Decreased Expression of Heat Shock Protein 72 In Skeletal Muscle of Patients With Type 2 Diabetes Correlates With Insulin Resistance

The clinical characteristics of the subjects participating in the two protocols are shown in Table 1. In protocol 1, six type 2 diabetic and six nondiabetic control subjects participated in a 3-h euglycemic-hyperinsulinemic (blood glucose 5 mmol/l, insulin 70 mU/l) clamp. A needle biopsy from the vastus lateralis muscle was performed before the clamp. In protocol 2, 12 monozygotic twin pairs discordant for type 2 diabetes and 12 healthy age- and sex-matched control subjects with normal glucose tolerance (NGT) and without a family history of diabetes and hypertension participated. In this protocol, the clamp was combined with indirect calorimetry (see below) to allow measurements of substrate oxidation as previously explained (16). Monozygosity of the twins was confirmed by genetic markers (17).

All studies were started in the morning after a 10-h overnight fast. A polyethylene catheter was inserted into an antecubital vein for infusion of test substances. Another polyethylene catheter was inserted into a contralateral wrist vein for blood sampling. This hand was placed and maintained in a heated plexiglas box to obtain arterialized venous blood. Plasma glucose was normalized in the diabetic patients before each study by a prior intravenous insulin infusion given in order to overcome the known compensatory effect of hyperglycemia per se, as previously explained (16,17). The insulin infusion was stopped when the plasma glucose concentration had declined to a near-normal level, and a further period of 30 min elapsed while the plasma glucose concentration continued to decline to normoglycemia (i.e., plasma glucose concentration ∼5 mmol/l). This time point marked the initiation of each experiment (time −120 min), when the tritiated glucose tracer bolus as described below was given. Except for the prior insulin infusion, the control subjects were studied in the fasting euglycemic state using a similar protocol with a primed continuous glucose tracer infusion from −120 min. Baseline measurements were performed during a predefined baseline “steady-state” period from −30 to 0 min, when tracer equilibrium was achieved. Insulin (Actrapid; Novo-Nordisk, Bagsvaerd, Denmark) was infused from 0 to +180 min at a constant rate of 40 mU · m⁻² · min⁻¹ in both the type 2 diabetic twins and the nondiabetic twins and control subjects. Plasma glucose concentration was maintained constant at euglycemia, using a variable glucose infusion. Plasma glucose concentration was monitored in arterialized blood every 5–10 min using an automated glucose oxidase method (Glucose Analyser 2, Beckman Instruments, Fullerton, CA). Measurements of insulin-stimulated glucose turnover rates were performed from +150 to +180 min during clamps (i.e., predefined insulin stimulated steady-state periods). The rate of nonoxidative glucose metabolism was calculated as the difference between total glucose disposal and glucose oxidation as measured by indirect calorimetry. Plasma concentrations of glucose and insulin were measured every 10 min during both baseline and insulin-stimulated steady-state periods. The glucose clamp studies were combined with a primed (22 μCi) continuous (0.22 μCi/min) infusion of [3-³H]glucose (New England Nuclear, Boston, MA) to obtain valid measurements of peripheral glucose turnover rates as previously explained (17). Glucose turnover rates were expressed as milligrams per kilogram of fat-free mass (FFM) per minute and are presented throughout the paper as the mean values calculated over the two 30-min steady-state periods. Total body fat content (and thus FFM) was measured using the bioimpedance method (18). Needle biopsies were obtained in the basal state (time 0 min) from the vastus lateralis muscle. The biopsies were immediately frozen in liquid nitrogen and stored at −80°C until analyzed. Informed consent was obtained from all subjects. The protocols were approved by the local ethics committees, and the study was conducted according to the principles of the Declaration of Helsinki.

Total RNA was prepared from the muscle biopsies by the method of Chomczynski and Sacchi (19). RNA was quantitated by measuring absorbance at 260 nm in a spectrophotometer, and its quality and integrity were estimated from the 260:280-nm absorbance ratios. RNA samples were stored at −80°C. Oligonucleotides were designed by using the sequence of human Hsp-72 cDNA, published by Milner and Campbell (20), coding for the 72-kDa-molecular weight inducible heat shock protein (GenBank accession no. [GBAN] M59830 or M34269). The sequences chosen as primers were 100% conserved between the rat, mouse, and human cDNAs (21) and GenBank accession nos. M59830, L16764, and M35021. The four oligonucleotides (listed below) were synthesized by DNA Technology (Aarhus, Denmark). The sequence of the oligonucleotides were as follows: primer 1: AGA TCA TCG CCA ACG ACC AG (565–584, sense); primer 2: GTC AGC ACC ATG GAC GAG ATC TCC TCA GTG CTT CAT GTC CGA CTG (854–835 + 754–731, antisense); primer 3: AAT TTA ATA CGA CTC ACT ATA GGG AAG ATC ATC GCC AAC GAC CAG (T7 promoter sequence [22] plus 565–584, sense); primer 4: CAC CAT GGA CGA GAT CTC CT (854–835, antisense). Reverse transcription was performed using M-MLV RTase (Life Technologies, Rockville, MD), according to the vendor’s recommendation. If the cDNA mixtures were not used immediately for the PCR, they were stored at −20°C. PCR was performed with standard mixtures using a Perkin-Elmer DNA Thermal Cycler. The amplification cycle profiles were 95°C (denaturing) for 1 min, 58°C (annealing) for 1 min, and 72°C (synthesis) for 30 s. The cycle number varied between 30 and 55, according to the actual requirement of the experiments.

For the preparation of mini-genes coding for different segments of HSP-70 cDNA, 1 μg of total cellular RNA (isolated from the heart of a heat-treated SPRD rat) was reverse-transcribed using primer 2 or 4. cDNAs from the two reactions were then amplified using primers 1 and 4, resulting in the production of a 210-bp-long (80 bp was deleted by primer 2 during reverse transcription) or a 290-bp-long product. The fragments were isolated from agarose gel and were used as templates for the subsequent rounds of PCR experiments. In these experiments, using primers 3 and 4, the PCR products were tagged 5-prime to the DNA constructs with the promoter of the T7 bacteriophage RNA polymerase. The pieces containing the T7 promoter were isolated from agarose gels and were stored at −20°C in known concentrations. Synthetic RNA (sRNA) standards were prepared by using the constructed mini-genes and reagents necessary for in vitro transcription (SP6/T7 Transcription Kit; Boehringer-Mannheim, Mannheim, Germany), as recommended by the manufacturer. Aliquots containing 1 μg sRNA in 10 μl diethyl pyrocarbonate-water were prepared and were stored at −80°C. The quantitative RT-PCR was performed as follows: 200 ng of total cellular RNA from each sample was mixed together with at least four different concentrations (10-fold and/or 3.16-fold serial dilutions) of the deleted-length sRNA template and was subjected to the RT-PCR procedure described above. PCR products were separated on agarose gels, stained with ethidium bromide (example is shown in Fig. 1A), and photographs were scanned by a GS-700 Imaging Densitometer (Bio-Rad, Hercules, CA). Ratios of the 210-bp (competitor) and 290-bp (target) bands were determined and were plotted against the known concentrations of the competitor on a double-log scale. Target RNA concentrations were determined from the x-axis intercept at the point where the ratio of the two products was equal (Fig. 1B). Measurements were repeated with each sample, and if the SD between the two measurements was <30%, the average results were taken as final. If the deviation was larger, the experiment was repeated again, until the range (SD <30%) was reached or until the correlation coefficient calculated from all the combined data was >0.9.

The inducible Hsp-72 and the constitutively expressed cognate Hsc-73 show significant homology at both nucleic acid and protein levels (20). Therefore, we determined the specificity of the assay to detect Hsp-72 in different biological samples. The nucleotide sequence of the PCR products made from RNA isolated from a nondiabetic subject and a patient with type 2 diabetes was determined with standard dideoxy-sequencing (ABI equipment; Institute of Biochemistry, BRC, Szeged, Hungary). In both subjects, the sequencing reaction showed the presence of only one PCR template, which had 98.4% sequence identity with the human Hsp-72 cDNA (GenBank accession no. M59830).

Accuracy was checked by determining the changes in Hsp-72 mRNA concentrations after heat shock in cultured NIH-3T3 cells. These cells express low basal levels of Hsp-72 but respond to heat stress robustly (23). After heat shock (43°C for 30 min), total RNA was isolated from the cells at different time intervals and Hsp-72 mRNA level was determined by the competitive RT-PCR assay. At 30 min after the heat shock, a significant elevation (80- to 120-fold) of Hsp-72 mRNA level was seen, which lasted for ∼60 min. At 4 h after the heat shock, the Hsp-72 mRNA had approached the baseline. In two independent experiments, highly comparable kinetics were obtained. Very similar kinetic data were reported by Deguchi et al. (24) by measuring the transcription of the hsp-70 gene with the nuclear run-on method.

Intra- and interassay variation of the competitive RT-PCR method was determined using known concentrations of target sRNA. Intra-assay variation was determined in three independent experiments, using three different concentrations of the target, and was found to be 6.6 ± 3.5, 9.7 ± 7.2, and 13.8 ± 12.1%, respectively. The interassay variation was 34.4% and was determined from five separate experiments, using new reagents each time.

Nonparametric statistical methods were used: Mann-Whitney U test for group comparison and Wilcoxon’s signed-rank test for paired data. Correlation analysis was done with the Spearman test; rho (R) correlation coefficients and probability values were calculated. P < 0.05 was considered statistically significant. All statistical calculations were performed with StatView software (SAS Institute, Cary, NC).

In the type 2 diabetic patients in protocol 1, insulin-stimulated glucose uptake rate was ∼60% lower than in the nondiabetic control subjects (P = 0.0163) (Table 1). In protocol 2, the fasting plasma glucose, insulin, and 2-h glucose concentrations were slightly higher in the nondiabetic co-twins than in the control subjects; seven of them had impaired glucose tolerance (IGT) and were slightly insulin resistant. The diabetic twins had 55% lower rates of insulin-stimulated glucose uptake (P = 0.0002), 63% lower rates of glucose storage (P = 0.0005), and 37% lower rates of glucose oxidation (P = 0.0003) compared with the control subjects. The nondiabetic monozygotic co-twins also had 25% lower rates of insulin-stimulated glucose uptake (8.5 ± 0.8 vs. 11.3 ± 0.9 mg · kg FFM⁻¹ · min⁻¹, P = 0.0404), which was mainly caused by a 37% decrease in the rate of glucose storage (4.8 ± 0.6 vs. 7.6 ± 0.9 mg · kg FFM⁻¹ · min⁻¹, P = 0.0243) compared with the control subjects. Lipid oxidation rates during insulin infusion were elevated in the type 2 diabetic twins compared with the other study groups, but they were not statistically different in nondiabetic twins and control subjects (Table 1).

In protocol 1, despite the substantial individual deviations within the groups (from 3 to 10 × 10⁶ copy/μg total RNA in the diabetic subjects and from 23 to 92 × 10⁶ copy/ μg total RNA in the control subjects), there was an ∼10-fold, statistically significant (P = 0.0039) difference between the two groups (Fig. 2). In protocol 2, the amount of Hsp-72 mRNA was also compared between monozygotic twins discordant for type 2 diabetes and healthy control subjects. The control subjects with no family history of diabetes showed the highest level of Hsp-72 mRNA, which was significantly different from the values in the diabetic (P = 0.0001) and nondiabetic twins (P = 0.0001) (Fig. 3A). No significant difference was found in Hsp-72 mRNA between the diabetic and nondiabetic groups of co-twins (P = 0.0901) (Fig. 3A); however, if the difference between the twin-pairs were analyzed with matched rank test, a statistically significant difference was found between the diabetic and nondiabetic (NGT or IGT) twin-pairs (P = 0.028) and also between diabetic twins and their co-twins with NGT (P = 0.0431). Twins with IGT did not show statistically significant alteration compared with their diabetic co-twins (P = 0.1755) (Fig. 3B).

When the Hsp-72 mRNA concentration was correlated with the rate of glucose uptake, as measured in patients participating in protocol 1, a tendency (r = 0.525) of higher heat shock protein mRNA level coinciding with stronger capacity to take up glucose could be observed, but this tendency was not statistically significant (P = 0.0814, data not shown). Correlation between Hsp-72 mRNA levels and different parameters of glucose and lipid metabolism were examined within different subgroups in protocol 2 (summarized in Figs. 4 and Table 2). In the subgroup of nondiabetic subjects (pooled control subjects and nondiabetic co-twins), strong positive correlations with the insulin-stimulated rate of glucose uptake (r = 0.581, P = 0.0053) (Fig. 4A) and glucose storage (r = 0.559, P = 0.0073) (Fig. 4C), and inverse correlations with the rate of lipid oxidation (r = − 0.425; P = 0.0416; Fig. 4d) and the 2-h glucose level (r = −0.716, P = 0.0006) (Table 2), were found. In the subgroup of control subjects without family background of diabetes (n = 12), the Hsp-72 mRNA levels showed negative correlations with the 2-h glucose levels (r = −0.744, P = 0.0236) (Table 2) and with the rate of lipid oxidation (r = −0.634, P = 0.0142) (Fig. 4D). The pooled data of twins showed that Hsp-72 mRNA level in this subgroup correlates positively with glucose uptake (r = 0.519, P = 0.0128) (Fig. 4A), glucose oxidation (r = 0.564, P = 0.0068) (Fig. 4B), and glucose storage (r = 0.514, P = 0.0137) (Fig. 4C). In the subgroup analysis of diabetic twins, there was positive correlation (r = 0.594, P = 0.0487) (Fig. 4B) between the amount of Hsp-72 mRNA and the rate of insulin-stimulated glucose oxidation, and there was negative correlation between Hsp-72 mRNA levels and oxidation of lipids (r = −0.643, P = 0.0329) (Fig. 4D). No statistically significant correlations were found between the stress protein transcript levels and either fasting plasma glucose or insulin concentrations in any of the subgroups (Table 2).

We provide the first evidence that there is decreased expression of the Hsp-72 mRNA in skeletal muscle from patients with type 2 diabetes, and that the level of Hsp-72 mRNA correlates with the rate of insulin-stimulated glucose and lipid turnover and glucose tolerance. In both of our protocols, we found an order of magnitude of difference between Hsp-72 mRNA levels measured in type 2 diabetic patients compared with control subjects without a family history of diabetes. Because we determined the amount of specific mRNA from the muscle biopsies, we have no proof as to whether the decreased Hsp-72 mRNA levels are also reflected by a decreased amount of protein. Measurement of expression at the level of transcript does not provide information about post-transcriptional control of gene expression, and, indeed, recent experimental evidence suggests that in general there is no obvious correlation between mRNA and protein expression levels (25,26). However, in the experiments cited above, the translation efficiency of different transcripts were compared where control mechanisms determining nuclear export, transcript stability, translational regulation, and protein degradation were acting on different templates, obviously with different outcomes. Comparing the expression of the same gene under different metabolic circumstances at both the transcriptional and the translational levels might result in a different conclusion. Saghizadeh et al. (27) studied the expression of tumor necrosis factor-α message in muscle biopsies of normal and diabetic subjects and found that the cytokine mRNA is upregulated in the diabetic tissue. The same upregulation of mRNA was found in cultured muscle biopsies of diabetic patients, with a concomitant increase in the production of the cytokine as well (27). The magnitude of the difference between Hsp-72 mRNA levels of the healthy and diabetic groups (10-fold) in our experiments, and the fact that stress protein expression in general is strongly regulated at the level of transcription (28,29), support the idea that in this case as well the differences in the levels of transcript were translated into differences in the amounts of protein.

Hsp-72 mRNA levels in skeletal muscle of the patients correlated with several parameters of carbohydrate and lipid metabolism. Already in protocol 1, with a limited amount of data, the rate of insulin-stimulated glucose uptake showed a tendency to increase with elevated heat shock protein mRNA expression. Analysis of the data of different subgroups of patients from protocol 2 showed that depending on the subgroup and on the number of patients in the subgroups, Hsp-72 mRNA level correlated strongly with different insulin-stimulated metabolic changes, such as the increase in the uptake, oxidation, and storage of glucose and the decrease in the oxidation of lipids. In addition, in the subgroup of pooled nondiabetic subjects and control subjects without family background of diabetes, the amount of the stress protein transcript showed strong correlation with the 2-h glucose levels during the oral glucose tolerance test. Although we cannot conclude from the data that the decreased Hsp-72 mRNA levels are the cause or consequence of the deranged metabolic state in type 2 diabetes, there is some support for a causal relationship. Because Hsp-72 mRNA expression did not show any correlation with either fasting plasma glucose or insulin level in any of the subgroups, it confirms our hypothesis that the altered stress protein mRNA level is not simply the consequence of the chronically elevated plasma glucose or insulin. The possible regulatory effect of glucose on the expression of the heat shock protein mRNA can also be ruled out by the fact that during our protocols, diabetic patients, at 1 h before the time of biopsy, were made normoglycemic. We cannot exclude the possibility that the acute administration of insulin (9.4 ± 2.1 IU added intravenously during a period of 52 ± 8 min) used for adjusting euglycemia in diabetic subjects affected Hsp-72 mRNA levels. However, this bolus was not given to nondiabetic co-twins, who possessed similarly low levels of the heat shock protein mRNA, and acute insulin treatment has been shown to increase the expression of the inducible 70-kDa-mass heat shock proteins in hepatocytes in vitro (30). This insulin-stimulated increase is opposed by our finding. Equally important is the fact that there is much less correlation between muscle Hsp-72 mRNA content and the basal rates of glucose and lipid metabolisms. Using the same subgroups, Hsp-72 mRNA showed correlation with basal-rate glucose oxidation or storage only in the group of twins or in the diabetic group, respectively (data not shown). These data strongly support the possible association between the stress protein and the action of insulin.

The possible nature of this potential association is speculative at this point. Stress proteins represent a highly conserved family of polypeptides with significant involvement in the creation and functional preservation of the protein machinery in living cells (31). Evidence has accumulated showing that stress proteins are indispensable for the cells not only under stressful conditions but also under physiological circumstances (32,33). The inability of an essential protein to form its native structure under physiological conditions may be the basis for human diseases (34).

The insulin receptor requires the contribution of different chaperones to its folding, assembly, and transport (35). It has been shown in fetal hepatocytes subjected to heat stress that HSP-70s associate with the insulin receptor to prevent recycling and damage (36). Given the conservation of the signaling cascade starting from the insulin receptor and the complexity of the downstream events involved in the transduction of the signal, it can easily be imagined that signal accuracy is maintained by scaffolding proteins, which preserve specificity of the reaction and assist in stabilizing the otherwise weak interaction between components. The scaffolding and catalytic roles of heat shock proteins is known for the glucocorticoid receptor (rev. in 37), and recently even catalytic activity has been shown to depend on the interaction between the enzyme and heat shock proteins (38).

GLUT4 is responsible for most of the insulin-stimulated increase in glucose transport (39, 40), and this transporter is another possible candidate for heat shock protein interference. The GLUT4 transporting vesicles contain and interact with several different proteins, some of which are likely to require constant chaperone inspection. Both GLUT4 (41) and Hsp-72 (42) are known to interact with F-actin, which plays an indispensable role in the relocalization of intracellular compartments. Although no heat shock protein associated with GLUT4 vesicles have been identified so far, several putative glucose transporter binding proteins have been recognized, and their molecular weights (28, 70, and 85 kDa) are very close to those of the most abundant stress proteins (43). Interestingly, physical exercise, known to increase insulin sensitivity in diabetic patients (44) by inducing translocation of the GLUT4 transporter to the plasma membrane (45), also elevates heat shock protein expression in skeletal muscle during workout (rev. in 46). Also relevant to the issue is the association that has been shown between the Hsp-72 protein and the peroxisome proliferator-activated receptors (PPARs) in that they form a complex in vivo in the rat. The possibility thus remains that Hsp-72 may play a role in the folding, subcellular localization, and/or signaling pathway of PPARs (47) as well.

We attempted to examine the origin of the defect in heat shock protein mRNA expression by studying monozygotic twins discordant for type 2 diabetes. This is a very special population, in which individuals possess some kind of type 2 diabetes genetic constitutions without having overt diabetes. This population is superior to first-degree relatives because relatives on average may only have a 50% chance of carrying the true type 2 diabetes genetic constitution, whereas identical twins are 100% predisposed to it. Defects in insulin action precede the overt disease, and in our study the nondiabetic co-twins were not entirely discordant for glucose tolerance and insulin sensitivity, i.e., the nondiabetic co-twins were significantly more insulin resistant and had higher glucose levels than the control subjects. Despite this, however, compared with control subjects, nondiabetic twins showed severely decreased expression of Hsp-72 mRNA, which was similar to that of frank diabetic twins, suggesting that the decrease in Hsp-72 mRNA in the nondiabetic co-twins was not the consequence of insulin resistance; rather, it might represent an inherited trait. On the other hand, when Hsp-72 mRNA content of twin-pairs with the same genetic background was compared, nondiabetic twins had higher message level than their diabetic pairs. Interestingly, when twins with IGT were compared with their co-twins, no notable deviations were found, whereas Hsp-72 mRNA levels in twins with NGT differed statistically significantly from those of their diabetic pairs. The results from the monozygotic twins discordant for type 2 diabetes are more consistent with a genetic as compared with a nongenetic origin of this defect; however, they are in harmony with the dogma that most forms of type 2 diabetes are polygenic, with superimposed environmental influences.

In conclusion, the finding of decreased levels of Hsp-72 mRNA in skeletal muscle of patients with type 2 diabetes and its relationship with insulin resistance raises the question of whether heat shock proteins are involved in the pathogenesis of skeletal muscle insulin resistance in type 2 diabetes. During the preparation of the revised version of this manuscript, preliminary data using microarrays to compare the expression profile between insulin-resistant and insulin-sensitive subjects reported decreased Hsp-70 expression in vastus lateralis muscle from the insulin-resistant individuals (48).

This study was supported in Malmö by grants from the Sigrid Juselius Foundation, the Swedish Medical Research Council, the Academy of Finland, the Swedish Diabetes Association, the European Community (BMH4-CT95-0662), the Novo Nordisk Foundation, and the Lundberg Foundation.