Characterization of a new member of the flavoprotein disulfide reductase family of enzymes from Mycobacterium tuberculosis.

The lpdA (Rv3303c) gene from Mycobacterium tuberculosis encoding a new member of the flavoprotein disulfide reductases was expressed in Escherichia coli, and the recombinant LpdA protein was purified to homogeneity. LpdA is a homotetramer and co-purifies with one molecule of tightly but noncovalently bound FAD and NADP+ per monomer. Although annotated as a probable lipoamide dehydrogenase in M. tuberculosis, LpdA cannot catalyze reduction of lipoyl substrates, because it lacks one of two cysteine residues involved in dithiol-disulfide interchange with lipoyl substrates and a His-Glu pair involved in general acid catalysis. The crystal structure of LpdA was solved by multiple isomorphous replacement with anomalous scattering, which confirmed the absence of these catalytic residues from the active site. Although LpdA cannot catalyze reduction of disulfide-bonded substrates, it catalyzes the NAD(P)H-dependent reduction of alternative electron acceptors such as 2,6-dimethyl-1,4-benzoquinone and 5-hydroxy-1,4-naphthaquinone. Significant primary deuterium kinetic isotope effects were observed with [4S-2H]NADH establishing that the enzyme promotes transfer of the C4-proS hydride of NADH. The absence of an isotope effect with [4S-2H]NADPH, the low Km value of 0.5 microm for NADPH, and the potent inhibition of the NADH-dependent reduction of 2,6-dimethyl-1,4-benzoquinone by NADP+ (Ki approximately 6 nm) and 2'-phospho-ADP-ribose (Ki approximately 800 nm), demonstrate the high affinity of LpdA for 2'-phosphorylated nucleotides and that the physiological substrate/product pair is NADPH/NADP+ rather than NADH/NAD+. Modeling of NADP+ in the active site revealed that LpdA achieves the high specificity for NADP+ through interactions involving the 2'-phosphate of NADP+ and amino acid residues that are different from those in glutathione reductase.

Most of the FDR enzymes are homodimers containing a tightly but noncovalently bound FAD per monomer (1,2). Reduction of the FAD cofactor by NAD(P)H on the re face of the FAD to generate a transient FADH 2 ⅐NAD(P) ϩ intermediate (3)(4)(5) is a common first step in these enzymes. Subsequent electron transfer from FADH 2 to a nonflavin redox center, located on the si face of the FAD, is also a common feature of these enzymes. An exception is STH, which appears to lack a nonflavin redox center. The nature of the nonflavin redox center varies, and three types have been identified (1,2): an enzymic disulfide (most commonly), a cysteine sulfenic acid (NADH oxidase and NADH peroxidase) (6), and a mixed enzymic Cys-S-S-CoA disulfide (coenzyme A disulfide reductase) (7,8).
LipDH, GR, trypanothione reductase, and mycothione reductase form a subgroup (Group 1) of the FDR enzyme family (1). The amino acid sequences of these proteins align well with one another along the entire polypeptide chain of ϳ50 kDa. Particularly noteworthy are the following four signature motifs (Fig.  1A). Two heptapeptide sequence motifs (GXGXXG), one at the N terminus and the other roughly in the middle of the polypeptide chain, are involved in recognizing the ADP portion of the FAD and pyridine nucleotide, respectively (9). The nonflavin redox center is an enzymic redox-active disulfide, which is invariably present in a CXXXXC motif and located near the N terminus of the polypeptide chain, whereas an essential His-Glu pair (HXXXXE), located at the C terminus of the adjacent monomer of the homodimeric enzyme, comprises the final motif (1).
The Mycobacterium tuberculosis genome encodes five open reading frames with significant amino acid sequence identity to the Group 1 FDR enzymes (Fig. 1A). Our previous studies have shown that lpdC (Rv0462) (10) and gorA (Rv2855) (11) code for LipDH and mycothione reductase, respectively. The sth (Rv2713) gene product bears high sequence identity (ϳ42%) with several well characterized STHs (12)(13)(14). The lpdA (Rv3303c) and lpdB (Rv0794c) genes were annotated as probable LipDHs based on sequence homology to LipDHs from other organisms (15). Although both lpdA and lpdB contain FAD and pyridine nucleotide binding motifs, they both lack essential catalytic residues known to be critical for LipDH catalysis. Specifically, lpdA lacks one of two cysteine residues that comprise the enzymic redox-active disulfide and involved in dithiol-disulfide interchange with lipoyl substrates (16,17), as well as the His-Glu pair involved in general acid catalysis (18), and lpdB lacks the His residue. Therefore, neither lpdA nor lpdB are expected to have LipDH activity.
In this report, we describe the cloning, expression in Escherichia coli, and purification of the LpdA protein. We show that, even though LpdA cannot catalyze reduction of disulfidebonded substrates, including lipoamide, lipoic acid, glutathione, cystine, 5,5Ј-dithiobis(2-nitrobenzoic acid), or 4,4Ј-dithiodipyridine, it catalyzes the reduction of nonspecific electron acceptors such as 2,6-dimethyl-1,4-benzoquinone (DMBQ) and 5-hydroxy-1,4-naphthaquinone (5-HNQ). We additionally crystallized and solved the three-dimensional crystal structure of LpdA in complex with FAD. The structure confirmed the absence of a redox-active disulfide and of a His-Glu pair in the active site and rationalizes the inability of this enzyme to catalyze reduction of disulfide-bonded substrates. We also show that LpdA co-purifies, not only with a tightly bound FAD, but with a tightly bound NADP ϩ per monomer as well. The kinetic consequences resulting from the tight binding of NADP ϩ to the enzyme are also described.
General Methods-Solution pH values were measured at 25°C with an Accumet® model 20 pH meter and Accumet® combination electrode standardized at pH 7.00 and 4.00 or 10.00. Protein purification was performed at 4°C using a fast protein liquid chromatography system (Amersham Biosciences). Column chromatography of small molecules was performed at room temperature on a Hewlett Packard 1100 HPLC system equipped with a UV-visible absorbance detector. Spectrophotometric assays were performed using a UVIKON XL double beam UVvisible spectrophotometer (BIO-TEK Instruments).
Cloning and Expression of M. tuberculosis LpdA-The lpdA gene (Rv3303c) from M. tuberculosis was amplified by PCR to generate blunt-ended DNA with NdeI and XhoI restriction sites at the ends. The oligonucleotide primers were: 5Ј-ATTCCATATGGTGACCCGCATCGT-GATCC-3Ј and 5Ј-CCGCTCGAGGTTACTCGTCGGTAGGTGGTG-3Ј. The amplified DNA product was ligated into the pCR®-Blunt plasmid and transformed into One Shot TM TOP10 cells following the instructions supplied by the manufacturer (Invitrogen). Plasmid DNA isolated from these cells was then digested with NdeI and XhoI, and the purified insert was ligated into purified plasmid pET-23a(ϩ) (Novagen) previously linearized with the same restriction enzymes. Cloning into these restriction sites adds a hexahistidine tag at the C terminus of LpdA. This recombinant plasmid was then transformed into competent E. coli BL21(DE3) cells (Novagen). The BL21(DE3) cells containing pET23a(ϩ):lpdA were grown at 37°C to an A 600 nm of 0.7 in Luria-Bertani medium containing 50 g/ml ampicillin. The cells were then cooled to 16°C, temperature was equilibrated for 2 h, and induction of protein expression by the addition of 0.5 mM isopropyl-1-thio-␤-D-galactopyranoside was allowed to proceed for 36 h at 16°C. DNA sequencing of the cloned lpdA gene showed that it was free of mutations.
FIG. 1. A, amino acid sequence alignment of residues that correspond to the redox-active disulfide and the His-Glu pair in five M. tuberculosis flavoprotein disulfide reductase homologues and human glutathione reductase (see introduction). B, active site of two-electron reduced human glutathione reductase with NADP ϩ bound to illustrate the structural arrangement of catalytic residues and NADP ϩ relative to the FAD (PDB code: 1GRB). C, active site of M. tuberculosis LpdA. Primed numbers indicate amino acids of the adjacent monomer of the dimer.
Purification of the LpdA Protein-All operations were carried out at 4°C. Cell paste (43 g) was suspended in 40 ml of 20 mM triethanolamine, pH 7.8, containing two tablets of Complete TM protease inhibitor mixture (Roche Applied Science), 15 mg of chicken egg white lysozyme (Sigma), 6 mg of bovine pancreas DNaseI (Roche Applied Science), 10 mM MgCl 2 , and stirred for 30 min. The cells were disrupted by sonication, and cell debris was removed by centrifugation at 20,000 ϫ g for 45 min. The supernatant was dialyzed for 4 h against 4 liters of 20 mM triethanolamine, pH 7.8, and the precipitate, which formed during dialysis, was removed by centrifugation as above. The dialysate (75 ml) was applied to a 50-ml column of nickel-nitrilotriacetic acid His⅐Bind Superflow (Novagen) equilibrated with 20 mM triethanolamine, pH 7.8 (Buffer A), containing 0.3 M NaCl. The column was washed with 150 ml of the same buffer and then eluted with a linear gradient (200 ml) from 0 to 0.25 M imidazole in Buffer A also containing 0.3 M NaCl. The yellow fractions containing the LpdA protein were pooled, dialyzed against 4 liters of Buffer A for 5 h, and applied to a 25-ml Mono Q column (Amersham Biosciences) equilibrated with Buffer A, and eluted with a linear gradient (200 ml) from 0 to 0.6 M NaCl in Buffer A. The fractions containing the LpdA protein, which eluted at ϳ0.3-0.4 M NaCl, were pooled, concentrated down to ϳ5 ml, and applied to a 320-ml calibrated HiTrap Sephacryl S-200 HR column (Amersham Biosciences) equilibrated with 50 mM Hepes, pH 7.5, containing 150 mM NaCl. The LpdA protein eluted at the volume calculated for a tetramer. The fractions containing the LpdA protein were pooled, concentrated down to 25 mg ml Ϫ1 (ϳ2.8 ml), and dialyzed against two changes of 4 liters of 10 mM Hepes, pH 7.5. Half of the protein was used to set up crystallization trays (see below), whereas the remainder was frozen at Ϫ70°C in 50% glycerol. Approximately 70 mg of purified LpdA was obtained from 43 g of cell paste.
Determination of Protein Concentration-The enzyme concentration was determined from ⑀ 458 nm ϭ 12,000 M Ϫ1 cm Ϫ1 for native LpdA, and turnover numbers were based on enzyme monomers. This value of the extinction coefficient was determined by normalizing the absorbance of the FAD cofactor at 450 nm (⑀ ϭ 11,300 M Ϫ1 cm Ϫ1 ) (19) of 1% SDSdenatured LpdA to that at 458 nm of native LpdA. The concentration of enzyme determined using the bicinchoninic acid protein assay (Pierce) with bovine serum albumin as standard agrees favorably (only 20% higher) with the value obtained using the above experimentally determined extinction coefficient, suggesting that the majority of the enzyme co-purifies with FAD bound.
Crystallization of LpdA and Data Collection-Crystals of LpdA were obtained using the hanging drop vapor diffusion method. Crystals were grown at room temperature from 5-l drops containing 3 l of protein solution (25 mg ml Ϫ1 in 10 mM Hepes, pH 7.5) and 2 l of precipitant solution (0.1 M trisodium citrate, pH 5.4 -5.8, 2.5-5.0% polyethylene glycol 6000). Crystals appeared in 3 (low pH) to 7 (high pH) days. Crystals were immersed in a solution of 10% polyethylene glycol 6000, 100 mM MES, pH 5.75, and 25% glycerol prior to vitrification by immersion in liquid nitrogen. X-ray diffraction data were measured at 75 K using an in-house Rigaku R-Axis IVϩϩ image plate detector and RU-H34 rotating anode x-ray generator equipped with Osmic Blue Optics and operating at 50 kV and 100 mA. All data were processed using DENZO/SCALEPACK (20). The LpdA crystals belong to space group R32 with unit cell dimensions a ϭ b ϭ 244.2 Å, c ϭ 104.8 Å, and there are two monomers per asymmetric unit yielding a solvent content of 58% (Table I).
Structure Determination and Refinement-The "quick soak" method (21) was used to screen potential heavy atom derivatives, four of which produced significant peaks in dispersive difference Patterson maps (KAu(CN) 2 , HgCl 2 , PtCl 6 , and trimethyllead acetate). Refinement of the heavy atom parameters, phase calculation, and solvent flattening were carried out with the program SOLVE/RESOLVE (22). The anomalous signal from the lead, mercury, and platinum data was also used in the phasing and improved the resultant electron density maps. Features within the initial SOLVE/RESOLVE MIRAS electron density map were used to determine the noncrystallographic 2-fold symmetry operator relating the two monomers per asymmetric unit. The SOLVE/RE-SOLVE MIRAS map was 2-fold-averaged and the auto-fitting program MAID (23) was used to get a start on the building of the LpdA model. Due to the limiting resolution (2.8 Å), a fair proportion of the structure was built manually within the program O (24). Refinement was conducted using CNS (version 1.0) (25) (see Table IV). Identification of Bound Cofactors-A 1-ml solution of 20 M LpdA in 50 mM sodium phosphate buffer, pH 6.0, was boiled for 3 min to denature the protein. Precipitated protein was removed by centrifugation followed by ultrafiltration of the yellow solution using an YM10 membrane (Millipore). The pH was then adjusted to 8.0 by adding 25 l of a 1 M Tris base solution. Aliquots (200 l) were either: 1) untreated, 2) treated with 0.05 unit/l CIP for 30 min, 3) treated with 0.03 unit/l T. acidophilum glucose dehydrogenase in the presence of 20 mM D-(ϩ)glucose for 30 min, or 4) treated with glucose dehydrogenase in the presence of D-(ϩ)-glucose for 30 min, followed by treatment with CIP for an additional 30 min. After removal of the enzymes by ultrafiltration using an YM10 membrane as above, aliquots (100 l) were applied to a 1 ml Mono Q column (Amersham Biosciences) equilibrated with 20 mM Tris, pH 8.0, and analyzed by HPLC. After an 8-ml wash with the same buffer, the bound cofactors were eluted with a linear gradient (30 ml) from 0 to 0.3 M NaCl in the same buffer. The pyridine nucleotides and FAD standards were prepared in 20 mM Tris, pH 8.0, and were standardized as described below.
Synthesis of 2Ј-Phospho-ADP-Ribose-2Ј-Phospho-ADP-ribose was prepared by enzymatic hydrolysis of NADP ϩ with N. crassa NAD ϩ nucleosidase. 8 mg of NADP ϩ was dissolved in 2 ml of 50 mM Tris, pH 7.5. N. crassa NAD ϩ nucleosidase (0.7 unit) was then added, and the reaction was allowed to proceed at 25°C for ϳ16 h. The reaction, after removal of the enzyme by ultrafiltration using an YM10 membrane (Millipore), was then applied to an 18 ml Mono Q column (Amersham Biosciences) and eluted with a 120-ml linear gradient from 0 to 0.5 M ammonium bicarbonate. The fractions containing 2Ј-phospho-ADP-ribose, which eluted at 0.3-0.35 M ammonium bicarbonate, were pooled. Excess ammonium bicarbonate was removed by repeatedly (three times) lyophilizing and re-dissolving the residue in water.
Data Analysis-Data were fitted using the nonlinear least squares curve-fitting programs of SigmaPlot 2000 for Windows version 6.00 (SPSS Inc.). Individual saturation curves were fitted to Equation 1, where V is the maximal velocity, A is the substrate concentration, and K is the Michaelis constant (K m ). Data showing parallel initial velocity patterns on reciprocal plots were fitted to Equation 2, where A and B, and K A and K B , are the concentrations, and Michaelis constants, respectively, for the substrates. Data displaying competitive inhibition were fitted to Equation 3, where I is the inhibitor concentration and K i is the inhibition constant in Equation 3.
Primary deuterium kinetic isotope effects were calculated from Equa- where E V/K and E V are the isotope effects minus 1 on V/K and V, respectively, and F i is the fraction of isotopic label.

RESULTS AND DISCUSSION
Cloning, Expression, and Purification of LpdA-To obtain large quantities of LpdA for structural and mechanistic studies, we cloned the lpdA gene into plasmid pET-23(a)ϩ and expressed the protein in E. coli BL21(DE3) cells. Although metal-chelation column chromatography was sufficient to achieve near-homogeneity, we found that additional chromatography on Mono Q and HiTrap Sephacryl S-200 HR columns gave a protein preparation that crystallized reproducibly and more readily. Approximately 70 mg of purified enzyme was obtained from 43 g of cell paste.
Although full-length LpdA does not readily ionize, electrospray/ionization mass spectrometry performed on separate samples and different protein preparations consistently gave two species with molecular masses of 52,543 Ϯ 3 Da and 52,413 Ϯ 3 Da compared with 52,554 Da expected for full-length LpdA and 52,422 Da with the N-terminal methionine post-translationally removed. To further characterize the LpdA polypeptide chain and to determine whether Glu 16 was post-translationally modified (see below), LpdA was treated with trypsin, and the resulting peptides were subjected to MALDI-TOF mass spectrometry. Most of the expected peptides were observed, including a peptide containing an unmodified Glu 16 .
Tightly-bound Cofactors-To determine the identity of the bound flavin cofactor (FAD or FMN), an aliquot of the soluble fraction resulting after heat-denaturation of LpdA was subjected to Mono Q ion exchange chromatography (see "Experimental Procedures"). A species with absorbance at 260 and 450 nm and that co-eluted with an FAD standard identifies the LpdA-bound flavin as FAD (see Fig. 2). Subsequent determination of the crystal structure of LpdA with FAD bound confirmed this result (see below). An additional peak with absorbance at 260 nm but not at 450 nm was also observed on the chromatograph, which co-eluted with an NADP ϩ standard. The following treatments prior to chromatography demonstrate that this species is NADP ϩ : 1) treatment with calf-intestinal alkaline phosphatase (CIP) converted this species to NAD ϩ , 2) incubation with glucose and T. acidophilum glucose dehydrogenase, which is NADP ϩ -specific, converted this species to NADPH, and 3) incubation with glucose and glucose dehydrogenase followed by CIP converted this species to NADH. The HPLC traces show that the area below the FAD and NADP ϩ peaks of the LpdA-bound species is similar to that in the standards establishing that the LpdA-bound FAD to NADP ϩ ratio is 1:1. Finally, the protein concentration of LpdA determined using ⑀ 458 nm of 12,000 M Ϫ1 cm Ϫ1 agrees well with that obtained using a protein assay (see "Experimental Procedures"). We conclude that homotetrameric LpdA contains one molecule of tightly but noncovalently bound FAD and NADP ϩ per monomer.
LpdA Cannot Reduce Disulfide-Bonded Substrates-No oxidation of NADH above the intrinsic NADH oxidase activity was observed when LpdA was incubated in the presence of 100 M NADH and Ն2 mM of the following disulfide-containing molecules: lipoamide, lipoic acid, oxidized dithiothreitol, glutathione, cystine, 5,5Ј-dithiobis(2-nitrobenzoic acid), or 4,4Ј-dithiodipyridine. Thus, LpdA cannot catalyze reduction of disulfide-bonded substrates. This was predicted and thus not surprising, because LpdA lacks one of two essential cysteine residues required for dithiol-disulfide interchange with disul- fide-bonded substrates as well as the His-Glu pair which functions as a general acid catalyst.
With NADH as the reductant and DMBQ or 5-HNQ as the oxidant, the activity of LpdA shows a nonlinear dependence on the concentration of the enzyme (Fig. 3). We considered two possibilities that could account for this behavior. 1) The activity of the enzyme might depend on its oligomeric state, which in turn depends on the enzyme concentration (40). For such a downward curvature on a plot of activity versus concentration of LpdA, the least active species would correspond to the higher order oligomer. However, we have no evidence to support such a change in oligomeric state. Gel-filtration chromatography, for example, showed a single homotetrameric species. 2) LpdA might co-purify with a tightly, but noncovalently, bound inhibitor (40,41). As the concentration of the enzyme in an assay is increased, the concentration of the inhibitor, which co-purifies with LpdA, increases proportionately. But, the ratio of free enzyme (active form) to inhibitor-bound enzyme (inactive form) decreases as the enzyme (and, hence, inhibitor) concentration is increased. The concentration range where such a deviation from linearity on a plot of activity versus concentration of LpdA is most pronounced is, of course, governed by the dissociation constant (K D ) that describes the following equilibrium (Reaction 1).

REACTION 1
We will show next that this explanation is sufficient to account for the observed, unusual kinetic behavior and that the tightly bound inhibitor is NADP ϩ .
Pre-treatment of native LpdA with CIP at 4°C for 4 h activates the NADH-dependent reduction of DMBQ or 5-HNQ catalyzed by LpdA significantly, and this activity now shows a linear dependence on enzyme concentration (Fig. 3). Activation results from CIP-catalyzed hydrolysis of the 2Ј-phosphate of NADP ϩ to generate NAD ϩ , which does not bind to LpdA as tightly. Therefore, with NADH as reductant, we used CIPtreated LpdA to obtain the kinetic parameters tabulated in Table II. With NADPH as reductant, CIP treatment was not necessary because the activity is linear with LpdA concentration (Fig. 3, inset). Thus, compared with NADH, NADPH can more effectively compete with NADP ϩ for the pyridine nucleotide-binding site, strongly suggesting that the physiological substrate/product pair is NADPH/NADP ϩ rather than NADH/NAD ϩ .
We also performed inhibition studies to determine the affinity of LpdA for NADP ϩ . We prepared LpdA that was free of NADP ϩ by treatment with CIP as above, followed by inhibition of CIP, which is Zn 2ϩ -dependent, by overnight incubation with 20 mM EDTA at 4°C. With NADH and DMBQ as the substrate pair, NADP ϩ was competitive with respect to NADH with a K i value of 6 Ϯ 2 nM (data not shown). This very low K i value demonstrates the high affinity of LpdA for NADP ϩ and explains why LpdA co-purifies with a stoichiometric quantity of bound NADP ϩ , even after three column chromatography steps and extensive dialysis.
To determine the significance of the nicotinamide portion of NADP ϩ to binding, we prepared 2Ј-phospho-ADP-ribose and determined its K i value using the same assay that was used for NADP ϩ . The K i value of 800 Ϯ 50 nM demonstrates that 2Јphospho-ADP-ribose is also a good inhibitor of LpdA, although not as potent as NADP ϩ . It should be noted that the binding interactions between the enzyme and the nicotinamide portion of NADP ϩ might not be fully responsible for the tighter binding of NADP ϩ to LpdA relative to 2Ј-phospho-ADP-ribose. The introduction of a hydroxyl group at the C 1 -position of the 2Јphospho-ADP-ribose ring by the action of NAD ϩ nucleosidase on NADP ϩ may additionally alter the steric, solvation, and associated conformational properties of the sugar.
Steady-state Kinetics-We began the steady-state kinetic characterization of LpdA using DMBQ as the oxidant. The low K m values of 11 and 1.6 M for DMBQ with NADH and NADPH as the reductant (Table II), respectively, suggested that DMBQ might resemble the physiological oxidant of LpdA and, thus, warranted further investigation. We next examined 5-HNQ as the oxidant, whose single-electron reduction potential of Ϫ90 mV (38) is close to that of DMBQ of Ϫ80 mV (37). The nearequality in the difference in reduction potential between NADH (Ϫ320 mV) and these two quinones, 230 versus 240 mV, essentially removes any thermodynamic bias in reactivity and should, therefore, allow any differences in binding to the enzyme to be expressed. Although 5-HNQ is bulkier than DMBQ (Scheme 1), the k cat and K m values for the two quinones are similar.
Oxidation of the reduced enzyme by quinones in the Group 1 FDR enzymes is thought to proceed from the reduced flavin to the quinone acceptor via one-or two-electron transfer (36,38,39). We have shown previously (36) by stopped-flow spectrophotometry that oxidation of four-electron reduced M. tuberculosis LipDH (lpdC) by DMBQ (i) shows bimolecular kinetics, that is, LipDH has no affinity for DMBQ, and (ii) occurs in the absence of NAD ϩ , as expected for a ping-pong mechanism. Because LpdA cannot distinguish between DMBQ and 5-HNQ, it is not expected to bind either of them with significant affinity. Therefore, ternary complexes such as NADH⅐E ox ⅐DMBQ ox and NAD ϩ ⅐E ox ⅐DMBQ red , which are required for sequential mechanisms, are unlikely to form. Thus, and as we will show next, a ping-pong mechanism for LpdA is more likely.
We observed a pattern of parallel lines on a reciprocal plot when NADH was varied at five fixed-levels of 5-HNQ (data not shown), which is consistent with a ping-pong mechanism. Because ping-pong mechanisms do not require the formation of ternary complexes before reaction, LpdA should catalyze the interconversion of NADH and NAD ϩ (or any of its analogs; e.g. thio-NAD ϩ ) in the absence of 5-HNQ or other oxidants. Incubation of CIP-treated LpdA with NADH and thio-NAD ϩ resulted in the formation of thio-NADH and NAD ϩ at a maximal rate (k cat ) of 3.6 s Ϫ1 , which is comparable to that of the NADH and 5-HNQ substrate pair (k cat ϭ 2.5 s Ϫ1 , Table II). Similarly, LpdA catalyzes the formation of thio-NADH and NADP ϩ from thio-NAD ϩ and NADPH at a rate of ϳ0.0090 s Ϫ1 , which is comparable to that of the NADPH and 5-HNQ (or DMBQ) substrate pair (k cat ϭ 0.0080 s Ϫ1 , Table II). These experiments provide strong evidence that the steady-state kinetic mechanism is ping-pong when NAD(P)H and 5-HNQ (or DMBQ) is the substrate pair.
Although the k cat /K m value for NADH is only ϳ5.3-fold higher than that for NADPH, the individual k cat and K m values are drastically different for these two reduced pyridine nucleotides; the k cat value with NADH as reductant is ϳ340-fold higher than that for NADPH, and the K m value for NADPH is ϳ66-fold lower than that for NADH. Although NADH has a higher k cat /K m value than NADPH, we propose that the physiological substrate/product pair of LpdA is NADPH/NADP ϩ because a number of observations suggest that they bind to the enzyme significantly tighter, and, hence, they will outcompete NADH/NAD ϩ for the pyridine nucleotide binding site in vivo: 1) the K m value for NADPH is significantly lower than that for NADH, and 2) LpdA has a high affinity for NADP ϩ .
Isotope Effects-An attractive explanation for the very low turnover number with NADPH as reductant is that release of NADP ϩ from the E red ⅐NADP ϩ complex before reaction with quinones, which is required for a ping-pong mechanism, is very slow. We addressed this question by comparing the magnitude of the isotope effects on V and V/K with [4S-2 H]NADH versus [4S-2 H]NADPH. Isotope effects on V and V/K probe different portions of the reaction for rate-limiting steps. For ping-pong mechanisms, which is the case for LpdA, D (V/K) NADH probes whether hydride transfer from NADH to the FAD limits the reductive half-reaction, whereas D V probes whether the hydride transfer step limits the overall reaction (excluding substrate binding steps) (42).
The first step in the catalytic cycle of the FDR enzymes involves transfer of the C 4 -proS hydride of the dihydronicotinamide ring of NAD(P)H to the N 5 position of the FAD (1, 2). Because LpdA belongs to the FDR family of enzymes, we measured the isotope effect arising from deuterium substitution at the C 4 -proS position of the dihydronicotinamide ring of NADH using a saturating concentration of DMBQ or 5-HNQ as the oxidant (Table III). The magnitude of the isotope effect on D (V/K) NADH of ϳ2 is in the normal range for primary deuterium kinetic isotope effects and is too large to be a secondary effect, thus establishing that the enzyme promotes transfer of the C 4 -proS hydride of NADH. This result also suggests that the oxidative half-reaction, oxidation of the reduced flavin by quinones, is also partially rate-limiting, because the magnitude of D V of 1.4 is smaller than that of D (V/K) NADH of ϳ2 (42).
With [4S-2 H]NADPH, no isotope effects were observed (Table III). Although the absence of an isotope effect on D (V/ K) NADPH could be explained by NADPH binding more tightly than NADH, the absence of an isotope effect on D V requires that other steps were also affected. We propose that release of NADP ϩ from the E red ⅐NADP ϩ complex is slow; otherwise D V would have remained at the value of 1.4 (42).
For ping-pong mechanisms, the k cat /K m value for one substrate is independent of the identity of the co-substrate. Although this is true for k cat /K m (NADH) using either DMBQ or 5-HNQ as the oxidant, it is not true for k cat /K m (DMBQ) and k cat /K m (5-HNQ) using either NADH or NADPH as reductant (see Table II). We suggest that the slow release of NADP ϩ from SCHEME 1. Chemical structures of 2,6-dimethyl-1,4-benzoquinone (DMBQ) and 5-hydroxy-1,4-naphthaquinone (5-HNQ). the E red ⅐NADP ϩ complex results in a significantly lowered steady-state concentration of E red (the form of the enzyme that reacts with quinones), relative to E red ⅐NADP ϩ , and that this may account for the 23-to 50-fold lower k cat /K m (DMBQ) and k cat /K m (5-HNQ) values with NADPH as reductant versus NADH as reductant.
Quinone Reduction and the Normal Catalytic Pathway-We present a working model, which is consistent with our findings from the steady-state kinetic experiments (Scheme 2). A number of observations mentioned previously strongly suggest that NADPH rather than NADH is the physiological reductant. NADPH binds to E ox and reduces the FAD to form an E red ⅐NADP ϩ intermediate. The low turnover number and absence of isotope effects on D (V/K) NADPH and D V for the NADPHdependent reduction of quinones is consistent with rate-limiting release of NADP ϩ from the E red ⅐NADP ϩ complex before rapid reaction with quinones. We think that NAD ϩ release from the E red ⅐NAD ϩ complex, on the other hand, is significantly faster than release of NADP ϩ from the E red ⅐NADP ϩ complex. As a result, the turnover number for the NADH-dependent reduction of quinones is ϳ340-fold higher, hydride transfer now becomes partially rate-limiting, thus giving rise to significant isotope effects on both D (V/K) NADH and D V. We speculate that the slow release of NADP ϩ from the E red ⅐NADP ϩ complex serves to protect the flavin from nonspecific oxidation by cellular oxidants until the physiological oxidant, X ox , binds to form a ternary E red ⅐NADP ϩ ⅐X ox complex. A somewhat similar protection occurs in the Group 1 FDR enzymes, where NAD(P) ϩ remains firmly bound to the enzyme until electrons are transferred to the redox-active disulfide. Electron transfer from the reduced flavin to X ox then occurs, followed by release of products. Thus, we think that oxidation of two-electron reduced enzyme by quinones is off the normal catalytic pathway (see Scheme 2).
Although all other Group 1 FDR enzymes are homodimers in solution (1,2), LpdA is a homotetramer. There is a dimer per asymmetric unit in the crystal form that was used to solve the structure. This dimer is similar to the physiologically relevant dimer of the Group 1 FDR enzymes and is stabilized by interactions involving the two interface domains of each monomer (Fig. 4A). The other dimer interface is generated by one of the crystallographic two-folds yielding a dimer of dimers with 222 symmetry (Fig. 4B). A total of 3530 Å 2 of solvent-accessible surface area is buried upon forming this second dimer interface. To our knowledge Enterococcus faecalis NADH peroxidase, which together with E. faecalis NADH oxidase and Staphylococcus aureus coenzyme A disulfide reductase form a separate subgroup (Group 3) of the FDR enzymes (1), is the only other FDR enzyme that forms tetramers (46). However, a comparison of the crystal structure of LpdA and NADH peroxidase indicates that the surfaces used to form the second dimerdimer interface are vastly different for the two enzymes. It should also be noted that the active sites of LpdA and NADH peroxidase are different and that LpdA cannot reduce hydrogen peroxide.
Active Site of LpdA-The active sites of human GR and LpdA are compared in Fig. 1 (B and C). As predicted by the amino acid sequence alignment, Ala 43 and Cys 48 of LpdA are located at positions that are normally occupied by the redox-active disulfide of the Group 1 FDR enzymes. Likewise, Tyr 450 and Gly 455 occupy positions that the His-Glu pair normally does. The absence of a redox-active disulfide and of a His-Glu pair in the active site of LpdA explains its inability to catalyze reduction of disulfide-bonded substrates. The only known FDR enzyme that uses a single enzymic cysteine residue to catalyze reduction of a disulfide-bonded substrate is coenzyme A disulfide reductase (7,8,47). It is unlikely that LpdA uses a mechanism analogous to that of coenzyme A disulfide reductase to catalyze reduction of an unknown disulfide-bonded substrate, because Cys 48 is not completely conserved among LpdA orthologues, which are only present in the actinomycetes. In the Streptomyces avermitilis and Tropheryma whipplei LpdA pro-  teins, which are 54 and 47% identical with M. tuberculosis LpdA, respectively; the corresponding residue is a valine.
We observed electron density on the si face of the FAD, ϳ10 Å away from the N 5 position of the FAD, which could not be accounted for by the protein (not shown). This density had a peak of 23 in the 2-fold averaged F o Ϫ F c map directly adjacent to the side chain of Glu 16 and significant electron density at the 4 -5 level directly surrounding that peak. This was suggestive of a modified Glu 16 and perhaps may have shed light on the physiological function of LpdA. To obtain independent evidence for a modified amino acid, LpdA was treated with trypsin, and the resulting peptides were subjected to MALDI-TOF mass spectrometry. Most of the expected peptides were observed, including a peptide containing an unmodified Glu 16 . We note that Glu 16 is not completely conserved among LpdA orthologues. In Tropheryma whipplei LpdA, the corresponding residue is a serine. Thus, we believe that the spurious electron density is most consistent with a metal ion, perhaps Ni 2ϩ from the nickel-nitrilotriacetic acid column chromatography step, and a bound citrate from the crystallization conditions. How-ever, due to the limited resolution of the data, this density was left unmodeled.
Structural Basis for the Specificity of LpdA for NADP ϩ -Even though NADP ϩ binds very tightly to LpdA in aqueous solution and neutral pH, no electron density corresponding to NADP ϩ was observed in the structure. Because we were interested in the structural basis for the tight binding of NADP ϩ to LpdA, we modeled NADP ϩ in the pyridine nucleotide-binding site based on the co-crystal structure of two-electron reduced human GR with bound NADP ϩ (48). The interactions of the 2Ј-phosphate of NADP ϩ with human GR and LpdA are compared in Fig. 5 (A and B).
Specificity for NADP ϩ versus NAD ϩ in many pyridine nucleotide dehydrogenases is brought about, in part, by the identity of the amino acid located 19 -21 residues downstream of the last glycine in the GXGXXG motif (9, 49 -51). NADP ϩ -specific dehydrogenases usually have an arginine at this position, whose side chain interacts with the 2Ј-phosphate of NADP ϩ by electrostatic and hydrogen bonding interactions. NAD ϩ -specific dehydrogenases, on the other hand, usually have an acidic amino acid at this position, the carboxylate side chain of which makes hydrogen bonds with the hydroxyl groups of the ribose containing the adenine ring. In human GR, which is NADP ϩspecific, this residue is Arg 218 . The structure of GR with NADP ϩ bound shows electrostatic and hydrogen-bonding interactions between the side chain of Arg 218 and the 2Ј-phosphate of NADP ϩ (Fig. 5A). Arg 224 and His 219 also contribute to the specificity by additional electrostatic and hydrogen bonding interactions.
The amino acid sequence alignment shows that the residues that correspond to Arg 218 , His 219 , and Arg 224 in human GR are Ser 213 , Gln 214 , and Pro 219 in LpdA, respectively, none of which are cationic (Fig. 5). The modeled LpdA⅐NADP ϩ complex (Fig.  5B) shows possible hydrogen-bonding interactions between Ser 213 and Gln 214 with the phosphate, whereas Pro 219 does not seem to participate in any direct interactions. We propose that a large part of the high affinity of LpdA for NADP ϩ arises from interactions between the side chain of Arg 245 (the corresponding residue in GR is Gln 250 ) and the phosphate. In the structure, the side chain of Arg 245 points away from the 2Ј-phosphate of the modeled NADP ϩ into solvent. However, Arg 245 along with Ser 213 and Gln 214 can be re-positioned without steric hindrance to interact favorably with the 2Ј-phosphate of the modeled NADP ϩ .
Conclusions-The functional characterization of the entire complement of enzymatic activities in any organism is a challenge that will take many decades. We have shown that the M. tuberculosis LpdA cannot catalyze reduction of lipoyl substrates or any of the disulfide-bonded substrates tested, despite its significant sequence homology to lipoamide dehydrogenase. LpdA lacks three residues that are essential for disulfide reduction chemistry, and therefore the annotated function of this enzyme was incorrect. We have demonstrated two relevant catalytic activities of LpdA. The first is a transhydrogenase activity that allows for the pools of reduced pyridine nucleotides to remain equilibrated under conditions in which one or the other pool becomes too oxidized, preventing efficient biosynthesis in pathways that require reducing equivalents from NADPH. Although the M. tuberculosis genome contains a putative soluble pyridine nucleotide transhydrogenase orthologue (Rv2713), the protein has never been expressed or enzymatically characterized. The second is a quinone reductase activity that could enable transfer of reducing equivalents from the reduced pyridine nucleotide pool to the electron transport chain. Although this activity is low with LpdA and watersoluble quinone acceptor substrates, the activity with membrane-associated ubiquinones may be physiologically relevant. Finally, there may be additional activities of LpdA that are novel to the flavoprotein disulfide reductase family. Future studies of LpdA and LpdB will address these issues of functional genomics in M. tuberculosis.