Interdependent Regulation of Insulin Receptor Kinase Activity by ADP and Hydrogen Peroxide*

Insulin signaling requires autophosphorylation of the insulin receptor kinase (IRK) domain. Using purified recombinant IRK fragments and the isolated intact insulin receptor, we show here that autophosphorylation is inhibited by ADP and that this effect is essentially reversed by hydrogen peroxide. Autophosphorylation was inhibited by hydrogen peroxide (60 (cid:1) M ) in the absence of ADP but enhanced in the presence of inhibitory concentrations of ADP (67 (cid:1) M ). Enhancement by hydrogen peroxide required direct interaction of hydrogen peroxide with the kinase domain and was not seen in insulin receptor mutants C1245A and C1308A. A similar enhancement was obtained in intact cells in the absence of insulin upon treatment with 1-(2-chloroethyl)-3-(2-hy-droxyethyl)-1-nitrosourea, indicating that IRK activity can be alternatively enhanced by a shift in the thiol/ disulfide redox status. Molecular modeling of the IRK domain indicated that the ATP-binding site becomes distorted after releasing the nucleotide unless the IRK domain is oxidatively derivatized at Cys 1245 . Recent clinical studies suggest that these effects may play a role in obesity due to the fact that cytoplasmic creatine kinase in combination with phosphocreatine


Insulin signaling requires autophosphorylation of the insulin receptor kinase (IRK) domain. Using purified recombinant IRK fragments and the isolated intact insulin receptor, we show here that autophosphorylation is inhibited by ADP and that this effect is essentially reversed by hydrogen peroxide. Autophosphorylation was inhibited by hydrogen peroxide (60 M) in the absence of ADP but enhanced in the presence of inhibitory concentrations of ADP (67 M). Enhancement by hydrogen peroxide required direct interaction of hydrogen peroxide with the kinase domain and was not seen in insulin receptor mutants C1245A and C1308A. A similar enhancement was obtained in intact cells in the absence of insulin upon treatment with 1-(2-chloroethyl)-3-(2-hydroxyethyl)-1-nitrosourea, indicating that IRK activity can be alternatively enhanced by a shift in the thiol/ disulfide redox status. Molecular modeling of the IRK domain indicated that the ATP-binding site becomes distorted after releasing the nucleotide unless the IRK domain is oxidatively derivatized at Cys 1245 . Recent clin-
ical studies suggest that these effects may play a role in obesity due to the fact that cytoplasmic creatine kinase in combination with phosphocreatine normally ensures rapid removal of ADP in muscle cells but not in fat cells.
Stimulation of insulin receptor kinase (IRK) 1 activity typically involves binding of insulin to the extracellular ␣-subunits of the insulin receptor, which leads to autophosphorylation of the intracellular ␤-subunits (1). The crystal structures of the non-phosphorylated IRK domain (IRK-0P; "gate-closed" conformation) and the trisphosphorylated IRK domain (IRK-3P; "gate-open" conformation) have been described (2,3). Dimerization of the IRK domain (4) or high concentrations of adenosine nucleotides (3,5,6) were found to favor the gate-open conformation even in the absence of autophosphorylation. The action of IRK can be down-regulated by a redox-sensitive protein-tyrosine phosphatase (7,8). Brief exposure of cells to hydrogen peroxide was shown to increase insulin receptor autophosphorylation. In some studies, this effect was shown to involve the oxidative inactivation of protein-tyrosine phosphatase activity (7)(8)(9)(10)(11)(12)(13). The physiological significance of the role of hydrogen peroxide was underscored by the finding that insulin stimulates the formation of superoxide radicals and their derivative hydrogen peroxide by NAD(P)H oxidase in adipocytes (7, 14 -17). An important role of hydrogen peroxide in insulin signaling was also suggested by the finding that mice overexpressing cellular glutathione peroxidase develop insulin resistance (18).
Earlier studies have shown that IRK activity is increased or suppressed by sulfhydryl-reactive agents such as iodoacetamide, maleimidobutyrylbiocytin, and N-ethylmaleimide (19 -22), suggesting that the IRK domain may contain one or more functionally important cysteine residues. Treatment of cells with the glutathione reductase inhibitor 1-(2-chloroethyl)-3-(2hydroxyethyl)-1-nitrosourea (BCNU), which is known to shift the intracellular thiol/disulfide redox status (23), causes a decrease in insulin receptor ␤-chain sulfhydryl groups (24). Using purified recombinant fragments of the IRK domain, we show here that hydrogen peroxide directly interacts with the IRK domain and enhances its autophosphorylation and kinase activity in the presence of ADP. In the absence of hydrogen peroxide, ADP was found to inhibit IRK activity in a dose-dependent way.
As creatine is utilized selectively by the brain and skeletal muscle tissues and mediates the rapid removal of ADP, i.e. a product of the kinase reaction, by cytoplasmic creatine kinase (25)(26)(27), the IRK activity of muscle tissues may normally have a potentially important advantage compared with the IRK activity of adipocytes. The experiments in this study show, however, that the inhibition of IRK activity by its product ADP is subject to redox regulation.

MATERIALS AND METHODS
Cells, Culture Conditions, and Oxidative Activation-Chinese hamster ovary (CHO) cells stably transfected with the human insulin receptor (IR; CHO-HIR cells) (28) were cultured in nutrient mixture F-12 (Invitrogen) with 10% fetal calf serum. Confluent cells were routinely used, as endogenous reactive oxygen species production is lower at higher cell density (29). One day before the experiment, the cells were cultured under serum-free conditions as described (30). The glutathione reductase inhibitor BCNU (80 M; Bristol-Myers Squibb Co., Munich, Germany) and hydrogen peroxide (50 M; Merck, Darmstadt, Germany) were added to the cells 2 h and 30 min, respectively, before harvest.
Sciences) in 15 l of kinase buffer. The reaction was stopped by boiling in Laemmli sample buffer, and the proteins were separated by reducing SDS-PAGE (7% gels) in glycine/Tris buffer (pH 8.3), blotted onto nitrocellulose membranes, and finally subjected to autoradiography. Subsequently, tyrosine phosphorylation was detected by Western blotting with anti-phosphotyrosine antibody 4G10 (BIOMOL Research Labs Inc., Hamburg, Germany). IR expression was determined with anti-␤-IR polyclonal antibody (Santa Cruz Biotechnology, Heidelberg, Germany) as described (24).
Transient Transfections-CHO cells were transiently transfected with pECE vectors containing the human wild-type IR or its mutants C1056A, C1234A, C1245A, and C1308A (kindly provided by Dr. C. W. Ward, CSIRO Health Sciences and Nutrition, Parkville, Australia). Transfections were performed with 1.5 g of DNA using the Polyfect reagent (QIAGEN Inc., Hilden, Germany) according to the manufacturer's instructions. After incubation for 30 h, cells were incubated for 17 h in serum-free nutrient mixture F-12 and finally incubated with or without 60 M hydrogen peroxide for 30 min. Cells were lysed, and the resulting immunoprecipitates were processed as described above.
Purified Recombinant IRK Preparations-The IRK-D fragment containing Val 978 -Lys 1283 of the human IR with mutations C981S and Y984F (35 kDa) was prepared by constructing a baculovirus transfer vector (pVL1393). The sequence was verified by automated sequencing. Cotransfection of Sf9 cells and isolation of recombinant baculoviruses were performed as described (4). Sf9 cells were lysed in 20 mM Tris-HCl(pH 7.5), 250 mM sucrose, 1 mM dithiothreitol, and Complete TM by sonication, and the recombinant protein was purified by sequential chromatography on Q (6-ml column; Amersham Biosciences), Superdex 75 (Amersham Biosciences), and Mono Q (1-ml column; Amersham Biosciences) and finally shock-frozen in liquid nitrogen and stored at Ϫ74°C. In addition, the entire cytosolic domain of IRK was expressed as a glutathione S-transferase (GST) fusion protein (GST-IRK, 71 kDa) in insect cells and purified to homogeneity by affinity chromatography as described previously (4).
Aliquots containing 0.4 g of GST-IRK or IRK-D were typically mixed with kinase buffer containing 1 mg/ml bovine serum albumin, 10 g/ml myelin basic protein (MBP), 1 mM cysteine, and 1 mM glutathione with or without 60 M hydrogen peroxide (final concentrations). Unless indicated otherwise, the samples were incubated for 15 min at 30°C, mixed with ATP substrate solution (0.4 mM ATP and 0.1 Ci/l [␥-32 P]ATP) in a final volume of 15 l/sample; and subjected to kinase assay at 30°C with shaking.
Native Gel Kinase Assay-Aliquots of 0.75 g of IRK-D were mixed with 1 g of MBP in 6 l of buffer containing 50 mM HEPES (pH 7.3), 0.1 mM dithiothreitol, 3 mM MgCl 2 , and 200 mM NaCl. After adding 1 l of 0.42 mM hydrogen peroxide (final concentration of 60 M), the sample was incubated for 10 min at room temperature and then mixed with 1 l of 20 mM ATP (final concentration of 2.5 mM). Aliquots of 4 l were incubated for another 5 or 30 min, subsequently mixed with 1.25 l of stop buffer (20 mM EDTA, 50 mM Tris (pH 7.5), and 40% glycerol), immediately frozen, and eventually applied to an 8% nondenaturing polyacrylamide gel containing 0.375 M Tris (pH 8.8). Finally, the samples were run for 60 min at 145 V in Tris/glycine buffer (pH 8.5) and blotted onto nitrocellulose membranes using a semidry blotting apparatus. The membranes were incubated overnight with anti-phosphotyrosine antibody 4G10 and treated with a chemiluminescent reagent (PerkinElmer Life Sciences). Subsequently, the membranes were washed with hot water and incubated with antibodies against phosphorylated Tyr 1158 of IRK (BIOSOURCE, Nivelles, Belgium). After incubation with alkaline phosphatase-coupled anti-rabbit IgG, the resulting bands were detected using a chemiluminescent alkaline phosphatase substrate (Bio-Rad, Munich).
Three-dimensional Modeling of the Phosphorylated IR ␤-Chain and of Derivatized Proteins-The three-dimensional structures of the IRK-3P domain and its derivatives were modeled on an Indigo II workstation (Silicon Graphics). The starting coordinates of non-hydrogen atoms were taken from Protein Data Bank entry 1ir3 (3). Selected atoms were removed; atoms not reported in the Protein Data Bank were inserted with the help of the INSIGHT II library; and hydrogen atoms were added. Energy minimization was performed using the DISCOVER module of the INSIGHT II program (Accelrys Inc.) as described (31). The figures of the minimized structures were obtained by RasMol.

Effect of Hydrogen Peroxide on Purified Recombinant
Fragments of the IRK Domain-To determine whether hydrogen peroxide may directly act upon the IR, we used purified recom-binant ␤-chain fragments. Two constructs were prepared, a GST-tagged fragment comprising the entire cytosolic domain of the ␤-chain (GST-IRK, 71 kDa) and a non-tagged fragment comprising the kinase domain from Val 978 to Lys 1283 (IRK-D, 35 kDa). The purified proteins were first incubated with or without 60 M hydrogen peroxide and with 1 mM ATP and then subjected to kinase assay with [ 32 P]ATP. The results show that hydrogen peroxide treatment enhanced the incorporation of 32 P (Fig. 1A). To further characterize the effect, we separated the different phosphorylated forms of IRK-D on native polyacrylamide gels on the basis of charge (32). The proteins were then transferred to nitrocellulose and detected by phospho-specific antibodies. All three phosphorylated forms of IRK-D were found to be markedly enhanced by hydrogen peroxide if tested after 30 min of incubation (Fig. 1B). The non-phosphorylated state (IRK-0P) reacted only weakly with both antibodies but accounted for 90% of the IRK-D fragment (data not shown).
Effect of Hydrogen Peroxide in Combination with ADP-If GST-IRK was incubated with hydrogen peroxide in the absence of ATP and then subjected to kinase assay with 0.4 mM ATP plus [ 32 P]ATP for different periods of time, hydrogen peroxide was found to enhance IRK activity only after 16 -32 min (data not shown). To determine whether this relatively late effect was mechanistically related to the endogenous production of ADP, we added graded concentrations of ADP at the start of the kinase reaction and determined the phosphorylation at an early time point (12 min). The results show that hydrogen peroxide indeed caused a significant increase in MBP substrate phosphorylation and IRK autophosphorylation at this early time point in the presence of 67 M ADP (Fig. 2, B and E). In the absence of ADP, hydrogen peroxide had practically no effect on MBP phosphorylation and even suppressed IRK autophosphorylation (Fig. 2E). As the relative enhancement of autophosphorylation at 22.5 M ADP was not significantly smaller than that at 7.5 M ADP (Fig. 2E), the difference between these two points must not be over-interpreted. In the absence of hydrogen peroxide, MBP phosphorylation and IRK autophosphorylation were inhibited by ADP in a dose-dependent manner (Fig. 2, C and F), implying that the relative enhancement by hydrogen peroxide had essentially the effect of reversing the inhibitory effect of ADP.
To determine whether hydrogen peroxide may alter the affinity for ATP, we analyzed the rate of autophosphorylation of GST-IRK during the first 6 min under conditions of graded ATP concentrations in the absence of ADP (data not shown). This analysis revealed no substantial change in K m(ATP) (221.1 versus 198.7 M).
Effect of Hydrogen Peroxide in Combination with the ATP Analog AMP-PNP-Like ADP, AMP-PNP facilitates the conversion of the gate-closed conformation into the gate-open conformation (3,5,6). But AMP-PNP is larger than ADP and, unlike ATP, is non-hydrolyzable. We therefore incubated GST-IRK with or without 60 M hydrogen peroxide for 15 min and with 2 mM AMP-PNP for another 30 min, precipitated the enzyme subsequently with anti-GST antibody-agarose, and de-termined the autophosphorylation activity using [␥-32 P]ATP and 0.2 mM ATP. The results show that, under these conditions, hydrogen peroxide enhanced IRK activity, on the average, by as much as 8.59-fold (p Ͻ 0.01) (data not shown).
IRK Activity of Hydrogen Peroxide-treated Cells in the Absence and Presence of Insulin-To study the effect of hydrogen peroxide on the IR in intact cells, CHO-HIR cells (28) were incubated with or without 50 M hydrogen peroxide for 30 min prior to harvest. The receptor was immunoprecipitated and then preincubated for 30 min with 0.2 mM unlabeled ATP with or without 5 mM AMP-PNP prior to kinase assay with 25 M ATP plus 5 Ci of [␥-32 P]ATP. As the kinase reaction was performed with 25 M ATP, the endogenous formation of ADP was relatively small in these experiments. If tested without preincubation, i.e. in the absence of AMP-PNP and ADP, receptor preparations from hydrogen peroxide-treated cells were not detectably different from control samples (data not shown). In contrast, after preincubation, kinase activity was strongly enhanced by hydrogen peroxide (Fig. 3, A-C). The strongest enhancement was again seen after preincubation with AMP-PNP. A similar enhancing effect of hydrogen peroxide was observed in insulin-stimulated cells (Fig. 3D).
In view of its structural similarity to ADP, AMP-PNP was also expected to compete for the ATP-binding site and thereby to inhibit kinase activity in the absence of hydrogen peroxide. To ensure that this was indeed the case under the conditions of our experiments, we compared kinase activities with and without  (Fig. 4). However, if the kinase reaction was performed without the addition of unlabeled ATP, i.e. with only a minute amount of radioactively labeled ATP, IR autophosphorylation was enhanced by AMP-PNP (Fig. 4A). This indicates that, under conditions of low nucleotide concentrations, the enhancing effect of AMP-PNP prevails over its inhibitory effect, in line with the notion that the optimal conversion of the gate-closed conformation into the gate-open conformation requires relatively high nucleotide concentrations.
Effect of Hydrogen Peroxide and ADP on the Immunoprecipitated Intact Insulin Receptor-If the immunoprecipitated receptor was treated with or without hydrogen peroxide and then directly subjected to kinase assay in the presence of graded concentrations of ADP for 15 min, it was again found that IRK activity was inhibited by ADP (67 M) in the absence of hydrogen peroxide (Fig. 3E) and that hydrogen peroxide enhanced IRK activity in the presence of ADP but not in its absence (Fig. 3F).
Alternative Mode of Redox Regulation of IRK Activity by BCNU-As several other signaling processes have been shown to be induced alternatively by hydrogen peroxide or by an oxidative shift in the thiol/disulfide redox status (reviewed in Refs. 33 and 34), we also studied the effect of the intracellular redox status on IRK activation. To induce an oxidative shift in the intracellular thiol/disulfide redox status, the cells were treated with the glutathione reductase inhibitor BCNU (see Ref. 23). The results show that BCNU treatment mediated an enhancing effect similar to that of hydrogen peroxide (Fig. 3,  A-C).
Effect of Hydrogen Peroxide on Mutant Insulin Receptor Proteins-To determine whether the enhancing effect of hydrogen peroxide may involve a cysteine residue of the IR ␤-chain, we performed experiments with CHO cells transiently transfected with DNA from mutant proteins. Cells were treated with or without 60 M hydrogen peroxide for 30 min prior to harvest and analyzed as described above. As mutant C1138A was previously found to be functionally defective (30), this mutant was not included in this study. The results show that hydrogen peroxide enhanced the relative autophosphorylation of the wild-type receptor and mutant C1056A (Fig. 5). The enhancing effect was statistically not significant with mutant C1234A, and it was completely abrogated with mutant C1245A. Autophosphorylation of C1308A was inhibited by hydrogen peroxide.
Molecular Modeling of Oxidized Forms of the IRK Domain-As crystal structures of selectively oxidized IRK proteins are not available, we used mass spectrometric analysis of hydrogen peroxide-treated IRK samples to characterize the structural modifications. However, in line with negative results of others (35), the critical cysteine-containing peptides could not be detected in the spectrum. We therefore performed molecular modeling studies. As hydrogen peroxide treatment typically converts cysteine into sulfenic acid residues, which spontaneously interact with glutathione to form mixed proteinglutathione disulfides (reviewed in 33,34), and because our experiments with hydrogen peroxide were routinely performed in the presence of glutathione, we performed molecular modeling with glutathionylated IRK structures. The limited accessibility of Cys 1056 , Cys 1234 , and Cys 1245 on the surface of the IRK domain practically excludes disulfide formation with another IRK domain. Three-dimensional models of the gate-open conformation of the IR ␤-chain were derived from Protein Data Bank entry 1ir3 (3). Certain atoms were removed, and other atoms not reported in entry 1ir3 were inserted with the help of the INSIGHT II library. Ser 981 and Phe 984 were replaced with the authentic Cys 981 and Tyr 984 , respectively, and AMP-PNP was replaced with ATP before energy minimization.
The resulting structure (Fig. 6A) (Fig. 6C) or Cys 1234 (Fig. 6D) prior to removal of ATP and peptide substrate, subsequent energy minimization yielded a structure in which all relevant contact points were arranged in a way similar to the original ATPloaded structure (Fig. 6A). The distances between the nitrogen of Gly 1005 and the nearest O-␦ of Asp 1150 were 7.11 and 7.55 Å, respectively, and the angles between Asn 1137 O-␦, Asp 1150 O-␦1, and Asp 1150 O-␦2 were 158°and 164°, respectively. The spacefilling models in the right panels show that the position of Gly 1005 relative to its neighbors is very similar in panels A, C,  amino acids 1002, 1003, 1004, 1005, 1006, 1082, 1083, 1136, 1170, 1171, and 1172 are stained green, except for the nitrogen and C-␣ atoms of Gly 1005 , which are stained cyan. Water molecules in the vicinity have been removed. B: shown is the structure obtained from IRK-3P ϩ ATP by energy minimization after removal of the two Mg 2ϩ atoms, the nucleotide, and the peptide. C and D: shown are the structures obtained by energy minimization after removal of the two Mg 2ϩ atoms, the nucleotide, and the peptide and after insertion of an S-linked glutathione (SG) at Cys 1245 or Cys 1234 , respectively. and D but markedly different in panel B. The structures glutathionylated at Cys 981 or Cys 1056 (data not shown) were similar to unmodified IRK-3P without ATP (Fig. 6B). The glutathionylated structures IRK-3P C1245S-SG and IRK-3P C1234S-SG with ATP and substrate (data not shown) were essentially indistinguishable from the structure in Fig. 6A. DISCUSSION Our experiments with the intact IR and purified GST-IRK fragment have shown that IRK activity was inhibited by ADP (67 M) in the absence of hydrogen peroxide. In the presence of inhibitory concentrations of ADP or the related nucleotide AMP-PNP, IRK activity was enhanced by hydrogen peroxide (60 M). This effect essentially reversed the inhibitory effect of ADP. As the enhancing effect requires direct interaction of hydrogen peroxide with the IRK domain, it is a novel mechanism of oxidative enhancement that may complement the enhancement of IR autophosphorylation by oxidative inhibition of protein-tyrosine phosphatase activity that has been described previously (7)(8)(9)(10)(11)(12).
The enhancing effect of hydrogen peroxide in the presence of ADP is tentatively explained by a mechanism involving oxidative modification of a critical cysteine residue. As mutations at Cys 1245 or Cys 1308 abrogated the enhancing effect of hydrogen peroxide (Fig. 5), we tentatively assumed that these cysteine residues may be involved. It is reasonable to assume that hydrogen peroxide causes the formation of a mixed disulfide between a certain cysteine residue of the IRK protein and a small molecular mass thiol compound in the vicinity. When intact cells were treated with hydrogen peroxide as in Fig. 3, the oxidized cytoplasmic IRK domain was inevitably exposed to millimolar concentrations of glutathione as the quantitatively most important intracellular thiol compound. Experiments with purified IRK fragments were therefore routinely performed in the presence of glutathione. Control experiments showed that hydrogen peroxide enhanced kinase activity even after omission of glutathione and cysteine (data not shown); but as IRK fragments were typically stabilized with dithiothreitol, the kinase reaction mixture contained, in this case, this small molecular mass compound at ϳ0.1 mM. We tentatively assume that the resulting mixed protein-dithiothreitol disulfide may have a structure equivalent to those of the glutathionylated IRK structures modeled in Fig. 6. It has previously been shown for other proteins that hydrogen peroxide converts certain cysteine residues into sulfenic acid residues, which spontaneously interact with glutathione to form mixed proteinglutathione disulfides (reviewed in Refs. 38 and 39). Similar mechanisms were shown to operate in other models of redox regulation. Molecular modeling studies indicated that removal of the nucleotide and substrate from the original IRK-3P structure (Fig. 6A) leads to a structural change (Fig. 6B) that compromises the accessibility of the binding site and the coordinated incorporation of the magnesium atoms. In contrast, the relevant contact points in the glutathionylated structures IRK-3P C1245S-SG (Fig. 6C) and IRK-3P C1234S-SG (Fig. 6D) were found to have distances and angles similar to those of the original ATP-loaded structure. This finding provides a satisfactory explanation for the positive effect of hydrogen peroxide in the presence of ADP or AMP-PNP. Whereas the inhibition of IRK activity by ADP in the absence of hydrogen peroxide is best explained by the competition for the ATP-binding site, ADP and AMP-PNP may enhance IRK activity in the presence of hydrogen peroxide by facilitating the conversion of the gateclosed conformation into the gate-open conformation (3,5,6). This positive effect would require, however, that the binding site remain essentially unaltered after the release of these nonproductive ATP analogs, i.e. before the next round of (pro-ductive) nucleotide binding. This is essentially the case for IRK-3P Cys1245S-SG and IRK-3P Cys1234S-SG but not for the unmodified structure or the structures IRK-3P Cys981S-SG and IRK-3P Cys1056S-SG. The effect of hydrogen peroxide on mutant C1245A was more severely compromised than that on mutant C1234A (Fig. 5), but this difference was not statistically significant and may be explained by the fact that Cys 1245 is more easily accessible for glutathionylation than Cys 1234 (data not shown). Equivalents of Cys 1245 in IRK such as Cys 475 in Lck, Cys 498 in v-Src, and Cys 376 in Ret-PTC-1 were shown previously to play an important role in the regulation of the catalytic or transforming activity of these kinase species (36 -38). The data of Fig. 5 also reveal that mutant C1308A was inhibited by hydrogen peroxide rather than enhanced. The mechanism of this effect is not known but is tentatively explained by the assumption that, after hydrogen peroxide treatment, Cys 1308 may also form a functionally important disulfide bridge. One possibility is that Cys 1308 may form a disulfide bridge with Cys 1245 and that this may be functionally equivalent or even superior to the glutathionylated structure. As Protein Data Bank entry 1ir3 contains amino acids up to position 1283, we were not able to model this intramolecular disulfide structure. Cys 1308 is located in the C-terminal tail of the ␤-chain.
These mechanistic models were further supported by the substantial enhancement of IRK activity by BCNU, an agent known to cause an oxidative intracellular shift in the glutathione redox status in various cell types (see Ref. 23). The enhancing effect of the glutathione reductase inhibitor BCNU on IRK activity (Fig. 3) is taken as an indication for an alternative mode of IRK activation by an oxidative shift in the glutathione redox status. This alternative mechanism may be pathophysiologically relevant in view of the age-related changes in the intracellular glutathione redox status that have been found in human and rat skeletal muscle tissues (39,40) and in the livers, kidneys, and brains of rats and mice (41)(42)(43). An ageand disease-related oxidative shift in the plasma thiol/disulfide redox status has also been observed in the blood plasma (44,45). An independent study of mice has shown that the plasma thiol/disulfide redox status directly correlates with the muscular glutathione redox status (46). Molecular modeling of double-glutathionylated IRK proteins, i.e. structures glutathionylated at Cys 1245 plus Cys 1056 , yielded structures with impaired accessibility of the nucleotide-binding site (data not shown). This effect is therefore one possibility to explain the observed hydrogen peroxide-mediated inhibition of kinase activity. However, this possibility is highly speculative, as one can think of various other mechanisms that potentially explain the oxidative destruction of the catalytic activity by hydrogen peroxide.
The physiological significance of the ADP effect is related to the fact that, in muscle tissue, high phosphocreatine concentrations in combination with cytoplasmic creatine kinase normally ensure rapid conversion of ADP into ATP (see the Introduction), whereas in adipose tissue, conversion of ADP into ATP depends on the relatively slow diffusion of ADP to the mitochondrial matrix. The cytoplasmic ADP concentration of adipose tissue is therefore expected to be markedly higher than that of resting muscle cells, which typically contain 13 M ADP, 4 mM ATP, and 25 mM phosphocreatine (47). Accordingly, the insulin reactivity of fat cells is expected to be weaker than that of skeletal muscle cells unless the ADP-mediated inhibition is ameliorated by an oxidative shift in redox status. Even relatively small changes in absolute autophosphorylation rates may significantly alter the delicate balance between muscle and adipose tissues with respect to insulin responsiveness and fuel deposition. This conclusion is supported by a recent clinical study of non-diabetic obese subjects (48) showing that (i) obesity is associated with abnormally low plasma thiol (mainly cysteine) levels; (ii) insulin reactivity is decreased in patients supplemented with N-acetylcysteine; and (iii) glucose clearance, i.e. the bona fide indicator of the insulin reactivity of muscle cells, is increased by simultaneous treatment with creatine if the patients are treated with N-acetylcysteine but not with placebo. Moreover, the mean HOMA-R index, which is defined as (fasting plasma glucose (mg/dl) ϫ fasting plasma insulin concentration (milliunits/ml)) ϫ 405 Ϫ1 and which is therefore an inverse indicator of insulin reactivity in the postabsorptive period, was found to be increased in the group with N-acetylcysteine plus placebo but decreased in the group with N-acetylcysteine plus creatine (48). N-Acetylcysteine treatment with or without creatine supplementation or treatment with a cysteine-rich protein has been shown to cause a decrease in body fat (48 -50), suggesting that the redox regulation of IRK activity may contribute to the development of obesity.
The finding that hydrogen peroxide or an oxidative shift in the thiol/disulfide redox status increases IRK autophosphorylation in the absence of insulin further suggests that the insulinindependent IR "basal activity" may increase in old age and other conditions associated with an oxidative shift in the plasma thiol/disulfide redox status (44,48). A decrease in IR signaling was recently found to increase longevity and stress resistance in various species (reviewed in Ref. 51). Mice with fat-specific IR knockout also show increased resistance to obesity (52). Importantly, the increased longevity of Caenorhabditis elegans mutants with defective DAF-2 protein, i.e. an IR analog, was shown to require an increase in autophagic activity (53). As autophagy is suppressed by the IR signaling cascade and therefore restricted to the fasted state (reviewed in Ref. 54), increased longevity appears to be the consequence of decreased IRK basal activity in the absence of insulin.