acyl substrate of InhA also contains a thioester carbonyl carbon that needs to be protected from being reduced. Because Phe 149 of InhA is fairly close (4 Å) to the nicotinamide ring of NAD ϩ , it is tempting to speculate that this aromatic residue may help direct hydride ion transfer to the correct location on the fatty acyl substrate by acting as an intermediate that passes the hydride ion from NADH to the substrate. The importance of the aromatic character of residue 149 is conserved in either a phenylalanine (mycobacterial InhAs) or a tyrosine (bacterial FabIs and plant ENRs).
The source of protons for reduction at position C2 of the fatty acyl substrate is not clear from our crystal structure. There are no polar/ionizable residues close enough to position C2 to be FIG. 5. Sequence alignment of FAS-II system enoyl-ACP reductases. The amino acid sequence information was obtained from GenBank TM , using the following accession numbers for the bacterial FabIs (E. coli, X78733; Salmonella typhimurium, M31806; and Helicobacter pylori, AE000539), the plant ENRs (Brassica napus, S60064; Arabidopsis thaliana, Y13860; and Nicotiana tabacum, Y13862), and the mycobacterial InhAs (Mycobacterium avium, AF002133; Mycobacterium bovis, U41388; Mycobacterium smegmatis, U02530; and M. tuberculosis, U02492). The numbering above the alignment corresponds to the M. tuberculosis InhA sequence and the lower-case letters below the alignment mark the location of ␣-helices (h) and ␤-strands (s). The substrate binding loop of the long chain substrate mycobacterial InhAs (residues 196 -219) is larger than that of the short chain substrate bacterial FabIs and plant ENRs. potential candidates, suggesting that the proton may in fact come from the solvent. The location of the C2 position of the fatty acyl substrate, at opposite ends of the major and minor solvent portals (Fig. 2B), eliminates these portals as a possible sources of proton donation. However, there is a hydrogenbonded network of ordered water molecules that leads from the closed end of every substrate binding crevice (Fig. 2B, lower  left) to the center of the homotetramer. The side chain of Phe 149 acts as a barrier between the active site and the water channel. Rotation of the side chain of Phe 149 would, however, result in exposure of the water channel to the active site. Within the crystal structure of an InhA binary complex (InhA and isonicotinic-acyl-NADH) (36), the bound form of the drug isoniazid causes the side chain of Phe 149 to move away from the water channel, indicating that rotation of the side chain of Phe 149 is sterically possible.