Uncoupling Protein-2 Controls Adiponectin Gene Expression in Adipose Tissue Through the Modulation of Reactive Oxygen Species Production

The care and use of mice were in accordance with the European Community Council Directive 86/609/EEC and approved by the Experimental Animal Ethics Committee, University of Barcelona. Seven- to 10-week-old UCP2-null male mice on the C57BL/6J genetic background (26) and control littermates were used. Fed mice were killed in the morning, at the beginning of the light cycle, by cervical dislocation, and inguinal adipose tissue and blood were collected. Serum was prepared, and adipose tissue was frozen in liquid nitrogen. Serum metabolites were quantified using a reflection photometry-based automatic system (Accutrend; Roche).

Medium, antibiotics, and serum were from GIBCO. The 3T3-L1 cells were grown in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% FCS and 5% penicillin/streptomycin (10,000 units/ml), at 37°C in 5% CO₂. Adipogenic differentiation of 3T3-L1 cells was achieved by the use of DMEM containing 10% FCS, 5 μg/ml insulin, and 125 μmol/l indomethacin, a method described to enhance endocrine capacities of 3T3-L1 cells (27). After 48 h, cells were washed with PBS, and insulin treatment was maintained for an additional 48 h to achieve differentiation. Oligomycin, antimycin A, parthenolide, AICAR (5-aminoimidazole-4-carboxamide-1-b-riboside) (Sigma), H89, and Trolox (Calbiochem) were dissolved in ethanol. The ROS-sensitive fluorescent probe 6-carboxy-2′,7′-dichlorodihydrofluorescein diacetate, diacetoxymethyl-ester (H₂-DCFDA) (Molecular Probes) was prepared in DMSO. The final concentration of ethanol or DMSO was kept <0.1%, and incubation with vehicle was performed as control.

AdCMV-UCP2 adenovirus driving human UCP2 expression (28) was provided by Dr. I.K. Lee (Keimyung University, Daegu, Korea). AdCMV-βGal expressing the Escherichia coli β-galactosidase gene was used as control. 3T3-L1 confluent cells were infected with adenovirus (either AdCMV-UCP2 or the control AdCMV-βGal) at a multiplicity of infection (MOI) of 50, for 2 h in the presence of polylysine at 0.5 μg/ml. This treatment led to an efficiency of transduction of ∼90%, as assessed in previous experiments analyzing fluorescence driven by AdCMV-GFP transduction in these same conditions. Cells were studied 3 days after transduction.

Adiponectin protein levels were determined in 3T3-L1 culture medium and in mice serum. After overnight serum deprivation, 3T3-L1 cells were incubated in fresh serum-free medium. At the end of the experiment, adiponectin release in cell culture medium was measured. Assays were performed using a Mouse/Rat Adiponectin ELISA Kit (B-Bridge International). Absorbance was detected at 450 nm using an automatic ELISA plate reader (Bio-Rad Benchmark Plus).

Nuclear extracts were prepared as reported elsewhere (29). The 3T3-L1 cells were washed twice with ice-cold PBS and scraped off the plates in buffer containing 10 mmol/l HEPES, pH 7.9, 10 mmol/l KCl, 0.1 mmol/l EDTA, 0.1 mmol/l EGTA, 1 mmol/l dithiothreitol, 0.5 mmol/l phenylmethylsulfonyl fluoride (PMSF), and protease inhibitors (5 mg/ml each of aprotinin, leupeptin, and pepstatin). Cells were allowed to swell on ice for 15 min; then 25 μl Nonidet P-40 (0.5%) was added, and the suspension was mixed for 10 s. The homogenate was centrifuged (16,000 × g, 4°C, 60 s), and the nuclear pellet was resuspended in 50 μl ice-cold buffer containing 20 mmol/l HEPES, pH 7.9, 0.4 mol/l NaCl, 1 mmol/l EDTA, 1 mmol/l EGTA, 1 mmol/l dithiothreitol, 1 mmol/l PMSF, and protease inhibitors. The nuclear extract was cleared (16,000 × g, 4°C, 5 min).

Mitochondrial proteins were isolated from 3T3-L1 cells as described previously (30). Cells were homogenized in a buffer containing 220 mmol/l mannitol, 70 mmol/l sucrose, 5 mmol/l MOPS, 2 mmol/l EGTA, and 0.2% BSA, pH 7.4. The homogenates were centrifuged for 10 min at 500 × g (4°C). The pellet was homogenized and centrifuged as above. Supernatants were mixed and then centrifuged at 10,000 × g (10 min, 4°C). The resulting pellet was washed twice in the same buffer and centrifuged at 5,000 × g (10 min, 4°C). Finally, the pellet was resuspended, and protein concentrations were determined by the Bradford assay (Bio-Rad).

Samples containing 40 μg cellular or mitochondrial protein or nuclear proteins after immunoprecipitation with C/EBPβ antibody (2-h incubation; SC-150; Santa Cruz) or 100 μg protein from adipose tissue were mixed with equal volumes of 2× SDS loading buffer, incubated at 90°C for 5 min, and electrophoresed on SDS/12% polyacrylamide gels. Proteins were transferred to Immobilon-P membranes (Millipore), and immunological detection was performed using rabbit or goat affinity-pure IgG against CHOP-10 or UCP2, respectively (SC-575 and SC-6526; Santa Cruz). Detection was achieved with horseradish peroxidase–coupled anti-rabbit and anti-goat (SC-2004 and SC-2354; Santa Cruz) antibodies and an enhanced chemiluminescence detection kit (Amersham). Mitochondrial extracts from HeLa cells previously transfected with a plasmid expression vector driving UCP2 cDNA were used as positive control for UCP2 detection. The sizes of proteins were estimated using protein molecular-mass standards (Bio-Rad).

RNA was prepared from frozen aliquots of adipose tissue or 3T3-L1 cells (RNeasy; Qiagen). Quantitative mRNA expression analysis was performed using TaqMan real-time RT-PCR (Applied Biosystems). The reverse transcriptase reaction was performed on 0.5 μg RNA using TaqMan standardized reagents, and the real-time PCR reaction was performed using TaqMan Universal PCR Master Mix and standardized gene expression primer probes (“Assay-on-Demand”): adiponectin (Mm00456425), glutathione peroxidase-1 (Mm00656767), superoxide dismutase-2 (Mm00449726), and PPARγ (Mm00440945). The amount of mRNA for the gene of interest in each sample was normalized to that of the housekeeping reference 18S rRNA (Hs99999901). Samples were run in duplicate on the ABI-PRISM 7700HT sequence detection system (Applied Biosystems).

The H₂-DCFDA probe was used to estimate intracellular ROS (31). After incubation with 10 mmol/l H₂-DCFDA for 30 min, 3T3-L1 cells were washed twice with ice-cold PBS and scraped on ice into water. After sonication, fluorescence was measured at 527 nm after excitation at 493 nm. Fluorescence data were normalized relative to protein concentrations in the cell lysates.

A double-stranded oligonucleotide containing the −117/−73 fragment of the mouse adiponectin promoter (5′-GCCCACTCATTGGCTATTGGCCTTGACTGGGTT GGCCAATGGTAA G-3′; Roche) was end-labeled using [α-³²P]dCTP and Klenow enzyme. The DNA probe (30,000 cpm) was added to 10 μg nuclear protein extract mixture and incubated for 20 min at room temperature in binding buffer (25 mmol/l HEPES, pH 7.6, 0.5 mmol/l dithiothreitol, 12.5 mmol/l ZnSO₄, 50 mmol/l KCl, 1 mg/ml BSA, 5% glycerol, 0.1% Nonidet P-40, and 2.5 μg poly[dI-dC]). Specific binding was assessed by competition tests; preincubating samples with a 100-fold molar excess of unlabeled doubled-stranded oligonucleotide. Samples were run on 5% nondenaturing polyacrylamide gels in 0.5× Tris-borate-EDTA at 130 V and 4°C for 60 min. Specificity of the C/EBPβ binding was further assessed by inducing a super-shifted band after incubating nuclear extracts with anti-C/EBPβ antibody (SC-150; Santa Cruz) for 30 min at room temperature.

A double-stranded oligonucleotide corresponding to the −117/−73 region of the mouse adiponectin promoter known to contain the C/EBPβ-response element (10) was inserted, using the HindIII and XhoI sites, upstream of the herpes simplex virus thymidine kinase basal promoter driving firefly luciferase expression (Tk109-Luc vector). This construct was transfected into confluent 3T3-L1 cells using Fugene reagent (Roche). PRL-TK (Promega), an expression vector of Renilla luciferase, was co-transfected. Luciferase activities were measured in a Turner-Designs luminometer (TD20/20) using the Dual-Luciferase Reporter kit (Promega). Luciferase activity elicited by the −117/−73 TK-Luc fragment construct was normalized for variation in transfection efficiency using Renilla luciferase. The CHOP-10 expression vector SH-GADD153 was provided by S. Ambrosio (University of Barcelona, Barcelona, Spain).

Data, shown as means ± SE, were analyzed with Student's t test, and differences were considered to be significant at a P value <0.05.

In a first attempt to determine whether changes in respiratory chain regulation elicited by mitochondrial uncoupling could be involved in the biological signals produced by adipose tissue, we measured the levels of adiponectin and metabolites in serum from control and UCP2-null mice. Table 1 shows that glucose and lactate levels were significantly lowered in UCP2-null mice, whereas triglyceride levels were not significantly altered. Overall, this pattern of circulating metabolites is in accordance with previous reports (20,21,32). Table 1 also shows that adiponectin levels in the serum of UCP2-null mice are significantly reduced with respect to control mice. Adiponectin mRNA levels were dramatically downregulated in adipose tissue from UCP2-null mice (a 65% reduction vs. controls) (Table 2). This suggested that the reduced serum adiponectin levels were caused by impaired adiponectin gene expression in adipose tissue due to the lack of UCP2. The reduction in adiponectin mRNA levels was not shared by other genes, and for instance, SOD2 mRNA levels were increased in UCP2-null mice (Table 2). Upregulation of the SOD2 gene suggests the occurrence of altered oxidative status in adipose tissues from UCP2-null mice. PPARγ mRNA levels were not significantly affected in UCP2-null mice.

To determine whether mitochondrial respiratory function and specifically UCP2 influence adiponectin production, adipogenic 3T3-L1 cells were used. Cells were transduced with an adenoviral vector driving UCP2 expression (AdCMV-UCP2) in conditions leading to an approximate twofold increase (2.16 ± 0.6–fold) in UCP2 protein levels in mitochondrial protein extracts, as assessed by immunoblot assays (Fig. 1A). The increase in UCP2 levels led to a rise in adiponectin mRNA levels in 3T3-L1 cells with respect to controls (Fig. 1B) and a significant increase in the release of adiponectin protein to the culture medium (1.44 ± 0.16–fold induction, P < 0.05). To ascertain the biochemical basis of the UCP2 effects, parallel assays were performed exposing 3T3-L1 cells to oligomycin or antimycin A, inhibitors of ATP synthase and respiratory chain complex III, respectively. These drugs are known to reduce ATP synthesis, as UCP2 does in 3T3-L1 adipocytes (33), but in this case, this is associated with an inhibition of the respiratory chain activity, in contrast with the action of mitochondrial uncoupling that increases respiration. Oligomycin and antimycin A produced effects opposite those of UCP2, leading to a significant reduction in adiponectin mRNA levels (Fig. 1B). Adiponectin protein release was downregulated in parallel (0.62 ± 0.05–fold decrease in oligomycin-treated cells, P < 0.05). These results indicated that a blockage of mitochondrial respiratory function represses adiponectin gene expression, whereas the mitochondrial modifications elicited by UCP2 enhance this expression. To ascertain the potential mechanisms underlying the observed effects, 3T3-L1 cells were incubated in the presence of inhibitors or activators of different signaling pathways potentially related to mitochondrial activity. Oligomycin or antimycin A treatment are expected to increase AMP levels, and therefore it could be hypothesized that AMP kinase activation could be involved in the regulation of adiponectin gene expression. However, Fig. 1C shows that AICAR, a drug that mimics AMP kinase, did not affect adiponectin mRNA levels, thus indicating that disturbances in the ATP/ADP/AMP relationship were not the main determinants of changes in adiponectin gene expression in response to altered mitochondrial respiratory function. Furthermore, when protein kinase A was inhibited by H89, no significant changes in the oligomycin effects on adiponectin mRNA levels were observed (data not shown), suggesting that changes in cAMP levels are not involved in the regulation of adiponectin mRNA expression. Treatment of 3T3-L1 cells with Trolox, an ROS scavenging drug, did not modify basal levels of adiponectin mRNA but alleviated oligomycin-dependent and antimycin A-dependent reduction in adiponectin mRNA expression (Fig. 1C). This suggests that a rise in ROS elicited by oligomycin treatment could mediate the downregulation of adiponectin gene expression. This would explain the upregulation of adiponectin mRNA levels due to UCP2, considering its known capacity to reduce mitochondrial ROS levels. Thus, a major role for mitochondrial ROS production in the control of adiponectin gene expression would explain the opposite effects of UCP2 and oligomycin.

To check directly whether UCP2, oligomycin, or antimycin A treatment modulates ROS production in accordance with the above hypothesis, we directly measured ROS levels in 3T3-L1 cells infected or not with AdCMV-UCP2 and treated or not with oligomycin or antimycin A. Figure 2 shows that in our conditions, UCP2 transduction induced a significant reduction in ROS levels in 3T3-L1 cells, whereas oligomycin and antimycin A caused a significant increase in ROS levels. The treatment with Trolox described above did not modify basal ROS levels but suppressed the rise in ROS caused by oligomycin or antimycin A (Fig. 2). These results are consistent with the proposal of ROS production as a main signal of mitochondrial origin controlling adiponectin gene expression and with the action of UCP2 upregulating adiponectin gene expression via a reduction in mitochondrial ROS production. A potential mechanism of adiponectin gene regulation could be the activation of the nuclear factor-κB (NF-κB) system, because the absence of UCP2 has been reported to increase NF-κB activation and cytokine release in macrophages (34). However, treatment of 3T3-L1 cells with parthenolide, which specifically inhibits NF-κB activation by preventing inhibitor of κB degradation (35), did not prevent either the oligomycin-dependent or the antimycin-dependent repression of adiponectin gene expression (Fig. 1C), thus indicating that mitochondrial activity impairment does not mediate adiponectin gene regulation via NF-κB activation

CHOP-10 transcription factor (also known as GADD153) is an inactive analog of C/EBP proteins, known to be induced by ROS and mediating some of the ROS-dependent effects on gene transcription (36–38). Because it has been demonstrated that the adiponectin gene promoter is regulated by C/EBPβ (10), we hypothesized that CHOP-10 activation could be a potential mechanism providing the signaling pathway linking ROS to adiponectin gene expression. To verify this possibility, we first determined CHOP-10 protein levels in nuclear extracts of 3T3-L1 after UCP2 transduction or oligomycin treatment. Figure 3A shows that CHOP-10 protein levels were decreased significantly in nuclear extracts from cells infected with AdCMV-UCP2 with respect to controls, whereas CHOP-10 levels were significantly increased with oligomycin treatment. This was fully consistent with variations in ROS production described above.

We next determined how the changes in CHOP-10 elicited as a consequence of mitochondrial activity influence the amounts of CHOP-10–C/EBPβ inactive heterodimers. Nuclear extracts from 3T3-L1 cells incubated or not with oligomycin were immunoprecipitated using an anti-C/EBPβ antibody, and the amount of CHOP-10 protein in the precipitate was immunodetected using an anti-CHOP-10 antibody. Results indicated that oligomycin treatment lead to a significant increase of CHOP-10–C/EBPβ inactive heterodimers in 3T3-L1 cells relative to untreated controls (Fig. 3C). This confirms that CHOP-10 protein variations result in parallel changes in the amounts of CHOP-10–C/EBPβ inactive heterodimers.

At this point, and to further establish whether the downregulation in adiponectin gene expression observed in UCP2-null mice may be associated with changes in CHOP-10 expression, immunoblot analysis was performed using adipose tissue protein extracts from wild-type and UCP2-null mice. The results indicated significantly higher levels of CHOP-10 protein in adipose tissue from UCP2-null mice (Fig. 4).

To establish whether changes in mitochondrial activity and formation of CHOP-10/CEBPβ heterodimers have consequences on C/EBPβ binding to the adiponectin promoter, electrophoretic mobility shift assays were performed using as probe the −117/−73 fragment of the adiponectin promoter, containing the C/EBPβ binding site (10), and nuclear protein extracts from 3T3-L1 cells after UCP2 transduction or oligomycin treatment. Results indicated that C/EBPβ binding to the −117/−73 region was significantly increased (1.33 ± 0.10–fold, P < 0.05) in extracts from UCP2-transduced cells, whereas it was reduced (0.53 ± 0.15–fold, P < 0.05) using extracts from oligomycin-treated cells (Fig. 5).

Finally, to establish the involvement of CHOP-10 in the regulation of the adiponectin promoter in response to ROS production and UCP2 activity, transient transfection assays were performed using a luciferase reporter construct driven by the −117/−73 fragment of mouse adiponectin promoter. Transduction of 3T3-L1 cells with UCP2 led to a dramatic increase in luciferase activity driven by the −117/−73 adiponectin gene element, whereas it did not affect the promoter activity of the control construct (Fig. 6). Oligomycin treatment had opposite effects and led to a significant reduction in −117/−73 TK-Luc activity (0.38 ± 0.11 in oligomycin-treated cells, P < 0.05) but not in the parental TK-Luc construct. The effects of overexpressing CHOP-10 via co-transfection with a CHOP-10 expression vector were analyzed. Results showed that CHOP-10 co-transfection suppressed the capacity of UCP2 to upregulate the −117/−73 adiponectin TK-Luc construct (Fig. 6). These results indicated that UCP2 effects leading to adiponectin gene upregulation are mediated via the −117/−73 element in the adiponectin promoter region and that CHOP-10 downregulation is involved in this effect.

The influence of adipose tissue upon systemic metabolism, including the control of glycemia and insulin sensitivity, has been recognized in recent years. Adiponectin is an insulin-sensitizing molecule released by adipose tissue, and the identification of the mechanisms of control of adiponectin gene expression and release are of utmost relevance for type 2 diabetes control and prospective therapies. In the present study, we established that the action of mitochondrial activity, and specifically of UCP2, on ROS production has a pivotal influence on adiponectin gene expression. Changes in mitochondrial activity and the action of UCP2 occur via changes in ROS levels and involve changes in the levels of CHOP-10, which acts as a negative regulator of adiponectin gene transcription by interfering with the C/EBP-binding site in the promoter. These findings are consistent with reports indicating that the intracellular status of adipocytes in relation to ROS influences adiponectin gene expression (39) and with the reported capacity of CHOP-10 to convey ROS signaling into overall regulation of gene expression in adipocytes (38).

Accumulating pieces of evidence have underlined the importance of mitochondrial activity in muscle and also in adipose tissue for insulin sensitivity. For instance, adipose tissue is the main target of thiazolidinediones, and the insulin-sensitizing effects of these drugs involve the induction of gene expression of mitochondrial proteins in the tissue (40), including UCP2 (41,42), as well as the induction of adiponectin gene expression (9). Although thiazolidinediones may favor adiponectin gene expression via direct activation of PPARγ-responsive elements in the adiponectin gene, the parallel induction of adiponectin and UCP2 gene expression in adipose tissue after thiazolidinedione treatment is consistent with a role of UCP2 in inducing adiponectin gene expression, as shown here. On the other hand, several experimental or clinical insulin-resistance states associated with low adiponectin levels and reduced UCP2 expression in adipose tissue have been reported. For instance, lipodystrophy in HIV-infected, antiretroviral-treated patients is associated with systemic insulin resistance, mitochondrial impairment in adipose tissue, including UCP2 downregulation, and reduced adiponectin in plasma and adipose tissue (43,44). Abnormal downregulation of adiponectin and UCP2 gene expression has been reported in adipose tissue from first-degree relatives of type 2 diabetic patients (45,46), whereas insulin sensitization in 11β-hydroxysteroid dehydrogenase type 1–deficient mice or in Crebbp heterozygous mice is associated with parallel induction of adiponectin and UCP2 mRNA levels in adipose tissue (47,48).

The involvement of UCP2 in the control of insulin/glucose homeostasis was first recognized from studies in pancreatic β-cells from UCP2-null mice, and UCP2 was identified as a negative regulator of insulin release (21), although there have been contradictory reports based on distinct experimental approaches (49). Our present findings indicate that UCP2 levels in adipose tissue are positive inducers of adiponectin gene expression and therefore may favor insulin sensitivity. This may help to explain the mild reduction in glycemia that, despite hyperinsulinemia (21), occurs in UCP2-null mice. On the other hand, insulin has been reported to repress adiponectin gene expression (8), and therefore high insulin levels may contribute to the adiponectin downregulation in UCP2-null mice. Further research will be required to establish whether the regulatory pathways of adiponectin gene regulation described here, involving ROS-dependent pathways under the control of UCP2, are related to the intracellular mechanisms of insulin action on the adiponectin gene.

In summary, we report here a novel pathway of regulation of adiponectin gene expression mediated by mitochondrial activity and specifically by mitochondrial UCP2 levels. This finding supports the notion that mitochondrial activity and ROS production in adipose tissue are important events influencing overall insulin sensitivity and systemic metabolism, via the control of adiponectin expression and release.

E.C. is the recipient of research fellowships from Nestle Foundation and Generalitat de Catalunya. This study has received support from Grants SAF2002-03648 and SAF2005-01722 from the Ministerio de Educacion y Ciencia, Spain, FISPI052336 from Ministerio de Sanidad y Consumo, Spain, and European Community 6th Framework Programme funding N LSHM-CT-2003-503041.

We thank Dr. I.K. Lee for the UCP2 adenoviral vector and S. Ambrosio and D. Haro (University of Barcelona) for SH-GADD153 and Tk109-Luc plasmid vectors, respectively.