Evaluation by Mutagenesis of the Importance of 3 Arginines in α, β, and γ Subunits of Human NAD-dependent Isocitrate Dehydrogenase*

Mammalian NAD-dependent isocitrate dehydrogenase is an allosteric enzyme, activated by ADP and composed of 3 distinct subunits in the ratio 2α:1β:1γ. Based on the crystal structure of NADP-dependent isocitrate dehydrogenases from Escherichia coli, Bacillus subtilis, and pig heart, and a comparison of their amino acid sequences, α-Arg88, β-Arg99, and γ-Arg97 of human NAD-dependent isocitrate dehydrogenase were chosen as candidates for mutagenesis to test their roles in catalytic activity and ADP activation. A plasmid harboring cDNA that encodes α, β, and γ subunits of the human isocitrate dehydrogenase (Kim, Y. O., Koh, H. J., Kim, S. H., Jo, S. H., Huh, J. W., Jeong, K. S., Lee, I. J., Song, B. J., and Huh, T. L. (1999) J. Biol. Chem. 274, 36866–36875) was used to express the enzyme in isocitrate dehydrogenase-deficient E. coli. Wild type (WT) and mutant enzymes (each containing 2 normal subunits plus a mutant subunit with α-R88Q, β-R99Q, or γ-R97Q) were purified to homogeneity yielding enzymes with 2α:1β:1γ subunit composition and a native molecular mass of 315 kDa. Specific activities of 22, 14, and 2 μmol of NADH/min/mg were measured, respectively, for WT, β-R99Q, and γ-R97Q enzymes. In contrast, mutant enzymes with normal β and γ subunits and α-R88Q mutant subunit has no detectable activity, demonstrating that, although β-Arg99 and γ-Arg97 contribute to activity, α-Arg88 is essential for catalysis. For WT enzyme, the Km for isocitrate is 2.2 mm, decreasing to 0.3 mm with added ADP. In contrast, for β-R99Q and γ-R97Q enzymes, the Km for isocitrate is the same in the absence or presence of ADP, although all the enzymes bind ADP. These results suggest that β-Arg99 and γ-Arg97 are needed for normal ADP activation. In addition, the γ-R97Q enzyme has a Km for NAD 10 times that of WT enzyme. This study indicates that a normal α subunit is required for catalytic activity and α-Arg88 likely participates in the isocitrate site, whereas the β and γ subunits have roles in the nucleotide functions of this allosteric enzyme.

The pig heart NAD-dependent isocitrate dehydrogenase has been shown to bind tightly, per mole of enzyme tetramer, 2 mol of every ligand (isocitrate, Mn 2ϩ , NAD, ADP, NADH, and NADPH), indicating that there are half as many sites as subunits (4,5). These observations have raised the question as to whether the subunits have specialized functions for particular ligands or if there is a functional similarity among the different subunits, and two subunits contribute to a given ligand site. Ehrlich and Colman (6) showed, by separating and recombining the subunits of the pig heart enzyme, that ␣␤ and ␣␥ dimers exhibit substantial catalytic activity, whereas, the separate ␤ monomer, ␥ monomer, and ␣ dimer are essentially inactive. This study indicated the importance of association of ␣ with either ␤ or ␥ subunits to generate a catalytic species.
No crystal structure has yet been determined for any NADdependent isocitrate dehydrogenase; however, the crystal structures of the bacterial NADP-specific isocitrate dehydrogenase of Escherichia coli (7)(8)(9) and of Bacillus subtilis (10) are known. Recently, a crystal structure was solved for a mammalian NADP-enzyme:porcine NADP-specific isocitrate dehydrogenase complexed with Mn 2ϩ and isocitrate (11,32). In these enzymes, the Arg 119 of E. coli (8), Arg 110 of B. subtilis (10), and Arg 101 of pig heart (11) interact with the ␣ and ␤ carboxylates of isocitrate bound to the active site. In contrast to the mammalian NAD-dependent isocitrate dehydrogenases, these NADP-specific enzymes are not allosterically regulated by ADP.
Among the human enzyme ␣, ␤, and ␥ subunits, alignment of the amino acid sequences of all 3 subunits indicates 34% identity plus 23% close similarity. If these three subunits are aligned with the E. coli enzyme, there is 13% identity plus 29% similarity; whereas only 5% identity plus 22% similarity is observed when the human NAD enzyme is compared with both the E. coli and pig heart NADP-dependent enzymes. However, there are certain amino acids that are conserved, including those known to interact with the isocitrate in the E. coli, B. subtilis, and pig heart NADP-enzymes. Fig. 1 shows a comparison of an important region of amino acid sequence alignment of the human NAD-isocitrate dehydrogenase subunits with the E. coli, B. subtilis, and pig heart NADP enzymes. Despite the relatively low amino acid sequence identity among these enzymes, the human NAD-enzyme ␣-Arg 88 , ␤-Arg 99 , and ␥-Arg 97 can be aligned with Arg 119 of E. coli, Arg 110 of B. subtilis, and Arg 101 of the pig heart NADP enzymes; all 3 arginines are conserved.
Although there have been numerous biochemical and kinetic studies of NAD-dependent isocitrate dehydrogenases isolated from bacteria, yeast, and animal tissues, the precise function of the individual subunits has not yet been determined for these allosteric mammalian enzymes. To elucidate the functional roles of mammalian enzyme subunits, Kim et al. (12) developed a co-expression system for the three subunits of the human NAD-isocitrate dehydrogenase in E. coli. They made a construct in which human isocitrate dehydrogenase (IDH) 1 ␣, ␤, and ␥ subunits were inserted into an expression vector to produce a complete and active human NAD-dependent isocitrate dehydrogenase.
In this paper, to evaluate the functional roles of ␣-Arg 88 , ␤-Arg 99 , and ␥-Arg 97 in catalysis and ADP activation, we have separately mutated the corresponding arginines of each of the three subunits of the human enzyme to neutral glutamine by site-directed mutagenesis. Enzyme containing one mutant subunit and two wild type subunits was then expressed. We report here the results of these first mutagenesis studies on a mammalian NAD-specific isocitrate dehydrogenase. A preliminary version of this work has been presented (13).
Site-directed Mutagenesis and Construction of Plasmid-The complete recombinant human NAD-dependent IDH was expressed in IDHdeficient E. coli using the plasmid pHIDH␣␤ 2 ␥ (7.0 kbp) for expression of wild type, ␣, ␤ 2 , and ␥ subunits, as described in Kim et al. (12). Point mutations were introduced into a single subunit, one at a time, by an overlap extension (PCR)-based site-directed mutagenesis (14). Subsequently, the DNA encoding one mutant subunit and two wild type subunits were assembled into the expression vector.
To construct the IDH(␣-Q88)␤ 2 ␥ mutant, two overlapping DNA fragments were produced by amplification of the human pHIDH␣␤ 2 ␥ plasmid DNA as template, using two sets of PCR primers, namely, P1/P3 and P2/P4 as a first PCR step. The two overlapping DNA fragments, 0.28 and 0.81 kbp with the IDH(␣-Q88) mutation, were then mixed and ligated in a second PCR step, using as end primers, the P1/P4 set.
Subsequently, the mutant IDH␣ DNA (1.1 kbp) obtained from PCR ligation was subcloned into pT7Blue T-vector. The occurrence of the desired mutation was confirmed by DNA sequencing in both directions. The IDH␣ DNA was excised with NdeI/BamHI digestions and then inserted into the NdeI/BamHI site of the expression vector pT7-7 (Amersham Biosciences) to construct the mutant pHIDH(␣-Q88) plasmid. To produce a complete plasmid DNA harboring IDH(␣-Q88)␤ 2 ␥ DNA, with wild type IDH␤ 2 and IDH␥ subunits, the IDH␤ 2 and IDH␥ DNA from plasmids pHIDH␤ 2 and pHIDH␥, respectively, were removed and each was sequentially added to the mutant pHIDH␣ plasmid, as previously described (12).
Transformation and Overexpression of Recombinant IDH Proteins in E. coli Cells-The wild type and mutant recombinant plasmids were used to transform E. coli EB 106 (DE 3) cells, deficient in isocitrate dehydrogenase activity, prepared as previously described (12). Bacterial colonies with positive inserts were subcultured by growing overnight at 37°C in LB media supplemented with ampicillin (0.05 mg/ml). Aliquots of E. coli cells harboring wild type and mutant plasmid were maintained in stock solution at Ϫ80°C in 40% glycerol.
E. coli cells harboring either wild type or mutant plasmid from glycerol stocks were subcultured overnight in 200 ml of LB media containing ampicillin. Four-liter flasks, each containing 2 liter of LB media with 0.05 mg/ml ampicillin (to a total volume of 8 liters), were inoculated with freshly grown E. coli cells (2% v/v) and were grown at 37°C, at 220 rpm, until the growth reached a mid-log phase or cell density of A 600 nm ϭ 0.4. The flasks were then placed in chilled water to lower the culture temperature to room temperature. Protein expression was induced in cells by addition of isopropyl-1-thio-␤-D-galactopyranoside to a final concentration of 0.4 mM and the flasks were shaken at the lower speed of 140 rpm with minimal aeration at 25°C for 22 h. The cell cultures were then centrifuged at 5000 ϫ g for 10 min and, to improve the effectiveness of cell lysis, the cells were suspended in the relatively small volume of 200 ml of cold 11 mM citrate Tris buffer, pH 7.2, containing 10% glycerol, 0.2 mM MnSO 4 , and 0.1 mM DTT and stored at Ϫ80°C. Frozen cells in a 100-ml volume were thawed partially and lysed, while keeping the preparation in cold water, by sonication with a large probe at 20 KHz and 475 W for 10 min with intermittent mixing. The lysate was centrifuged at 16,000 ϫ g for 10 min and the clear supernatant (crude extract) was separated. The protein concentration and isocitrate dehydrogenase activity were determined in this crude extract.
FIG. 1. Comparison of selected regions of amino acid sequences of ␣, ␤, and ␥ subunits of human NAD-dependent isocitrate dehydrogenase with NADP-dependent isocitrate dehydrogenases from E. coli, B. subtilis, and pig (heart). The star indicates that the amino acids at the position are identical and the dot indicates that the amino acids are similar, using ClustalW. The Arg residues of the human enzyme subunits shown in bold letters were mutated to Gln residues. The Arg residues of the E. coli, B. subtilis, and porcine enzymes numbered in superscript are aligned with the bolded Arg of the human enzyme subunits. dependent increase in UV absorbance of NADH produced by reduction of NAD, at 25°C in a 1-ml standard assay mixture containing 33 mM Tris acetate buffer, pH 7.2, 20 mM DL-isocitrate, 1 mM MnSO 4 , and 1 mM NAD as the final concentrations. During the purification, the protein concentration was determined from the A 280 nm after correcting for the ratio of A 280 /A 260 (15). For the pure preparations, enzyme concentrations in milligrams/ml were calculated using E 280 nm 1% ϭ 6.45 (4). The specific activity was expressed as micromole/min/mg of protein. In the crude extract and ammonium sulfate fractions, to assay isocitrate dehydrogenase activity, the NADH oxidase inhibitor, rotenone (10 l in 100% ethanol), was added to the 1-ml standard assay solution to give a final concentration of 2.5 M rotenone.

SDS-PAGE of Wild Type and Mutant
Enzymes-To determine the purity of protein samples during the purification steps, aliquots of fractions were analyzed in 15% polyacrylamide gels containing 0.1% SDS in a discontinuous pH electrophoresis system (16). The preparation of stacking and resolving gel, electrophoresis running conditions, protein staining, and destaining solutions were followed as described (17). Pure mammalian NAD-dependent isocitrate dehydrogenase was recognized by the appearance of two close bands in approximately equal intensity, with an upper band (␤, ␥ subunits) of 39,000 Da and a lower band of 37,000 Da (␣ subunits) (3,18).
Purification of Wild Type and Mutant Isocitrate Dehydrogenases from Crude Extract-The wild type IDH was isolated from crude extract by ammonium sulfate fractionation: the solution was brought to 30% saturated ammonium sulfate and centrifuged to remove the precipitate. The supernatant was then brought to 50% ammonium sulfate and centrifuged. The resultant precipitate was dissolved in 12 mM sodium citrate buffer, pH 7.4, containing 20% glycerol, 0.2 mM MnSO 4 , and 0.1 mM DTT (Buffer A). This crude extract was dialyzed against 6 liters of Buffer A with 3 changes, for 6 h each. The enzyme was applied to a DE-52 column (3 ϫ 13 cm), which had previously been equilibrated with Buffer A. The column was eluted with the same buffer (525 ml) until A 280 nm reached baseline. The bound enzyme was eluted from DE-52 in a linear gradient formed from 100 ml of Buffer A and 100 ml of 50 mM sodium citrate buffer, pH 7.4, containing 20% glycerol, 0.2 mM MnSO 4 , and 0.1 mM DTT.
Fractions (5 ml) exhibiting specific activity more than 6.0 units/mg were pooled (210 ml) and concentrated to 32 ml by ultrafiltration. The pool was dialyzed with two changes, against 6 liters of 10 mM citrate Tris buffer, pH 6.0, containing 20% glycerol and 0.1 mM DTT (Buffer B). The preparation was further fractionated by chromatography on a cellulose phosphate cation-exchange column (2.5 ϫ 11.0 cm). The column was washed with 250 ml of Buffer B, and then a linear gradient consisting of 100 ml of Buffer B and 100 ml of 0.15 M citrate Tris buffer, pH 6.0, containing 20% glycerol, and 0.1 mM DTT was used to elute the enzyme. The fractions (2 ml) were collected in tubes containing MnSO 4 (final concentration, 2 mM) to maintain enzyme stability. Aliquots of fractions showing high specific activity were screened by SDS-PAGE as described above.
The mutants ␣-R88Q, ␤-R99Q, and ␥-R97Q were purified as described above except for the following modifications. In the case of the ␣-R88Q mutant, because the crude extract did not have any observable activity, fractions were evaluated by SDS-PAGE. The redissolved ammonium sulfate precipitate was dialyzed and applied to the DE-52 column under the same conditions described for wild type enzyme. Because ␣-R88Q did not bind to the DE-52 column, the fractions eluting in the starting buffer were concentrated and dialyzed against 6 liters of 12 mM sodium citrate, pH 6.3, buffer containing 20% glycerol, 0.2 mM MnSO 4 , and 0.1 mM DTT (Buffer C) with 2 changes, and was applied to a Matrex Gel Blue A affinity column (2 ϫ 11 cm) equilibrated with Buffer C. After elution with 180 ml of the initial buffer, the bound ␣-R88Q protein was eluted first in a linear gradient consisting of 100 ml of Buffer C and 100 ml of 0.2 M triethanolamine chloride buffer, pH 7.4, containing 20% glycerol, 0.2 mM MnSO 4 , and 0.1 mM DTT. The ␣-R88Q enzyme was eluted after 60 ml of the elution volume of the gradient. The remaining protein was eluted with 0.2 M triethanolamine chloride buffer, pH 7.4, containing 20% glycerol, 0.2 mM MnSO 4 , and 0.1 mM DTT. The gradient fractions (2 ml) containing pure ␣-R88Q protein were pooled and concentrated for further characterization.
For purification of the ␤-R99Q mutant enzyme, the crude extract (after ammonium sulfate fractionation) was applied to the DE-52 column under the same conditions used for the wild type enzyme. The ␤-R99Q mutant enzyme did not bind to the DE-52 column at pH 7.4. The wash fractions (250 ml) from the DE-52 column were pooled, concentrated to 12.5 ml, and dialyzed against 6 liters of 12 mM sodium citrate, pH 6.0, buffer, containing 20% glycerol, 0.1 mM MnSO 4 , and 0.1 mM DTT (Buffer D) with two changes, and were applied to a cellulose phosphate column (2.5 ϫ 11.0 cm), which was equilibrated with the same buffer used for dialysis but without MnSO 4 . After the column was eluted with 200 ml of the initial buffer, the ␤-R99Q mutant enzyme bound to the cellulose phosphate column was eluted after ϳ70 ml, in a linear gradient consisting of 200 ml of Buffer D and 200 ml of 50 mM sodium citrate buffer, pH 6.0, containing 20% glycerol and 0.1 mM DTT. Gradient fractions (6 ml) with the highest specific activity were pooled and found to be homogeneous on SDS-PAGE.
Because the ␥-R97Q mutant enzyme did not bind to either the DE-52 or cellulose phosphate at pH 7.4, chromatography on DE-52 was subsequently conducted at pH 8.0. The enzyme eluted from the cellulose phosphate column was pooled and dialyzed against 6 liters of 15 mM triethanolamine chloride buffer, pH 8.0, containing 20% glycerol and 0.1 mM DTT (Buffer E). The dialyzed ␥-R97Q enzyme was applied to a DE-52 column (3 ϫ 13 cm) equilibrated with Buffer E, followed by elution with 370 ml of initial buffer. The ␥-R97Q enzyme was eluted in a linear gradient consisting of 300 ml of Buffer E and 300 ml of 0.3 M triethanolamine chloride buffer, pH 8.0, containing 20% glycerol and 0.1 mM DTT. The fractions (7 ml) were collected into tubes containing MnSO 4 (final concentration, 2 mM) to stabilize the enzyme. The ␥-R97Q enzyme, eluted after 175 ml in the gradient, appeared to be homogeneous on SDS-PAGE.
Secondary Structure Determination by Circular Dichroism-A Jasco model J-710 spectropolarimeter was used to measure the ellipticity as a function of wavelength between 250 and 200 nm in a 0.1-cm path length quartz cell. The wild type and mutant proteins (0.1 mg/ml) were dialyzed against 2 liters of 25 mM triethanolamine chloride buffer, pH 7.4, containing 10% glycerol and 0.2 mM MnSO 4 and the same buffer was used to dilute the enzymes. The mean residue molar ellipticity [] (degree cm 2 dmol Ϫ1 ) was calculated from the equation: [] ϭ /(10 nCl), where is the measured ellipticity in millidegrees, C is the molar concentration of protein determined, as described earlier, using an average subunit molecular weight of 37,722 calculated from the sequences of the ␣ (19), ␤ and ␥ (12) subunits of human NAD-dependent isocitrate dehydrogenase, l is the cell path length (0.1 cm), and n is the number of residues per enzyme subunit (n ϭ 346, an average per subunit calculated from the total number of residues present in a tetramer of 2␣, ␤, and ␥ subunits (12,19) for human NAD-dependent isocitrate dehydrogenase).
Native Molecular Mass Determination by Gel Filtration Chromatography-Wild type and mutant enzymes, 0.4 ml of 2.0 mg/ml were applied to a Superose-12 (1 ϫ 30 cm) column of a fast performance liquid chromatography system (Amersham Biosciences). The enzymes had been dialyzed and the column was equilibrated and eluted with 50 mM HEPES buffer, pH 7.0, containing 20% glycerol, 0.2 mM MnSO 4 , and 0.1 mM DTT. The proteins were eluted at a flow rate of 0.5 ml/min and 0.5-ml fractions were collected. The column has a total volume of 24.0 ml (V t ) and a void volume of 7.14 ml (V 0 ), determined with blue dextran 2000. The elution volume (V e ) for each protein was measured from A 280 nm of the fractions. The native molecular mass of the wild type and mutant enzymes were determined from a plot of log (M r ) against Amino Acid Sequence Determination-The NH 2 -terminal amino acid sequences of wild type and mutant enzymes were determined using an Applied Biosystems Protein/Peptide Sequencer (model Procise) equipped with an on-line Microgradient delivery system (model 140 C) and a Macintosh computer (model 610).
Determination of Subunit Composition and Separation of Subunits by High Performance Liquid Chromatography (HPLC)-Wild type or mutant enzyme, 0.2 ml of 1 mg/ml, was denatured by addition of 0.05 ml of 20% SDS and the solution was incubated at 35°C for 24 h. The sample was diluted to 1.0 ml with 0.75 ml of a mixture containing 0.1% trifluoroacetic acid and acetonitrile (CH 3 CN) (4:1), and applied to a C 4 reverse phase column (Vydac) using a Varian 5000 liquid chromatograph system. The column was previously equilibrated with 0.1% trifluoroacetic acid in water and the proteins (1 ml fractions) were eluted with the same solution for 10 min followed by a linear gradient to 0.1% trifluoroacetic acid in CH 3 CN in 200 min. The fractions collected in each distinct peak were pooled, lyophilized, and subjected to NH 2 -terminal amino acid sequencing. The subunits in these protein samples were identified by comparing their NH 2 -terminal sequences with those of the known sequences of the enzyme subunits (12,19). Subunit composition was determined by estimating the area under each peak obtained from a whole protein sample applied to HPLC separation. Because the amino acids at positions 1, 6, and 7 are different in ␣, ␤, and ␥ subunits, the subunit composition of wild type and mutant enzymes was also determined from the molar ratios of amino acids in these cycles upon sequencing the whole protein.
Kinetic Studies of Wild Type and Mutant Enzymes-For K m determinations, the concentration of either coenzyme, DL-isocitrate or Mn 2ϩ , was varied, whereas the other substrates were maintained at the standard assay concentration. For the isocitrate K m determination, the substrate concentration varied between 0.05 and 20 mM; for the Mn 2ϩ K m determination, the metal ion concentration was 0.02-5.0 mM; whereas, for the NAD K m determination, 0.01-8.0 mM coenzyme was used. For the ADP activation studies, 1 mM ADP and 1.65 mM MnSO 4 were included, whereas the concentration of DL-isocitrate varied. For measurements with the ␥-R97Q mutant enzyme, 5 mM NAD was included (instead of 1 mM) in the determination of the K m -isocitrate and in the ADP activation studies. To measure the ADP concentration dependence of V max for the ␥-R97Q enzyme, the ADP concentration was varied from 5 M to 2 mM under the standard conditions at pH 7.2 except that the MnSO 4 concentration was 1.65 mM.
ADP Binding Studies by Ultrafiltration-Ultrafiltration binding experiments (equivalent to equilibrium dialysis) were carried out at room temperature (20 -23°C) in 50 mM PIPES buffer, pH 7.0, containing 20% glycerol and 0.3 mM MnSO 4 . The wild type and mutant enzymes were previously dialyzed against the same buffer. Samples containing 25 M enzyme subunits and about 80 -120 M total ADP in 0.5 ml, were mixed and incubated at room temperature for 5 min followed by filtration through Microcon YM-10 (Millipore Corp.) filter devices by centrifuging at 6500 rpm for 30 min in a microcentrifuge. (Before loading enzyme samples into the Microcon filter devices, 0.5 ml of buffer containing the particular ADP concentration, but no enzyme, was filtered completely; this "membrane conditioning" treatment was repeated three times.) Enzyme and ADP (0.5 ml) solution was filtered in the same device treated previously with the respective ADP solutions. The ADP concentration in the filtrate was determined spectrophotometrically from A 260 nm and was used as the free ADP. The bound [ADP] was calculated from the difference between total and free [ADP].

Mutagenesis, Expression, and Purification of Wild Type and
Mutant Enzymes-The human NAD-dependent isocitrate dehydrogenase, with glutamine substituted for arginine at positions 88 in ␣, 99 in ␤, or 97 in ␥ subunit, was generated by an oligonucleotide-directed PCR method and expressed in E. coli using the protein expression vector pHIDH␣␤ 2 ␥ (12, 14). There are human IDH␤ 1 (1.6 kbp) and IDH␤ 2 (1.3 kbp) cDNA sequences corresponding to two isoforms of ␤ subunit (12), of which the ␤ 2 subunit has been used here. Each subunit (with the introduction of one point mutation at a time) was assembled into the expression vector, while maintaining the other two subunits as wild type, to express a complete recombinant enzyme. The desired mutation that had been introduced in the plasmid DNA was confirmed by nucleotide sequencing analysis. The cDNA encoding wild type or mutant enzymes was expressed in IDH-deficient E. coli, and the enzymes were purified as described under "Experimental Procedures." The yield of purified human wild type and mutant enzymes was 9 -36 mg from 8 liters of cell culture. The highest specific activity of the wild type human enzyme purified by this procedure was 22 mol/min/mg, comparable with that of the NADspecific isocitrate dehydrogenase purified from pig hearts (4). Whereas ␣-R88Q mutant enzyme preparation has no detectable activity (i.e. Ͻ6.4 ϫ 10 Ϫ3 mol/min/mg), the ␤-R99Q and ␥-R97Q mutant enzymes exhibited specific activities of 14 and 2 mol/min/mg, respectively, as shown in the Table II. The purity of the enzyme preparations was assessed by SDS-PAGE. Fig. 2 shows that all four of the purified human enzyme samples exhibit only the two bands characteristic of mammalian NAD-dependent isocitrate dehydrogenase (3,18), with an upper band of 39 kDa (␤, ␥ subunits) and a lower band of 37 kDa (two ␣ subunits). The two bands exhibit approximately equal intensity as was shown for the pig heart NAD-dependent IDH enzyme (3,18). Subunits ␣, ␤, and ␥ have distinct sequences in the region of the amino terminus.
Human IDH␣, TGGVQTVTLIPGDGI Human IDH␤, ASRSQAEDVRVEGSF Human IDH␥, FSEQTIPPSAKYGGR Edman degradation of each of the enzyme preparations reveals the presence of only the sequences of the ␣, ␤, and ␥ subunits of human NAD-dependent isocitrate dehydrogenase (12,19).
Circular Dichroism Spectra of Expressed Enzymes-Circular dichroism spectra of the wild type and mutant isocitrate dehydrogenases were determined to ascertain whether the mutations have caused an appreciable conformational change. The CD spectra of the three mutant enzymes are superimposable on the spectrum of the wild type enzyme. All the spectra exhibit minima at 208 and 220 nm characteristic of proteins with appreciable amounts of ␣-helix. These results indicate that the mutations do not result in a detectable change in the secondary structure of the enzyme.
Determination of Molecular Mass of Native Enzymes by Gel Filtration-The wild type and mutant enzymes were subjected to gel filtration on a Superose-12 column equilibrated and eluted in 50 mM HEPES, pH 7.0, buffer containing 20% glycerol, 0.2 mM MnSO 4 , and 0.1 mM DTT. Fig. 3 shows that the wild type and mutant enzymes elute at about the same elution volume and each enzyme exhibits only one peak, indicating that these proteins have similar molecular weights under native conditions. As determined from a plot of log M r against K d , using standard proteins, the molecular masses of the wild type and three mutant enzymes are ϳ315 kDa. These data indicate that the native enzyme exists in solution as an octamer. The native molecular weight for human NAD-dependent isocitrate dehydrogenase as measured here is comparable with those of the purified enzymes from beef heart (20) and pig heart (2, 21); these enzymes have been reported to have molecular masses of 333 to 340 kDa as determined by gel filtration.
Determination of Subunit Composition of Enzymes by Reverse Phase Chromatography-The SDS-denatured wild type and mutant enzymes were separated and eluted from a C 4 column equilibrated with 0.1% trifluoroacetic acid in water, using a gradient in acetonitrile, as illustrated in Fig. 4. The subunits of porcine NAD-dependent IDH have previously been shown to elute in the order: ␥ then ␤ then ␣ under similar conditions (22). (The microheterogeneity observed in Fig. 4, within the region designated as ␤ subunit and within the region designated as ␣ subunit, corresponds to the observation on isoelectric focusing of several bands within a given subunit group, each of which has the same NH 2 -terminal amino acid sequence. This microheterogeneity has been attributed to differences in the extent of amidation of the acidic amino acid residues (3).) To confirm the identification of the enzyme subunits, the separated subunits were subjected to NH 2 -terminal sequencing. Table I shows the subunit composition of wild type and mutant human NAD-dependent isocitrate dehydrogenase. In each case, the ratio of ␣:␤:␥ is approximately 2:1:1, as has been shown for the NAD-dependent isocitrate dehydrogenase isolated from pig hearts (2,3). Thus the expression system described in this paper yields complete enzyme with each subunit type present in the correct proportion. Table II, first column, shows the specific activities of the purified wild type and mutant NAD-dependent IDH enzymes under standard assay conditions, as described under "Experimental Procedures." No activity was detected for the ␣-R88Q mutant enzyme either in crude extract or with the purified enzymes (i.e. activity was Ͻ6.4 ϫ 10 Ϫ3 mol/min/mg); these results indicate the Arg 88 in the ␣ subunit is essential for activity. The mutants ␤-R99Q and ␥-R97Q have measurable catalytic activity, but the specific activity is lower than that of the wild type enzyme. This result demonstrates that Arg 99 of ␤ and Arg 97 of ␥ subunits contribute to catalytic activity. The kinetic constants determined for Mn 2ϩ and NAD of wild type and mutant enzymes are shown in Table II. For the ␤-R99Q mutant enzyme, the K m values for Mn 2ϩ and NAD are similar to those of wild type enzyme, indicating that Arg 99 in ␤ subunit is not needed for Mn 2ϩ or NAD binding. For the ␥-R97Q mutant enzyme, the K m for Mn 2ϩ of the ␥-R97Q mutant is similar to that of wild type enzyme, indicating that Arg 97 in the ␥ subunit is not required for Mn 2ϩ binding. However, the K m for NAD is about 10-fold higher in the ␥ subunit mutant as com-pared with wild type enzyme, indicating that ␥-Arg 97 contributes to the apparent affinity for NAD. The V max value of 3.7 mol/min/mg (obtained by extrapolating to infinitely high concentration of NAD) is the maximum activity that could be measured for the ␥-R97Q mutant and is higher than its "specific activity" obtained with the 1 mM NAD concentration in the standard assay.

Determination of Specific Activities and Kinetic Constants for Mn 2ϩ and NAD ϩ -
Kinetic Parameters for Isocitrate and ADP Activation of IDH Enzymes-ADP is known to activate mammalian allosteric NADdependent IDH enzymes by decreasing the K m for isocitrate without changing the V max (1). Table III shows that for the human wild type enzyme, the K m for isocitrate and the 7-fold decrease in K m -isocitrate in the presence of 1 mM ADP is comparable with that determined for the pig heart enzyme (1, The protein (0.2 mg) was treated with 4% SDS, as described under "Experimental Procedures." The ␣, ␤, and ␥ subunits were separated by HPLC (C 4 column). The column was initially eluted for 10 min with 0.1% trifluoroacetic acid in water (1 ml/min), followed by a linear gradient from the initial solvent to 100% acetonitrile containing 0.07% trifluoroacetic acid in 200 min. The elution of proteins monitored at A 220 nm (-) and the gradient (--) are shown.

TABLE I
Subunit composition for wild type and mutant isocitrate dehydrogenases As described under "Experimental Procedures," the subunit ratio was determined by estimating the relative distribution of peak areas in HPLC separation of subunits obtained from the whole enzyme, as illustrated in Fig. 4, and from the molar ratios of amino acid yields (cycles 1, 6, and 7) obtained by Edman degradation of the whole protein.  23). For ␤-R99Q mutant enzyme, the K m for isocitrate is not appreciably different from that of the wild type enzyme, indicating that Arg 99 in the ␤ subunit is not required for isocitrate binding. Because the addition of 1 mM ADP does not affect the K m for isocitrate, Arg 99 of the ␤ subunit is critical for ADP activation; however, one cannot distinguish from this result whether this Arg 99 is required for ADP to bind or to activate the enzyme. For the ␥-R97 mutant enzyme, the K m -isocitrate, V max and ADP activation were determined both at 1 mM NAD and at the saturating 5 mM NAD. The results show that K m -isocitrate remains the same as that of wild type enzyme, indicating that Arg 97 in the ␥ subunit is not involved in isocitrate binding. Addition of 1 mM ADP has no effect on the K m for isocitrate, but there is a 2-fold increase in V max observed at both concentrations of NAD, showing that Arg 97 in the ␥ subunit influences the ADP activation of the ␥-R97Q mutant enzyme. In contrast to wild type or the ␤-R99Q mutant enzyme, for the ␥-R97Q enzyme, there is an effect on V max , although there is no influence of ADP on K m -isocitrate; therefore, it is apparent that ADP can still bind to the mutant enzyme.
Determination of ADP Binding of Enzymes by Ultrafiltration-Because the kinetic data do not yield information on ADP binding by the ␣ subunit mutant enzyme (which lacks detectable catalytic activity), or ␤ subunit mutant enzyme (because this nucleotide fails to affect the kinetics), we tested the ability of wild type and mutant enzymes to directly bind ADP using an ultrafiltration technique that is equivalent to equilibrium dialysis. Table IV shows the number of moles of ADP bound per mol of enzyme tetramer when 80 -120 M total ADP was added to the enzyme (at subunit concentration of 25 M). For wild type enzyme, the result indicates that up to 2 mol of ADP/mol of enzyme tetramer are bound, similar to the results previously reported for pig heart native enzyme (6). The ␣-R88Q mutant is still capable of binding ADP, indicating that the Arg 88 mutation in ␣ subunit does not prevent the ADP binding. The results indicate that the ␤-R99Q and ␥-R97Q mutants can also bind nucleotide at 80 -120 M total ADP, although less ADP is bound by these mutant enzymes than by wild type enzyme at the same total concentrations of ADP. If it is assumed that the mutant enzymes can also bind up to 2 mol of ADP/mol of enzyme tetramer, it can be estimated that the dissociation constant for enzyme-ADP is 15.9 Ϯ 7.7, 73.4 Ϯ 21.1, 234 Ϯ 97, and 513 Ϯ 156 M for wild type, ␣-R88Q, ␤-R99Q, and ␥-R97Q enzymes, respectively. (Values of greater precision cannot be determined by the method used when the affinity between enzyme and ligand is this weak.) Because ADP increases 2-fold the V max of the ␥-R97Q enzyme, the affinity between this mutant enzyme and ADP can also be estimated kinetically from the ADP concentration dependence of V max ; the dissociation constant for the ␥-R97Q enzyme-ADP complex is estimated as 302 Ϯ 44 M by this method. All three mutant enzymes retain the ability to bind ADP, albeit with weaker affinity than wild type; however, substitution of Arg by Gln at ␤-Arg 99 and ␥-Arg 97 clearly changes the allosteric effect of ADP on this isocitrate dehydrogenase.

DISCUSSION
Both NADP-and NAD-dependent isocitrate dehydrogenases catalyze the conversion of isocitrate to ␣-ketoglutarate and carbon dioxide, although only the eukaryotic NAD-dependent enzymes are allosterically regulated. Despite the very low amino acid sequence identity plus similarity for the human NAD-dependent enzyme as compared with the NADP-dependent enzymes from bacterial as well as mammalian sources, certain critical amino acids involved in isocitrate binding are well conserved. Based on the known crystal structures of the NADP-dependent enzymes from E. coli (7)(8)(9), B. subtilis (10), and pig heart (11), we chose as candidates for mutagenesis one arginine from each subunit of the human NAD-isocitrate dehydrogenase that was equivalent to Arg 101 of the pig heart NADP enzyme: ␣-Arg 88 , ␤-Arg 99 , and ␥-Arg 97 (Fig. 1). Arg 101 of the porcine enzyme is less than 2.9 Å away from the negatively charged ␣ and ␤ carboxylates of bound isocitrate (11), where it may help to position the isocitrate correctly for the catalytic reaction. The aim of the present study was to evaluate in the NAD-dependent isocitrate dehydrogenase whether the corresponding arginine in each type of subunit was important in catalysis, or whether, in some of the subunits, the arginine had a different function.
For each of the three mutant enzymes we constructed, only   one subunit contained an amino acid substitution, whereas the other two subunit types were wild type. Thus, the effect of replacing the equivalent arginine on each subunit could be separately examined. Neutral glutamines were used as replacements for the positively charged arginines, because they are similar in size and are therefore expected to produce minimal effects on the overall structure of the mutant enzyme. In fact, the circular dichroism spectra of the three mutant enzymes are similar to that of wild type enzyme, indicating that none of the mutations causes any appreciable effect on the secondary structure of the enzymes. Furthermore, all expressed mutant enzymes have the same native molecular mass as wild type enzyme, as indicated by gel filtration. The average molecular mass observed (315 kDa) corresponds to that of an octamer, as has previously been reported for the porcine and bovine NAD enzymes (20,21). Most significantly, the expression procedure did not affect the subunit composition of the enzymes, because both mutant and wild type human enzymes have similar ratios of 2:1:1 of ␣:␤:␥ subunits, the same characteristic ratios documented earlier for complete porcine NAD enzyme (3).
We now find that intact, purified mutant human enzyme, with only Arg 88 of the ␣ subunit replaced by Gln, is totally inactive, indicating that Arg 88 in the ␣ subunit is essential for catalytic activity. Although this mutant enzyme is inactive, it retains its ability to bind the allosteric activator, ADP, but with an affinity for the nucleotide that is a little less than that of wild type enzyme. These results show that Arg 88 in the ␣ subunit is not required for ADP binding. Because Arg 88 is positively charged, it is likely that it interacts with the negatively charged ␣ and ␤ carboxylates of isocitrate as does Arg 101 in the metal-isocitrate crystalline complex of the porcine NADP-isocitrate dehydrogenase (11). The role of Arg 101 of the porcine NADP enzyme was supported by our mutagenesis study in which replacement of Arg 101 by Gln caused an 18-fold increase in the K m for isocitrate as well as a decrease of V max to 4% of that of the wild type enzyme (24).
The importance for catalytic activity of Arg 88 in the ␣ subunit was presaged by affinity labeling studies of the NAD-dependent pig heart isocitrate dehydrogenase in our laboratory. The compound adenosine 5Ј-O-[S-(4-bromo-2,3-dioxobutyl)thiophosphate] was shown to react at Arg 88 and Arg 98 residues in the ␣ subunit concomitant with inactivation and, based on the ligands that protect against inactivation by the reagent, we concluded that the target site was the isocitrate substrate site (25). The same residues in the ␣-subunit (Arg 88 and Arg 98 ) were also labeled by 8-(4-bromo-2,3-dioxobutylthio)NAD; in that case, protection by both isocitrate and NADPH against inactivation led to the conclusion that the isocitrate and allosteric NADPH sites were overlapping (22,23). Both of these studies with the pig NAD enzyme not only demonstrated that the modification of Arg 88 in the ␣ subunit led to loss of activity, but also that the ␣ subunit contained the isocitrate binding site. These affinity labeling studies of the pig NAD enzyme are undoubtedly relevant to the human enzyme, because the ␣ subunit sequence is highly conserved among the mammalian species: a comparison of the ␣ subunit sequences of monkey, human, bovine, and pig shows that there is 94% identity plus similarity (22), and Arg 88 of the ␣ subunit is among the residues completely conserved among these species (22).
In contrast to the mutation in the ␣ subunit, the mutant enzyme with ␤-R99Q as the only alteration retains about 63% of the V max of wild type enzyme. Clearly, Arg 99 in the ␤ subunit is not required for catalysis, although it has some influence on activity. Furthermore, the K m values for isocitrate, Mn 2ϩ , and NAD are similar to the values for wild type enzyme, indicating that ␤-Arg 99 does not contribute to the apparent affinity of the enzyme for active site ligands. The most striking change in the kinetic properties of the ␤ subunit mutant enzyme is the loss of response to the ADP activator. Because this ␤ subunit mutant is still able to bind ADP (albeit more weakly than wild type enzyme), it may be that ␤-Arg 99 facilitates communication between ADP bound to the ␤ subunit and isocitrate bound to the ␣ subunits to effect a lowering of the K m for isocitrate.
The mutant enzyme featuring ␥-R97Q retains about 17% of the maximum activity of wild type enzyme when saturated with the coenzyme, Mn 2ϩ , and isocitrate. We conclude that Arg 97 of the ␥ subunit makes some contribution but is also not essential for catalysis. Whereas the K m values for isocitrate and Mn 2ϩ are unchanged, the K m for NAD is 10 times higher than that of wild type enzyme. Therefore, Arg 97 of ␥ subunit must strengthen the affinity of the enzyme for the coenzyme. Perhaps the positively charged guanido group of arginine interacts with the negatively charged pyrophosphate moiety of NAD. We conclude that the ␥ subunit must contain at least part of the NAD binding site.
Another notable feature of the kinetics of the ␥-R97Q mutant enzyme is that, similar to the ␤ subunit mutant enzyme, it exhibits the same K m for isocitrate in the absence and presence of 1 mM ADP. In fact, 1 mM ADP increases the V max of the enzyme 2-fold (with an enzyme-ADP dissociation constant of about 300 M) in contrast to the absence of an effect of ADP on V max for wild type, as well as the other mutant enzymes. Thus, the ␥-R97Q mutant enzyme still can bind ADP. It appears that Arg 97 of the ␥ subunit is a determinant of the mode of ADP activation; it may also modulate interaction between the ADP bound to the ␥ subunit and isocitrate bound to the ␣ subunit.
Additional evidence for the location of the ADP sites on the ␤ and ␥ subunits comes from affinity labeling of the ADP site of NAD-dependent isocitrate dehydrogenase by the reactive ADP analogue, 2-BDB-TADP (26). In this chemically modified enzyme, the K m for isocitrate is also insensitive to the addition of ADP. The modified amino acid was identified as the aspartate within the sequence LGDGLF, which was first located in the ␥ subunit (27) and later was also found to be in the ␤ subunit (28). The target aspartates can now be identified as ␤ subunit Asp 192 and ␥ subunit Asp 191 (22), indicating that ␤ and ␥ are the two subunits that contain the ADP sites.
The earlier studies in which the individual subunits were separated as monomers and then recombined to ␣␤ and ␣␥ dimers suggested that the ␤ and ␥ subunits are functionally equivalent, although they are distinguishable in sequence (6). (The ␤ and ␥ subunits are actually 53% identical in sequence and, in addition, have 17% strong similarities; they are much closer in sequence than either is to the ␣ subunit.) The basic units of the enzyme, each containing one catalytic site and one ADP site per two subunits (4,5), are proposed to be the ␣␤ and ␣␥ dimers. These dimers are then assembled into tetramers and octamers. The allosteric yeast NAD-dependent isocitrate dehydrogenase, relatively distant in evolution from the human enzyme, is an octameric enzyme composed of only two types of subunits, termed IDH1 and IDH2. In a series of studies involving mutagenesis and gene disruption (e.g. Refs. 29 -31), the laboratory of McAlister-Henn has attributed catalytic activity primarily to IDH2, and AMP activation to IDH1, but noted that there are important interactions between the subunits necessary for a functional enzyme.
The ␣, ␤, and ␥ subunits of the human NAD-dependent isocitrate dehydrogenase are clearly distinct but exhibit a resemblance in sequence: comparison of all three yields 34% identity plus 23% similarity. The present study, in which glutamine replaces a homologous arginine in each subunit in separate mutant enzymes, revealed that ␣-Arg 88 is essential for catalysis and likely participates in the isocitrate site; whereas ␤-Arg 99 and ␥-Arg 97 contribute to the allosteric activation by ADP and (in the case of ␥-Arg 97 ) the binding of NAD. We propose that in the course of evolution, gene duplication led to the diversity of subunits found in the mammalian NAD-dependent isocitrate dehydrogenase, and the arginine that was originally part of the isocitrate site was then recruited (in the new subunits) to play a role in the nucleotide functions of this allosteric enzyme.