INTRODUCTION

Glucose transport is considered rate limiting for glucose metabolism by skeletal muscle. Glucose enters muscle cells by facilitated diffusion, mediated by two glucose transporter isoforms, GLUT1 and GLUT4. GLUT1 is expressed in most cells, is localized primarily at the cell membrane, and is thought to participate mainly in basal glucose transport in muscle. GLUT4 is expressed only by cells that accelerate glucose transport in response to insulin (skeletal muscle, heart muscle, and adipose cells). In contradistinction to GLUT1, under basal conditions GLUT4 is localized mainly in intracellular vesicles and is translocated to the cell membrane in response to insulin. The predominant glucose transporter in skeletal muscle is GLUT4, and its translocation to the plasmalemma is thought to be the primary mechanism by which insulin stimulates muscle glucose uptake (reviewed in Refs. 1, 2, 3).

Recently, transgenic mice have been developed that overexpress GLUT4 in skeletal muscles, heart, and adipose cells (4) or GLUT1 in skeletal muscles (5). Both models exhibit mild fasting hypoglycemia, without significant changes in circulating insulin or glucagon, and enhanced glucose tolerance after challenge with an oral glucose load (5, 6). Basal glucose transport and glycogen deposition are also increased in both models, but their enhancement is much greater in muscles overexpressing GLUT1 (6, 7) rather than GLUT4 (8). The latter show enhanced glucose transport stimulation by insulin in vitro (8), whereas insulin fails to further stimulate glucose transport in muscles overexpressing GLUT1 (9). Other stimuli, e.g. insulin like growth factor-1, hypoxia, and contractile activity, which normally stimulate muscle glucose transport and GLUT4 translocation, also fail to stimulate glucose transport in GLUT1-overexpressing muscles (9), although total GLUT4 protein expression is unchanged (5). During euglycemic hyperinsulinemic clamp studies, the difference between the two types of transgenic mice was even more striking. Insulin stimulated total body glucose utilization was ~40% higher in GLUT4-overexpressing mice than in controls, but it was 50% lower than that of controls in GLUT1 transgenic mice (10). Similar studies using isotope tracer dilution methods during the euglycemic hyperinsulinemic clamp revealed that the decreased glucose utilization observed in previous studies (10) in GLUT1 transgenic mice reflected, in part, increased hepatic glucose output. Nevertheless, isotopically determined basal glucose utilization was increased in GLUT1 transgenic mice, and the insulin-stimulated increment in glucose disposal was decreased, indicating insulin resistance, in agreement with data in isolated muscles (9).¹ Since skeletal muscle is the major site of insulin stimulated glucose utilization in vivo (11), the data indicate that chronic overexpression of GLUT1 in muscle leads to insulin resistance (10).

Insulin resistance is a major feature of non-insulin-dependent diabetes and of uncontrolled insulin-dependent diabetes (11). Sustained hyperglycemia causes insulin resistance in humans and rodents; the major site of glucose-induced insulin resistance is skeletal muscle (reviewed in Refs. 11 and 12). Studies in adipocytes in primary culture suggested that increased flux of glucose via the hexosamine-synthetic pathway may cause glucose-induced insulin resistance of glucose transport (13, 14). Subsequent studies in isolated skeletal muscle (15) and in glucose-infused (12) or glucosamine-infused (16, 17) rats were consistent with this hypothesis.

Glutamine:fructose-6-phosphate amidotransferase (GFAT)² is the rate-limiting enzyme that catalyses the entry of glucose into the hexosamine-synthetic pathway. The products of the reaction are glucosamine 6-phosphate (GlcN-6-P) and glutamate. Upon further metabolism, the former yields essential substrates for glycosylation of proteins and lipids. Major products of the pathway that accumulate in cells are UDP-N-acetyl-hexosamines (UDP-HexNAc), representing UDP-N-acetylglucosamine (UDP-GlcNAc) and UDP-N-acetylgalactosamine (UDP-GalNAc), usually in a 3:1 ratio (14, 15, 18, 19). A schematic diagram of the hexosamine biosynthetic pathway in the context of glucose metabolism in muscle is shown in Fig. 1.

GLUT1-overexpressing muscles are subjected to a chronically increased flux of glucose, without major changes in substrate or hormone concentrations in the extracellular milieu. To test the hypothesis that the insulin resistance of these muscles may reflect increased flux of glucose via the hexosamine-synthetic pathway, we measured the concentrations of the major products of the pathway, UDP-HexNAc, and the activity of the rate-limiting enzyme, GFAT. We also measured the concentrations of other glucose-derived nucleotide-linked sugars: UDP-hexoses (UDP-Hex = UDP-Glc + UDP-Gal) and GDP-mannose. UDP-Hex is derived from glucose 1-phosphate, and UDP-Glc is the obligatory substrate of glycogen synthase.

EXPERIMENTAL PROCEDURES

The transgenic mouse lines that overexpress GLUT1 in skeletal muscle (5, 7, 9, 10) or GLUT4 in skeletal muscle, adipose tissue and heart (4, 6, 8, 10) have been described previously. Transcription of the human GLUT4 gene was controlled by its own promoter (4), whereas that of the human GLUT1 gene was controlled by the rat myosin light chain 2 promoter (5). For experiments, mice expressing a single GLUT1 or GLUT4 gene were mated with wild-type B65JLF1/J and C57BL/6J mice (Jackson Laboratories, Bar Harbor, ME) respectively. The offspring of these matings consisted of an ~50:50 mixture of heterozygous transgenic mice and wild-type controls (10). Transgenic mice and controls from the same litter were used in experiments. There was no difference in the parameters measured between males and females in either the transgenic or the control groups, and the data were pooled for analysis. The levels of total glucose transporter protein expressed in skeletal muscle are similar in the GLUT1 and GLUT4 transgenic lines. What varies is the proportion of GLUT4:GLUT1 in the two lines. This ratio is ~1:1 in the GLUT1 line and ~50:1 in the GLUT4 line.

At the time of study, mice were 8-15 weeks old. They were housed in a facility equipped with a 12-h light cycle, fed ad libitum (Rodent Blox, Ralston Purina, St. Louis, MO) or fasted for 16 h before experiments, and killed between 9:00 and 11:00 a.m. under methoxyflurane (Metofane, Pittman Moore, Washington Crossing, NJ) anesthesia. Hind limb muscles (including calf, thigh, and hip muscles) and the liver were rapidly removed and frozen in liquid N₂. Muscles from one hind limb (~0.5 g) were used for the assay of GFAT activity and muscles from the other (~0.5 g) were used for the analysis of nucleotide-linked sugars; ~0.3 g of liver was prepared for either analysis. The frozen tissues were powdered in a mortar under liquid N₂, and weighed aliquots of the frozen tissue powder were processed immediately for analyses of GFAT activity and nucelotide sugars as described below. Frozen hind limb muscles, from a separate experiment, were stored at 80 °C for mRNA analysis.

The method used and its validation have been described in detail (12). Briefly, frozen tissue powder was homogenized at 4 °C in 3 volumes of 0.3  perchloric acid (PCA), precipitates were pelleted by centrifugation, and PCA was extracted from the supernatants with 2 volumes of 1:4 trioctylamine:1,1,2-trichloro-trifluoroethane. The aqueous phase was stored at 70 °C until analysis by high pressure liquid chromatography (HPLC) within 5 days. The extracts were filtered, and HPLC was performed on a Whatman Partisil anion exchange column (4.6 × 250 mm) eluted with a concave gradient of ammonium phosphate from 15 m, pH 3.8, to 1 , pH 4.5, over 50 min, at a flow rate of 1 ml/min. UDP-HexNAc, UDP-Hex, GDP-mannose, and UDP were quantified by UV absorption (A₂₅₄) and comparison to external standards. With this method UDP-GlcNAc coelutes with UDP-GalNAc and UDP-Glc with UDP-Gal. Since the absorption coefficient of glucose or galactose-containing sugar nucleotides was very similar, UDP-GlcNAc was used as a standard for UDP-HexNAc and UDP-Glc for UDP-Hex.

PCA extracts were prepared from powdered, frozen muscles as described above and stored at 70 °C until analysis within 24 h.. Most nucleotides were separated on a Beckman Ultrasphere ODS 5 µm C₁₈ reverse phase HPLC column eluted isocratically with 100 m triethylamine phosphate (TEA-Pi), pH 5.8/1% acetonitrile, using a modification of the method of Pogalotti and Santi (20). Adequate separation of UTP required a second injection of extract, which was eluted isocratically with TEA-Pi, pH 5.8/0.5% acetonitrile. Peaks were quantified by UV absorption and comparison to external standards.

The GFAT enzyme activity assay was carried out as described previously (12). Frozen tissue powder was homogenized in 4-5 volumes of extraction buffer (25 m HEPES, pH 7.5, 4 °C, 5 m EDTA, 100 m KCl, 5 m glucose 6-phosphate, and a mixture of protease inhibitors (12)). Extracts were centrifuged at 4 °C (60,000 × g for 15 min and the supernatants at 100,000 × g for 60 min). The supernatants were spin-filtered over Sephadex G-25 columns, preequilibrated with assay buffer (25 m K₂PO₄, pH 7.5, 1 m EDTA, 50 m KCl). Aliquots of the gel-filtered cytosolic extracts were incubated for 60 min at 37 °C in the presence of 12 m glutamine and 6 m fructose 6-phosphate (Fru-6-P). In studies of muscle extracts, aliquots were also assayed at intermediate concentrations of Fru-6-P (2.4, 3.0, and 4.0 m). Reactions were stopped with PCA (final concentration 0.3 ), and after centrifugation, PCA was extracted from the supernatants as described above. The aqueous phase was stored at 70 °C until analysis within 1 week. GlcN-6-P, the product of GFAT activity, was measured fluorometrically, after derivatization of the extracts with o-phthalaldehyde and separation of GlcN-6-P by HPLC on a Beckman Ultrasphere ODS 5-µm C₁₈ column as described (12). For each extract, a blank was analyzed in parallel, consisting of all the ingredients of the assay but with PCA added immediately at 4 °C. Blank absorption was typically <10% of that of GlcN-6-P and was subtracted. In some experiments, an additional blank was generated, i.e. 0.5 m N-acetylglucosamine (GlcNAc) was included in the assay in the presence of 6 m Fru-6-P, and the samples were incubated for 60 min at 37 °C. Addition of GlcNAc completely inhibited GlcN-6-P generation, and the blanks obtained with the two methods were indistinguishable. The protein concentration in the gel-filtered extracts was determined spectrophotometrically with Coomassie protein assay reagent (Pierce) against bovine serum albumin standards, and GFAT enzyme activity was expressed in pmols of GlcN-6-P generated·mg of protein¹·min ¹.

Total RNA was isolated from skeletal muscle by the procedure of Chomczynski and Sacchi (21). GFAT mRNA was quantified by RNAse protection assay (RPA) as described previously (12). A plasmid (modification of pGEM-57f (+); Promega, Madison, WI) containing an insert with a PCR-derived sequence from the rat GFAT cDNA was a gift from Dr. Gary L. McKnight (Zymogenetics, Seattle, WA). The plasmid was linearized with NotI. The probe for the RPA was synthesized in the presence of [-³²P]UTP (DuPont NEN) using T7 RNA polymerase according to the protocol for the Ambion (Austin, TX) MAX1script kit. Template DNA was removed by digestion with DNase I, and the full length probe (363 bp) was purified on a 5% sequencing gel and eluted. Muscle RNA samples (10 µg) were hybridized overnight at 37 °C with 10⁵ cpm of the probe and then digested with RNase A/T1 according to the protocol of the Ambion RPA II kit. After precipitation, the pellets were dissolved in loading buffer, the products were separated on a 5% sequencing gel, and the protected bands (292 bp) were identified by autoradiography, excised, and quantified by liquid scintillation counting. The data were normalized to -actin mRNA levels and determined in parallel RPA assays, as described (12).

Results are presented as means ± S.E. The significance of differences between means was analyzed by unpaired two-tailed Student's t test.

All reagents and standards were of the highest purity available and were obtained from Sigma unless otherwise noted.

RESULTS

GLUT1-overexpressing Mice

The concentrations of UDP-HexNAc, UDP-Hex, and GDP-mannose were increased 2-3-fold in muscles of GLUT1-overexpressing mice compared to the wild-type littermate controls (Fig. 2). The marked increases in nucleotide-linked sugars were observed when mice were fed ad libitum or after an overnight fast. The differences between muscles of GLUT1 transgenics and controls were highly significant (p < 0.001 for UDP-HexNAc and UDP-Hex, in the fed and fasted states, respectively, and p < 0.02 and p < 0.001 for GDP-mannose, in the fed and fasted states, respectively). Fasting tended to decrease the concentrations of the three nucleotide-linked sugars slightly (~25%) in muscles of control mice, but the differences were not statistically significant. The ratio of UDP-HexNAc/UDP-Hex was not significantly different between muscles from control and GLUT1 transgenic mice. The concentrations of UDP were higher in muscles of GLUT1 transgenic mice than in controls (p < 0.05), and the increases essentially paralleled those of the UDP-linked sugars.

In view of the ~2-fold increase in UDP concentrations in GLUT1-overexpressing muscles, concentrations of other nucleotides were also measured. There were no significant differences in the concentrations of ATP, ADP, GTP, GDP, or UTP between muscles overexpressing GLUT1 and controls. UTP was 12% lower in the transgenic muscles, but the difference did not reach statistical significance (Table I).

Table I. Nucleotide concentrations in muscles of GLUT1 heterozygous transgenic mice and control littermate Mice were fed ad libidum. Muscle extracts were prepared as described in the legend of Fig. 2; nucleotides were separated by HPLC and quantitated fluorometrically as described under ``Experimental Procedures.'' n = 3 controls and 4 transgenics, except for UDP, which was measured in the experiments shown in Fig. 2A, where n = 6-7. Nucleotide Control GLUT1 nmol/g of muscle ATP 2776  ± 14 2771  ± 112 ADP 375  ± 18 325  ± 19 GTP 535  ± 23 540  ± 37 GDP 12.2  ± 0.37 12.6  ± 0.82 UTP 52.5  ± 4.2 46.2  ± 6.1 UDP 1.7  ± 0.12 3.3  ± 0.63a a  p < 0.05.

In control mice, UDP-HexNAc, UDP-Hex, and UDP concentrations (nmol/g of tissue) were ~20-fold higher in liver than in muscle, whereas GDP-mannose was increased 6-fold (compare Figs. 2 and 3). GLUT1 transgenic mice do not express the transgene in liver (5). As expected, there was no significant difference in the concentrations of nucleotide-linked hexoses or hexosamines between livers of control or GLUT1 transgenic mice, either in the fed or in the fasted state (Fig. 3).

GFAT activity in hind limb muscles of GLUT1 transgenic mice was 60 and 70% greater than in their wild-type littermates in the fasted and fed states, respectively (p < 0.01; Fig. 4). The increase in activity in transgenic mice was similar when it was assayed at different concentrations of Fru-6-P (2.4-6 m; the latter is a near saturating substrate concentration). GFAT activity was slightly (~30%) higher in muscles of ad libitum fed versus overnight fasted mice, in both control and GLUT1 transgenics, but the increase in the fed versus fasted state was significant (p < 0.05) only in the transgenic group.

As previously reported in rats (12), GFAT activity in mouse liver was much greater than that in muscle. Hepatic GFAT activity was similar in transgenic mice and controls. In liver extracts prepared from overnight fasted mice assayed in the presence of 6 m Fru-6-P, GFAT activity was 244.6 ± 46.3 pmol/mg protein/min in controls and 238.5 ± 24.5 (n = 5) in GLUT1 transgenic mice.

To establish whether the increased GFAT activity observed in muscles of GLUT1-overexpressing mice reflected increased mRNA expression, GFAT mRNA in muscle was quantified by RPA. Data were normalized to -actin mRNA expression, assayed in parallel. No significant difference in GFAT mRNA abundance was detected between RNA extracted from muscles of GLUT1-overexpressing mice and controls (Fig. 5).

GLUT4-overexpressing Mice

Because overexpression of GLUT1 (9, 10), but not that of GLUT4 (8, 10), caused insulin resistance, we also measured nucleotide-linked sugars and GFAT activity in muscles of mice overexpressing GLUT4. As shown in Fig. 6, the concentrations of UDP-HexNAc, UDP-Hex, GDP-mannose, and UDP were essentially identical in muscles of GLUT4 transgenic mice and their control littermates, in the fed and the fasting state. Furthermore, the concentrations of nucleotide sugars were similar in muscles of C57BL/6J (Fig. 6) and the B65JLFI/J controls (Fig. 2), except that in the fed state UDP-HexNAc concentrations tended to be lower (~40%; p < 0.01) in the C57BL/6J strain.

GFAT activity in muscle was not significantly different between controls and GLUT4-overexpressing mice in the fed or the fasted state (Fig. 7). GFAT activity in muscle tended to be higher in the fed versus the fasted state in both control and GLUT4 transgenic mice, but the nutritional effect was only significant in the latter group, assayed in the presence of 6 m Fru-6-P (p < 0.01).

Comparing GFAT activity measurements shown in Fig. 7 to those in Fig. 4, it is noteworthy that the activity in muscles of control mice was ~50% lower (p < 0.01) in C57BL/6J mice (Fig. 7) than in the B65JLF1/J strain (Fig. 4). This correlates with the UDP-HexNAc measurements discussed above and may reflect differences in genetic background since it was not accounted for by differences in age or sex, was consistently observed, and could not be attributed to differences in the assay procedure.

GFAT activity in liver was not significantly different between GLUT4 transgenic mice and controls. When assayed in the presence of 6 m Fru-6-P in liver extracts from fasted mice, it was 118.7 ± 12.7 pmol/mg of protein/min in controls and 145.3 ± 8.9 in transgenics (n = 4).

DISCUSSION

Muscles overexpressing GLUT1 demonstrated marked (2-3-fold) increases in the concentrations of all nucleotide sugars measured (UDP-Hex, UDP-HexNAc, and GDP-mannose), whereas these concentrations remained unchanged in muscles overexpressing GLUT4. Basal glucose transport is increased 2-8-fold in the former (9), but it is only mildly increased (20-50%) in the latter (8). Intracellular free glucose is increased in GLUT1-overexpressing (7) but not in GLUT4-overexpressing (8) muscles. Previous observations in GLUT1-overexpressing mice suggested that glucose transport is rate-limiting for glycolysis and glycogen synthesis in skeletal muscle (7). Our data suggest that glucose transport is also rate-limiting for the synthesis of nucleotide-linked hexoses and hexosamines.

In GLUT1-overexpressing mice, circulating insulin and glucagon concentrations are unchanged and glycemia is mildly reduced (5). In rats rendered markedly hyperglycemic and hyperinsulinemic by infusion of glucose, UDP-Hex concentrations in muscle are actually reduced (12). In the latter condition, enhanced UDP-glucose utilization, secondary to insulin-mediated glycogen synthase activation, likely exceeds the synthesis of UDP-glucose, in spite of accelerated glucose transport into the cell. In muscles overexpressing GLUT1, glycogen concentrations are markedly increased (~10-fold), exceeding that observed in any physiological model (7). Glycogen content is only mildly increased (~30%) in muscles overexpressing GLUT4 (8). Glucose 6-phosphate concentrations in muscle are unchanged in both transgenic models (7, 8). In GLUT1-overexpressing muscles, glycogen synthase activity (both total and activated) is decreased and glycogen phosphorylase is unchanged (7). Thus, the marked accumulation of glycogen in GLUT1-overexpressing muscles likely reflects the increased availability of UDP-Glc, the substrate of glycogen synthase. Taken together, the data suggest that UDP-Glc concentrations play an important role in modulating the rate of glycogen synthesis and may limit glycogen deposition under certain physiological conditions, e.g. when glycogen synthesis is activated by insulin.

Of particular interest was the observation that muscles overexpressing GLUT1 (9, 10) but not those overexpressing GLUT4 (8, 10) developed glucose transport insulin resistance although total GLUT4 expression was unchanged in the former (5). Since increased glucose flux via the hexosamine-synthetic pathway has been implicated in glucose-induced insulin resistance (12, 14, 15, 16, 17), we examined this parameter in muscles of both transgenic models. Our data support the concept that glucose flux via the hexosamine-synthetic pathway is increased in muscles overexpressing GLUT1 and not in those overexpressing GLUT4. In the former, but not in the latter, the concentration of the major product of the pathway, UDP-HexNAc, was increased 2-3-fold, and the activity of the rate-limiting enzyme for glucose entry into the pathway, GFAT, was increased by 60-70%. By RPA analysis, we were unable to detect a significant change in GFAT mRNA abundance in GLUT1-overexpressing muscles, suggesting that the increase in GFAT activity likely represents posttranscriptional regulation. Whether this reflects an increase in GFAT protein (i.e. enhanced translation or decreased degradation) or a posttranslational modification that activates the enzyme will have to be determined when suitable specific antibodies to GFAT become available. In the fungus Blastocladiella emersonii, the enzyme is regulated by reversible phosphorylation/dephosphorylation on serine (22, 23). Preliminary data indicate that mammalian GFAT activity may also be regulated by this mechanism (24). Our data suggest that chronic elevation of glucose flux can increase GFAT activity in muscle in vivo. Mechanisms of GFAT regulation need further study; this seems especially warranted by a recent report indicating that in skeletal muscles of patients with non-insulin-dependent diabetes who are insulin resistant, GFAT activity is increased (25).

Although the co-existence of two phenomena does not prove causality, our data are consistent with the hypothesis that the insulin resistance observed in GLUT1-overexpressing mice may be mediated by increased glucose flux via the hexosamine-synthetic pathway. Indeed the GLUT1-overexpressing mouse is the first in vivo model in which the two phenomena co-exist under physiological conditions. We have made similar observations in muscles of ob/ob mice, in which insulin resistance is accompanied by increased GFAT activity and markedly increased concentrations of UDP-HexNAc, but in contrast to GLUT1-overexpressing mice, UDP-Hex and GDP-mannose are minimally or not increased (26). In vitro, GFAT activity was found to be increased in myocytes cultured in the presence of high glucose or high insulin, and the two effects were additive (27).

The mechanism by which hexosamine products may cause insulin resistance is conjectural. As discussed previously (12, 15), products of the pathway are essential substrates for both N- and O-linked glycosylation, and changes in their absolute or relative concentrations may affect posttranscriptional processing of glycoproteins and/or glycolipids involved in signaling. The K_m of several glycosyl transferases is in the range of 10⁴-10⁵ , and their activity could be regulated in part by substrate concentrations (reviewed in Ref. 28). A group of cytosolic proteins which are modified by O-glycosylation with a single GlcNAc appear to be particularly interesting, since they include several transcription factors (29). Indeed, glucose-induced transforming growth factor- expression in vascular smooth muscle cells appears to be mediated by products of the hexosamine synthesis pathway (30).

The reason that there are such profound phenotypic differences between GLUT1-overexpressing and GLUT4-overexpressing muscles is unclear. The absolute magnitude of overexpression of the respective glucose transporters in muscle was similar in the two models; because skeletal muscle natively expresses much more GLUT4 than GLUT1, expression of the transgenes resulted in a >10-fold increase in GLUT1 versus a 2-4-fold increase in GLUT4 (10). However, since GLUT1 is predominantly localized on the cell membrane and under basal conditions GLUT4 is largely sequestered in an intracellular compartment, muscles of GLUT1-overexpressing mice exhibit increased glucose transport continuously, whereas GLUT4 overexpressers manifest it intermittently in response to meal-induced insulin secretion. Although the overexpression of both transgenes was observed in all skeletal muscles examined (10), the transgenes were controlled by different promoters, and their relative expression in muscles of different fiber type may vary. The human GLUT4 transgene was controlled by its own promoter, which is natively more highly expressed in muscles rich in slow oxidative fibers (31, 32), whereas GLUT1 overexpression was under the control of the rat myosin light chain 2 promoter, which may be preferentially expressed by fast twitch, glycolytic fibers (33). We have no information concerning possible differences in the activity of the hexosamine-synthetic pathway among muscles with different fiber types. The different genetic background of the mice used may also have contributed to the phenotypic differences between the two transgenic models. Finally, as discussed previously (3, 10), there may be subcellular compartmentation, and the metabolic fate of glucose transported on GLUT4 may differ from that transported on GLUT1.

In summary, our data indicate that the chronic, sustained acceleration of glucose transport into muscle, as observed with GLUT1 overexpression, results in marked accumulation of nucleotide-linked sugars and enhanced GFAT activity. The increase in UDP-Hex likely accounts for the marked accumulation of muscle glycogen, whereas the accumulation of products of the hexosamine-synthetic pathway may contribute to the insulin resistance observed in this model.