Crystal Structure and Functional Analysis of Lipoamide Dehydrogenase from Mycobacterium tuberculosis*

We report the 2.4 Å crystal structure for lipoamide dehydrogenase encoded by lpdC from Mycobacterium tuberculosis. Based on the Lpd structure and sequence alignment between bacterial and eukaryotic Lpd sequences, we generated single point mutations in Lpd and assayed the resulting proteins for their ability to catalyze lipoamide reduction/oxidation alone and in complex with other proteins that participate in pyruvate dehydrogenase and peroxidase activities. The results suggest that amino acid residues conserved in mycobacterial species but not conserved in eukaryotic Lpd family members modulate either or both activities and include Arg-93, His-98, Lys-103, and His-386. In addition, Arg-93 and His-386 are involved in forming both “open” and “closed” active site conformations, suggesting that these residues play a role in dynamically regulating Lpd function. Taken together, these data suggest protein surfaces that should be considered while developing strategies for inhibiting this enzyme.

Over the last decade, tuberculosis has reemerged as one of the leading causes of death (1). Despite the substantive response of the host immune system that includes both oxidative and nitrosative stress, Mycobacterium tuberculosis (Mtb) 2 evades these insults, in part by adopting metabolic enzymes to catabolize these toxic compounds (2). Lipoamide dehydrogenase (Lpd), dihydrolipoamide acyltransferase (DlaT; formerly termed succinyl transferase), an alkylhydroperoxidase termed AhpC, and the protein (AhpD) encoded by an adjacent gene, have been shown to take part in this defense pathway (3,4). All of these enzymes are oxidoreductases, and each contains redox centers that reduce or oxidize adjacent partners in the pathway.
The Lpd redox center consists of two conserved cysteine residues and a non-covalently bound FAD molecule (5). DlaT, encoded by the gene Rv2215, was originally annotated as the E2 component of ␣-ketoglutarate dehydrogenase, but subsequent investigation revealed that Mtb lacks ␣-ketoglutarate dehydrogenase. Rv2215 instead encodes DlaT, the E2 component of pyruvate dehydrogenase (PDH) (6). DlaT contains a redox center in the form of dihydrolipoyl cofactors that are covalently attached to lysine ⑀-amino groups within lipoyl domains that form the so-called "swinging arms" of this enzyme (7). Both AhpD and AhpC contain redox centers consisting of two conserved cysteine residues (3). Each of these redox centers can oxidize or reduce the redox center of the adjacent enzyme in the cascade. For example, AhpC metabolizes peroxides and peroxynitrite (ROI/RNI) (3,8,9), AhpD regenerates the redox center of AhpC, DlaT regenerates the redox center of AhpD, and Lpd regenerates the redox center of DlaT. Lpd serves a critical function in this cascade because its redox center is regenerated directly by NADH. The process can be represented as a flow of electrons from NADH 3 Lpd 3 DlaT 3 AhpD 3 AhpC 3 ROOH. The absence or perturbation of any member in this cascade has been shown to negatively impact the electron flow (3), 3 thus making this pathway an attractive target for therapeutic intervention against Mtb. Such intervention might render the organism more susceptible to the oxidative and nitrosative stress imposed by the host immune system, as demonstrated for Mtb in which dlaT has been disrupted. 3 The inhibition of Lpd might also have adverse effects on Mtb unrelated to the defense of Mtb against the host immune response. Lpd and DlaT are both components of the multienzyme PDH complex (6). Lpd constitutes the E3 component of PDH, catalyzing the NAD ϩ -dependent oxidation of the dihydrolipoyl cofactor that is covalently linked to the lipoyl domains of DlaT, the E2 component. Lpd, DlaT, and AceE, the E1 component of PDH (6), synthesize acetyl-CoA, a component of the Krebs cycle critical for the generation of precursors for energetic and/or biosynthetic functions. These pathways are essential and conserved across evolution.
Lpd belongs to the group I flavoprotein disulfide reductase protein family (11). Members of the flavoprotein disulfide reductase family share low sequence identity (25-35%), and although the structure of Lpd from Mtb was not known, significant structural information is known for other group I flavoprotein disulfide reductase proteins whose three-dimensional structures are comparable with root mean square deviations of less than 2 Å over nearly 400 C␣ atoms. The structure of human Lpd has also been recently reported (12). The intact structure of DlaT is yet to be determined although structures of a homologous catalytic DlaT core (13)(14)(15)(16) and a DlaT lipoyl domain have been reported (17,18). The structures of AhpD (3,4,19) and AhpC (20) have recently been determined.
The pET-based plasmid was transformed into E. coli BL21(DE3) CodonPlus RIL (Stratagene). A 5-liter culture was grown by fermentation at 37°C to an A 600 of 2, adjusted to 30°C and 1 mM isopropyl-␤-Dthiogalactopyranoside and incubated for 4 h. Cells were harvested by centrifugation and resuspended in 20 mM Tris-HCl (pH 8.0), 350 mM NaCl, 10 mM imidazole, 20% sucrose, 1 mM ␤-mercaptoethanol, and 20 g/ml lysozyme and sonicated. After the insoluble material was removed by centrifugation, His 6 -Smt3-Lpd was purified by metal-affinity and gel filtration chromatography (Superdex200). The His-Smt3 tag was removed by the Smt3-specific protease Ulp1 and Lpd was further purified by gel filtration (Superdex200). Lpd eluted from gel filtration at a position consistent with dimer formation. Lpd was concentrated to 8 mg/ml, frozen in liquid nitrogen, and stored at Ϫ80°C.
Crystallographic Analysis-Crystals of Lpd were grown using the hanging drop vapor diffusion method at 6°C. The drop contained 2 l of protein (4 mg/ml) and 2 l of reservoir solution (50% 2-methyl-2,4pentanediol (MPD)). Crystals grew over 8 -10 days in a final crystallization condition that included 150 mM NaCl, 20 mM Tris (pH 8.0), 1 mM ␤-mercaptoethanol, and 50% MPD. Crystals were flash-frozen in the solution in which they grew using liquid nitrogen. Lpd crystals were obtained from protein containing selenomethionine (22) and used for the structure determination by single wavelength anomalous diffraction analysis. Lpd crystals belong to orthorhombic space group P2 1 2 1 2 1 with a ϭ 83.59, b ϭ 96.67, c ϭ 122.89 Å. X-ray diffraction data were collected at NSLS beamline X9A at wavelength 0.979 Å. Data were reduced with DENZO, SCALEPACK (23), and CCP4 (24). The structure of Lpd was determined using single wavelength anomalous diffraction technique employing the anomalous signal from selenium atoms as implemented in the software SOLVE (25). Twelve selenium atoms were located, and phases derived from these anomalous centers were used for density modification as implemented in the software RESOLVE (26). Non-crystallographic symmetry was used for further phase improvement. Manual model building was accomplished using O (27). Refinement of the model was accomplished using CNS (28). Water molecules were located and refined. The final model contains 6938 protein atoms, 106 cofactor atoms (FAD), one MPD molecule (8 atoms), and 333 solvent atoms and was refined to an R value of 19.9% and an R free value of 24.6%. Crystallographic details are in TABLE ONE.
Construction of Lpd Mutations-Single amino acid substitutions were introduced into the pSMT3-Lpd vector (see above) using PCR- where F o and F c are observed and calculated structure factors, respectively. e Calculated with PROCHECK.
based mutagenesis techniques using the QuikChange kit and protocols (Stratagene) with minor modifications. Complementary sets of oligonucleotide primers of length 23-39 containing respective codon substitutions were obtained from Invitrogen. Plasmid DNA containing the Lpd mutations were extracted from a culture grown from a single colony of transformed XL1-Blue cells. All mutations were confirmed by sequencing. Expression strains containing each mutation were used to obtain mutant proteins as described above using a 1-liter culture of E. coli. Cells were harvested and lysed using BugBuster HT (Novogen). The insoluble material was removed by centrifugation. The supernatant containing His 6 -Smt3-Lpd was mixed with nickel-nitrilotriacetic acid resin (Qiagen), incubated for 30 min, and washed five times with 1 volume of wash buffer (500 mM NaCl, 20 mM Tris (pH 8) 1 mM ␤-mercaptoethanol, 10 mM imidazole). The resin was resuspended in 5 ml containing 20 mM triethanolamine (pH 7.8). Lpd was cleaved from the His 6 -Smt3 tag using Smt3-specific protease Ulp1 (21), thus liberating soluble Lpd into the supernatant. Protein concentration was assessed, and the material was frozen in liquid nitrogen and stored at Ϫ80°C for later use. Biochemical Assays-Enzymatic activities of wild-type and mutant Lpd proteins were assessed by three different assays: 1) a lipoamide dehydrogenase (LDH) assay, in which Lpd proteins were tested alone for their ability to catalyze lipoamide reduction; 2) a PDH assay, in which Lpd proteins were tested in complex with AceE (E1 component of PDH) and DlaT (E2 component of PDH) recombinant proteins for pyruvate-and CoA-dependent NADH production; 3) and a dithionitrobenzene (DTNB) assay, in which Lpd proteins were tested in complex with DlaT and AhpD recombinant proteins for NADH-dependent DTNB reduction. All assays were done in triplicate in 96-well plates using a SpectraMAX 340PC plate reader (Molecular Devices) at room temperature. LDH assay reaction mixtures (0.

RESULTS
The overall structure of the Mtb Lpd protomer is similar to other known group I flavoprotein disulfide reductase proteins (Fig. 1) (11). These include glutathione reductase (PDB code 1GRB) (29), trypanothione reductase (PDB code 1AOG) (30), and human Lpd (hE3, PDB code 1ZMC, 1ZMD) (12), as well as the inactive homologue of Lpd from Mtb (LpdA, PDB code 1XDI) (31) with sequence identities ranging between 25 and 35%. Despite low sequence identity, the three-dimensional structures align well with root mean square differences of less than 1.75 Å over ϳ400 C␣ atoms. As expected from previous structures, Mtb Lpd is a dimer and similar in overall organization to lipoamide dehydrogenase structures from other organisms. The Lpd fold consists of an N-terminal FAD-binding domain (residues 1-47 and residues 105-144), a NADH-binding domain (residues 145-268), a central domain (residues 269 -343), and two segments of the structure that mediate dimer formation (residues 48 -104 and 344 -464) (Fig. 1). The Lpd dimer is formed via extensive interactions between protomers that buries a total surface area of ϳ3600 Å 2 for each protomer. Each protomer includes a non-covalently bound FAD molecule. FAD was copurified from E. coli lysate, as it was not added during purification of the recombinant protein. NADH was not observed bound within our struc- and eukaryotic (human; accession P09622, pig; accession P09623, dog; accession P49819, and mouse; accession O08749). Secondary structure from the crystal structure is shown above the aligned sequences. Active site residues are indicated by * below the alignment and residues for which mutational analysis was conducted are indicated by a solid triangle below the alignment. Amino acid residue number pertains to the sequence from Mtb. Gaps indicated by (.). B, matrix of sequence identity obtained by pairwise sequence alignments between the sequences presented in A. Accession number and name are indicated on both x and y axes. Multiple sequence alignment was carried out using ClustalW (35). The figure was prepared using ESPript (10).
ture, although a clear NADH binding pocket was apparent on the surface of the protein where NADH has been observed to bind previously (12,29,32). Mtb Lpd crystallizes as the biological dimer in the crystallographic asymmetric unit, and although the protomer structures appear similar, there were small but important structural differences observed when comparing the two protomers (see below).
The two Lpd active sites are formed by interfacial residues between protomers (Fig. 1). The primary constituents of the active site include a disulfide-based redox center (Cys-41 and Cys-46). These residues are located proximal to the non-covalently bound FAD and a His-Glu pair (His-443, Glu-448) that has been postulated as a general acid for catalysis (Fig. 1B) (33). Cys-41 and Cys-46 were observed as an oxidized disulfide. The redox center is accessed through a channel ϳ15 Å deep and is encompassed by residues that include Pro-13, Tyr-16, Leu-42, Ile-47, Arg-93, Gly-96, Phe-99, Leu-100, Glu-321, Ala-383Ј, His-443Ј, Asn-463Ј, and Phe-464Ј (apostrophes indicate residues from the second protomer). This channel is likely utilized in interactions with the lipoylated side chain from DlaT.
Most active site residues were observed in nearly identical conformations in both promoter structures; however, one considerable difference was noted. In one active site, Arg-93 was observed interacting with an MPD molecule in a conformation that would block access to the deep channel. In the other active site, Arg-93 interacts with His-386 in an open conformation that would permit full access into the active site and redox center. Both Arg-93 and His-386 are conserved among mycobacterial Lpd family members but not conserved among other eukaryotic family members (see below) ( Fig. 2A).
To establish the functional basis for these and other differences observed in the structure, we selected several candidate residues for mutational analysis based on the Lpd structure and sequence alignments of representative family members from both bacterial and eukaryotic organisms. The goals for the mutagenesis and functional analysis were to identify residues essential for catalysis that were conserved among mycobacteria but not conserved in eukaryotic Lpd family members. We reasoned that if surfaces could be identified that were essential for Mtb Lpd function, they might be exploited for therapeutic intervention.  Based on the amino acid sequence alignment of Lpd sequences from M. tuberculosis (Mtb, accession P66004), Mycobacterium leprae (Mle) (Q50068), pig (P09623), human (P09622), dog (P49819), and mouse (O08749), we selected several single point mutations based on their location around the Lpd active site and on surfaces postulated to be involved in interaction with the lipoylated DlaT domain (Fig. 2). These mutations were placed into two classes. Class I included residues conserved in prokaryotic sequences (Mtb and Mle) but not conserved between prokaryotic and eukaryotic sequences (human, pig, dog, and mouse). Class II residues were conserved in all five sequences listed above. Using these criteria, 20 candidate residues were selected for mutation. In addition to these residues, we also selected several lysine residues for mutagenesis (Lys-216, Lys-220, Lys-223, and Lys-224), because they formed a cluster on surface that we speculated might interact with the lipoylated domain of DlaT (Fig. 3A). Although 29 mutants were constructed, only 23 produced suitable protein for biochemical analysis.
Wild-type and mutant Lpd proteins were evaluated in three different assays; (a) the LDH assay evaluated whether mutants were effective in binding lipoamide and reducing it to dihydrolipoamide, (b) the PDH assay evaluated the ability of mutant Lpd proteins to oxidize dihydrolipoamide on the swinging arms of DlaT in complex with E1 (AceE) and E2 (DlaT) components of PDH complex, and (c) the DTNB assay evaluated the ability of mutant Lpd proteins to reduce lipoamide on the swinging arms of DlaT and then supply electrons to AhpD. Results of these assays are presented in TABLE TWO and graphically in Fig. 3. Of the 23 mutants assayed, 9 were observed to have significantly diminished activity when compared with wild-type, whereas 14 had negligible defects in activity.

DISCUSSION
Three of the six mutations selected from class II residues exhibited little to no activity in any of the assays utilized in this study, not surprising because they are conserved throughout the Lpd protein family. These residues included Asp-5, Asn-43, and Phe-464 (TABLE TWO). Asp-5 is located far from the active site and is not thought to directly participate in catalysis. As such, it remains unclear why alanine substitution at this position would preclude activity in the LDH assay, although it might play an indirect role in coordinating either FAD or NADH. Asn-43 is located between the two active site cysteine residues and is likely important for positioning this critical region of the protein for productive catalysis. It may also play a role in positioning the helix that coordinates FAD. Phe-464 is the most C-terminal amino acid. It is positioned at the periphery of the deep channel that is thought to participate in binding or coordinating the lipoylated side chain from DlaT.
Substitution of this position to alanine completely abrogates activity in all three assays.
Of the mutations made within class I residues (conserved in Mtb, but not conserved with eukaryotic Lpd family members), only four gave rise to defects greater than 50% in any of the three assays and included mutation at Arg-93, His-98, Lys-103, and His-386 (TABLE TWO; Figs.  2 and 3). Lys-103 is proximal to Phe-464 (see above), and glutamate substitution at this position would be predicted to disrupt its interaction with the Phe-464 C terminus. Lys-103 is conserved as a glutamine in other eukaryotic Lpd family members (Fig. 2). His-98 is conserved as alanine in other eukaryotic Lpd family members. His-98 substitution to alanine in Mtb Lpd diminishes its activity to only 33% in the LDH assay, but preserved over 60% activity in both PDH and DTNB assays. His-98 is located within the helix and nearby a conserved glycine that forms one side of the deep channel leading to the active site. It is also observed within hydrogen bonding distance to Glu-108. Position 108 is conserved as either Arg or His in other eukaryotic family members, suggesting that the interaction between His-98 and Glu-108 is specific to mycobacterial Lpd family members.
Perhaps the most interesting class I mutations were at positions 93 and 386, as these two residues interact in the Mtb Lpd structure (Fig. 4). Arg-93 is conserved as leucine in eukaryotic family members. Alanine substitution at position 93 abolishes LDH activity but results in 40 or 80% activity in PDH or DTNB assays, respectively. Not surprisingly, charge reversal at position 93 completely abrogates activity in all three of the assays. Although position 386 is conserved as lysine in eukaryotic Lpd family members, substitution of His386 with either lysine or alanine resulted in diminished activities in the three assays that were tested.
As stated in previous sections, Arg-93 was observed in two distinct conformations in each of the two Lpd active sites (Fig. 4). In one active site, Arg-93 is stacked above His-386Ј across the dimer interface (Fig.  4B). In this "open" conformation, the deep channel is fully accessible to solvent and to the presumed interaction with dihydrolipoamide. In the "closed" configuration, Arg-93 is positioned in the channel such that the access is partially blocked by the arginine side chain (Fig. 4B). A single molecule of MPD was observed in the channel interacting with Arg-93, a molecule present in our crystallization condition (see "Experimental Procedures"). Although MPD binding in the active site is likely a crystallization artifact, it is interesting for several reasons. The first is that alanine substitution at Arg-93 abrogates LDH activity but only exhibits diminished activity in PDH or DTNB assays, suggesting that defects resulting from diminished dihydrolipoamide interaction can be overcome when the lipoyl group is presented by DlaT. This suggests that Arg-93 may interact directly with dihydrolipoamide in these assays. Although MPD and lipoamide are structurally distinct, they do share some common functionality that could be exploited when trying to understand how lipoamide may interact with the Lpd active site. At the very least, it can be speculated that if small molecules could be engineered to interact with the pocket and Arg-93 with high affinity, they might lock the channel in a closed conformation as observed in our structure. This approach might lead to specific inhibition of the Mtb enzyme insomuch as position 93 is conserved as a leucine residue in human and other eukaryotic sequences and analysis of the human Lpd structure suggests that leucine could not occupy the position observed for Mtb Lpd Arg-93 in the "closed" conformation.