Adipocytes play an integrative role in the regulation of energy metabolism and glucose homeostasis in the human body. Functional defects in adipocytes may cause systemic disturbance of glucose homeostasis. Recent studies revealed mitochondrial abnormalities in the adipose tissue of patients with type 2 diabetes. In addition, patients with mitochondrial diseases usually manifest systemic metabolic disorder. However, it is unclear how mitochondrial dysfunction in adipocytes affects the regulation of glucose homeostasis. In this study, we induced mitochondrial dysfunction and overproduction of reactive oxygen species (ROS) by addition of respiratory inhibitors oligomycin A and antimycin A and by knockdown of mitochondrial transcription factor A (mtTFA), respectively. We found an attenuation of the insulin response as indicated by lower glucose uptake and decreased phosphorylation of Akt upon insulin stimulation of adipocytes with mitochondrial dysfunction. Furthermore, the expression of glucose transporter 4 (Glut4) and secretion of adiponectin were decreased in adipocytes with increased ROS generated by defective mitochondria. Moreover, the severity of insulin insensitivity was correlated with the extent of mitochondrial dysfunction. These results suggest that higher intracellular ROS levels elicited by mitochondrial dysfunction resulted in impairment of the function of adipocytes in the maintenance of glucose homeostasis through attenuation of insulin signaling, downregulation of Glut4 expression, and decrease in adiponectin secretion. Our findings substantiate the important role of mitochondria in the regulation of glucose homeostasis in adipocytes and also provide a molecular basis for the explanation of the manifestation of diabetes mellitus or insulin insensitivity in a portion of patients with mitochondrial diseases such as MELAS or MERRF syndrome.
Type 2 diabetes mellitus (T2DM), also known as non-insulin-dependent diabetes mellitus, is characterized by insulin resistance of insulin-responsive tissues such as skeletal muscle and adipose tissues or reduced insulin secretion of β cells. T2DM has received increasing attention from clinicians and biomedical researchers in recent decades due to its high prevalence and social and economic burden on our society . It has been held that the pathogenesis of T2DM is quite complicated and involves multiple etiological factors including genetic mutation, sedentary lifestyle, unhealthy dietary habit and environmental toxins and is aggravated by complications such as hypertension, dyslipidemia and cardiovascular disease .
Mitochondria are involved in the regulation of energy metabolism and their defects are associated with aging and a variety of diseases including cardiovascular diseases, neurological disorders, myopathies, muscle weakness and cancer . It has been documented that a portion of patients with mitochondrial diseases harboring an A to G transition at the 3243th nucleotide position of mitochondrial DNA (mtDNA) (termed the A3243G mutation) had the symptoms of T2DM . Recent studies have revealed a decrease in the copy number of mtDNA and mitochondrial density, downregulation of mitochondrial genes, abnormalities of mitochondrial morphology, impairment of oxidative phosphorylation and β oxidation of fatty acids in skeletal muscles or adipose tissues of T2DM patients or mice with diabetes [5-11]. However, the molecular mechanisms underlying these metabolic disorders and insulin resistance caused by mitochondrial dysfunction have remained elusive. Furthermore, most of the previous studies have focused on insulin insensitivity of the skeletal muscle, and little attention has been paid to the abnormalities of adipose tissues.
It has been established that adipose tissues are indispensable for the maintenance of glucose homeostasis in the human body. Recently, glucose intolerance and insulin insensitivity were observed in mice knocked out of the adipocyte-specific glucose transporter 4 (Glut4) . Minamino et al.  demonstrated that the removal of dysfunctional epididymal fats from insulin-insensitive mice reversed physiological function in the maintenance of glucose homeostasis. However, the healthy mice that had received subcutaneous implantation of dysfunctional adipose tissues developed insulin resistance and hyperglycemia . Adipose tissues are able to secrete adipokines to regulate glucose metabolism in other tissues such as muscle, liver and pancreas to maintain glucose homeostasis . It was reported that dysregulation of adipokine secretion in adipose tissues is associated with T2DM [15, 16]. The above-mentioned findings clearly indicate that adipose tissues play a pivotal role in the integration and maintenance of glucose homeostasis in the human body.
In the present study, we investigated the involvement of mitochondrial dysfunction in adipocytes on the pathogenesis of insulin insensitivity and T2DM. We examined the roles of mitochondria in the insulin sensitivity and adiponectin secretion of mature adipocytes by treatment of cells with respiratory inhibitors and genetic knockdown of mitochondrial transcription factor A (mtTFA), respectively. The results revealed that an increase of intracellular H2O2 elicited by mitochondrial dysfunction could lead to lower insulin response of adipocytes through a decrease in Akt activation and Glut4 expression. In addition, the decreased secretion of adiponectin from adipocytes with defective mitochondria may impair the glucose utilization of muscle cells leading to systemic disturbance of glucose homeostasis.
Mitochondrial dysfunction induced by respiratory inhibitors leads to insulin resistance of adipocytes
To investigate whether mitochondrial dysfunction may lead to insulin resistance, we treated the adipocytes with inhibitors of mitochondrial respiratory enzymes such as antimycin A and oligomycin A. The results showed that the mitochondrial oxygen consumption rate and intracellular ATP content of adipocytes were decreased after treatment for 24 h with oligomycin A and antimycin A, respectively (Fig. 1A,B). Moreover, the intracellular levels of H2O2 of the treated adipocytes were significantly higher than those of the controls (Fig. 1C). We then measured the insulin-stimulated glucose uptake in the adipocytes with mitochondrial dysfunction induced by treatment with respiratory inhibitors. The results showed a significant increase in the basal glucose uptake in adipocytes treated with oligomycin A or antimycin A while there was no stimulation in glucose uptake after addition of insulin compared with the control (Fig. 1D).
Knockdown of mtTFA leads to insulin resistance of adipocytes
To avoid possible side effects of the respiratory inhibitors, we knocked down mtTFA by lentivirus infection to suppress the biogenesis and bioenergetic function of mitochondria. The results revealed 40% decrease of mtTFA at the protein level (Fig. 2A). The copy number of mtDNA (Fig. 2B) and the expression levels of mtDNA-encoded polypeptides (ND6, COX1 and COX2) and nuclear DNA-encoded polypeptides (NDUFA9 and UQCRC2) that constitute the respiratory enzyme complexes (Fig. 2C) were all decreased in adipocytes with mtTFA knockdown. This indicates a decrease in mitochondrial biogenesis. In order to know whether suppressed expression of mtTFA also affects the respiratory function of mitochondria per se, we measured the electron transport activities of Complexes I and IV in isolated mitochondria. The results showed that, in addition to a reduction of mitochondrial biogenesis, the activities of respiratory enzyme Complexes I and IV of mitochondria were decreased in the adipocytes with mtTFA knockdown (Fig. S1). Moreover, knockdown of mtTFA led to a decrease in the overall bioenergetic function of mitochondria in adipocytes indicated by a significant decline of State 3 respiration and intracellular ATP level (Fig. 2D,E). Similar to the results obtained from treatment of cells with respiratory inhibitors, we observed an attenuation of the insulin-stimulated glucose uptake in the adipocytes with mtTFA knockdown (Fig. 2F). These results together suggest that mitochondrial dysfunction leads to a decline in the capability of insulin-stimulated glucose uptake and thereby results in insulin resistance in adipocytes.
Impairment of insulin signaling and decrease of Glut4 in adipocytes with mitochondrial dysfunction
To unravel the mechanisms involved in insulin resistance of adipocytes, we investigated the effects of mitochondrial dysfunction on the insulin signaling pathway including phosphorylation of insulin receptor substrate 1 (IRS1) on Tyr612, one of several tyrosine residues involved in PI3K binding, and phosphorylation of Akt on Ser473 and its downstream target AS160 on Thr642. The results showed that the phosphorylation of IRS1, Akt and AS160 upon insulin stimulation was diminished in adipocytes that had been treated for 24 h with oligomycin A and antimycin A, respectively (Fig. 1F). In addition, there was a decrease of insulin-activated protein phosphorylation in the adipocytes with mtTFA knockdown (Fig. 2H). Together, these results suggest that mitochondrial dysfunction inhibits insulin signaling and subsequent events which culminate in insulin resistance in adipocytes.
On the other hand, we investigated the expression levels of glucose transporters in adipocytes with mitochondrial dysfunction. After treatment of adipocytes with 10 μg·mL–1 of oligomycin A or 1 μm of antimycin A for 24 h, the protein expression level of Glut4 was significantly decreased (Fig. 1E). Similarly, the adipocytes with a suppressed expression of mtTFA also expressed lower levels of Glut4 (Fig. 2G). In addition, we observed that the Glut4 was significantly downregulated by oligomycin A in a time-dependent manner (Fig. 3A). Interestingly, glucose transporter 1 (Glut1), which is responsible for basal glucose uptake in most cells, showed an opposite change (Figs 1E and 3A). Further study on the expressions of Glut4 and Glut1 revealed that the mRNA level of Glut4 was significantly decreased at 4 h and that of Glut1 was increased at 8 h after oligomycin A treatment (Fig. 3B). These findings indicate that mitochondrial dysfunction inhibits the expression of the Glut4 gene at the transcriptional level. Taken together, the results led us to suggest that these dual effects resulting from mitochondrial dysfunction may account, at least in part, for insulin resistance of adipocytes.
Mitochondrial-dysfunction-elicited reactive oxygen species (ROS) are involved in the downregulation of Glut4 and insulin insensitivity in adipocytes
Because mitochondrial dysfunction in adipocytes resulted in an overproduction of intracellular H2O2, we then investigated whether excess ROS in adipocytes is a possible contributory factor to insulin insensitivity. We analyzed the changes of intracellular H2O2 in adipocytes that had been treated with 10 μg·mL–1 oligomycin A for 24 h, and the results revealed that the intracellular H2O2 level was significantly increased at 4 h and maintained at a high level until 24 h (Fig. 3C). In this time course experiment, we found that the increase of intracellular H2O2 occurred concurrently with the downregulation of Glut4 expression. To examine whether ROS can regulate the Glut4 expression in adipocytes, we treated adipocytes with 200 μm H2O2 for 3, 6 and 12 h, respectively, and determined the protein level of Glut4. We found that the ROS level was increased 30%–40% and the protein level of Glut4 was downregulated in adipocytes treated with exogenous H2O2 for 12 h (Fig. 3D,E). In order to investigate whether ROS elicited by mitochondrial dysfunction plays a crucial role in insulin insensitivity of adipocytes, we pretreated cells with N-acetyl cysteine (NAC), an antioxidant. The data showed that the ROS level was decreased and the insulin response was recovered by NAC in adipocytes treated with respiratory inhibitors (Fig. 3F,G). These results suggest that the increased production of ROS in adipocytes with mitochondrial dysfunction contributes to the decrease of the expression of Glut4 and insulin insensitivity.
Decrease of adiponectin secretion in adipocytes with mitochondrial dysfunction
Adiponectin has been considered the most important adipokine due to its higher concentration than the others and the observation that its serum level is negatively correlated with T2DM . These findings prompted us to investigate whether mitochondrial dysfunction would lead to any change of the adiponectin level in adipocytes. The results showed that the mRNA expression and the secreted amount of adiponectin were decreased in adipocytes treated with oligomycin A for 24 h and small hairpin RNA (shRNA) against mtTFA for 48 h, respectively (Fig. 4A–D). Moreover, we also demonstrated that the expression of adiponectin was downregulated in adipocytes with mitochondrial dysfunction induced by treatment of oligomycin A or sh-mtTFA in a dose-dependent manner (Fig. S2). In order to examine the effect of adiponectin on glucose utilization in muscle cells, we measured the glucose uptake of C2C12 myotubes after treatment with different concentrations of adiponectin for 1 h. The results showed that adiponectin stimulated the basal and insulin-stimulated glucose uptake of C2C12 myotubes, respectively, in a dose-dependent manner (Fig. 4E). Furthermore, adiponectin increased insulin-induced phosphorylation of Akt on Ser473 in a dose-dependent manner (Fig. 4F). These results clearly indicate that the adiponectin secreted by adipocytes has an effect on the glucose metabolism of muscle cells. Taken together, these findings strongly suggest that adipocytes with mitochondrial dysfunction exert decreased glucose utilization in other tissues through the decrease of adiponectin secretion.
Accumulated evidence has established that adipocytes play an important role in the regulation of energy metabolism and glucose homeostasis, and that dysfunction of adipocytes may play a role in the pathogenesis of metabolic diseases such as T2DM and metabolic syndrome . Insulin resistance was observed when mitochondrial function was impaired by depleting mtDNA  or by adding a respiratory inhibitor to myocytes  or by knocking out the enzymes required for β oxidation of fatty acids in liver . However, very few studies focused on the possible impact of mitochondrial dysfunction on insulin sensitivity of adipocytes. Moreover, depletion of mtDNA with ethidium bromide and inhibition of respiratory function with inhibitors might affect the stability of the nuclear genome or damage other proteins and thereby lead to complicated results [18, 19]. In the present study, we showed that mitochondrial dysfunction directly affects insulin sensitivity of adipocytes. To avoid possible toxic effects of oligomycin A and antimycin A on cells, the dose and duration of treatment we chose did not lead to detectable cell death (data not shown). On the other hand, we knocked down mtTFA to inhibit the biogenesis and function of the mitochondria more specifically in adipocytes and avoid possible effects on the nuclear DNA. Although it has been shown that mtTFA knockdown during the process of adipocyte differentiation led to a decrease of insulin-stimulated glucose uptake , it is difficult to exclude the effect from the fewer adipocytes matured in the process of differentiation because mitochondrial biogenesis is essential for adipogenesis [22, 23]. In this study, we first knocked down mtTFA in mature adipocytes and demonstrated that insulin sensitivity and adiponectin secretion were concurrently decreased. Most importantly, we found that the decreased level of insulin sensitivity was actually reflected by the decrease of mitochondrial respiratory function. The more severe the mitochondrial defects caused by respiratory inhibitors the more pronounced the decrease in the induction of insulin signaling and glucose uptake of adipocytes upon insulin stimulation (Figs 1 and 2). These results have provided solid evidence to substantiate that the extent of mitochondrial dysfunction is related to the severity of diabetes and insulin resistance observed in T2DM patients .
Glut4 is highly expressed in adipose tissues and skeletal muscles and is also called the ‘insulin-responsive glucose transporter’ because it is translocated to the plasma membrane and executes its function upon insulin stimulation . The decrease in the amount of Glut4 is one of the factors involved in the pathogenesis of T2DM. Manipulation of Glut4 levels in transgenic mice has revealed that glucose homeostasis is highly dependent on the level of Glut4 expression . In this study, we showed that the protein expression of Glut4 in adipocytes was downregulated by treatment with oligomycin A in a time-dependent manner (Fig. 3A). This is mostly a result of inhibition of Glut4 gene expression at the transcriptional level because the same amplitude of change in the mRNA level was observed (Fig. 3B). Lim et al.  also observed a decline of Glut4 expression after oligomycin A treatment of myotubes for 24 h. It was reported that oxidative stress could impair the binding of transcription factors to the promoter of the Glut4 gene . We observed concurrent increase of H2O2 and downregulation of Glut4 upon oligomycin A treatment and a decrease in the protein expression of Glut4 in adipocytes exposed to oxidative stress (Fig. 3C–E). These findings suggest that oxidative stress elicited by mitochondrial dysfunction is the major factor that leads to insulin insensitivity of adipocytes. Interestingly, the expression of Glut1 was upregulated when Glut4 was downregulated in adipocytes with mitochondrial dysfunction (Fig. 3A,B), and this effect increased the basal glucose uptake in adipocytes (Fig. 1D, black bar). These observations were consistent with a report of the enhancement of glycolysis and increased Glut1 expression and glucose utilization in human cells with mitochondrial dysfunction . AMPK has been reported as a key mediator that upregulates glycolysis in human skin fibroblasts with mitochondrial defects caused by a pathogenic mtDNA mutation . However, there was no significant difference in the phosphorylation of AMPK in our adipocytes with mitochondrial dysfunction (data not shown). Further investigation is warranted to explore the signaling pathways that are involved in the upregulation of glycolysis in adipocytes with mitochondrial dysfunction. It has been reported that Glut1 translocation is partially involved in insulin-stimulated glucose uptake . By contrast, there was no obvious increase in glucose uptake of oligomycin-A-treated adipocytes upon addition of insulin (Fig. 1D). This might be due to impairment in the mechanism of Glut1 translocation to the plasma membrane.
On the other hand, we demonstrated that mitochondrial dysfunction of adipocytes not only leads to insulin insensitivity but also impairs the insulin response or glucose utilization in other tissues (e.g. muscle cells) through a decrease in the expression and secretion of adiponectin from adipocytes (Figs 4 and S2). In addition to serving a role in increasing glucose uptake and fatty acid oxidation in muscle and liver, adiponectin has been reported to assume another role in target organs to alleviate the insulin resistance and associated complications. A recent study also showed that adiponectin can act on muscle cells to increase the activity of PGC-1α and enhance mitochondrial oxidative metabolism . Moreover, adiponectin may increase nitric oxide bioavailability to reduce the ROS levels and inhibit the inflammatory response of vascular epithelial cells in mice with T2DM [31, 32]. These results indicate that adiponectin may serve as a critical physiological factor in the protection of epithelial cells from oxidative damage induced by oxidative stress and prevent the pathogenesis of atherosclerosis. Because of the multiple function of adiponectin, strategies to increase its level in blood circulation may be a potential therapy for T2DM. Similar to the results obtained in the present study, some studies showed that overproduction of ROS by mitochondrial dysfunction is involved in the downregulation of adiponectin expression. Furthermore, it was shown that a loss of mitochondrial antioxidant defense by knockout of peroxiredoxin 3 (Prx3), a thioredoxin-dependent mitochondrial peroxidase, or uncoupling protein 2 (UCP2) could decrease the expression of adiponectin via an aberrant regulation of oxidative stress [33, 34].
Based on the results obtained in this study, we propose a molecular mechanism (Fig. 5) that illustrates the important role that mitochondrial dysfunction of adipocytes plays in the pathogenesis of systemic impairment of glucose homeostasis in T2DM. The increase of intracellular H2O2 induced by mitochondrial dysfunction may result in insulin insensitivity through attenuation of the activation of insulin signaling, blockade of the gene expression and translocation of Glut4, and insufficient adiponectin secretion by adipocytes. These results have substantiated the role of mitochondrial dysfunction in the dysregulation of glucose homeostasis in adipose tissues and have provided useful information for the development of a new therapeutic strategy for the treatment of T2DM.
Materials and methods
Chemicals and antibodies
Respiratory inhibitors, including antimycine A, oligomycin A and KCN, were all purchased from Merck Inc. (Darmstadt, Germany). [3H]-2-deoxyglucose was supplied by Perkin-Elmer Life Sciences (Wellesley, MA, USA). 2′,7′-Dichlorodihydrofluorescein diacetate (DCFH2-DA) was purchased from Molecular Probes (Eugene, OR, USA). The primary antibodies used in this study were from the following sources: monoclonal antibodies against p-Akt (S473), Akt, p-AS160 (T642), AS160 and IRS1 were from Cell Signaling Technology (Danvers, MA, USA), p-IRS1 (Y612) was purchased from AbboMax (San Jose, CA, USA), the antibody against mtTFA was from LifeSpan BioSciences (Seattle, WA, USA), monoclonal antibodies against Glut1 and Glut4 were supplied by Abcam (Cambridge, UK), those against the subunits of respiratory enzymes were from Molecular Probes and the monoclonal antibody against β-actin was from Chemicon (Temecula, CA, USA). All the other chemicals were purchased in the highest purity available from Sigma-Aldrich Chemical Co. (St Louis, MO, USA).
Differentiation of 3T3-L1 preadipocytes to adipocytes
Two days after confluence (day 0), 3T3-L1 cells were grown in the differentiation medium composed of 10% DMEM, 0.25 μm dexamethasone, 0.5 mm 3-isobutyl-1-methylxanthine, 10 μg·mL–1 insulin, 100 units·mL–1 penicillin G and 100 μg·mL–1 streptomycin sulfate at 37 °C containing 10% CO2 for 3 days. After induction of differentiation, the adipocytes were cultured in 10% DMEM containing 1 μg·mL–1 insulin for the first 2 days and then in fresh 10% DMEM without insulin in the following 4 days. All the experiments were conducted by using adipocytes 9 days after differentiation.
Differentiation of C2C12 myoblasts to myotubes
To induce differentiation of C2C12 cells, the growth medium (DMEM with 10% fetal bovine serum) was changed to DMEM containing 2% horse serum (Gibco, Invitrogen, Grand Island, NY, USA) when cells had been grown to 80% confluence, and the differentiation medium (DMEM containing 2% horse serum) was changed continually every 2 days. After 7 days, mature myotubes were formed and the cells were used to measure glucose uptake.
Knockdown of mtTFA
The lentiviruses containing shRNA constructs of the mtTFA gene (sh-mtTFA) and luciferase control gene (sh-scramble) were obtained from the RNAi Core Facility at Academia Sinica, Taipei, Taiwan. To achieve the maximum efficiency of knockdown of mtTFA, the adipocytes infected with lentivirus for 24 h were selected by 0.8 μg·mL–1 puromycin for another 24 h. The targeting sequences of sh-scramble and sh-mtTFA in the pLKO.1 plasmid were 5′- CAAATCACAGAATCGTCGTAT-3′ and 5′-CCTCAGATTAAGTGCTGAGAT-3′, respectively.
Measurement of oxygen consumption rate
The oxygen consumption rate of adipocytes was measured with a 782 Oxygen Meter (Strathkelvin Instruments, North Lanarkshire, Scotland, UK). An aliquot of 330 μL assay buffer (125 mm sucrose, 65 mm KCl, 2 mm MgCl2, 20 mm phosphate buffer, pH 7.2) containing 106 cells was delivered into the closed chamber of the oxygen meter at 37 °C. After addition of 0.002% purified digitonin to permeabilize the plasma membrane, 10 mm each of glutamate and malate, and then 10 mm ADP, were subsequently added to initiate the substrate-supported respiration and state 3 respiration, respectively. Finally, 2.5 mm KCN was added to inhibit mitochondrial respiration for the measurement of the rate of non-mitochondrial oxygen consumption .
Measurement of intracellular ATP content
The intracellular content of ATP was measured with the Bioluminescent Somatic Cell Assay Kit (Sigma-Aldrich). An aliquot of 50 μL suspension of viable cells (containing 10 000–20 000 cells) was mixed with 150 μL Somatic Cell Releasing Reagent to disintegrate the cells and release the intracellular ATP. Half of the lysate was mixed with an equal volume of the ATP Assay Mix and the luminescence intensity was measured on a Victor2TM 1420 Multilabel Counter (Perkin-Elmer Life Sciences, Boston, MA, USA) .
Determination of intracellular hydrogen peroxide
The fluorescent dye DCFH2-DA was used to measure the intracellular content of H2O2 . The cells were incubated with 80 μm DCFH2-DA in the medium at 37 °C in the dark for 20 min. The relative fluorescence intensity of DCF in 10 000 cells per sample was determined on a flow cytometer (Model EPICS XL-MCL; Beckman-Coulter, Miami, FL, USA) with an excitation wavelength of 488 nm and emission wavelength of 535 nm.
Glucose uptake assay
Glucose uptake was measured according to the method described by Fong et al. . Adipocytes were grown in a serum-free medium for 2 h and the medium was then replaced by a KRH buffer (KRP buffer containing 20 mm HEPES, pH 7.2). The adipocytes were stimulated with (insulin-stimulated group) or without (basal group) addition of 100 nm insulin for 30 min before incubation with 0.2 mm [3H]-2-deoxyglucose ([3H]-2DG) for another 10 min. The reaction was terminated by the addition of 20 mm glucose in cold NaCl/Pi. Finally, the cells were lysed with 2% SDS and mixed with a liquid scintillation cocktail (Beckman-Coulter) to determine the radioactivity with a liquid scintillation counter (Perkin-Elmer Life Sciences).
Secretion of adiponectin by adipocytes
Adipocytes were cultured in DMEM containing 10 μg·mL–1 oligomycin A for 24 h or incubated with a suitable amount of shRNA of mtTFA for 48 h, and the medium was then changed to DMEM without serum. After further incubation for 12 h, the conditioned medium was collected and centrifuged to remove the cells. The concentration of adiponectin secreted by adipocytes was measured by using a mouse adiponectin ELISA kit (Millipore, Billerica, MA, USA) according to the manufacturer's instructions.
Statistical analysis was performed using the Microsoft® Office Excel 2003 statistical package. Experimental data are presented as means ± standard deviation of the results obtained from three independent experiments. The significance level was determined by Student's t test. A difference is considered to be statistically significant for P < 0.05.
This work was supported by research grants (NSC100-2320-B-010-024-MY3 and NSC97-2320-B-010-038-MY3) from the National Science Council, the Executive Yuan, Taiwan. We also acknowledge the support of an intramural research grant 982A01 from Mackay Medical College. The authors wish to express their appreciation of Professor Jim C. Fong for his kind gift of the 3T3-L1 preadipocytes used in this study.
Conflict of interest
The authors declare that they have no conflict of interest.