Interactions Among Glucose Delivery, Transport, and Phosphorylation That Underlie Skeletal Muscle Insulin Resistance in Obesity and Type 2 Diabetes: Studies With Dynamic PET Imaging

The rate of glucose uptake into skeletal muscle during insulin-stimulated conditions is a strong determinant of systemic insulin sensitivity (and insulin resistance [IR]). Much of the control of glucose uptake into skeletal muscle is located at the steps of glucose delivery, transport, and phosphorylation (1–4). Among these loci of metabolic control, stimulation of glucose transport mediated by translocation of GLUT4 in response to insulin is crucial (5–7). The small accumulation of glucose within insulin-stimulated muscle has provided supporting evidence that defects in glucose transport are primarily responsible for skeletal muscle IR (4,8–10). The small intracellular accumulation of glucose, however, could also be consistent with defects in glucose delivery (11). Studies using positron emission tomography (PET) have revealed that most of the insulin-stimulated control of glucose uptake in lean, nondiabetic human skeletal muscle is nearly equally distributed between glucose delivery and transport (12). The relative contribution of glucose delivery and transport within skeletal muscle in obesity and type 2 diabetes (T2D) remains uncertain.

Defects in glucose phosphorylation in response to insulin can also potentially contribute to IR. Animal models suggest that glucose phosphorylation capacity may limit glucose uptake when the capacity for glucose transport is high (13,14), although genetic manipulation of the glucose phosphorylation capacity by hexokinase overexpression appears to be less influential during insulin stimulation (1,14). In dynamic PET imaging studies quantifying rates of glucose delivery, transport, and phosphorylation using [¹⁵O]H₂O, [¹¹C]3-O-methylglucose (3-OMG), and [¹⁸F]fluorodeoxyglucose (FDG), respectively, Bertoldo et al. (12) reported that glucose phosphorylation represented an additional locus of control governing insulin-stimulated rates of glucose uptake in skeletal muscle of lean, healthy subjects. It is not clear, however, the degree to which skeletal muscle IR in obesity or T2D can be attributed to the control of glucose transport and phosphorylation.

The purpose of the current clinical investigation was to use dynamic in vivo PET imaging to dissect out potential loci of skeletal muscle IR in obesity and T2D. We used the triple-tracer method developed by Bertoldo et al. (12) to quantify rates of glucose delivery, transport, and phosphorylation during fasting (basal) and insulin-stimulated conditions. The present study provides novel in vivo quantitative information regarding the contribution and interaction of these proximal steps to skeletal muscle IR.

Informed, written consent was obtained from all participants, and the protocol was approved by the University of Pittsburgh Institutional Review Board. Three groups of research volunteers were recruited by advertisement: normal weight, nondiabetic (NW); overweight or obese nondiabetic (OB); and patients with T2D. NW is considered the control group in this study. The clinical characteristics are shown in Table 1; groups were well matched for age and sex distribution, and OB and T2D groups were closely matched for BMI. Blood lipoprotein levels were similar among groups, although plasma triglycerides tended to be higher in T2D. All participants had a medical examination to verify good health. Participants with T2D were treated with lifestyle measures alone or with metformin, which was withheld 3 days before the metabolic studies.

All participants underwent two sessions of triple-tracer PET imaging: one during fasting conditions (basal) and another, in random order, during steady-state insulin-stimulated conditions (insulin), as described below. On the evening before a study, research volunteers were admitted to the University of Pittsburgh Clinical Translational Research Center and fasted overnight. In the morning, catheters were placed in an antecubital vein for infusions and in a radial artery for blood sampling. During basal studies, volunteers did not receive insulin or dextrose infusions. During insulin studies, insulin was infused for 6 h at 40 mU ⋅ m⁻² ⋅ min⁻¹, arterial glucose was measured at 5-min intervals, and euglycemia was maintained by an adjustable infusion of 20% dextrose. Arterial glucose was measured with a YSI glucose analyzer (Yellow Springs, OH). Plasma insulin was measured by radioimmunoassay, and plasma fatty acids were measured by colorimetric enzymatic assay (Wako NEFA C test kit; Wako Chemicals, Richmond, VA).

PET and magnetic resonance imaging (MRI) was performed at the University of Pittsburgh PET Center and the Magnetic Resonance Research Center using procedures previously described (12). Briefly, a Siemens/CTI ECAT HR+ PET scanner was used in the three-dimensional imaging mode (63 parallel planes; axial field-of-view = 15.2 cm, slice width = 2.4 mm), and the final reconstructed PET image resolution was ∼6 mm. Participants were positioned in the PET scanner with the midcalf area in the center of the field.

[¹⁵O]H₂O was given as an ∼30-mCi bolus, and arterial blood sampling was obtained with a Siemens liquid activity monitor. PET scanning began with the injection and lasted 3 min (18 frames of 10 s duration each). PET data were corrected for radioactive decay and scatter (15). During insulin studies, the insulin infusion was given for at least 2 h before PET imaging.

[¹¹C]3-OMG was administered as a slow bolus of ∼5 mCi ∼30 min after completing PET imaging of [¹⁵O]H₂O, and dynamic PET scan was started (90 min; 36 frames; 8 of 15 s duration, 8 of 30 s duration, 4 of 1 min duration, and 16 of 5 min duration). Arterial [¹¹C]3-OMG was measured in 0.5-mL samples obtained manually (10 samples every 6 s, 8 samples every 15 s, 7 samples every 1 min, 10 samples every 5 min, and 3 samples every 10 min). Blood samples were immediately centrifuged, and 200 μL of plasma was removed for immediate [¹¹C] counting with a COBRA Auto-Gamma model 5003 gamma counter (Packard Instruments, Meriden, CT).

Two hours after injection of [¹¹C]3-OMG, ∼6 mCi [¹⁸F]FDG was given as a slow intravenous bolus, and dynamic PET scanning was started (90 min; 28 frames; 8 of 30 s duration, 8 of 60 s duration, 8 of 4 min duration, and 4 of 8 min duration). Arterial blood draws were made for determining [¹⁸F]FDG activity (10 × 6 s, 8 × 15 s, 7 × 1 min, 10 × 5 min, and 3 × 10 min). Blood was immediately centrifuged, and 100 μL of plasma was used for [¹⁸F] counting (>350 KeV).

A T1-weighted MRI of the midcalf was obtained on the same day as the PET imaging session and consisted of a series of T1-weighted images (acquisition parameters: axial plane, repetition time = 31 ms, echo time = 15 ms, flip angle = 10°, field of view = 40 × 20 cm, matrix size = 256 × 192 × 120, partition thickness = 1.5 mm, contiguous partitions, 1 average, total acquisition time = 12.75 min) acquired for high-resolution anatomy. The MRIs were aligned to PET images by a previously described method (16,17). Briefly, summation PET images were created from the frames of the initial 15 min of the scanning period and coregistered with MRI so that precise anatomical alignment was achieved.

A region of interest (ROI) was placed over soleus muscles on the summation PET images and MRIs, and these ROIs were applied to each frame (18 for [¹⁵O]H₂O, 36 for [¹¹C]3-OMG, and 28 for [¹⁸F]FDG) across 58 of the 68 planes of each leg, omitting the proximal and distal 5 planes to reduce the influence of scatter. Tracer activity within an ROI was converted to radioactivity concentration (µCi/mL) using an empiric phantom-based calibration factor (µCi/mL/PET counts per pixel).

The model of Kety (18) was used to estimate tissue perfusion of skeletal muscle.

Compartmental modeling of [¹¹C]3-OMG kinetics was performed, as previously described in detail (19,20). Briefly, this model has three compartments (plasma and two tissue pools) with four rate constants as shown in Fig. 1A. The rate constants (mL/mL/min) and (min⁻¹) describe a reversible exchange between plasma and the first tissue compartment and are ascribed physiologically to an exchange of [¹¹C]3-OMG between the capillary and the muscle interstitial space. The rate constants (min⁻¹) and (min⁻¹) describe a reversible exchange between the first and second tissue compartments and are ascribed physiologically to bidirectional transmembrane glucose transport. From these rate constants, an overall partition coefficient V_d for uptake of [¹¹C]3-OMG from plasma to tissue can be calculated and is divisible into a partition coefficient for the first compartment, muscle interstitial space V_ic, and a partition coefficient for the tissue compartment V_ec, as shown in Eq. 1:

Compartmental modeling of [¹⁸F]FDG kinetics was performed as previously described (21–23). Briefly, a model of three compartments with five rate constants was used that is similar to that described for [¹¹C]3-OMG but with an additional tissue compartment and rate constant to represent phosphorylation of [¹⁸F]FDG. In this article, to avoid confusion with the rate constants used for [¹¹C]3-OMG, those pertaining to [¹⁸F]FDG are designated with an asterisk (k_n*). Thus, and describe the delivery of [¹⁸F]-FDG from the capillary to the muscle interstitial space and reserve process, respectively; and describe inward and outward transport of [¹⁸F]FDG; and describes the phosphorylation of [¹⁸F]FDG. From the five rate constants, the fractional uptake of [¹⁸F]FDG can be determined, as represented by the macroparameter K (mL/mL/min) shown in Eq. 2:

A specific glucose tracer was not used in this study; instead, a glucose model was developed from the kinetics determined using [¹¹C]3-OMG and [¹⁸F]FDG, as previously described (12,21–25). Rate constants from compartmental modeling of [¹¹C]3-OMG were used for that of glucose, and a value for the rate constant of glucose phosphorylation was additionally derived from the lumped constant (LC), as shown in Eq. 3:

From glucose k₁–k₅ parameters, intracellular glucose concentration (Ce) was estimated, using Eq. 4 (16):

Following the theory of metabolic control analysis, the control coefficients (CC) of delivery, transmembrane transport, and phosphorylation were calculated as shown in Eq. 5 (17,26,27):

The higher the value of a control coefficient, the more that this process is rate limiting and according a locus of control.

Data are presented as mean ± SEM. One-way ANOVA was used to examine for significance of differences among subject groups. One-way repeated-measures ANOVA was used to compare basal and insulin group parameters. A P < 0.05 was considered significant.

The clinical characteristics of the research subjects are presented in Table 1, and the rates of exogenous glucose infusion during insulin infusion clamp studies are shown in Fig. 1A. Compared with NW, both OB and T2D manifested systemic IR, which was most severe in T2D (P < 0.001). Fasting glucose values were significantly different among T2D (144.0 ± 13.0 mg/dL), NW (96.5 ± 1.8 mg/dL), and OB (94.4 ± 1.5 mg/dL) (P < 0.05). Insulin levels were not significantly different among the three groups at fasting (NW 9.81 ± 0.94, OB 18.01 ± 3.54, T2D 18.39 ± 1.80 µU/mL) or insulin stimulation (NW 87.95 ± 2.18, OB 102.30 ± 8.89, T2D 102.52 ± 6.02 µU/mL). Free fatty acids were also not significantly different between the groups at fasting (NW 592 ± 65, OB 577 ± 30, T2D 532 ± 63 µmol/L) or insulin stimulation (NW 106 ± 33, OB 127 ± 31, T2D 167 ± 35 µU/mL). Skeletal muscle (soleus) insulin sensitivity, as measured by K, the macroscopic PET imaging parameter for [¹⁸F]FDG uptake, was strongly correlated with respective glucose infusion rates as shown in Fig. 1B.

Tissue perfusion in soleus muscle was lower (P < 0.01) in OB (0.0194 ± 0.0055 mL/min-mL) and T2D (0.0125 ± 0.0022 mL/min-mL) than in NW (0.0316 ± 0.0139 mL/min-mL) during basal conditions. During insulin infusion, muscle tissue perfusion did not change significantly from basal conditions in any group. The k₁ parameter obtained from [¹¹C]3-OMG modeling, which describes tracer delivery from plasma to tissue, has values that are concordant with those obtained using [¹⁵O]H₂O, a tracer that is used to measure solely tissue perfusion kinetics. This concordance of k₁ values for these two tracers is consistent with earlier findings (12,20,28).

The mean group soleus muscle tissue time–activity curves for [¹¹C]3-OMG are shown in Fig. 2A and B for basal and insulin-stimulated conditions, respectively. Using these data in conjunction with arterial plasma time–activity data, tissue data were modeled to estimate parameters describing both delivery (plasma to extracellular space) and bidirectional glucose transport for [¹¹C]3-OMG, and these parameters of the tracer kinetics that underlie the shapes of the tissue–activity curves are shown in Table 2. The kinetic parameter of great interest for this investigation of muscle IR is k₃, which described the inward transport for [¹¹C]3-OMG, a nonmetabolized glucose analog that has nearly identical affinity as glucose for glucose transporters (25). Across groups, the basal values for k₃ were similar (Fig. 3). During insulin infusion, k₃ increased sevenfold in NW participants, which is similar to earlier findings in lean nondiabetic subjects (12,19). In the OB subjects, the group mean value for k₃ increased sixfold over the basal condition, which is a similar response to that found in NW. However, in T2D, the group mean value for k₃ during insulin-stimulated conditions was significantly lower than in NW or OB (P < 0.01), and although the value increased during the insulin condition compared with the basal condition, the change was a twofold increase, which was a blunted response to insulin compared with the changes observed in the NW and OB nondiabetic groups. Because the transport of [¹¹C]3-OMG is bidirectional, the higher tissue activities and k₃ values in the NW and OB groups were associated with higher k₄ values (and greater changes in k₄ during insulin compared with basal conditions). As well, the higher efficiency of inward glucose transport during insulin-stimulated conditions was associated with reduced values for k₂, the parameter describing efflux from tissue (extracellular space) to plasma. In T2D, the insulin-stimulated versus basal changes in k₂ and k₄ were blunted compared with NW and OB. The insulin-resistant response of glucose transport in T2D is also evident within the volume of distribution parameters for [¹¹C]3-OMG (Table 2), which reveals that although there was evidence of soleus response to insulin in T2D compared with basal conditions, the amplitude of this response was attenuated in T2D, indicating that impaired insulin stimulation of glucose transport contributes strongly to muscle IR.

The key difference between [¹¹C]3-OMG and [¹⁸F]FDG is that the latter glucose analog can be a substrate for hexokinase and undergo phosphorylation, which over the short time course of these PET imaging studies causes [¹⁸F]FDG-6-P to be formed and irreversibly trapped within soleus muscle. The soleus muscle time–tissue activity curves for [¹⁸F]FDG during basal and insulin-stimulated conditions are shown in Fig. 4A and B. During basal conditions, there was close similarities across groups, but during insulin-stimulated conditions, marked group differences became evident with sharply diminished tissue activity in T2D and a discernibly altered shape. For the goals of this investigation, to study the potential contribution of glucose phosphorylation to IR, the arterial and tissue activity for [¹⁸F]FDG were used to determine K, the macroscopic parameter of irreversible tracer uptake. As described in the research design and methods, K was used in conjunction with kinetic parameters for [¹¹C]3-OMG and with the lumped constant to estimate the kinetics of actual glucose delivery, bidirectional transport, and phosphorylation. The kinetic parameters for [¹⁸F]FDG are presented in Supplementary Table 1. These data will not be further discussed here other than to note that the findings across groups and between basal and insulin studies are similar to those for [¹¹C]3-OMG, with differences reflecting known analog effects for [¹⁸F]FDG (as reflected in the lumped constant) and with the addition of k₅, the kinetic parameter ascribed to the formation of [¹⁸F]FDG-6-P (Fig. 5).

The partition coefficient V_ec for glucose in the tissue extracellular compartment relative to plasma did not change from the basal to the insulin conditions. The partition coefficient for glucose in the tissue intracellular compartment relative to plasma increased during the insulin condition (P < 0.05) in NW and OB but not in T2D. was then used to calculate the intracellular glucose values when multiplied by the plasma glucose levels. Intracellular glucose levels were not significantly different at basal conditions, but under insulin stimulation, NW had significantly higher intracellular glucose levels than T2D (NW 33.7 ± 4.8, OB 31.3 ± 12.5, T2D 10.1 ± 2.0 mg/dL; P < 0.05).

The rate constants describing glucose kinetics were used to calculate the control coefficient for delivery (CC_delivery), transport (CC_transport), and phosphorylation (CC_phosphor) during the basal and insulin conditions are presented in Fig. 6. Glucose transport and phosphorylation (hexokinase) largely account for control of glucose kinetics in skeletal muscle during basal conditions. Of these, transport exerted stronger control than phosphorylation. Thus, glucose delivery was not found to substantially control glucose uptake into skeletal muscle during basal conditions. This pattern was consistent across all groups. For NW subjects, this pattern of distribution of control shifted during insulin conditions such that glucose delivery and transport contributed nearly equally to the control of glucose uptake and together accounted for 90% of control, whereas phosphorylation was calculated to manifest only the remaining 10% of rate control for the kinetics of glucose uptake. This was in contrast to that observed for T2D in which insulin-stimulated glucose uptake was predominantly, but not exclusively controlled, that is, limited by glucose transport. Thus, the distribution of insulin-mediated glucose uptake was vastly different in T2D muscle.

To our knowledge, this study is the first to use three tracers ([¹⁵O]H₂O, [¹¹C]3-OMG, and [¹⁸F]FDG) during dynamic PET imaging to ascertain separate and specific defects in glucose metabolism in obesity and T2D. The primary findings were that lower insulin-stimulated glucose transport accounts for a quantifiable and substantial degree of IR in skeletal muscle of T2D and that additional defects in glucose phosphorylation characterize the IR of T2D.

Following an overnight fast, neither glucose transport nor phosphorylation was different in the skeletal muscle of OB or T2D compared with lean, healthy muscle. The predominant control of glucose uptake in the fasting state was at the level of glucose transport across all study groups, with some additional control posited by glucose phosphorylation. Glucose delivery exerted only negligible control, although this was quantitatively lower in muscle in T2D. The lower basal muscle blood flow in diabetes could perhaps reflect a lower capillary density or perfusion (29). These observations are consistent with other studies showing that basal glucose uptake is not impaired in IR and diabetes (30) and that glucose transport is the rate-limiting step in basal glucose uptake (30).

The glucose infusion rates in response to physiological levels of insulin were lower for OB subjects, and this systemic IR was more severe in T2D, which was strongly associated with the macroscopic index of muscle glucose flux assessed by PET. The lower rates of glucose transport and fractional phosphorylation in OB than in NW, although not as markedly low as in T2D, clearly indicate that skeletal muscle IR in OB, along with normal fasting glycemia, helps to define the prediabetic state. This is consistent with values reported in earlier studies in normal weight subjects using similar protocols (12,19). However, despite the nearly 50% lower systemic glucose infusion rates in OB subjects, their lower insulin-stimulated increase in glucose transport was not statistically different, suggesting that obesity alone has less dramatic effects of inward glucose transport despite fairly marked systemic IR. The small number of subjects could have limited our ability to quantify the obesity-related defects in glucose transport, particularly given the heterogeneity of insulin sensitivity in obesity. Moreover, the systemic insulin sensitivity could have been confounded by hepatic IR, which again, highlights the utility of PET to quantify skeletal muscle glucose metabolism. We also note that a lean, T2D group was not included, which could have further quantified the effects of both obesity and T2D. Additionally, we recognize that even though our sex ratios were similar, there could be sex-specific differences in metabolic disease that may have influenced the findings. In stark contrast, skeletal muscle in T2D had much lower fractional glucose transport rates and intracellular glucose levels. These results are consistent with lower muscle GLUT4 content and translocation in T2D (30–33) and with the slight accumulation of free glucose in response to insulin (4,8,10), all of which have been interpreted as defects in glucose transport and comprising a major limitation to glucose uptake in response to insulin. Although bidirectional glucose transport within muscle can be assessed directly in vitro, the in vivo glucose transport also depends on delivery to the tissue. The time–tissue activity curves for 3-OMG reflect an integration of these two reversible processes (12,20,28), which the current study distinguished using sequential imaging with [¹⁵O]H₂O and [¹¹C]3-OMG. Insulin-stimulated rates of muscle tissue perfusion were similar across subject groups and are similar to those reported previously for normal weight subjects (34). These rates corresponded very well with kinetic values for tissue perfusion attributed to k₁ ascertained for [¹¹C]3-OMG and [¹⁸F]FDG. Rates of tissue perfusion did not change significantly in response to insulin compared with fasting. This does not discount the hemodynamic effects of insulin to stimulate capillary recruitment, as has been demonstrated (35,36), which is beyond the spatial resolution of PET. Improving spatial resolution with PET may provide additional insight into the dynamic interaction between glucose delivery and transport in IR. NW subjects in the current study had a redistribution of control of glucose uptake from predominantly glucose transport in the fasting state to greater control by delivery in response to insulin, which was not observed for muscle in OB or T2D subjects. Thus, the greater control of glucose uptake by insulin in muscle of NW subjects is consistent with in vivo studies in rodents, indicating that insulin-stimulated glucose uptake is constrained by delivery (11,14,37). This is also in accord with a major defect in insulin-stimulated glucose transport in diabetic muscle in the current study.

The insulin-stimulated fractional phosphorylation of glucose was significantly lower in muscle in diabetes. Thus, after accounting for glucose delivery and lower glucose transport in diabetes ascertained specifically with [¹¹C]3-OMG, an additional defect in glucose phosphorylation exists in insulin-resistant diabetic muscle. This defect was quite pronounced in diabetic muscle and was not as obvious in muscle of OB subjects. This finding could reflect the heterogeneity or wider range in the severity of IR of obesity and suggests that obesity per se does not unequivocally cause defects in skeletal muscle glucose metabolism but, nevertheless, contributes to the underlying progression of IR eventually causing T2D. Dynamic PET imaging performed using [¹⁸F]FDG indicated that glucose phosphorylation does not exert strong rate control on glucose uptake during insulin-stimulated conditions in muscle of NW subjects but still distinguishes skeletal muscle IR in T2D. This is similar to a previous article that used dynamic PET studies modeling [¹⁸F]FDG uptake kinetics to observe a major defect in glucose transmembrane transport with an additional, albeit smaller, defect in glucose phosphorylation (23). The current study extends these findings in two very important ways. First, the use of multiple PET tracers in the same subject has allowed us to separate and quantify the proximal steps of glucose transport and phosphorylation with considerably less dependence on mathematical modeling of kinetic data. Second, we have now more clearly shown for the first time using these multiple PET tracers that glucose phosphorylation is impaired in skeletal muscle in T2D after accounting for defects in glucose transport. This is consistent with reports in animal models demonstrating additional defects in insulin-stimulated glucose phosphorylation (1,11,13,14,38–40). Both GLUT4 and hexokinase are coordinately regulated (41), and both track strongly with oxidative capacity and capillary density in muscle, all of which have been consistently demonstrated to be affected in T2D (30,31,33,42).

In summary, dynamic PET imaging of skeletal muscle glucose metabolism reveals a major defect in glucose transport, which defines much of the severity of IR of T2D, and that an additional impairment within the efficiency of glucose phosphorylation is present. In contrast to lean, healthy skeletal muscle in which the control of insulin-stimulated glucose uptake is more broadly distributed across each of the proximal steps of glucose delivery, transport, and phosphorylation, the rate-limiting control of glucose metabolism in muscle in T2D is posited mostly at glucose transport, with smaller control being manifest at glucose delivery and phosphorylation. These triple-tracer PET imaging data have important implications for clinical treatment of T2D, and these methodologies could be valuable in determining whether these defects persist at higher doses of insulin and which specific control points may be rectified with insulin-sensitizing interventions.

Acknowledgments. The authors acknowledge the efforts and cooperation of the research volunteers and the support from the staffs of the University of Pittsburgh General Clinical Research Center and the PET Center.

Funding. These studies were supported by the National Institutes of Health, National Institute of Diabetes and Digestive and Kidney Diseases (NIH-NIDDK) (DK-60555-02), by the University of Pittsburgh General Clinical Research Center (5MO1-RR-00056), and by the Obesity and Nutrition Research Center (NIH-NIDDK, P30-DK-46204). This work was also supported by NIH grant EB-01975.

Duality of Interest. R.R.P. is currently employed by Amylin Pharmaceuticals. D.E.K. is currently employed by Merck Sharp & Dohme Corp. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. B.H.G. supervised the project and wrote the manuscript. A.B. performed the modeling analysis. J.M.N. researched the data, contributed to the discussion, and reviewed and edited the manuscript. K.A., R.R.P., and C.K. researched the data. J.C.P. and C.C. contributed to the modeling analysis. D.E.K. contributed to the discussion and reviewed and edited the manuscript. B.H.G. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.