IRS-1 Mediates Inhibition of Ca 2 (cid:1) Mobilization by Insulin via the Inhibitory G-protein G i *

Platelet agonists initiate aggregation and secretion by activating receptors coupled to the G-protein G q , thereby raising cytosolic Ca 2 (cid:1) , [Ca 2 (cid:1) ] i . The rise in [Ca 2 (cid:1) ] i is facilitated via inhibition of cAMP formation by the inhibitory G-protein of adenylyl cyclase, G i . Since insulin attenuates platelet activation, we investigated whether insulin interferes with cAMP regulation. Here we report that insulin (0.5–200 nmol/liter) interferes with agonist-induced increases in [Ca 2 (cid:1) ] i (ADP, throm-bin), cAMP suppression (thrombin), and aggregation (ADP). The effects of insulin are as follows: (i) independ-ent of the P2Y 12 receptor, which mediates ADP-induced cAMP lowering; (ii) not observed during G s -mediated cAMP formation; (iii) unaffected by treatments that af-fect phosphodiesterases (3-isobutyl-1-methylxanthine); and (iv) not changed by interfering with NO-mediated regulation of cAMP degradation ( N G -monomethyl- L -ar-ginine).

tients with diabetes mellitus type I and II have platelets that show increased adhesion, aggregation, thromboxane production, and P-selectin expression (3). The hyperactivity might be caused by the absence of insulin inhibition, since intensive insulin treatment in diabetic patients reduced platelet aggregation (4).
The insulin receptor is a heterotetrameric transmembrane glycoprotein composed of two extracellular ␣ subunits (135 kDa each) and two transmembrane ␤ subunits (95 kDa each) that function as allosteric enzymes in which the ␣ subunit inhibits the tyrosine kinase activity of the ␤ subunit. Insulin binding to the ␣ subunit relieves the inhibition of the kinase activity in the ␤ subunit leading to autophosphorylation of the ␤ subunits and a conformational change that further increases the kinase activity. The insulin receptor tyrosine kinase phosphorylates proteins such as Shc and the insulin receptor substrates IRS-1 (165-185 kDa) and IRS-2 (180 -190 kDa). IRS-1 and IRS-2 have a highly conserved amino terminus, which contains a pleckstrin homology domain, a phosphotyrosine binding domain, and a carboxyl terminus with several tyrosine phosphorylation sites. IRS-1 and IRS-2 are complementary and act as "docking sites" to several Src homology 2 domains containing proteins, such as the regulatory subunits of phosphatidylinositol 3-kinase (PI3K) 1 (5).
GTP-binding proteins (G-proteins) can also act as signal transducers for the insulin receptor. G-proteins are guanine nucleotide-binding regulatory proteins that function as molecular switches between a GTP-bound "on state" and a GDPbound "off state." These proteins amplify, transmit, and integrate signals. The major G-proteins involved in platelet aggregation and secretion are G q ␣, which mediates increases in cytosolic Ca 2ϩ concentration, [Ca 2ϩ ] i , and G i ␣, which inhibits adenylyl cyclase thereby suppressing cAMP that is an inhibitor of platelets (6). Receptors that couple to G-proteins are generally seven-transmembrane proteins, but there are important exceptions. The insulin-like growth factor II receptor has a single transmembrane domain and couples directly to G i2 in a manner similar to that of conventional G-protein-coupled receptors (7). Studies have been reported suggesting that the insulin receptor binds G i ␣ 2 (8,9).
The insulin receptor is present on muscle, liver, and adipose tissue but also on endothelial cells, lymphocytes, erythrocytes, and platelets. A human platelet contains ϳ570 insulin receptors (10). Insulin binding induces phosphorylation of the ␤ subunits (11,12), demonstrating that the receptor is functional. In insulin-resistant patients, platelet inhibition by insulin is attenuated or absent (13,14), suggesting that in healthy individuals insulin signals to mechanisms that suppress platelet functions.
Many platelet agonists initiate aggregation and secretion of granule contents via G q which induces Ca 2ϩ release from intracellular storage sites and an increase in [Ca 2ϩ ] i . A rise in [Ca 2ϩ ] i is a key step in platelet activation. It triggers granule FIG. 1. Insulin inhibits ADP-induced calcium mobilization. Fura 2-AM-loaded platelets were incubated with and without insulin and stimulated with 10 mol/liter ADP at 20°C. A, curve a represents ADP-induced Ca 2ϩ mobilization in the absence of insulin. Curve b represents the Ca 2ϩ mobilization after preincubation with 1 nmol/liter insulin for 5 min. Insulin alone did not change the basal [Ca 2ϩ ] i (19.4 Ϯ 2.7 nmol/liter). B, platelets were incubated with 1 nmol/liter insulin for 0 -15 min prior to stimulation with ADP. The ADP-induced Ca 2ϩ mobilization (218.0 Ϯ 72.8 nmol/liter) was expressed as 100% and taken as control. Insulin inhibited the ADP-induced Ca 2ϩ mobilization optimally when incubated for 5 min (25 Ϯ 6%, p Ͻ 0.001). C, platelets were incubated for 5 min with different concentrations of insulin prior to stimulation with ADP. Insulin inhibited dose-dependently the ADP-induced Ca 2ϩ response at 0.5 nmol/liter and more (p Ͻ 0.01). Data (means Ϯ S.D., n ϭ 5) are expressed as percentages of control. The asterisk indicates a significant difference compared with controls (p Ͻ 0.05). secretion thereby releasing ADP, activates the fibrinogen receptor (integrin ␣ IIb ␤ 3 ) forming aggregates, and induces a procoagulant surface that facilitates the formation of thrombin. One of the mechanisms that suppress platelet activation is an increase in cAMP. The inhibition is mediated by the cAMP-dependent protein kinase A. This kinase interferes with multiple steps in platelet activation cascades, such as receptor-ligand interaction, the activity of G-proteins, and the activation of phospholipase C␤, protein kinase C, and mitogen-activated protein kinases. Protein kinase A also interferes with the elevation of [Ca 2ϩ ] i and inhibits actin-binding protein and caldesmon that are involved in cytoskeletal reorganization (15). Because a small rise in cAMP already leads to a strong activation of protein kinase A, platelet-activating sequences are extremely sensitive to increases in cAMP (16,17).
Optimal platelet functions require maximal stimulation of the G q pathway and the concomitant inhibition of cAMP production. cAMP is formed from ATP through the action of adenylyl cyclase and is subsequently metabolized by phosphodiesterases (PDEs). Adenylyl cyclase is inhibited by G i , which makes this G-protein a key factor in the control of cAMP formation. G i is activated either by direct interaction with the agonist receptor or via secreted granule ADP, which activates the P2Y 12 receptor via an extracellular feedback loop (18 -21). Since G i is involved in insulin signaling, we addressed the question whether insulin inhibits platelet functions by interfering with the activity of G i .
Immunoprecipitation-Washed platelets were incubated at 20°C, and samples were collected in 10ϫ lysis buffer containing 1% w/v SDS, 5% w/v n-octyl glucoside, 0.5 M EDTA, and 10% v/v Nonidet P-40, with 10% v/v protease inhibitor mixture and 1 mmol/liter Na 3 VO 4 . Detergent-insoluble material was sedimented by centrifugation for 1 min at FIG. 2. Effect of ADP receptor blockage on inhibition by insulin. Fura 2-AM-loaded platelets were incubated for 5 min with and without 100 nmol/liter insulin prior to stimulation with 10 mol/liter ADP in the absence and presence of 50 nmol/liter AR-C69931MX, an inhibitor of the P2Y 12 receptor, at 20°C. The ADP-induced Ca 2ϩ mobilization was expressed as 100%. AR-C69931MX decreased the Ca 2ϩ mobilization to 68 Ϯ 13% (n ϭ 3, p Ͻ 0.001) which is in the range of inhibition by insulin. A combination of insulin and AR-C69931MX did not induce further inhibition. Further details are in Fig. 1.
Immunoblotting-Washed platelets were incubated at 20°C, and samples were collected in 3ϫ sample buffer. Aliquots were subjected to SDS-PAGE on 5% gels. Proteins were transferred from the gel to nitrocellulose sheets and blocked in either 5% Protifar, 0.1% TBST or 4% FIG. 3. Insulin inhibits thrombin-induced calcium mobilization. Fura 2-AM-loaded platelets were incubated with and without insulin prior to stimulation with 0.25 units/ml thrombin at 20°C. A, curve a represents thrombin-induced Ca 2ϩ mobilization. Curve b represents Ca 2ϩ mobilization after preincubation with 100 nmol/liter insulin for 10 min. B, platelets were treated with 100 nmol/liter insulin for 0 -20 min. The Ca 2ϩ response by thrombin (592.6 Ϯ 69.4 nmol/liter) was expressed as 100%. Insulin optimally inhibited the thrombin-induced Ca 2ϩ response after 10 min of preincubation (25 Ϯ 8%, p Ͻ 0.001). C, platelets were incubated for 10 min with different concentrations of insulin prior to stimulation with thrombin. Insulin dose-dependently decreased the thrombin-induced Ca 2ϩ mobilization at 100 nmol/liter and more (p Ͻ 0.001). Further details are in Fig. 1.
PY-BSA, 0.1% TBST. Blots were probed separately with a primary antibody (4G10 anti-phosphotyrosine, IRS-1 (C-20), G i ␣ 2 (T-19), G z ␣ (I-20), G s ␣ or G q ␣ antibodies for immunoprecipitations and anti-insulin receptor phospho-Tyr 1158 for whole cell lysates) according to the recommendations of the manufacturer. The proteins were detected by enhanced chemiluminescence with horseradish peroxidase-labeled secondary antibodies (GAMPO for 4G10 anti-phosphotyrosine and anti-␣-rabbit horseradish peroxidase for all other antibodies, respectively). As control for lane loading, the blots were stripped by a 30-min incubation in 0.1% TBST with 2% SDS at 80 o C. After extensive washing, the blots were subjected to the same procedure as described above. Since the anti-insulin receptor phospho-Tyr 1158 antibodies recognize an aspecific band at 250 kDa, this band was taken as a control for lane loading. The intensity of the bands was quantitated with ImageQuant software.
Measurement of Platelet Aggregation-PRP was prepared as described above, and the platelet concentration was adjusted to 2.0 ϫ 10 11 cells/liter with platelet-poor plasma. Aliquots of 0.5 ml were warmed to 37°C for 5 min, followed by stimulation. Platelet aggregation was monitored continuously for 7 min under 900 rpm in an aggregometer (model 570 VS, Chrono-Log Corp., Havertown, PA).
Statistical Analysis-Statistical analysis was performed using oneway analysis of variance with Tukey's multiple comparisons test as post-test for repeated measurements unless mentioned otherwise. Results are expressed as means Ϯ S.D. of n observations. Differences were considered to be significant at p Ͻ 0.05.

Insulin Inhibits Agonist-induced Ca 2ϩ
Mobilization-Fura 2-AM-loaded platelets were incubated at 20°C with 1 nmol/ liter insulin and stimulated with 10 mol/liter ADP. Insulin alone did not change the basal [Ca 2ϩ ] i (19.4 Ϯ 2.7 nmol/liter), but a preincubation with insulin led to a 25 Ϯ 6% reduction of the ADP-induced Ca 2ϩ mobilization (Fig. 1A). Studies with different incubation times showed that optimal inhibition was observed after 5 min of preincubation. Simultaneous addition of insulin and ADP or incubation times of 15 min or longer failed to reveal the inhibition by insulin (Fig. 1B). To investigate the threshold above which insulin inhibited Ca 2ϩ mobilization, platelets were incubated for 5 min with different concentrations of insulin followed by stimulation with ADP. There was a dose-dependent increase in the inhibition by insulin, which became significant at 0.5 nmol/liter insulin and more ( Fig. 1C).
ADP activates platelets via the P2Y 1 receptor, which is coupled to G q and signals to [Ca 2ϩ ] i , while concurrently suppressing cAMP formation via the P2Y 12 receptor and G i (18). As shown in Fig. 2, the P2Y 12 antagonist AR-C69931MX induced the same degree of inhibition of ADP-induced Ca 2ϩ mobilization as insulin (100 nmol/liter). A combination of AR-C69931MX and insulin did not inhibit Ca 2ϩ signaling stronger than each of these factors alone. Thus, insulin appears to inhibit ADP-induced Ca 2ϩ mobilization by interfering with the regulation of G i .
Insulin also inhibited Ca 2ϩ mobilization induced by thrombin. Stimulation with 0.25 units/ml thrombin induced a larger increase in [Ca 2ϩ ] i than 10 mol/liter ADP, and more insulin was required to inhibit this response. When platelets were incubated with 100 nmol/liter insulin prior to stimulation with 0.25 units/ml thrombin, a 25 Ϯ 8% fall in the Ca 2ϩ response [cAMP] in the presence of IBMX (12.5 Ϯ 2.9 nmol/10 11 platelets) was expressed as 100%. Insulin did not change the raised [cAMP] by IBMX. Insulin abolished the thrombin-induced decrease of [cAMP] (61.3 Ϯ 9.1%). D, washed platelets were preincubated with 100 nmol/liter insulin for 10 min, followed by thrombin for 5 min, and subsequently stimulated with 10 ng/ml PGI 2 for 5 min in the absence and presence of 100 mol/liter L-NMMA (20 min preincubation). [cAMP] in the presence of PGI 2 was expressed as 100%. Thrombin abolished the increase in [cAMP] stimulated by PGI 2 (31 Ϯ 8%, n ϭ 4, p Ͻ 0.0001). Insulin reversed the effects of thrombin on the PGI 2 -stimulated [cAMP] (50 Ϯ 6%, asterisk indicates p Ͻ 0.001). Treatment with L-NMMA had no effect. ns indicates not significant. Further details are in Fig. 1. was observed (Fig. 3A). Optimal inhibition required a preincubation of 10 -15 min, and again there was a tendency to normalize after longer incubation (Fig. 3B). Insulin attenuated the thrombin-induced Ca 2ϩ response dose-dependently leading to optimal inhibition at 100 nmol/liter or more after 10 min of preincubation (Fig. 3C).
Effect of Insulin on the Thrombin-induced Decrease of cAMP-To investigate whether insulin interfered with the regulation of adenylyl cyclase, platelets were incubated with insulin, and [cAMP] was determined. Insulin alone (1, 10, and 100 nmol/liter) did not change the basal [cAMP] (6.7 Ϯ 1.3 nmol/10 11 platelets) during 10 min of incubation. Stimulation with thrombin (0.25 units/ml) induced a 60% decrease of basal cAMP, as in agreement with a previous publication (23). This reduction was smaller when insulin (100 nmol/liter) and thrombin were added simultaneously and completely disappeared after 10 min of preincubation with insulin (Fig. 4A). Addition of PGI 2 (10 ng/ml) raised [cAMP] to 27.3 Ϯ 2.0 nmol/ 10 11 platelets in 10 min. A similar rise was found when platelets were preincubated for 10 min with insulin (1, 10, and 100 nmol/liter). Thus, insulin failed to interfere with the activation of adenylyl cyclase by PGI 2 (Fig. 4B). To assess whether insulin interfered with cAMP degradation, platelets were treated with IBMX, an inhibitor of PDEs. IBMX raised the basal [cAMP] to 12.5 Ϯ 2.9 nmol/10 11 platelets in 5 min, and this effect was not disturbed by insulin. Again, thrombin interfered with the rise in cAMP, and insulin abolished the effect of thrombin. Thus, insulin only interfered with cAMP regulation in the presence of an agonist that activates G i (Fig. 4C). Earlier studies with platelets suspended in plasma suggested that insulin inhibited platelets by raising [cAMP] (24). This effect was attributed to insulin-induced formation of nitric oxide (NO) via NO synthase and subsequent inhibition of PDE3b. To investigate whether a similar mechanism was present in isolated platelet suspensions, platelets were treated with L-NMMA, an inhibitor of NO synthase. L-NMMA did not change the basal [cAMP] or the rise induced by PGI 2 . Also the suppression of the [cAMP] increase by thrombin was left undisturbed. Again, insulin interfered with the fall in cAMP induced by thrombin, and this effect was the same in the absence and presence of L-NMMA (Fig. 4D). These data argue against a role for NO-mediated cAMP control in the present studies.
Insulin Increases the Tyrosine Phosphorylation of G i ␣ 2 -To address the question whether insulin interfered with the regulation of cAMP formation via G i , immunoprecipitation studies were performed using an anti-G i ␣ 2 antibody, and the tyrosine phosphorylation of G i ␣ 2 on Western blot was measured using a 4G10 anti-phosphotyrosine antibody. Treatment of platelets with 1 nmol/liter insulin induced a transient increase in the tyrosine phosphorylation of G i ␣ 2 , with an optimal effect between 2 and 5 min (Fig. 5, A and C). Similar results were obtained in immunoprecipitates with 4G10 anti-phosphotyrosine antibody on blots with a G i ␣ 2 antibody (Fig. 5B). A 5-min incubation period with increasing concentrations of insulin (0.5, 1, and 10 nmol/liter) showed a dose-dependent increase in the tyrosine phosphorylation of G i ␣ 2 (see below). These results suggest that insulin interfered with the fall in cAMP via tyrosine phosphorylation of G i ␣ 2 .
IRS-1 Co-precipitates with G i ␣ 2 -To investigate whether IRS-1 played a role in insulin signaling to G i ␣ 2 , platelets were FIG. 5. Insulin increases the tyrosine phosphorylation of G i ␣ 2 . Washed platelets were treated with 1 nmol/liter insulin. At the indicated times, cells were lysed, and samples were collected for immunoprecipitation (IP) and Western blotting (WB). A, G i ␣ 2 was immunoprecipitated followed by immunoblotting with 4G10 anti-phosphotyrosine or anti-G i ␣ 2 antibody. The figures are representative for three observations with similar results. B, tyrosine-phosphorylated proteins were immunoprecipitated with 4G10 anti-phosphotyrosine followed by immunoblotting with an antibody against G i ␣ 2 . C, bands were scanned and quantified with ImageQuant software. Data are expressed as percentage of control. A dose of 1 nmol/liter insulin increased the tyrosine phosphorylation of G i ␣ 2 , being optimal at 2-5 min of incubation (281 Ϯ 106 and 276 Ϯ 58%, respectively, n ϭ 3). Further details as in Fig. 1. treated with insulin, and immunoprecipitates of G i ␣ 2 were analyzed for IRS-1. As shown in Fig. 6A, preincubation with 1 nmol/liter insulin induced association of G i ␣ 2 with IRS-1 reaching a maximum after 5 min. Thereafter was a rapid decline in the appearance of the IRS-1 band, and after 10 min it had disappeared. Increasing the insulin concentration to 100 nmol/ liter led to a stronger co-association between G i ␣ 2 and IRS-1 and a slower dissociation of the two components. Here, complex formation was apparent between 5 and 10 min. These findings suggest that IRS-1 takes part in insulin signaling to G i ␣ 2 by direct binding to the G-protein subunit and demonstrate that the association depends on the insulin concentration showing a weak and short interaction at 1 nmol/liter insulin and a stronger and more persistent interaction at 100 nmol/liter insulin. The co-association between IRS-1 and G i ␣ 2 found in G i ␣ 2 immunoprecipitates (Fig. 6A) was also present in immunoprecipitates of IRS-1 analyzed with a G i ␣ 2 antibody (Fig. 6B). A survey for the presence of G q ␣, G s ␣, and G z ␣ was negative, indicating that the interaction between IRS-1 and a G␣ subunit was specific for G i ␣ 2 .
Phosphorylation of the Insulin Receptor and IRS-1-The transient nature of the tyrosine phosphorylation of G i ␣ 2 (Fig. 5) and the co-association of IRS-1 with G i ␣ 2 (Fig. 6) suggested that signal generation by the insulin receptor was equally transient. To clarify how the insulin receptor was activated, the phosphorylation of the receptor ␤ subunit was measured with an antibody against phospho-Tyr 1158 of the insulin receptor. Treatment of platelets with increasing concentrations of insulin for 15 min induced a dose-dependent phosphorylation of the ␤ subunit (Fig. 7A). Time courses over a 15-min incubation period showed that 1 nmol/liter insulin induced an increase in ␤ subunit phosphorylation that reached a plateau after 5 min and that 100 nmol/liter insulin induced a 5-fold stronger phosphorylation that reached a plateau after 25 min (Fig. 7, B and C and not shown). Importantly, there was no indication of receptor dephosphorylation during this period at either insulin concentration. Concurrent analysis of IRS-1 showed a dose-dependent phosphorylation induced by insulin. Both at 1 and 100 nmol/liter insulin, this phosphorylation was transient showing an optimum between 5 and 10 min and decreasing to prestimulating values after 15 min. Thus, the transient phosphorylation of IRS-1 and G i ␣ 2 was not caused by receptor inactivation but was the result of interference with insulin signaling at a step downstream of receptor activation and upstream of the formation of an IRS-1-G i ␣ 2 complex.
Effect of Epinephrine on Insulin-induced Platelet Inhibition-Epinephrine is known to enhance the sensitivity of platelets to activating agents by reducing the level of cAMP (23,25,26) and to antagonize the effect of insulin in rat skeletal muscle by decreasing the IRS-1-associated activity of PI3K (27). To investigate whether epinephrine interfered with the effects of insulin on [Ca 2ϩ ] i , platelets were incubated for 5 min with insulin (1 and 10 nmol/liter), and epinephrine (10 mol/liter) was added 1 min prior to stimulation with ADP. Epinephrine increased ADP-induced Ca 2ϩ mobilization by about 20% (Fig.  8A). The lower Ca 2ϩ increase in the presence of insulin (1 and 10 nmol/liter) completely normalized in the presence of epinephrine. Thus, epinephrine abolished the inhibition by insu-FIG. 6. IRS-1 co-precipitates with G i ␣ 2 . A, washed platelets were treated with 1 and 100 nmol/liter insulin. At the indicated times, cells were lysed, and samples were collected for G i ␣ 2 immunoprecipitation (IP) and Western blotting (WB) for IRS-1. The figures are representative for four observations with similar results. Data are expressed as percentage of control. IRS-1 co-precipitated transiently with G i ␣ 2 upon incubation with insulin, being optimal after 5 min for 1 and 100 nmol/liter insulin (321 Ϯ 128 and 540 Ϯ 187%, respectively, n ϭ 4). Co-precipitation between IRS-1 and G i ␣ 2 was still present after 10 min of incubation with 100 nmol/liter insulin. B, washed platelets were treated with 1 nmol/liter insulin for 5 min; cells were lysed, and samples were collected for IRS-1 immunoprecipitation and Western blotting for G i ␣ 2 , G q ␣, G s ␣, and G z ␣. The figures are representative for three observations with similar results. G i ␣ 2 co-precipitated IRS-1 upon insulin incubation, whereas no association of other G␣-proteins with IRS-1 could be detected. Further details are in Fig. 5. lin. To investigate whether a similar neutralization was found at the level of cAMP, platelets were treated with epinephrine and thrombin followed by PGI 2 with or without 5 min of preincubation with insulin (100 nmol/liter) (Fig. 8B). Epinephrine alone or in combination with insulin abolished the rise in [cAMP] induced by PGI 2 . Since insulin alone did not change PGI 2 -induced cAMP increases (Fig. 4B), these changes are caused by epinephrine. Thrombin induced a 73% fall in [cAMP], and this fall increased further in the presence of epinephrine. Again, the effect of insulin on the decrease in [cAMP] induced by thrombin completely disappeared when epinephrine was present. Collectively, these data illustrate that epinephrine abolishes the effect of insulin on the regulation of cAMP. To investigate whether epinephrine interfered with insulin at the level of G i ␣ 2 , an immunoprecipitation was performed using a 4G10 anti-phosphotyrosine antibody, and the tyrosine phosphorylation of G i ␣ 2 was measured on Western blot. Insulin induced a dose-dependent increase in G i ␣ 2 tyrosine phosphorylation. This increase was completely abolished by epinephrine (Fig. 8C). To determine whether epinephrine interfered with insulin signaling at the level of the insulin receptor, the tyrosine phosphorylation of the insulin receptor ␤ subunit was measured. Insulin increased the tyrosine phosphorylation of the ␤ subunit, which was abolished by the addition of epinephrine. Preincubation with sodium vanadate, an inhibitor of protein-tyrosine phosphatases, partially reversed the effects of epinephrine on the tyrosine phosphorylation of the insulin receptor (Fig. 8D). These findings indicate that epinephrine neutralizes the inhibition by insulin by blocking the tyrosine phosphorylation of both G i ␣ 2 and the insulin receptor ␤ subunit.
Effect of Insulin on Platelet Aggregation in PRP-To address the question whether this mechanism was also functional in PRP, an aggregation assay was performed at 37°C. PRP was preincubated for 2 min with 1 nmol/liter insulin and 10 mol/ liter epinephrine was added 1 min prior to initiation of aggregation with 10 mol/liter ADP. Epinephrine enhanced the aggregation response to ADP. Insulin inhibited ADP-induced platelet aggregation without affecting the shape change response. The addition of epinephrine neutralized the inhibition of insulin such that the difference in platelet aggregation proved not to be significant (Fig. 9). These results illustrate that platelet inhibition by insulin is also present in PRP.
Sensitivity of Ca 2ϩ Increases to Inhibition by cAMP-As illustrated in Figs. 1 and 3, ADP-induced Ca 2ϩ rises were inhibited following a short (about 5 min) incubation with 1 nmol/liter insulin. In contrast, inhibition of thrombin-induced Ca 2ϩ rises required a longer (about 10 min) incubation with a higher dose of insulin (100 nmol/liter). To investigate whether this difference was caused by a factor downstream of the formation of cAMP, Ca 2ϩ increases induced by ADP and thrombin were measured in platelets preincubated with increasing concentrations of PGI 2 . As shown in Fig. 10, Ca 2ϩ rises induced by 10 mol/liter ADP were strongly inhibited by small increases in PGI 2 leading to complete inhibition at 1 ng/ml PGI 2 . In contrast, Ca 2ϩ rises induced by 0.25 units/ml thrombin were resistant to these PGI 2 concentrations, although at higher concentrations (10 ng/ml) complete inhibition was observed (data not shown). When the thrombin concentration was lowered to the range where a similar Ca 2ϩ increase was found as induced by ADP, both responses were equally sensitive to PGI 2 . The presence of the P2Y 12 receptor blocker AR-C69931MX led to a 45% fall in Ca 2ϩ response illustrating a major contribution of secreted ADP in thrombin-induced Ca 2ϩ rises. Collectively, these data indicate that the differences in preincubation time and insulin concentration required for inhibition of Ca 2ϩ rises reflect the weaker activation by ADP compared with a more persistent activation by thrombin in combination with granulereleased ADP. DISCUSSION The present study reveals a novel mechanism by which insulin inhibits the responsiveness of platelets for activating agents. The decrease in responsiveness is illustrated by a 17% lower ADP-induced Ca 2ϩ mobilization (0.5 nmol/liter insulin) and a 25% lower thrombin-induced Ca 2ϩ mobilization (100 nmol/liter insulin). The Ca 2ϩ mobilization induced by thrombin is substantially larger than the ADP-induced Ca 2ϩ mobilization, and consequently more insulin is required to inhibit the rise in [Ca 2ϩ ] i . In addition, there is the release of ADP from thrombin-stimulated platelets which contributes to the increase in [Ca 2ϩ ] i . The effect of insulin is transient and depends on association and tyrosine phosphorylation of IRS-1 and G i ␣ 2 . Apparently, the result is loss of G i activity as expressed by an impaired reduction of cAMP and a weaker Ca 2ϩ response than observed in the absence of insulin. Interestingly, epinephrine, an activator of G i proteins and inhibitor of IRS-1/PI3K activity in rat skeletal muscle (27), abolishes the effect of insulin on cAMP regulation, Ca 2ϩ mobilization, and aggregation. It also abolishes the insulin-induced tyrosine phosphorylation of G i ␣ 2 , again suggesting that tyrosine phosphorylation of G i leads to inhibition of the G-protein. Epinephrine abolishes insulin signaling in platelets by interfering with the phosphorylation of the insulin receptor ␤ subunit.
ADP is known to activate platelets via the P2Y 1 receptor,

FIG. 8. Effect of epinephrine on insulin-induced platelet inhibition.
A, Fura 2-AM-loaded platelets were incubated for 5 min with 1 and 10 nmol/liter insulin prior to stimulation with 10 mol/liter ADP; 10 mol/liter epinephrine was added 1 min prior to ADP stimulation. The Ca 2ϩ response induced by ADP was expressed as 100%. Epinephrine increased the ADP-induced Ca 2ϩ response to 120 Ϯ 11% (p Ͻ 0.05) and completely abolished the effect of insulin on ADP-induced Ca 2ϩ mobilization. Further details are in Fig. 1. B, platelets were stimulated as indicated with insulin (100 nmol/liter, 15 min), epinephrine (10 mol/liter, 11 min), and/or thrombin (0.25 units/ml, 10 min), and PGI 2 (10 ng/ml, 5 min) prior to analysis of [cAMP]. Epinephrine reduced [cAMP] by 74% as did thrombin. Epinephrine enhanced the effect of thrombin on [cAMP]. Insulin did not reverse the effects of epinephrine on [cAMP]. C, washed platelets were treated with insulin (0.5, 1, 10 nmol/liter) for 5 min with or without 1 min of incubation with 10 mol/liter epinephrine, and cells were lysed. Tyrosine-phosphorylated proteins were immunoprecipitated (IP) with 4G10 anti-phosphotyrosine, followed by immunoblotting with a G i ␣ 2 antibody. The blot is representative of three observations with similar results. Insulin dose-dependently increased the tyrosine phosphorylation of G i ␣ 2 . Epinephrine blunted the effects of insulin on the tyrosine phosphorylation of G i ␣ 2 . D, platelets were treated with 100 nmol/liter insulin for 5 min with or without 1 min incubation with 10 mol/liter epinephrine in the presence or absence of sodium vanadate (100 mol/liter, 30 min). Cells were lysed in 3ϫ sample buffer for Western blotting (WB) for insulin receptor phospho-Tyr 1158 . The blot is representative of three observations with similar results. Insulin increased the tyrosine phosphorylation of the insulin receptor ␤ subunit. Epinephrine abolished the tyrosine phosphorylation by insulin and was partially reversed by sodium vanadate. Further details are in Fig. 4. which is coupled to G q and signals to Ca 2ϩ mobilization, aggregation, and secretion. These responses are facilitated by concurrent binding of ADP to the P2Y 12 receptor, which is coupled to G i and suppresses cAMP formation (18). Thrombin activates platelets by binding to members of the protease-activated receptors 1 and 4. It is a potent inducer of secretion of dense granule contents, leading to liberation of ADP and subsequent activation of the P2Y 12 receptor (28). Our present findings are in accord with this concept and show qualitatively similar effects of insulin on Ca 2ϩ mobilization by ADP and thrombin. The differences in dose inhibition studies of insulin for ADP-and thrombin-induced Ca 2ϩ mobilization reflect the stronger and more persistent activation by 0.25 units/ml thrombin compared with 10 mol/liter ADP. Consequently, the fall in cAMP is also stronger with thrombin than with ADP, thus making it possible to evaluate interference by insulin. It was impossible to analyze the effect of insulin on cAMP in platelets stimulated with ADP or with the low thrombin (0.08 units/ml) concentration, but in view of the similarities with the effects induced by the high thrombin concentration a similar mechanism is likely to be operational.
For inhibition of ADP-induced Ca 2ϩ mobilization a 5-min preincubation with 1 nmol/liter insulin was sufficient. For inhibition of Ca 2ϩ mobilization induced by a relatively high thrombin concentration (to reveal the effect of insulin on cAMP), a longer preincubation with a high insulin concentration was required. These differences are reflected in the coassociation of IRS-1 with G i ␣ 2 . Under conditions that interfered with ADP signaling, IRS-1-G i ␣ 2 interaction was optimal after 5 min and rapidly declined thereafter. Under conditions that interfered with thrombin signaling, the association was more pronounced and lasted longer. Tyrosine phosphorylation of G i ␣ 2 correlated with binding of IRS-1 to G i ␣ 2 . Thus, differences in Ca 2ϩ inhibition were the result of differences in the binding of IRS-1 to G i ␣ 2 and the resulting tyrosine phosphorylation of the G␣ subunit caused by this association.
The catalytic loops within the tyrosine kinase domain of the insulin receptor contain the three tyrosine motifs Tyr 1158 , Tyr 1162 , and Tyr 1163 . The general concept is that autophosphorylation within the activation loop of the insulin receptor involves the initial phosphorylation of Tyr 1162 followed by Tyr 1158 and Tyr 1163 , upon which the insulin receptor becomes fully active. Insulin induced a dose-dependent phosphorylation of the receptor ␤ subunit, which reached a plateau after 5 (1 nmol/liter insulin) to 25 min (100 nmol/liter insulin). These kinetics differ strongly with the transient nature of the phosphorylation of IRS-1, the formation of an IRS-1-G i ␣ 2 complex, and the phosphorylation of G i ␣ 2 . Apparently, there is a crucial role for a tyrosine phosphatase that dephosphorylates IRS-1. It is known from animal studies that disruption of the gene encoding PTP1B leads to a state of increased insulin-dependent tyrosine phosphorylation of the insulin receptor and IRS proteins (5), suggesting that a single phosphatase controls the phosphorylation state of both proteins. In addition, both the insulin receptor and IRS proteins undergo serine phosphorylation, which may attenuate signaling by decreasing the tyrosine phosphorylation. Several kinases have been implicated in this process, including PI3K, protein kinase B, protein kinase FIG. 9. Effect of insulin on platelet aggregation in PRP. PRP was incubated with insulin (ins) (1 nmol/liter, 2 min) with or without 1 min of preincubation with 10 mol/liter epinephrine (epi) and stimulated with 10 mol/liter ADP at 37°C. A, the curve shown is representative for four observations with similar results. B, platelet aggregation induced by ADP was expressed as 100%. Insulin reduced platelet aggregation in PRP to 75 Ϯ 8% (p Ͻ 0.05), and epinephrine enhanced aggregation to 128 Ϯ 12% (p Ͻ 0.01). Epinephrine abolished the inhibition by insulin. Note that there was no significant difference between platelets treated with epinephrine in the presence and absence of insulin. Further details are in Fig. 1.   FIG. 10. Sensitivity of Ca 2؉ increases to inhibition by cAMP. Fura 2-AM-loaded platelets were incubated with different concentrations of PGI 2 (0.2-10 ng/ml) for 1 min prior to stimulation with 10 mol/liter ADP, 0.08 units/ml (low dose), or 0.25 units/ml (high dose) thrombin. Low dose of thrombin raised [Ca 2ϩ ] i in the same range as 10 mol/liter ADP. Both responses were equally sensitive to small increases in [cAMP] by PGI 2 resulting in 75% inhibition at 1 ng/ml PGI 2 . The higher Ca 2ϩ rises induced by high dose of thrombin were less sensitive to cAMP since only 10% inhibition was seen at 1 ng/ml PGI 2 . Complete inhibition of high dose of thrombin-induced Ca 2ϩ mobilization was observed at 10 ng/ml PGI 2 (data not shown). Incubation with AR-C69931MX prior to stimulation with low dose of thrombin inhibited the Ca 2ϩ mobilization with 45% (p Ͻ 0.002; inset). Further details as in Figs. 1-3. C, glycogen synthase kinase-3, and mammalian target of rapamycin (5). Similar mechanisms may operate in platelets with the important restriction that they leave receptor phosphorylation undisturbed.
Trovati et al. (24) reported earlier that platelets suspended in plasma are inhibited by insulin through a rise in [cAMP]. Also cAMP production induced by the stable PGI 2 analogue iloprost and the adenylyl cyclase activator forskolin was enhanced by insulin. The effect was attributed to insulin-induced production of NO, which would activate guanylyl cyclase and raise cGMP. In turn, cGMP would inhibit cAMP degradation by inhibiting PDE3b. Our present studies based on platelets suspended in buffer do not support these observations. First, inhibition of [Ca 2ϩ ] i increases by insulin is not affected by IBMX, an inhibitor of PDEs. Second, changes in [Ca 2ϩ ] i regulation and [cAMP] are unaffected by L-NMMA, an inhibitor of NO synthases. A further difference is that insulin does not change [cAMP] increases caused by stimulation of cAMP production (PGI 2 ) or inhibition of its degradation (IBMX). Instead, inhibition by insulin only becomes apparent during agonist-stimulated activation of G i , indicating that it is restricted to conditions where G i is activated by the P2Y 12 receptor.
Other reports already indicated that G i might play an important role in the signaling effects by insulin. Studies in mice with genetically compromised G i ␣ 2 expression showed hyperinsulinemia, impaired glucose tolerance, and resistance to insulin, which are characteristic for diabetes mellitus type II. In addition, there was abolished counterregulation of lipolysis by insulin, insulin-stimulated glucose transport and recruitment of GLUT-4, impaired insulin-stimulated tyrosine phosphorylation of IRS-1, and an elevated cellular phosphotyrosine phosphatase activity (29). In human adipocytes the synergistic activation of NADPH-dependent H 2 O 2 generation in vitro by Mn 2ϩ and insulin was mediated by a co-association of the insulin receptor with G i ␣ 2 (30).
Epinephrine enhances platelet activation by other agonists via binding to ␣ 2a -adrenergic receptors and G i -mediated inhibition of adenylyl cyclase (31). The G i ␣ family includes the ubiquitously expressed G i ␣ 1,2,3 as well as several members with a more restricted expression, such as G z ␣. There is ample evidence that G i ␣ 2 is a major mediator in epinephrine-induced adenylyl cyclase inhibition (31,32), but recent evidence suggests that also G z contributes to cAMP control. In G i ␣ 2 knockout mice other G i proteins can functionally replace G i ␣ 2 -mediated signal transduction (21). Comparisons between wild type and G z ␣ knockout mice reveal a role for both G z and other G i members in the regulation of cAMP (25). Thus, it is important to establish the relative contributions of the G i members in cAMP regulation with respect to the inhibitory role of insulin. In rat skeletal muscle epinephrine suppresses insulin-induced glucose uptake by decreasing the IRS-1 associated activity of PI3K (27). By analogy, IRS-1 might be a target for epinephrine in platelets especially since IRS-1 can bind to G i proteins via pleckstrin homology domains (33). The present data show that instead of interfering with IRS-1, epinephrine interferes with the phosphorylation of the insulin receptor, thereby preventing tyrosine phosphorylation of G i ␣ 2 and attenuating the rise in [Ca 2ϩ ] i .
Patients with a defect in the P2Y 12 receptor have an increased tendency to bleed, indicating that suppression of cAMP is vital for normal hemostasis. The present study reveals a similar but transient modulation of cAMP regulation by insulin. The inhibition of G i activity by insulin is in the same range as found with a P2Y 12 receptor antagonist and results in a decrease in Ca 2ϩ mobilization of about 20% (34) and reduced adhesion and aggregation (18,19). Conversely, one might speculate that the hyperresponsiveness of platelets in diabetes mellitus type I and II illustrates the absence of the platelet inhibition by insulin. These findings illustrate the importance of G i -mediated suppression of cAMP accumulation for optimal platelet function in vivo.