Discordant Effects of Glucosamine on Insulin-stimulated Glucose Metabolism and Phosphatidylinositol 3-Kinase Activity*

The impact of increased GlcN availability on insulin-stimulated p85/p110 phosphatidylinositol 3-kinase (PI3K) activity in skeletal muscle was examined in relation to GlcN-induced defects in peripheral insulin action. Primed continuous GlcN infusion (750 μmol/kg bolus; 30 μmol/kg·min) in conscious rats limited both maximal stimulation of muscle PI3K by acute insulin (I) (1 unit/kg) bolus (I + GlcN = 1.9-fold versussaline = 3.3-fold above fasting levels; p < 0.01) and chronic activation of PI3K following 3-h euglycemic, hyperinsulinemic (18 milliunits/kg·min) clamp studies (I + GlcN = 1.2-fold versus saline = 2.6-fold stimulation;p < 0.01). To determine the time course of GlcN-induced defects in insulin-stimulated PI3K activity and peripheral insulin action, GlcN was administered for 30, 60, 90, or 120 min during 2-h euglycemic, hyperinsulinemic clamp studies. Activation of muscle PI3K by insulin was attenuated following only 30 min of GlcN infusion (GlcN 30 min = 1.5-fold versus saline = 2.5-fold stimulation; p < 0.05). In contrast, the first impairment in insulin-mediated glucose uptake (Rd) developed following 110 min of GlcN infusion (110 min = 39.9 ± 1.8versus 30 min = 42.8 ± 1.4 mg/kg·min,p < 0.05). However, the ability of insulin to stimulate phosphatidylinositol 3,4,5-trisphosphate production and to activate glycogen synthase in skeletal muscle was preserved following up to 180 min of GlcN infusion. Thus, increased GlcN availability induced (a) profound and early inhibition of proximal insulin signaling at the level of PI3K and (b) delayed effects on insulin-mediated glucose uptake, yet (c) complete sparing of insulin-mediated glycogen synthase activation. The pattern and time sequence of GlcN-induced defects suggest that the etiology of peripheral insulin resistance may be distinct from the rapid and marked impairment in insulin signaling.

The hexosamine biosynthetic pathway in skeletal muscle serves a vital role in the production of the amino sugars that are utilized in multiple glycosylation pathways. Increased biosynthetic activity within the hexosamine pathway is associated with the development of insulin resistance (1)(2)(3)(4)(5). In fact, increasing the amount of flux into the GlcN pathway by various means has been shown to induce defects in insulin-stimulated glucose uptake (1, 2, 4 -6), GLUT4 translocation (3), and glycogen synthase activation (5,7,8).
Entry into the hexosamine pathway involves the conversion of fructose 6-phosphate to glucosamine 6-phosphate via the rate-limiting enzyme glutamine fructose-6-P-amidotransferase (9). The principal end product of the pathway is UDP-GlcNAc (10), which modifies intracellular proteins by glycosylation. Thus, even modest perturbations of the amount of flux through the hexosamine pathway could have diverse effects on protein functions. The amount of flux into the pathway, estimated by the accumulation of UDP-GlcNAc in muscle, is strongly correlated with the degree of impairment in peripheral insulin action (5).
The mechanism(s) of GlcN-induced defects in insulin action and the early sequence of events resulting in peripheral insulin resistance are still uncertain (4,11). The variety of observed effects of GlcN may be compatible either with a proximal defect in the insulin signaling pathway or defects at more than one downstream site of insulin action. The p85/p110 phosphatidylinositol 3-kinase, (PI3K), 1 is an important proximal effector in the insulin signaling cascade (12,13). The metabolic actions of insulin mediated by PI3K include glucose uptake (14), GLUT4 translocation (15), and glycogen synthase activation (16). Additionally, PI3K binds to all four insulin receptor substrates (IRS-1 to 4) as yet identified (17), highlighting its vital role in insulin signaling. Decreased PI3K activity in skeletal muscle has been observed in in vivo models of insulin resistance (18 -20).
Acute stimulation of PI3K by bolus insulin is known to achieve maximal levels within the first few minutes of a large bolus dose (11,18,21). However, prolonged stimulation of PI3K by insulin for 100 min has recently been demonstrated in human skeletal muscle (22). Consequently, we examined the effect of increased GlcN availability on both acute and sustained activation of skeletal muscle PI3K. Additionally, we examined the effect of GlcN on insulin-stimulated production of phosphatidylinositol 3,4,5-trisphosphate (PtdIns(3,4,5)P 3 ), probably the most metabolically important end product of PI3K (23). We thereby defined the pattern and time sequence of GlcN-induced defects in proximal insulin signaling in skeletal muscle, particularly in the activation of the intracellular PI3K pool and on the peripheral metabolic actions of insulin.

Animals
Eighty nine normal male Harlan Sprague-Dawley rats (Charles River Breeding Laboratories, Inc., Wilmington, MA) were housed in individual cages and subjected to a standard light (6 a.m. to 6 p.m.)dark (6 p.m. to 6 a.m.) cycle. The rats were anesthetized with intraperitoneal injection of pentobarbital (50 mg/kg body weight), and indwelling catheters were inserted into the right internal jugular vein and the left carotid artery, as described previously (1, 24 -26). The venous catheter was extended to the level of the right atrium, and the arterial catheter was advanced to the level of the aortic arch.

In Vivo Studies
All in vivo studies were performed 5-7 days following catheter placement in awake, fasted, and unstressed rats. At the end of all the in vivo studies, rats were anesthetized (pentobarbital 60 mg/kg body weight, intravenously); the abdomen was quickly opened, and the rectus abdominal muscle was freeze-clamped in situ with aluminum tongs precooled in liquid nitrogen (1,24). The time from injection of the anesthetic to freeze clamping of the muscle was approximately 30 s. All tissue samples were stored at Ϫ80°C for subsequent analysis. The study protocol was reviewed and approved by the Institutional Animal Care and Use Committees of the Albert Einstein College of Medicine.
Insulin Bolus Studies (Fig. 1A)-Initial pilot studies (n ϭ 12) using bolus doses of insulin between 0.5 and 1.0 unit/kg administered 1, 2, or 5 min prior to sacrifice indicated that maximal stimulation of PI3K by insulin was observed 2 min following a 1 unit/kg bolus of insulin (data not shown). Thus, in the current protocol, fasting animals received intra-arterial boluses of insulin (1 unit/kg) followed 1.5 min later by the intravenous administration of pentobarbital. Freeze-clamped rectus muscle was then obtained 2 min following the insulin bolus. The effect of primed continuous infusions of GlcN (750 mol/kg bolus, 30 mol/ kg⅐min) on the ability of insulin to maximally stimulate PI3K activity was assessed by infusion of GlcN (n ϭ 6) versus saline (n ϭ 8) for 2 h, prior to the administration of insulin. Plasma samples were obtained via the venous catheter both before the insulin bolus and at the time of sacrifice for measurement of plasma glucose, insulin, and GlcN concentrations.
3-h Insulin Clamp Studies (Fig. 1B)-Euglycemic, hyperinsulinemic (18 milliunits/kg⅐min) clamp studies were performed for 3 h in combination with [3-3 H]glucose infusion as described previously (1,24,26). GlcN (750 mol/kg bolus, 30 mol/kg⅐min) was administered throughout the 3-h insulin clamp studies (n ϭ 5), while time control euglycemic, hyperinsulinemic clamp studies were performed with infusion of saline for 3 h in additional age-and weight-matched control rats (n ϭ 6). A primed continuous infusion of HPLC-purified [ 3 H-3]glucose (NEN Life Science Products; 8 Ci bolus, 0.4 Ci/min) was infused throughout the studies in order to measure the rates of peripheral glucose uptake, glycolysis, and glycogen synthesis. A variable infusion of 25% glucose solution was started at time 0 and adjusted every 10 min, maintaining basal plasma glucose concentrations (ϳ7 mM) throughout.
Plasma samples were obtained for determination of [ 3 H]glucosespecific activity at 10-min intervals throughout the insulin infusions. Samples for measurement of plasma insulin and GlcN concentrations were obtained at times 0, 60, 120, and 180 min. The total volume of blood sampled was ϳ3.0 ml/study; to prevent volume depletion and anemia, a solution (1:1 v/v) of ϳ3.0 ml of fresh blood (obtained by heart puncture from a littermate of the test animal) and heparinized saline (10 units/ml) was infused throughout.
Time Course Studies (Fig. 1C)-Primed continuous infusions of GlcN were administered for variable time intervals during the 2-h euglycemic, hyperinsulinemic (18 milliunits/kg⅐min) clamp studies to establish a time course for the effects of GlcN on insulin-mediated glucose uptake and on insulin-stimulated PI3K activity. Thus, GlcN (750 mol/kg bolus, 30 mol/kg⅐min) was administered throughout the 2-h insulin clamp studies (designated 120 min, n ϭ 5) or during the final 30 (30 min, n ϭ 5), 60 (60 min, n ϭ 5), or 90 min (90 min, n ϭ 5) of the protocols. Time control euglycemic hyperinsulinemic (18 milliunits/ kg⅐min) clamp studies were performed with infusion of saline for 2 h in an additional n ϭ 12 age-and weight-matched control rats. Primed continuous infusions of HPLC-purified [3-3 H]glucose (8 Ci bolus, 0.4 Ci/min) were infused throughout the studies, and plasma samples were obtained for determination of [ 3 H]glucose-specific activity, insulin, and GlcN concentrations as described above.
Basal Studies-Assessment of basal PI3K activity was performed in additional fasting animals in the presence of 2-h infusions of GlcN (750 mol/kg bolus, then 30 mol/kg⅐min; n ϭ 8) or saline (n ϭ 12).

Whole Body Glycolysis
The rates of glycolysis were estimated as described previously (24 -26). Briefly, plasma-tritiated water-specific activity was determined by liquid scintillation counting of the protein-free supernatant (Somogyi filtrate) before and after evaporation to dryness. Since tritium on the C-3 position of glucose is lost to water during glycolysis, it can be assumed that plasma tritium is present either in the form of tritiated water or [3-3 H]glucose (24).

Glycogen Formation in Vivo and Glycogen Synthase Activity
The rates of peripheral glycogen synthesis during the insulin clamp studies were estimated as the difference between the rates of glucose uptake and glycolysis. Muscle glycogen concentration was determined following digestion with amyloglucosidase as described previously (24). Muscle glycogen synthase activity was measured by a modification (24,26) of the method of Thomas et al. (27) and is based on the measurement of the incorporation of radioactivity into glycogen from UDP-[U-14 C]glucose.

PI 3-Kinase Activity
Frozen rectus muscle samples were pulverized in liquid nitrogen and placed in ice-cold lysis buffer (NaCl 140 mM, Tris⅐HCl 10 mM, CaCl 2 1 mM, MgCl 2 1 mM, aprotinin 10 g/ml, leupeptin 50 M, sodium vanadate 2 mM, phenylmethylsulfonyl fluoride 1 mM, glycerol 10%; total dilution 1:3). Samples were immediately homogenized on ice with a Tissumizer at moderate speed, using 3 cycles of 20 s each. Nonidet P-40 1% by volume was added to each tube, and the mixture was rotated for 1 h at 4°C. Post-13,000 ϫ g supernatants were immunoprecipitated with ␣p85 antibody (28) overnight and then assayed for PI3K activity by the method of Ruderman (12). Each set of assayed samples included tissues from at least 2 fasted and insulin-stimulated animals, respectively, and PI3K activity in each sample was expressed as a multiple of the average fasting activity in that sample set.
Preparation of skeletal muscle samples involved alkaline hydrolysis of tissue phospholipid extract to generate Ins(1,3,4,5)P 4 from PtdIns(3,4,5)P 3 . Briefly, 200 mg of frozen muscle was homogenized in 2 ml of 10% trichloroacetic acid and then centrifuged at 3,000 rpm for 10 min. 3 ml of EDTA 10 mM were added to the pellet, which was then centrifuged at 3,000 rpm for 10 min. The pellet was washed twice with chloroform/MeOH and then extracted with chloroform/MeOH/HCl (40: 80:1). Following centrifugation at 3,000 rpm for 20 min, the chloroform was taken to dryness under N 2 . 200 l of 1 M KOH was added to the pellet, which was put in a boiling water bath for 30 min. The pH was then raised to 5.0 with 1 M acetic acid. Two extractions were performed with 2 ml of butanol/petroleum ether/ethyl acetate (20:4:1) to remove fatty acids, and then the samples were dried and stored at Ϫ20°C. On the day of the assay the samples were resuspended in 200 l of 40 mM acetic acid.
The assay was performed by measurement of displaced labeled

Analytical Procedures
Plasma glucose was measured by the glucose oxidase method (Glucose Analyzer II, Beckman Instruments, Inc., Palo Alto, CA) and plasma insulin by radioimmunoassay using rat and porcine insulin standards. Plasma [ 3 H]glucose radioactivity was measured in duplicate on the supernatants of Ba(OH) 2 and ZnSO 4 precipitates of plasma samples after evaporation to dryness to eliminate tritiated water. Regression analysis of the slopes of 3 H 2 O rate of appearance (used in the calculation of the rates of glycolysis) was performed at 60-min intervals throughout the insulin clamp studies.
The retention times for UDP-Glc, UDP-Gal, UDP-GlcNAc, and UDP-GalNAc were 28.5, 30.7, 33.9, and 35.4 min, respectively. Plasma GlcN concentrations were determined by high performance liquid chromatography (HPLC) following quantitative derivatization with phenylisothiocyanate as described by Anumula and Taylor (32). All HPLC analyses were performed on a Waters HPLC system using a reverse-phase, ion pairing isocratic method, on two C18T (Supelco) reverse-phase columns (0.46 ϫ 25 cm) in series.
ATP concentrations in skeletal muscle were measured as follows: freeze-clamped skeletal muscle samples were finely pulverized in liquid nitrogen and then stirred into ice-cold 3 M HClO 4 (3 ml/g) in an alcohol bath maintained at Ϫ10°C. Following complete penetration of acid into tissue powder, tissue extracts were diluted with 1 ml of H 2 O, 0.3 ml of extract, rotated at 4°C for 10 min, then centrifuged at 5,000 ϫ g for 10 min. Supernatants were removed and neutralized to pH 7 with 2 M KOH and 0.4 M imidazole base. ATP concentrations in neutralized supernatants were calculated from fluorometric measurements of NADPH generated in a two-step reaction, following the method of Passonneau and Lowry (33).
All values are presented as the mean Ϯ S.E. Comparisons between groups were made using repeated measures analysis of variance where appropriate. Where F ratios were significant, further comparisons were made using Student's t tests (paired difference test and small sample test for independent samples, as appropriate).

General and Biochemical Characteristics of the Experimental
Animals-At the time of the study, the mean body weight of the animals was similar in all groups and averaged 309 Ϯ 16 g. The mean plasma glucose, free fatty acid, and insulin concentrations at base line (0 h) were also similar in all groups.
Insulin Bolus Studies-We assessed the effect of 2-h infusions of GlcN versus saline on acute insulin stimulation of PI3K, 2 min following intra-arterial bolus of insulin (1 units/kg) in normal rats (Fig. 2).
The administration of bolus insulin resulted in rapid elevations in plasma insulin levels from fasting levels of 25 Ϯ 5 microunits/ml to Ͼ1000 microunits/ml after 2 min. Whereas in saline-infused animals, bolus insulin induced a 3.3-fold stimulation of PI3K activity, there was only a 1.9-fold stimulation by insulin in the glucosamine-treated group. Of note, there was a tendency toward small increases in basal (non-insulin-stimulated) PI3K activity following 2 h of GlcN infusion, but these did not achieve statistical significance.
Following the observed decreases in Rd and GIR in the presence of increased GlcN availability, more prolonged infusion of GlcN also resulted in impairments in the rates of peripheral glycogen synthesis (180 min ϭ 21.7 Ϯ 1.7 versus 30 min ϭ 26.3 Ϯ 2.0 mg/kg⅐min; p Ͻ 0.01) which first became significant by 160 min of GlcN infusion (p Ͻ 0.05).
PI 3-Kinase Activity (Fig. 3B)-We measured PI3K activity following 3-h euglycemic, hyperinsulinemic clamp studies with infusions of either GlcN or saline throughout. The infusion of GlcN for 3 h throughout insulin clamp studies resulted in a marked impairment in the ability of insulin (ϳ600 microunits/ ml) to stimulate PI3K activity (Fig. 3B), such that there was only a 1.2-fold stimulation of PI3K activity with GlcN infusion, relative to a 2.4-fold stimulation in the time control group.
Time Course Insulin Clamp Studies (Fig. 4)-Once we established that the defect in insulin-stimulated PI3K activity was also detectable following prolonged insulin infusions, we performed time course studies to determine the sequence of various GlcN-induced defects in insulin action and the length of time required for their onset. These observations allowed us to examine the time course of the effect of GlcN on PI3K vis à vis other metabolic actions of insulin.
Primed continuous infusion of GlcN for variable durations during the course of 2-h insulin clamp studies resulted in very rapid elevations in plasma GlcN levels, to 1.1 Ϯ 0.2 mM by 60 min of primed continuous infusion (Table I), again reaching a plateau by 90 min of infusion. Fig. 4A depicts the GIR at 10-min intervals throughout the 2-h insulin clamp studies with GlcN infusion over variable time intervals and throughout time control studies with saline infusion. Again, significant decreases in Rd and GIR were observed following 110 min of GlcN infusion (p Ͻ 0.05 versus 30 min) and were thus only seen in the group which received GlcN infusion throughout the 2-h studies (GlcN 120 min).
The time course of the defect in glucose uptake was comparable with that in previous studies, as was the ϳ17% decrease in rates of peripheral glucose disposal following 3 h of GlcN infusion (1). Indeed, these impairments were about half-maximal relative to previously reported decreases of ϳ35% following 5 h of GlcN infusion (1). PI 3-Kinase Activity (Fig. 4B)-Thus, by having demonstrated that GlcN impairs maximal stimulation PI3K by bolus insulin, we examined whether this defect might be relevant in the context of more chronic insulin action. We first determined that activation of PI3K in skeletal muscle was sustained following 3-h euglycemic, hyperinsulinemic clamp studies. However, prolonged increases in GlcN availability throughout the clamp studies again resulted in marked inhibition of PI3K stimulation.
Relative to fasting levels, there was a 2.5-fold stimulation of PI3K activity by elevations in circulating insulin levels for 2 h. Following 30 min of GlcN infusion, there was a significant decrease in the ability of insulin to activate PI3K (1.5-fold increase, p Ͻ 0.05 for 30Ј versus saline-infused), whereas beyond 60 min of GlcN infusion insulin failed to activate PI3K.
PtdIns(3,4,5)P 3 (Fig. 5)-We measured PtdIns(3,4,5)P 3 levels in skeletal muscle following the 2-and 3-h insulin clamp studies, with and without the infusion of GlcN for variable time intervals. Whereas acute bolus insulin increased production of PtdIns(3,4,5)P 3 by ϳ2.4-fold above fasting levels (results not shown), levels were raised ϳ1.4-fold following insulin clamp studies. GlcN infusion for up to 3 h did not affect the ability of insulin to increase PtdIns(3,4,5)P 3 levels. Since there was no difference between any of the time intervals (results not shown), we combined the results from all 2-and 3-h insulin clamp studies, with (GlcN) and without (Sal) the infusion of GlcN continuously throughout the studies (Fig. 5). Skeletal muscle samples from these studies were assayed on a total of nine separate occasions. Each assay compared at least 3 studies for each of the following conditions: GlcN, Sal, and fasting (Fast), and tissue from each study was assayed on at least three separate occasions.
In the time course studies, elevated skeletal muscle concentrations of UDP-GlcNAc were already evident by 30 min of primed continuous GlcN infusion (Fig. 6A). The skeletal muscle concentrations of UDP-Glc and UDP-galactose were similar throughout the various GlcN infusions and comparable with those in saline-infused controls. However, the ratios of the skeletal muscle concentrations of UDP-GlcNAc to UDP-Glc progressively increased with GlcN infusion over all time courses (Fig. 6B).
The levels of skeletal muscle UDP-GlcNAc that had accumulated following 30 min of primed continuous GlcN infusion were similar to those associated with fat-induced insulin resistance and other models of increased flux into the hexosamine pathway, namely hyperglycemia and the infusion of glucosamine at much lower infusion rates than those currently used (1,2). However, the comparable elevations in UDP-Glc-NAc previously observed with these other experimental conditions are associated with a time delay of several hours.
Glycogen Formation and Glycogen Synthase Activity (Fig.  7)-As noted above, there was a decrease in the rate of glycogen synthesis following 160 min of GlcN infusion in the 3-h insulin clamp studies. There was, however, complete sparing of the activation of glycogen synthase by insulin following 3 h of GlcN infusion. This suggests that the decrease in the rate of glycogen synthesis is secondary to the decrease in glucose uptake.
Skeletal Muscle ATP Concentrations-To determine whether increased availability of GlcN might result in intracellular ATP depletion, ATP concentrations were measured in skeletal muscle following 3-h insulin clamp studies with infusion of either saline or GlcN throughout. Following infusion of GlcN for 3 h, muscle ATP levels averaged 4.7 Ϯ 0.4 mol/g wet weight, identical to levels of 4.6 Ϯ 0.2 mol/g in the saline-infused controls.

DISCUSSION
These protocols were designed to advance our understanding of the pathogenesis of GlcN-induced insulin resistance by establishing the temporal relationship of defects in insulin signaling and peripheral insulin action. We examined the effect of increased glucosamine availability on the ability of insulin to stimulate PI3K, peripheral glucose uptake, and glycogen synthase in skeletal muscle. There was a substantial delay between the rapid and marked impairment in PI3K activation and the gradually progressive defect in glucose uptake. Sustained elevations in plasma GlcN levels for up to 3 h did not affect the activation of skeletal muscle glycogen synthase by insulin.
Data presented in two recent reports suggest that prolonged infusion of GlcN in normal rats has inhibitory effects on insulin-stimulated IRS-1-associated PI3K activity in skeletal muscle (34,35). Our findings have revealed the earliest effects of GlcN on PI3K activation yet demonstrated. Since the primary goal of these studies was to establish a time course of the inhibitory effects of GlcN, primed continuous infusions of GlcN were designed to achieve rapid elevations of plasma GlcN levels into the range previously associated with defects in in vivo  (GlcN 30 -120Ј), or following control euglycemic hyperinsulinemic clamp studies with saline infusion (Sal). The activity of muscle PI3K is expressed relative to fasting tissues included in each assay set. The pairs of representative images depict 32 P incorporation into PI-3-P. *, p Ͻ 0.01; #, p Ͻ 0.001 versus saline. arb units, arbitrary units.
insulin action, and thereby to permit careful analysis of the earliest sites of GlcN-induced defects in insulin action. Another novel contribution of this work was in examining the effect of GlcN on the ability of insulin to activate the entire intracellular PI3K pool in skeletal muscle. By only examining the activation of IRS-1-associated PI3K, one could potentially overlook important effects of GlcN on PI3K mediated via other insulin receptor substrates (17). The association of IRS-1 with PI3K represents a very early step in insulin signaling, with significant interactions occurring within seconds of the initial activation of the insulin receptor. Since this is a dynamic interaction, the majority of activated PI3K may no longer be recoverable with ␣IRS-1 antibodies following several hours of stimulation by insulin.
Given the complexities of insulin signaling and the interplay between initial activation of the signaling cascade by insulin and subsequent feedback inhibition (36), substantial time might be required for a defect in proximal signaling to affect glucose transport. Alternatively, the GlcN-induced defect in PI3K activation may not be responsible for the subsequent defects in peripheral glucose disposal. Indeed, several models in which proximal insulin signaling was reduced or defective have revealed a remarkable sparing of glucose uptake. Insulin receptor kinase activity was reduced by 42%, yet insulin-stimulated 2-deoxyglucose uptake was unaffected in soleus muscle of transgenic mice with stably expressed dominant-negative (Ala-1134 3 Thr) mutant human insulin receptors (37). In addition, overexpression of protein tyrosine phosphatase 1B in 3T3-L1 adipocytes reduced insulin-stimulated, IRS-1-associated PI3K activity by ϳ60% without altering insulin-stimulated glucose transport (38). Conversely, stable expression of mutant insulin receptors from patients with type A insulin resistance in Chinese hamster ovary cells resulted in markedly impaired insulin-stimulated glycogen synthesis despite normal tyrosine phosphorylation of IRS-1 by insulin (39).
These complementary models all show discrepancies between defects in proximal insulin signaling and insulin-stimulated glucose uptake. Together, they suggest that marked impairment of insulin-mediated PI3K activation may be compatible with normal insulin action on glucose uptake. Alternatively, the residual PI3K activity may be sufficient to stimulate and sustain glucose uptake in skeletal muscle, since impaired glucose uptake has only been demonstrated in the presence of complete abolition of insulin-stimulated PI3K activity with wortmannin (14). Thus, there may be an independent defect to account for the impairment in insulin-mediated glucose uptake.
Another apparent paradox is the complete preservation of PtdIns(3,4,5)P 3 levels in skeletal muscle despite the inhibitory effect of GlcN on insulin-stimulated PI3K activity. A potential explanation both for our results and the in vitro findings de-  scribed above might be provided by recent characterizations of the Class II PI 3-kinases. PI 3-kinases in this novel group are distinguished from Class I PI 3-kinases by a C-terminal C2 domain and markedly lower sensitivity to wortmannin (40,41). The widely expressed ␣-isoform can be activated by insulin in cell culture models of insulin-sensitive tissues, probably by insulin-induced phosphorylation that does not involve association with IRS-1 (40). Furthermore, phosphorylation of PtdIns(4,5)P 2 to PtdIns(3,4,5)P 3 by a Class II PI3K has been observed when lipids were presented together with phosphatidylserine as a carrier (41). Hence, persistent stimulation of Class II PI3K by insulin could occur despite GlcN-induced impairment of IRS-1 activation (34), and thus the PtdIns(3,4,5)P 3 pool could remain unperturbed.
There was also a discrepancy between the inhibitory effects of GlcN on insulin-mediated glucose uptake and the complete sparing of insulin-stimulated glycogen synthase activity. Glucose transport and glycogen synthase activation appear to share many steps in the insulin signaling pathway, particularly PI3K (16) and Akt (protein kinase B), a serine/threonine kinase downstream of PI3K in the insulin signaling cascade (42,43). However, Akt has a dominant role in the regulation of glycogen synthase activity via inhibition of glycogen synthase kinase 3 (42). It is noteworthy that there was apparent sparing of insulin-mediated Akt activation despite GlcN-induced impairment of PI3K activity in two recent reports (34,35). Additionally, expression of constitutively active PI3K in 3T3-L1 adipocytes resulted in differential effects on glucose uptake and glycogen synthesis, with some ability to simulate the effects of insulin on the former but no effect on the latter (44). The above data suggest that the observed defects in insulin-stimulated PI3K activity would not be expected to affect glycogen synthase activation.
However, the ability of insulin to activate glycogen synthase was inhibited in vitro by overexpression of glutamine:fructose-6-phosphate amidotransferase in Rat-1 fibroblasts (8) and thereby chronically enhancing flux into the hexosamine pathway. Insulin was also unable to activate glycogen synthase in the presence of massively (ϳ7-fold) increased flux into the GlcN pathway following the combined infusion of GlcN and uridine for several hours (5). Thus, a greater degree of flux into the GlcN pathway may be required for inhibition of insulin-stimulated glycogen synthase activity, relative to other GlcN-induced defects.
The O-linked pathway of cytosolic glycosylation offers a po-tential mechanism whereby increased flux into the hexosamine pathway might induce diverse metabolic effects. This process involves the addition of an O-linked GlcNAc moiety to the hydroxyl group of either serine or threonine and is regulated by the dynamic interplay between the enzymes O-GlcNAc-transferase (45) and N-acetylglucosaminidase (46). O-Linked glycosylation thereby induces rapid post-translational modifications of a wide range of proteins, including cytoskeletal proteins, tumor suppressor peptides, and transcription factors (e.g. Sp1) (45,47,48). Presumably, increased flux into the GlcN pathway could quantitatively or qualitatively alter the O-linked glycosylation state of various proteins involved in insulin action and thereby affect their function via numerous potential mechanisms. The time frame of the observed defect in glucose uptake is compatible with an effect on transcription. Indeed, increased availability of GlcN has been shown to increase the transcription of certain genes, including transforming growth factor-␤ (49) and leptin (50). In addition, O-linked glycosylation of serine or threonine residues can competitively inhibit phosphorylation at the same sites (51). The reciprocal glycosylation/phosphorylation state of a key serine/threonine kinase could affect the tyrosine phosphorylation state and thus the activation of IRS-1 or other upstream signaling molecules. O-Linked glycosylation can also affect the intracellular localization of proteins (51) and might in such a way determine the intracellular localization of PI3K and thus its activation by insulin. Alternatively, increased GlcN availability might directly affect glucose transport due to abnormal O-linked glycosylation of the GLUT4 vesicles themselves (5).
Recently, Hresko et al. (11) reported massive depletion of intracellular ATP together with impaired 2-deoxyglucose uptake and decreased insulin-stimulated phosphorylation/activation of multiple sites in the insulin signaling pathway in 3T3-L1 adipocytes exposed to variable concentrations of GlcN in glucose-free media. However, such high ratios of GlcN to glucose would cause nearly complete blockade of hexokinase by competitive inhibition (1), which could itself account for substantial inhibitions of 2-deoxyglucose uptake. The lack of GlcNinduced ATP depletion in the current studies supports the notion that skeletal muscle, unlike cultured cells, has a great capacity to regenerate ATP.
In conclusion, the impairment in insulin-stimulated PI3K activity induced by increased GlcN availability is very rapid and precedes the subsequent impairment in peripheral glucose FIG. 7. The effect of increased GlcN availability on insulin-stimulated glycogen synthetase activity (K m ) following 2-h euglycemic, hyperinsulinemic clamp studies with primed continuous infusions of GlcN for the indicated time intervals, or following control euglycemic, hyperinsulinemic clamp studies with saline infusion. Glycogen synthetase (GSase) activation following these studies was compared with that in fasting tissues following 2-h infusions of saline (Fast). Ј indicates minutes. uptake. However, normal stimulation of skeletal muscle glycogen synthase by insulin still occurred despite the GlcN-induced impairment in PI3K activation. The diverse mechanisms whereby O-linked glycosylation can affect intracellular processes, and the variable times required for these effects, could be compatible with more than one GlcN-induced defect in the insulin effector cascade. Thus, although an early defect in PI3K activation could result from a rapid effect on either its phosphorylation state or intracellular localization, there may be a subsequent defect(s) in transcriptional regulation to account for the progressive decline in glucose uptake. Indeed, the observed discrepancies between the effects of GlcN on different metabolic actions of insulin suggests that GlcN-induced insulin resistance in skeletal muscle may involve the existence of defects at more than one site in the insulin effector cascade.