Adipose Mitochondrial Biogenesis Is Suppressed in db/db and High-Fat Diet–Fed Mice and Improved by Rosiglitazone

Male, 4-week-old C57BL/6J and C57BL/Ks db/db and db/+ mice were obtained from The Jackson Laboratory (Bar Harbor, ME) and housed in a temperature- and humidity-controlled environment with a 12-h light/dark cycle. db/db and db/+ mice were fed a standard chow (Purina 5001; TestDiet, Richmond, IN) throughout the experiments. For the HFD model, 5-week-old C57BL/6J mice were placed on either a high-fat diet (D12331, 58% kcal from fat; Research Diets, New Brunswick, NJ) or a standard-fat diet (d12329, 11% kcal), with coconut oil as the main fat source, matching that in the high-fat diet (15). All animals were allowed free access to food and water. Care and use of the animals were in accordance with protocols approved by the GlaxoSmithKline Institutional Animal Care and Use Committee.

RSG maleate (BRL49653C) was obtained from GlaxoSmithKline (Harlow, U.K.). For the HFD model, RSG treatment began after 8 weeks on either the high-fat or standard-fat diet. For the db/db model, treatment began in db/db and db/+ mice at 8 weeks of age. In both models, mice received RSG (1, 3, or 10 mg/kg) or vehicle (0.5% methylcellulose) by oral gavage twice daily for 14 days. Blood samples were taken on day 0 and 14 after 6-h fasting. Ten mice were used in each treatment group in each model.

Plasma glucose and insulin was measured using an automated clinical chemistry analyzer (olympus au400; Olympus America, Melville, NY) and an enzyme-linked immunosorbent assay kit (ALPCO Diagnostics, Windham, NH), respectively. A1C levels were measured from whole blood using the Variant II (Bio-Rad; Hercules, CA).

All adipose tissues used in the microarray experiment were inguinal subcutaneous adipose tissue. RNA isolation, DNase I treatment, reverse transcription and cRNA synthesis, labeling, and hybridization with Genechips mouse genome 430a array (Affymetrix Microarray Suite; Affymetrix, Santa Clara, CA) were conducted following the standard affymetrix protocol (online appendix [available at http://dx.doi.org/10.2337/db06-1135]). Total RNA (5 μg) from single animals was used for each chip. The RNA samples were randomized before cRNA synthesis and rerandomized before hybridization to avoid batch effects. Array intensity data were captured by the Genechip Computer Operating System using the algorithm (MAS 5.0; Affymetrix).

Inguinal adipose tissue fixed in 10% neutral buffered formalin was paraffin embedded and sectioned (5 μm in thickness). After deparaffinization and antigen retrieval (16), the sections were stained with goat anti–cytochrome c (1:500 dilution; Santa Cruz), goat anti–mitochondrial heat shock protein 60 (HSP60) (1:200; Santa Cruz), or rabbit anti–uncoupling protein (UCP)-1 (1:2,000; Abcam, Cambridge, U.K) polyclonal antibodies (online appendix).

Quantification of cytochrome c–positive area was performed by two blinded pathologists using eCognition (Definiens, München, Germany), following manufacturer's instructions. Details of image acquisition and morphometric analysis were provided in the online appendix. Cytochrome c–positive area was calculated as cytochrome c–positive area divided by the total area of examined images and expressed as μm²/mm². UCP-1–positive multilocular cells on acquired images were identified and counted by blinded pathologists and expressed as number of cells per 0.705 mm².

Transcriptomics results from the microarray analysis were confirmed using representative genes by quantitative RT-PCR (online appendix). Relative amounts of nuclear and mitochondrial DNA were determined by quantitative real-time PCR using glyceraldehydes-3-phosphate dehydrogenase and cytochrome b, respectively (17) (online appendix). The ratio of mitochondrial DNA to nuclear DNA reflects the tissue concentration of mitochondria per cell.

Citrate synthase activity was measured similarly to Bogacka et al. (3) using a Sigma kit (online appendix). Protein concentrations in the lysate were determined using a micro biochronic acid protein assay kit (Pierce, Rockford, IL). The citrate synthase activity was expressed as μmol · min⁻¹ · g protein⁻¹.

Plasma measurements are presented as means ± SE and were analyzed using two-way repeated-measures ANOVA with a post hoc analysis using the Student Newman Keuls multiple comparison test. P < 0.05 was considered statistically significant.

Microarray analysis was based on 10 chips per treatment group (1 chip for each mouse), five groups in each mouse model. Probe sets with intensity values ≥22.1, which represented the 95 percentile of the built-in Affymetrix negative control probe sets for all the chips, were considered present. Probe set intensity values were log10 transformed and analyzed using both ANOVA and the false discovery rate (18). P < 0.01 from ANOVA was used to select genes with a significant change because, in our analyses, this criterion was found to be more stringent than that using an false discovery rate q <0.1 (10% false discovery rate), an acceptable false-positive rate in microarray literature (e.g., 19). In addition, a 1.2-fold cutoff for median fold changes, previously used by others (e.g., 20) and consistently generating reproducible results in our microarray experiments (M.K.H., Y.Q., J.X.R., personal communications), was used for further gene selection.

Cytochrome c–positive areas, UCP-1–positive cell numbers, relative mitochondrial DNA content, and citrate synthase activity are presented as means ± SE. Statistical analyses of the differences between vehicle- and rosiglitazone-treated db/db mice and between vehicle-treated db/+ and vehicle-treated db/db mice were conducted using one-way ANOVA followed by Dunnett test, using vehicle-treated db/db mice as the control, and Student's t test, respectively. P < 0.05 was considered statistically significant.

db/db mice had significantly higher glucose, insulin, and A1C levels than db/+ mice (Table 1). Fourteen days of RSG treatment significantly decreased glucose and insulin levels at all three doses. A1C levels were reduced with significant changes occurring with the 3 and 10 mg/kg of RSG. RSG significantly increased body weight and inguinal adipose weight (data not shown).

HFD mice had significantly higher glucose, insulin, and A1C levels than standard-fat diet–fed mice (SFD) (Table 2), but the levels were generally lower than those in db/db mice. In HFD mice, only the highest dose of RSG (10 mg/kg) significantly decreased glucose after 14 days of treatment, but all RSG doses significantly decreased insulin levels. There were no significant reductions in A1C with RSG treatment in the HFD mice. Body weight and inguinal adipose weight were significantly increased in mice treated with 3 and 10 mg/kg RSG (data not shown).

Mitochondria host enzymes necessary for the tricarboxylic acid (TCA) cycle, fatty acid oxidation (FAO), and oxidative phosphorylation (OXPHOS) pathways, all of which contribute to ATP production. As listed in online appendix Table 1, of 22,626 probe sets (each identifying one gene transcript) on Affymetrix mouse chip 430A, 248 represent genes encoding proteins localized to mitochondria and related to OXPHOS/TCA/FAO, based on Affymetrix chip annotation and confirmed by Entrez Gene (http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?CMD = search&DB = gene).

Of 248 probe sets, 241 (97%) were called “present” in the inguinal adipose tissue, composed of BAT and white adipose tissue (WAT) (21). Figure 1 shows that the expression of the majority of the OXPHOS/TCA/FAO genes in the tissue was suppressed in db/db and HFD mice compared with those in db/+ and SFD mice, respectively. In a dose-dependent manner, RSG induced the expression of most of the OXPHOS/TCA/FAO genes in both db/db and HFD mice. At 10 mg/kg, RSG increased the expression of the majority of the genes to a level equivalent to or higher than that in db/+ and SFD mice. This suggests that RSG is capable of correcting the compromised adipose mitochondrial gene expression in db/db and HFD mice.

Among the regulated genes were those encoding key subunits of enzymes regulating the control points (22) in the TCA cycle, namely, pyruvate dehydrogenase, isocitrate dehydrogenase 3, and α-glutarate dehydrogenase, and those encoding electron transporters in OXPHOS, including cytochrome c and cytochrome c1. The rate of FAO is controlled by carnitinepalmitoyl transferase I (CPT-1) (22). CPT-1b, the major CPT-1 isoform in BAT (23), was significantly decreased in HFD mice (to 29% of that in SFD) and induced by RSG at a 10-mg/kg dose by 6.2- and 4.9-fold in db/db and HFD mice, respectively. In addition to enzymes/proteins directly involved in OXPHOS/TCA/FAO, those responsible for the transport of metabolites and intermediates for ATP production and FAO across mitochondrial membranes, including voltage-dependent anion channels, ATP/ADP shuttles, and carnitine/acyl carnitine shuttles, were suppressed in the adipose tissue from db/db and HFD mice and were dose-dependently induced by RSG (Fig. 1).

To date, three UCP isoforms have been identified in mammalian cells (24): UCP-1, UCP-2, and UCP-3, which share >57% homology (25). In mice, UCP-1 is specifically expressed in mitochondria-rich BAT and plays a key role in nonshivering thermogenesis (25,26). Ectopic expression of UCP-1 in WAT mitigates obesity induced by genetic or dietary factors (25,26). UCP-3 is predominantly expressed in skeletal muscle, heart, and BAT and has been shown to facilitate FAO (27), whereas UCP-2 is localized to the pancreas, immune system, WAT, and the brain and has been linked to insulin secretion and modulation of inflammatory response (24,25).

UCP-1 was detected in the inguinal adipose tissue of db/+ mice and SFD mice, reaffirming the presence of BAT in the tissue (21). Interestingly, UCP-1 expression in the tissue from SFD mice was ∼10-fold higher than that in db/+ mice (online appendix Table 1), indicating that the tissue from SFD mice had a greater proportion of the BAT-like component. UCP-1 expression in the tissue was lower in db/db than in db/+ mice and was even more dramatically decreased in HFD compared with SFD mice (Table 3). RSG treatment remarkably increased UCP-1 expression (by 37- and 34-fold, respectively) in both db/db and HFD mice.

Interestingly, the expression of cell death–inducing DFFA-like effector a (CIDEA), a modulator of UCP-1 and another marker of BAT-like tissue (5,28,29), was decreased in db/db and HFD mice compared with db/+ and SFD mice, respectively, and was increased by RSG in both models (Table 3). UCP-3 levels also decreased in db/db and HFD compared with db/+ and SFD mice, respectively, and were increased by RSG (Table 3). UCP-2 levels were not affected by diabetes/obesity or RSG treatment.

To investigate whether the observed changes were associated with changed mitochondrial mass, the regulation of mitochondrial structural proteins not directly participating in metabolic activities was examined. These proteins include mitofilin, localized to mitochondrial inner membrane and regulating morphology of mitochondrial cristae (30); optic atrophy 1, anchored to mitochondrial inner membrane and localized to the intermembrane space and implicated in maintenance of mitochondrial genome and network (31); surfeit 1, localized to the inner membrane and involved in assembly of cytochrome c oxidase complex (32); and BCS1L, a human homologue of yeast BCS1 localized to the inner membrane and involved in the assembly of complex III of mitochondrial respiratory chain (33). Figure 2A shows that the transcription of these proteins was induced by RSG in db/db mice and so was that of mitofilin and optic atrophy 1 in HFD mice.

To further investigate the roles of diabetes/obesity and RSG on mitochondrial mass/biogenesis, transcription of genes directly participating in mitochondrial replication was examined. Figure 2B shows that 28 and 26% of transcripts representing mitochondrial ribosomal proteins were downregulated in db/db and HFD mice, respectively, whereas the rest of the mitochondrial ribosomal protein transcripts, with a few exceptions, were unchanged. Similarly, 27% transcripts encoding translocases that transport cytosolic proteins across mitochondria outer and inner membrane (translocases of outer mitochondrial membranes and translocases of inner mitochondrial membranes, respectively) were suppressed in both db/db and HFD mice. Furthermore, 42% transcripts for mitochondrial heat-shock proteins (HSPs)/chaperones, responsible for proper protein folding and transport, were downregulated, whereas the rest were unchanged, in db/db mice.

At 10 mg/kg, RSG treatment induced the transcription of 76 and 60% of mitochondrial ribosomal proteins and 80 and 70% of translocases of outer mitochondrial membranes and translocases of inner mitochondrial membranes, in db/db and HFD mice, respectively. The RSG effect was dose dependent, and the dose dependency was more evident in db/db than in HFD mice. In addition, RSG (10 mg/kg) induced in db/db and HFD mice 71 and 58% of the transcripts, respectively, for mitochondrial HSPs, including mitochondrial HSP70 (Hspa9a) and HSP60 (Hspd1), whereas there was no overall effect on cytosolic HSPs (such as Hspa4 and Tcp1, which belong to the cytosolic HSP70 and HSP60 family, respectively) or endoplasmic reticulum chaperones (e.g., calnexin, calreticulin, and GRP94), indicating that the induction by RSG was selective for mitochondrial proteins.

Mitochondrial biogenesis involves the integration of multiple transcriptional pathways controlling both nuclear and mitochondrial gene expression. PPARγ coactivators (PGC)-1α and -1β have been implicated as key regulators of mitochondrial biogenesis in adipose tissue (3,6,34–37) and muscle (11,36–40). Table 3 shows that PGC-1β transcription was suppressed in the inguinal fat in both db/db and HFD mice and was induced by RSG in both models. PGC-1α transcription in C57BL/6 mice was not influenced by diet (i.e., there was no difference between HFD and SDF mice); however, its transcription was suppressed in db/db compared with db/+ mice. RSG elicited expression of PGC-1α in HFD (P < 0.01) and db/db (P = 0.06) mice.

The expression of nuclear receptors ERRα and PPARα was suppressed in db/db and HFD mice and induced by RSG. Receptor interacting protein (RIP) 140, a recently identified transcription corepressor negatively regulating mitochondrial biogenesis and metabolic functions in adipocytes (6,28), was induced in HFD but not in db/db mice. Surprisingly, RIP140 expression was induced by RSG in db/db mice and not affected in HFD mice. This indicates that RSG-induced mitochondrial biogenesis might not involve suppression of RIP140.

Increased mitochondrial metabolic activity is associated with an increased chance of leaking O₂⁻, leading to formation of other reactive oxygen species, followed by oxidative stress, cell injury, and eventually cell death. Figure 3 shows the transcription of antioxidant enzymes catalase, superoxide dismutase 1, superoxide dismutase 2, glutathione peroxidase 1, and methionine sulfoxide reductase A was generally increased by RSG. Meanwhile, RSG did not affect the transcription of mitochondrial proteins favoring apoptosis, such as Bax and Bad, or cytosolic regulators of apoptosis, such as caspases, in db/db and HFD mice.

The transcription of 18 genes representing control point enzymes and structural proteins or those involved in mitochondrial replication or antioxidant enzymes were examined by QRT-PCR using inguinal adipose RNA samples from the db/db model. The results, generally consistent with the microarray data (online appendix Table 2A), supported the notion that mitochondrial transcription was downregulated in diabetes/obesity and induced by RSG. The microarray findings were also confirmed at the protein level by immunohistochemistry and histomorphometric analysis of a few representative mitochondrial markers (i.e., cytochrome C, HSP60, and UCP-1) (see below).

To test whether diabetes/obesity and RSG also affected mitochondrial gene expression in epididymal adipose tissue, predominantly WAT (21), the transcription of the 18 representative genes was examined by QRT-PCR. As expected, CIDEA was not detected, and UCP-1 expression was very low in epididymal adipose tissue. UCP-1 was surprisingly induced under the diabetes/obesity condition but was dramatically induced with RSG treatment. The regulation of the other genes was generally consistent with that in inguinal adipose tissue (online appendix Table 2B).

Staining of inguinal adipose tissue showed that adipocyte size in db/db mice was generally larger than that in db/+ mice (Fig. 4A) and decreased with RSG treatment (41). Cytochrome c, a mitochondrial marker, was significantly decreased in the db/db mice. RSG increased cytochrome c staining in virtually all adipocytes, indicating the conversion of mitochondria-poor to mitochondria-rich adipocytes. There was also increased presence of smaller, multilocular cells enriched with cytochrome c (Fig. 4A, arrows). The increase of total cytochrome c staining was RSG dose dependent (Fig. 4B), consistent with the microarray results. Similar observations were made with immunostaining of mitochondrial HSP60, another mitochondrial marker (Fig. 4C). UCP-1 staining was limited to the multilocular cells, confirming their brown adipocyte nature (Fig. 4D). RSG dose-dependently increased brown adipocyte numbers in inguinal adipose tissue (Fig. 4F).

Mitochondrial DNA content in inguinal adipose tissue from db/db mice was slightly decreased compared with db/+ mice, but this did not reach statistical significance. However, mitochondrial DNA content in the tissue was increased by RSG dose dependently and, at 3- and 10-mg/kg doses, was significantly greater than vehicle-treated mice (Fig. 5A).

Citrate synthase, localized exclusively to the mitochondrial matrix, catalyzes the conversion of acetyl-CoA to citrate, the initial step in the TCA cycle. Citrate synthase activity has been used as a reference to determine changed mitochondrial function associated with mitochondrial biogenesis in adipose (3) and muscle (42). The activity in inguinal adipose tissue from db/db mice was significantly lower than that from db/+ mice (Fig. 5B). RSG dose-dependently increased the activity, reaching a plateau at 3 and 10 mg/kg.

The effects of diabetes/obesity and insulin sensitizers, such as RSG, on adipose mitochondrial biogenesis are only beginning to be understood and have not been systematically investigated in either db/db or HFD mice. The current study comprehensively evaluated mitochondrial nuclear DNA–encoded proteins involved in important metabolic functional pathways, including the control-point enzymes of each individual pathway. In addition, the regulation of transcription factors, mitochondrial structural proteins, and the machinery for mitochondrial replication were evaluated. In both mouse models, adipose mitochondrial gene transcription was overwhelmingly decreased in the disease states, and RSG dose-dependently increased the expression of many of the mitochondrial genes. The current results are consistent with recent observations in several in vivo models (3,8,9). The results on gene expression were extended by mitochondrial staining, quantification of mitochondrial DNA, and citrate synthase activity and further support a link between adipose mitochondria and the pathogenesis of diabetes/obesity.

The exact mechanisms by which adipose mitochondrial dysfunction and RSG-induced mitochondrial biogenesis contribute to whole-body energy homeostasis are unclear. Impaired mitochondria may lead to a lack of ATP (9), thereby attenuating triglyceride synthesis from fatty acids and carbohydrates in adipocytes, leading to a reduced adipose capacity for energy storage. Indeed, the insulin-sensitizing action of PPARγ agonists in muscle and liver is believed to be due to promoting triglyceride synthesis in adipose tissue and, thus, partitioning of circulating lipids away from muscle and the liver (1). A decrease in ATP may also impair the synthesis and secretion of adipokines that serve as endocrine signals. In fact, low serum levels of adiponectin, the most abundant secretory protein from adipocytes (43) and an adipokine that improves insulin sensitivity in other organs including the liver and muscle (44), is associated with diabetes/obesity in vivo. PPARγ activation increased circulating adiponectin levels in others’ (rev. in 1,44) and the current study (data not shown).

Adipose mitochondria also regulate whole-body energy metabolism via dissipating energy and generating heat in BAT. The inguinal adipose tissue consists of both BAT and WAT (21). The expression of BAT markers UCP-1, Cidea, and CPT-1b, were coordinately suppressed in the disease states and coordinately induced by RSG treatment in both models (Table 3). This indicates that the plasticity of the adipose tissue (diabetes/obesity induces the conversion of adipose from a semi-BAT phenotype to a more classical WAT phenotype) and RSG reverses the process by promoting the increase of mitochondria in pre-existing WAT (Fig. 4A; online appendix Table 2B) and the expansion of the BAT component (Fig. 4D and E). This supports the long-standing interest (e.g., 5,26) in converting WAT to BAT and transforming adipose tissue into fat-oxidizing machinery as a therapeutic approach for obesity and/or diabetes.

In the current study, RSG significantly increased mitochondrial activity. Inherent to increased mitochondrial activity is the potential of increased reactive oxygen species generation. Interestingly, RSG induced the transcription of antioxidant enzymes, responsible for removal of excess reactive oxygen species. This effect may prevent potential cell injury resulting from RSG-induced mitochondrial metabolic activity. The dual RSG actions on mitochondria and antioxidant enzymes could work through the upregulation of one or both of the PGC-1 proteins (Table 3) that have been shown to regulate both mitochondrial biogenesis and the antioxidant machinery (13,45). A recent study (46) also showed upregulation of antioxidant enzymes per se may be beneficial for improving insulin resistance.

In addition to genes involved in mitochondrial biology, a number of inflammation genes were coordinately regulated, but in an opposite direction to the mitochondrial genes, by diabetes/obesity and RSG treatment (data not shown), which is consistent with previous reports (47–49). Therefore, anti-inflammatory properties of PPARγ agonists represent yet another potential mechanism that may be involved in their efficacy in type 2 diabetes. The present study cannot discriminate whether mitochondrial effect is responsible for inflammation changes or vice versa. In addition, the study cannot rule out that RSG-induced improvements in mitochondrial function are an indirect consequence of improved glycemia; however, separate studies have indicated metformin, which decreased glucose levels in db/db mice, did not cause similar changes in adipose mitochondrial gene expression (online appendix Table 3). Furthermore, a similar effect on mitochondrial genes was observed with RSG in murine 3T3 L1, C3H10T1/2, and human primary adipocyte cultures that were free of inflammatory cell types and exposed to a constant concentration of glucose in the medium (J.X.R., Y.Q., unpublished data) (34,35). Taken together, this suggests that RSG is likely to have direct actions on adipose mitochondrial function.

In conclusion, in this study transcriptional profiling combined with phenotypic, molecular, and functional measurements indicates a deficiency in adipose mitochondrial function under diabetes/obesity conditions, and RSG treatment improves mitochondrial function. The results suggest a new area for the investigation of RSG action, that of promoting mitochondrial activity and health in metabolic tissues. These findings may bring new insights on the mechanisms of RSG action on diabetes/obesity treatment and new treatment approaches targeting adipose mitochondria.

The authors are thankful to Katherine Michelle Waters for providing excellent data from QRT-PCR, validating the microarray results; Matthew J. McVey for excellent assistance with the animal procedures; Drs. Randall Smith, Amit Bhattacharyya, and Dana Vanderwall for their excellent assistance in data analysis; Drs. Kirby Gottschalk and Richard Buckholz for kindly reviewing the manuscript and excellent comments; and Drs. Guanghui Hu, Qing Xie, Aixue Liu, Gregory Pahel, and Suzanne Neill for helpful discussions.