The Tuberculosis Prodrug Isoniazid Bound to Activating Peroxidases*

Isoniazid (INH, isonicotinic acid hydrazine) is one of only two therapeutic agents effective in treating tuberculosis. This prodrug is activated by the heme enzyme catalase peroxidase (KatG) endogenous to Mycobacterium tuberculosis but the mechanism of activation is poorly understood, in part because the binding interaction has not been properly established. The class I peroxidases ascorbate peroxidase (APX) and cytochrome c peroxidase (CcP) have active site structures very similar to KatG and are also capable of activating isoniazid. We report here the first crystal structures of complexes of isoniazid bound to APX and CcP. These are the first structures of isoniazid bound to any activating enzymes. The structures show that isoniazid binds close to the δ-heme edge in both APX and CcP, although the precise binding orientation varies slightly in the two cases. A second binding site for INH is found in APX at the γ-heme edge close to the established ascorbate binding site, indicating that the γ-heme edge can also support the binding of aromatic substrates. We also show that in an active site mutant of soybean APX (W41A) INH can bind directly to the heme iron to become an inhibitor and in a different mode when the distal histidine is replaced by alanine (H42A). These structures provide the first unambiguous evidence for the location of the isoniazid binding site in the class I peroxidases and provide rationalization of isoniazid resistance in naturally occurring KatG mutant strains of M. tuberculosis.

8 million people per year are diagnosed with tuberculosis in both developed and developing nations, resulting in more than 2 million deaths per year (2). However, despite the worldwide success of INH in treating tuberculosis since the 1950s, very little is understood about the mode of action or the mechanism of activation of this prodrug. In recent times the emergence of INH-and multidrug-resistant strains of tuberculosis and the scale of the tuberculosis epidemic has highlighted the need for this fundamental lack of knowledge to be addressed with a view to discovering new antimicrobial targets and treatment regimens.
In vitro and in vivo studies have established that a catalase peroxidase, KatG, which is endogenous to M. tuberculosis (the principal causative agent of tuberculosis), is essential for prodrug activation (3). KatGs are bifunctional heme enzymes that exhibit both catalase activity and broad spectrum peroxidatic activity comparable with monofunctional peroxidases (4,5). The peroxidatic activity involves the formation of an oxidized ferryl intermediate (Equation 1, Compound I) that is subsequently reduced by substrate. This reduction usually occurs in two successive single-electron transfer steps as follows (Equations 2 and 3) (where P ϭ peroxidase, HS ϭ substrate, S ⅐ ϭ 1-electron oxidized form of substrate).
KatG is a dimeric heme-containing enzyme of ϳ160-kDa molecular mass; its structure ( Fig. 1A) (6) and function place it in the class I superfamily of peroxidases along with cytochrome c peroxidase (CcP) (Fig. 1B) and ascorbate peroxidase (APX) (Fig. 1C) (7). The determination of the crystal structure of M. tuberculosis KatG (6) provided a major breakthrough in understanding the molecular mechanism of INH activation and highlighted the remarkable structural similarity both in the overall structures and at the active sites of KatG and the monofunctional class I peroxidases such as CcP (8) (1). We establish that the INH occupies the ␦-meso edge of the heme, and we also show that in an active site mutant of sAPX (W41A) INH can bind directly to the heme iron to become an inhibitor. Furthermore, we show in a complex of INH and the H42A mutant of sAPX that the indole nitrogen of Trp-41 is important for ligand orientation. The data are discussed in the context of our current understanding of INH binding and activation and provide an explanation for the molecular mechanism of mutation-acquired resistance.

EXPERIMENTAL PROCEDURES
Enzyme Expression and Purification-The Y39A/N184R mutant of recombinant CcP from Baker's yeast (optimized for crystallization) was prepared and isolated with modifications to SCHEME 1. A, the proposed catalytic mechanism of INH oxidation by the class 1 peroxidases. The first stage (steps 1 and 2) forms the isonicotinoyl radical, which is then thought to react with the liberated hydrazine to form the primary product isonicotinamide (38). The INH numbering scheme is also shown. B, the active site residues of KatG that are proposed to be involved in the catalytic mechanism of INH oxidation (1). Possible stabilizing interactions are shown as dotted lines, and the heme is represented as a rhombus. The equivalent residues in CcP and sAPX are as follows: Arg-104 ' Arg-48, Arg-38; Trp-107 ' Trp-51, Trp-41; His-108 ' His-52, His-42; Asp-137 ' Ser-81, Ala-70. Asp-137 of KatG is thought to play a major role in stabilizing the catalytic intermediate of INH oxidation.
published procedures (12). Recombinant cytosolic sAPX and the W41A and H42A mutants were prepared and isolated according to published procedures (13,14). All protein preparations were checked for homogeneity by SDS-PAGE.
Protein Crystallography-Crystals of CcP were prepared by microdialysis with 100 l of a 10-30 mg/ml solution of CcP in 500 mM potassium phosphate, pH 6.0, against 10 ml of 50 mM potassium phosphate, pH 6.0, containing 30% 2-methyl-2,4-pentanediol by volume. Crystals were grown at 4°C. Once formed, the crystals of CcP were soaked in mother liquor saturated with INH for 5 min prior to rapid cooling to 100 K.
Crystals of sAPX, sAPX (W41A), and sAPX (H42A) were prepared as described previously (9). Once formed, crystals were soaked in mother liquor containing INH (100 mM) overnight prior to freezing in liquid nitrogen for storage and transport.
Data Collection and Refinement-Diffraction data were collected for the CcP⅐INH complex in-house using a Rigaku RU2HB x-ray generator with copper anode and Xenocs multilayer optics and an R-Axis IV detector. Diffraction data were collected for sAPX⅐INH and sAPX (W41A) INH complexes FIGURE 1. Stereo diagrams of the key active site residues, ␦-heme edge hydrophobic binding pocket, and bound waters of the unliganded class I peroxidases discussed in this study. KatG (6) (A), CcP(B), and sAPX (9) (C). The protein residues are shown in green, the heme group in blue, and waters are represented as red spheres. on station ID14 -3 at the European Synchrotron Radiation Facility (ESRF) (Grenoble, France) using an ADSC Q4R detector. Data from sAPX⅐(H42A)⅐INH were also collected at ESRF using ID14 -1 and an ADSC Q4R detector. All data were collected at 100 K. Data were indexed, integrated, and scaled using MOSFLM (15) and SCALA (16). 5% of the data were flagged for the calculation of R free and excluded from subsequent refinement. Data collection statistics are shown in Table 1. The CcP⅐INH structure was refined from the 1.70 Å wild-type CcP structure (17)

Identification of the Hydrophobic Binding Pockets in CcP and sAPX-
Numerous studies have shown that aromatic substrate oxidation in the class I peroxidases occurs at the ␦-heme edge (6, 11, 23, 24). This site  is compared in KatG, CcP, and sAPX in Fig. 1. In the unliganded CcP structure the hydrophobic pocket itself is defined, as predicted from the crystal structure (17), by residues Pro-145, Ser-81, Ser-185, Arg-48, Trp-51, and His-52, with the "bottom" of the pocket being defined by pyrrole ring IV of the heme while one "side" is left open to solvent (Fig. 1B). Examination of electron density within this pocket in CcP reveals water molecules hydrogen-bonded within the pocket and leading out into bulk solvent. This open side forms a channel to allow the substrate to access the binding pocket.
The unliganded sAPX structure (Fig. 1C) shows an almost identical active site architecture with the hydrophobic pocket defined by Trp-41, His-42, Pro-132, Ser-173, and Ala-70, and again the bottom of the pocket is defined by pyrrole ring IV of the heme. Like CcP (17) and KatG (6), ordered water molecules occupy this pocket in sAPX with a substrate access channel leading to bulk solvent. Thus, the pockets in all three of these class 1 peroxidases are similar and available to bind hydrophobic aromatic substrates.
The Structures of INH Bound to CcP and sAPX-INH binds in place of four of the water molecules within the hydropho-bic pocket of CcP ( Fig. 2A). The structure allows hydrogen bonds between N2 of the substrate and the main chain oxygen of Pro-145, the carbonyl oxygen on the substrate and the side chain of Arg-48, and the pyridinyl nitrogen on the substrate and a water molecule leading out to bulk solvent. The guanidinium group of Arg-48 swings out to accommodate the substrate (C moves by 2 Å) and displaces a further water molecule. The overall position of the INH in the crystal structure reported here is similar to the NMR-derived model (Fig. 2B) (1) except for the interactions of N2 and N3. Notably, the crystal structure is inconsistent with a predicted (1) hydrogen bond between the globally conserved distal tryptophan (Trp-51 in the case of CcP) and N3 of INH.
The structure of sAPX in complex with INH also shows INH bound in the hydrophobic pocket at the ␦-meso heme edge (Fig. 3A). However, in this case the INH is rotated through ϳ90°compared with the orientation in CcP and the predicted sAPX⅐INH model (Fig.  3B) (1). The pyridine ring of INH is positioned directly over Ser-173, and the complex is locked into position via three hydrogen bonds, two of which are to the same residues that are used to bind INH in CcP. The main chain carbonyl of Pro-132 hydrogen bonds to N3 of the INH, the carbonyl oxygen of INH hydrogen bonds via a water molecule to the N⑀ group of Arg-48, and the N2 of INH is hydrogen-bonded to a water molecule.
Previous mechanistic studies have proposed a KatG-catalyzed activation mechanism that involves splitting of the C-N bond of the hydrazide moiety of INH, yielding a diazene that is stabilized by Asp-137 (Scheme 1) (1, 6). In M. tuberculosis KatG the acid group of Asp-137 is predicted to be located directly above the carbonyl of INH and is ideally orientated to provide a stabilizing acid-base interaction with the reacting INH. However, in CcP and sAPX, Asp-137 is replaced with Ser-81 and Ala-70, respectively. Ser-81 could still offer some rudimentary stabilization for the activated INH, whereas this is unlikely with Ala-70. This hypothesis is supported by the relative reaction rates of INH oxidation, KatG Ͼ ϾCcp ϾsAPX (1).  tions of KatG have been identified in INH-resistant strains (25), with one of the mutation hotspots being Ser-315 (mutated to asparagine, isoleucine, arginine, and glycine, but most commonly threonine) (26 -28). Ser-315 is conserved in both CcP and sAPX (Fig. 1) as Ser-185 and Ser-173, respectively. In the CcP⅐INH structure the C␣ of Ser-185 is ϳ4.0 Å from the aromatic ring of INH, and in sAPX the C␤ of Ser173 is ϳ3.8 Å away and located directly below it. All of these mutations would introduce considerably more steric bulk than serine, thereby placing the side chain of the mutated residue at position 315 closer than allowed Van der Waal's contact distances, blocking the binding of the INH molecule. These observations are consistent with predictions that these mutations produce a steric interference to INH binding (6,29).
Active site residues His-108 and Ala-110 are also frequently mutated in KatG sequenced from isolates demonstrating resistance to INH (30 -32). Mutation of active site residues of peroxidases has been shown (14) to introduce conformational mobility into the distal cavity, possibly promoting alternative inhibitive INH binding orientations. To probe this, we have determined the structures of INH bound to two active site mutations of sAPX (W41A and H42A). Removal of the distal tryptophan in the W41A mutant of sAPX opens up the distal cavity and imparts a degree of conformational mobility (14). Electron density from crystals of INH-soaked sAPX (W41A) shows a molecule of INH occupying the hydrophobic pocket in an identical position to that found in the wild-type sAPX⅐INH complex. A second INH molecule is also seen coordinated directly to the heme iron through the NH 2 of the acyl hydrazide, with the aromatic ring occupying the pocket created by removal of the indole of Trp-41 (Fig. 4A). This mode of binding has been seen previously (10) in APX for aromatic acids such as SHA, which is an inhibitor of APX peroxidase activity. However, coordination of INH to the iron is only possible when an additional pocket is created in the W41A mutant, whereas coordination of SHA is possible with Trp-41 in situ (10). This is because SHA forms a hydrogen bond between its iron coordinating O Ϫ and the NH of Trp-41 (Fig. 5). However, INH has an NH 2 group coordinated to the iron that is unable to hydrogen bond to the NH of Trp-41; furthermore, its bulk would also result in a steric clash. The removal of Trp-41 allows the INH to move around through 90°t o coordinate to the iron without this steric conflict (Fig. 5). Kinetic studies show that INH becomes an inhibitor of sAPX (W41A) peroxidase activity, determined by competition with guaiacol (k cat for guaiacol oxidation is 98.3 Ϯ 1.   The structure of INH bound to the H42A mutant of sAPX reveals a third binding orientation of INH within the distal cavity (Fig. 4B). In this structure the INH again occupies the ␦-heme edge but is rotated 90°anticlockwise compared with the orientation in sAPX. The removal of His-42 creates enough space for the aromatic ring of INH to sit directly above the iron and hydrogen bond through N1 to the NH of Trp-41. This provides further evidence that Trp-41 influences coordination to the heme iron. Although the cavity in sAPX⅐H42A can contain the INH molecule, the steric clash with Trp-41 (see Fig. 5) prevents coordination. Kinetic studies show no INH turnover by sAPX (H42A) (data not shown); however, it has been previously shown that peroxidase activity in this mutant (21) is severely reduced. A reduction in INH activation is consistent with the observation of INH resistance in M. tuberculosis strains with mutations at the distal His-108 residue of KatG (30,31).
Occupation of the Ascorbate Binding Site of sAPX by INH-The electron density maps of INH-soaked sAPX (Fig. 6) and mutants (data not shown) also show density consistent with a second INH molecule bound within the ascorbate binding pocket at the ␥-heme edge, which has previously been identified as the primary site of catalysis within the enzyme (9) (Fig.  6). The INH molecule binds in place of six well ordered water molecules that occupy the pocket in the absence of ascorbate. The NH 2 group of INH forms two hydrogen bonds, one to the side chain of Arg-172 and one via a water molecule to one of the heme propionate groups. The flexible Lys-31 also partially swings in from solvent to form a hydrogen bond to N1 of INH. Movement of Lys-31 on binding of ascorbate has also been observed (9). The role of the heme propionates has traditionally been believed to be in stabilizing the heme prosthetic group within the protein structure. However, there are now several examples where the heme propionates are involved in substrate or cofactor binding, notably in manganese peroxidase (33), ascorbate peroxidase (9), and in nitric-oxide synthase (34 -36); this has led to the suggestion (37) that there might be a broader role for the proprionates than merely keeping the heme in place. The data above add to this by indicating, for the first time, that the ␥-heme edge might also be capable of binding aromatic substrates.
In summary, we have provided unambiguous structural evidence for the location of INH binding in the class I peroxidases CcP and sAPX. These data expand on previous predictions and provide a molecular understanding of prodrug binding and activation. Furthermore, by comparing mutations engineered into our peroxidase models with naturally occurring INH-resistant variants of M. tuberculosis, we have identified three separate means of drug resistance: steric hindrance of the ␦-heme edge INH binding site, direct binding to the heme iron (and consequent inhibition of compound I formation), and the loss of peroxidatic activity due to the mutation of essential catalytic residues. Finally, we have expanded the framework of understanding for aromatic substrate binding and activation in the heme peroxidases. We have shown that single point mutations can dramatically alter substrate binding and oxidation, indicating that the sites may be more promiscuous than previously thought. These observations provide a fundamental platform upon which our understanding of the enzyme-catalyzed activation of this prodrug can now be developed for more effective tuberculosis therapies in the future.