Insulin resistance (IR) is a common disorder frequently observed in different metabolic diseases, such as obesity, dyslipidemia, arterial hypertension, and type 2 diabetes. The development of the euglycemic-hyperinsulinemic clamp technique (1) allowed determination of peripheral insulin-stimulated glucose uptake and the degree of IR present primarily in skeletal muscle. This technique has contributed further evidence that peripheral IR is a primary factor in the development of type 2 diabetes (2–4) and is considered the “gold standard” for the assessment of insulin sensitivity (IS) (5, 6).
The molecular mechanisms causing IR are still not unequivocally elucidated, but there is increasing evidence that glucose and fatty acid metabolism are closely linked and that alterations of this connection are related to IR (7). One feature of this link is the accumulation of lipids in muscle. In vivo proton magnetic resonance spectroscopy (MRS) has been successfully applied to study the intramyocellular lipid storage in intact human muscle noninvasively. The lipid methylene and methyl proton resonances (between 1.0 and 1.6 ppm) have been assigned to two lipid compartments within muscle tissue (intramyocellular and extramyocellular lipids, IMCL and EMCL, respectively) (8, 9) and can be observed separately (9).
IMCL content represents a good indicator of whole-body IS in nondiabetic, nonobese humans (10, 11), and elevated IMCL levels are related to impaired insulin-stimulated glucose uptake in obese humans and in type 2 diabetic patients (12–14). In animals, however, only a few studies using IMCL as a biomarker for IR have been performed (15–17). Additionally, it was recently shown that IMCL content varies significantly between different muscle types (18, 19) and that it also depends on other factors, like age and gender (19).
The aim of our study was to investigate the interrelation between IS in terms of glucose disposal (GD) as well as exogenous glucose infusion rates (GIR) and IMCL content both in a predominantly glycolytic (M. tibialis anterior, TIB) and a more oxidative muscle type (M. soleus, SOL) in different animal models of IR and to identify the respective parameters that correlate best.
Well-established animal models for IR were chosen: (a) diet-induced ones (fructose- and “cafeteria”-fed Wistar rats); and (b) as such with a genetic background: female obese Zucker Diabetic Fatty (ZDF) rats, which are insulin resistant while maintaining normoglycemia, and male obese ZDF rats, which become spontaneously diabetic at an age of about 10 weeks. Furthermore, we examined the effect of long-term treatment with the PPARγ-agonist rosiglitazone, which is known to increase IS in IR individuals, on the correlation coefficients.
MATERIALS AND METHODS
Animal Models and Study Design
All animal handling and surgery was performed in accordance with German Animal Protection Law.
Experiments were performed in male Wistar rats (HsdCpb:WU; Harlan Winkelmann, Borchen, Germany, n = 35), male and female obese ZDF rats (ZDF/Gmi-fa/fa; male: n = 12, female: n = 12) and their lean littermates (ZDF/Gmi-+/?; male: n = 7, female: n = 16; Genetic Models Inc., Indianapolis, USA). All rats were fed a standard rat chow, if not indicated otherwise. Wistar rats were divided into three groups: a normal fed group (n = 12), a second group fed a high-fructose diet (60% fructose, obtained from Ssniff, Soest, Germany; n = 12), and a third group fed a “cafeteria” diet (standard rat chow plus noodles, biscuits, cheese, and chocolate ad libitum; n = 11). One group of female lean ZDF rats was fed a high-fat diet (26% fat, Ssniff; n = 8).
Additionally, subgroups of obese male (n = 5) and female (n = 6) ZDF rats were treated with rosiglitazone (“Avandia”, Smith Kline Beecham, Munich, Germany) at a dose of 3 mg/kg/day by food admixture. Treatment was started at the age of 8 weeks and continued for the whole experimental period. Untreated animals served as control groups. The detailed study design is given in Fig. 1.
Animals were housed two per cage in a room maintained at 22°C and at a 12-h light–dark cycle with ad libitum access to water and food.
One to 3 days prior to the terminal euglycemic-hyperinsulinemic glucose clamp study (1), IMCL content was monitored by MRS, body weight of all animals was recorded, and blood samples were obtained retroorbitally during isoflurane anesthesia for the determination of nonesterified fatty acids (NEFA), triglycerides, and plasma insulin as described in Ref. (16). Blood samples (10 μL) for the determination of blood glucose were taken from the tip of the tail.
MR measurements were performed in fed animals under isoflurane anesthesia on a 7-T Biospec scanner (Bruker BioSpin GmbH, Ettlingen, Germany) as described previously (16, 19). Briefly, localized spectroscopy was achieved using a PRESS sequence (voxel size: 1.6 × 3.2 mm3, TE = 17 ms, TR = 1000 ms, 1024 acquisitions, CHESS water suppression). All spectra were acquired at an angle of 0° between muscle fiber and magnetic field to maximize the chemical shift of EMCL (9) and to optimize the separation of IMCL and EMCL. The doublet structure of the creatine methylene signal (3.9 ppm) characteristic for parallel alignment (20) was used to verify proper positioning.
Spectra analysis was performed blinded using the MRUI software package (the MRUI Home Page, see: http://www.mrui.uab.es/mrui/mruiHomePage.html). All data were subjected to the HLSVD method in order to remove the residual water signal. Further spectra analysis implied the time-domain fitting algorithm AMARES (21). The resonances of total creatine (tCr; 3.05 ppm) and IMCL (1.3 ppm) were fitted as singlets (the triplet structure of tCr at 3.05 ppm was never detectable with our experimental design) using prior knowledge of frequencies and equivalent linewidths. IMCL levels were expressed relative to the tCr signal intensity.
Euglycemic-Hyperinsulinemic Glucose Clamp Study
The study design consisted of a 120-min basal period followed by a 2-hr euglycemic-hyperinsulinemic clamp with a constant insulin infusion (4.8 mU/kg/min). The study was performed in overnight fasted animals under pentobarbital anesthesia as described previously (16). Additionally, the rats were given a constant infusion of [U-13C]glucose (1 mg/kg/min) to estimate rates of glucose production and utilization (22). Every 15 min, blood samples were obtained via tail-tip bleeds for determination of glucose enrichment. Enrichments were calculated from the ratio of [U-13C]glucose/12C-glucose during the last 30 min of the basal period and during the last 60 min of the clamp (i.e., under steady-state conditions). This ratio was determined by gas chromatography-mass spectrometry analyses of derivatized glucose from blood samples following literature protocols (22, 23).
Blood samples were drawn at 30 min for determination of baseline insulin levels and at 180, 210, and 240 min for determination of insulin levels during hyperinsulinemia. At 120 and at 240 min, additional blood samples for determination of NEFA were taken. Animals were euthanized by pentobarbital overdose.
All calculations were carried out during steady-state conditions, when the rate of appearance of glucose (Ra) is equal to the rate of glucose disappearance (Rd). Whole body glucose uptake (GD = Ra = Rd, in mg/kg/min) was calculated as the ratio of the [U-13C]glucose infusion rate to the enrichment of plasma glucose. Under these conditions, GD is equal to the sum of the rates of exogenous glucose infusion (GIR) and of endogenous glucose production (EGP), i.e., Ra = Rd = GD = GIR + EGP. From this equation, the rate of EGP can be calculated. In the basal state, when no exogenous glucose is infused, EGP is assumed to be equal to GD, i.e., Ra = Rd = GD = EGP (24, 25).
Overt diabetes is predominantly characterized by hyperglycemia leading to the excretion of significant amounts of glucose via the kidneys, since the renal threshold is exceeded. Male obese ZDF rats older than 14 weeks display a manifest diabetic condition with marked hyperglycemia and glucosuria. Therefore, basal GD in this animal model musto be corrected for urinary glucose loss (UGL) by subtracting the amount of glucose excreted via urine from the rate of EGP. In order to determine UGL, animals were placed in metabolic cages with free access to water and food, and their urine was collected over 24 hr. UGL was calculated in milligrams per kilogram per minute.
Since the glucose clamp is a highly sophisticated labor-intensive method for the determination of in vivo insulin sensitivity, we performed glucose clamp studies in subgroups of rats. We considered group numbers of 4–7 animals sufficient for a proper statistical evaluation. The selection was effected randomly.
Results are given as means ± SD. Depending on data normality, statistical differences were determined using an unpaired bilateral t test or a Mann–Whitney rank sum test (SigmaStat, Jandel GmbH, Germany). When testing for differences among three or more experimental groups, a one-way ANOVA followed by a post hoc analysis with Bonferroni's correction or a Kruskal–Wallis one-way ANOVA on ranks followed by a post hoc analysis with Dunn's correction was used. Parameters with values P < 0.05 were considered to differ significantly.
Simple correlation analysis using all individual pair values was applied to examine the relationships between IMCL and IS, and Spearman's regression coefficient was calculated.
Plasma parameters and body weight of all groups are collected in Table 1. Metabolic plasma parameters of high-fructose-fed Wistar rats were comparable to those of the control group, except for slightly elevated plasma triglycerides. Cafeteria-diet-fed Wistar rats displayed a slight hyperinsulinemia, significantly increased body weight, and triglyceride levels but normal blood glucose values compared to controls, reflecting an insulin resistant condition without manifestation of diabetes. On high-fat feeding, female lean ZDF rats showed unchanged blood glucose levels, but significantly higher body weight, NEFA, and triglyceride values, whereas in female obese rats a stronger rise in body weight, insulin, and triglycerides was observed. Blood glucose remained normal, demonstrating the insulin resistant but nondiabetic state of obese female ZDF rats. Old male obese ZDF rats displayed extremely elevated blood glucose values indicative of their overt diabetic state. Body weight, insulin, NEFA, and triglyceride levels were significantly higher than in their lean littermates. In male and female obese ZDF rats, rosiglitazone treatment caused a clear amelioration of most plasma parameters, thus improving the insulin-resistant condition, and additionally prevented the progression to overt diabetes in males. However, a pronounced body weight gain was observed in rosiglitazone-treated animals.
Table 1. Body Weight and Metabolic Plasma Parameters of Investigated Rats
Body weight (g)
Blood glucose (mmol/l)
Note. Values are given as mean ± SD. n = 6–12;
P < 0.05,
P < 0.001 versus respective insulin-sensitive control group;
P < 0.05,
P < 0.001 versus untreated group. RGZ, rosiglitazone.
IMCL levels and glucose clamp results are shown in Table 2. IMCL levels of both investigated muscles were significantly increased in both diet-induced models compared to the Wistar control group, as well as in male and female obese ZDF rats compared to their lean littermates (Table 2; Fig. 2). In female lean ZDF rats fed a high-fat diet IMCL values increased compared to lean control rats, but no significant level was reached. IMCL content in the TIB was slightly reduced in rosiglitazone-treated female, but not in male ZDF rats, compared to controls. IMCL content in SOL was higher in both male and female obese ZDF rats compared to untreated control rats.
Table 2. IMCL Levels in TIB and SOL and Data from Euglycemic-Hyperinsulinemic Clamp at 4.8 mU/kg/min Insulin (GIR, GD, and EGP; in mg/kg/min) of Investigated Rats
(EGP basal) GD basal
Note. Values are given as means ± SD.
P < 0.05,
P < 0.001 versus respective insulin-sensitive control group;
P < 0.05,
P < 0.001 versus untreated group.
High SD in this group resulted from calculated negative values (5).
In this group, values for basal GD were corrected for urinary glucose loss (mean basal GD = mean calculated EGP − mean UGL = 21.9–15.1; in mg/kg/min). RGZ, rosiglitazone.
Basal rates of GD, which correspond to basal EGP, were unchanged in fructose- and cafeteria-fed Wistar rats compared to controls (Table 2). In male obese ZDF rats, however, EGP was three times higher than in lean littermates, reflecting their diabetic condition. In these overtly diabetic hyperglycemic rats, basal GD was corrected for UGL. Their basal GD rates were then comparable to lean controls, although exhibiting fivefold higher blood glucose levels. Female obese ZDF rats demonstrated a two times elevated GD in the basal state, whereas high-fat feeding did not affect basal GD rates in female lean rats. Rosiglitazone treatment had no effect on basal rates of glucose turnover (after correction for UGL in male diabetic controls), either in male or in female ZDF rats.
The amount of exogenous glucose required to maintain normoglycemia (GIR) was lower in both diet-induced models than in Wistar controls (Table 2). GD rates were also reduced in diet-fed rats compared to controls, and rates of glucose production were elevated. Obese male and female ZDF rats also displayed significantly reduced GIR and GD, accompanied by two times higher rates of EGP in male animals. High-fat feeding of lean female ZDF rats resulted in significantly lower GIR and GD compared to respective control animals, which were similar to those of diet-fed Wistar rats. Rosiglitazone-treated male and female obese ZDF rats displayed higher values for GIR, indicating an improvement of IS. There were tendencies toward increased GD, predominantly in females, and toward decreased EGP, which was significant in male animals.
In order to investigate the correlation between IMCL content and IS, IMCL levels measured in both muscle types and GIR as well as GD as indicators for peripheral IS were used. IMCL detected in the TIB was found to correlate better with both parameters of IS, GIR, and GD (Table 3; Fig. 3 a and 4 a), than IMCL content in the SOL (Table 3, Fig. 3b and 4b). Surprisingly, TIB-GIR correlated best (r = −0.69, P < 0.001; slightly better than TIB-GD: r = −0.64, P < 0.001), although per definitionem we would have expected GD to be the better marker for IS, as proven in the SOL muscle (SOL-GD: r = −0.66, P < 0.001; SOL-GIR: r = −0.56, P < 0.001).
Table 3. Spearman's Correlation Coefficients Including Rosiglitazone-Treated Rats (A) and Excluding Treatment Groups (B)
A. Including treated animals
Note. P < 0.001 for all comparisons.
B. Excluding treatment groups
It is evident from Fig. 3 (GIR and IMCL) that the values obtained from diabetic rats (male obese ZDF) did not fit to the data, independent of the investigated muscle. Excluding these data sets from the correlation analysis, the fit improved to significant levels (TIB-GIR: r = −0.72, P < 0.001; SOL-GIR: r = −0.72, P < 0.001). When correlating IMCL and GD, exclusion of male diabetic ZDF rats resulted in an improvement of the correlation coefficient in SOL (TIB-GD: r = −0.59, P < 0.001; SOL-GD: r = −0.73, P < 0.001).
The effect of treatment with rosiglitazone was reviewed separately. On rosiglitazone treatment, male obese ZDF rats exhibited values for IMCL and for GD similar to those of their untreated diabetic littermates, but required significantly higher GIR and showed an improved insulin-mediated suppression of EGP. Female obese ZDF rats treated with rosiglitazone displayed slightly lower IMCL levels than the untreated group in the glycolytic TIB, while IMCL in the oxidative SOL tended to be elevated. However, GIR and GD were improved in treated animals. In this group, exclusively IMCL values measured in the TIB correlated inversely with IS and thus matched with the postulated interrelation between IMCL and IS. However, when excluding rosiglitazone-treated animals, all correlation coefficients deteriorated. The effect of rosiglitazone treatment on the correlation coefficients is shown in Table 3B.
In the present study, IMCL content in two different skeletal muscle types, the predominantly nonoxidative TIB and the predominantly oxidative SOL, was correlated with IS in terms of peripheral insulin-stimulated GD as well as of exogenous GIR in various animal models for IR. Our purpose was (i) to identify the parameters (muscle type as well as IS index) that correlate best and (ii) to validate this correlation in models of diet-induced IR as well as in models with a genetic background, with and without treatment with an insulin-sensitizing agent. Our study demonstrated that the inverse correlation between IMCL content and IS found in humans (10, 11, 14, 26, 27) is also present in these animal models of IR. Additionally, we were able to show that: (i) IMCL and GIR values of overtly diabetic obese ZDF rats exhibit the largest deviation from all other data from insulin-sensitive or insulin-resistant animals, and (ii) the relationship between IMCL and peripheral IS was found to exist also in animals treated with rosiglitazone.
IMCL deposits are found in predominantly glycolytic skeletal muscles as well as in more oxidative types. These lipid stores turned out to be a reliable marker of IR in obese (28, 29) and nonobese adults (10) as well as in nondiabetic offspring of type 2 diabetic subjects (11, 26). Hence, the deposition of lipid in nonadipose tissue has been considered an important factor in the development of IR by several groups (7, 10). Although the precise mechanisms by which intramyocellular fatty acid species (possibly in form of long-chain acyl coenzyme A, LCACoA) might cause IR are still unclear, LCACoA have been reported to inhibit the activity of hexokinase, pyruvate dehydrogenase, and glycogen synthesis in the liver (30), and it is possible that similar inhibitory effects of these activated fatty acids may also occur in skeletal muscle. Furthermore, it has been shown that abnormalities in insulin signaling may arise from overaccumulation of lipids in skeletal muscle cells (31, 32). But until now it has not been indisputably proven that high IMCL levels are the cause and not the consequence of IR.
The equivalent correlation of IMCL levels with GD compared with GIR, predominantly observed in TIB, was unexpected, as GD is, by definition, the more accurate marker for peripheral IS. While GIR represents the amount of exogenous glucose required to maintain euglycemia, GD reflects the total amount of glucose (exogenous as well as endogenous glucose produced by the liver) that is finally taken up by the tissues and thus removed from circulation. As muscle is largely responsible for insulin-mediated glucose disposal (more than 80%; (33)), the GD index should reflect muscle IS best.
Male obese ZDF rats showed the largest deviation from the correlation. These rats were about 20 weeks old and exhibited a manifest diabetic condition. They were functionally highly insulin resistant, as reflected by a low GIR and elevated rates of EGP during the clamp, despite exhibiting similar, or rather lower, IMCL levels compared to rosiglitazone-treated littermates. Most likely, the “paradoxically” low IMCL content in these overtly diabetic animals is a consequence of their metabolic decompensation characterized by stagnant body weight development due to glucosuria and a catabolic lipid metabolism (16).
Rosiglitazone is used for the treatment of insulin-resistant states in type 2 diabetes and is known to exert its insulin sensitizing effects, among others, via redistribution of fat from nonadipose tissue back into the adipocytes (34, 35). This reduction of tissue lipid availability was reported to be accompanied by increases in body fat mass and body weight (36), whereas metabolic plasma parameters as well as peripheral IS were ameliorated both in patients and in animal models (37, 38). Literature data regarding the effect of rosiglitazone on IMCL content are inconsistent. Both a marked reduction of IMCL in animals (16, 17, 39) and no changes in IMCL levels in humans due to rosiglitazone treatment were reported (38).
In our study, treatment of male obese ZDF rats with rosiglitazone resulted in a marked improvement of IS in terms of elevated GIR and diminished EGP during hyperinsulinemia, whereas IMCL values were not significantly different from that of obese diabetic controls. In this context, it is important to mention that in several additional studies ((16); our data, not yet published) we performed longitudinal IMCL measurements in untreated as well as in rosiglitazone-treated obese male ZDF rats. These experiments revealed a rapid (within several days), pronounced (fourfold), and persistent (throughout the whole study) reduction of IMCL levels following long-term rosiglitazone treatment, indicating a pronounced effect of rosiglitazone on lipids stored within muscle cells in this animal model. The similar IMCL levels in untreated diabetic and in rosiglitazone-treated, nondiabetic obese ZDF rats in the present study are due to two different physiologic conditions: the catabolic state of the untreated diabetic group and the “healthy-on-treatment” situation of the treated. Thus, the pronounced effect of rosiglitazone on IMCL content is masked by the impact of overt diabetes on the very same parameter in the respective control group.
Rosiglitazone treatment did not significantly increase insulin-mediated muscle glucose uptake, as represented by GD. However, whole-body IS in terms of GIR was clearly improved in males and females, as well as insulin-mediated suppression of EGP in male obese rats. These findings confirm the observation that rosiglitazone exerts an insulin-sensitizing effect on the liver (36).
The correlation between IMCL and parameters of IS was confirmed when the data set for IMCL measured in rosiglitazone-treated ZDF rats was included. In female ZDF rats, despite improved values for GIR and GD, IMCL content in SOL tended to be unchanged or even elevated in this group when compared to untreated controls. However, in TIB IMCL was lower in the treatment group and thus matched better to the correlation analysis, confirming the inverse correlation between IMCL and IS. Interpretation of the inconsistent data for IMCL in this animal model turned out to be difficult, because, to our knowledge, no literature data regarding IMCL levels in female ZDF rats exist.
In our study, IMCL measured both in TIB and in SOL correlates well with IS in various rat models of IR. In literature, however, the muscle fiber type in which the lipid is stored is considered of a certain importance. Kelley and his group (40) postulate that skeletal muscle with sufficient oxidative enzyme equipment as the SOL can easily cope with lipid overflow, whereas lipid stored within a predominantly glycolytic muscle type may be fatal. As a consequence, IMCL determined in the glycolytic TIB might be more relevant with regard to IS than IMCL content in the oxidative SOL. This conclusion is confirmed by other publications. While in the TIB of Zucker fatty rats reduced IMCL levels following rosiglitazone treatment were found (17, 39), Mayerson and colleagues (38) observed no changes in IMCL content in the SOL of type 2 diabetics after 3 months treatment with rosiglitazone, although functional IR as assessed by glucose clamp technique was ameliorated. In our study, however, no significant superiority of one muscle type was found.
In summary, the present data show that IMCL content correlates well with IS in different animal models of IR. The validity of IMCL as a tissue biomarker for IS was confirmed in diet-induced models as well as in genetic ones—with the exception of overtly diabetic animals, due to their metabolic decompensation. Thus, we conclude that in preclinical biomedicine MR spectroscopically determined IMCL content is helpful to estimate the degree of IR in various rat models and to characterize the insulin-sensitizing potential of new drugs. Combined with the advantage of repeated, noninvasive measurements in longitudinal pharmacological studies allowing intraindividual monitoring and comparisons, this technical refinement promises a reduction of animal numbers as well as a more in-depth knowledge of the close interrelation of lipid and carbohydrate metabolism in IR.
We thank M. Roden, Vienna, for helpful discussions during the preparation of the manuscript.