Influence of Allosteric Regulators on Individual Steps in the Reaction Catalyzed by Mycobacterium tuberculosis 2-Hydroxy-3-oxoadipate Synthase*

Background: 2-Hydroxy-3-oxoadipate synthase is predicted to be essential for growth of Mycobacterium tuberculosis and is subject to allosteric regulation. Results: The role of regulators was characterized by kinetics and detection of thiamin diphosphate-bound covalent intermediate distribution. Conclusion: AcCoA activates steps leading to a predecarboxylation intermediate; GarA chiefly inhibits postdecarboxylation steps. Significance: This novel regulatory regime in thiamin diphosphate enzymes provides new leads to inhibition. Allosteric regulation often controls key branch points in metabolic processes. Mycobacterium tuberculosis 2-hydroxy-3-oxoadipate synthase (HOAS), a thiamin diphosphate (ThDP)-dependent enzyme, produces 2-hydroxy-3-oxoadipate using 2-ketoglutarate and glyoxylate. The proposed chemical mechanism in analogy with other ThDP-dependent carboligases involves multiple ThDP-bound covalent intermediates. Acetyl coenzyme A is an activator, and GarA, a forkhead association domain-containing protein known to regulate glutamate metabolism, is an allosteric inhibitor of HOAS. Steady state kinetics using assays to study the first half and the full catalytic cycle suggested that the regulators act at different steps in the overall mechanism. To explore the modes of regulation and to test the effects on individual catalytic steps, we performed circular dichroism (CD) studies using a non-decarboxylatable 2-ketoglutarate analog and determined the distribution of ThDP-bound covalent intermediates during the steady state of the HOAS reaction using one-dimensional 1H gradient carbon heteronuclear single quantum coherence NMR. The results suggest that acetyl coenzyme A acts as a mixed V and K type activator and predominantly affects the predecarboxylation steps. GarA does not inhibit the formation of the predecarboxylation analog and does not affect the accumulation of the postdecarboxylation covalent intermediate derived from 2-ketoglutarate; however, it decreases the abundance of the product ThDP adduct in the HOAS pathway. Thus, the two regulators act on different halves of the catalytic cycle in an unusual regulatory regime.


Allosteric regulation often controls key branch points in metabolic processes. Mycobacterium tuberculosis 2-hydroxy-3oxoadipate synthase (HOAS), a thiamin diphosphate (ThDP)dependent enzyme, produces 2-hydroxy-3-oxoadipate using 2-ketoglutarate and glyoxylate. The proposed chemical mechanism in analogy with other ThDP-dependent carboligases involves multiple ThDP-bound covalent intermediates. Acetyl coenzyme A is an activator, and
GarA, a forkhead association domain-containing protein known to regulate glutamate metabolism, is an allosteric inhibitor of HOAS. Steady state kinetics using assays to study the first half and the full catalytic cycle suggested that the regulators act at different steps in the overall mechanism. To explore the modes of regulation and to test the effects on individual catalytic steps, we performed circular dichroism (CD) studies using a non-decarboxylatable 2-ketoglutarate analog and determined the distribution of ThDP-bound covalent intermediates during the steady state of the HOAS reaction using one-dimensional 1 H gradient carbon heteronuclear single quantum coherence NMR. The results suggest that acetyl coenzyme A acts as a mixed V and K type activator and predominantly affects the predecarboxylation steps. GarA does not inhibit the formation of the predecarboxylation analog and does not affect the accumulation of the postdecarboxylation covalent intermediate derived from 2-ketoglutarate; however, it decreases the abundance of the product ThDP adduct in the HOAS pathway. Thus, the two regulators act on different halves of the catalytic cycle in an unusual regulatory regime.
Thiamin diphosphate (ThDP) 2 -dependent 2-ketoglutarate: glyoxylate carboligase activity and the resultant production of 2-hydroxy-3-oxoadipate (HOA) reported in both bacterial and mammalian cell extracts (1)(2)(3)(4)(5) are assigned to the E1 component of the 2-ketoglutarate dehydrogenase complex (E1o) based on experiments with the purified enzymes in vitro (see Scheme 1). In mycobacterial cell extracts, the 2-ketoglutarate dehydrogenase complex activity could not be detected, and the E2o component could not be identified (6). However, the product of the gene Rv1248c (annotated as the E1o of Mycobacterium tuberculosis 2-ketoglutarate dehydrogenase complex) could catalyze the synthesis of HOA in vitro using either 2-ketoglutarate and glyoxylate or a small molecule extract of mycobacteria as a source of substrates. Overexpression of Rv1248c in M. tuberculosis led to an increase in the abundance of HOA (detected as 2-hydroxylevulinate by mass spectrometry), suggesting that the HOA synthase activity arises from Rv1248c in vivo (7). The metabolic fate of HOA is unknown. HOA synthase activity has been implicated in glyoxylate detoxification in mouse liver (8).
Acetyl coenzyme A (AcCoA) was recently identified as an allosteric activator, and GarA (Rv1827) was identified as an allosteric inhibitor of Mycobacterium smegmatis 2-ketoglutarate decarboxylase, a homolog of HOAS (9), with respect to both its 2-ketoglutarate decarboxylase activity and its HOA synthase activity. However, GarA inhibited the two activities to a different extent. To gain more insight into the function of this still enigmatic enzyme, we attempted to identify the individual chemical steps affected by these allosteric modulators. We used circular dichroism (CD) spectroscopy to monitor accumulation of a stable predecarboxylation intermediate analog and chemical quench NMR to assess rate-limiting steps to gain white suspension was centrifuged at 35,000 ϫ g for 25 min, and the pellets were collected, dissolved in Buffer A, and dialyzed against 4 liters of Buffer A overnight. Batch purification was performed using a Q Sepharose FF column (20 ml). The column was washed with Buffer A, and bound proteins were eluted with Buffer B (Buffer A ϩ 1 M NaCl) with a linear gradient in 15 column volumes. Fractions containing HOAS (as determined by SDS-PAGE) were pooled, concentrated, and dialyzed against Buffer A overnight. Additional purification was achieved by using a Mono Q (8-ml) column equilibrated with Buffer A and eluted using Buffer B with a linear gradient in 11 column volumes. Fractions containing HOAS were pooled and purified further on Sephacryl S-100 HR (320 ml) equilibrated with Buffer A and eluted with Buffer A. Fractions containing HOAS were pooled, concentrated, flash frozen in liquid N 2 , and stored at Ϫ80°C until use. In a typical batch, protein was purified to ϳ90% homogeneity according to Coomassie staining and densitometry analysis. The N-terminal sequence of a sample of the purified recombinant protein was confirmed by Edman degradation.
Cloning, Overexpression, and Purification of GarA-The M. tuberculosis garA gene (Rv1827) was PCR-amplified from H37Rv DNA using the forward primer 5Ј-TCAGTGACGGA-CATGGAACCCGG-3Ј and the reverse primer 5Ј-TCACGGG-CCCCCGGTACT-3Ј. The PCR fragment containing additional TCA coding for N-terminal Ser (to assist in complete Ulp protease cleavage of the SUMO tag) was cloned into the pET-SUMO vector using the TA Cloning method. After the sequence had been verified, the resulting plasmid was transformed into E. coli BL21(DE3) CodonPlus-RILP competent cells. Three liters of LB containing 50 mg/liter kanamycin were inoculated with 60 ml of overnight starter culture and incubated at 37°C with shaking at 220 rpm. At an A 600 of ϳ0.6 -0.8, target protein overexpression was induced with 1.0 mM isopropyl ␤-D-1-thiogalactopyranoside, and the culture was incubated at 37°C while shaking at 220 rpm for 6 h. The harvested cells were washed with PBS and resuspended in 40 ml of 25 mM Tris (pH 8.0) (Buffer C). PMSF (1 mM final concentration), lysozyme (0.6 mg/ml final concentration), one protease inhibitor mixture tablet, and 1 g/ml DNase were added to the suspension and incubated in a 37°C water bath for 30 min. The lysozyme-treated cell suspension was passed twice through a French press operating at 1000 p.s.i. in an ice-cooled chamber. All further operations were performed at 4°C unless otherwise noted. The lysed cells were clarified by centrifugation at 35,000 ϫ g for 25 min. Batch purification was performed with a His-Trap FF (5 ml) column washed with Buffer C, and the protein was eluted with Buffer D (Buffer C ϩ 500 mM imidazole) with a linear gradient in 7 column volumes. Fractions containing the target SUMO fusion protein were pooled and treated with ubiquitin-like protein 1 (Invitrogen) at 1 g/mg protein and incubated at ambient temperature for 6 h. Next, the mixture was loaded onto a His-Trap FF column (5 ml) equilibrated with Buffer C and eluted with Buffer D. The flow-through containing GarA without the fusion SUMO tag was concentrated and dialyzed against Buffer C overnight. Aliquots of the dialyzed protein were flash frozen in liquid N 2 and stored at Ϫ80°C until use.
Enzyme Assays-Protein concentrations were determined using the Bradford assay method (12). Ferricyanide reductase assays were performed as described (6). Succinate semialdehyde production was monitored using horse liver alcohol dehydrogenase in the presence of NADH, and the disappearance of NADH was monitored at 340 nm. The activity was measured in the presence of 20 mM 2-ketoglutarate in a 1-ml volume of standard assay buffer (20 mM Bis-Tris, 1 mM ThDP, 5 mM MgCl 2 , 0.2 mg/ml NADH, 0.08 mg/ml alcohol dehydrogenase (pH 6.5)) at 37°C. The reaction was started with addition of 1 M HOAS. Assays were performed using a Uvikon XL UVvisible spectrophotometer (SI Analytics GmbH, Mainz, Germany) equipped with a circulating water bath and thermospacers operating at 37°C. Assays were performed in triplicate, and the experiment was performed at least twice.
pH Dependence of Enzyme Activity-A three-component buffer system (50 mM acetic acid, 50 mM MES, and 100 mM Tris) (13) containing additional 500 M ThDP and 5 mM MgCl 2 was used in the pH range 4.5-8.5 to assay for the ferricyanide reductase and carboligase activities. The pH dependence of the kinetic parameters for the ferricyanide reductase activity was determined using the assay modified for the plate reader format described in the following section.
CD Spectroscopy-CD spectra were recorded on a Chirascan CD spectrometer (Applied Photophysics, Leatherhead, UK) in a 1-cm-path length cell in the near-UV (280 -400-nm) wavelength region.
Steady State Kinetics of 2-Hydroxy-3-oxoadipate Synthase Activity-Time-dependent synthesis of chiral 2-hydroxy-3oxoadipate by HOAS from 2-ketoglutarate and glyoxylate was monitored continuously at 278 nm in the kinetics mode by CD. A typical reaction mixture in a 2.4-ml cuvette contained 20 mM Bis-Tris (pH 6.5), 200 M ThDP, 5 mM MgCl 2 , and varying concentrations of either 2-ketoglutarate or glyoxylate in the presence of a fixed (10 mM) concentration of the second substrate. The reaction was started by addition of 500 nM HOAS and was monitored for 500 s at 37°C, and data points were averaged every 10 s. Lineweaver-Burk plots were obtained by varying 2-ketoglutarate (0.
In this expression, v is the steady state linear velocity, V max is the maximal velocity at saturating substrate, E is the enzyme concentration, S is the concentration of the varied substrate, K m is the Michaelis constant for the varied substrate, and K i is the apparent substrate inhibition constant at high substrate concentration.
Plate Reader Assays-A typical ferricyanide reductase assay reaction mixture (200 l) in 20 mM Bis-Tris (pH 6.5) contained 5 mM MgCl 2 , 1.6 mM K 3 Fe(CN) 6 In these expressions, L is the concentration of the varied regulator, ␤ is the value of the kinetic parameters at various L, ␤ 0 is the value in the absence of the regulator, ␤ max is the maximal value, and ␤ min is the minimal value. K L is the concentration of the regulator at which the half-maximal regulatory effect is observed on each of the parameters. K R is the constant for loss of activation, and n is the Hill coefficient. K i of MSP was estimated by global fit of inhibition performed using Equation 6 where the varied substrate S is 2-ketoglutarate.
The effect of regulators on the steady state linear velocity in the (R)-HOA synthase assay was determined by fitting the data to Equation 3 for AcCoA and Equation 5 for GarA. Here, ␤ is the steady state linear velocity v/E where E is the enzyme concentration. L is the concentration of the regulator, and K L is the apparent activation or inactivation constant. Data for GarA, plotted in the fractional inhibition form, were fitted to Equation 3. Here, ␤ is the percentage of inhibition, ␤ 0 was set to zero, and ␤ max is the maximal percentage of inhibition. L is the concentration of GarA, and K L is the apparent inactivation constant, K inact . K i of GarA was estimated by global fit of inhibition performed using Equation 7 where the varied substrate S is glyoxylate. All data were fit by nonlinear regression using SigmaPlot v.10.0.
Titration of HOAS with MSP-To a 2.4-ml cuvette containing HOAS (2.75 mg/ml; 20 M) in 20 mM KH 2 PO 4 (pH 7.0) with additional 200 M ThDP and 5 mM MgCl 2 , incremental amounts of MSP were added from a 1 M stock solution to obtain the desired final concentrations (0 -20 mM). After mixing for 10 s and incubation for an additional 30 s, near-UV (280 -400 nm) CD spectra were collected at 25°C. The procedure was essentially the same for experiments with HOAS (20 M) in the presence of saturating concentrations of the allosteric regulator, either AcCoA (500 M) or GarA (50 M). For quantitative analysis, the amplitude of the CD band pertaining to the 1Ј,4Јiminopyrimidine tautomer of the MSP-ThDP complex at 302 nm corrected for dilution was plotted against MSP concentration. Apparent dissociation constants (K d app ) were calculated by fitting the data to the Hill function described in Equation 8 using SigmaPlot v.10.0.
In this expression, CD 302 is the observed CD signal at the given wavelength, CD 0 is the CD signal of the protein at this wavelength in the absence of MSP, CD 302 max is the maximum CD signal at saturation with MSP, [MSP] is the concentration of substrate analog, and n is the Hill coefficient. were analyzed by a one-dimensional 1 H gradient carbon heteronuclear single quantum coherence experiment on a Varian 600-MHz NMR instrument. Data acquisition was at 25°C, and sample pH was ϳ0.75. 2816 transients in the 1 H dimension were collected with a recycle delay time of 2 s. Data processing (exponential window function and zero filling) and analyses (base-line correction, integration, and signal-to-noise ratio estimates) were performed using ACD/NMR processor academic edition (v12.01, Advanced Chemistry Development, Inc., Toronto, Canada). The relative integrals of C6Ј-H signals of respective ThDP-bound covalent intermediates represent their relative ratios after correction for the amount of unbound ThDP estimated using Equations 9 and 10. A, near-UV CD spectra acquired at the indicated times to detect accumulation of (R)-HOA produced by HOAS (500 nM) from 2-ketoglutarate (10 mM) and glyoxylate (10 mM) at 37°C in Bis-Tris (pH 6.5) containing additional 200 M ThDP and 5 mM Mg 2ϩ . B, steady state progress curves of the reaction at the indicated concentrations of 2-ketoglutarate (2KG) in the presence of 10 mM glyoxylate. C, disappearance of (R)-HOA signal at 278 nm at longer time scales. After the negative maximum was reached in the presence of 1 mM 2-ketoglutarate and 10 mM glyoxylate, the progress curve for decarboxylation was acquired by monitoring the timedependent changes in CD at 278 nm. The blue line is the experimental data. The black curve is the regression fit of the data to CD 278 (t) ϭ CD 278 (0) ϩ A(1 Ϫ e Ϫkt ) where CD 278 (t) is CD (millidegrees (mdeg)) at time t (s) and k (s Ϫ1 ) is the first-order rate constant.

Cloning, Expression, and Purification
The bioinformatics prediction of the N terminus of HOAS has been revised since the M. tuberculosis enzyme was first studied (14). To ensure that our recombinant protein included the native N terminus, we identified the N-terminal start sequence of endogenous HOAS. We immunoprecipitated HOAS from M. tuberculosis H37Rv lysates using ␣-HOAS antibodies (7) and subjected the band corresponding to the target protein to Edman sequencing. The first 5 amino acids were determined to be ANISS in agreement with recent observations (15). Next, HOAS was cloned into the pET-11c vector, and tagless recombinant HOAS was overexpressed and purified to Ͼ90% purity as judged by SDS-PAGE. The N-terminal start sequence of the recombinant enzyme was confirmed by Edman sequencing. GarA was PCR-amplified and cloned into the pET-SUMO vector, and recombinant GarA was overexpressed and purified to Ͼ95% purity. Approximately 15 mg of pure HOAS and 15 mg of pure GarA were obtained per liter of culture.

Detection and Characterization of HOA
Carboligase products of ThDP-dependent enzymes can be detected by CD spectroscopy in the near-UV region (250 -400 nm) because the products are chiral. A negative band at max ϭ 278 nm showed a time-dependent increase in the presence of HOAS (500 nM), 2-ketoglutarate, and glyoxylate (Fig. 1A). This band, suggestive of accumulation of a chiral product, was observed in the supernatant after enzyme removal and did not appear when one of the three components was omitted. Earlier, 1 H NMR and LC-MS analyses of this reaction assigned the product as 2-hydroxy-3-oxoadipate, which is decarboxylated non-enzymatically to hydroxylevulinate (7). By analogy with carboligase reactions of other ThDP-dependent enzymes (16,17), we assigned the observed band to the (R)-HOA enantiomer. The progress curves measured at 278 nm allowed determination of the steady state linear velocities of HOAS-catalyzed production of (R)-HOA (Fig. 1B) (17). At longer times, the negative CD band decreased, consistent with earlier observations of decarboxylation of (R)-HOA to achiral 2-hydroxylevulinate and similar to the decarboxylation of acetolactate to acetoin or of tartronate semialdehyde to glycolaldehyde (17). Fitting this latter curve to a first-order equation yielded a halflife of 28.9 min for (R)-HOA with a first-order rate constant for decarboxylation of (4 Ϯ 0.05) ϫ 10 Ϫ4 s Ϫ1 , which is within the error range of the measured steady state linear velocities (Fig. 1C).

Steady State Kinetic Studies of (R)-HOA Synthase (HOAS) Activity
pH Dependence of Activity-The pH dependence of 2-ketoglutarate-dependent ferricyanide reductase activity (assay for formation of the postdecarboxylation intermediate; Scheme 1) and HOAS activity showed similar bell-shaped profiles in the range of 4.5-8.5 with maxima at ϳ6.5, suggesting that both reactions proceed through a common pH-dependent rate-limiting step ( Fig. 2A).
Steady State Kinetic Parameters-Substrate dependence of the HOAS reaction in the presence of a 10 mM concentration of the fixed substrate and a 0 -50 mM concentration of the varied substrate yielded hyperbolic Michaelis-Menten plots for 2-ketoglutarate, and saturation plots with glyoxylate showed linear substrate inhibition at high concentrations (Fig. 2, B and C). The kinetic parameters obtained after fitting the data to Equations 1 and 2 are summarized in Table 1. Because of substrate inhibition at high concentrations, Lineweaver-Burk plots obtained by varying the concentration of 2-ketoglutarate at fixed variable concentrations of glyoxylate were nonlinear (Fig.  3, A and B), and global fitting of data to deduce the kinetic mechanism and the order of substrate addition was not performed. Linear regions in Fig. 3, A and B, were used to obtain secondary plots (Fig. 3, C and D). In the secondary plot in Fig.  3C, the intercept drops, reaching saturation, whereas the slope increases linearly with glyoxylate. This is evocative of a ping-pong kinetic mechanism where glyoxylate displays competitive substrate inhibition with respect to 2-ketoglutarate as described by Cleland (18).

Allosteric Regulation
AcCoA was shown to be a potent activator of M. smegmatis ketoglutarate decarboxylase that binds at a shallow pocket on the surface of the enzyme some 40 Å away from the active site (9). Also, GarA was shown to inhibit HOAS and glutamate dehydrogenase allosterically and to activate glutamine synthetase, thus acting as a general regulator of glutamate metabolism in mycobacteria (19). We tested the effect of these allosteric regulators on steps leading to (i) addition of ThDP to 2-ketoglutarate and decarboxylation by the ferricyanide reductase assay and (ii) (R)-HOA synthesis using the CD assay.
Analyses of the steady state parameters from the ferricyanide reductase assay revealed that AcCoA increases the k cat by over 3-fold and decreases the K m 2KG by 6-fold, suggesting that it acts as a mixed V and K type activator to produce ϳ20-fold enhancement in k cat /K m 2KG . With respect to ThDP, both k cat and K m ThDP increased and k cat /K m ThDP decreased with increasing AcCoA. Within the range tested, k cat /K m ThDP Ͼ k cat /K m 2KG (Fig. 4,  A, B, and C).
In the presence of GarA, similar to earlier observations (20), only a 2.2-fold reduction of activity with K inact ϭ 2.44 Ϯ 0.43 M was observed in the ferricyanide reductase assay (Fig. 5A and see Table 3). Analyses of the steady state parameters revealed that GarA decreases the k cat by 2-fold and increases the K m 2KG by 2.4-fold, acting as a mixed V and K type inactivator to produce a 4.3-fold decrease in k cat /K m 2KG (Fig. 5, B and C). The activation/ inactivation constants (concentrations of the regulator producing a half-maximal effect on the observed kinetic parameters) are summarized in Table 2.
Next, in the (R)-HOA synthase assay, the K act was 9.92 Ϯ 0.44 M for activation by AcCoA with a 4.7-fold rate enhancement (Fig. 6A). In the presence of GarA, the K inact was 2.86 Ϯ 0.36 M with a 17-fold rate reduction (Fig. 6B). In addition, the inhibition by GarA was non-competitive with respect to glyoxylate (Fig. 6C). Although the K inact for GarA was similar from both assays, the different rate reductions suggested that GarA possibly affected the second half of the catalytic cycle. We hypothesized that this could be due to the differential role of the allosteric regulators at the individual catalytic steps. To test this hypothesis, we used recently developed CD and NMR methods (21).

Allosteric Regulation of Predecarboxylation Steps
Substrate Analog as a Mechanism-based Inhibitor-Phosphonate analogs of 2-ketoacids were shown earlier to be potent mechanism-based inhibitors of ThDP-dependent decarboxylases (22,23). They form a covalent ThDP adduct in a mechanism similar to the natural substrates and accumulate as non-decarboxylatable analogs of the predecarboxylation intermediate (Fig. 7E). The enzyme-bound form of this intermediate is also exclusively detected by near-UV (280 -400 nm) CD spectroscopy due to the induced chirality of active site-bound ThDP (V conformation) and stabilization of the 1Ј,4Ј-iminopyrimidine chromophore during the catalytic cycle (24).  We tested MSP using the ferricyanide reductase assay in the absence and presence of AcCoA. The mechanism of inhibition was competitive with respect to 2-ketoglutarate under the different conditions tested (Fig. 7, A and B, and Table 3 Table 2. Error bars represent S.E. in parameters. 169 Ϯ 22 M in the ferricyanide reductase assay (Table 3), suggesting that the activation observed in the presence of AcCoA is achieved by improving the affinity for the substrate or by stabilizing the incipient lactyl-type intermediate (Fig. 7D). In the presence of saturating GarA (50 M), a positive CD band at 302 nm saturating at a maximal amplitude similar to the GarA free enzyme was again observed, suggesting that the binding of GarA does not inhibit the formation of MSP-ThDP, the predecarboxylation adduct. However, the increased K d of 2300 Ϯ 235 M suggests that the affinity for the substrate is weaker compared with the unregulated enzyme (Fig. 7D), consistent with our observations with the ferricyanide reductase assay ( Table 3).

Detection of ThDP-bound Covalent Intermediates during Steady State of Various Reactions
To determine the effect of the presence of allosteric regulators on the first half and full catalytic cycle of the HOAS reaction, steady state accumulation of ThDP-bound covalent intermediates was estimated using HOAS reconstituted with [C2,C6Ј-13 C 2 ]ThDP. The assignment of NMR resonances to the covalent ThDP intermediates was based on the following observations. 1) The E. coli ketoglutarate dehydrogenase complex E1o component accepts succinate semialdehyde in the reverse direction, and the adduct C2-hydroxybutanoate-ThDP could be observed at 7.37 ppm (data not shown). The 2-ketoglutarate-ThDP adduct could not be detected in the forward direction, so the C6Ј-H chemical shift is not available for this intermediate. The assignment of C6Ј-H chemical shifts for glyoxylate-ThDP adducts was achieved by Kaplun et al. (25). Finally, we assigned the new resonance at 7.41 ppm to the HOA-ThDP adduct (product-ThDP adduct) based on 1) the product formed, 2) the expected similarity of the chemical mechanism to other ThDP carboligases, and 3) the comparison with the known trend of C6Ј-H chemical shift of various other C2-substituted ThDP intermediates.
M. tuberculosis HOAS-catalyzed conversion of 2-ketoglutarate to succinate semialdehyde is slower (0.02-0.05 s Ϫ1 ) than the rate of oxidation of the enamine in the K 3 Fe(CN) 6 assay (ϳ1.0 s Ϫ1 ). In the presence of 2-ketoglutarate, the 2-ketoglutarate-ThDP adduct was not detected, whereas C2-hydroxybutanoate-ThDP (HBThDP) occupied 35.2% of the active sites, and ThDP occupied 64.8% of the active sites ( Fig. 8A and Table 4). Taken together, these observations suggest that (i) the decarboxylation step is much faster than the addition of 2-ketoglutarate to ThDP, and (ii) succinate semialdehyde release is the rate-limiting step for the first half of the catalytic cycle (Fig.   8, reaction scheme). In the presence of AcCoA, again the 2-ketoglutarate-ThDP adduct was not detected, and HBThDP and ThDP were detected at 39.3 and 60.6%, respectively (Fig. 8B). Similar results (HBThDP, 34.5%; ThDP, 65.5%) were observed in the presence of saturating GarA, suggesting that GarA does not significantly affect the decarboxylation step and the steady state concentration of HBThDP (Fig. 8C).

DISCUSSION
There is complexity to the mechanism of this enzyme as it could result in one of three outcomes: 1) formation of succinate semialdehyde via 2-ketoglutarate decarboxylation, a non-oxidative decarboxylase pathway; 2) formation of HOA, a carboligase activity; and 3) formation of a putative precursor to reductive succinylation of a putative E2 component in a manner expected in most prokaryotic and eukaryotic cells for 2-oxoacid dehydrogenase complexes. In view of results so far accumulated (6,7,26), it appears that pathways 1 and 2 are predominant. All three pathways are derived from ThDP-catalyzed decarboxylation of 2-ketoglutarate and the reaction of the enamine intermediate with either a H ϩ , glyoxylate, or lipoamide bound to a putative E2, respectively.
Slow aldehyde release (pathway 1) is a familiar outcome in thiamin-dependent 2-oxoacid dehydrogenases and carboligases in the absence of acceptor substrates. Our earlier work had shown that the gene product of Rv1248c does not catalyze pathway 3. Based on work with M. smegmatis ketoglutarate decarboxylase, it was suggested that our use of recombinant enzymes based on a misannotated sequence and the absence of AcCoA could explain our observations (9). To test these ideas, we cloned tagless HOAS based on immunoprecipitation from M. tuberculosis lysates and N-terminal sequencing. We were

TABLE 2 Effect of regulators on 2-ketoglutarate decarboxylation
Activation/inactivation constants of regulators for the steady state kinetic parameters k cat , K m , and k cat /K m were obtained from the K 3 Fe(CN) 6 assay (Figs. 4 and 5). (k cat /K m ) is the value in the presence of maximal AcCoA or GarA, and (k cat /K m ) 0 is in the absence of the regulators.

Regulator
Varied substrate showed lower turnover numbers with M. smegmatis ketoglutarate decarboxylase in earlier work by Wagner et al. (9) (k cat ϳ 0.1 s Ϫ1 for the AcCoA-activated enzyme) compared with our assays. As this study deals with the influence of allosteric regulators on HOAS catalysis, we chose pathway 2 for a detailed steady state kinetic characterization. However, the detection of intermediates and the role of allosteric regulators as elucidated here would bear significance for any of the pathways considered.
No chemical models and 1 H NMR chemical shifts for the ThDP covalent intermediates in the catalytic cycle of 2-ketoglutarate-dependent enzymes are available to date. Hence, we used [C2,C6Ј-13 C 2 ]ThDP-reconstituted HOAS for unambiguous assignment of ThDP-derived covalent intermediates using one-dimensional 1 H gradient carbon heteronuclear single quantum coherence NMR, a method that selectively detects 1 H attached to 13 C atoms. In the various experiments, we did not detect resonances related to buffer, substrate, or other low level contaminants; hence, new resonances can be assigned with confidence to ThDP covalent intermediates. However, a full NMR characterization of the HOA-ThDP adduct was not possible given the unstable nature of HOA (t1 ⁄ 2 ϳ 29 min), the anticipated starting material for the chemical synthesis of HOA-ThDP. We did not undertake NMR investigations of the ferricyanide reductase pathway because the use of K 3 Fe(CN) 6 is incompatible with our NMR methods.
The (R)-HOA detected by our CD assay was assumed to be enantiopure for calculation of linear velocities. We were unable to trap and purify the unstable (R)-HOA to determine the true enantiomeric excess by chiral gas chromatography-coupled mass spectrometry. Thus, the k cat values reported here are the lower limits. The k cat values of the unactivated (4 s Ϫ1 ) and the AcCoA-activated enzyme (21 s Ϫ1 ) are comparable with those of other carboligases that use 2-ketoacid substrates, such as glyoxylate carboligase and isozymes of acetohydroxyacid synthase from various bacteria (17,25,(27)(28)(29)(30). The rates on the unactivated enzyme are similar to those of M. tuberculosis acetohydroxyacid synthase, whereas the activated rates are comparable with those of E. coli enzymes. Variants of 2-ketoacid decarboxylases can be engineered to enhance rates of production of carboligase products. Examples include the E636A E1p variant of E. coli pyruvate dehydrogenase (k cat ϭ 1 s Ϫ1 ) and the D28A variant of Saccharomyces cerevisiae pyruvate decarboxylase (k cat ϭ 0.05 s Ϫ1 ), which behave like acetolactate synthase (31), and the E473Q variant of Zymomonas mobilis pyruvate decarboxylase, which catalyzes the enantioselective conversion of pyruvate and benzaldehyde to (R)-phenylacetylcarbinol at 2.5 s Ϫ1 (32). This comparison suggests that HOAS is indeed catalytically competent toward C-C bond-forming reactions and even surpasses the engineered variants when AcCoA activation is taken into account. Although (R)-HOA production has been identified in other species, kinetic characterization of the other enzymes (6,26) will be necessary to determine whether the M. tuberculosis enzyme exemplifies a general feature of the E1o components or whether it has diverged to become a particularly competent HOA synthase in connection with a variant tricarboxylic acid cycle in M. tuberculosis.    The structure of (R)-HOA as determined by NMR and LC-MS (5, 7) and structures of various carboligase products of E. coli E1o, HOAS, and 2-succinyl-5-enolpyruvyl-6-hydroxy-3cyclohexene-1-carboxylate synthase (MenD) provide strong footing for the suggested chemical mechanism of HOAS (Scheme 1) (33), similar to other ThDP-dependent carboligases. In addition, 2-ketoglutarate-ThDP adducts could be detected, whereas glyoxylate-ThDP adducts could not be detected by NMR under the various conditions tested. Although a ping-pong kinetic mechanism can be anticipated based on the substrate addition steps being separated by an irreversible decarboxylation step, the Lineweaver-Burk plots suggest a more complex kinetic mechanism. Analysis of secondary plots for 2-ketoglutarate suggests a competitive substrate inhibition by glyoxylate as described by Cleland (18). The hysteretic lag phase in the progress curves that is indicative of the switching of the dimeric enzyme from a lower activity state to a higher activity state in the presence of 2-ketoglutarate as seen in the x-ray structures of M. smegmatis ketoglutarate decarboxylase (9) could contribute to the complexity of the Lineweaver-Burk plots. Pre-steady state and steady state kinetic experiments with acetohydroxyacid synthase I mutants and deoxy-D-xylylose-5-phosphate synthase indicate that these enzymes utilize a ThDP-substrate-acceptor ternary complex where the substrate is covalently bound to ThDP, and the presence of the acceptor enhances decarboxylation rates (34 -36). In sum, these examples show that ThDP enzymes can perform similar chemical steps through different kinetic mechanisms.
M. tuberculosis HOAS, its homolog M. smegmatis ketoglutarate decarboxylase, and the SucA components from other actinomycetes share sequence similarity and domain organization. OdhA from Corynebacterium glutamicum is similarly regulated by OdhI (GarA homolog), but its allosteric activation by AcCoA is not yet established (37,38). ThDP-dependent decarboxylases are subject to homotropic allosteric regulation by their 2-ketoacid substrates, and in M. tuberculosis, a ketoacid decarboxylase involved in amino acid metabolism via the Ehrlich pathway is regulated by amino acids (39). Mammalian ␣ 2 ␤ 2 heterotetrameric E1 components of pyruvate dehydrogenase are regulated by kinases that phosphorylate a residue in the active site loop, leading to inhibition of the postdecarboxylation steps. However, this does not interfere with the predecarboxylation steps (40).
In contrast, the mycobacterial enzymes are unique among the ThDP superfamily of enzymes as they are subject to two distinct modes of allosteric regulation by different regulators.
The inhibitory activity of GarA is regulated by phosphorylation; binding of unphosphorylated GarA regulates HOAS. The phosphorylation state of GarA is dependent on the presence of PknB and PknG, whose expression is regulated by external stresses (19,41,42).
AcCoA increases the k cat in predecarboxylation and (R)-HOA synthase steps accompanied by a reduction in K m for 2-ketoglutarate and a 22-fold increase in k cat /K m for the predecarboxylation steps. In contrast, GarA produces a 2.2-fold reduction in k cat , a 2.4-fold reduction in K m , and a 4-fold reduction in k cat /K m in the predecarboxylation steps. However, a more substantial 17-fold reduction in k cat is observed in the (R)-HOA synthase assay, suggesting that the major ϳ8-fold impact on k cat reduction by GarA occurred on postdecarboxylation steps.
The development of modern techniques to interrogate ThDP catalysis during individual catalytic steps provided a unique opportunity to test and understand the role of various allosteric regulatory mechanisms at the individual chemical steps. First, from CD studies using substrate analogs, it is evident that AcCoA activation is pronounced in the first half of the catalytic cycle, whereas strikingly GarA does not inhibit formation of the predecarboxylation intermediate, although the substrate affinity is decreased. Second, from NMR experiments, the accumulation of the postdecarboxylation intermediate is evident in the presence of GarA, suggesting that GarA does not inhibit decarboxylation either. There is a measurable effect only on the covalent intermediate levels in the second half of the catalytic cycle (addition of glyoxylate to the enamine) in direct correlation with loss of activity in steady state kinetics.
Taken together, these results suggest that although the two regulators influence K m for 2-ketoglutarate they exert their maximal influence on different covalent bond formation steps in the catalytic cycle. AcCoA activates the predecarboxylation steps, and GarA inhibits addition of the second substrate to the central enamine intermediate. Non-competitive inhibition by GarA with respect to glyoxylate suggests that binding of GarA at a distal region elicits changes that interfere with glyoxylate binding.
M. tuberculosis HOAS expression is predicted to be essential for the optimal growth of M. tuberculosis (43). Thus, M. tuberculosis HOAS could be a potential target for antituberculosis drugs. Enzymes of the central carbon metabolism are difficult targets for drug development because of the evolutionarily conserved nature of the function and of the structures of the enzymes that perform them. However, even among enzymes a 2 eq of ͓C2,C6Ј-13 C 2 ͔ThDP were used to ensure saturation and serve as internal standard. Percent occupancies of active sites were determined from relative integrals in Fig.  8 after correction for excess unbound ThDP and estimation of unbound enzyme using Equations 9 and 10. b Predecarboxylation adduct of 2-ketoglutarate with ThDP at the C2 position (2KG-ThDP) was not detected (ND). that perform conserved chemical transformations by the same chemical mechanisms, the electrostatic stabilization of the transition state could differ. In addition, the presence of allosteric activators could change the nature of the transition state as in AMP nucleosidase (44,45). Based on the dependence of K i of MSP and K d of MSP-ThDP adduct on the presence of AcCoA, we suggest that accounting for allosteric regulators could aid in developing mechanism-based inhibitors with improved potency and enhanced specificity against homologs that use different mechanisms of allostery. In addition, our findings might be of interest for studies of the regulation of other ThDP-dependent enzymes involved in the metabolism of glucose and glutamate.