CRYSTAL STRUCTURE OF MYCOBACTERIUM TUBERCULOSIS CATALASE-PEROXIDASE*

The Mycobacterium tuberculosis catalase-peroxidase is a multifunctional heme-dependent enzyme that activates the core anti-tuberculosis drug isoniazid. Numerous studies have been undertaken to elucidate the enzyme-dependent mechanism of isoniazid activation, and it is well documented that mutations that reduce activity or inactivate the catalase-peroxidase lead to increased levels of isoniazid resistance in M. tuberculosis. Interpretation of the catalytic activities and the effects of mutations upon the action of the enzyme to date have been limited due to the lack of a three-dimensional structure for this enzyme. In order to provide a more accurate model of the three-dimensional structure of the M. tuberculosis catalase-peroxidase, we have crystallized the enzyme and now report its crystal structure refined to 2.4-A resolution. The structure reveals new information about dimer assembly and provides information about the location of residues that may play a role in catalysis including candidates for protein-based radical formation. Modeling and computational studies suggest that the binding site for isoniazid is located near the delta-meso heme edge rather than in a surface loop structure as currently proposed. The availability of a crystal structure for the M. tuberculosis catalase-peroxidase also permits structural and functional effects of mutations implicated in causing elevated levels of isoniazid resistance in clinical isolates to be interpreted with improved confidence.


INTRODUCTION
Catalase-peroxidases (CPs) 1 are bi-functional, heme-dependent enzymes which exhibit a strong catalatic activity, comparable to that of monofunctional catalases, and a broad spectrum peroxidatic activity. The proposed role of the enzyme is to protect bacteria from toxic molecules including hydroperoxides and hydroxyl radicals that are present in an aerobic environment. CPs are generally homodimers or homotetramers with subunits of about 80 kDa. The single polypetide chain containing two domains, each of approximately 40 kDa, is proposed to have arisen from a gene duplication event (1). Sequence analysis shows that the N-terminal domain contains a heme-binding motif whilst the C-terminal domain lacks this feature. In spite of the strong catalatic activity displayed by CPs, the sequence shows no homology with catalases. However, both CP domains show highest sequence similarity with yeast cytochrome c peroxidase (CcP) and ascorbate peroxidase (APX) and CPs are therefore classified as belonging to class I of the superfamily of plant, fungal and bacterial peroxidases (2).
The Mycobacterium tuberculosis CP has been the subject of numerous studies because of its ability to activate isoniazid (INH), a core compound used to treat tuberculosis. In particular, it has long been observed that INH resistance in tuberculosis-causing mycobacteria has often been correlated with reduced levels of catalase activity (3)(4)(5).
Subsequently, it was confirmed that the presence of an active CP, encoded by a single gene, katG, is sufficient to confer INH sensitivity in M. tuberculosis, the organism principally responsible for tuberculosis (6). Studies using M. tuberculosis CP obtained either from the organism or in a recombinant form demonstrated that the enzyme is capable of oxidizing INH (7,8); however, the mechanism of oxidation and the precise mode of action of the drug are still subjects for debate. Of the evidence to date, there is some indication that activated INH may inhibit the action of InhA (9) and KasA (10), enzymes which are proposed to be involved in the synthesis of mycolic acids. In vitro assays have demonstrated that by guest on August 29, 2020 http://www.jbc.org/ Downloaded from Structure of M. tuberculosis CP 4 inactivation of the InhA enzyme by INH only occurs in the presence of NADH or NAD + (11), and the crystal structure of InhA containing an INH-NADH adduct present in its active site has been determined (12). Recently published data have also shown that an INH-NADH adduct is an effective inhibitor of InhA (13).
It is established that in vitro activation of INH by M. tuberculosis CP yields acid, amide and aldehyde products (8,14,15). Generation of the amide product has been shown to proceed via the cleavage of a C-N bond (14) which is thought to yield an acyl radical (8,14) that is capable of reacting with NADH or NAD + , for example. Residues  tuberculosis CP (mtCP) to 2.4-Å resolution. We discuss how mainchain and side chain features relate to the assembly and function of the enzyme. Comparative analyses of sequences and structures of related class I and class III peroxidases predict that the binding site for INH in mtCP will be near the δ-meso edge of the heme, rather than in a surface loop as currently proposed by others (16). Computational analyses of interaction sites for INH, in combination with the availability of an NMR-derived model of a complex between horseradish peroxidase C (HRPC) and INH (17), have enabled us to model the location of the drug within the active site of mtCP. The significance of the identification of this binding site is discussed in the context of the interpretation of radical sites and mutations related to INH resistance in clinical isolates.

EXPERIMENTAL PROCEDURES
Protein Overproduction and Purification mtCP was purified from Escherichia coli strain UM255 (18) transformed with the pTrc99A derivative, pTBCP, using a modified version of a previously reported protocol (19). To maximise heme incorporation, excess hemin was added to the crude cell extract prior to sonication. Briefly, hemin chloride (Sigma Aldrich, Poole, UK) was dissolved in 1 ml Milli-Q water to a final concentration of 24.5 mM. Addition of 1 µl 10 M NaOH was sufficient to dissolve the hemin chloride yielding a final solution at about pH 6.0. Reconstituted holoenzyme was achieved by adding the hemin solution in a 1 to 25 (v/v) ratio to the cell pellet of 1 l of culture resuspended in 10-15 ml 100 mM KH 2 PO 4 /K 2 HPO 4 (pH 6.5) at 4 °C.
The suspension was mixed well by vortexing and incubated at 4 o C for 30 min before sonication. Cells were lysed by sonication with 3 x 30 s bursts on power level 8, using a microtip and an XL2020 sonicator (Labcaire Systems, Clevedon, UK). Insoluble material was removed by centrifugation at 10000 g at 4 ˚C for 50 min. The supernatant was treated with 100 µg/ml DNAse and 10 µg/ml RNAse for 1 h at 4 ˚C and then recentrifuged as just described.
The supernatant was applied to a DEAE-Sepharose anion exchange column (19 (19). Pooled, active fractions were then loaded, at 2.0 ml/min, onto an Amersham Biosciences Resource Q (6 ml) anion exchange column preequilibrated with 30 ml 100 mM KH 2 PO 4 /K 2 HPO 4 (pH 6.5). The column was washed with 12 ml 100 mM KH 2 PO 4 /K 2 HPO 4 (pH 6.5) and mtCP was eluted with a 60-ml linear gradient of 0-1.0 M NaCl in 100 mM KH 2 PO 4 /K 2 HPO 4 (pH 6.5) at a flow rate of 2.0 ml/min. 1.0-ml by guest on August 29, 2020 http://www.jbc.org/ Downloaded from Structure of M. tuberculosis CP 6 fractions were collected, assayed and pooled. An Amersham Biosciences HiTrap Desalt column was used to remove NaCl and change buffer conditions for crystallization studies.
The column was equilibrated with 25 ml 0.1 M sodium acetate (pH 7.0) using a syringe.
mtCP was loaded onto the column in 1.5-ml aliquots and eluted with 2 ml 0.1 M sodium acetate (pH 7.0).

Structure Solution, Refinement, and Superpositions
The diffraction data were processed using the program DENZO and scaled with program SCALEPACK (20). A portion of the data set (10%) was flagged for the calculation of R free and excluded from subsequent refinement. The structure of mtCP was solved by molecular replacement using the crystal structure of the Burkholderia pseudomallei CP (bpCP) (

GRID calculations
GRID, a computational program by P. J. Goodford (30), was designed to detect energetically favorable binding sites on proteins, structurally determined by X-ray crystallography or NMR, with the view that this information could be used in the rational design of drugs. In GRID, the target molecule is the species being evaluated and the probe molecule is a small biologically active species such as water, amine nitrogen, carboxy oxygen, or hydroxyl. GRID calculates the interaction energies between the probe molecule and the target molecule accounting for both the shape and energy of the two species. Analysis of GRID data yields threedimensional maps which can be contoured to identify the location of energetically favorable interaction sites.

Fold and assembly of the M. tuberculosis CP
The mtCP crystal structure was determined to a resolution of 2.4 Å, with R cryst and R free values of 21.1% and 26.8 %, respectively. Full data collection and refinement statistics are summarized in Table 1. The mtCP crystal structure shows that the enzyme assembles as a functional homodimer. The electron density map of each monomer includes residues 24-740 of the polypeptide chain, one heme b moiety and three glycerol molecules. In addition 703 water molecules associated with the homodimer were also identified and refined. Each The crystal structure of mtCP shows that the enzyme adopts a similar fold to those reported for the crystal structures of hmCP (31) and bpCP (16). mtCP shares 55% identity and 69% similarity with hmCP, and 66% identity and 77% similarity with bpCP. N-terminal residues of mtCP as was previously proposed using the yeast two-hybrid assay (35) and as was observed in the bpCP crystal structure (16). The N-terminal residues appear to "hook" around each other in a manner which may further stabilize the formation of the dimer, shown schematically in Fig. 2B, a detail not previously observed in the hmCP (31) or bpCP (16) crystal structures. This stabilization appears to be mediated by the presence of stacking interactions between Tyr28 and Tyr197, and Trp38 and Trp204 from opposite monomers (Fig. 2C). This long N-terminal extension appears unique to the CPs and is not present in other members of the class I peroxidase family e.g. CcP (26) and APX (27,32) ( Fig. 3A). Hence, although inter-domain contacts between monomers of mtCP appear to be important in mediating homodimer formation, it is also interesting to note this "hook" motif may also play a significant role in dimer assembly.

Active site of M. tuberculosis CP
The architecture of the heme-containing active site of mtCP (Fig. 4) is similar to those of hmCP and bpCP, and closely resembles other class I peroxidases.
The heme protoporphyrin IX moiety embedded within the active site of mtCP is fully occupied and unmodified. The level of heme incorporation in CPs has been shown to vary from substoichiometric to stoichiometric values (e.g. (19,33,36)). In the case of mtCP, occupancy of heme was raised from 0.5 heme/dimer (19)  activity. In CPs, the most dramatic effects upon activity are observed for mutations involving the covalently modified distal site tryptophan residue, Trp107 in mtCP (Fig. 4A), which severely impair the catalatic activity with, in general, limited effects upon the peroxidatic activity of the enzyme (45,46). Mutation of mtCP Tyr229, or its equivalent in the Synechocystis CP, to phenylalanine also results in the loss of catalatic activity (47,48).
Taken together these data further support the role of this adduct in maintaining the catalatic activity of the enzyme.
Four well-ordered water molecules can also be identified above the heme, within the distal pocket of the mtCP crystal structure (Fig. 4). Wat7, Wat235 and Wat427 occupy very similar positions to those observed for water molecules in the bpCP (16) and hmCP (31) crystal structures. mtCP also contains an additional water molecule, Wat352, located near the adduct and hydrogen-bonded to Wat427. In the hmCP crystal structure (31) a similar water molecule, Wat376, is also found near the adduct. In the bpCP crystal structure (16), the position of this water is occupied by a postulated perhydroxy modification on the vinyl group of the porphyrin ring of the heme. Notably, a water molecule is observed ligated to the heme iron at the sixth coordination position in the distal pocket of bpCP (16) and in one of the monomers of hmCP (31). In the mtCP crystal structure, however, no equivalent water molecule is observed and the heme appears only in a 5-coordinate state (Fig. 4), which is consistent with a number of spectroscopic studies of the enzyme (36,49,50).

Spectroscopic correlation: sites of radical formation in M. tuberculosis CP
In heme-containing peroxidases, a common catalytic cycle is observed which involves reaction of the ferric iron with hydrogen peroxide resulting in the formation of an oxyferryl species known as Compound I (51,52). In most peroxidases, Compound I is characterized by an Fe(IV)=O heme and a porphyrin π-cation radical (reviewed by (53)). The most notable exception is CcP where the radical is not localised on the porphyrin ring but is instead on the proximal Trp191 residue (54)(55)(56), equivalent to Trp321 in mtCP (Fig. 4). The Compound I spectrum for CP based on studies of alkyl peroxide-mediated oxidation of CPs is similar to that for other peroxidases and therefore was initially attributed to a protoporphyrin radical species (57,58). Although initial EPR studies drew similar conclusions (36), recent studies have suggested that protein-based radicals may indeed be generated involving tyrosine and/or tryptophan residues (47,(59)(60)(61). mtCP contains 21 tyrosine residues and 24 tryptophan residues which are scattered throughout the structure of the enzyme as shown in Fig. 5. The most recent evidence for the presence of a protein-based radical in mtCP has been the identification of a tyrosyl radical (47,59) proposed to occur at Tyr353 based upon EPR, spin trapping and mutational studies (60). As shown in Fig. 5, Tyr353 is located approximately 15 Å from the heme Fe, on the surface of the structure and now seems an unlikely candidate when considered in the context of the crystal structure. The side chain of Tyr353 points away from the surface of the protein, making it perhaps a likely target for radical trapping; however, given the number of tyrosines in the structure, the presence of a tyrosyl radical is still possible but could perhaps involve other tyrosine residues which are less solvent exposed and/or are located nearer to the heme (Fig. 5). EPR studies of a Synechocystis CP enzyme have also demonstrated the presence of a tyrosyl radical and a tryptophanyl radical in addition to a porphyrin-based radical (61). Although no residue has yet been assigned to the tyrosyl radical, Trp106 in this enzyme has been proposed as the possible site for the tryptophanyl radical. In mtCP the equivalent residue is Trp91, shown in The location of a binding site for aromatic compounds in CPs has been postulated to be in the conserved surface loop, highlighted in Fig. 1, based upon the presence of a peak of electron density observed within this loop region in the bpCP crystal structure (16). The authors speculate that this density might be occupied by a pyridine-like metabolite, although no biochemical data are yet published to support this hypothesis. In mtCP, a small peak of electron density is also observed in the same site (Fig. 6A)   model is representative of the complex formed during oxidation of the drug (see also (17)). production of mycolic acids which are essential to the survival of the organism (reviewed in (72)). Using the mtCP crystal structure the effects of a number of point mutations located in the proposed INH binding site in the distal pocket can now be rationalized to some extent (Fig. 8).

Role of mutations in M. tuberculosis CP associated with INH resistance
For example, mutations affecting Ser315 are amongst the most commonly occurring (69,(73)(74)(75) resulting in up to a 200-fold increase in the Minimum Inhibitory Concentration (MIC) for the drug (76). Ser315 has been reported to be mutated to asparagine, isoleucine, arginine and glycine although the most frequently occurring mutation is to threonine. As a result, a number of in vitro studies have been undertaken to understand the origins of resistance using the S315T mutant as a model (50,(77)(78)(79)(80)(81)(82)(83)(84)(85)(86). In the absence of a structure for mtCP it was recently postulated that Ser315 forms hydrogen bonds to one of the heme proprionate groups and that mutation to threonine would therefore modify the heme pocket altering INH binding (16,50). Based upon the proposed binding site for INH in the bpCP crystal structure (INH2 in Fig. 6B), the S315T mutation was predicted to alter the binding site for the hydrazinyl moiety of INH and/or affect the transfer of electrons to the heme. As shown in which were attributed to conformational changes in the heme pocket (50). Ser315 is not predicted to play a significant role in the proposed mechanism for the enzyme-catalyzed activation of the drug (see (17)). This is consistent with the observation that the S315T mutant displays a reduced affinity for INH (78) but is able to oxidize INH with a rate equivalent to the wild-type enzyme (80). Furthermore, with the exception of glycine, the other mutations at this position would also increase steric bulk and further limit access to the Based upon the availability of the mtCP crystal structure, certain residues in its active site have been postulated to be involved in enzyme-catalyzed activation of INH (17) but only the distal His108 residue has been a site for mutations conferring increased resistance to the drug (Fig. 8). This residue is proposed to be involved in the binding and enzyme-catalyzed activation of INH via interactions with the hydrazinyl moiety of the drug (17). In monofunctional peroxidases, this distal histidine has been shown to be particularly important in the formation of the oxyferryl Compound I (reviewed by (53)). In CcP, mutation of this residue, His52, to leucine results in a 10 5 reduction of activity (87). In CPs, mutation of this residue results in a substantial reduction in catalatic and peroxidatic activities (45,46). With regard to INH resistance, His108 in mtCP has been reported to be mutated to glutamic acid and glutamine (e.g. (70,71)). These mutations may reduce the affinity of the enzyme for INH but glutamic acid/glutamine do have hydrogen bond donor and acceptor groups which could allow INH to bind in the distal pocket based upon the model presented here. However, neither of these residues would be predicted to shuttle protons in the same manner as His108 in the enzyme-catalyzed activation pathway (17). In addition, the A110V mutant associated with increased levels of INH resistance (88) may in fact, be exerting its effects by altering the local conformation of His108 which could in turn alter its ability to bind and/or activate INH.
Based upon the mtCP structure, Asp137 has also been identified to play a key role in the binding and activation of INH as this residue appears to be a CP-specific proton donor in the enzyme-catalyzed activation pathway (discussed in (17)). No mutants associated with INH resistance have been reported to involve Asp137 although a number have been observed nearby including N138S, A139P, S140N or D142A (6,69,70,73,88) (Fig. 8). These mutants might be exerting their effects through local conformational changes which could alter the orientation of the Asp137 side chain, making it less effective in binding INH and/or catalyzing the turnover of this drug. It is interesting to note that in the catalatic reaction carried out by the enzyme, Asp137 is also postulated to act as a proton donor (89). Mutations at this position would therefore have the potential to drastically alter the enzyme's ability to function as a catalase. It is therefore possible that slight repositioning of the side chain of Asp137 through mutations nearby may be sufficient to significantly reduce INH binding and/or activation whilst retaining sufficient levels of catalase activity in vivo.
In         Ramachandran statistics e (%) most favored regions 88.2 additional allowed regions 11.3 generously allowed regions 0.5 a Numbers in parenthesis represent values in the highest resolution shell (last of 10 shells) b R sym = Σ h Σ i |I(h,i) -<I(h)>| / Σ h Σ i I(h,i) where I(h,i) is the intensity value of the ith measurement of h and <I(h)> is the corresponding mean value of I(h) for all i measurements. c R cryst = Σ ||F obs | -|F calc || / Σ |F obs |, where |F obs | and |F calc | are the observed and calculated structure factor amplitudes respectively. d R free is the same as R cryst but calculated with 10% subset of all reflections that was never used in crystallographic refinement. e As evaluated by PROCHECK (90). by guest on August 29, 2020