Human 2-Oxoglutarate Dehydrogenase Complex E1 Component Forms a Thiamin-derived Radical by Aerobic Oxidation of the Enamine Intermediate*

Background: The human 2-oxoglutarate dehydrogenase complex, comprising E1o, E2o, and E3 components, catalyzes conversion of 2-oxoglutarate to succinyl-CoA. Results: Human E1o generates both a thiamin-enamine-derived radical and the reactive oxygen species, superoxide, and hydrogen peroxide. Conclusion: Human E1o produces reactive oxygen species at a rate of <1% of succinyl-CoA under physiological conditions. Significance: This work presents the novel discovery that the 5-carboxyl group affects enzymatic reactivity of 2-oxoglutarate. Herein are reported unique properties of the human 2-oxoglutarate dehydrogenase multienzyme complex (OGDHc), a rate-limiting enzyme in the Krebs (citric acid) cycle. (a) Functionally competent 2-oxoglutarate dehydrogenase (E1o-h) and dihydrolipoyl succinyltransferase components have been expressed according to kinetic and spectroscopic evidence. (b) A stable free radical, consistent with the C2-(C2α-hydroxy)-γ-carboxypropylidene thiamin diphosphate (ThDP) cation radical was detected by electron spin resonance upon reaction of the E1o-h with 2-oxoglutarate (OG) by itself or when assembled from individual components into OGDHc. (c) An unusual stability of the E1o-h-bound C2-(2α-hydroxy)-γ-carboxypropylidene thiamin diphosphate (the “ThDP-enamine”/C2α-carbanion, the first postdecarboxylation intermediate) was observed, probably stabilized by the 5-carboxyl group of OG, not reported before. (d) The reaction of OG with the E1o-h gave rise to superoxide anion and hydrogen peroxide (reactive oxygen species (ROS)). (e) The relatively stable enzyme-bound enamine is the likely substrate for oxidation by O2, leading to the superoxide anion radical (in d) and the radical (in b). (f) The specific activity assessed for ROS formation compared with the NADH (overall complex) activity, as well as the fraction of radical intermediate occupying active centers of E1o-h are consistent with each other and indicate that radical/ROS formation is an “off-pathway” side reaction comprising less than 1% of the “on-pathway” reactivity. However, the nearly ubiquitous presence of OGDHc in human tissues, including the brain, makes these findings of considerable importance in human metabolism and perhaps disease.

Functionally competent E1o-h and E2o-h components were created for the first time. There are no structures for these components or, for that matter, for any intact E2 component from any source. Structures are available for the E3 component (7)(8)(9)(10)(11)(12) and for N-terminally truncated (some 77 amino acids) Escherichia coli 2-oxoglutarate dehydrogenase (E1o-ec) (13). The availability of the human E1o and E2o components enabled us to characterize longer-lived ThDP-bound intermediates on E1o-h.
Although typical ThDP-assisted decarboxylation mechanisms are ionic in nature, there are important known exceptions; (a) pyruvate-ferredoxin oxidoreductase (14,15) utilizes three tandem Fe 4 S 4 clusters to oxidize the enamine intermediate to acetyl-CoA, and (b) pyruvate oxidase produces acetate in E. coli and acetyl phosphate in Lactobacillus plantarum, and both pyruvate-ferredoxin oxidoreductase and pyruvate oxidase (16) have been shown to support a postdecarboxylation-related radical that has spin density not only at the C2␣ atom but also around the thiazolium ring (henceforth referred to as the ThDP-enamine radical) (14 -17). The observation of a similar radical on E1o-ec (18) raised the possibility that such a radical also exists on the E1o-h. There have been several reports that mammalian OGDHc could be a major source of the reactive oxygen species (ROS), superoxide radical anion and hydrogen peroxide, in isolated brain and skeletal muscle mitochondria (5,19) and in the isolated synaptosomes (6), as was also confirmed with the isolated mammalian OGDHc (5,6). Based on experiments with spin-trapping techniques and using the isolated mammalian OGDHc with a modified flavin cofactor of E3, it was concluded that E3-bound FAD is responsible for the production of radical species by OGDHc (20). The above conclusions were supported further by experiments with brain mitochondria isolated from E3-deficient mice that produce less H 2 O 2 as compared with wild-type mice (5). Based on these studies, ROS generation by OGDHc was attributed to the E3 component (6,(21)(22)(23)(24). It has been documented that OGDHc could produce H 2 O 2 in the forward physiological reaction by oxidation of OG in the presence CoA and in the reverse reaction at high NADH/NAD ϩ ratio (no OG is needed). With the E3 component being shared by OGDHc, PDHc, and branched-chain 2-oxo acid dehydrogenase complex, OGDHc was reported to produce H 2 O 2 at about twice the rate produced by PDHc and 4 times the rate produced by branched-chain 2-oxo acid dehydrogenase complex in isolated skeletal muscle mitochondria, and the rates were dependent on concentration of their specific substrates rather than on NADH/NAD ratio (19).
The relatively stable enzyme-bound enamine is the likely substrate for oxidation by O 2 , leading to the generation of superoxide anion and ThDP-derived radical. The results indicate that E1o-h does produce both the ThDPenamine radical and superoxide/H 2 O 2 . Generation of superoxide/H 2 O 2 by E1o-h may have pathological relevance, especially under conditions of genetic deficiencies of the E2o-h due to mutations (25), polymorphism (26), or down-regulation (27,28). In cultured human embryonic kidney cells, the reduction in E2o-h level augmented ROS generation and cell death (27). E2o-h deficiency also elevated the amyloid plaque burden, the levels of A␤ oligomers and nitrotyrosine, and the occurrence of spatial learning and memory deficits in mice (29), whereas decreased OGDHc activity in general and resultant prolonged oxidative stress are known to be part of the pathogenesis in Alzheimer disease (30 -32).
Although the E3 component is common to all 2-oxo acid dehydrogenase complexes in a particular cell type, the chemistry on E1o is apparently significantly different from that on the PDHc. The findings with both human and bacterial E1os place on firm ground the dramatic differences between the pyruvate and OG-derived intermediates on ThDP, and all point to and are consistent with a role of the side chain carboxyl group (carbon 5) in the series of reactions taking place on ThDP.
Protein Expression and Purification-Expression and purification of the E. coli E1o-ec, E2o-ec, and E3-ec was as reported earlier (33).
Construction of Plasmid and Expression and Purification of E1o-h and E2o-h-The genes encoding His 6 -tagged E1o-h and His 6 -tagged E2o-h were synthesized by DNA2.0, Inc. (Menlo Park, CA). The E1o-h gene optimized for expression in E. coli cells and the E2o-h gene were inserted into pET-22b(ϩ) and pET-15b vectors, respectively, and the resulting plasmids were expressed in BL21 (DE3) cells. Cells were grown in LB medium supplemented with 50 g/ml ampicillin containing 1 mM MgCl 2 and 0.5 mM thiamin for E1o-h expression or 0.3 mM DL-␣-lipoic acid for E2o-h expression. Protein expression was induced by 0.5 mM isopropyl 1-thio-␤-D-galactopyranoside for 5 h at 37°C for E2o-h and for 15 h at 25°C for E1o-h. Cells were washed with 0.20 M KH 2 PO 4 (pH 7.0) containing 0.15 M NaCl and stored at Ϫ20°C. The harvested cells were dissolved in 50 -70 ml of 50 mM KH 2 PO 4 (pH 7.5) containing 0.3 M KCl, 2.0 mM ThDP, 5.0 mM MgCl 2 , and one protease inhibitor mixture tablet (Roche Applied Science). Cells were treated with lysozyme (0.60 mg/ml) at 4°C for 20 min and were disrupted using a sonic dismembrator. The clarified lysate was loaded onto a nickel-Sepharose column (GE Healthcare) equilibrated with 50 mM KH 2 PO 4 containing 0.30 M KCl, 0.50 mM ThDP, 1.0 mM MgCl 2 , and 30 mM imidazole (pH 7.5). The E1o-h was eluted with 300 mM imidazole in the same buffer. The E1o-h was dialyzed against 50 mM KH 2 PO 4 (pH 7.5) containing 0.30 M NaCl, 0.50 mM ThDP, 2.0 mM MgCl 2 , and 1.0 mM benzamidine⅐HCl at 4°C for 15 h. The protein was concentrated, and buffer was exchanged to 50 mM HEPES (pH 7.5) containing 0.15 M NaCl, 0.50 mM ThDP, 1.0 mM MgCl 2 , and 1 mM benzamidine⅐HCl and was stored at Ϫ20°C. The His 6tagged E2o-h was purified in a manner similar to that reported for E2p-ec (34).
Expression and Purification of E3-h-For expression and purification of E3h, E. coli M15 cells were used as well as pQE-9 vector with the gene encoding E3-h as reported earlier (35) with some modifications. Cells were grown in LB medium supplemented with 100 g/ml ampicillin, and E3-h expression was induced by 0.5 mM isopropyl 1-thio-␤-D-galactopyranoside at 25°C for 15 h. E3-h was purified using a nickel-Sepharose column (GE Healthcare 52 M subunits) with a mass ratio (g/g/g) of 1:5:5 in 0.25 ml of 50 mM KH 2 PO 4 (pH 7.5) containing 0.50 mM ThDP, 1.0 mM MgCl 2 , 0.15 M NaCl, and 1.0 mM benzamidine⅐HCl at 25°C. A 0.010-ml aliquot of the reaction mixture was withdrawn after a 1-h incubation to start the reaction. The stoichiometry of components used for activity measurement was chosen from kinetic experiments where, at constant concentrations of E1o-h subunits (0.046 M) and E3-h subunits (0.52 M), the concentration of E2o-h subunits was varied (0.059 -0.80 M), leading to S 0.5,E2o-h ϭ 0.18 Ϯ 0.08 M. In a similar experiment with varied concentrations of E3-h subunits (0.026 -0.53 M) and constant concentrations of E1o-h (0.046 M) and E2o-h (0.24 M) subunits, an S 0.5,E3-h of 0.17 Ϯ 0.02 M was calculated. The reaction medium contained the following in 1.0 ml: 0.10 M Tris-HCl (pH 7.5), 0.50 mM ThDP, 2.0 mM MgCl 2 , 2.0 mM OG, 2.0 mM DTT, 5 mM NAD ϩ , 0.1-0.20 mM CoA at 37°C. The reaction was initiated by the addition of CoA and OGDHc. Steady-state velocities were taken from the linear portion of the progress curve. One unit of activity is defined as the amount of NADH produced (mol⅐min Ϫ1 ⅐mg E1o Ϫ1 ).
E1o-h Specific Activity-The E1o-h specific activity was measured in the reaction medium with the external oxidizing agent DCPIP (36) at 600 nm in the following reaction medium contained in 1 ml: 50 mM KH 2 PO 4 (pH 7.0), 0.50 mM ThDP, 1.0 mM MgCl 2 , 2 mM OG, and DCPIP (0.08 mM) at 37°C. The reaction was initiated by the addition of 0.01-0.015 mg of E1o-h. Steady-state velocities were taken from the linear portion of the progress curve recorded at 600 nm. One unit of REACTION 1 Thiamin Intermediates on Human 2-Oxoglutarate Dehydrogenase OCTOBER 24, 2014 • VOLUME 289 • NUMBER 43 activity is defined as the amount of reduced DCPIP produced (mol⅐min Ϫ1 ⅐mg E1o-h Ϫ1 ).
Superoxide Production by E1o-h-Superoxide production by E1o-h was measured via reduction of acetylated cytochrome c (cyt c), as described earlier, with minor modifications, at pH 6.3, where the sensitivity of the method was optimal (22,(37)(38)(39)(40). The initial reaction velocity was measured and converted to generated superoxide amount (nmol)⅐min Ϫ1 ⅐mg E1o-h Ϫ1 , taking into account that the stoichiometry of the reaction between cyt c and superoxide is 1:1 (37). Assay conditions were the following in a 200-l final volume in a microtiter plate-based experiment: 50 mM KH 2 PO 4 (pH 6.3), 0.2 mM ThDP, 1 mM Mg 2ϩ , 0.1 mM CoA, 2 mM OG, 50 M cyt c (all in final concentrations), and 64.44 g of E1o-h. The concentration of cyt c was chosen to determine a linear dynamic range for the measurements (22,37). Protein concentration was determined by the Bradford method (41). The reaction was initiated by the addition of OG (10 l, 5%) after a 15-min incubation at 37°C and after the baseline was stabilized. Measurements were carried out at a max of 550 nm of the reduced form of acetylated cyt c in enzyme immunoassay/radioimmune assay 96-well, flat-bottom, polystyrene plates (Corning Inc.) using a Spectramax M2 (Molecular Devices, Sunnyvale, CA) or a Victor 3 (PerkinElmer Life Sciences) multilabel counter spectrophotometer/fluorimeter. An extinction coefficient of 10,170 M Ϫ1 cm Ϫ1 ( max of 550 nm) was determined for the fully reduced cyt c by sodium dithionite under the assay conditions. The superoxide dismutase from bovine erythrocytes was used to verify that the cyt c reduction was indeed caused by superoxide generated by the E1o-h; 100 units of superoxide dismutase was added to 200 l of reaction mixture, where applied.
H 2 O 2 Production by E1o-h-To detect H 2 O 2 generated by E1o-h, the Amplex Red fluorescent dye (Molecular Probes) was used as reported earlier (6,21,42). In the presence of horseradish peroxidase (HRP), the Amplex Red reacts with H 2 O 2 with a 1:1 stoichiometry, producing highly fluorescent resorufin. Fluorescence was detected in a Photon Technology International (Lawrenceville, NJ) Deltascan fluorescence spectrophotometer; the excitation wavelength was 550 nm, and the fluorescence emission was detected at 585 nm. A calibration curve was generated with known amounts of H 2 O 2 at the end of each experiment. The reaction was initiated by the addition of the E1o-h or OG after a 15-min incubation at 37°C of Amplex Red and HRP and after the stabilization of the fluorescence signal, generally 100 -200 s after the addition of all required components. For calculation of the reaction rate, slopes of the curves detected in the absence of E1o-h were subtracted from the initial slopes measured in the presence of both E1o-h and OG. The reaction assay in 2 ml of 50 mM KH 2 PO 4 (pH 6.3 or 7.3) contained the following components at the final concentrations: 1 unit of HRP, 1 M Amplex Red, 0.1 mM EGTA, 0.2 mM ThDP, 0.4 mM ADP, 1 mM Mg 2ϩ , 0.12 mM CoA, 1 mM OG, and 64.4 g of E1o-h, at 37°C under continuous stirring. Statistical differences were evaluated using the Excel program with two-tailed Student's t tests assuming unequal variances and were accepted to be significant when p Ͻ 0.05.
Stopped-flow CD Spectroscopy-Kinetic traces were recorded on a Pi*-180 stopped-flow CD spectrometer (Applied Photo-physics) using a 10-mm path length. The intermediates seen on steady-state CD at ϳ298 and ϳ352 nm were detected on stopped-flow CD at 297 and 365 nm, respectively, as dictated by the high sensitivity of the lamp at those wavelengths. Data from six or seven repetitive shots were averaged and fit to the appropriate equations (see the figure legends). For pre-steady-state rate determination of ThDP-bound intermediates, E1o-ec (38 M concentration of active centers) in one syringe was mixed in the reaction cell with OG (4 mM) from the second syringe, and the reaction was monitored at 297 or 365 nm for 50 s at 20°C. For CD experiments at 365 nm in the absence and presence of enamine acceptors, E1o-ec (38 M concentration of active centers) from one syringe was mixed with OG (200 M) placed in the second syringe, and the reaction was monitored for 50 s at 20°C. The alternative enamine acceptors, E2o-ec(1-176) didomain (60 M), DCPIP (50 M), and glyoxylate (GA, 2 mM) were placed in the second syringe together with OG.
Rapid Chemical Quench and NMR Methods-The rapid reaction mixing experiments were carried out using a rapid chemical quench apparatus (Kintek RQF-3 model, Kintek Corp.), and NMR spectra were acquired on a Varian INOVA 600-MHz instrument using gradient carbon homonuclear single quantum coherence (gCHSQC; to enable observation of only those protons attached to 13 C nuclei (43)). The water signal was suppressed by presaturation. For reconstitution with [C 2 ,C 6 Ј-13 C 2 ]ThDP, the E1o-ec (0.12 mM, free of ThDP) was incubated with 1 eq of [C 2 ,C 6 Ј-13 C 2 ]ThDP (0.12 mM) and 5 mM MgCl 2 for 30 min on ice. [C 2 ,C 6 Ј-13 C 2 ]ThDP-labeled E1o-ec was loaded into syringe A, and OG was loaded into syringe B in a rapid chemical quench apparatus. A quench solution (12.5% TCA in 1 M DCl/D 2 O) was loaded into syringe C. Enzyme and OG solutions were mixed rapidly in a 1:1 volume ratio by a computer-controlled drive-plate and incubated for predetermined times in a reaction loop. After this incubation period, the reaction mixture was mixed rapidly with quench solution from syringe C, and the resulting mixture was collected. For longer time scales (Ͼ20 s), the samples were assembled similarly into three different solutions. Next, equal volumes of [C 2 ,C 6 Ј-13 C 2 ]ThDP-labeled E1o-ec (200 l) and OG (200 l) were mixed in an Eppendorf tube, and after predetermined times, the reaction was stopped by the addition of quench solution (200 l). The contents of the syringes in 20 mM KH 2 PO 4 (pH 7.4) were varied to monitor the following different reactions. (a) For E1o-ec with OG single turnover, syringe A contained [C 2 ,C 6 Ј- 13  The quenched reaction mixture was collected and centrifuged at 15,700 ϫ g for 20 min, the supernatant was separated and filtered through Gelman 0.45 M discs, and then the filtrate was analyzed by gCHSQC NMR at 25°C. Using [C 2 ,C 6 Ј-13 C 2 ]ThDP in conjunction with the gCHSQC method enables observation of only those protons bound to 13 C atoms, very useful for intermediate detection in the aromatic spectral region. The interpretation of NMR spectra was carried out as with pyruvate earlier (43).
Electron Paramagnetic Resonance Spectroscopy-For sample preparation, the following protocol was used: (a) E1o-h (50 mg/ml concentration of active centers ϭ 0.449 mM) in 0.4 ml of 50 mM HEPES (pH 7.5) containing 0.15 M NaCl, 0.50 mM ThDP, and 1.0 mM MgCl 2 was mixed with 10 mM OG at room temperature. The mixture was immediately transferred into an EPR tube and was flash-frozen in liquid nitrogen within 40 s of mixing all components. (b) E1o-ec (0.150 mM concentration of subunits) in 0.20 ml of 50 mM Tris-HCl (pH 7.4) containing 2 mM ThDP, 2 mM MgCl 2 , and 10% glycerol was mixed with 10 mM OG or 2-oxoadipate at room temperature and was transferred into an EPR tube. (c) For OGDHc-h assembly, E1o-h, E2o-h, and E3-h were mixed in 50 mM HEPES (pH 7.5) containing 0.50 mM ThDP, 1.0 mM MgCl 2 , and 0.15 M NaCl with a 0.222 mM concentration of subunits for each protein (ratio of E1o-h/E2oh/E3-h subunits ϭ 1:1:1). The reaction was started by the addition of 10 mM OG. After ϳ30 s, the samples were flash-frozen in liquid nitrogen, and the sample was inserted into the EPR resonator. The following conditions were used for EPR experiments. X-band (9 GHz) EPR measurements were made using a Varian E-112 spectrometer interfaced to a PC using custom written software. A finger Dewar filled with liquid nitrogen and inserted into a TE 102 rectangular EPR cavity maintained the sample at 77 K throughout the measurements. Spectrometer parameters used to acquire the spectrum in Fig. 4 are as follows: modulation amplitude, 5 G; microwave power, 0.05 milliwatt; receiver gain, 3.2 ϫ 10 4 ; microwave frequency, 9.119 GHz, scan time, 0.5 min; time constant, 0.128 s; number of scans, 49. The field was calibrated using a standard sample of manganese doped in MgO (44). The concentration of the radical species was estimated by comparing the double integrated intensity of the radical signal with that of a 400-mol Cu(II)-EDTA standard sample. Spectral simulations were carried out using previously described software (45) in which the Zeeman interaction is treated exactly and nuclear hyperfine coupling interactions are treated to first order. Parameters used for the simulation displayed under "Results and Discussion" are as follows: principal g-values g 1 ϭ 2.008, g 2 ϭ 2.005, g 3 ϭ 2.002; 14 N (nuclear spin I ϭ 1) hyperfine coupling principal values (in Gauss) ϭ 1, 1, 13; unique principal value of 14 N hyperfine coupling interaction oriented along g 3 ; four isotropic 1 H hyperfine coupling interactions (in gauss) ϭ 5, 4, 3, and 3; Gaussian broadening function ϭ 2.5 G half-width at half-height.

Functional Competence of the E1o-h
Our interest in the catalytic mechanism of ThDP-dependent enzymes and the involvement of OGDHc in human pathology known in the literature prompted us to synthesize genes expressing E1o-h and E2o-h. Because data on OGDHc reported in the literature are mostly limited by mammalian sources, this is the first report on OGDHc-h assembled from individual recombinant components that produce active OGDHc-h. The identity of E1o-h and E2o-h was confirmed by Fourier transform MS analysis of peptic peptides resulting from pepsin treatment of E1o-h and E2o-h components (see supplemental Tables S1 and S2). The functional competence of E1o-h and E2o-h was confirmed by the following experiments.
Kinetic Studies-An E1o-h specific activity of 1.78 Ϯ 0.40 mol⅐min Ϫ1 ⅐mg E1o Ϫ1 (k cat ϭ 6.7 s Ϫ1 ) was measured by monitoring the reduction of the external oxidizing agent 2,6-DCPIP at 600 nm (36) in the assay containing E1o-h and OG as compared with 0.34 mol⅐min Ϫ1 ⅐mg E1o Ϫ1 reported by us earlier for E1o-ec (33) ( Table 1) (for the composition of assay medium see "Experimental Procedures"). In this reaction, the enamine intermediate produced on E1o-h from OG is oxidized by DCPIP with the formation of succinate and reduced DCPIP (36). Although there is a recent report of an E1o-h specific activity of 0.30 mol⅐min Ϫ1 ⅐mg protein Ϫ1 , measured as the rate of carboligase reactivity with glyoxylate as an acceptor instead DCPIP, we cannot directly compare the results of the two assays (46).
An overall OGDHc activity of 7.86 Ϯ 0.74 mol⅐min Ϫ1 ⅐mg E1o-h Ϫ1 (k cat ϭ 30 s Ϫ1 ) was measured for E1o-h reconstituted with E2o-h and E3h components once the stoichiometry of the components was optimized (for the composition of assay medium and the stoichiometry of the components see "Experimental Procedures" and the legend to Fig. 1). To our knowledge, this is the first report on the successful expression of recombinant human E1o and E2o components that could be assembled into active OGDHc-h. As a comparison, an E1o activity of 20 mol⅐min Ϫ1 ⅐mg E1o Ϫ1 within the isolated pig heart OGDHc was calculated (46), in good accord with our data above (7.86 Ϯ 0.74 mol⅐min Ϫ1 ⅐mg E1o-h Ϫ1 ) and confirming the functional competence of the recombinant E1o-h and E2o-h components.
We estimate a S 0.5 , OG of 0.15 mM for E1o-h in the overall OGDHc assay (Fig. 1, top), similar to that reported for mammalian oxoglutarate dehydrogenase complexes (47)(48)(49)(50)(51)(52). Also, values of S 0.5,OG of 30 M and S 0.5 , OG of 120 M were reported for OG binding in two active centers of human heart OGDHc with an overall OGDHc activity of ϳ6 mol⅐min Ϫ1 ⅐mg protein Ϫ1 (53).
The pH dependence of the overall OGDHc-h activity displayed a pH-independent plateau near physiological pH (6.5-7.5) but reduced activity with increasing pH, with an apparent pK a of 8.48 (Fig. 1, bottom), which correlates well with the pH optimum of 6.65 over a relatively broad range (ϳ2 pH units) (51) and a pH optimum of 6.5-7.0 reported for mammalian OGDHcs (54). Our subsequent experiments were carried out in this pH-independent plateau region. Circular Dichroism Studies with OG-CD titration of E1o-h by OG in the absence of E2o-h and E3 revealed the appearance of two CD bands (Fig. 2, top) similar to those observed with E1o-ec (Fig. 2, bottom), leading to the following conclusions: (a) a positive CD band at 302 nm, earlier assigned to the 1Ј,4Јiminopyrimidine tautomer (IP form) of either the first predecarboxylation or the second postdecarboxylation ThDP-bound intermediate (C2-(␣-hydroxy)-␥-carboxypropyl-ThDP in Scheme 1) because it had been concluded that all ThDP intermediates with C2␣ tetrahedral substitution exist in their IP form at a pH near and above the pK a of the 4Ј-aminopyrimidinium conjugate acid form (55) and (b) a positive CD band at 347 nm (Fig. 2) that could tentatively be assigned to the enamine intermediate (the first postdecarboxylation intermediate in Scheme 1). Both E1o-h CD bands displayed dependence on the concentration of OG, resulting in an S 0.5 , OG of 0.14 mM (302 nm) and S 0.5,OG of 0.09 mM (347 nm). Unlike on E1o-ec, according to the behavior of this CD band, this inter-mediate is not as stabilized on E1o-h as on E1o-ec, making its detection more difficult.
Circular Dichroism Studies on E1o-h with Phosphonate and Phosphinate Analogues-Formation of the IP tautomer of ThDP-bound predecarboxylation intermediate in the active centers of E1o-h was evident from CD spectra of E1o-h titrated by phosphonate analogues of OG succinylphosphonate methyl ester (CD maximum at 296 nm, S 0.5,succinylphosphonate methyl ester ϭ 48 Ϯ 19 M) and succinylphosphonate (CD maximum at 304 nm) and also from acetylphosphinate, an analogue of pyruvate (CD maximum at 300 nm, S 0.5,acetylphosphinate ϭ 28 Ϯ 2.6 M), known to be a potent inhibitor of ThDP-dependent decarboxylases (33, 47, 56 -63). In addition, the CD spectra of all tested phosphonate analogues displayed a the negative CD band around 330 nm, earlier assigned at Rutgers to the Michaelis complex (64), suggesting half-of-the site reactivity with one active center filled with ThDP-bound predecarboxylation intermediate and the second active center filled with Michaelis complex. Based on the data presented, we can conclude that the tested phosphonate and phosphinate analogues bind in the active centers of E1o-h according to the CD bands formed at 300, 302, and 304 nm, respectively (see Fig. 3

EPR Detection of the ThDP-Enamine Radical on E1o-h
The X-band continuous wave EPR spectrum of the radical generated by reaction of E1o-h (445-mol concentration of active centers) aerobically with 0.5 mM ThDP and 10 mM OG is shown in Fig. 4 (top). The spectrum has an overall line shape reminiscent of that previously assigned to thiaminderived radical species, specifically C2␣-hydroxyethylidenethiamin diphosphate (15) and ThDP-enamine radicals (15). In order to determine the validity of the assignment to a thiaminenamine radical for the experimental spectrum shown in Fig. 4, we generated spectral simulations using parameters (see "Experimental Procedures") previously assigned for the ThDPenamine radical (15). Although the simulation depicted in Fig. 4 (bottom) cannot be considered a unique fit at this stage, several points can be made regarding the parameters that provide support for the structural assignment of the radical. (a) The principal g-values used in the simulation are those measured for the ThDP-enamine radical using high frequency EPR (130 GHz) (15). Although the X-band (9 GHz) spectrum in Fig. 4 does not resolve the peaks associated with those g-values, the spectral position and shape require a g-matrix with values and anisotropy similar to those determined for the ThDP-enamine radical. Further confirmation of the g-values for the radical observed on E1o-h requires a high frequency EPR study. (b) A large, highly anisotropic nuclear spin ϭ 1 hyperfine interaction is required to simultaneously fit both the low field and high field spectral features. The value used in the simulation in Fig. 4 is consistent with that used for the 14 N in the ThDP-enamine radical and is diagnostic for hyperfine interaction expected for the 14 N in the thiazolium ring. (c) 3-5 isotropic hyperfine couplings originating from spin ϭ 1 ⁄ 2 nuclei (most likely 1 H) are required to adequately reproduce the spectral shape. This fit cannot be achieved by simply convolving a line shape function with the spectrum generated using the g-values and 14 N hyperfine interaction mentioned previously. The requirement of   OCTOBER 24, 2014 • VOLUME 289 • NUMBER 43 including 3-5 protons with isotropic couplings in the range of 3-5 G is consistent with what would be expected for a -type radical mainly localized to a thiazolium cation ring that has ␤-type protons at the various ring positions. The resolution observed in the X-band spectrum (Fig. 4) does not allow for a more detailed assignment of the protons contributing to the spectrum (e.g. determination of whether a CH 2 -CH 2 moiety has replaced a methyl group at the C2␣ position following OG decarboxylation). Indeed, a more rigorous assignment of g-values, 14 N hyperfine values, and 1 H hyperfine values will require experiments using high frequency EPR, isotopic substitution, electron nuclear double resonance, and/or electron spin echo envelope modulation. However, the X-band CW spectrum and simulation displayed in Fig. 4, together with the arguments presented above, provide strong support for the assignment of the radical species to a ThDP-derived species, probably a ThDPenamine radical.

Thiamin Intermediates on Human 2-Oxoglutarate Dehydrogenase
Upon assembly of OGDHc from equal numbers of subunits of E1o-h, E2o-h, and E3 components, the thiamin-enamine radical was once more readily detected from OG with no CoA present (Fig. 4, middle), indicating that assembly to OGDHc per se does not affect formation of the thiamin-enamine radical.
Radical species were also detected in samples of E1o-ec reacted aerobically with both OG and 2-oxoadipate in the presence of ThDP. The spectra (not shown), although lower in intensity, displayed a line shape similar to that reported previously for E1o-ec (18) and for the E1o-h/OG/ThDP system studied here (Fig. 4). The E1o-ec EPR spectra are consistent with a ThDP-enamine radical species, in agreement with the assignment made in the previous EPR/electron nuclear double resonance study of E1o-ec (18) and for E1o-h (Fig. 4).
An important further result drawn from the EPR studies is the assessment of the concentration of ThDP-enamine radicals at the E1o-h and E1o-ec active centers. Spin quantification measurements determined the amount of radical to be approximately (a) 0.9 M for the E1o-h/OG/ThDP system (0.449 mM concentration of E1o-h active centers; 0.2% occupancy), (b) ϳ0.3 M for both the E1o-ec/OG/ThDP and E1o-ec/2-oxoadipate/ThDP systems (0.150 mM concentration of E1o-ec active centers; 0.2% occupancy), and (c) ϳ1.3 M for the OGDHc-h assembled from E1o-h, E2o-h, and E3 components (0.222 mM concentration of E1o-h active centers; 0.59% occupancy). It is evident that less than 1% of the active centers of E1o-h and E1o-ec are occupied by the radical species.

The ThDP-bound Intermediate Preceding the Thiamin-Enamine Radical
Assignment of the Positive CD Band Centered near 300 nm to the Second Postdecarboxylation Intermediate-Because both the E1o-h and E1o-ec display the same CD behavior with OG (Fig. 2), and given the greater stability of the intermediates on E1o-ec, this latter enzyme was used to characterize the ThDPbound intermediate responsible for the newly observed CD/UVvisible signal. Upon the incremental addition of OG to E1o-ec, there developed two positive CD signals, one at ϳ300 and the other at ϳ350 nm (Fig. 2, bottom). The positive band near 300 nm could pertain to either pre-or postdecarboxylation ThDP-bound intermediate with C2␣-tetrahedral substitution (55).
Formation of C2-(␣-Hydroxy)-␥-carboxypropyl-ThDP was further confirmed by carrying out the reverse reaction. Succinic semialdehyde (SSA) is the expected product of non-oxidative decarboxylation of OG in the absence of enamine oxidation. The [C 2 ,C 6 Ј-13 C 2 ]ThDP-labeled E1o-ec was reacted with SSA, followed by acid quench and produced a resonance at 7.37 ppm (not shown), consistent with that observed in the forward direction and assigned to the second postdecarboxylation product (C2-(␣-hydroxy)␥-carboxypropyl-ThDP; Scheme 1). The presence of C2-(␣-hydroxy)-␥-carboxypropyl-ThDP in the NMR spectrum even after a 3-min incubation of E1o-ec with OG suggests stability of the second postdecarboxylation intermediate and correlates well with the CD results (stability of the positive 300 nm CD band).
Assignment of the Positive CD Band at 350 nm to the ThDP-Enamine Intermediate-Evidence is presented below consistent with the assignment of this band to the enamine, the experiments being necessitated by our own earlier observations, where generation of the enamine derived from pyruvate decarboxylation was modeled and gave a UV spectrum with max of 290 -295 nm (68), at a much shorter wavelength than here seen for OG. Pre-steady-state rates of formation of ThDP-bound intermediates on E1o-ec with OG were determined by a stopped-flow CD experiment. To determine which of the species giving rise to the CD bands at 300 and 350 nm is formed first, the E1o-ec (38 M concentration of active centers) from one syringe was mixed in a reaction cell with OG (4 mM) from the second syringe, and the reaction was monitored for 50 s at 297 nm or at 365 nm (two different experiments; the optimal intensity of the lamp in the instrument dictates these wavelengths as giving the strongest lines nearest to the wavelength of interest) (Fig. 6). The observed rate of formation of the second postdecarboxylation intermediate (C2-(␣-hydroxy)␥-carboxypropyl-ThDP in Scheme 1) at 297 nm is 0.78 Ϯ 0.02 s Ϫ1 (Fig.  6A), whereas at 365 nm, the rates are k 1 ϭ 2.8 Ϯ 0.1 s Ϫ1 and kЈ 1 ϭ 0.3 Ϯ 0.01 s Ϫ1 for the biphasic buildup of the enamine intermediate and k 2 ϭ 0.011 Ϯ 0.008 s Ϫ1 for its depletion (Fig. 6B). These results suggest that formation of the 350-nm species is 3-4 times faster than that of the now proven second postdecarboxylation intermediate at 300 nm, further suggesting that the species represented by the CD band at 350 nm is formed before the second (C2␣-tetrahedral) postdecarboxylation intermediate, consistent with the assignment of the 350 nm band to the enamine intermediate. This positive CD band has never been observed before with either E1o or any other ThDP enzymes.
Additional Evidence Supporting the Assignment of the Positive CD Band at 350 nm to the Enamine Intermediate-One could hypothesize three plausible candidates for this transient species produced upon the addition of OG to E1o-h or to E1o-ec (Scheme 1, bottom right): the enamine intermediate; the one-electron oxidation product of the enamine, C2-(␣-hydroxy)-␥-carboxypropylidene-ThDP radical; or the two-electron oxidation product, 2-succinyl-ThDP (see Scheme 1 for an explanation of intermediates). The following arguments support the formation of the enamine intermediate. (a) Observation of the ThDP-enamine radical with less than 1% active center occupancy (i.e. submicromolar concentration) rules out the radical because the CD instrumentation employed is incapable of observing such low amounts by either steady-state of timeresolved methods. (b) The recent report of the observation of a CD band near 390 -395 nm for 2-acetyl-ThDP tends to rule out 2-succinyl-ThDP because it should have a characteristic CD signal near that of 2-acetyl-ThDP (69). (c) The 1 H NMR spectrum of [C 2 ,C 6 Ј-13 C 2 ]ThDP-labeled E1o-ec with OG revealed a chemical shift of C6ЈH at 7.37 ppm (Fig. 5, left) corresponding to a C2␣-hydroxyalkyl-ThDP-like compound, C2-(␣-hydroxy)-␥-carboxypropyl-ThDP (second postdecarboxylation product), a result of C2␣ protonation of the enamine (Scheme 1, right). (d) We then hypothesized that the presence of an alternate electrophilic acceptor for the putative enamine would change the kinetics of build-up and/or depletion of the CD band corresponding to the intermediate. The following potential enamine acceptors were tested: (i) as a mimic of the physiological reaction, the E2o-ec-derived didomain was used (E2o-ec(1-176), comprising the N terminus, the lipoyl domain, the peripheral subunit binding domain, and linkers); (ii) DCPIP was used as an external oxidizing agent; and finally (iii) glyoxylate was used as a known enamine acceptor in a so-called "carboligase" reaction (33, 70, 71). As control, a stopped-flow CD The data were fitted using the triple exponential equation (CD ( t) ϭ CD 1 ⅐e Ϫk1t ϩ CD 2 ⅐e Ϫk1Јt Ϫ CD 2 ⅐e Ϫk2t ϩ c), and the line is the regression fit trace. k 1 , k Ϫ1 , and k 2 are the apparent rate constants; c and CD max in the exponential rise to maximum model. All stopped-flow CD experiments were performed in 20 mM KH 2 PO 4 (pH 7.0) with 0.5 mM ThDP and 2 mM MgCl 2 at 20°C. experiment of E1o-ec with OG was carried out and displayed relatively good stability of the 365 nm signal (Fig. 7a). It is evident from comparing the data in Fig. 7 (ad) that the three acceptors (DCPIP, glyoxylate, and E2o-ec(1-176) didomain) shortened the lifetime of the species corresponding to the 365 nm CD band, consistent with it being pertinent to the ThDPenamine. (e) A UV-visible stopped-flow photodiode array (PDA) experiment was carried out by mixing E1o-ec with OG (Fig. 8) and resolved several issues. (i) From the experience gained in CD studies of 12 ThDP enzymes so far, when there is a strong UV-visible signal observable along with the CD band at the same wavelength, a conjugated system is the source of the electronic transition rather than a charge transfer band, whose max is not predictable with current theory. (ii) Unexpectedly, the PDA experiment could resolve two species with max at 348 and 361 nm and a clean isosbestic point (Fig. 8, top); with time, the species monitored at 348 nm diminishes, whereas the one at 361 nm increases in amplitude (Fig. 8, bottom). Recognizing that these species correspond to a postdecarboxylation intermediate(s), two possible explanations could be envisioned once more: (i) perhaps the ThDP-enamine and the ThDP-enamine radical (Scheme 1) have uncannily similar max values, where the max at 348 nm corresponds to the former, which is converted to the radical with longer max , or (ii) both species are ThDP-enamines, perhaps in different conformations/configurations. Several arguments favor the second explanation. The species at 348 nm is fully formed at the first time point (Fig. 8,  bottom), consistent with the time-resolved CD behavior, whereas the formation of the species at 361 nm has a significant lag time. Also, the recently published x-ray structure of 2-oxoglutarate decarboxylase (72) from Mycobacterium smegmatis with OG (this enzyme is sometimes referred to as E1o in M. smegmatis) and 2-hydroxy-3-ketoadipate synthase in Mycobacterium tuberculosis (66) suggested the existence of two enamines (postdecarboxylation intermediates), which the authors termed "early" and "late" (72), the early one converted with time to the late one. (f) Although OG added to E1o-ec gave rise to this CD band near 350 nm, the related 2-oxo acids with no side chain carboxyl groups, 2-oxovalerate and 2-oxoisovalerate, did not (data not shown) (70).
The cumulative evidence points to the ThDP-enamine as the species responsible for the newly identified electronic transition near 350 nm produced by OG on E1o-h and E1o-ec. Possible explanations of the max observed are shown in Scheme 2, but whatever the correct explanation is, it must involve a longer conjugated system than present in the ThDP-enamine derived from pyruvate.

Evidence for Reactive Oxygen Species Production by E1o-h
Generation of ROS by E1o-h was detected by two independent and widely used methods: the Amplex Red fluorescence assay for detection of H 2 O 2 and the acetylated cyt c reduction assay for detection of superoxide anion. The assays verified superoxide and H 2 O 2 generation by E1o-h and exhibited quantitatively the same order of magnitude specific activity (Fig. 9). An E1o-h specific activity of 4.45 Ϯ 0.41 nmol⅐min Ϫ1 mg E1oh Ϫ1 (n ϭ 3) (pH 7.3) and 4.05 Ϯ 0.10 nmol⅐min Ϫ1 ⅐mg E1o-h Ϫ1 (n ϭ 3) (pH 6.3) was determined for H 2 O 2 generation, with no statistically significant difference detected between the results at the two pH values. The E1o-h specific activity for superoxide generation was 2.06 Ϯ 0.07 nmol⅐min Ϫ1 ⅐mg E1o-h Ϫ1 (n ϭ 5) (pH 6.3). No statistically significant difference in activity was found when the pH was changed from 7.3 to 6.3 (measured in the Amplex Red assay). These values were compared with the specific activity of the E1o-h as an isolated component or as part of OGDHc assembled from individual components ( Table 1). The comparison shows that the ROS generation capacity by E1o-h is 0.23% for H 2 O 2 and 0.12% for superoxide as compared with the E1o-h specific activity measured in the DCPIP assay.
Because the primary product is proposed to be superoxide (a one-electron reduction of O 2 to O 2 . according to Reaction 2), the H 2 O 2 present in the assay could be considered to be the product of spontaneous dismutation (disproportionation) of superoxide (6,21,23,73), similar to that reported for other enzymes (74), rather than two-electron reduction of O 2 to H 2 O 2 according to Reaction 3.

REACTIONS 2 AND 3
In support of this conclusion we present the following. (a) Because one-electron enamine oxidation on E1o-h leading to enamine-derived radical was detected by the EPR experiment, a one-electron reduction of O 2 leading to superoxide is logical.
(b) The production of superoxide by E1o-h was directly detected by the cyt c assay. (c) In a control experiment, the superoxide dismutase, when added in the cyt c assay, completely eliminated the signal by trapping superoxide produced by E1o-h. (d) When catalase was added to the Amplex Red assay in a similar control experiment, the signal was also eliminated, confirming H 2 O 2 accumulation in the Amplex Red assay.
Additional control experiments were carried out to confirm ROS generation by E1o-h. (a) In the Amplex Red assay, neither OG nor E1o-h revealed H 2 O 2 production by itself in the presence of Mg 2ϩ , ThDP, ADP, and CoA. Further, the order of addition of OG or E1o-h did not change the experimental results beyond experimental error. (b) The Amplex Red assay was also tested using commercial porcine heart OGDHc (Sigma) and the protocol reported earlier (6), leading to the observation that the rate of H 2 O 2 generation by porcine heart OGDHc was in good accord with that reported earlier. Nevertheless, a direct quantitative comparison was not reliable because of the impurities present in the commercial OGDHc preparations.  From these simple estimates we conclude that superoxide formation by E1o-h is the rate-limiting step (7.6 ϫ 10 Ϫ3 s Ϫ1 ), whereas spontaneous dismutation to H 2 O 2 (2 s Ϫ1 ) and reaction of HO 2 /O 2 . with acetylated cyt c (3.5 s Ϫ1 ) are both much faster and have nearly equal probability. At present, the functional relevance of ROS generation by OGDHc-h under various pathological conditions has been proven and is discussed in detail in the literature, including by the Semmelweis group and others (4 -6, 19 -22, 30, 31, 75-78). In theory, E1o-h could generate the same amount of superoxide as the entire OGDHc functioning in the mitochondrion based on data presented in this paper and in the literature. However, a direct comparison of ROS production by OGDHc on isolated mitochondria with data presented in this paper on E1o-h is not possible because these values were calculated per total protein but not per amount of OGDHc content. A reasonably relevant comparison is with OGDHc isolated from pig heart, providing a superoxide-producing activity of 1 nmol⅐min Ϫ1 ⅐mg protein Ϫ1 in both the physiological reaction (with OG and CoA) and in the reverse reaction with NADH. This reaction was attributed to E3-h in the OGDHc and constituted 0.3-0.4% of the overall OGDHc reaction (20).
Recently, the relative maximum capacities for O 2 . /H 2 O 2 generation were compared for NADH/NAD ϩ -dependent enzyme complexes in intact skeletal muscle mitochondria. The studies revealed that under optimal conditions, the OGDHc produced superoxide/H 2 O 2 at about twice the rate of PDHc and at 4 times the rate of BCOADHc (when oxidizing their specific 2-oxo acid substrates in the physiological reaction) and at 8 times the rate of respiratory complex site I F , indicating that OGDHc has the highest capacity in this group (19).

Conclusions
The first successful expression and purification of active, fulllength E1o and E2o components of the human 2-oxoglutarate dehydrogenase multienzyme complex enabled this collaborative effort to undertake studies directed to an elucidation of the fundamental reactions carried out by this enzyme complex. (a) Formation of a ThDP-enamine radical was demonstrated by EPR spectroscopy generated from OG for both the human and E. coli E1o components and for OGDHc-h reconstituted from the three component enzymes, as shown for the E1o-ec earlier (18). How and whether this radical, formed by reaction of dioxygen with the enamine, receives special stabilization remains to be determined. (b) Upon the addition of OG to E1o-h and E1o-ec, a new electronic transition was observed centered at 350 nm by both UV-visible and CD spectroscopy. Multiple experiments are suggestive of the transition being pertinent to the ThDP-enamine intermediate, the first ThDPbound postdecarboxylation intermediate in Scheme 1. The stability of the ThDP-enamine, compared with that derived from pyruvate, is more than likely due to some stabilization afforded by the side chain C5 carboxylate group (Scheme 2), perhaps of the enaminolate oxyanionic forming an intramolecular hydrogen bond or of the lone pair of electrons at C2␣ in the enamine forming an intramolecular hydrogen bond. Either scenario is difficult to model, unlike the enaminol form derived from pyruvate, where it was modeled by replacing an OH group by an OCH 3 group (68). (c) Of greatest import to human physiology and pathology, the E1o-h component now joins the E3-h component as being able to generate the superoxide anion and . Superoxide/H 2 O 2 production by E1o-h. A, superoxide/H 2 O 2 production by E1o-h as measured by the cyt c or the Amplex Red assay and expressed as the amount of superoxide or H 2 O 2 produced (pmol)⅐min Ϫ1 ⅐mg E1o-h Ϫ1 . Linear fitting to initial velocity data points generated always an R 2 Ͼ 0.982 in the cyt c assay or greater in the Amplex Red assay. The primary data were corrected with the average slopes of the blanks (for both assays). Error bars, S.E. values for five (cyt c assay) or three (Amplex Red assay) parallel measurements. No statistically significant difference was found between the rates of H 2 O 2 production measured at two different pH values by the Amplex Red assay. For experimental conditions, see "Experimental Procedures." B, representative Amplex Red fluorescence traces for H 2 O 2 generation by E1o-h at different pH values with additions indicated. For experimental conditions, see "Experimental Procedures." Curves have been offset for clarity. For calibration of all experiments (designated as cal at the bottom), 100 pmol of H 2 O 2 was applied in duplicates. 1 mM NAD ϩ was added to the reaction assay at pH 7.3. thence hydrogen peroxide, probably by dismutation, shown here for the first time. The specific activity on this pathway is 0.1-0.2% of the physiological activity, certainly significant. Here both products formed in Reaction 1 are clearly demonstrated for E1o-h, adding to the important differences in the aerobic behavior between the pyruvate and 2-oxoglutarate dehydrogenase complexes. The importance of superoxide formation from E1o-h under physiological conditions remains to be demonstrated in the human OGDHc.
Finally, the ability to assess the fractional occupancy of active sites by the ThDP-enamine radical, as well as the fractional specific activity of ROS formation compared with NADH production, allows us to conclude that the O 2 -induced oxidationreduction chemistry in Reaction 1 is "off-pathway" at much less than 1% of the "on-pathway" reactivity of the enamine toward reductive succinylation of E2o and thence succinyl-CoA formation. Therefore, there is little likelihood that the principal pathway follows a radical mechanism for the reductive acylation of the lipoyl-E2. The low occupancy of E1o active centers by the thiamin enamine radical suggests that the on-pathway mechanism produces succinyl-CoA via 2-electron enamine oxidation (ionic mechanism) proposed in a model reaction (79).