Activated function of the pyruvate dehydrogenase phosphatase through Ca2+-facilitated binding to the inner lipoyl domain of the dihydrolipoyl acetyltransferase.

Micromolar Ca2+ facilitates ∼10-fold enhancement of pyruvate dehydrogenase phosphatase (PDP) activity by aiding the association of PDP with the dihydrolipoyl acetyltransferase (E2) component. Connected by linker regions, E2 consists of two lipoyl domains, the NH2-lipoyl domain (L1) and the interior lipoyl domain (L2), and a pyruvate dehydrogenase component binding domain surrounding a 60-mer inner core. Using recombinant constructs of L1 or L2, E2-enhanced PDP activity was markedly decreased by L2 but not by L1, effectively competing with intact E2 in Ca2+-dependent binding of PDP (half-maximal reduction at 2.0 μM L2 versus 6.7 μM E2 subunit). Using L2 fused to glutathione S-transferase resulted in direct Ca2+-dependent binding of PDP to L2 (Kd, ∼1.7 μM L2). Affinity-bound glutathione S-transferase-L2 was used to purify PDP to homogeneity by selective binding and elution by Ca2+ chelation. The large activity enhancement of PDP by E2 was eliminated by enzymatic removal of lipoates from E2 and restored by their enzymatic reintroduction. The critical role of the L2 lipoate is not in binding of PDP to E2, since PDP was still bound by delipoylated L2, and delipoylated L2 inhibited E2-enhanced PDP activity, although lipoylated L2 was more effective in each of these tests. Thus, pyruvate dehydrogenase complex activity is increased by enhanced availability of PDP to its E2-bound, phosphorylated pyruvate dehydrogenase substrate as a consequence of the Ca2+-facilitated interchange of PDP among the mobile L2 domains and an essential (undetermined) step engaging the L2 lipoate.

of the COOH-terminal domain of E2 produces a central cavity in the shape of a dodecahedron. Exterior to this inner core assemblage, after the first linker region, each E2 has a 5-kDa domain that binds the pyruvate dehydrogenase (E1) component (4,5); 20 -30 E1 tetramers (␣ 2 ␤ 2 ) bind per E2 60 . Then, set off by two more linker regions are two ϳ10-kDa lipoyl domains, an interior one (L2), and an NH 2 -terminal one (L1). An E3-binding protein (E3BP) is similar to E2 in consisting of three linker connected domains (6) in which the distinct inner domain of E3BP binds the inner domain of E2 (7,8) apparently inside the dodecahedron cavity 2 and connects by a linker region to an exterior E3 binding domain (10), followed by a linker-connected lipoyl domain.
PDC is regulated by interconversion of E1 between a nonphosphorylated, active form and a phosphorylated, inactive form (E1b). Pyruvate dehydrogenase phosphatase (PDP) catalyzes the Mg 2ϩ -requiring, Ca 2ϩ -stimulated dephosphorylation and activation of E1 (11)(12)(13)(14). Removal of phosphates from the ␣ subunit of E1 can occur with resolved E1b but is enhanced manyfold when E1b and PDP associate with the E2 core. PDP binds E2 via an interaction that requires Ca 2ϩ (15) to be provided at a micromolar level (16). The magnitude of this enhancement (typically 7-16-fold) depends on the level of E1b used, since kinetically it results primarily from a large decrease in the K m of PDP for E1b that results from concentrating PDP along with E1b at the surface of the E2 core (15). Ca 2ϩ also causes about a 2-fold decrease in the K m of PDP for Mg 2ϩ (17).
PDP is composed of two subunits (16,18). Its catalytic subunit (M r 52,600) is in the phosphatase 2C class but shares only about 20% sequence identity with rat cytosolic ␣ and ␤ isoforms of phosphatase 2C (19). The other subunit (M r 96,000) serves a regulatory role by affecting the concentration of Mg 2ϩ required for PDP activity (20). This subunit is a flavoprotein (FAD), and its sequence is distantly related to the mitochondrial flavoprotein dimethylglycine dehydrogenase (14), which functions in choline degradation. Spermine and, to a lesser degree, other polyamines reduce the K m of PDP for Mg 2ϩ (21), probably by reversing the effect of the regulatory PDP subunit (20). The effect of spermine appears to be a direct effect on the phosphatase (22). The concentration dependence of PDP for Mg 2ϩ is also reduced in permeabilized mitochondria prepared from insulin-treated adipose tissue (23).
A major interest of our laboratory is how the E2 core binds and aids the function of the kinase and phosphatase components. We have recently established that the kinase preferentially binds to the L2 domain of E2 and that L2 must retain its lipoyl cofactor for that association to occur (24). We have also found that the reduction and acetylation of the L2 lipoate * This work was supported by National Institutes of Health Grant DK18320, a grant from the Kansas Affiliate of the American Heart Association, and Kansas Agricultural Experiment Station Contribution 95-561-J. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  markedly enhances kinase activity (25). Previously, we have used protease Arg C to progressively remove the outer domain regions of E3BP (then termed protein X) and then those of the E2 core and observed a loss of the capacity for enhanced PDP function correlated with removal of the outer domains of E2 (22). In cleaving E2, protease Arg C removed the lipoyl domain region of E2 and in a separate cleavage removed the E1 binding domain. Here we have found that E1b binding to E2 makes it a better substrate for PDP, established that Ca 2ϩ -dependent binding of PDP to E2 occurs at the L2 domain of E2 and characterized the affinity of this interaction, uncovered a critical role for the L2 lipoate for E2-enhanced PDP activity, and used the specific PDP-L2 interaction to purify PDP.

EXPERIMENTAL PROCEDURES
Materials-The bovine kidney PDC and, from the complex, the resolved E1 and E2⅐E3BP⅐kinase subcomplex were prepared by standard procedures (26,27). The E2 lipoyl domain region was selectively removed from E2•E3BP•kinase by treatment with Clostridium histoliticum collagenase and a kinase and lipoyl domain-free E2 IB •E3BP subcomplex recovered as described previously (5). Bovine kidney PDP was purified through the DEAE chromatography step according to the procedure of Pratt et al. (18) to yield a stable preparations in which PDP constituted 20 -50% of the protein. The preparation used in most studies had an initial specific activity of 370 units/mg (units defined below); PDP activity decreased following freezing and thawing, but the reduced activity was very stable. Recombinant constructs of the lipoyl domains of human PDC E2 were prepared both fused to glutathione S-transferase (GST) and as free structures as described previously (28). The E2 amino acid sequences of the constructs used were L1(1-98), L2(120 -233) L1•H1•L2(1-233) and H1•L2(98 -233), with H1 designating the first hinge (or linker region) connecting L1 and L2. Enterococcus faecalis lipoamidase was prepared by minor modification of the procedures of Suzuki and Reed (29), and lipoyl protein ligase, made as described by Green et al. (30), was kindly provided by John Guest (University of Sheffield, Sheffield, United Kingdom).
PDP Assays-Phosphatase activity was determined by measuring the increase in PDC activity or the release of [ 32 P]phosphate from PDCb or resolved E1b. To prepare phosphorylated PDC, 5-10 mg of complex was incubated at 30°C with 0.5 mM ATP in 50 mM MOPS-K (pH 7.5), 0.5 mM EDTA, 2 mM dithiothreitol, and 1.2 mM MgCl 2 for 5 min, followed by at least 30 min at 4°C. 32 P-E1 (6 -8 mg) was prepared in the same buffer by incubation of 50 g of E2•kinase/mg of E1 for 30 min at 30°C, followed by overnight on ice. [␥-32 P]ATP was initially at least 400 cpm/pmol, and, for either PDC or resolved E1, at least three [ 32 P]phosphates were incorporated per E1 tetramer. ATP was then hydrolyzed by treatment with hexokinase and glucose, followed by further dialysis with EDTA reduced to 0.1 mM. To remove E2•kinase from E1b, this structure (saturated with bound E1b) was pelleted by centrifugation in a Ti-50.2 rotor at 35,000 rpm for 90 min and excess E1b recovered in the supernatant. The protein concentrations of phosphorylated substrates were measured by the BCA technique as described previously (28).
The rate of increase in PDC activity by PDP (1 unit ϭ an increase of 1 mol NADH/min) was estimated by minor modifications of the procedure of Pratt et al. (18). In a final volume of 25 l of buffer A (40 mM MOPS-K, final pH 7.4, 0.1 mM EDTA, 0.4 mM dithiothreitol, 1.2 mM Ca 2ϩ , 1.0 mM EGTA, and 2 mg/ml Pluronic F-68 (BASF Corp.)), 20 g of PDCb with 0.02-0.03 units of PDP (about 0.1 g of DEAE fraction or 0.04 g of purified PDP) were incubated for 60 s at 30°C followed by addition of Mg 2ϩ to a final concentration of 10 mM (unless otherwise indicated) to initiate PDP activity. After 120 s, a 20-l sample of this reaction mixture was added to a standard 1-ml NAD reduction assay (18,31), which served to quench the phosphatase reaction and measure active PDC. This rate was corrected for endogenous PDC activity, which was evaluated by conducting the same assay in the absence of added PDP or with PDP but in the absence of added Mg 2ϩ . These assays gave similar results, but very low endogenous PDP activity was sometimes associated with the PDC substrate (plus Mg 2ϩ higher than minus Mg 2ϩ in the absence of added PDP). For assays measuring the specific activity of PDP preparations, the plus Mg 2ϩ -minus PDP blank was used, and trace endogenous PDP activity was subtracted; for assays testing effects of added PDP effectors, the minus Mg 2ϩ blank was used. Endogenous PDP underwent similar activity changes as added PDP. In the absence of Mg 2ϩ , no increase in PDC activity with time was detected with or without added PDP, indicating no contamination by the mitochondrial Mg 2ϩ -independent phosphatase (32) in either PDP or PDC preparations.
Release of phosphate from 32 P-PDC was assessed in reaction mixtures of the same composition, using the same sequence of steps and same control assays as described above; however, reaction mixtures also contained 0.2 mg/ml bovine serum albumin and were quenched 120 s after addition of Mg 2ϩ by the addition of 200 l of 10% (w/v) trichloroacetic acid, followed by immediate vortexing and incubation on ice for at least 30 min, followed by centrifugation for 10 min at 14,000 ϫ g. Assays performed in the absence of added Ca 2ϩ but with 1.0 mM EGTA added reduce free Ca 2ϩ to Յ1 nM even in the presence of 10 mM Mg 2ϩ . When either the Mg 2ϩ concentration was varied or further additions were included in reaction mixtures, those are indicated in the text or figure legends. The radioactivity of a 100-l aliquot or two 80-l aliquots and of 1 nmol of the original [␥-32 P]ATP also in 100 l of 10% trichloroacetic acid were counted using standard liquid scintillation techniques. In assays of the release of 32 P from E1b and E1b with other components added at the levels indicated in the legends, somewhat higher levels of phosphatase (0.2-0.4 g) were used, and reaction times were sometimes increased to 3 min. All assays were performed at least in duplicate, and standard deviations from average values are shown.
Binding of PDP to GST-Lipoyl Domain Construct-The indicated levels of GST-H1⅐L2(98 -233) were combined with 0.26 units of PDP (1.8 g of a stable DEAE fraction of PDP) in 60 l of buffer containing 40 mM MOPS-K, pH 7.5, 1.2 mM Ca 2ϩ , 1.0 mM EGTA, 0.4 mM dithiothreitol, and 2 mg/ml Pluronic F-68. Studies were conducted in the absence and presence of Mg 2ϩ (2.0 mM Mg 2ϩ plus 0.2 mM EDTA when added). After 120 s, 50-l samples held at held at room temperature (ϳ24°C) or on ice were transferred to calibrated 0.5-ml Eppendorf tubes containing 25 l of packed GSH-Sepharose prepared in the same buffer and at the same temperature. These were mixed by vortex for 10 s and during the next 10 s transferred to an Eppendorf 5415 C centrifuge, and then the gel beads with the bound GST construct were pelleted at 14,000 rpm for 10 s. PDP activity was determined for duplicate 5-l samples of each supernatant. Nonspecific binding was evaluated under the same conditions using the same range of GST levels as GST-H1⅐L2. We found a wide variation in the capacity of different GSH-Sepharose lots to rapidly and nearly completely bind GST or the GST construct (residual GST activity in the supernatant measured as described previously; Ref. 28). The preparation used (Pharmacia Biotech Inc., lot 236820) reproducibly bound Ͼ92% of the highest level of the GST construct used under the above conditions, whereas most lots tested bound only 65-90% of the GST activity. Variation was primarily in the rate of binding rather than the extent of binding; the basis of the difference is not understood. Minor corrections of data were made assuming the same proportion of PDP was bound to the small portion of GST-H1⅐L2 (determined as GST activity that remained in the supernatant) and did not pellet with the GSH-Sepharose.
Purification of PDP-GST-H1⅐L2 (3 mg) was bound to 1.5 ml of GSH-Sepharose and then equilibrated in 40 mM MOPS-K, 1.2 mM Ca 2ϩ , 1.0 mM EGTA, 2.0 mM MgCl 2 , 0.2 mM EDTA, and 10 mg/ml Pluronic F68. To 1 ml of this lipoyl domain-bearing gel, 80 -160 units of PDP, typically 0.5-1 mg protein, prepared through the DEAE concentrator column step previously described, were incubated at 4°C in 2 ml of the above buffer for 30 min with mild shaking. Free PDP activity was Ͻ2% of the original level. This gel was then layered over 0.5 ml of GST-H1⅐L2-GSH Sepharose washed with 10 ml of the above buffer and then eluted at 25 l/min with the above buffer modified to contain no Ca 2ϩ , 1.5 mM EGTA, and 10% glycerol (v/v). PDP activity was determined for each fraction by spectrophotometric assay, and individual fractions were frozen and stored at Ϫ80°C.
Lipoylation Status-The removal and reintroduction of lipoyl groups of phosphorylated PDC were executed by minor modifications of developed procedures (24,25). 32 P-PDCb was treated with lipoamidase (300 g/g) in 50 mM sodium phosphate (pH 7.5) at 30°C. Samples of PDCb were removed at the indicated times and evaluated for changes in phosphatase activity (20 g) and lipoyl content (5 g). Lipoyl content was estimated by determining the change in E3 activity using a cyclic assay (26) carried out in microplate wells containing 50 mM sodium phosphate (pH 7.5), 1 mM EDTA, 0.3 mM NADH, 0.1 mM NAD ϩ , and 0.2 mM 5,5Ј-dithiobis(2-nitobenzoate) in a final volume of 200 l. The change in absorbance at 405 mn (⑀ ϭ 27.2 A min Ϫ1 cm Ϫ1 ) was measured with time using a UV Max microplate reader. The very low residual E3 activity observed following extended delipoylation was comparable to the E3 activity obtained by addition of a level of lipoate equivalent to the lipoyl domains in PDCb. (Preliminary experiments using PDC not labeled with 32 P confirmed that loss of E3 activity correlated with loss of acetylation sites.) After 150 min, 0.4 mM phenylmethylsulfonyl fluoride was added, the incubation continued for 12 min to inactivate lipoamidase, followed by addition of dithiothreitol to 0.5 mM to quench the reactivity of residual phenylmethylsulfonyl fluoride, and dialyzed into 40 mM MOPS-K (pH 7.5) and 0.2. mM EDTA.
Restoration of lipoyl groups onto 180 g of the delipoylated PDCb was conducted by incubating 1.8 g of Escherichia coli lipoyl protein ligase for 20 min in the presence of 0.25 mM [␥-32 P]ATP (preparation used to make PDCb to maintain the radiospecific activity due to any further phosphorylation), 0.1 mM lipoate, and 1.5 mM MgCl 2 (25). EDTA was added to 1.5 mM to terminate the reaction along with leupeptin and aprotinin to 1 g/ml, followed by dialysis for 48 h against 40 mM MOPS-K (pH 7.5) and 0.2 mM EDTA. SDS-PAGE analysis (33) showed no change in the pattern of the delipoylated and relipoylated PDCb preparations, except that delipoylated E2 had a mobility slightly slower than the original or relipoylated E2. The level of lipoylation was evaluated by comparing acetylation of lipoyl moieties of 8.7-g samples of the PDCb preparations using [1-14 C]acetyl-CoA in the presence of NADH along with removal of CoA formed by the ␣-ketoglutarate dehydrogenase complex reaction as described previously (25,28). PDC samples were stored Ϫ80°C for 8 weeks prior to measurements of acetylation, so 32 P counts were low; corrections for 32 P were made from analysis of equivalent samples not exposed to [1-14 C]acetyl-CoA by count differentials in the 14 C energy range and comfirmation of counting equivalency in the energy range of 32 P decay above that for 14 C.
The delipoylation of 31 g of L2 or 82.5 g of GST-H1⅐L2 was carried out by incubation with 0.45 g of lipoamidase for 2.5 h at 30°C. Full delipoylation was confirmed by complete conversion to a form with a decrease in mobility in native gel electrophoresis as described previously (28).

Effect on PDP Activity of Removal of the E2 Lipoyl Domain
Region but not Its E1 binding Domain-E2 enhances the rate of dephosphorylation of 32 P-E1b (M) by ϳ10-fold, but reduction of free Ca 2ϩ to a subnanamolar level by chelation with EGTA eliminates nearly all of this activation. Rahmatullah et al. (22) found that loss of E2-activated PDP activity correlated with removal of the outer domains of E2 and not with removal of the outer domains of E3BP. However, although the approach used (protease Arg C) first removed the outer domain of E3BP, it cleaved linker regions on both sides of the E1 binding domain of E2, so E1 binding was also being progressively lost. Treatment with collagenase selectively removes the bilipoyl domain region of E2, and the resulting E2 IB ⅐E3BP subcomplex binds E1. E2 IB ⅐E3BP (5 g) gave a 43% stimulation of PDP activity at 10 mM Mg 2ϩ (from 30 Ϯ 1.5 to 43 Ϯ 1.1 nmol of [ 32 P]PO 4 released/min/mg of PDP) and nearly a 2-fold stimulation at 0.5 mM Mg 2ϩ (from 5.35 Ϯ 0.7 to 11.9Ϯ.2 nmol of [ 32 P]PO 4 released/min/mg). Accordingly, binding of E1b to E2 improves the presentation of E1b as a substrate to PDP. Mechanisms involving restriction of the movement and orientations of bound E1b (entropic effect) or a conformational change in bound E1b could contribute to this enhanced dephosphorylation.
Inhibition of PDP by Lipoyl Domains-Since removal of the outer lipoyl domain region of E2 prevents most of the E2activated PDP function, it seemed likely that the Ca 2ϩ -dependent interaction of PDP with E2 was at a lipoyl domain. Fig. 1a shows the effects of 5 and 15 M L1, L2, H1⅐L2, and L1⅐H1⅐L2 on the rate of PDP dephosphorylation of 32 P-PDC in the presence of a saturating level of Ca 2ϩ . L2-containing structures (L2, H1⅐L2, and L1⅐H1⅐L2) reduced E2-activated phosphatase activity, with Ն67% reduction due to 5 M L2 or H1⅐L2 and Ն60% due to the bilipoyl domain L1⅐H1⅐L2 structure, and 15 M L2 or H1⅐L2 reduced the rate of dephosphorylation to ϳ16%, with a slightly higher activity (ϳ19%) with this level of L1⅐H1⅐L2. In marked contrast, L1 had little if any effect on E2-enhanced phosphatase activity at either concentration. Assuming inhibition resulted solely from free L2 competing in binding PDP with the L2 domains of the assembled PDCb E2 (subunit concentration, ϳ6.4 M), the finding that 5 M free L2 gives more than 50% inhibition would indicate that free L2 competes very effectively. Fig. 1b also shows the effect of L2 on PDP activity when Ca 2ϩ was chelated by EGTA. Removal of free Ca 2ϩ greatly diminished PDP activity, and little, if any, inhibition was detected with 15 M L2, supporting the prospect that L2 inhibition results from selective binding of PDP to L2 in a Ca 2ϩ -requiring process.
Concentration Dependence for L2 Inhibition- Fig. 2 shows the concentration dependence for L2 inhibition with 21 g of 32 P-PDC (ϳ10 g of E2, 6.7 M E2 subunit). Fig. 2, inset, presents the same data in a Dixon plot. Half-maximal inhibition occurred at ϳ2.0 M L2, which is below the E2 subunit concentration. The solid lines are theoretical fits derived as- suming that L2 inhibition resulted from competitive binding of L2 and intact E2 subunits. The excellent fit supports the operation of a simple competition between E2 and L2 and yields a computed optimum ratio for the equilibrium dissociation constants for the formation of L2⅐PDP and E2⅐PDP complexes of 0.335 (the ratio is calculated as described in the legend to Fig.  2). The results strongly suggest that free L2 competes very effectively with the intact E2 subunits of PDC for the binding of the PDP, thereby eliminating rapid dephosphorylation of E1b bound to the E2 in PDC.
Inhibition by GST-H1⅐L2-The preparation of lipoyl domains fused to GST affords a simple method for evaluating direct binding of PDP to L2. The use of GST-H1⅐L2 was selected because the L2 domain is separated from GST by an intervening mobile linker region (H1). Fig. 1a shows that inhibition of E2-activated PDP activity by 5 and 15 M GST-H1⅐L2 (monomer concentration) was reduced (55 and 77% inhibition, respectively), significantly less than by L2 alone. Not only might GST (and L1 in L1⅐H1⅐L2) physically interfere with PDP-L2 encounters, but, because the fusion protein is a dimeric structure ((GST-H1⅐L2) 2 ), it would lower the effective concentration and slow the rate of diffusion. Nevertheless, GST-H1⅐L2 was clearly still very effective and was used in tests of direct binding of PDP to L2.

FIG. 2. Change in E2-activated PDP activity with increasing
L2. PDP assays were conducted with 21 g of 32 P-PDC in the presence of 10 mM Mg 2ϩ with the indicated levels of L2 (construct amino acid sequence 120 -233). Other conditions were as described in "Experimental Procedures." Inset, data replotted in a Dixon plot. The lines shown in both figures are theoretical fits for the ratio of the equilibrium binding constants of PDP with L2 over that of PDP with E2 subunits: ). This equation assumes that free PDP (not bound to L2 or E2) is not significant (at a fixed E1, the level of E2 present in PDCb used gives near maximal stimulation of the kinase). The observed activity is calculated as: where V max is the activity in the absence of L2, and V min is the extrapolation of the best fit line to infinite L2. L2⅐PDP is derived from the best r value fit of the data obtained by iteratively solving for L2⅐PDP at all L2 levels from the quadratic form of the first equation; r ϭ 0.335 gave the best fit. Ca 2ϩ -dependent Binding of PDP to L2-Using a buffer containing 1.2 mM Ca 2ϩ plus 1.0 mM EGTA, PDP was selectively bound by GST-H1⅐L2 anchored on GSH-Sepharose but not by GST so anchored. PDP was then selectively eluted by washing with 1.0 mM EGTA. Thus, direct Ca 2ϩ -dependent binding was demonstrated. Furthermore, we found tight binding not only required Ca 2ϩ but was markedly enhanced by the presence of 2.0 mM Mg 2ϩ (e.g. the fractional binding of PDP by 5 M GST-H1⅐L2 was reduced from 72 to 41% when Mg 2ϩ was not included). Fig. 3 shows the change in PDP binding with the level of GST-H1⅐L2 in the presence of 0.2 mM free Ca 2ϩ and 1.8 mM free Mg 2ϩ . In this experiment, the fusion protein and PDP were incubated together, followed by rapid mixing with GSH-Sepharose and pelleting of the Sepharose beads. An apparent binding constant of about 1.75 Ϯ 0.4 M L2 is obtained from Fig. 3. Thus we have obtained direct support for PDP binding to L2; however, this binding is weaker than expected from analysis of activity studies or Fig. 2 (see "Discussion"). Our conditions of rapidly executing the steps of binding of the fusion protein to beads and removing the GST structures while minimizing the volume of the beads prevented use of higher concentrations of GST-H1⅐L2 at room temperature. In one study on ice, GST-H1⅐L2 was rapidly and nearly completely bound by the affinity gel at a somewhat higher fusion protein level. At the reduced temperature, the binding was somewhat tighter over the same concentration range as Fig. 3, (K d , ϳ0.8 M), and a weaker class of binding sites was detected at higher levels of the fusion protein (K d , ϳ3 M). The many data points in Fig. 2 are closely fit by a single binding constant and not fit well by two classes of sites; therefore, we surmise that a technical deficiency in the binding tests is occurring at the higher levels of fusion protein. Consistent with somewhat tighter binding at reduced temperatures, slower dissociation of PDP from GST-H1⅐L2 at 6°C than room temperature was found in developing the conditions for PDP purification.
Purification of PDP-Tests revealed that PDP could be maintained fully bound to the L2-bearing column after washing with 5 column volumes in the presence of Ca 2ϩ and Mg 2ϩ at 6°C, and that a transition to buffer containing EGTA and no added Ca 2ϩ , followed by slow elution, gave good recovery of PDP activity in a fairly sharp peak (Fig. 4a). This process was used in the purification of the PDP. Fig. 4b shows the pattern of SDS-PAGE-separated PDP eluted from GSH-Sepharose-GST-H1⅐L2. Only PDP subunits were observed, indicating that essentially homogeneous PDP (specific activity, 660 units/mg) 3 was obtained using this affinity purification step.
Effect of Delipoylation on PDP Activity and Binding to L2- Fig. 5a shows that delipoylation of 32 P-PDC on PDP activity leads to a marked reduction in PDP activity. Fig. 5b shows the associated loss in the reduction of lipoates in PDCb by E3 catalysis as the PDCb is delipoylated. The loss of E2-activated PDP activity was more pronounced as the last portion of lipoates was removed. Reintroduction of lipoates through lipoyl protein ligase catalysis to a level that gave two-thirds as high an acetylation capacity as the control PDCb (Fig. 6b) restored high PDP activity (Fig. 6a). The only change in SDS-PAGE patterns was a slight decrease in E2 mobility following delipoylation that was reversed on relipoylation. 4 This suggests that the lipoyl group of the L2 domain plays an essential role in PDP function.
To assess whether delipoylation led to a loss of PDP binding to the L2 domain following delipoylation, the effects of delipoylated L2 versus lipoylated L2 in inhibiting dephosphorylation of PDCb were compared. Fig. 7a shows that 13 M delipoylated L2 was somewhat less effective than 13 M L2 but still significantly inhibited PDP dephosphorylation of E1b in the intact complex. Fig. 7b compares the binding by 4 M lipoylated and delipoylated GST-H1⅐L2. Delipoylation reduced but did not prevent binding of the phosphatase. The marked reduction in the capacity of delipoylated E2 to activate PDP activity cannot be explained by reduced PDP binding. E1b binding was fully retained in delipoylated PDCb (data not shown). Thus, lipoyl groups of E2 must have a specialized role in facilitating high PDP activity (see "Discussion").

DISCUSSION
Hormones that signal cellular events that place increased energy demands often initiate signal transduction pathways that increase cellular Ca 2ϩ derived either from intracellular stores or from transient extracellular import. The rise in cytoplasmic Ca 2ϩ to a near micromolar level is coupled by specialized import-export transporters that generate a similar in- 3 We have observed specific activities as high as 1500 units/mg in freshly prepared PDP, but this specific activity is not maintained under storage conditions that we have found to date. 4 Unusual mobility decreases with removal of lipoates from lipoyl domain structures are readily observed for native gels, but mobility decreases have also been observed by SDS-PAGE analyses in some cases (28). The latter shifts probably result from a different cause than those occurring in native gel patterns as previously discussed (28). crease in free mitochondrial Ca 2ϩ (35,36). This signal then up-regulates NADH production within mitochondria through activating three mitochondrial dehydrogenases (36 -38). The ␣-ketoglutarate dehydrogenase complex and isocitrate dehydrogenase are activated by direct allosteric mechanisms in which the S 0.5 for their respective keto acid substrates are substantially decreased (36 -41), and in the case of ␣-ketoglutarate dehydrogenase, micromolar Ca 2ϩ significantly reduces the potency of NADH inhibition (42). Here, we have further investigated the indirect mechanism whereby Ca 2ϩ increases PDC activity. Although a marked decrease in the K m of PDP for E1b and a lowering of its K m for Mg 2ϩ (15,17,20) were known to ensue from Ca 2ϩ -dependent binding of PDP to E2, little was known about the specific site of PDP binding. It was reported that Ca 2ϩ does not enhance PDP dephosphorylation of the peptide substrate (43), suggesting that both PDP and E1b needed to be bound to the E2 60 core and that this juxtapositioning is the likely cause of the reduced K m of PDP for E1b.
We have demonstrated by inhibition and direct binding studies that the Ca 2ϩ -dependent association of PDP selectively occurs at the E2 inner lipoyl domain. Thus, this domain has a special role in supporting the regulation of PDC, since it also serves as the binding site for the kinase (22). The data for L2 inhibition of E2-activated PDP activity (Fig. 2) fit well when analyzed as a simple competition between E2 subunits and the L2 domain. Assuming that the competitive binding operates at equilibrium prior to PDP catalysis, the results suggest about 3-fold tighter binding to the free L2 domain than to E2 subunits, and extrapolation of the PDP activity to infinite L2 gives only 2% residual PDP activity. Yang et al., 5 using recombinant mature human E2, observed half-maximal activation of PDP activity by 0.8 -1.2 M E2 subunits. Using this range to approximate the equilibrium binding constant for PDP associating with E2, an analysis 6 of the fit in Fig. 2 that kinetically treats L2 as a competitive inhibitor of activation by E2 yields K i values for L2 in the range of 0.12-0.18 M.
The studies on direct binding GST-H1⅐L2 support a weaker equilibrium binding constant of ϳ1.75 M. We observed that the fusion protein was somewhat less effective than free L2 in inhibiting the dephosphorylation of PDCb, but GST-H1⅐L2 is no more than a 2-fold poorer inhibitor. The short time involved in capturing the GST-H1⅐L2⅐PDP complex on GSH-Sepharose may lead to some re-equilibration after immobilization on the gel, altering access of PDP to L2. Thus, the weaker binding of PDP to L2 than predicted from the above analysis of L2 inhibition may be partially explained by these factors but may also reflect limitations of that analysis. Nervertheless, the data support highly specific binding of L2 to PDP with a near micromolar binding constant. In vivo, more than just changes in free Ca 2ϩ may influence this association within the mitochondrial matrix, since a much more confined and organized state is likely (and suggested by toluene-permeabilized mitochondria 5 D. Yang, J. Song, T. W., and T. E. Roche, submitted for publication. 6 Since, using the theoretical fit for Fig. 2, extrapolation of PDP activity to infinite L2 gives ϳ2% PDP activity remaining, and since free PDP levels (not bound to L2 or E2) are very low, the data can be fit by an approximation that treats L2 as a competitive inhibitor of E2 subunits functioning as an essential activator. Then, at the x intercept in the Dixon plot, [L2] Ϸ ϪK I (1 ϩ A/K a [1 ϩ S/K as ]) where K I ϭ K d for the L2⅐PDP complex, [E2] ϭ A, K a is assumed to equal the subunit concentration for E2 found to give half-maximal activation of PDP (0.8 -1.2 M (Yang et al. 5 ), presumably equivalent to K d for the E2⅐PDP complex), S ϭ 2.2 M E1b, and K as ϭ K m of E2-activated PDP for E1b, which is about 3 M (15). Relipoylation of delipoylated PDCb, PDP assays (a), and acetylation assays (b) were conducted as described under "Experimental Procedures." not releasing PDP; Ref. 23). Furthermore, PDP may reversibly interact with sites other than PDC E2, particularly when the intramitochondrial free Ca 2ϩ concentration is low.
For the L2-PDP interaction to yield a large enhancement in PDP activity, we have found that the lipoyl domains of the assembled E2 core must retain their lipoyl moieties. Loss of PDP binding by L2 was a possible explanation for the loss of E2-activated PDP function. However, significant (albeit reduced) Ca 2ϩ -dependent inhibition and binding of PDP were still observed with delipoylated L2 structures. Therefore, the complete removal of E2 enhancement of PDP activity after delipoylating the E2 core implies that processes other than simply binding and concentrating PDP and E1b are facilitated by the E2 core. Since L1 did not affect PDP activity, it seems unlikely that the L1 lipoyl group contributes to E2-activated PDP activity. Yang and Roche 7 have recently found that a recombinant oligomeric E2 structure lacking L1 can activate PDP activity, establishing that the L2 lipoate can serve. An interesting and reasonable possibility is that the L2 lipoate somehow augments encounters of PDP with E1b at the surface of E2. A potential role of the lipoate would be to support efficient delivery of the phosphatase to E1b by the lipoyl cofactor of the PDP-carrying L2 interacting with E1b in a manner similar to the E1 lipoyl domain interaction occurring in the reductive acetylation reaction (step 2 catalyzed by E1 in the overall PDC reaction). This would require that PDP be tethered to L2 in a way that does not interfere with the highly specific L2-E1 interaction while simultaneously aiding the positioning of the active site of the P c subunit of PDP to efficiently dephosphorylate E1b.
The small enhancement in PDP activity observed when E1b is bound by E2 IB , which lacks the L2 binding site, implies a consequential effect of this binding, since localizing E1 on this assembled structure may actually reduce the frequency of diffusion-based encounters relative to free E1b interacting with free PDP. Binding of E1b to E2 may induce a conformational change in E1b that exposes its sites of phosphorylation or may restrict E1b positions in a way that increases the probability of productive encounters with a freely diffusing PDP.
Schemes involving anchoring the entire E2 core were developed for purification of PDP (16,18). However, these were expensive preparations and had a lower capacity for binding PDP (possibly because multiple cross-linking between the matrix and E2 60 aggregates interfered with PDP interactions). The use of GST-H1⅐L2 affords high recovery of essentially homogeneous PDP through selective elution by Ca 2ϩ chelation. GST-H1⅐L2 can be economically and rapidly prepared, and the GSH-Sepharose-bound protein can be repeatedly used.
Since Ca 2ϩ normally binds to proteins via carboxylate side chains, we find it particularly interesting that the COOHterminal region of the L2 domain, which constitutes extended structure not found in bacterial lipoyl domains, is enriched in acidic residues (Glu-209, Glu-211, Asp-213, Asp-219, Glu-224, and Asp-227). There are also three basic residues in the last 20 residues (209 -219) and limited hydrophobic residues (Ile-214, Phe-217, Tyr-220, Val-225, and Leu-228) to participate in anchoring this region of L2 by participating in an extended hydrophobic core. Of particular interest is the prospect that this carboxylate-rich region may contribute to binding of PDP by binding Ca 2ϩ . The presence of this region raises the prospect that Ca 2ϩ -PDP may interact with L2 and induce formation of a bridge structure involving a weakly held COOH-terminal segment of L2 becoming partly or fully disengaged from the rest of the L2 domain while binding PDP. This region may not be essential to forming a globular domain, since a lipoylated L2 structure in which this region was deleted served as a substrate for E3; however, it was a very poor E1 substrate and did not bind the kinase. 8 Our findings that the L2 domain of E2 serves as the specific binding site of PDP and that the L2 lipoyl group has a critical role in facilitating the large enhancement of PDP dephosphorylation of E2-bound E1b raises further questions about the nature of these interactions and the series of steps that complete the transduction of the Ca 2ϩ signal. Future studies will seek to answer these questions.