Cytosolic Aspartate Aminotransferase, a New Partner in Adipocyte Glyceroneogenesis and an Atypical Target of Thiazolidinedione*

We show that cytosolic aspartate aminotransferase (cAspAT) is involved in adipocyte glyceroneogenesis, a regulated pathway that controls fatty acid homeostasis by promoting glycerol 3-phosphate formation for fatty acid re-esterification during fasting. cAspAT activity, as well as the incorporation of [14C]aspartate into the neutral lipid fraction of 3T3-F442A adipocytes was stimulated by the thiazolidinedione (TZD) rosiglitazone. Conversely, the ratio of fatty acid to glycerol released into the medium decreased. Regulation of cAspAT gene expression was specific to differentiated adipocytes and did not require any peroxisome proliferator-activated receptor γ (PPARγ)/retinoid X receptor-α direct binding. Nevertheless, PPARγ is indirectly necessary for both cAspAT basal expression and TZD responsiveness because they are, respectively, diminished and abolished by ectopic overexpression of a dominant negative PPARγ. The cAspAT TZD-responsive site was restricted to a single AGGACA hexanucleotide located at -381 to -376 bp whose mutation impaired the specific RORα binding. RORα ectopic expression activated the cAspAT gene transcription in absence of rosiglitazone, and its protein amount in nuclear extracts is 1.8-fold increased by rosiglitazone treatment of adipocytes. Finally, the amounts of RORα and cAspAT mRNAs were similarly increased by TZD treatment of human adipose tissue explants, confirming coordinated regulation. Our data identify cAspAT as a new member of glyceroneogenesis, transcriptionally regulated by TZD via the control of RORα expression by PPARγ in adipocytes.

Glyceroneogenesis (a largely unexplored pathway) is an abbreviated version of gluconeogenesis that produces glycerol-3-P from precursors other than glycerol or glucose in adipose tissue when glucose utilization is reduced (1)(2)(3). Glyceroneogenesis is increased during fasting and allows the re-esterification of fatty acids (FAs) 5 for triacylglycerol synthesis at a time when their breakdown is occurring through lipolysis (4). This re-esterification pathway was proposed several decades ago as a regulatory control process developed by adipocytes to prevent an all-out release of FAs into the blood (5)(6)(7). Cytosolic phosphoenolpyruvate carboxykinase (PEPCK-C) is a critical step in glyceroneogenesis (4). The physiological importance of PEPCK-C in FA homeostasis and storage was definitively confirmed by the phenotypes of mice in which adipocyte PEPCK-C gene expression was either invalidated or overexpressed. Invalidation of the PEPCK-C gene in white adipose tissue of mice results in lipodystrophy and predisposition to insulin resistance (8). In contrast, mice that constitutively overexpress the PEPCK-C gene in adipose tissue become obese because of increased FA storage, without exhibiting any increase in their plasma FA levels (9), unless they are fed a normolipidic diet (10). These results suggest that the control of FA homeostasis exerted by glyceroneogenesis in adipose tissue, together with an equilibrated diet, could participate in the prevention of type 2 diabetes. Indeed, a sustained increase in plasma non-esterified FAs, a situation that occurs with a high fat diet and in lipodystrophy, can be a leading cause of insulin resistance and ultimately of type 2 diabetes (11).
Thiazolidinediones (TZDs) are the latest generation of antidiabetic drugs that are able to normalize several plasma parameters. However, their mechanism of action is not yet completely understood. Because TZDs are the most well known ligands for peroxisome proliferator-activated receptor-␥ (PPAR␥), a transcription factor that is expressed at high levels in adipose tissue (12), and because TZDs are potent stimulators of selective PEPCK-C gene expression in adipocytes (13), we hypothesized that TZDs could exert some of their hypolipidemic action by inducing glyceroneogenesis, thereby reducing FA output from adipose tissue. Indeed, we identified this metabolic pathway as a TZD target in cultured adipocytes and fat tissue of Wistar rats (14) and, more recently, in human subcutaneous adipose tissue (15), thus confirming that the glyceroneogenesis-dependent FA-lowering effect of TZDs could be an essential aspect of the antidiabetic action of these drugs.
Cytosolic aspartate aminotransferase (cAspAT; EC 2.6.1.1) is coupled with PEPCK-C for liver gluconeogenesis because it produces its substrate (16,17). Furthermore, cAspAT gene expression is regulated by glucocorticoids, as is the expression of other genes belonging to this pathway (18,19). Similar to PEPCK-C, cAspAT is down-regulated by glucose and up-regulated by a low glucose/high protein diet in adipose tissue (20,21). Thus, we postulated that cAspAT might be involved in glyceroneogenesis in adipose tissue under conditions of restrained glucose transport and availability.
In this work, we show that cAspAT is functionally involved in glyceroneogenesis in adipocytes in which transcription of its gene is induced by TZDs, thus allowing increased glycerol-3-P production and, ultimately, triacylglycerol synthesis from aspartate. Surprisingly, the TZD responsiveness of cAspAT was found to be dependent upon both PPAR␥ and protein synthesis. The TZD-responsive region of the gene promoter was identified as a unique AGGTCA sequence that bound retinoic acidrelated orphan receptor-␣ (ROR␣), but not PPAR␥. ROR␣4 was able to transactivate the cAspAT promoter, and its amount was increased in nuclear extracts from TZD-treated adipocytes. ROR␣4 is thus suggested to be a transcription factor involved in the regulation of cAspAT gene expression by TZD because its own expression is controlled by rosiglitazone.

EXPERIMENTAL PROCEDURES
Cell Culture and Treatment-3T3-F442A, 3T3-C2, and BFC-115 adipocytes were cultured in Dulbecco's modified Eagle's medium (DMEM) containing glucose (25 mM), penicillin (200 IU/ml), streptomycin (50 mg/liter), biotin (2 mg/liter), and pantothenate (4 mg/liter) with 10% newborn calf serum at 37°C in a humidified atmosphere of 10% CO 2 and 90% air. At confluence, newborn calf serum was changed to 10% fetal calf serum, and insulin (20 nM) was added to the medium to initiate adipocyte differentiation. The medium was changed every 2-3 days. Experiments were carried out using differentiated adipocytes (8 -10 days after confluence) unless described otherwise, and drug treatments were performed as described in the legends to the figures. Differentiation of 3T3-L1 adipocytes was induced with isobutylmethylxanthine, insulin, and dexamethasone as described previously (22). COS-7 cells were cultured in DMEM with 10% fetal calf serum and then transfected (FuGENE 6, Roche Applied Science) by PPAR␥ and retinoid X receptor-␣ (RXR␣) expression vectors before nuclear extract isolation. The human adipose tissue came from normal weight patients who participated in the study described by Leroyer et al. (15). Human adipose tissue explants were cut into small pieces (ϳ20 mg), rinsed, and incubated in the same medium as that used for 3T3-F442 adipocytes but with the addition of 1 M rosiglitazone or ethanol for 24 h before RNA extraction.
Functional Analysis-For glyceroneogenesis studies, mature 3T3-F442A adipocytes in 6-well dishes were treated for 72 h with either ethanol or 1 M rosiglitazone in DMEM containing 10% fetal calf serum. The medium was then replaced with glucose-and serum-free DMEM supplemented with 0.3% FA-free bovine serum albumin. Four hours later, the incorporation of a glyceroneogenic substrate was evaluated by adding 2 ml of Krebs-Ringer phosphate buffer containing 0.3% FA-free bovine serum albumin (Krebs-Ringer phosphate buffer/bovine serum albumin) with either 2 mM aspartate or pyruvate and 1 Ci of [ 14 C]aspartate or [ 14 C 1 ]pyruvate in the presence of 150 M 3-mercaptopicolinic acid as indicated. One hour later, the cells were washed twice with ice-cold 1ϫ phosphate-buffered saline and homogenized in radioimmune precipitation assay buffer (50 mM Tris (pH 8), 0.150 mM NaCl, 1% Nonidet P-40, 0.1% SDS, and 0.5% deoxycholate). Lipids were extracted by the simplified method of Bligh and Dyer (23), and [ 14 C]aspartate or [ 14 C 1 ]pyruvate incorporation was estimated by counting the radioactivity in the lipid fraction. Fatty acid and glycerol release was measured under lipolytic conditions as follows: 1 h after adding 1 M isoproterenol to the cells, glycerol and fatty acid concentrations were measured in the cultured medium using kits from Sigma and Roche Applied Science, respectively.
For cAspAT activity, 3T3-F442A adipocytes were collected in extraction buffer (50 mM HEPES (pH 7.8) containing 40 mM KCl, 11 mM MgCl 2 , 1 mM EDTA, and 1 mM dithiothreitol). The cell lysate was adjusted to 4 mM ␣-ketoglutarate to protect the cytosolic isoenzyme, and the cell lysate was incubated at 70°C for 8 min to destroy mitochondrial aspartate aminotransferase according to Parli et al. (24). After centrifugation at 15,000 rpm for 15 min at 4°C, cAspAT activity was measured according the instructions provided by ABX Diagnostics (Montpellier, France), and protein content was determined by the BCA assay (Sigma).
RNA Extraction and Analysis-The cells were treated for 24 h with either 1 M rosiglitazone or 5 M pioglitazone in DMEM containing 5% dialyzed serum. RNA was extracted by the method of Chomczynski and Sacchi (25). Total RNA was electrophoresed on a 1% agarose gel containing 2.2 M formaldehyde and blotted onto a nylon membrane (Hybond-N ϩ , Amersham Biosciences). The integrity and relative amounts of RNA were assessed by methylene blue staining. Prehybridization and hybridization of the blots were carried out using QuickHyb solution (Stratagene). Membranes were hybridized for 1 h at 68°C with 100 g/ml sonicated salmon sperm DNA and 10 6 cpm/ml cDNA labeled with [␣-32 P]dATP by random priming according to the manufacturer's instructions. Membranes were washed twice for 15 min at room temperature with 2ϫ SSC and 0.1% SDS and then for 30 min at 60°C with 0.1% SDS and 0.1 ϫ SSC. The specific probes used were PC116, a rat PEPCK-C cDNA fragment (26), and cAspAT and mitochondrial aspartate aminotransferase cDNA fragments as described (18). An oligonucleotide specific for the 18 S rRNA was 32 Plabeled and used as a control as described previously (20). mRNA signals were quantified using an Instant Imager (Packard Instrument Co.) and corrected for differences in RNA load-ing by comparison with the signals generated by the 18 S cDNA probe.
Total RNA from 200 mg of human adipose tissue was isolated with an RNeasy kit (Qiagen Inc.). cDNA was synthesized with random hexamers and avian myeloblastosis virus reverse transcriptase (Promega Corp.). Real-time PCR was performed using a Roche Diagnostics LightCycler and LightCycler SYBR Green fluorophore. The following primers were used: ROR␣, GTCA-GCAGCTTCTACCTGGAC and GTGTTGTTCTGAGAGT-GAAAGGCACG; cAspAT, TTCTGTTCCCCTTCTTTGAC and GATTCCCCACTCTCTCATTG; and 18 S rRNA, GAGC-GAAAGCATTTGCCAAG and GGCATCGTTTATGGTCG-GAA. Each PCR experiment was normalized to the amount of 18 S rRNA.
Transient Transfection and Chloramphenicol Acetyltransferase (CAT) Assays-The experiments were carried out on 8-day post-confluent differentiated 3T3-F442A adipocytes using the polyethyleneimine (PEI) adenofection method described previously (27). Briefly, for each 25-mm plate, 2.5 g of plasmid DNA and an amount of 25-kDa PEI corresponding to a PEI nitrogen/DNA phosphate ratio of 1.5 were diluted into 16.6 l of 0.15 M NaCl and then mixed together for 15 min with intermittent vigorous vortexing. Polycation⅐DNA complexes and a non-replication-deficient adenoviral vector containing the Rous sarcoma virus nuclear localization sequence and the lacZ gene (20 -400 plaque-forming units) were then diluted into 2 ml of serum-free medium and added to culture dishes. Seven hours later, the medium was discarded, and cells were rinsed twice with serum-free medium and allowed to recover for 16 h before the addition of various effectors in DMEM supplemented with 5% dialyzed fetal calf serum for 48 h. Cells were then scraped and homogenized as detailed (28). CAT assays were performed as described by Seed and Sheen (29), and protein content was measured by the BCA assay. The rat gene promoter-CAT constructs used were Ϫ2405-cAspAT, Ϫ553-cAspAT, Ϫ398-cAspAT, Ϫ286-cAspAT, and Ϫ177-cAspAT (30) and Ϫ1800-PEPCK-C (26). The three mutations of the AGGTCA hexanucleotides in the Ϫ553-cAspAT construct were as described (31). The expression vector encoding dominant-negative PPAR␥ (pSG5-DN-PPAR␥) was constructed according to Gurnell et al. (32). The mouse vectors pCMX-ROR␣4 and pCMX-ROR␣1 were from Dr. V. Giguère (McGill University, Quebec, Canada) (33) and kindly provided by Dr. J.-L. Danan (CNRS UPR 9078). Transfection of 3T3-C2 cells was performed 8 days post-confluence by the FuGENE 6 technique according to the manufacturer's instructions.
Nuclear Extract Preparation and Gel Mobility Shift Assays-Nuclear proteins were extracted from 3T3-F442A adipocytes (control or 18-h TZD-treated) at different stages of differentiation and from COS-7 cells in which PPAR␥ and RXR␣ were ectopically expressed. Cells were scraped on ice and washed twice with phosphate-buffered saline. After centrifugation at 800 ϫ g for 5 min at 4°C, the pellet was dissolved in 1ϫ buffer (10 mM Tris-HCl (pH 7.8), 10 mM NaCl, 3 mM MgCl 2 , and 0.002% NaN 3 ) and 0.5% Nonidet P-40 containing a mixture of protease inhibitors (Roche Applied Science). The cell lysate was centrifuged at 5100 ϫ g for 5 min at 4°C. The pellet was resuspended in 20 mM HEPES (pH 7.4), 420 mM NaCl, 25% glycerol, 1.5 mM MgCl 2 , 0.2 mM EDTA, and 0.01% NaN 3 . After a 30-min incubation, the solution was centrifuged at 100,000 ϫ g for 15 min at 4°C. The supernatant containing nuclear protein was mixed (v/v) with 20 mM HEPES, 50 mM KCl, 0.2 mM EDTA, and 0.01% NaN 3 , aliquoted, and stored in liquid nitrogen.
For gel mobility shift assays, double-stranded oligonucleotides corresponding to the cAspAT sequences and PPAR-responsive element (PPRE) sequences (PCK2) from the rat PEPCK-C gene (13) and the acyl-CoA oxidase gene (34) were end-labeled with [␣-32 P]dCTP using Klenow polymerase. Nuclear extracts containing 10 g of protein were incubated on ice for 30 min with 6 l of electrophoretic mobility shift assay (EMSA) mixture (17.5 mM HEPES (pH 7.6), 80 mM KCl, 0.1 mM EDTA, 8 mM MgCl 2 , 8% glycerol, and 500 ng poly(dI-dC)) and 0.5 ng of the labeled probe; VF-15 l. DNA⅐protein complexes were separated on a 5% polyacrylamide gel in 0.5ϫ TBE ( 0.5 M Tris, 0.5 M boric acid, and 5 mM EDTA). The gel was dried and exposed to x-ray film (Amersham Biosciences). For supershift analysis, 1 or 2 l of anti-PPAR␥ antibody (catalog no. sc-1984X) or anti-ROR␣ antibody (catalog no. sc-6062X) (both from Santa Cruz Biotechnology, Inc.) was preincubated at room temperature for 15 min with the nuclear extracts before the binding reaction.
Unlabeled ROR␣1 and ROR␣4 proteins were obtained in vitro from the pCMX-hROR␣1 or pCMX-hROR␣4 vector (kind gifts of Dr. V. Giguère) using the TNT T7 quick coupled transcription/translation kit (Promega Corp.). DNA⅐protein complexes were allowed to form by incubating aliquots of the programmed lysates for 20 min on ice as described by Chauvet et al. (35). For competition experiments, double-stranded oligonucleotides were added simultaneously with the labeled probe.
Oligonucleotide sequences were as follows: Western Blot Analysis-Aliquots of 3T3-F442A adipocyte nuclear extracts (25 g) and ROR␣-programmed lysates were analyzed by Western blotting with antiserum against ROR␣ (catalog no. sc-6062, Santa Cruz Biotechnology, Inc.) as described previously (36). Amounts of protein were quantified with a Typhoon 9400 (Amersham Biosciences) and standardized versus respective actin amounts.

Involvement of cAspAT in Adipocyte Glyceroneogenesis:
Induction by Rosiglitazone-We have shown previously that rosiglitazone is able to stimulate [ 14 C 1 ]pyruvate incorporation into adipocyte triacylglycerols (14). This leads to an increase in FA re-esterification and thus in the reduction of FA release into the culture medium (14) when the cells are in a lipolytic situation. This effect depends on PEPCK-C activity because it is prevented by 3-mercaptopicolinic acid, an inhibitor of this enzyme. To determine whether cAspAT is also involved in glyceroneogenesis, we carried out a similar experiment using [ 14 C]aspartate as a potential precursor of glycerol-3-P. Aspartate is the substrate for cAspAT and allows the generation of oxalacetate, the substrate of PEPCK-C. Thus, cAspAT could act upstream of PEPCK-C in the glyceroneogenic pathway. Fig. 1A shows that [ 14 C]aspartate was in fact incorporated into the neutral lipid fraction of 3T3-F442 adipocytes, although to a lower extent than pyruvate, and thus could be a precursor of glycerol-3-P for FA re-esterification. Pretreatment of the cells with rosiglitazone for 72 h increased aspartate incorporation by 2-fold compared with the 2.5-fold increase for pyruvate and was abolished by 3-mercaptopicolinic acid, suggesting a role for PEPCK-C.
To further evaluate the association of these results with cAspAT, we measured its enzyme activity. As soon as 12 h after the beginning of the rosiglitazone treatment, activity increased, reaching a maximum at 48 h (1.9-fold) (Fig. 1B). To determine whether the increased cAspAT activity was related to a physiological benefit (i.e. a decrease in FA output from adipocytes placed in a lipolytic situation), we measured the FA and glycerol contents of the culture medium of the cells that had been incubated for 1 h with 2 mM aspartate and 1 M isoproterenol.
Glycerol release has long been considered as a lipolytic index because of the very low glycerol kinase activity in adipocytes (5). However, its bioavailability was recently shown to be affected by TZD treatment in 3T3-F442A adipocytes (14) and in adipose tissue from lean human patients (15). Indeed, TZD is able to induce glycerol kinase gene expression (37), thus allowing the use of glycerol for the re-esterification of FA at a ratio of 3 mol of FA to 1 mol of glycerol. Thus, in response to TZD treatment, two FA re-esterification pathways appear to co-exist in adipocytes: 1) the glycerol kinase-dependent path, which allows glycerol phosphorylation and does not modify the FA/glycerol lipolytic ratio, and 2) the glyceroneogenesis-dependent FA re-esterification path, which decreases this ratio because it affects only FA (14,15). Fig. 1C illustrates such a decrease in the FA/glycerol ratio when aspartate was provided as a substrate for glyceroneogenesis. This decrease in FA release was time-dependent and inversely proportional to the increase in cAspAT activity generated by the rosiglitazone treatment. These data confirm that cAspAT is involved in the glyceroneogenic pathway by providing glycerol-3-P from aspartate to further re-esterify FA into triacylglycerol and that cAspAT, as well as PEPCK-C, is a potential target for TZD in adipocytes.
TZD Induction of cAspAT Gene Expression Is Specific for Mature Adipocytes-To determine whether the increase in cAspAT enzyme activity following rosiglitazone treatment was due to an effect at the gene level, we performed Northern blot analysis in various adipocyte cell lines (Fig. 2). In differentiated 3T3-F442A adipocytes, cAspAT mRNA levels were increased by rosiglitazone treatment and by another TZD, pioglitazone ( Fig. 2A). The same increase by rosiglitazone treatment was also observed in two other adipocyte cell lines, 3T3-L1 (Fig. 2C) and BFC-115 (Fig. 2D), a cell line derived from mouse brown adipose tissue (38), in which glyceroneogenesis is very active (39). In contrast, cAspAT mRNA levels were not changed by rosiglitazone treatment of 3T3-F442A preadipocytes (Fig. 2B). Thus, our results show that TZDs increase cAspAT mRNA levels only in fully differentiated adipocytes, the maximal -fold increase versus 18 S rRNA (3.2 Ϯ 0.8-fold) being observed at day (D) 8 post-confluence. Ci of 14 C-radiolabeled substrate in the presence or absence of the PEPCK-C inhibitor 3-mercaptopicolinic acid (3MP; 0.15 mM). Lipids were extracted from the cells, and radioactivity was measured as described under "Experimental Procedures." The results are expressed as nmol/mg of protein (prot)/h and represent the mean Ϯ S.E. of at least three independent experiments. G3P, glycerol-3-P. B and C, 3T3-F442A adipocytes were treated with 1 M rosiglitazone, and both cAspAT activity and glyceroneogenesis-dependent FA release were analyzed over time. cAspAT activity was measured in the cells at the indicated times. FA and glycerol release was simultaneously evaluated in the medium during 1 h of lipolysis (with 1 M isoproterenol) and in the presence of 2 mM aspartate. The glyceroneogenesis-dependent FA re-esterification process decreased the FA/glycerol ratio because it affected only one parameter, the FA level. Results are the mean Ϯ S.E. of three independent experiments.

cAspAT in Adipocyte Glyceroneogenesis
Involvement of PPAR␥ in the Rosiglitazone Effect on cAspAT Gene Expression-PPAR␥ is a nuclear receptor that binds TZDs with high affinity (12). Therefore, we expect that concentrations of TZDs as low as 10 nM would induce cAspAT gene expression if PPAR␥ is really involved in the regulation. We first carried out a time course experiment in 3T3-F442A adipocytes and determined that induction of cAspAT mRNA by rosiglitazone treatment was maximal at 24 h of exposure (data not shown). 3T3-F442A adipocytes were then treated with increasing concentrations of rosiglitazone (1 nM to 10 M); Fig. 2F illustrates the maximal observed effect between 10 nM and 10 M.
We next compared the differentiation-dependent expression of the cAspAT and PPAR␥ genes in the 3T2-F442A cell line (Fig. 2E). In contrast to the PPAR␥ mRNA level, which greatly increased from D0 to D6, the cAspAT mRNA level did not significantly change during the differentiation process. Thus, during adipogenesis, cAspAT gene expression does not fit the same pattern as PPAR␥ and its dependent genes such as PEPCK-C.
We next assessed the effect of rosiglitazone treatment on cAspAT gene transcription. To address this question, we transiently transfected 3T3-F442A adipocytes by PEI adenofection (27) with the 5Ј-deleted rat cAspAT gene promoter upstream of the CAT reporter gene (30). The transfected adipocytes were then treated with 1 M rosiglitazone for 48 h before assaying CAT activity. The Ϫ2405, Ϫ553, and Ϫ398 bp constructs responded to rosiglitazone treatment with a 2-fold increase in CAT activity (Fig. 3A), indicating that the latter sequences encompass the TZD-responsive region. In contrast, rosiglitazone responsiveness was completely abolished when the Ϫ286 and Ϫ177 bp constructs were used. Thus, the TZD-responsive sequence of the cAspAT regulatory region appears to be located between Ϫ398 and Ϫ286 bp upstream of the translation initiation site (position ϩ1). Using cells transfected with the Ϫ398 bp construct, we tested the effect of other PPAR and RXR ligands. Fig. 3B shows that, whereas pioglitazone was as efficient as rosiglitazone, neither the RXR ligand 9-cis-retinoic acid nor the PPAR␣ activator clofibrate acted as a transactivator. Hence, in adipocytes, this region responds selectively to PPAR␥ agonists. Furthermore, the lack of responsiveness to 9-cis-retinoic acid also suggests that cAspAT is not a direct PPAR␥ target.
To obtain further support for the involvement of PPAR␥ in the regulation of cAspAT gene expression, dominant-negative (DN) PPAR␥ was tested. This mutated PPAR binds DNA and its ligand, but has impaired recruitment of steroid receptor   AUGUST 10, 2007 • VOLUME 282 • NUMBER 32 coactivator-1 and CAAT-binding protein and is constitutively linked to corepressors (32). We transfected increasing concentrations of DN-PPAR␥ into 3T3-F442A adipocytes together with the Ϫ398-cAspAT construct (Fig. 4A, left panel). Cells were then treated or not with rosiglitazone for 48 h before assaying CAT activity. In the absence of rosiglitazone, DN-PPAR␥ decreased CAT activity by 2-fold at all concentrations tested. This effect was specific for the Ϫ398-cAspAT construct and was not found for the TZD-unresponsive Ϫ177-cAspAT construct. As expected, rosiglitazone induced CAT activity by ϳ2-fold in the absence of the DN-PPAR␥ expression vector. Interestingly, DN-PPAR␥ completely abolished rosiglitazone induction of CAT expression (Fig. 4A, left panel), thereby supporting the involvement of PPAR␥ not only in the basal but also in the TZD stimulation of cAspAT gene transcription. In the same cells and in the presence of rosiglitazone, increasing concentrations of DN-PPAR␥ progressively inhibited the Ϫ1800-PEPCK-C-CAT construct (Fig. 4B), used as a positive control, whereas the maximal inhibition of the Ϫ398-cAspAT construct was obtained even at the lowest concentration used and had no effect on the Ϫ177-cAspAT-CAT construct (Fig. 4A), used as a negative control. Moreover, although overexpression of wild-type PPAR␥ increased transcription of the Ϫ1800-PEPCK-C-CAT construct in 3T3-F442A adipocytes by 13-fold (Fig. 4B), it had a limited effect (1.5-fold) on the Ϫ398-cAspAT construct and no effect on the Ϫ177-cAspAT construct (Fig. 4A). These results confirm that PPAR␥ acts differently on the cAspAT promoter compared with a direct PPAR␥ target such as the PEPCK-C gene.

cAspAT in Adipocyte Glyceroneogenesis
Induction of cAspAT Gene Transcription by Rosiglitazone Is Indirect-We hypothesized that a direct target via PPAR␥ would be insensitive to protein synthesis inhibition (34). However, we observed that co-treatment of cycloheximide (10 M) and rosiglitazone totally prevented rosiglitazone induction of cAspAT in 3T3-F442A adipocytes without affecting its basal expression (data not shown). This suggested that the synthesis of a protein with a rapid turnover would be required for rosiglitazone-induced stimulation of cAspAT gene expression.
TZD-activated PPAR␥ transactivates responsive genes by binding as a heterodimer with RXR␣ to direct repeat (DR)-1 consensus sequences (AGGTCAnAGGTCA) named PPREs, which are present in the regulatory regions of the target genes (34). We found no DR1 consensus sequence in the Ϫ398/Ϫ286 bp sequence of the cAspAT gene (Fig. 5A), which is involved in the response to rosiglitazone treatment. However, close analysis of this region revealed a potential DR2 sequence, which has been shown recently to be a target of PPAR␥ in the Rev-Erb␣ gene promoter (40). It also contains three nuclear receptorresponsive elements (NRREs) (31), which could be potential binding sites for members of the nuclear hormone receptor superfamily such as ROR and Rev-Erb␣ (Fig. 5A). As shown by EMSA, three oligonucleotides covering this region (probes 7, 9, and 11) bound proteins from 3T3-F442A adipocyte nuclear extracts. However, none of the complexes was recognized by the anti-PPAR␥ antibody under experimental conditions that allowed the disappearance of the PPAR␥⅐RXR␣ complex bound to the PPRE sequence (PCK2) of the PEPCK-C gene (Fig. 5B). Consequently, we tested the ability of PPAR␥ and RXR␣ (ectopically overexpressed in COS-7 cells) to bind to the Ϫ398/ Ϫ286 bp fragment (Fig. 5C). None of the three oligonucleotides (probes 7, 9, and 11) was able to bind the PPAR␥/RXR␣ heterodimer, whereas the PPRE of the acyl-CoA oxidase gene displayed a major retarded band that was supershifted by the anti-PPAR␥ antibody, as expected. These results indicate that TZD induction of cAspAT gene transcription does not require direct PPAR␥/RXR␣ binding to the Ϫ398/Ϫ286 bp sequence.

Identification of the Binding Site That Confers Rosiglitazone Responsiveness to the cAspAT Gene Promoter and Analysis of Protein Binding to This NRRE-rich Region-
We then focused our study on the Ϫ416/Ϫ364 bp promoter region, which displays three NRRE sequences potentially able to bind the ROR subfamily and its negative form, Rev-Erb␣. The expression of these transcription factors is increased by PPAR␥ agonists in adipocytes (40,41). In an attempt to more precisely identify the TZD-responsive sequence on the cAspAT gene promoter, 3T3-F442A adipocytes were transfected with the Ϫ553-cAspAT construct in which each NRRE had been mutated (named Mut1, Mut2, and Mut3, from 5Ј to 3Ј) (Fig. 6) (31). Mut3 completely abolished the rosiglitazone responsiveness of the construct, whereas neither Mut1 nor Mut2 significantly affected the regulation. Thus, the NRRE located at Ϫ381 to Ϫ376 bp, but

cAspAT in Adipocyte Glyceroneogenesis
not the potential DR2 sequence (invalidated in Mut1 or Mut2), is necessary for the rosiglitazone responsiveness of the cAspAT gene promoter.
We then performed EMSA as described previously (31) using either the wild-type Ϫ404/Ϫ374 bp sequence or its mutant (Mut3). We used nuclear extracts prepared from differentiating 3T3-F442A adipocytes, both untreated and treated for the last 18 h at D1, D4, and D8 postconfluence with rosiglitazone. We first observed that the electrophoretic pattern of control nuclear extracts obtained with the Ϫ404/Ϫ374 bp sequence varied with the differentiation state of the adipocytes (Fig. 7,  compare lanes 1, 3, and 5). Complexes A and F appeared only at D4 and further increased in amount at D8. Complex B, faintly present in untreated D1 adipocytes (lane 1), was also enhanced at D4 (lane 3). Complexes C and D were apparently not modified. Finally, complex E seemed to completely disappear at D8 (lane 5).
After rosiglitazone treatment, the amounts of complexes C and E were increased only at D1 (referred to as D1R) (Fig. 7, lane 2), and the binding pattern obtained at D4 (D4R) (lane 4) was similar to that obtained with untreated D8 adipocytes (lane 5). This change between D4 and D4R was not thought to be related to the TZD control of the cAspAT gene that occurred later only at D8; it may be the result of the acceleration of the differentiation process by rosiglitazone. We thus focused on nuclear extracts from D8, where the Ϫ404/Ϫ374 bp cAspAT sequence specifically bound five complexes (named A, B, C, D, and F) (lane 5). It was specifically complexes A, B, and F, the binding of which increased throughout adipocyte differentiation, that seemed to bind this sequence more strongly after rosiglitazone treatment (lane 6). Indeed, the amount of binding (as measured using an Instant Imager) was always found to be greater than that for complexes C and D (1.8-fold for complexes A, B, and F versus 1.3fold for complexes C and D) in rosiglitazone-treated (D8R) versus untreated (D8) nuclear extracts. Furthermore, our data confirmed our above observations that there was apparently no new binding when cells were treated with rosiglitazone ( lanes 5 and 6). FIGURE 5. EMSA of the ؊398/؊296 bp fragment of the cAspAT gene. A, shown is the sequence of the Ϫ398/Ϫ296 bp region of the cAspAT regulatory region. AspAT7, AspAT9, and AspAT11 are the three probes used for EMSAs. Potential NNREs are indicated by arrows. B, nuclear extracts were prepared as described under "Experimental Procedures." In gel shift assays, 32 P-radiolabeled oligonucleotides containing either the PPRE (PCK2) of the PEPCK-C gene or the three different sequences of the cAspAT gene (AspAT7, AspAT9, and AspAT11) were incubated with 10 g of 3T3-F442A adipocyte nuclear extracts (NE) for 30 min. The anti-PPAR␥ antibody (Ab) was added 15 min before the binding reaction. The results shown were obtained with the AspAT7 (lanes 1 and 2), AspAT9 (lanes 3 and 4), AspAT11 (lanes 5 and 6), and PCK2 (lanes 7 and 8) probes. C, COS-7 cells were transfected with the PPAR␥ and RXR␣ expression vectors (10 g/dish). Nuclear extracts were prepared as described under "Experimental Procedures." Gel shift assays were carried out with 10 g of nuclear extracts from COS-7 cells and 1 g of nuclear extracts from 3T3-F442A adipocytes (to prevent the lack of coactivators). Lanes 1-3, results obtained with the AspAT7, AspAT9, and AspAT11 probes, respectively; lanes 4 -6, the complex obtained with the acyl-CoA oxidase (ACOX) PPRE probe. The arrowhead shows the band corresponding to the supershift induced by the anti-PPAR␥ antibody.  AUGUST 10, 2007 • VOLUME 282 • NUMBER 32

cAspAT in Adipocyte Glyceroneogenesis
Finally, we compared the protein binding pattern of the Mut3 (TZD-unresponsive) sequence versus the wild-type Ϫ404/Ϫ374 bp cAspAT (TZD-responsive) sequence. In control nuclear extracts (D8), the Mut3 sequence showed diminished binding for some complexes (complexes C and F), but, more surprisingly, an increased binding for some others (complexes A and B) (Fig. 7, compare lanes 10 and 5). After rosiglitazone treatment, the intensity of these DNA complexes A and B varied differently according to the probe used (decreasing with Mut3 (lanes 10 and 11) while increasing with the wild-type probe (compare lanes 10 and 11 versus lanes 5 and 6)).
Finally, cross-competition experiments with a 50-fold molar excess of unlabeled oligonucleotides (wild-type or Mut3) revealed that both sequences competed rather similarly with the complexes obtained with both probes (Fig. 7, lanes 7-9 and  12-14). Indeed, to be visualized, their exposition times shown in Fig. 7 were longer than that used for lanes 5 and 6 and lanes 10 and 11. Taken together, our data show that (i) the rosiglitazone responsiveness of the cAspAT gene requires the integrity of sequence AGGACA (located between Ϫ381 and Ϫ376 bp), the mutation of which deals with a decrease in some complex binding already in untreated adipocytes, thus suggesting impairment of transcription factor binding to this site; and (ii) the TZD-responsive cAspAT sequence specifically binds some nuclear factors, the amount of which increases with the differentiation process. Thus, our data suggest the involvement of adipocyte-enriched transcription factors able to bind a AGGACA sequence in mediating the TZD responsiveness of the cAspAT gene.
ROR␣ Is Involved in the Complexes That Bind the Ϫ404/ Ϫ374-bp Region of the cAspAT Gene Promoter-To determine whether ROR␣ could be this transcription factor, new EMSAs were performed using the Ϫ404/Ϫ374 bp region of the cAspAT gene promoter, new nuclear extracts prepared from D10 postconfluent adipocytes that had been treated with rosiglitazone for the last 18 h, and antibodies directed against ROR␣ and PPAR␥. Rather similar protein binding patterns were observed with the Ϫ404/Ϫ374 bp region for D10 and D8 adipocyte nuclear extracts, except that the intensity of complex F was increased at D10 (compare Fig. 7,  lane 6, and Fig. 8A, lane 1). This suggested that the amount of complex F, which decreased with the Mut3 probe, still increased upon differentiation when the Ϫ404/Ϫ374 bp region was used.
We then confirmed that the anti-PPAR␥ antibody did not supershift any of the complexes bound to the Ϫ404/Ϫ374 bp region (Fig. 8A,  lanes 2 and 2Ј), as shown previously for probe 7 (Fig. 5B), whereas the same antibody supershifted the complex obtained with the acyl-CoA oxidase PPRE (data not shown). However, a faint supershift with the anti-ROR␣ antibody was visualized only when the gel with the Ϫ404/Ϫ374 bp cAspAT gene promoter used a probe was overexposed (Fig. 8A, lanes 1Ј, 2Ј, and 3Ј). This suggests that ROR␣ might be one of the proteins that binds to this TZDresponsive region of the cAspAT gene. Furthermore, when a nuclear extract from control cells was used, this faint supershift could not be seen (data not shown), suggesting that rosiglitazone treatment probably increased the amount of ROR␣ in the extract, as also shown by Austin et al. (41). Finally, we could not detect any supershift of the complexes that bound to Mut3 with the anti-ROR␣ antibody (data not shown).
To definitively confirm the ROR␣ binding capacity, we performed EMSA with in vitro transcribed/translated human ROR␣1 (hROR␣). Fig. 8B shows that the Ϫ404/Ϫ374 bp cAspAT sequence bound ROR␣1 (third lane), as found with the fibrinogen ROR␣-responsive element-binding site used as a positive control (first lane). The specificity of ROR␣ binding to both probes was determined first by using a specific antibody against ROR␣, resulting in a complete supershift of the DNA⅐ROR complexes (second and fourth lanes). Furthermore, competition with the unlabeled human fibrinogen ROR␣-responsive element resulted in rapid inhibition of binding (fifth through seventh lanes), which was greatly less effective with the unlabeled mutated human fibrinogen ROR␣-responsive element (eighth through tenth lanes). Identical results were obtained when the competitions were performed with increasing excesses (15, 50-, and 100-fold) of the unlabeled wild-type and Mut3 Ϫ404/Ϫ374 bp cAspAT sequence (eleventh through thirteenth and fifteenth through seventeenth lanes). These data indicate that the TZD-responsive sequence of the cAspAT gene specifically binds ROR␣ and that mutation of sequence AGGACA significantly impairs such binding.
ROR␣ Increases Expression of the Ϫ398-cAspAT-CAT Construct in 3T3-C2 Cells and Is a Rosiglitazone Target-To test the functional involvement of the ROR␣ subfamily in cAspAT gene transcription and particularly the hypothesis that ROR␣ could be the transcription factor that is synthesized in a PPAR␥dependent way, we compared the effects of the ectopic expression of ROR␣4 with those of PPAR␥/RXR␣ on the Ϫ398- FIGURE 7. Protein binding to the wild-type and mutated (Mut3) ؊404/؊374 bp cAspAT DNA sequences that confer rosiglitazone responsiveness to the cAspAT gene promoter. Radiolabeled Ϫ404/Ϫ374 bp cAspAT (wild-type and Mut3) oligonucleotides were incubated with nuclear extracts prepared from adipocytes at D1, D4, and D8 post-confluence and with or without an 18-h treatment with 1 M rosiglitazone (D1R, D4R, and D8R). Self-and cross-competitions (Comp.) were performed in the presence of a 50-fold molar excess of the indicated unlabeled oligonucleotide (lanes 7-9 and 12-14).
cAspAT construct expression in 3T3-C2 cells. We chose this preadipocyte subclone, which is unable to correctly differentiate into adipocytes, as the recipient cell because it expresses PPAR␥ weakly (40) and does not express Rev-Erb␣ (41). This avoids potential competition with ROR␣. Indeed, Rev-Erb␣, which has no transcription activity, would act as a negative competitor on this NRRE-rich region. The ectopic expression of ROR␣ increased expression of the Ϫ398-cAspAT construct by 2-fold whether in the absence or presence of rosiglitazone, and this effect was specific for the TZD-responsive region because the Ϫ177-cAspAT construct was not affected (Fig.  9A). This confirmed that ROR␣ was able to activate cAspAT gene transcription by itself and that it was not able to restore sensitivity to rosiglitazone in the absence of PPAR␥. Furthermore, ectopic expression of PPAR␥/RXR␣ increased the Ϫ398-cAspAT construct basal expression by 5-fold, whereas it was augmented only by 1.7-fold when 3T3-F442A adipocytes were used (compare Figs. 4A and 9A). This confirms that 3T3-C2 cells have much lower amounts of PPAR␥ compared with 3T3-F442 cells, as demonstrated previously by Eubank et al. (42). As expected, rosiglitazone treatment increased transcription of the Ϫ398-cAspAT construct directed by the ectopic expression of PPAR␥/RXR␣ by 1.5fold. This ligand effect is similar to the TZD response rate of the cAspAT gene in PPAR␥-expressing adipocytes. That this TZD-responsive promoter sequence did not bind PPAR␥ confirms that PPAR␥ acts by stimulating the neosynthesis of another transcription factor both by itself and, even more efficiently, when activated by its ligand.
We then examined the level of ROR␣ protein in nuclear extracts from preadipocytes (D1 post-confluence) and from either untreated or 18-h rosiglitazone-treated D10 adipocytes. The Western blot shown in Fig. 9B indicates that the ROR␣4 protein amount, not detectable at D1, was increased by rosiglitazone (1.8-fold versus the actin amount). The presence of ROR␣1 protein was found to be rather weak and did not lead to any significant conclusion.
To strengthen the hypothesis that ROR␣ could be the transcription factor responsible for cAspAT expression in response to PPAR␥ agonists, we measured the amounts of ROR␣ and cAspAT mRNAs in six human adipose tissue explants untreated or treated for 24 h with rosiglitazone. The amounts of both ROR␣ and cAspAT mRNAs increased in response to the PPAR␥ ligand (Fig. 9C). However, the increases were 3-4-fold in some individuals but 1.5-fold in others. These results support the existence of interindividual variation in the TZD responsiveness of human adipose tissue, as shown previously for the PEPCK-C gene (15).
Taken together, our data show that (i) both PPAR␥ and ROR␣4 can significantly transactivate cAspAT gene expression in recipient cells that do not endogenously express these nuclear factors; (ii) the ROR␣4 protein level is specifically increased by rosiglitazone in adipocyte nuclei; and (iii) ROR␣ and cAspAT gene expression seems to be co-regulated by rosiglitazone in human adipose tissue. Thus, the data support our hypothesis that PPAR␥ is indirectly involved in cAspAT gene transcription in adipocytes by controlling the synthesis of ROR␣4 as a transactivator that is involved both in basal expression and in the response to rosiglitazone treatment.

DISCUSSION
In this work, we have provided evidence for the presence of a new partner of glyceroneogenesis in adipose tissue, cAspAT. The product of this enzyme (oxalacetate) is the substrate for the key enzyme of the pathway, PEPCK-C. But unlike PEPCK-C, cAspAT is an indirect transcriptional target of TZD. ROR␣ was identified as the possible direct target of this PPAR␥ ligand in adipocytes. FIGURE 8. ROR␣ specifically binds to the ؊404/؊374-bp region of the cAspAT gene promoter. A, radiolabeled Ϫ404/Ϫ374 bp cAspAT oligonucleotides were incubated with nuclear extracts prepared from D10 post-confluent 3T3-F442A adipocytes that had been treated with 1 M rosiglitazone for the last 18 h. For supershift analysis, the anti-PPAR␥ or anti-ROR␣ antibody (Ab) was added to the reaction 5 min before the labeled probes. ACOX, acyl-CoA oxidase. B, specificity of hROR␣1 binding to the Ϫ404/Ϫ374 bp region of the cAspAT gene promoter. Double-stranded oligonucleotides corresponding to the Ϫ404/Ϫ374 bp region of the cAspAT gene promoter and to the human fibrinogen-␤ ROR␣-responsive element (FgRORE) were labeled as described under "Experimental Procedures" and incubated with 1 l of in vitro translated hROR␣1 protein lysate before DNA⅐protein complexes were resolved by nondenaturing PAGE. For supershift analysis, the anti-PPAR␥ or anti-ROR␣ antibody was added to the reaction 5 min before the labeled probes. For competition (Comp.) studies, in vitro translated hROR␣1 protein lysate was incubated with 15-, 50-, and 100-fold molar excesses of the indicated unlabeled double-stranded oligonucleotides and labeled probes for 15 min at 4°C. mRORE, mutated human fibrinogen-␤ ROR␣-responsive element.
One major role of adipose tissue is to store energy as triglycerides, which are then lipolyzed to provide energy in the form of free FAs to the organism under fasting conditions. Hormonal control of the lipolysis rate via the hormone-sensitive lipase determines this output of FAs into the blood. We (2,3,44) and others (1,45) have recently highlighted the existence of a little known pathway, glyceroneogenesis, which is also increased under the same physiological conditions as the hormone-sensitive lipase, thus exerting a simultaneous control on FA homeostasis. Glyceroneogenesis slows down the release of lipolyzed FAs into the blood by allowing the re-esterification of 30 -60% of these FAs in adipocytes (3,14,45,46). Glyceroneogenesis has been found to be the major pathway involved in FA recycling in human adipose tissue (15). However, its efficiency, which is decreased in the adipose tissue of obese women, can be corrected by ex vivo rosiglitazone treatment (15). Pyruvate is generally used for testing glyceroneogenic capacities because it is a common intermediate in the metabolism of both lactate and alanine, the major physiological glyceroneogenic precursors (47). Aspartate is another intermediate specific for lactate metabolism (48,49). It has never been tested as a "potential" glyceroneogenic substrate. We have shown here that, in addition to pyruvate, aspartate is used for re-esterifying FAs in adipocytes, the involvement of which implies the participation of cAspAT in the production of glycerol-3-P. As described for PEPCK-C, the inhibition of substrate incorporation ([ 14 C]as-partate) into triglycerides by 3-mercaptopicolinic acid was observed, as well as an increase in cAspAT activity by TZD and a corresponding decrease in FA output from adipocytes in response to rosiglitazone treatment. Thus, our data highlight, for the first time, the functional contribution of this enzyme to the adipocyte glyceroneogenic pathway. Identification of cAspAT as a new pharmacological target for TZD suggests that cAspAT may contribute to the hypolipidemic action of the antidiabetic drug by increasing FA cycling in adipose tissue. We have confirmed that this control by TZD also occurs in human adipose tissue, although with interindividual variations.
Surprisingly, the molecular mechanism by which rosiglitazone increases cAspAT gene expression is different from the classical regulation of the PEPCK-C gene via a PPRE named PCK2 (13). Indeed, TZDs are ligands for the transcription factor PPAR␥, which binds DNA through a DR1 consensus sequence (PPRE). Whereas cAspAT mRNA levels were enhanced by two different PPAR␥ ligands in adipocytes derived from both white and brown tissue, this regulation occurred only at the end of the differentiation process, i.e. later than the PPAR␥ "plateau," and in a protein synthesis-dependent manner. The cAspAT 112-bp sequence identified as TZD-responsive contains no DR1 consensus sequence and only three RGGTCA hexanucleotide sequences (NRREs) that are conserved between rat and human promoters (31). Two of these sequences are organized as a DR2 sequence, recently shown to bind PPAR␥/ RXR␣ in the Rev-Erb␣ gene promoter (40). However, EMSA confirmed that there was no direct binding of PPAR␥ to the 112-bp fragment of the cAspAT gene (Figs. 5 and 8).
A similar situation was reported in mouse liver, where fibrates down-regulate the expression of the cAspAT gene in a PPAR␣-dependent way, but without any direct binding of PPAR␣ to its promoter (31,50). In our work, the TZD-responsive region (Ϫ398 to Ϫ286 bp) that we delineated in adipocytes was exactly the same as that mediating the response to PPAR␣ ligands in hepatocytes and did not bind PPAR␥. Nevertheless, the indirect involvement of PPAR␥ in TZD-responsive cAspAT gene expression was suggested by several criteria: 1) the fact that rosiglitazone is already active at a concentration of 10 nM, 2) the timing of the TZD responsiveness during the differentiation process from preadipocytes to adipocytes, 3) the selectivity of the transcriptional response to PPAR␥ ligands but not to RXR␣ or PPAR␣ ligands, and 4) the loss of TZD responsiveness in the Ϫ398- cAspAT construct in a PPAR␥-non-expressing cell line (3T3-C2). PPAR␥ implication was definitively confirmed in adipocytes in which a negative transdominant form of PPAR␥ was ectopically expressed, leading both to the decreased basal expression of the Ϫ398-cAspAT construct and to the complete abolition of its response to TZD. Thus, our results suggest that PPAR␥ is necessary, but not sufficient, for the TZD responsiveness of cAspAT gene expression in adipocytes. Indeed, PPAR␥ did not seem to affect cAspAT expression as much as PEPCK-C gene expression because ectopic expression of PPAR␥ had only a minor effect on cAspAT expression but significantly increased PEPCK-C gene transcription in differentiated adipocytes. In the same way, cAspAT gene expression did not vary in parallel with PPAR␥ during the adipogenic process, as direct PPAR␥ target genes do, such as aP2 and PEPCK-C. Thus, our data support the idea of cAspAT being an indirect target of TZD via PPAR␥.
To more precisely analyze this atypical PPAR␥-dependent way of regulation, we tried to more precisely define the DNA-binding site involved. In the liver, the mutation of each of the three NRREs (Mut1, Mut2, and Mut3) has been described to counteract the functional response of the cAspAT gene to fibrates, and more particularly, one of them (Mut3) completely impairs the binding of an unidentified protein to the Ϫ404/Ϫ374 bp promoter region (31). We made the assumption that the same region of the cAspAT promoter is involved in the rosiglitazone responsiveness in adipose tissue. We tested the functional involvement of each of these NRREs by point mutations and identified a unique AGGTCA sequence located between Ϫ381 and Ϫ376 bp (altered in Mut3) as necessary for the transcriptional response of cAspAT to rosiglitazone in differentiated adipocytes. We also showed that a shorter TZD-responsive region (Ϫ404 to Ϫ374 bp) did not bind the same protein complexes in adipocytes as in preadipocytes and that mutation of the AGGTCA sequence located between Ϫ381 and Ϫ376 bp (Mut3) impaired the binding of such an adipocyte-specific nuclear factor.
The AGGTCA sequence binds monomers of members of the ROR and Rev-Erb␣ orphan nuclear receptor family (51). Rev-Erb␣ acts as a negative regulator of transcription and can also bind to DNA via DR2 sequences as a homodimer (52). Considering that the members of this orphan nuclear receptor family accumulate during the adipocyte differentiation process (40,41,43), we hypothesized that ROR␣ or ROR␥ was a potential candidate for mediating cAspAT gene transcription. In fact, by EMSA and using in vitro transcribed/translated hROR␣1, we identified ROR␣ as one of the proteins that specifically bind to the Ϫ404/Ϫ374 bp region of the cAspAT gene promoter, but not to our mutated probe Mut3. Furthermore, we demonstrated that the ectopic overexpression of ROR␣4 in 3T3-C2 cells specifically increases transcription of the Ϫ398-cAspAT construct independently of rosiglitazone, as expected in these PPAR␥-deficient cells. Thus, our data indicate that the cAspAT gene is a new target gene of the ROR␣ nuclear receptor family.
We next tried to determine how ROR␣ could be involved in mediating the TZD responsiveness of the cAspAT gene in adi-pocytes. It had already been shown that the expression of members of this orphan nuclear receptor family can be increased by PPAR␥ overexpression or by PPAR␥ ligands in adipocytes (38,39,53). Our data confirmed the increase in the ROR␣4 protein level in nuclear extracts from rosiglitazone-treated versus untreated adipocytes. We also observed that a parallel increase in the amounts of ROR␣ and cAspAT mRNAs was achieved by ex vivo treatment of human adipose tissue samples with rosiglitazone.
Taken together, these data strongly suggest that ROR␣ could be involved in mediating the TZD responsiveness of the cAspAT gene in adipocytes and thus would explain the paradoxical protein synthesis dependence of this PPAR␥-dependent regulation. Treatments that increase intracellular calcium concentrations (such as ionomycin) have been reported to enhance the transcription of proteins belonging to the ROR family (54). Thus, our previous observation of increased cAspAT expression by the calcium ionophore A 23187 in adipocytes (55) still argues for the possible involvement of the ROR family in controlling transcription of the cAspAT gene in adipocytes.
In conclusion, this study describes, for the first time, the involvement of cAspAT in glyceroneogenesis, a pathway recently recognized as important in adipocytes for the regulation of FA levels in blood. We have shown that, similarly to the key enzyme of glyceroneogenesis, PEPCK-C, transcription of the cAspAT gene is increased by treatment of cells with rosiglitazone, an antidiabetic drug that decreases plasma FAs. Whereas the direct involvement of PPAR␥ in this regulation was ruled out by EMSA experiments, PPAR␥ was shown to be necessary in both basal and rosiglitazoneinduced cAspAT gene transcription in adipocytes. By elucidating the TZD-responsive element as a unique AGGTCA sequence that can specifically bind ROR and then by showing that ROR␣ (the amount of which is increased by rosiglitazone) is capable of transactivating cAspAT, we suggest that ROR␣ gene transcription could be a direct target of PPAR␥, although its promoter has not been well characterized yet. These data have provided new insights into the complex mechanisms of PPAR␥ interactions with other nuclear transcription factors, which have already been described in adipocytes: cooperation with upstream stimulatory factor and Sp1 for Ucp2 (56) and hormone-sensitive lipase (57) gene transcription, respectively, and activation of liver X receptor transcription (58). Ultimately, we may hypothesize that pharmacological ligands to ROR might exhibit the favorable effects of PPAR␥ agonists and none (or at least fewer) of their adverse effects.