Weight Loss-Induced Plasticity of Glucose Transport and Phosphorylation in the Insulin Resistance of Obesity and Type 2 Diabetes

Obese subjects with and without type 2 diabetes with no medical contraindication to a very-low-calorie diet were recruited after obtaining informed consent. Subjects taking insulin, taking medications known to adversely affect glucose homeostasis, or reporting weight change of >6 kg during the prior 6 months were excluded. Subjects underwent a screening medical visit including laboratory testing. Type 2 diabetic subjects discontinued oral diabetic agents and participated if their glucose levels remained <300 mg/dl. The University of Pittsburgh institutional review board approved all study procedures.

Type 2 diabetic subjects (8 women, 1 man) were 51 ± 2 years old, and subjects with normal glucose tolerance (7 women, 2 men) were 45 ± 2 years old (P = 0.07 compared with diabetes). At baseline (Table 1), both groups were of comparable weight and BMI.

Subjects met weekly with a dietitian, received an 800 kcal/day liquid meal replacement (Optifast Formula) for 12 weeks, and consumed 1,200 kcal per day during weeks 13–14. Before clamp studies, subjects followed a weight maintenance diet for at least 2 weeks, maintained an intake of at least 200 g carbohydrate for at least 3 days, and avoided exercise or strenuous exertion for 2 days.

Subjects were admitted to the University of Pittsburgh General Clinical Research Center, received a standardized dinner, and fasted overnight. An antecubital venous catheter was placed for infusion of insulin, dextrose, and PET tracers, and a radial artery catheter was placed for arterial samples. During a 4-h insulin infusion (40 mU/m² per min), euglycemia (90 mg/dl) was maintained with 20% dextrose. Plasma insulin levels were obtained every 30 min. Subjects were moved into the PET/computed tomography (CT) scanner 2 h after starting the insulin infusion.

The SMART scanner, a combined PET (ECAT ART) and helical CT tomograph (SoMatom AR.SP; Siemens) scanned a 16.2-cm-long segment of both calves (25). The CT was used for attenuation correction of the PET image (26) and placement of regions of interest within corresponding regions of skeletal muscle imaged with PET as previously described (17).

For blood flow assessment, a bolus injection of [¹⁵O]H₂O (45 mCi) was given, and PET imaging lasted 4 min (24 × 10 s time frames). Arterial blood sampling, controlled by a peristalsis pump and quantitated with a blood radioactivity monitor, was 10 ml/min for 90 s and then reduced to 2 ml/min.

Allowing 10 min for decay of the [¹⁵O]H₂O signal, a bolus injection of [¹⁸F]FDG (4 mCi) was given, and 28 PET frames (2 × 30 s, 8 × 15 s, 4 × 30 s, 3 × 1 min, 1 × 2 min, 10 × 5 min) were obtained over 60 min. Arterial blood sampling for [¹⁸F]FDG consisted of 29 hand-drawn 0.5 ml samples from a radial artery catheter (10 × 6 s, 8 × 15 s, 3 × 3 min, 5 × 10 min) over 62 min.

The [¹⁵O]H₂O data were analyzed using a one-tissue compartment model (27). Two parameters were estimated: 1) muscle blood flow (MBF; ml · min⁻¹ · ml⁻¹) and 2) water efflux from the tissue, expressed as λ/MBF (min⁻¹), where λ is the distribution constant of water in the tissue. The [¹⁸F]FDG kinetics were assessed with both the 5K muscle-specific model and the three-compartment/three-rate constant (3K) model originally developed for dynamic PET studies of the brain and often used for skeletal muscle studies.

The 5K model (Fig. 1A) more specifically delineates the parameters of [¹⁸F]FDG transport and phosphorylation in skeletal muscle by accounting for three kinetic steps: 1) transport from capillaries to extracellular space; 2) transport from extracellular space to intracellular space; and 3) intracellular phosphorylation (20,21). The five-rate constants of the model represent the following: k₁ (ml · ml⁻¹ · min⁻¹), delivery to extracellular space; k₂ (min⁻¹), efflux from the extracellular space to plasma; k₃ (min⁻¹), transport into skeletal muscle; k₄ (min⁻¹), efflux from skeletal muscle to extracellular space; and k₅ (min⁻¹), phosphorylation within skeletal muscle.

From the estimates of the five-rate constants, three additional parameters were calculated: fractional uptake of [¹⁸F]FDG, K (ml · ml⁻¹ · min⁻¹),

and control coefficients of transmembrane transport, C^T = k₅/(k₄ + k₅), and transmembrane phosphorylation, C^P = k₄/(k₄ + k₅). The biochemical principle behind C^T and C^P is that after entry, glucose can either be trapped by phosphorylation (k₅) or exit (k₄) the cell as free glucose. If all glucose entering a myocyte is phosphorylated, transport is considered to limit the overall rate of uptake and the locus of control. In contrast, if the majority of glucose that enters the cell exits the cell without being phosphorylated, then phosphorylation can be considered the rate-limiting step or locus of control.

The 3K model (Fig. 1B), originally applied to [¹⁸F]FDG kinetics in the brain, can be described by using three-rate constants: k₁′ (ml · ml⁻¹ · min⁻¹), delivery to tissue; k₂′ (min⁻¹), efflux of nonmetabolized [¹⁸F]FDG from tissue to plasma; and k₃′ (min⁻¹), phosphorylation within tissue. From the model one can also calculate K (ml · ml⁻¹ · min⁻¹),

As previously reported (20,21), the k₁′ parameter primarily reflects blood flow, and the k₃′ parameter reflects the combined steps of glucose transport and phosphorylation.

Compartmental modeling assumes the arterial plasma time course of tracer activity as a model input function. The parameters of the [¹⁵O]H₂O and the 5K and the 3K models were estimated by weighted nonlinear least squares (28), with weights optimally chosen (20,28). The 5K model was successfully identified in five type 2 diabetic and four obese subjects before weight loss, and in four type 2 diabetic and nine obese subjects after weight loss. In the remaining subjects, k₄ was imprecisely resolved because of data with a high signal-to-noise ratio. In these subjects, a Bayesian approach incorporated available a priori knowledge of k₄ and identified the 5K model with optimal weights by maximum a posteriori probability Bayes estimation (29,30), assuming k₄ as a Gaussian variable with mean and standard deviation equal to the mean and standard deviation values of the k₄ estimates obtained in the remaining subjects.

Maximal aerobic capacity (Vo_2max) was assessed using a standard incremental protocol with an electronically braked cycle ergometer (ERG 601; Bosch, Stuttgart, Germany) (7).

Data are expressed as the means ± SE. ANOVA was used to compare subjects with normal glucose tolerance and those with type 2 diabetes. Weight loss changes were compared using paired t tests. All statistics were performed using SPSS 10.0 for Windows. A P value of <0.05 was considered significant.

Weight loss was comparable in type 2 diabetic and obese subjects (15 ± 2 and 13 ± 2 kg for diabetic and obese subjects, respectively; P = 0.40 for comparison between groups), representing a mean weight loss of 16 ± 2 and14 ± 2%, respectively. Pre- and post-weight loss clinical characteristics are shown in Table 1. Both groups remained overweight after weight loss (BMI 30.0 ± 0.5 and 29.6 ± 1.5 kg/m² in diabetic and obese subjects, respectively), and no individual achieved a BMI of <25 kg/m². In subjects with type 2 diabetes, levels of HbA_1c, fasting glucose, and fasting insulin were elevated compared with obese subjects and declined with weight loss. Basal free fatty acid levels were slightly elevated compared with obese subjects (P = 0.06) and declined significantly with weight loss. In obese subjects, weight loss did not change values of HbA_1c, fasting glucose, or free fatty acids, but fasting levels of insulin declined significantly (P < 0.05). Maximal aerobic capacity was lower in type 2 diabetic subjects at baseline and did not change with weight loss in either group. During the euglycemic-hyperinsulinemic clamp (Table 2), type 2 diabetic and obese subjects had comparable levels of glucose and insulin before and after weight loss. Free fatty acids were elevated in type 2 diabetic subjects, and they declined significantly in both groups after weight loss. Total-body glucose disposal was reduced in type 2 diabetic subjects before weight loss (P < 0.001) and improved significantly (P < 0.01) with weight loss. Total-body glucose disposal also improved in obese subjects, but this improvement did not reach statistical significance (P = 0.14).

The dynamic PET imaging data upon which any modeling results are based are the tissue time-activity curves. Representative curves for [¹⁸F]FDG in skeletal muscle are shown in Fig. 2. When comparing the patterns of [¹⁸F]FDG activity at baseline, uptake occurred more slowly in type 2 diabetes and reached a lower value (Fig. 2A) compared with obesity (Fig. 2B). After weight loss, the tissue curves were markedly increased and reshaped in type 2 diabetes (Fig. 2C) and obesity (Fig. 2D). These changes were particularly evident in type 2 diabetes after weight loss because the curves had an upward convexity that was largely absent in the pre-weight loss study.

Effects of weight loss on skeletal muscle blood flow and [¹⁸F]FDG parameters are shown in Fig. 3. Fractional skeletal muscle [¹⁸F]FDG uptake, K, (Fig. 3A) was reduced at baseline in type 2 diabetic compared with obese subjects (P < 0.01). With weight loss, K tripled in type 2 diabetes (P < 0.01) and doubled in obesity (P < 0.01). Despite these marked differences in skeletal muscle [¹⁸F]FDG uptake, values for skeletal muscle blood flow assessed with [¹⁵O]H₂O (Fig. 3B) were comparable at baseline and did not change with weight loss in either group. Likewise, weight loss did not significantly affect k₁ (Fig. 3C) or k₂ (0.33 ± 0.06 and 0.25 ± 0.07 min⁻¹ in type 2 diabetic subjects before and after weight loss; 0.34 ± 0.09 and 0.32 ± 0.07 min⁻¹ in obese subjects before and after weight loss), parameters representing inward and outward delivery of [¹⁸F]FDG from the vascular to the extracellular space. Values for blood flow determined by [¹⁵O]H₂O highly correlated with k₁ from [¹⁸F]FDG analysis (r = 0.63, P < 0.01), consistent with the concept that k₁ derived from 5K [¹⁸F]FDG data quantification reflects blood flow (20,21).

The effects of weight loss on transport (k₃) and phosphorylation (k₅) parameters are shown in Fig. 3D and 3F. Before weight loss, k₃ was reduced in type 2 diabetic subjects compared with obese subjects, but it did not reach statistical significance (P = 0.10). After weight loss, k₃ increased approximately twofold in both type 2 diabetes and obesity, with the increase in obesity reaching statistical significance (P < 0.01). The k₃ after weight loss in type 2 diabetes remained lower than in obesity, although this difference was not statistically significant. The rate constant for phosphorylation, k₅, was reduced in type 2 diabetic subjects at baseline compared with obese subjects (P < 0.01) (Fig. 3F). After weight loss, k₅ increased threefold in type 2 diabetes (P < 0.001) and by >50% in obesity (P = 0.05). After weight loss, k₅ was comparable in type 2 diabetes and obesity. The rate constant for outward glucose transport from skeletal muscle, k₄, was not significantly different in type 2 diabetes and obesity at baseline (Fig. 3E), but it decreased by more than threefold in both groups after weight loss (P = 0.08 for type 2 diabetic and P < 0.05 for obese subjects).

Before weight loss, control of glucose uptake was approximately equally distributed between transport and phosphorylation in type 2 diabetic subjects (50 ± 8% C^T, 50 ± 8% C^P) (Fig. 4A), whereas in obesity the control of glucose uptake resided more at the step of glucose transport (66 ± 6% C^T, 34 ± 6% C^P) (Fig. 4B). After weight loss, both groups had marked improvements in skeletal muscle glucose transport and phosphorylation, and the distribution of control of skeletal muscle glucose metabolism shifted strongly toward transport in both type 2 diabetes (91 ± 10% C^T, 9 ± 6% C^P) (Fig. 4C) and obesity (93 ± 1% C^T, 7 ± 1% C^P) (Fig. 4D). These findings suggest that after weight loss, phosphorylation efficiency improved so much that phosphorylation was probably no longer rate constraining, and the rate-limiting step for glucose metabolism after weight loss could largely be attributed to glucose transport.

Using the same [¹⁸F]FDG activity curves, the 3K model was also applied. The fractional uptake of [¹⁸F]FDG (Fig. 5A) exhibits the same trend demonstrated with the 5K analysis of the data. The parameters k₁′ and k₂′ (Fig. 5B and C) did not change with weight loss in type 2 diabetic or obese subjects. In addition, the k₁′ values were highly correlated with skeletal blood flow values obtained from [¹⁵O]H₂O analysis (r = 0.71, P < 0.001), consistent with the interpretation that these values largely reflect blood flow in skeletal muscle. Parameter k₃′ (Fig. 5D), interpreted as an aggregate parameter reflecting both skeletal muscle glucose transport and phosphorylation, was reduced by 50% in type 2 diabetic subjects compared with those with obesity alone at pre-weight loss studies (P < 0.05) (Fig. 5). With weight loss, k₃′ increased fivefold in type 2 diabetic subjects (P < 0.05) and threefold in obese subjects (P < 0.01).

In skeletal muscle, impaired glucose transport and phosphorylation are prominent manifestations of insulin resistance (13,15–17,19). The current investigation assessed the plasticity of these impairments in response to weight loss. Blood flow and glucose metabolism assessments with PET imaging provided a unique integrated physiological profile of the dynamics of substrate delivery, glucose transport, and phosphorylation in human skeletal muscle in response to substantial weight loss. Weight loss in obese subjects with normal glucose tolerance induced a twofold increase in insulin-stimulated [¹⁸F]FDG metabolism, whereas in type 2 diabetic subjects, there was a threefold increase. Equally as dramatic were the marked changes in the configuration of the tissue time-activity curves for [¹⁸F]FDG metabolism in skeletal muscle. To examine whether changes in tissue perfusion induced changes in glucose metabolism with weight loss, we used PET imaging with [¹⁵O]H₂O. Consistent with our prior studies using venous occlusion plethysmography (8), weight loss increased glucose uptake in the absence of an increase in blood flow. Thus, the changes in [¹⁸F]FDG tissue time-activity curves can be deduced to reflect substantial effects of weight loss on transport and phosphorylation because the metabolism of [¹⁸F]FDG is substantially limited to these proximal steps of metabolism.

Two physiologically based compartmental modeling approaches of the tissue time-activity curves were used to better understand the specific effects of skeletal muscle glucose transport and phosphorylation. The 5K model, specifically developed for quantification of [¹⁸F]FDG skeletal muscle images, indicated that weight loss had effects on both glucose transport and phosphorylation. The phosphorylation parameter was decreased in type 2 diabetic subjects before weight loss, and it was fully corrected to levels observed in obese subjects after weight loss. The glucose transport parameter also improved after weight loss, reaching statistical significance in obesity but not in type 2 diabetes. Control coefficients were estimated as an index of the distribution of control between glucose transport and phosphorylation. In healthy lean subjects during similar insulin stimulation, we find transport to be the primary loci of control, reflecting a high efficiency of glucose phosphorylation (21). In contrast, in type 2 diabetes, control was approximately equally distributed between transport and phosphorylation before weight loss. After weight loss, the rate-limiting influence of glucose transport emerges more clearly, reflecting a sharp improvement in the efficiency of glucose phosphorylation. This finding suggests that improvement in the efficiency of glucose phosphorylation is an integral aspect of the improvement of insulin resistance after weight loss in type 2 diabetes, and it also suggests that further alleviation of insulin resistance in type 2 diabetes could require additional improvement in the capacity for glucose transport. In those with obesity alone, in which the pre-weight loss impairment in the efficiency of glucose phosphorylation was not as great, a chief effect of weight loss was upon glucose transport, although control analyses also indicated some improvements in phosphorylation.

A three-compartment model for [¹⁸F]FDG tissue kinetics, originally described for studies of deoxy-glucose metabolism by brain (31,32) and the first model used for quantification for [¹⁸F]FDG kinetics in skeletal muscle (17,19,33), was also used to analyze the data. Before weight loss in obese subjects with and without type 2 diabetes, this model indicated severe impairment in the k₃′ parameter, the rate constant that primarily reflects efficiency of glucose transport and phosphorylation (21). Like the k₁ values obtained using the 5K model, the k₁′ parameter defined by this model largely reflects the transfer of [¹⁸F]FDG from the plasma to a tissue compartment, and it was strongly correlated with blood flow measured independently with [¹⁵O]H₂O. This further clarifies that the k₁′ parameter of the 3K model, when applied to skeletal muscle, pertains primarily to substrate delivery, which is consistent with our prior observations (18). After weight loss, the k₁′ parameter did not change; in contrast, the k₃′ parameter improved threefold in obesity and five-fold in type 2 diabetes, reflecting a strong improvement in glucose transport and phosphorylation.

Several in vivo methods for the study of glucose transport and phosphorylation in skeletal muscle in humans have revealed a crucial role in the pathogenesis of insulin resistance in type 2 diabetes and obesity (13,15–17,19,20). What is unique about the current study is that we showed that inefficiencies in glucose transport and phosphorylation in type 2 diabetes and obesity can be substantially reversed with weight loss in the absence of changes in maximal aerobic capacity. In healthy subjects, induction of insulin resistance by experimental elevation of fatty acids has been shown to become manifest as impairments of glucose transport and/or phosphorylation (1,34). In a like manner, albeit in the opposite direction, the current study reveals that glucose transport and phosphorylation are central in mediating the reversal of insulin resistance. One interpretation is that impairments in glucose transport and phosphorylation may not be primary defects of insulin resistance but instead are induced impairments, or, at least, these defects have a substantial component of induced impairment.

Other studies also reveal malleability of glucose transport and phosphorylation in the reversal of insulin resistance. Recently, Greco et al. (11), found that weight loss after bariatric surgery restored GLUT4 expression in skeletal muscle and doubled insulin sensitivity in morbid obesity without type 2 diabetes. This is consistent with work from a decade ago by Friedman et al. (12), who found that after weight loss induced by gastric bypass surgery, in vitro insulin-stimulated uptake of 2-deoxyglucose in muscle fiber strips increased twofold. Henry et al. (10) also reported that weight loss doubled peripheral tissue glucose transport in type 2 diabetes by assessing ex vivo insulin-stimulated 3-O-methylglucose transport in adipocytes. Although it was not a weight loss study, Perseghin et al. (22) found, in an elegant study using in vivo NMR of muscle, that insulin-stimulated G6P accumulation within skeletal muscle increased twofold in the offspring of type 2 diabetic subjects in response to an exercise intervention, yet muscle glycogen synthesis remained impaired. In the current study, we cannot exclude the possibility that steps of glucose metabolism distal to blood flow, glucose transport, or its phosphorylation might contribute to some persistence of insulin resistance.

Thus, cross-sectional studies of type 2 diabetes and obesity reveal the importance of impaired glucose transport and phosphorylation in the pathophysiology of skeletal muscle insulin resistance, and they reveal that these defects can be induced quickly by elevated fatty acids. However, a more limited body of intervention-based investigations (22,23,35), of which ours is among the first to be performed with weight loss in vivo, suggest that these impairments may not be primary defects in the pathogenesis of insulin resistance. Consistent with the effects of changes in fatty acid levels to induce or alleviate insulin resistance, insulin-suppressed fatty acid levels declined in both groups in response to weight loss. The nature of the primary defect causing insulin resistance cannot be identified on the basis of the current study; however, it might be speculated based on the robust response to weight loss that was observed, and against the emerging profile of nutrient signaling that can trigger insulin resistance, that a substantial portion of the defects in glucose transport and phosphorylation arise as homeostatic adaptations to energy surplus, and they are corrected when the energy surplus is alleviated.

A limitation of the current study is the challenge of discerning roles of glucose transport and phosphorylation from single-tracer ([¹⁸F]FDG) tissue time-activity curves. To address this limitation, we used [¹⁵O]H₂O and were able to clarify that differences in tissue time-activity curves for [¹⁸F]FDG after weight loss were not caused by changes in patterns of blood flow and tissue perfusion. Furthermore, to address the limitations of the k₁′ and k₃′ parameters of the 3K model described above, we used a recently developed muscle-specific model to more clearly distinguish between transport and phosphorylation (21). One rationale for the 5K addition of a separate compartment for the kinetics of the exchange of [¹⁸F]FDG between the plasma compartment and the extracellular space of skeletal muscle is that skeletal muscle has a larger interstitial space than brain (36).

In summary, the current study shows that weight loss markedly improves the efficiency of skeletal muscle glucose transport and phosphorylation. The marked improvement in glucose phosphorylation in type 2 diabetes is an observation that will require further investigation to discern potential mechanisms. Weight loss had little effect on insulin-stimulated tissue perfusion or blood flow in skeletal muscle. These effects of weight loss underscore the importance of glucose transport and phosphorylation as a fulcrum of insulin resistance in skeletal muscle in type 2 diabetes and obesity, and they clearly reveal the plasticity of defects at these steps. We conclude that based on the response to weight loss, impairment in glucose transport and phosphorylation in obese subjects with and without type 2 diabetes has a substantial reactive component of impairment in the pathogenesis and treatment of insulin resistance in skeletal muscle.

These studies were supported by a Mentored Patient-Oriented Research Career Development Award from the National Institutes of Health (NIH) National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK; grant no. K23-DK02782), a Mid-Career Development Award for Patient Oriented Research (NIH-NIDDK Grant K24-DK02647), NIH Grant RO1-DK60555, the University of Pittsburgh General Clinical Research Center (#5MO1-RR00056), and the Obesity and Nutrition Research Center (NIH-NIDDK Grant P30-DK-46204-01). Optifast Formula was provided by Novartis Nutrition.

We gratefully acknowledge the efforts and cooperation of the research volunteers and the valuable help from the staffs of the University of Pittsburgh General Clinical Research Center and PET Center. In particular, we would like to express special appreciation to Rena R. Wing, PhD; Bret Goodpaster, PhD; David Townsend, PhD; Mary Lou Klem, PhD; Julie C. Price, PhD; Patricia H. Harper, MS, RD; Tracey Lawrence; Therese McKolanis, MPH; Christy Matan; and Jan Beattie, RN, BSN.