Marked Differences between Two Isoforms of Human Pyruvate Dehydrogenase Kinase*

Pyruvate dehydrogenase kinase (PDK) isoforms 2 and 3 were produced via co-expression with the chaperonins GroEL and GroES and purified with high specific activities in affinity tag-free forms. By using human components, we have evaluated how binding to the lipoyl domains of the dihydrolipoyl acetyltransferase (E2) produces the predominant changes in the rates of phosphorylation of the pyruvate dehydrogenase (E1) component by PDK2 and PDK3. E2 assembles as a 60-mer via its C-terminal domain and has mobile connections to an E1-binding domain and then two lipoyl domains, L2 and L1 at the N terminus. PDK3 was activated 17-fold by E2; the majority of this activation was facilitated by the free L2 domain (half-maximal activation at 3.3 μm L2). The direct activation of PDK3 by the L2 domain resulted in a 12.8-fold increase in k catalong with about a 2-fold decrease in the K m of PDK3 for E1. PDK3 was poorly inhibited by pyruvate or dichloroacetate (DCA). PDK3 activity was stimulated upon reductive acetylation of L1 and L2 when full activation of PDK3 by E2 was avoided (e.g.using free lipoyl domains or ADP-inhibited E2-activated PDK3). In marked contrast, PDK2 was not responsive to free lipoyl domains, but the E2–60-mer enhanced PDK2 activity by 10-fold. E2 activation of PDK2 resulted in a greatly enhanced sensitivity to inhibition by pyruvate or DCA; pyruvate was effective at significantly lower levels than DCA. E2-activated PDK2 activity was stimulated ≥3-fold by reductive acetylation of E2; stimulated PDK2 retained high sensitivity to inhibition by ADP and DCA. Thus, PDK3 is directly activated by the L2 domain, and fully activated PDK3 is relatively insensitive to feed-forward (pyruvate) and feed-back (acetylating) effectors. PDK2 was activated only by assembled E2, and this activated state beget high responsiveness to those effectors.

The pyruvate dehydrogenase complex (PDC) 1 catalyzes the irreversible conversion of pyruvate to acetyl-CoA and NADH with the departure of CO 2 . The inactivation of PDC by phosphorylation (1) limits the commitment of glucose-connected fuels to undergoing complete oxidation or to being transformed to fatty acids (2). The fractional PDC activity is set by the competing steady state activities of two classes of dedicated enzymes, the pyruvate dehydrogenase kinase (PDK) and pyruvate dehydrogenase phosphatase (PDP). These highly regulated enzymes catalyze the phosphorylation and dephosphorylation, respectively, of the pyruvate dehydrogenase (E1) component of PDC (1)(2)(3).
PDC is regulated by at least four related PDK isoforms (4 -7) and two PDP isoforms, with related catalytic subunits (8). The PDK isozymes and the related branched chain dehydrogenase kinase (9) form a unique family of serine kinases. They are distantly related to bacterial histidine kinases, sharing five conserved motifs (4 -6). The bacterial kinases form a stable histidine-phosphate intermediate and generally transfer this phosphate to an aspartic acid side chain (10,11). The mitochondrial kinases phosphorylate serine substrates without forming a stable histidine phosphate; indeed, slow (relative to kinase turnover) autophosphorylation of a serine and no autophosphorylation of the conserved histidine was found with the branched chain kinase (12).
The availability of the individual PDK isoforms allows determination of their unique functional and regulatory properties, a step toward understanding how the required tissuespecific regulation of PDC activity is achieved. Kinasecatalyzed inactivation of PDC plays a key role in limiting glucose oxidation when more abundant fatty acids are used to provide oxidative energy (2). This routinely occurs in many tissues but is particularly important during starvation (2,(13)(14)(15)(16)(17)(18), when limited glucose must be conserved for glucose-utilizing tissues such as brain. PDC is similarly down-regulated due to high PDK activity in the diabetic state (2,(15)(16)(17)(18)(19)(20)(21)(22). In both cases, this occurs with PDK overexpression (1, 14 -22). When fatty acid oxidation is not used by a tissue or when glucose is being converted to fatty acids, the activity of PDC must be regulated very differently. Consequently, PDC limits the nearly exclusive use of glucose as an oxidative energy source in neural tissues and facilitates the conversion of glucose to fatty acids in adipose tissue when there is surplus glucose. Thus, it seems likely that different PDK (and PDP) isoforms have developed to meet the distinct tissue-specific and metabolic statespecific requirements for proper tuning of PDC activity. Variation in the distribution of specific mRNAs for the PDK isoforms supports this conclusion (22)(23)(24).
The organization of PDC plays an important role in supporting PDC activity and in the regulation of PDC by PDKs and PDPs. The core structure of PDC is formed by association of 60 dihydrolipoyl acetyltransferase (E2) subunits. E2 is a segmented protein with four domains connected by mobile linker regions (25,26). Twenty trimers of the C-terminal inner domain of E2 assemble at the vertices of dodecahedron; these trimers catalyze the transacetylation reaction (27,28). Via linker regions, this inner core domain is first connected to an E1-binding domain and then two lipoate-bearing domains, an inner domain (L2) and an N-terminal domain (L1). The dihydrolipoyl dehydrogenase-binding proteins (E3BP) has a similar segmented structure (29,30); about 12 E3BP associates via its C-terminal domain with the inner core of E2 (29 -32). E3BP then contains an E3-binding domain and a lipoyl domain (designated L3) set off by linker regions. The assembled E2-E3BP are estimated to bind 6 -12 E3 and 20 -30 E1 ␣ 2 ␤ 2 tetramers (3). Bovine PDK and PDP activities have also been shown to be markedly enhanced via binding to this central E2-E3BP core structure via the lipoyl domain region of E2 (30,34,35). By using recombinant constructs of the L1 and L2 domains of E2, unspecified isoforms of bovine PDK were shown to bind preferentially to the L2 domain via an interaction that requires the lipoyl prosthetic group (36,37). Here we begin the characterization of purified human PDKs and evaluate the functional activation and the changes in regulatory properties of two kinase isoforms, PDK2 and PDK3, as a result of their specific interactions with the lipoyl domains of human E2 and human E3BP. Our ability to produce individual lipoyl domains and E2-60-mer structures with or without E3BP allow us to sort the direct effects of individual lipoyl domains and unique contributions of the assembled complexes to activated PDK function and regulation.
Acetyl-CoA and NADH, common products of the PDC reaction and the catabolism of fatty acids, stimulate bovine PDK activity resulting in the feed-back throttling down of PDC activity (2,38,39). Increases in the NADH/NAD ϩ and acetyl-CoA/CoA ratio stimulate PDK activities by increasing the proportion of reduced and acetylated lipoyl groups on the lipoyl domain of E2 (39 -43). Pyruvate and ADP, serving as signals indicating abundant substrate and low energy, act synergistically to inhibit PDK activity by direct binding to PDKs (44,45). Evidence has been presented that sensitivity to these effectors varies with the particular PDK isoform (23). By using an all human system, we establish that marked changes in catalytic efficiency and effector responsiveness of human PDK2 and PDK3 occur as a consequence of their association with E2.

EXPERIMENTAL PROCEDURES
Materials-Recombinant human E2, K46AE2, and K173AE2 were prepared as described previously (37,46). Recombinant human E2-E3BP was prepared as described elsewhere. 2 Recombinant human E1 (47) was engineered with the polyhistidine purification tag easily removed by PreScission protease treatment; this E1 was prepared by modification of the method by Korotchkina et al. (47). 2 The acetyltransferase-catalyzing inner dodecahedron core of E2, E2 I , was prepared by tryptic removal of the exterior E1 binding and lipoyl domain region from bovine kidney E2, followed by pelleting the E2 I through sucrose layers (42). Homogeneous lipoyl domain constructs were prepared in a fully lipoylated state either as free domains or fused to glutathione S-transferase (GST) as described elsewhere (49) by log phase expression in Escherichia coli BL21 that prevented modifications of the lipoyl domains introduced by E. coli JM109 (33). The lipoyl domain constructs of E2 used include the following: outer lipoyl domain, L1; inner lipoyl domain, L2; bilipoyl structure, L1-L2; and the lipoyl domain of E3BP, designated L3. The specific design for constructing, expressing, and preparing purified GST-L3 and, from this, preparing free L3 (residues 1-98) will be described elsewhere. 2 Porcine heart E3 and thrombin were from Sigma, [␥-32 P]ATP from NEN Life Science Products, and TALON affinity purification resin from CLONTECH. Human E3 expression system was provided by M. Patel, and human E3 was prepared as described previously (50). PreScission protease is a product of Amer-sham Pharmacia Biotech; BL21(DE3) E. coli and pET28a vector were obtained from Novagen. GroESL plasmid was kindly provided by Dr. Anthony Gatenby at DuPont (51). Synthetic DNA was obtained from Integrated DNA Technologies, Inc., Coralville, IA.
Expression Vector Construction-The cloning vectors harboring the human PDK2 and PDK3 cDNA inserts were a kind gift from Kirill M. Popov and Robert A. Harris (6). For PDK2, PCR was performed with Pfu DNA polymerase (Stratagene) using primers that produced a 1.2kilobase pair product encoding mature PDK2 with a 5Ј NheI site and a 3Ј HindIII site introduced. The product was ligated into pET28a vector at the NheI and HindIII sites using T4 DNA ligase (New England Biolabs) by standard procedures (52). The resulting plasmid was cotransformed along with a plasmid encoding GroEL and GroES into E. coli BL21(DE3) made competent by the method of Inoue et al. (53). Selection for the presence of both plasmids was done on LB plates containing 50 g/ml kanamycin and 50 g/ml chloramphenicol. The pET28a vector provides a start codon and encodes an N-terminal polyhistidine sequence followed by a sequence encoding a thrombin cleavage site. Along with this construct, which provided high quality PDK2, a second expression vector was constructed in which the thrombin cut site coding region was replaced by a region coding for the PreScission protease cleavage site. This was made by removing DNA from the plasmid between NcoI and NdeI sites and replacing it with a synthesized DNA segment to encoding a cut site specific for human rhinovirus 3C protease (54).
For construction of PDK3 expression plasmid, a 270-base pair PCR fragment was generated using primers that matched the 5Ј SacI site at the beginning of the mature sequence and a 3Ј BamHI site matching the unique internal BamHI site in PDK3 cDNA. This PCR-amplified region of PDK3 was then recovered and used as a template for a second round of PCR that changed the 5Ј restriction site to NheI and maintained the 3Ј BamHI site. After restriction treatment, this was ligated with the remaining PDK3 coding region produced by digestion with BamHI and XhoI and pET28a plasmid opened from NheI to XhoI sites. This was followed by transformation and subsequent re-engineering, as described for PDK2, to produce two expression vectors, one encoding a thrombin cut site and the other a PreScission protease site, with each expressing N-terminal polyhistidine tags. All constructs were confirmed by DNA sequencing performed by the Automated Sequencing Facility at Kansas State University.
Expression and Purification of Recombinant PDK2 and PDK3-PDK plasmid-containing bacteria were grown at 37°C to mid-log phase (A 600 Ϸ0.6) in LB media containing 50 g/ml kanamycin and 50 g/ml chloramphenicol. Then expression was induced with 0.5 mM isopropyl-␤-Dthiogalactopyranoside at 22-24°C for 16 h. Bacteria were harvested by centrifugation at 4,000 ϫ g for 20 min at 4°C and frozen at Ϫ80°C. Following thawing, bacterial pellets were resuspended to 10% (w/v) in HN buffer (20 mM Hepes-Na, pH 8.0, 0.5 M NaCl, 1% (v/v) ethylene glycol). Ice water-cooled suspensions were sonicated by six repetitions of 50% pulsing at 250 watts for 30 s followed by at least 1 min of cooling. Supernatants were cleared by centrifugation at 10,000 ϫ g for 20 min at 4°C, and Pluronic-F68 was added to the supernatant to a level of 0.1% (w/v). 1 ml of equilibrated TALON resin was added per 50 ml of supernatant, and this suspension was gently mixed for 60 min at 4°C. The mixture was transferred to a column, and the gel resin was washed first with 4 column volumes of HN buffer ϩ 0.1% Pluronic-F68 ϩ 20 mM imidazole and then with 3 column volumes of HN buffer containing 0.1% Pluronic-F68 plus 25 mM imidazole. PDK was then eluted with buffer containing 100 mM imidazole. PDK containing fractions were pooled, and 1 mM dithiothreitol, 1 mM EDTA, and either 50 units of thrombin plus 2 mM Ca 2ϩ or 100 units of PreScission protease were added for a 1-or a 3-h incubation on ice, respectively. Following thrombin digestion, EGTA was added to 2 mM and glycerol was then added to 20%. Following PreScission protease digestion, protease was removed by passing the mixture through a column with 0.25 ml of GSH-Sepharose equilibrated with HG buffer (16 mM Hepes-Na, pH 8.0, 0.5 mM EDTA, 0.1% Pluronic F-68, 1% ethylene glycol, 20% glycerol). PDK preparations were desalted on a Sephadex G-25 column equilibrated with HG buffer for PDK2 and HG buffer containing 0.15 M NaCl for PDK3. PDK preparations were stored unfrozen at Ϫ20°C. Final PDK recovery was typically 5-10 mg per liter of bacterial growth media for PDK2 and 2-4 mg per liter of growth media with PDK3. Much greater purity was obtained using the Co 2ϩ -containing Talon system than with nickel-affinity columns. PDKs were stored for over 9 months in an unfrozen state at Ϫ20°C with Ͻ30% loss of activity. However, aged preparations of PDK3 needed to be preincubated with E2 at 4°C for 60 min to exhibit maximal activity. PDK2 could be concentrated to Ͼ10 mg/ml, whereas PDK3 had to be maintained at Ͻ0.3 mg/ml to avoid development of insoluble aggregates.
Kinase Activity Assays-PDK activity was measured in duplicate as the initial rate of incorporation of [ 32 P]phosphate into E1 (42, 46) using 0.1 mM [␥-32 P]ATP (150 -500 cpm/pmol) at 30°C, unless otherwise stated. For comparative purposes with assays of prior kinase preparations, kinase activities were evaluated in 60 mM Tris-Hepes, pH 7.3, with no inorganic ions and in the three buffer formulations used with purified PDKs as follows: MOPS-K ϩ buffer, 50 mM MOPS-K ϩ , pH 7.3, 20 mM K x PO 4 , pH 7.3, 60 mM KCl, 0.4 mM DTT, 0.4 mM EDTA, 2 mM MgCl 2 (42,46); phosphate buffer, 20 mM K x PO 4 , pH 7.0, 2 mM MgCl 2 , 0.2 mM EDTA, 2 mM DTT, 0.1% Triton X-100, 0.1% Pluronic F-68 (55); Tris buffer, 20 mM Tris-HCl, pH 7.4, 5 mM MgCl 2 , 50 mM KCl, 1 mM ␤-mercaptoethanol (23). Subsequently, the MOPS-K ϩ buffer system was used as described previously (42,46). With the standard levels of kinase and E1, elevated (E2-activated and stimulated) kinase initial velocities were highly linear for 2 min and lower kinase rates for over 3 min. Most data are average values of duplicate or triplicate assays measured after a 60-s reaction time; however, for assays with reduced E1 levels, the reaction time was decreased to 40 s, and in some cases, the PDK level was reduced to 0.1-0.2 g. For kinase specific activities Ͼ15 nmol/min/mg rates were nearly always within Ϯ6% of the average value reported with lower rates occasionally having somewhat higher deviations (specific activities: 8 -15 less than Ϯ10%; 3-7 less than Ϯ15%). When we report a change in the fractional activity of a PDK due to an experimental condition that falls within these experimental errors, the result (small inhibition or activation) has been qualitatively observed in repeated (at least three) experiments using multiple kinase preparations. However, variation in the specific activities of different kinase preparations (see Table II) were too large to allow the averaging of results from repetitions of the same experiment.
In studies of PDK regulation, 1 mM dichloroacetate (DCA) was added 1 min prior to ATP, and when included, ADP was introduced as mixture of 0.2 mM ADP ϩ 0.1 mM ATP to prevent trace ATP contaminating the ADP from inappropriately giving some PDK reaction before desired. NADH plus NAD ϩ (0.6/0.2 mM) was added 60 s and 50 M acetyl-CoA 20 s prior to ATP. Unless otherwise indicated, the effects of these PDC products were measured in assays with 1 g of E3 and, when including free lipoyl domains, 2 g of E2 I to catalyze the reduction and acetylation of lipoyl moieties, respectively. Acetylation of E2 was measured as described previously (41)  For studies of pyruvate inhibition of PDK activities in the presence of E2, human E1 was prepared completely free of thiamine pyrophosphate (TPP) by extensive washing of E1 bound to a nickel affinity column with buffer lacking Mg 2ϩ and containing 0.2 mM EDTA. After elution of E1 and removal of the His tag, E1 was dialyzed over a 24-h period with three changes of 50 mM MOPS-K ϩ , pH 7.2, containing 1.0 mM EDTA and then equilibration in this buffer with 0.1 mM EDTA. Complete removal of TPP was verified by a complete lack of PDC activity in reconstitution assays using high levels of E1 and excess levels of E2-E3BP and E3. Addition of TPP to this E1 gave a specific activity for E1 of 18 mol/min/mg in reconstituted PDC assays performed at 30°C. This indicates that this E1 retained full catalytic capacity. Kinase assay conditions were otherwise the same as used with DCA, above.

Expression and
Purification of PDK Isozymes-The purified PDK preparations are nearly homogeneous (96 to 99%) based on SDS-PAGE analyses. After removal of the N-terminal His tags with thrombin (monitored by SDS-PAGE), PDK2 and PDK3 have calculated molecular masses of 45,806 and 46,504, respectively, with the non-native amino acid sequence GSH-MAS retained at each N termini. The kinase monomers were 10 Da larger when prepared with by the Precision protease and retained the sequence GPHMAS. Problems encountered in studies using high levels of thrombin-detagged PDKs 3 were avoided with kinases containing the PreScission protease for detagging followed by removal of the GST-fused human rhinovirus 3C protease.
Activity of PDK Isoforms-In Tris-Hepes buffer, which lack inorganic ions, PDK2 and PDK3 had specific activities of 690 Ϯ 7 and 437 Ϯ 4 nmol⅐min Ϫ1 ⅐mg Ϫ1 , respectively, at 30°C using 0.1 mM ATP, 0.23 nM human E2-60-mer, and 28 E1 tetramers per E2-60-mer. The specific activity of PDK2 preparation in the presence of E2 in Tris at 37°C was 4-fold higher than reported for rat PDK2 (23). The standard kinase assay buffers (23,36,55) differ greatly in their ion content ("Experimental Procedures"). The patterns for variation kinase activities with the assay buffer including variations in the extent of E2 activation were notably different for PDK2 and PDK3 (Table I). In all cases, E2 markedly increased PDK2 activity. PDK2 had the highest activity at 30°C in the phosphate buffer system both in the absence of E2 (44 nmol⅐min Ϫ1 ⅐mg Ϫ1 ) and in the presence of E2 (184 nmol⅐min Ϫ1 ⅐mg Ϫ1 ). However, in the phosphate buffer system, E2 facilitated only a 4.2-fold increase in PDK2 activity, whereas a 10-fold increase in activity was observed in both the MOPS-K ϩ and Tris buffers (Table I). When PDK2 activity is measured with a fresh E2 that contains reduced lipoyl groups, specific activities in MOPS-K ϩ can rise to 130 -150 nmol⅐min Ϫ1 ⅐mg Ϫ1 at 30°C (e.g. control values Figs. 4A and 5A). The substantial decrease in PDK2 specific activity from the Tris-Hepes buffer (690 nmol⅐min Ϫ1 ⅐mg Ϫ1 ) versus the MOPS-K ϩ buffer (140 nmol⅐min Ϫ1 ⅐mg Ϫ1 for the same preparation) is primarily due to inhibition by anions. The marked influences of ions in altering PDK2 effector responses will be described elsewhere. 4 PDK3 activity at 30°C was also highest in the absence of E2 in the phosphate buffer system at 40 nmol⅐min Ϫ1 ⅐mg Ϫ1 PDK3, yet in the presence of E2 the highest activity, 308 nmol⅐min Ϫ1 ⅐mg Ϫ1 , was obtained in the Tris buffer (Table I). As with PDK2, E2 caused only a 4-fold increase in PDK3 activity in the phosphate buffer yet facilitated 17-and an 11-fold increases in the MOPS-K ϩ and Tris buffers, respectively. Phosphate lowers PDK3 activity observed in the phosphate buffer and contributes to the even lower activity in the MOPS buffer. Phosphate reduced PDK3 activity both by direct inhibition and by elevating the K m value of PDK3 for ATP (below). In Tris-Hepes, 5, 10, and 15 mM phosphate anion (Tris as counter ion) lowered PDK3 activity by 10, 16, and 29% and by 15, 33, and 49% in the presence of 0.1 mM ADP (ADP, alone, reduced PDK3 activity by 39%). Thus, ADP enhanced the inhibition of PDK3 by P i . Table II reports maximal specific activities and apparent K m values for PDK2 and PDK3 obtained at 30°C in the presence of 6.4 M E2 and 24 g of E1 with ATP varied from 0.05 to 0.3 mM. With the MOPS-K ϩ buffer system, the apparent K m value of 3 The use of commercially available thrombin preparations to remove His tags from the PDKs caused problems in subsequent studies analyzing PDK binding to GST-lipoyl domains (J. C. Baker, X. Yan, and T. E. Roche, unpublished observations.). Even the presence of thrombin inhibitors (a mixture of EGTA, aprotinin, leupeptin, benzamidine, and 4-(2-aminoethyl)-benzenesulfonyl fluoride), lipoyl domains were released from GST. Apparently, a contaminating protease in the thrombin preparations utilized the thrombin site between GST and L2. 4 X. Yan and T. E. Roche, manuscript in preparation. PDK2 for ATP is substantially lower, 5.7 M, than with the other buffer systems, and the apparent V max of 93.1 nmol/ min/mg is also reduced. However, the apparent K m value of PDK3 for ATP was lowest in the Tris buffer system, 6.3 M, and this buffer also gave the highest apparent V max , 505 nmol/ min/mg of PDK3. With more than double the maximum velocity and a lower K m , the catalytic efficiency of PDK3 is estimated to be at least 5.5-fold higher in Tris buffer than in the other buffer systems. At 37°C, the maximum velocity of PDK3 in this buffer exceeded 800 nmol/min/mg. The apparent K m values of PDK2 and PDK3 for ATP were elevated in the phosphate buffer system to 40 and 29 M, respectively. Although providing a lower specific activity for each PDK, the MOPS-K ϩ buffer system was used in further characterizing these kinases based on the trend with bovine PDKs in which elevated K ϩ and phosphate levels enhanced the regulatory responses with known effectors (38,40,45,56).
L2 Activation-Bovine kidney PDKs have been shown to bind primarily to the L2 domain of E2 (36,37). Thus, we first tested the effect of variation of the level of free L2 on PDK2 and PDK3 activity. PDK2 activity was not activated by L2 and was inhibited Ͻ30% by high levels (30 -60 M) L2. Dimeric GST-L2 modestly activated PDK2 activity with a maximum increase obtained with 16 -19 M GST-L2 subunit (8 -9.5 M GST dimer) followed by a decrease in PDK2 activity at higher GST-L2 levels until inhibition rather than activation was observed with Ͼ75 M GST-L2 subunit. In marked contrast, PDK3 was profoundly activated by L2. Fig. 1 shows the effects of variation in the level of the L2 domain on PDK3 activity. PDK3 was halfmaximally activated by 3.8 M L2 and approached maximal activation (8.3-fold) with 15 M L2. There was a sharp transition to maximal PDK3 activity (109 nmol⅐min Ϫ1 mg Ϫ1 ) at the higher L2 levels. As shown in Fig. 2A, PDK2 was not activated by 15 M L1, L2, or L1-L2, whereas GST-L1 and GST-L2 reproducibly increased PDK2 activity by about 20% (see Table IV). GST-L3 caused a small decrease in PDK2 activity, and E2-E3BP gave 7.8-fold activation, about 85% as high an activation as by E2. Thus, 15 M levels of the free lipoyl domain constructs gave little direct activation of PDK2. Apparently the 8 -10-fold increase in PDK2 activity primarily results from greatly increased encounters due to co-localization of PDK2 and E1 at the E2 surface (see "Discussion").
In marked contrast, PDK3 activity was directly and substantially increased by lipoyl domain structures (Fig. 2B). L2 or GST-L2 enhanced PDK3 activity by 7-8-fold in comparison to the over 17-fold activation by E2 and 15-fold enhancement by E2-E3BP. Interestingly, L1 caused little or no activity increase, yet the bifunctional GST-L1 supported a 1.5-2-fold enhancement in activity and GST-L3 activated to a similar extent. The bilipoyl structure, L1-L2 facilitated a substantial increase in PDK3 activity; the level of activation suggests a preferential interaction with the L2 domain. So the L2 domain is far more effective than the L1 or L3 domains in activating PDK3.
Basis for L2 Activation of PDK3-The major effect of E2 on bovine PDK1 was to lower the K m for E1 (44). We evaluated the kinetic consequences of the direct L2 activation of PDK3 activity with variation in the E1 concentration using 0.5 mM ATP (Fig. 3). 15 M L2 increased the apparent V m by 13.8-fold from 9.1 Ϯ 0.5 in the absence of L2 to 124 Ϯ 6 nmol/min/mg; the apparent K m for E1 was also reduced from 10 Ϯ 1.6 to 5.2 Ϯ 0.8 M. Under these conditions (E1 saturating, ATP 40-fold Ͼ K m (app)), L2 evidently induces a conformational change in PDK3 that markedly increases k cat . Thus, a major portion of E2 activation of this kinase is achieved by a direct effect of the L2 domain in enhancing the catalytic efficiency of PDK3. Beyond the large direct activation by L2, the additional activation (Ն1.6-fold) by E2 probably results from E2-mediated co-localization of PDK3 with E1.
Effects of Dichloroacetate (DCA) and ADP on PDK Activity- Table III shows the effects of DCA on PDK2 activity in the absence of E2 and presence of E2. PDK2 was transformed to being poorly inhibited by DCA in the absence of E2 (2.0 mM giving Ͻ50% inhibition) to being strongly inhibited in the presence of E2 (0.24 mM DCA giving 67% inhibition). The fractional inhibition (50 -55%) by 0.2 mM ADP in the absence of E2 was not significantly changed by E2. A conformational change in TABLE II Kinetic parameters for E2-activated kinase reactions ATP varied from 0.05 to 0.3 mM. Assays were conducted at 30°C using 19 g of E2 (6.4 M E2 subunit) and 24 g of E1 in 50 l of the indicated buffer system. Other conditions were as described under "Experimental Procedures." The deviations of K m (app) and V max (app) values from least square analyses of the reciprocally related x and y intercepts of plots of 1/v versus 1/(ATP) were ϽϮ7 and ϽϮ4%, respectively. a With different PDK2 preparations tested in the MOPS-K ϩ buffer system, specific activities have ranged from 85 to 155 nmol ⅐ min Ϫ1 ⅐ mg. Higher activities were in part due to reduced lipoyl groups in the E2 source used for activation (e.g. Fig. 4A and Fig 5A, control in the absence of E3).
b As PDK3 preparations were stored, it was increasingly important to preincubate PDK3 with E2 for at least 60 min at 4°C prior to assays to maintain the specific activity (Ͼ85% retained after several months). The range of specific activities of different PDK3 preparations measured in the MOPS buffer system was from 161 to 191 nmol ⅐ min Ϫ1 ⅐ mg Ϫ1 . E2-bound PDK2 may enhance the sensitivity to DCA inhibition, although an alternative explanation is considered under "Discussion." In marked contrast to PDK2, DCA inhibition of PDK3 was not enhanced by E2 with 1 mM DCA repeatedly giving Յ13% inhibition but greater inhibition by higher levels of DCA. At 0.2 mM, ADP alone gave 50 -60% inhibition of PDK3 in the presence (Fig. 4B) or absence of E2 and the combination of ADP and DCA reduced PDK3 activity by ϳ70%. Because of the lack of sensitivity to DCA (and pyruvate, below), the response of PDK3 to alternative effectors was tested. Carnitine, acetylcarnitine, malate, spermine, and calcium had no effect on PDK3 activity in the presence of E2 (data not shown). Binding of a saturating level of TPP (0.2 mM) to the E1 substrate reduced free PDK3 activity by 30%, but the activities of L2-or E2-activated PDK3 were reduced Ͻ10% by addition of 0.2 mM TPP to E1 lacking TPP (below).
Pyruvate Inhibition-DCA is normally used in regulatory studies since, unlike pyruvate, it cannot serve as a substrate in E1-catalyzed reductive acetylation of lipoyl groups that stimulate kinase activity (40). To evaluate pyruvate inhibition in the presence of E2, human E1 was prepared completely free of TPP by steps described under "Experimental Procedures." Fig. 4B shows that, as with DCA, PDK3 was not significantly inhibited by pyruvate in the presence of E2 with or without ADP. Pyruvate inhibition was also very weak in the absence of E2 (data not shown). In contrast, PDK2 was markedly inhibited by pyruvate in the presence of E2; indeed, 0.1 or 0.3 mM pyruvate was as effective, respectively, as 0.3 or 1.0 mM DCA (Fig. 4A). Pyruvate was a somewhat more effective inhibitor than DCA in the absence of E2, but E2 also greatly enhanced pyruvate inhibition. Unlike E2, GST-L2 at 15 M did not enhance pyruvate or DCA inhibition of PDK2.
Stimulation of PDK2 and PDK3 by NADH and Acetyl-CoA-As indicated in the Introduction, stimulation by NADH and acetyl-CoA are mediated by reduction and acetylation of lipoyl prosthetic groups. No stimulation of PDK2 was observed in the absence of a lipoyl domain source. Fig. 5 shows the stimulation of PDK2 and PDK3 activities after being preincubated with E2 and E3 and also presents results when PDK2 was preincubated with E2 without E3. In the presence of E3, NADH stimulated PDK2 activity by 1.6-fold and the combina-  tion of NADH and acetyl-CoA stimulated PDK2 by more than 3.5-fold. Inclusion of ADP greatly inhibited PDK2 activity but did not change the fold stimulation by the combination of products as was found with bovine PDC kinases (57) (see "Discussion"). E3 caused a decrease in E2-activated PDK2 activity; simultaneously, there was a time-dependent (probably oxygen aided) decrease in reduced lipoyl groups on the E2-60-mer. This was observed as a decrease in the capacity to acetylate sites with [1-14 C]acetyl-CoA in the absence of added NADH; this decreased from nearly 30% of lipoyl domains of E2 after preincubated with E1 for 120 min to Ͻ10% of the lipoyl domains for E2-60-mer incubated with E1 and E3. Thiol exposure leads to the reduced lipoyl groups on E2. (Recombinant E2 was purified until the last step in the presence of 10 mM ␤-mercaptoethanol; the preincubation of concentrated E1 and E2 for 120 min was in the presence of 0.8 mM DTT.) The addition of acetyl-CoA in the absence of E3 or NADH (added in all cases at a ratio of 3:1 with NAD ϩ ) gave appreciable stimulation of PDK2 by acetylating these reduced lipoyl groups. In the absence of E3, acetyl-CoA stimulation was not further enhanced by inclusion of NADH (Fig. 5, left panel). Our studies (below) with free lipoyl domains definitively support the requirement that NADH reduce lipoyl groups via the E3 reaction to stimulate kinase activity (42,43); however, the lack of elevation of PDK2 activity by NADH (Fig. 5) in the presence of E3 beyond that achieved with E2 in the absence of E3 indicates that reduction of 30 -35% of the lipoyl groups of E2 is enough to maximize the increase in PDK2 activity due to lipoate reduction.
In marked contrast, PDK3 did not significantly respond to the addition of NADH and acetyl-CoA in the presence of E2 (Fig. 5, right panel). Therefore, the high PDK3 specific activity in the presence of E2 was not further increased by reduction and acetylation of lipoyl groups. This result was somewhat unexpected from previous work on resolved bovine PDK but agrees with a previous study on PDK3 in the Tris buffer system (23). Interestingly, the addition of acetyl-CoA, alone, provided a small enhancement of PDK3 activity to yield the highest activity. This occurred in conjunction with acetylation of the low proportion of available reduced lipoyl groups of E2 (not estimated in this separate experiment from that with PDK2).
Product Stimulation of PDK2 and PDK3 Facilitated by Free Lipoyl Domain Sources-In agreement with previous studies (38 -43), in the absence of a lipoyl-bearing domain source, PDK2 and PDK3 activities were not enhanced even in the presence of E3 and E2 I (acetyltransferase-catalyzing inner core of E2, lacking the bilipoyl domain and E1 binding domain). Similarly, in the absence of E3, NADH failed to stimulate with all free lipoyl domain sources, and without inclusion of E2 I , acetyl-CoA had no effect on PDK2 and PDK3 activities (data not shown). Table IV shows the capacities of individual lipoyl domains to mediate NADH and acetyl-CoA stimulation of PDK2 and PDK3 activities. With free lipoyl domain sources, only the dimeric GST-L2 mediated a significant (2.4-fold) stimulation of PDK2. Free L2 supported a 34% increase, and L1 was ineffective; however, dimeric GST-L1 facilitated a small (but reproducible) stimulation of PDK2 (24%). These results indicate that interaction of PDK2 with a lipoyl domain source with the potential for bifunctional (GST-held) or multivalent (E2) binding aids stimulation of PDK2 by reductive acetylation (cf. "Discussion").
Based on results with E2, which failed to support PDK3 stimulation, the responses of PDK3 with various free lipoyl domain constructs were unexpected (Table IV). L1 facilitated a substantial stimulation by NADH and acetyl-CoA, and the dimeric GST-L1 supported a 6.9-fold stimulation of PDK3 upon being reductively acetylated by reaction of these products. With L2-containing structures, higher absolute PDK3 activities were obtained, but the fractional stimulation of PDK3 was lower than that obtained with L1-containing structures. Results with the bilipoyl domain L1-L2 structure indicate that PDK3 preferentially binds L2. However, L1 gains in its capacity to enhance PDK3 activity upon being reductively acetylated. Overall, our results suggest that PDK3 activity reaches a maximum in the presence of E2 such that reduction and acetylation of lipoyl groups cannot further increase the limiting catalytic step rate (see "Discussion"). Below, we evaluate the related prospect that inhibition of E2-activated PDK3 allows product stimulation of PDK3 activity to be observed, and we further evaluate the roles of L1 and L2 in E2-60-mers when only one lipoyl domain can undergo reductive acetylation.
Effects of Mutant E2 Structures on PDK Regulation-Since lipoylation is required for PDK binding to E2, we evaluated PDK function and regulation with E2 mutant structures that have an alanine individually substituted for the Lys that undergoes lipoylation, Lys-46 in L1 and Lys-173 in L2 (37). With K46A-E2 and K173A-E2, the capacity of lipoylated L2 or L1, respectively, to support activated function and product stimulation was determined in the presence and absence of inhibitors (ADP ϩ DCA). As shown in Fig. 6, left panel, when only the L2 was lipoylated (K46A-E2) or with native E2, PDK2 underwent ϳ4-fold stimulation by NADH ϩ acetyl-CoA. Strong inhibition was retained by DCA/ADP (ϳ80%) in the absence or presence FIG. 5. Stimulatory effects of NADH and acetyl-CoA on PDK2 and PDK3 activities. PDK2 and PDK3 were preincubated for 120 min with E1, E2, and E3, and samples of PDK2 were also preincubated with E1 and E2 in the absence of E3. NADH plus NAD ϩ (0.6 mM/0.2 mM) were added 60 s prior to ATP, and 50 M acetyl-CoA was added for the final 20 s prior to ATP. Assay mixtures contained 6.4 M E2 subunit and, when included, 1 g of E3 and 0.2 mM ADP. Other conditions were as described under "Experimental Procedures." of the stimulatory effectors. Thus, the regulatory effects operate in a nearly independent fashion with PDK2, and L2 is very effective in supporting enhanced activity and product stimulation. When only L1 had a functional lipoylation site (K173A-E2), all PDK2 rates for each equivalent condition were about half that of E2 or K46A-E2, but the fold changes were similar in magnitude. This indicates that, within E2, the L1 domain can directly facilitate enhanced PDK2 activity, mediate stimulation of PDK2 by acetylation of its lipoyl group, and enhance the sensitivity of PDK2 to inhibition by DCA. Nevertheless, the near equivalence of results with E2 and K46A-E2 supports a dominant role of L2 in native E2-60-mers.
In the absence of effectors, both K46A-E2 and K173A-E2 substantially activated PDK3 (Fig. 6, right panel); K46A-E2 was somewhat less effective than native E2 (in the absence of E2, PDK3 activity was 4% of E2-activated activity). K173A-E2 gave only 60% the activity wild-type E2. This agrees with the capacity of both GST-L1 and GST-L2 to activate PDK3 activity ( Fig. 2B and Table IV). When PDK3 activity fell well below full E2 stimulation, NADH plus acetyl-CoA stimulated PDK3. Acetylation of K173A-E2 gave nearly a 2-fold stimulation. Inhibition of PDK3 activity by ADP/DCA led to a stimulation with all three E2 sources, thereby producing PDK3 activities that were 22-24-fold higher than in the absence of an E2 source. With PDK3, the stimulatory effects of NADH/acetyl-CoA dominated the inhibitory effects of ADP (DCA was ineffective, above). These results establish that both acetylation of L2 and L1 in E2 can enhance PDK3 activity but that E2 activation, particularly by binding to L2, is so effective in the absence of inhibitors that there is little or no further capacity for product stimulation (cf. "Discussion"). DISCUSSION We have characterized two PDKs using an all human component system in which the PDKs, the E1, and all E2 constructs were purified with affinity tags removed, thereby leaving only six or fewer non-native amino acids on their N termini. Our preparations have high specific activities compared with prior preparations of purified PDKs (23,55). PDK2 and PDK3 specific activities varied with standard buffer conditions as did the magnitudes by which E2 enhanced the kinase activities. We have characterized effector regulation of the PDK activities in a MOPS-K ϩ buffer that contains ions that lower activity, but these ions were required for full expression of the regulatory responses of bovine PDK (38,40,45,56). Under the same conditions, our human PDK2 had a specific activity at least 4-fold higher than rat PDK2 (23). Their use of kinase-depleted porcine heart PDC as a source of E2 and E1 supported very high human PDK3 activity (23).
Our focus has been on how E2 via its lipoyl domains facilitates the largest changes in kinase activity. We have dissected the roles of individual lipoyl domains and gained insights into the importance of having the lipoyl domains of E2 housed in an E2-60-mer structure. Besides employing free lipoyl domains as monomers, we tested dimer structures in which the flexibly held domains are fused to opposing corners on one side of the GST dimer; that spacing should allow the two lipoyl domains to interact with a PDK oligomer. PDK2 and PDK3 behaved quite differently in virtually every response evaluated. In common, E2 greatly enhanced the rates at which PDK2 and PDK3 phosphorylate E1. However, the enhancements appear to result from different mechanisms.
A major contribution to elevation of PDK3 activity is the capacity for the lipoyl domains of E2, most effectively the L2 domain, to facilitate directly a very large increase in PDK3 activity, primarily through increasing the k cat of PDK3. This would seem to require that the L2 domain induces critical structural changes in PDK3 that speeds up or transforms the rate-limiting step(s) in PDK3 catalysis. Although most of the 17-fold activation of PDK3 by E2 might be attributed to this direct activation by binding to the L2 domain, binding of PDK3 to E2 is not the equivalent to binding to L2 since E2-bound PDK3 must have rapid access to many E2-bound E1. In marked contrast, free monomeric lipoyl domains had no direct effect on the catalytic efficiency of PDK2. Dimeric GST-L2 gave a small increase in activity suggesting the importance of bifunctional binding a PDK2 oligomer by L2 domains. In comparison to the free domain, L2 domains in the E2-60-mer may be in a conformation that, in binding PDK2, induces a critical conformational change in PDK2 to thereby enhance kinase activity. However, E2 activation of PDK2 may entirely result from increased productive encounters produced by the restricted orientations of the E2-bound E1 and PDK2 with access being sustained as a dissociative or direct transfer mechanism. Previous studies on bovine PDK(s), which are tightly bound to E2 (58), led to the suggestion that required encounters of a kinase with many tightly bound E1 (59,60) are facilitated by "hand over hand" transfer on the surface of E2 by either E1 tetramers (61) or by the kinase oligomer (36).
Contrary to the trend found in past studies on bovine kidney and heart PDKs (e.g. Refs. 43 and 62), E2-activated PDK2 was inhibited more potently by pyruvate than by DCA. When even a small portion of E1 retains a bound TPP, pyruvate is used by E1 to acetylate the lipoyl groups of E2 and, thereby, stimulate PDK activity (40). This prevents detection of the full potency of pyruvate inhibition through direct binding to a PDK (44). By using TPP-free E1, pyruvate inhibited PDK2 activity to the same extent as 3-fold higher levels of DCA. E2 transformed PDK2 from being poorly to potently inhibited by these inhibitors. Free lipoyl domains failed to enhance the inhibition and, therefore, to support a mechanism in which this gain in sensitivity was elicited by the conformation of PDK2 upon binding to the lipoyl domains of E2 (data not shown). Within the crowded environment at the surface of E2, the PDK2-lipoyl domain interaction may be distinct and support a conformational change in PDK2 that enhances pyruvate binding to PDK2. An alternative mechanism seems likely. PDK2 probably shares the property of bovine PDKs wherein DCA or pyruvate inhibits by binding to the PDK⅐ADP reaction intermediate and not to free PDK (45). Then, a capacity of an E2-60-mer to remove E1 availability as a limitation for the rate of PDK catalysis could increase the steady state level of PDK⅐ADP. Such an E2-facilitated transition would then increase pyruvate binding and inhibition. Systematic kinetic and binding studies should allow that mechanism to be evaluated.
In contrast, although binding to L2 had an enormous impact in enhancing catalysis by PDK3, it failed to alter the weak inhibition of PDK3 by DCA or pyruvate. To date, we have found no condition that elevates the inhibition by these ligands including addition of ADP or ions that enhance pyruvate and ADP inhibition of bovine PDKs. The very high activity of PDK3 and lack of pyruvate inhibition suggest a need for alternative inhibitors. E2-activated PDK3 activity was not significantly inhibited by loading saturating TPP on the E1 substrate, and PDK3 was insensitive to known activators of the opposing phosphatase (Ca 2ϩ or spermine). ADP and phosphate were inhibitors of PDK3 suggesting this kinase would respond to the intramitochondrial phosphate potential (ATP/(ADP ϩP i )); however, healthy tissues generally maintain an elevated phosphate potential. There would appear to be a need for an additional means of reducing PDK3 activity.
The extensive studies on bovine kidney PDKs need to be re-evaluated based on the properties of specific PDK isoforms. The sequences associated with bovine kidney E2 closely match those for PDK2 (upper band) and PDK3 (lower band). 5 Pyruvate and DCA were found to only partially inhibit the PDKs associated with bovine kidney PDC (45). This partial inhibition is apparently a consequence of differences in sensitivity of two PDKs, one (PDK2) very sensitive to inhibition and the other (PDK3) poorly inhibited by DCA or pyruvate.
As with rat PDK2 (23), PDK2 was effectively stimulated by NADH and acetyl-CoA. We have demonstrated that GST-L2 can support a much stronger stimulation of PDK2 activity (2.4-fold, Table IV) than the L2 monomer (34%). Because GST-L2 is a dimer, this indicates that the multimeric nature of lipoyl domains around the surface of the assembled E2 core contributes to facilitating the elevation of PDK2 activity due to reductive acetylation. Stimulation by free lipoyl domains required E3 to use NADH in reducing lipoyl groups and the assembled inner domain of E2 to catalyze the transacetylase reaction as was found in earlier studies (42,43). Mutation to prevent acetylation of the L1 domain in E2 allowed high stimulation of PDK2; unexpectedly, when only L1 could undergo reductive acetylation (K173A-E2), substantial stimulation of PDK2 was observed in marked contrast to the results with free L1 structures. It seems likely that acetylated L1 is more effective because it is so concentrated at the surface of E2 (Ͼ1.0 mM), although alternative explanations (altered L1 structure or a role of nonlipoylated L2) are possible. Within experimental error, the inhibitory effects of DCA plus ADP and the stimulatory effects of products operated independently in affecting PDK2 activity (Fig. 6). Given the high responsiveness of PDK2 to activation by feed-back control and to inhibition by pyruvate and ADP, it is significant that, among the PDK isoforms, PDK2 was found to be the most broadly distributed in animal tissues (23).
With just E2 present, PDK3 was not stimulated by products.
However, PDK3 was shown to be effectively stimulated when conditions were used that prevented the full, direct E2 activation of PDK3 by using a free lipoyl domain source, inhibiting E2-activated PDK3 with ADP or preventing lipoylation of the L2 domain of E2 by mutation. With E2-60-mers, independent of these changes, reductive acetylation raised PDK3 activity to nearly the same level. In combination these results indicate that fully E2-activated PDK3 is operating with reaction rate "ceiling" that cannot be elevated by reductive acetylation of the lipoyl group E2s. When activity is reduced, stimulation can occur, but the enhanced rate cannot rise above the ceiling rate. Apparently, E2-activated PDK3 catalysis becomes limited by a step that is not influenced by feed-back stimulation.
With bovine kidney PDC, ADP-inhibited kinase activity underwent a greater fractional stimulation by PDC products (57). Again this can be explained only by invoking the response of the combination of both PDKs. Upon being inhibited by ADP (not shown) or ADP plus DCA (Fig. 6), the fractional stimulation of PDK2 by NADH plus acetyl-CoA is unchanged, but PDK3 undergoes a transition from not being stimulated in the absence of ADP to undergoing stimulation of its ADP inhibited activity. Therefore, the enhanced fractional stimulation of bovine kidney PDC kinase by NADH plus acetyl-CoA may reflect the portion of kinase activity supplied by PDK3 with product stimulation reversing PDK3 inhibition by ADP.
In general, our results support the past trend of studies with the E1-associated bovine PDK in which the L2 domain was found to be the most effective in supporting PDK activity (36,37,40). We have detected a greater capacity for the L1 domain to support enhanced function of human PDK2 and PDK3. There may be conditions that enhance kinase-L1 interactions. Inclusion of the third lipoyl domain (L3) by using the assembled E2-E3BP subcomplex only modestly reduced PDK2 and PDK3 activities. GST-L3 inhibited PDK2 and modestly activated PDK3 supporting some interaction with this domain. Further studies will be needed to evaluate the consequences of an ongoing binding competition between the lipoyl domains toward the modulation of kinase activities. Chen et al. (63) presented evidence with nematode PDKs from Ascaris suum and Caenorhabditis elegans that C-terminal regions (final 84 -87 amino acids, respectively) of these PDKs are required for binding to E2, and activity is retained following removal of this region. However, mutation of a highly conserved glycine residue at position 319 of rat PDK2 within this C-terminal region caused more than a 90-fold increase in the K m value for ATP (48). Further studies are needed to establish the roles of this region and the basis for binding to the lipoyl domains of E2.
By using an all human system, we have found tremendous differences between PDK2 and PDK3. For the most part, these ensue from marked differences in how binding to the lipoyl domains of E2 alters the catalytic efficiency and the responsiveness to known effector control. The combination of PDK2 and PDK3 activities appears to explain many of the regulatory responses observed with bovine kidney PDC that contains the related isoforms.