Normalization of Skeletal Muscle Glycogen Synthesis and Glycolysis in Rosiglitazone-Treated Zucker Fatty Rats: An In Vivo Nuclear Magnetic Resonance Study

Zucker obese (fa/fa) rats and their lean littermates (Harlan, Indianapolis, IN) were housed in an environmentally controlled room with a 12-h light/dark cycle. Obese and nonobese rats (8–10 weeks old) were placed on a pair-fed diet (#5L35; LabDiet, Richmond, IN) consisting of 100 kcal/day for 5–10 days before surgery. Rats were chronically catheterized (carotid artery and jugular vein) as described elsewhere (22), and the feeding regimen was continued for an additional 7 days. Rats were placed into three groups: fatty rosiglitazone (FRSG) (n = 12), fatty control (FC) (n = 9), and lean control (LC) (n = 8). Immediately upon completion of the surgical procedure, rats were begun on a 7-day drug treatment protocol. The FRSG group received 3 mg/kg rosiglitazone (rosiglitazone maleate) via oral gavage once daily, and the control groups received vehicle (sterile water). Catheters were kept patent by flushing 1–3 days before the infusion experiment.

All rats were fasted overnight before the infusion experiment. Rats were acclimated to restraint by being positioned in restraining tubes for 1 and 2 h, respectively, in the 2 days before the experiment. On the day of the experiment, rats were placed into a customized restraining tube that allowed their left hindlimb to be secured to the outside of the tube in a manner to limit free movement of the leg for NMR measurements. The rats were transiently anesthetized (<30 s) with 2.0% isoflurane (Abbott Laboratories, Chicago) so they could be placed in the restraining tube. A euglycemic-hyperinsulinemic (10 mU · kg⁻¹ · min⁻¹ Humulin Regular; Eli Lilly, Indianapolis, IN) clamp was begun using 1,6-¹³C₂ glucose (99% enriched, 20% wt/ vol; Cambridge Isotope Laboratories, Cambridge, MA) at 2.5 min after the commencement of the primed continuous insulin infusion. Plasma glucose concentrations were clamped between 7 and 8 mmol/l. All clamps lasted for 150 min. Blood samples were drawn during the baseline NMR measurement, at 7.5 min, 15 min, and every 15 min thereafter for immediate assessment of plasma glucose and lactate concentrations. At the end of the in vivo NMR experiment, rats were anesthetized with an intravenous dose of 50 mg/kg Nembutal (Abbott Laboratories). Superficial skin was rapidly removed from the left hindquarter followed by in situ freeze clamping of the gastrocnemius and biceps femoris muscles. Rats were killed with a lethal dose of Nembutal.

All in vivo ¹³C and ³¹P NMR experiments were performed on a Bruker ABX system (horizontal/40-cm diameter bore, 4.7-Tesla field strength magnet; Bruker Medical, Billerica, MA). Both ¹³C observe/¹H decouple and ³¹P observe NMR spectroscopy were performed using a triple tune radio frequency probe consisting of concentric surface coils (the outer ¹H coil [42 mm] tuned to 200.21 MHz, and the inner dual frequency ¹³C or ³¹P coil [25 mm] tuned to 50.34 and 81.05 MHz, respectively). The rat hindlimb was positioned over the ¹³C/³¹P coil (vertical in plane) and placed in the magnet isocenter. Because of the intrinsic low ¹³C/³¹P sensitivity in the hindlimb experiments, it was necessary to measure the bulk signal from the larger tissue beds of mixed fiber type including the gastrocnemius and biceps femoris. ¹H decoupled-¹³C NMR spectroscopy was performed in the following manner: an initial frequency selective sinc pulse (20 ms) set on the low-field side of the methylene carbon of lipids at 30 ppm was immediately followed by a nonselective hard pulse (∼70° flip angle, 7 mm from surface coil). The sinc pulse power was adjusted to eliminate most of the signals in that region. Broadband ¹H Waltz-8 decoupling was applied during acquisition, and additional Nuclear Overhauser Enhancement was achieved using low-power decoupling during the relaxation delay (TR = 0.5 s, SW = 10 KHz, 2 K data). A 15-min baseline spectrum was followed by subsequent 15-min acquisitions throughout the duration of the experiment. ¹³C NMR spectra were processed semiautomatically using a Gaussian filter and baseline flattening before peak integration (Nuts NMR processing software; Acorn NMR, Fremont, CA).

³¹P NMR was performed using a hard pulse (45° flip angle) that was optimized 7 mm from the surface coil (TR = 1 s, SW = 3 KHz, 2 K data). Two 10-min spectra were acquired: one at baseline and the other at 140–150 min. ³¹P NMR spectra were processed using Gaussian resolution enhancement. In all spectra, the phosphocreatine (PCr) peak was set to 0 ppm, and β-ATP and G-6-P peaks were integrated using Nuts NMR processing software (21). [G-6-P] was extrapolated from the in vivo spectra after correction for differential saturation with respect to the β-ATP peak and enzyme assay-derived ATP concentrations (modified diagnostic kit no. 366; Sigma, St. Louis, MO).

The creatine/PCr peak (54.4 ppm) in the ¹³C spectrum and γ-ATP (−16.0 ppm) in the ³¹P spectrum were used as an internal reference standard for movement of the leg. If severe movement compromised these peak integrals, the experiment was terminated.

Muscle tissue extracts were prepared for high-field NMR analysis by homogenizing ∼0.5 g skeletal muscle as previously described (21). NMR analysis was performed at 9.4 Tesla (Bruker WB-400 AMX NMR spectrometer; Bruker). Proton-observed carbon-enhanced spectroscopy was performed on tissue extract samples for fractional enrichment calculations (21). For quantitation, a correction factor was calculated when a TR (pulse repetition time) equaling 30 s was used. Alanine concentration was quantitated by comparing its signal intensity with a known internal concentration standard (lactate), which was measured in tissue extracts using a 2300 STAT PLUS analyzer (Yellow Springs Instruments, Yellow Springs, OH).

Plasma glucose concentrations were measured using a 2300 STAT PLUS analyzer. Plasma immunoreactive free insulin was measured with a double-antibody radioimmunoassay technique (Linco Research, St. Charles, MO). The 1,6-¹³C₂ enrichment of plasma glucose was determined using an LC/MS (liquid chromatography/mass spectrometry) system (Agilent Technologies LC/MSD 1100) (23). The MS electrospray chamber conditions were set for positive ionization. Selected ion monitoring was used to detect the sodium adducts (M⁺23⁺) of each glucose isotope, with a m/z = 203 ion corresponding to the sodium adduct of normal glucose and a m/z = 205 ion corresponding to the sodium adduct of 1,6-¹³C₂ glucose. Standard curves were generated (0–97% 1,6-¹³C₂ glucose atom percent excess [APE]). Plasma samples were extracted using ice-cold acetone, and 5 μl was injected into the column.

Glycogen 1,6-¹³C₂ fractional enrichments were determined using the isolated glycogen after perchloric acid extraction (21), and absolute glycogen concentrations were measured using a separate portion of muscle as previously described (24). Intramyocellular glucose was calculated by measuring free glucose in the tissue extracts and correcting for extracellular space using an intra-/extracellular space ratio of 9.0 and plasma glucose assumed to be equal to the extracellular space glucose (25). Plasma triglyceride concentrations were determined enzymatically using an Olympus AU600 analyzer (Olympus Optical, Melville, NY).

Endogenous glucose production (EGP) was estimated by taking the difference of the glucose infusion rate (G_inf) and the tracer-determined whole-body glucose disposal rate (GDR). Whole-body GDR was calculated from G_inf and the steady-state plasma 1,6-¹³C₂ glucose enrichment at 120 min.

The incremental change in C-1 glycogen peak intensity from 1,6-¹³C₂ glucose incorporation was measured at 100.5 ppm. Incremental plasma glucose ¹³C fractional enrichment as well as final glycogen ¹³C enrichment and concentrations were used to back-extrapolate the glycogen concentration (μmol/g, which represents micromoles glucosyl units per gram muscle wet weight) at each measured time point to baseline, as described by Bloch et al. (26). Glycogen synthesis rates were determined using a linear regression analysis over the individual time point glycogen concentrations.

Metabolic steady-state conditions were assumed for calculating carbon flux through the glycolytic pathway into the intermediate triose pool of lactate, pyruvate, and alanine (Fig. 1).

Differential equations were developed from steady-state rate equations and solved for glycolytic flux (V_gly) (27). ¹³C label incorporation from 1,6-¹³C₂ glucose into 3-¹³C lactate and 3-¹³C alanine in the hindlimb muscles may be used as an indirect marker of pyruvate labeling. Therefore, the 3-¹³C lactate and 3-¹³C alanine turnover rate provides a qualitative index of V_gly. The absolute flux was calculated by simultaneously solving the differential equations describing the V_gly using CWave software (Dr. Graeme F. Mason, Yale University). The parameters required for the modeling of V_gly include plasma 1,6-¹³C₂ glucose, tissue 3-¹³C lactate and 3-¹³C alanine turnover curves, and lactate and alanine concentrations. A limitation of this measurement is that the turnover of the plasma glucose pool must be faster than that of the lactate and alanine pools. Therefore, these measurements can only be made under hyperinsulinemic conditions when plasma glucose turnover is increased (27).

The traditional method by which skeletal muscle V_gly is measured in vivo is an indirect one. It is calculated as the difference between glucose uptake rates as assessed using the 2-deoxyglucose method and glycogen synthesis rate as assessed using ³H or ¹⁴C glucose (28). Whole-body glycolysis may be calculated using the ³H glucose washout method (29).

All data are reported as means ± SE. ANOVA was performed on data to determine the significance at a minimum threshold of P < 0.05 between the three groups. A multiple comparison Newman-Keuls post hoc test was used when necessary to determine significance between the groups.

As a result of the pair-feeding protocol, there were no differences in weight between the FC (350 ± 4 g) and FRSG (345 ± 5 g) groups. However, both groups weighed significantly more than the LC group (291 ± 3 g, P < 0.001). The weight gain during the treatment period was 29 ± 3, 30 ± 2, and 11 ± 1 g in the FC, FRSG, and LC groups, respectively (P < 0.001 LC vs. FC and FRSG). Surprisingly, there were no differences in basal insulin concentrations between the FRSG (428 ± 77 pmol/l) and FC (475 ± 111 pmol/l) groups, but both groups had significantly higher concentrations than the LC group (125 ± 28 pmol/l, P < 0.05). Basal plasma glucose in the FRSG group (8.7 ± 0.6 mmol/l) was significantly reduced compared with that of the FC group (13.5 ± 0.8 mmol/l, P < 0.001) and was normalized to the LC group (7.5 ± 0.4 mmol/l). Basal plasma triglyceride concentration in the FRSG group (114 ± 12 mg/dl) was also lower than that in the FC group (273 ± 37 mg/dl, P < 0.001) but higher than that in the LC group (32 ± 3 mg/dl, P < 0.01).

During the euglycemic-hyperinsulinemic clamp experiment, plasma glucose concentrations were maintained at ∼7–8 mmol/l (Table 1). Although steady-state insulin concentrations during the euglycemic-hyperinsulinemic clamp were similar in the FC and FRSG groups, they were significantly higher than those in the LC group (Table 1). Glucose turnover was increased in the FRSG group after treatment. This may be visualized in Fig. 2 because the plasma 1,6-¹³C₂ glucose APE increased rapidly during the clamp in the LC and FRSG groups but was slower in the FC group. G_inf at 120 min was significantly higher in the FRSG group (20.3 ± 1.9 mg · kg⁻¹ · min⁻¹) than in the FC group (13.8 ± 1.4 mg · kg⁻¹ · min⁻¹, P < 0.05) but lower than in the LC group (30.1 ± 1.8 mg · kg⁻¹ · min^{− 1}, P < 0.01; Table 1). EGP did not differ in the three groups; therefore, whole-body GDR was also higher in the FRSG group than in the FC group. However, whole-body GDR in the FRSG group was not normalized to the LC group (Table 1).

Baseline subtracted ¹³C NMR spectra depicting ¹³C label turnover in a LC rat during a euglycemic-hyperinsulinemic clamp is shown in Fig. 3. The β and α anomer peaks of 1-¹³C glucose appear at 96.8 and 93.0 ppm, respectively, and the large peak slightly downfield at 100.5 ppm corresponds to the C-1 glucosyl unit of the glycogen polymer. The C-6 glucosyl unit of the glycogen polymer resides at 61.4 ppm and is significantly larger than C-1 because of the overlapping 6-¹³C peak. 3-¹³C lactate and 3-¹³C alanine may also be observed at 21.0 and 16.9 ppm, respectively. The absolute steady-state V_glyc was significantly reduced in the FC group (16.6 ± 12.8 nmol · g⁻¹ · min^{− 1}) versus the LC group (75.2 ± 15.3 nmol · g^{− 1} · min⁻¹, P < 0.05) but was normalized in the FRSG group (56.3 ± 11.5 nmol · g^{− 1} · min⁻¹, P < 0.05 vs. FC, Table 1). The best-fit label turnover curves to 3-¹³C pyruvate are shown in Fig. 4. Empirically, the 3-¹³C pyruvate turnover in the LC and FRSG groups is more rapid than in the FC group. This increased label turnover in the LC and FRSG groups versus the FC group may be used as a qualitative marker of increase flux through glycolysis in these groups (24). Absolute V_gly was calculated using the metabolic steady-state turnover kinetics of ¹³C label in 3-¹³C lactate and 3-¹³C alanine pools. The benefit of using the doubly labeled (1,6-¹³C₂) glucose precursor is to double the 3-¹³C enrichment in lactate and alanine, thereby providing greater NMR sensitivity. V_gly was also significantly reduced in the FC group (18.8 ± 8.6 nmol · g⁻¹ · min⁻¹) versus the LC group (56.2 ± 16.6, P < 0.02) but was normalized in the FRSG group (52.9 ± 9.1 nmol · g⁻¹ · min⁻¹, P < 0.02 vs. FC, Table 1). Therefore, the decreased glucose disposal observed in the FC group was associated with an ∼78% reduction in glycogen synthesis flux and an ∼67% reduction in V_gly, whereas these fluxes in the FRSG group were normalized to the LC group. Although muscle glycogen synthesis and glycolysis fluxes are normalized after rosiglitazone treatment, whole-body GDR is only partially normalized.

Figure 5A illustrates a ³¹P NMR baseline spectrum with G-6-P (part of the phosphomonoester region), inorganic phosphate (both extracellular and intracellular), PCr (assigned to 0 ppm), and α-, β-, and γ-ATP indicated. The basal G-6-P concentrations were the same in all groups before the euglycemic-hyperinsulinemic clamp (165 ± 12, 152 ± 29, and 147 ± 29 nmol/g in the LC, FRSG, and FC groups, respectively; Fig. 5B). At 150 min, [G-6-P] increased in both the LC and FRSG groups (16.9 ± 5.8% and 13.0 ± 11.1% vs. baseline, respectively) but decreased in the FC group (−23.3 ± 13.4% vs. baseline, P < 0.05 vs. LC and FRSG groups, Fig. 5B).

There were no differences in intramyocellular glucose after correction for intra-/extracellular glucose space (0.88 ± 0.32, 0.38 ± 0.18, and 0.48 ± 0.26 mmol/l in LC, FRSG, and FC groups, respectively).

The data reported here show that after 7-day rosiglitazone treatment in the Zucker fatty rat, insulin-stimulated whole-body glucose disposal was increased and muscle glycogen synthesis and glycolysis were normalized. Although TZDs, including rosiglitazone, have been shown to improve glycemic control in animal models of type 2 diabetes by reducing insulin resistance in muscle (1,3,4), liver (1,4), and adipose tissue (5,6), in vivo characterization of the glucose metabolic fate in these tissues has been difficult to achieve. In the present study, an in vivo NMR spectroscopic technique was used to noninvasively measure insulin-stimulated glucose disposal in the hindlimb muscles of an awake rat.

Using in vivo ¹³C NMR spectroscopy, glycogen synthesis was shown to be increased in skeletal muscle of either type 2 diabetic (12) or insulin-resistant obese (30) subjects treated with troglitazone. Although whole-body glucose oxidation as measured by indirect calorimetry was increased in the troglitazone-treated type 2 diabetic subjects, the skeletal muscle V_gly was not ascertained (12). The direct in vivo effect of acute troglitazone stimulation on glycolysis was measured in perfused Sprague-Dawley rat hindlimb (31). The investigators showed that in the presence of troglitazone alone (no exogenous insulin), O₂ consumption and lactate/pyruvate release were increased with no exogenous fatty acids present in the perfusate. Therefore, it was postulated that both aerobic and nonaerobic glycolysis was stimulated by the TZD. However, these hindlimb perfusion experiments were not performed in chronically TZD-treated or insulin-resistant animal models.

The necessity for an in vivo noninvasive measurement of both glycogen synthesis and glycolysis becomes increasingly apparent in the presence of differing results with regard to the glucose metabolic fate as measured in isolated muscle strips after in vitro TZD incubation (32,33) or chronic oral TZD treatment (34–37). Fürnsinn et al. (32) showed that in vitro troglitazone incubation elicited a glucose catabolic response in excised Sprague-Dawley soleus muscle strips. Using U-¹⁴C glucose tracer techniques for flux measurements, both glycogen synthesis and glucose oxidation were decreased after insulin administration, whereas anaerobic glycolysis (measured as lactate release) increased to the extent that the anaerobic/aerobic glycolysis ratio was ∼50:1. However, these data contrast with a number of studies that have examined the in vivo effects of chronic oral TZD treatment on skeletal muscle glucose metabolism (glycogen synthesis and glycolysis) in insulin-resistant rodent models (34–37). Although direct in vivo measurements of these fluxes were not performed, these studies nevertheless show that TZDs can act by various degrees to normalize both glycogen synthesis and glycolysis. After 4-day englitazone treatment in ob/ob mice, both insulin-stimulated glycogen synthesis and glycolysis were normalized, as assessed in incubated soleus muscle strips using 1-¹⁴C glucose to measure glycogen synthesis and glucose oxidation and 3-³H glucose to measure glycolysis (34). An interesting result of this study was that basal glucose oxidation was increased after englitazone treatment and normalized to the lean group. More recently, Sreenan et al. (35) showed that the increased insulin-stimulated glucose disposal, as measured using 2-deoxyglucose in 6-week troglitazone-treated ZDF rat soleus muscle, was also due to an increase in both glycogen synthesis and glycolysis. Also of note was that troglitazone almost doubled the basal glucose oxidation flux in these rats. This result correlated with an enhancement of muscle pyruvate dehydrogenase activity under basal conditions in the study. Fürnsinn et al. (36) examined the effects of treatment with the TZDs BM13.1258 and BM15.2054 on glucose intermediary metabolism. In this study, insulin-stimulated glycogen synthesis, as measured in isolated soleus, was increased after 10-day oral treatment of either TZD in Zucker fatty rats, yet glucose transport and oxidation was only increased in the BM13.1258-treated rats. These results suggest that different TZDs exhibit different responses with regard to glycogen synthesis and glucose oxidation and therefore may control glucose metabolism via multiple mechanisms.

Further to the notion that TZDs may exhibit control of glucose disposal through multiple mechanisms is that TZDs appear not to exhibit the same degree of insulin sensitivity in all tissue. Although both glycogen synthesis and glucose metabolism in hindlimb skeletal muscle of Zucker fatty rats were normalized after 7-day rosiglitazone treatment in the present study, whole-body GDR was only partially normalized (73% of control). This discrepancy is most likely due to a noneffect of rosiglitazone on insulin-stimulated glucose disposal in other tissue (e.g., heart and adipose tissue) (1).

Measurements of [G-6-P] along with glycogen synthesis and V_gly can be used to provide an index of metabolic control (12). For example, as insulin stimulated hexokinase and glycogen synthase in the FC group, glucose transport could not keep up with the demand, which resulted in a decrease in [G-6-P] versus the LC and FRSG groups. This hypothesis holds true if intracellular glucose were held constant or lower in the FC versus the LC and FRSG groups. Although there were no significant differences between the groups, there was large variability in our intracellular glucose measurement. Therefore, it is difficult to delineate the role of glucose transport versus hexokinase in the control of glucose uptake in this model. Although there is as yet no direct evidence for an effect of rosiglitazone on GLUT4 expression or function in skeletal muscle, rosiglitazone corrects expression levels of the skeletal muscle proteins cellubrevin, VAMP-2, and syntaxin-4, which are known to be associated with GLUT4 vesicle trafficking in skeletal muscle of hyperglycemic ZDF rats (38).

In summary, a combined in vivo ¹³C and ³¹P NMR spectroscopic approach was used to measure skeletal muscle glucose uptake and its distributed fluxes (glycogen synthesis and glycolysis) and G-6-P (a key metabolic intermediate) during a euglycemic-hyperinsulinemic clamp in awake Zucker fatty rats. The data suggest that the increased insulin-stimulated glucose disposal in skeletal muscle associated with rosiglitazone treatment was attributed to a normalization of both glycogen synthesis and glycolysis.

We are indebted to Kimberly Brown, Diane McGee, and Kathleen Morasco for their expert technical assistance. We are also grateful to Peter Brown, Yale University School of Medicine, for radio frequency antenna design and construction.