Adipose Tissue Integrity as a Prerequisite for Systemic Energy Balance

Peroxisome proliferator-activated receptor γ (PPARγ) is an essential regulator of adipocyte differentiation, maintenance, and survival. Deregulations of its functions are associated with metabolic diseases. We show here that deletion of one PPARγ allele not only affected lipid storage but, more surprisingly, also the expression of genes involved in glucose uptake and utilization, the pentose phosphate pathway, fatty acid synthesis, lipolysis, and glycerol export as well as in IR/IGF-1 signaling. These deregulations led to reduced circulating adiponectin levels and an energy crisis in the WAT, reflected in a decrease to nearly half of its intracellular ATP content. In addition, there was a decrease in the metabolic rate and physical activity of the PPARγ+/- mice, which was abolished by thiazolidinedione treatment, thereby linking regulation of the metabolic rate and physical activity to PPARγ. It is likely that the PPARγ+/- phenotype was due to the observed WAT dysfunction, since the gene expression profiles associated with metabolic pathways were not affected either in the liver or the skeletal muscle. These findings highlight novel roles of PPARγ in the adipose tissue and underscore the multifaceted action of this receptor in the functional fine tuning of a tissue that is crucial for maintaining the organism in good health.

White adipose tissue (WAT) 2 plays a dual role in regulating energy homeostasis (1). First, it is a tissue that responds to nutrient intake by storing excess energy in the form of triglycerides (TG) and to metabolic demands associated with fasting or exercise by releasing the stored TG as free fatty acids (FFAs) and glycerol (2). Second, WAT is an endocrine organ in addition to its energy reserve functions. In fact, it integrates meta-bolic signals and secretes molecules, called adipokines, which in turn impact on multiple target organs, such as the liver, muscle, or brain. Therefore, it contributes significantly to the control of whole body energy homeostasis (3)(4)(5).
Deregulation of WAT functions in obesity or lipodystrophy is often linked to metabolic disorders, such as dyslipidemia, atherosclerosis, hypertension, insulin resistance, glucose intolerance, and prothrombotic and proinflammatory states (6 -10). Thus, WAT functional integrity is required for the balanced body metabolism of a healthy organism.
PPAR␥ (NR1C3) is highly expressed in the WAT, where it plays an important role in adipogenesis and in lipid metabolism (11)(12)(13). Suppression of PPAR␥ expression in preadipocytes impairs their differentiation (14,15). Furthermore, specific deletion of PPAR␥ in mature adipocytes causes their death, accompanied by macrophage infiltration in the affected WAT (16). In humans, heterozygous PPAR␥ mutations are responsible for partial lipodystrophy, severe insulin resistance, steatosis, and hypertension (17)(18)(19)(20)(21). The activation of PPAR␥ improves insulin sensitivity in both humans and mice. Agonists of PPAR␥, such as the thiazolidinedione Pioglitazone, are used clinically and are effective in reducing hyperglycemia, hyperinsulinemia, and hyperlipidemia in patients suffering from type 2 diabetes (22)(23)(24). Together, these facts underline the functions of PPAR␥ in adipocyte differentiation and survival and underscore its role in WAT integrity and whole body homeostasis.
Although the homozygous deletion of PPAR␥ in a mouse model was shown to be embryonic lethal, the survival of PPAR␥ Ϫ/Ϫ mice by inactivation of PP〈R␥ in all tissues except the trophoblasts was successful. These animals suffered from lipodystrophy, insulin resistance, and hypotension (25). However, deletion of only one PPAR␥ allele had some intriguing effects (15,26). In fact, PPAR␥ ϩ/Ϫ mice were resistant to obesity induced by a high fat diet (HFD) and, under these conditions, remained more sensitive to insulin then their WT counterparts (27). Decreased PPAR␥ activity under HFD conditions had a positive outcome on the development of obesity and diabetes. Based on these observations, a novel approach in type 2 diabetes therapy would include the use of PPAR␥ antagonists, potentially with fewer side effects compared with the present day synthetic agonists (thiazolidinediones).
Taking advantage of our PPAR␥ ϩ/Ϫ mouse model, we aimed at understanding how deletion of one allele of PPAR␥, which significantly reduces the activity of the receptor via a gene dosage effect, would affect WAT function and whole body metabolism, when the mice are fed with a standard diet (SD), a condition which does not exacerbate the lipid storage function of the WAT. The results reported herein show that deletion of one PPAR␥ allele affects specifically the expression of genes associated with metabolic pathways in the WAT. In addition to genes involved in lipid storage, genes involved in glycolysis, de novo fatty acid synthesis, and lipolysis were also down-regulated in the PPAR␥ heterozygous mice, creating a strong energy deficit in these animals. These defects in WAT functions correlated with a lowering of the metabolic rate of the whole body and were accompanied by a reduction in physical activity. These results cast doubt on a potential long term use of PPAR␥ antagonists for the treatment of type 2 diabetes.

EXPERIMENTAL PROCEDURES
In Vivo Animal Study-WT and PPAR␥ ϩ/Ϫ male mice, of a mixed background Sv129/C56BL/6, were maintained at 23°C on a 12-h light-dark cycle. The animals studied were between 10 and 12 weeks of age. They had free access to water and to an SD, except during fasting, when they had free access to water only, food being withdrawn for 24 h. In some experiments, 5-6week-old animals were fed with an SD containing 0.004% of Pioglitazone (w/w) for 5 weeks. The Pioglitazone treatment protocol was adapted from Ref. 15, a study that involved PPAR␥ ϩ/Ϫ animals too. Pioglitazone was kindly provided by Takeda Chemical Industries (Switzerland). The standard food pellets containing the Pioglitazone as well as the control pellets were produced by Provimi-Kliba (Switzerland). For analysis, the animals were killed in the morning between 9 and 11 a.m. by cervical dislocation, and tissues were rapidly frozen in liquid nitrogen. The animal experimentation protocols were approved by the Commission de Surveillance de l'Expérimention Animale of the Canton de Vaud (Switzerland).
RNA Preparation-The RNA from epidydymal WAT, gastrocnemius skeletal muscle, and liver was extracted from the frozen tissues using the Trizol reagent (Invitrogen) according to the manufacturer's instructions. The RNA for microarray analyses was further purified using Qiagen RNeasy columns (Qiagen). The RNA quality was assessed by capillary electrophoresis on a 2100 Bioanalyzer (Agilent Technologies).
Microarray Experiment and Data Processing-To minimize interindividual variation due to the mixed background of the mouse strain, each PPAR␥ ϩ/Ϫ animal had a WT counterpart coming from the same litter.
Three independent sets of total RNA samples (three WT and three PPAR␥ ϩ/Ϫ animals) from epidydymal WAT and gastrocnemius skeletal muscle were isolated. cRNA was synthesized from 5 g of total RNA, according to Ref. 71. After purification using a Qiagen RNeasy column, aliquots of 20 g of cRNA were fragmented. Each fragmented cRNA (15 g) was then hybridized to an Affymetrix "Mouse Genome 430 2.0 Array" Gene-Chip microarray. Hybridization, washing, and scanning were according to Affymetrix instructions.
Data from the scanned chips were analyzed using the Affymetrix MAS 5.0 software (28,29). To identify differentially expressed transcripts, pairwise comparison analyses were car-ried out. Each experimental sample was compared with each reference sample, resulting in nine pairwise comparisons. Transcripts were considered to be differentially expressed if their levels changed in the same direction in seven of nine comparisons. Further data filtering and analyses were performed with the Genespring (Agilent) and the Ingenuity Pathway Analysis 4.0 software.
Quantitative RT-PCR-Single-stranded cDNA templates for quantitative real time (qRT)-PCR analysis were synthesized using Superscript II reverse transcriptase and random priming, starting from the same RNAs used for the microarray analysis, and from additional independent experiments as described above. Amplicons were designed using the Primer Express software (Applied Biosystems), and their sequences were checked by BLAST against the mouse genome to ensure that they were specific for the gene being assayed. The efficiency of each primer pair was tested in a cDNA dilution series. The list of primers is available on demand.
Real time PCR was carried out in optical 384-well plates and labeled by using the SYBR green master mix (Applied Biosystems), and the fluorescence was quantified with a 7900HT SDS system (Applied Biosystems). The relative expression level of target genes was normalized according to geNorm, using ␤actin, tubulin ␣2, and hypoxanthine guanine phosphoribosyltransferase 1 as references to determine the normalization factor (30). Fold changes were calculated from the ratio of means of the normalized quantities and their statistical significance was determined by a paired Student's t test.
ATP Level Measurements-Frozen WAT homogenate was transferred into a plastic tube containing 6% HClO 4 . Following centrifugation, the supernatant was recovered and neutralized with 5.5 M KOH. The ATP concentration was measured with an ATP determination kit, a time-stable assay from Biaffin GmbH&Co KG (Germany). The kit allows quantitative determination of small amounts of ATP by a bioluminescence assay involving the oxidation of the firefly luciferase depending on the ATP present in the extracts. The ATP concentration was derived according to the manufacturer's instructions.
Glycerol Level Measurements-The glycerol content was measured with a glycerol measuring kit (Randox). Briefly, the glycerol present in the samples was converted into a colored product measured at a wavelength of 520 nm. The glycerol concentration was then determined according to the manufacturer's instructions.
Metabolic Measurements-Metabolic cage studies were conducted in a comprehensive laboratory animal monitoring System (8-chamber CLAMS system; Columbus Instruments, Columbus, OH). The mice were adapted to powdered food for 24 h before they were introduced into the metabolic cages, where a 48-h acclimation preceded the 24-h recording time. Information was collected on the metabolic activity, food intake, water drinking, and physical activity.
Blood was collected from the orbital sinus between 9:00 and 11:00 a.m., using heparinized microcapillary tubes and immediately centrifuged. The serum fraction was frozen immediately. Depending on the experiment, the animals were either normally fed or fasted for 24 h.
The plasma concentrations of TG, free fatty acids (FFAs), glycerol, and ketone bodies were measured at the Mouse Clinic Institute (ICS; Strasbourg, France) on a Olympus AU-400 automated laboratory work station (Olympus-SA France) using commercial reagents (Olympus Diagnostica GmbH, Lismeehan, Ireland).
The plasma leptin and adiponectin concentrations were measured using the mouse leptin enzymelinked immunosorbent assay kit and the mouse adiponectin enzymelinked immunosorbent assay (Linco Reserach).
The plasma glucose levels were measured with an Accu-Chek Sensor glucometer (Roche Applied Science), and the plasma insulin concentrations were measured with an Ultra mouse insulin enzyme-linked immunosorbent assay kit (Mercodia SA).

RESULTS
Decreased Metabolic Rate in PPAR␥ ϩ/Ϫ Mice-In PPAR␥ ϩ/Ϫ animals, PPAR␥ mRNA and protein (PPAR␥1 and PPAR␥2) levels were reduced by half compared with those of WT mice (31). This prompted us to explore the impact of this reduced PPAR␥ expression on whole body metabolism in the absence of any excess energy challenge, as is usually done with HFD feeding in assessing the role of PPAR␥ in lipid storage. Instead, the PPAR␥ ϩ/Ϫ mice were fed with an SD. Metabolic parameters of the PPAR␥ ϩ/Ϫ mice and their WT littermates were determined using metabolic cages. As expected, both mutated and WT animals consumed more O 2 and produced more CO 2 during the dark cycle, when they are generally more active (Fig. 1A, left). Although the PPAR␥ ϩ/Ϫ mice had a similar weight (Table S1) and ate an equal amount of food (data not shown), they consumed less oxygen and produced less CO 2 during both the light and dark cycles when compared with their WT counterparts, a difference reflected in a decrease of 14% in the metabolic rate (heat production) of PPAR␥ ϩ/Ϫ animals (Fig. 1A). This effect was clearly PPAR␥-dependent, since a 5-week treatment with SD containing the PPAR␥ agonist Pioglitazone, at 0.004%, alleviated the metabolic rate difference between the two genotypes ( Fig. 1A, right). Moreover, there was a trend, not statistically significant, for increased O 2 consumption, CO 2 production, and a higher metabolic rate in Pioglitazone-treated PPAR␥ ϩ/Ϫ mice, whereas such a tendency was not observed in WT animals.
Deletion of One PPAR␥ Allele Does Not Affect the Carbohydrate to Lipid Ratio in Metabolic Fuel Utilization-To determine whether the decrease in the metabolic rate of PPAR␥ ϩ/Ϫ animals was associated with alterations in the use of carbohydrates versus lipids as fuel molecules, we calculated the FIGURE 1. Metabolic rate, fuel consumption, and total physical activity in PPAR␥ ؉/؊ and control mice. A, oxygen (O 2 ) consumption, carbon dioxide (CO 2 ) production, and metabolic rate in the PPAR␥ ϩ/Ϫ and control mice. O 2 and CO 2 were measured by indirect calorimetry and expressed as average VO 2 and VCO 2 /kg of body weight/h during a 24-h monitoring session (light/dark). Metabolic rate (heat) is calculated from the oxygen production and the respiratory exchange ratio (RER) and is expressed as average kcal/h/kg of body weight during a 24-h monitoring session (light/dark). B, the fuel consumption (RER) is the ratio of CO 2 produced to the amount of O 2 consumed and serves as a guide of the fuel type consumption (carbohydrate (RER ϭ 1.0) or fat (RER ϭ 0.7)). C, the total physical activity was measured as the horizontal and rearing movements during the 24-h monitoring period (total of light and dark movements). The activity is expressed as the average number of times a mouse crosses both the x and y axes at least twice. n ϭ 12 (CTL experiments); n ϭ 5 (Pioglitazone experiments). Values are expressed as mean Ϯ S.E.; *, p Յ 0.05. respiratory exchange ratio (RER). The RER (equal to VO 2 / VCO 2 ) indicates whether lipids (RER ϭ 0.7) or carbohydrates (RER ϭ 1.0) are being oxidized to produce energy. Both genotypes consumed carbohydrates as the main energy source (Fig.  1B, left). This result disagrees with the notion of a metabolic compensation through increased fat oxidation in PPAR␥ ϩ/Ϫ mice (9). Moreover, the Pioglitazone treatment had no significant effect on the choice of fuel type (Fig. 1B, right).
Decreased Physical Activity in PPAR␥ ϩ/Ϫ Mice-Since a decreased metabolic rate in the mutated animals may correlate with a change in behavior, we measured their physical activity (horizontal and rearing movements). Interestingly, the PPAR␥ ϩ/Ϫ mice presented a 23% decrease in total activity (Fig.  1C, left). This phenotype correlated with a strong decrease in the plasma adiponectin concentration in PPAR␥ ϩ/Ϫ mice, whereas the leptin level remained unchanged (Figs. 2A and S1). This observation is in agreement with the reduced spontaneous motor activity of transgenic mice overexpressing an antisense adiponectin RNA, resulting in decreased circulating adiponectin levels (32). The Pioglitazone treatment corrected this decrease in physical activity, suggesting an implication of PPAR␥ (Fig. 1C, right). In brief, reduced PPAR␥ levels decreased the metabolic rate and the physical activity of mice without changing their fuel preference.
PPAR␥ ϩ/Ϫ and WT Mice Have Similar Plasma Insulin and Glucose Profiles-Since a decreased metabolic rate might impact on glucose and lipid homeostasis, we analyzed the plasma profile of the WT and PPAR␥ ϩ/Ϫ mice. The plasma insulin concentration was normal in unchallenged animals and was decreased after a 24-h fast as expected, but no difference was observed between WT and mutant mice (Fig. 2B). Moreover, the glycemia was also normal in PPAR␥ ϩ/Ϫ mice, which however had a significantly attenuated response to fasting (Fig. 2C). In fact, the fasting glycemia was higher in the PPAR␥ ϩ/Ϫ mice compared with that of the WT animals. Fasting for 24 h decreased the glucose level by 45% in WT mice, whereas it was decreased by only 30% in the PPAR␥ ϩ/Ϫ animals. Thus, after fasting, PPAR␥ ϩ/Ϫ mice presented a less pronounced hypoglycemia.
The Plasma Lipid Profile of PPAR␥ ϩ/Ϫ Mice Reveals an Alteration in Lipolytic Activity-In fed conditions, the plasma FFA concentrations were normal, and no deregulation was observed in PPAR␥ ϩ/Ϫ mice (Fig. 2D). WT animals responded normally to fasting by liberating FFAs from the WAT into the circulation, thus increasing the plasma FFA concentration. Remarkably, no significant increase was observed in the PPAR␥ ϩ/Ϫ mice, suggesting a deregulation of the lipolytic activity of the PPAR␥ ϩ/Ϫ WAT. This defect was confirmed by measuring the circulating glycerol concentration. As for the FFAs, there was no difference in the fed glycerol concentration between WT and PPAR␥ ϩ/Ϫ mice (Fig. 2E). However, the PPAR␥ ϩ/Ϫ mice responded less well to fasting, since they increased their plasma glycerol concentration by only 32%, compared with the 62% monitored in WT mice. Given that the fasting glycerol and FFA concentrations are indicators of the lipolytic activity in the WAT, we concluded that PPAR␥ ϩ/Ϫ mice might have a decreased lipolytic activity. This alteration should also be detectable in the WAT itself, in which the total glycerol (glycerol ϩ glycerol-3-P) originates from glycolysis, glyceroneogenesis, and lipolysis. There was a 23% decrease in total glycerol content of the PPAR␥ ϩ/Ϫ WAT, suggesting that at least one of the three above functions or all of them were impaired (Fig. 3).
Reduced circulating FFA levels should have consequences for ketone body synthesis in the liver, which depends on FFA availability. We measured the ketone body concentrations after fasting in both WT and PPAR␥ ϩ/Ϫ mice (Fig. 2F). PPAR␥ ϩ/Ϫ mice were less efficient in producing ketone bodies, since their plasma concentration of this peripheral organ fuel was 33% lower than in WT animals. Thus, this decreased supply in ketone bodies might reflect the reduced availability of FFAs in PPAR␥ ϩ/Ϫ mice. The TG concentrations were increased by 38% after fasting in WT mice, which reflects the recycling to TG-very low density lipoprotein by the liver of a portion of the FFA liberated by the WAT during fasting (Fig. 2G). In agree-  OCTOBER 12, 2007 • VOLUME 282 • NUMBER 41 ment with the observations reported above, the plasma TG concentration was not increased in PPAR␥ ϩ/Ϫ mice, in contrast to that measured in WT animals. The reason why fed PPAR␥ ϩ/Ϫ animals also presented reduced ketone body levels remains to be elucidated (Fig. 2F).

PPAR␥ Controls WAT Integrity and Metabolic Rate
Based on the results described so far, we hypothesized that the PPAR␥ ϩ/Ϫ mice decreased their metabolic rate and their physical activity to adapt to a diminished energy supply. Three organs, the liver, skeletal muscle, and WAT, are primarily involved in energy supply and consumption. Deletion of one PPAR␥ allele might have affected the expression pattern of PPAR␥ target genes. This possibility was tested by assessing the expression of genes involved in metabolic pathways of the three key organs mentioned above by microarray analysis (WAT and skeletal muscle) and/or qRT-PCR (WAT and liver). The expression of genes not represented in the microarray was analyzed by qRT-PCR.
The Expression of Metabolic Genes Is Not Affected in the Liver of PPAR␥ ϩ/Ϫ Mice-The liver is one of the major organs responsible for whole body energy balance. PPAR␥ is expressed at low levels in the liver under normal conditions but is increased in steatosis induced by HFD or other pathophysiological conditions (11)(12)(13). The expression pattern of several metabolic genes, which are known to be transcriptionally regulated by PPARs, was tested by qRT-PCR. We found that the expression in the liver of the metabolic genes listed in Table 1 was not affected in the PPAR␥ ϩ/Ϫ animals fed an SD. In fact, genes involved in the glycolytic or FA oxidation pathways, such as very long chain acyl-CoA dehydrogenase, long chain acyl-CoA dehydrogenase, and carnitine palmitoyltransferase 1, were expressed at comparable levels in PPAR␥ ϩ/Ϫ and WT mice. Similarly, the gluconeogenic pathway was not affected, even during fasting (Fig. S2). In particular, known PPAR target genes, such as those encoding phosphoenolpyruvate carboxykinase, glycerol-3phosphate dehydrogenase 1, glycerol kinase, aquaporin 3, and glucose-6-phosphatase, were not deregulated. Since PPAR␣ is a regulator of these genes in liver, these results suggest that the PPAR␥ insufficiency did not induce a compensatory activation of PPAR␣. In addition, these results are in agreement with the RER measurements (Fig. 1B, left), which showed no fuel source switch between PPAR␥ ϩ/Ϫ and WT mice. We concluded that the liver of PPAR␥ ϩ/Ϫ mice on SD is not the organ primarily responsible for the deficiency in energy supply, the lowering of the metabolic rate, and the reduction in motor activity of these animals.

The Expression of Metabolic Genes Is Not Affected in Skeletal
Muscle of PPAR␥ ϩ/Ϫ Mice-The skeletal muscle is a key site of energy consumption. PPAR␥ is weakly expressed in this organ, and its selective deletion in skeletal muscle can affect insulin sensitivity of the liver (33) as well as muscle and WAT insulin sensitivity (34).
In our study, deletion of one PPAR␥ allele did not affect the expression of metabolic genes in the gastrocnemius skeletal muscle. Large scale analysis of PPAR␥ ϩ/Ϫ mice and their WT counterparts on SD showed that of the 39,000 transcripts analyzed, only a few were deregulated in the PPAR␥ ϩ/Ϫ mice (data not shown), and none of those was among the metabolic genes listed in Table 1. Thus, it is worth noting that, at least at the mRNA expression level, there was no deregulation of the expression of glycogenolysis, glycolysis, and FA oxidation in the PPAR␥ ϩ/Ϫ mice. In conclusion, the skeletal muscle is appar-ently not responsible for the energy deficit observed in the PPAR␥ ϩ/Ϫ mice, which argues against a significant role of PPAR␥ in this tissue under SD conditions.
The Expression of Metabolic Genes Is Affected in PPAR␥ ϩ/Ϫ WAT-Since PPAR␥ is expressed at high levels in WAT, it was of interest to analyze the transcriptional profile of the PPAR␥ ϩ/Ϫ and WT WAT. The majority of genes deregulated in PPAR␥ ϩ/Ϫ WAT were classified into two major categories. First, genes involved in cell stress, detoxification, and inflammatory pathways were up-regulated; second, genes involved in metabolic and energy production pathways were down-regulated (Fig. 4). The present study concentrated on the second category only. Taken together, these results validated our hypothesis of a marked metabolic deregulation and energy production impairment in the PPAR␥ ϩ/Ϫ WAT.
The Insulin and IGF-1 Signaling Pathways Are Down-regulated in the PPAR␥ ϩ/Ϫ WAT-The microarray analysis revealed that the insulin as well as the insulin growth factor 1 (IGF-1) signaling pathways were down-regulated in PPAR␥ ϩ/Ϫ WAT. Insulin is a key regulator of fuel metabolism, whereas IGF-1 is mostly involved in cell survival, growth, and differentiation. However, much remains to be clarified, since IGF-1 was also shown to control, at least in part, the expression of metabolic enzymes (35,36). IGF-1 expression was down-regulated by 46% in the PPAR␥ ϩ/Ϫ WAT, a result that is in agreement with our previous work ( Table 2) (31). The present analysis went further by showing that not only IGF-1 expression was deregulated in PPAR␥ ϩ/Ϫ WAT but also several other genes in the activation cascade downstream from IGF-1, such as c-Jun and 3-monooxygenase tryptophan 5-monooxygenase activation protein ␥ (14-3-3-␥), along with the insulin-like growth factor-binding protein, which were down-regulated by 39, 36,

TABLE 1 Metabolic pathways investigated in skeletal muscle and liver isolated from fed animals
A search for differences in gene expression levels between PPAR␥ ϩ/Ϫ and control mice was performed either by microarray in the skeletal muscle or by qRT-PCR in liver. ϫ identifies the genes analyzed. The number of animals tested was as follows: skeletal muscle, n ϭ 3 for each genotype; liver, n ϭ 6 for each genotype. and 46%, respectively (Table 2). Thus, processes activated by IGF-1 were affected in the mutated WAT, namely adipocyte differentiation and survival and stimulation of glucose and lipid uptake (37)(38)(39)(40). One major actor in the insulin pathway is the insulin receptor substrate 1; its mRNA was down-regulated by 44% in the PPAR␥ ϩ/Ϫ WAT (Table 2). Furthermore, the mRNAs of the eukaryotic translation initiation factor 4E-binding protein 1, phosphodiesterase 3B, and Rap guanine nucleotide exchange factor 1 belonging to the same pathway were also down-regulated by 41, 40, and 34%, respectively. These results extend those of a previous study, which showed that the expression of phosphodiesterase 3B as well as its insulin-induced activation were decreased in the PPAR␥ ϩ/Ϫ WAT (41). On the contrary, an inhibitor of the insulin pathway, called protein-tyrosine phosphatase receptor-type F was up-regulated by 2-fold in the PPAR␥ ϩ/Ϫ WAT (Table 2).

GenBank
Glucose Uptake and Glycolysis Are Decreased in the PPAR␥ ϩ/Ϫ WAT-Gene expression changes identified by microarray and/or assessed by qRT-PCR indicated that glucose uptake, glycolysis, and its associated lipogenesis, which are controlled by insulin, were deregulated in the PPAR␥ ϩ/Ϫ WAT ( Table 2 and Fig. 5). The expression of the insulin-independent glucose transporter 1 was decreased by 65%, suggesting a decreased basal glucose entry into the adipocytes (Fig. 5). This effect was corrected by Pioglitazone treatment, showing the involvement of PPAR␥ in this deregulation (Fig S3). In contrast, the expression of glucose transporter 4 was not affected.
Once taken up by the adipocytes, glucose is processed via three main pathways. In brief, 35% of the glucose is used for TG synthesis, 50% is used for lactate production, and finally 15% is used for ATP production that provides the energy for adipocyte maintenance and survival (42). One of the main enzymes of the glycolytic pathway, hexokinase 2, was down-regulated by 40% in PPAR␥ ϩ/Ϫ WAT (Table 2 and Fig. 5). Its expression was also restored by Pioglitazone treatment (Fig S3). The level of this

TABLE 2 Genes affected in WAT by the deletion of one PPAR␥ allele
Their expression level reflects the decreased metabolic activity of the WAT (R*, fold change in heterozygous animals compared with WT mice, whose value is set to 1; threshold 1.3; validated result: reproducibility in seven of nine comparisons). RNA samples from three animals of each genotype were analyzed. enzyme is stimulated by insulin via sterol regulatory element-binding protein-1c (43, 44) (Fig. 6). In line with this effect, fasting decreases and refeeding restores hexokinase 2 levels in different tissues, including WAT (44). Therefore, the decreased expression of hexokinase 2 in the PPAR␥ ϩ/Ϫ WAT is consistent with impaired insulin signaling in the mutated animals. 6-Phosphofructo-2-kinase/fructose-2,6-bisphosphatase is also part of the glycolytic pathway. Four genes encode 6-Phosphofructo-2-kinase/fructose-2,6-bisphosphatase (PFKFB1 to -4), and it is PFKFB3 that was identified as the isoform that is probably responsible for the activation of glycolysis in adipocytes. PFKFB3 was down-regulated by 40% in the PPAR␥ ϩ/Ϫ WAT (Table 2). Data so far available support a role of this enzyme in TG synthesis in adipocytes (45).

GenBank
Based on these combined results, we conclude that glucose uptake, as well as the glycolytic pathway, are defective in the PPAR␥ ϩ/Ϫ WAT (Fig. 6), and this alteration is linked to the decreased activity of PPAR␥.
De Novo Fatty Acid Synthesis Is Down-regulated in the PPAR␥ ϩ/Ϫ WAT-Acetyl-CoA, produced by catabolic pathways, is used for FA synthesis (Fig. 6). Pyruvate carboxylase, whose activity increases during adipocyte differentiation, is thought to have a lipogenic function by providing amounts of acetyl groups and NADPH for fatty acid synthesis. Its gene is a direct PPAR␥ target in adipose tissue (46). Consistent with this finding, expression of the pyruvate carboxylase gene was decreased by 32% in the PPAR␥ ϩ/Ϫ WAT. An even more pronounced down-regulation was observed for acetyl-CoA carboxylase, which carboxylates acetyl-CoA to malonyl-CoA, and for fatty acid synthase, which produces palmitate from malonyl-CoA. Their levels were reduced by 47 and 39%, respectively (Table 2 and Fig. 5), and their expression was restored by Pioglitazone treatment (Fig. S3). Interestingly and in line with our findings, the expression of these enzymes was previously shown to be increased by the PPAR␥ agonist GW1929 (47).
In addition to carbon supply (acetyl-CoA), the de novo production of fatty acids requires reducing power, in the form of NADPH. NADPH is generated by the oxidation of glucose 6-phosphate via an alternative pathway to glycolysis, the pentose phosphate pathway (Fig. 6). Treatment of 3T3L1 adipocytes with the PPAR␥ agonist rosiglitazone was shown to upregulate the expression of enzymes involved in this pathway (48). It is regulated mainly by the activation of glucose-1-dehydrogenase, a deficiency of which results in a decrease of NADPH production (48,49). Glucose-1-dehydrogenase expression was down-regulated by 41% in PPAR␥ ϩ/Ϫ WAT. Transketolase is another enzyme linking the pentose phosphate pathway to the glycolytic pathway. Its expression was downregulated by 49% in PPAR␥ ϩ/Ϫ WAT ( Table 2). Transketolase expression was shown to be decreased in insulin receptor knock-out mice, and transketolase ϩ/Ϫ mice suffer from growth retardation and preferential reduction of the WAT (42,50). Thus, any reduction in the expression of these two key enzymes reflects decreased FA synthesis in the PPAR␥ ϩ/Ϫ WAT (Fig. 6).

Deletion of PPAR␥ Decreases the Expression of Enzymes Involved in Lipid
Storage-It is thought that the major source of activated glycerol (glycerol-3-P) used by the WAT to esterify FA is produced by the glycolytic pathway during feeding and by the glyceroneogenetic pathway during fasting (51,52). The production of glycerol-3-P from gluconeogenic precursors, such as amino acids and lactate, is called glyceroneogenesis, and it is one of the most important pathways in the adipocytes (53)(54)(55). Glyceroneogenesis is controlled by two enzymes, phosphoenolpyruvate carboxykinase and glycerol-3-phosphate dehydrogenase 1. Phosphoenolpyruvate carboxykinase regulation was shown to be exclusively transcriptional, with a participation of PPAR␥ in WAT and PPAR␣ in the liver, through two peroxisome proliferator response element sites present in the promoter (56,57). Glycerol-3-phosphate dehydrogenase 1 was also shown to be a PPAR␥ target in WAT (58). The expression of these two enzymes was decreased in the PPAR␥ ϩ/Ϫ WAT, by 45 and 29%, respectively (Table 2 and Fig. 5). Although increasing their expression in both genotypes, Pioglitazone treatment was unable to completely alleviate the difference between PPAR␥ ϩ/Ϫ and WT mice, suggesting an additional regulation, which may depend on the promoter context (Fig. S3) (48).
Diacylglycerol acyltransferase 1 catalyzes the terminal and only committed step in triacylglycerol esterification by using activated glycerol (glycerol-3-P) and fatty acyl-CoA as substrates. Diacylglycerol acyltransferase 1 is necessary for adipose tissue formation and essential for its survival (59). Diacylglycerol acyltransferase 1 was showed to be up-regulated by the PPAR␥ agonist GW1929 (47) and was consistently down-regulated by 29% in PPAR␥ ϩ/Ϫ WAT.
The reduction in phosphoenolpyruvate carboxykinase, glycerol-3-phosphate dehydrogenase 1, and diacylglycerol acyltransferase 1 expression as well as the decrease in glycerol content measured in PPAR␥ ϩ/Ϫ WAT are in agreement with a reduced glyceroneogenesis and lipid accumulation in the

Reduced Expression of Genes
Encoding Lipolytic Enzymes in the PPAR␥ ϩ/Ϫ WAT-Fasting induces a breakdown to FFA and glycerol of the lipids stored in the WAT through lipolysis. These products are released into the circulation by FFA and glycerol transporters. The expression of proteins involved in lipolysis, such as the hormone-sensitive lipase, which catalyzes the rate-limiting step in adipose tissue lipolysis, and aquaporine adipose, which transports glycerol from adipocytes into circulation, were decreased in PPAR␥ ϩ/Ϫ WAT by 47 and 39%, respectively ( Table 2 and Figs. 5 and 6). The genes encoding these enzymes are known PPAR␥ target genes (58,60). Decreased lipolytic activity in PPAR␥ ϩ/Ϫ WAT might be a mechanism for sparing the TG stocks. Moreover, these results are in line with the observed plasma levels of FA and TG (Fig. 2,  C and F).
ATP Crisis in PPAR␥ ϩ/Ϫ WAT-The gene expression profile of PPAR␥ ϩ/Ϫ WAT showed a global decrease in metabolic pathways. This deficit affects the use of glucose for de novo fatty acid synthesis and their subsequent storage. Glucose is also used by the adipocytes as the source of ATP that is required for their survival (61)(62)(63). Since the glycolytic pathway is down-regulated in the PPAR␥ ϩ/Ϫ WAT, we suspected an ATP deficit in this organ. Moreover, our microarray profile showed that genes of cell stress were up-regulated in the PPAR␥ ϩ/Ϫ WAT (not shown), suggesting that an ATP deficit might be causing this cellular stress. Thus, we measured the ATP concentration in the WT and PPAR␥ ϩ/Ϫ WAT. As expected by the gene expression profiles reported herein, the ATP level was decreased by 45% in the PPAR␥ ϩ/Ϫ WAT (Fig.  7). This result is in agreement with a known PPAR␥ role in lymphocytes, where it attenuates the decline in ATP levels (64) and with the cell type-specific deletion of PPAR␥ in mature adipocytes, which leads to necrosis possibly by ATP depletion (16). Finally, enzymes involved in FA ␤-oxidation were not deregulated in the mutant WAT. This is in agreement with the minor role of this pathway in the adipose tissue and argues against compensation by increased fat utilization for the lack of energy due to decreased glycolytic activity. Clearly, the adipose tissue in PPAR␥ ϩ/Ϫ mice suffers from a significant energy deficit due to decreased glycolytic activity. FIGURE 6. Schematic representation of the metabolic pathways influenced by the deletion of one PPAR␥ allele in the WAT. The enzymes that were shown to be down-regulated in PPAR␥ ϩ/Ϫ WAT either on the microarray experiment or by qRT-PCR or both are shown in the scheme. The enzymes already described as being regulated by PPAR␥ or by its ligands are in red, and those shown to be regulated by the insulin pathway are in green.

DISCUSSION
Adipose tissue integrity is crucial for whole body energy homeostasis. Its functional deregulation is usually found associated with the metabolic syndrome. Among transcription factors, PPAR␥ occupies key functions in WAT, where it participates in adipocyte differentiation and maintenance, including promotion of lipid storage (65). In the absence of PPAR␥, preadipocytes do not differentiate into adipocytes, and its deletion in mature adipocytes leads to their death by necrosis (16).
We suspected that lipid storage is by far not the sole metabolic function of PPAR␥. However, in previous studies, challenging the PPAR␥ ϩ/Ϫ mice with an HFD most likely exacerbated the lipid storage function of PPAR␥, and therefore other key functions went unnoticed. For this very reason, we chose to study the metabolic role of PPAR␥ in the absence of any specific nutritional challenge by investigating young adult PPAR␥ ϩ/Ϫ male mice fed on an SD.
The plasma analyses showed that if insulin and glucose concentrations were normal in basal feeding conditions, there was an alteration of the glycemic control in the mutated mice during fasting. Furthermore, the plasma lipid profile of these animals, especially FFA and glycerol, suggested the possibility of an abnormal lipolysis. This presumption was strengthened by the fact that the plasma concentrations of ketone bodies, which are produced in the liver from FFA that are released by WAT during lipolysis, were low. Although this would designate the adipose tissue as the main culprit for these deregulations, we nevertheless extended our study to the liver and skeletal muscle. All three organs are either important consumers (skeletal muscle) or producers (WAT and liver) of energy. Furthermore, all three express PPAR␥, although at different levels, WAT presenting by far the highest levels of this receptor. Based on their gene expression profiles, we concluded that the liver and skeletal muscle in PPAR␥ϩ/Ϫ mice are most likely not involved in the metabolic deregulation of these animals when fed an SD. Of all of the genes tested that are involved in glycolysis and FA oxidation in both tissues, as well as in gluconeogenesis in the liver, none presented a modified expression in the PPAR␥ ϩ/Ϫ mice. These results are in agreement with liver and muscle gene expression profiles in fatless mice or mice with an organ-specific deletion of PPAR␥ in skeletal muscle. In these studies, no modification was observed in the expression of genes involved in ␤-oxidation and gluconeogenesis (10,33,34,66).
On the contrary, deletion of one PPAR␥ allele had unexpectedly profound effects on the expression of genes involved in WAT differentiation and energy production. The present study shows that genes involved in the IGF-1 signaling pathway, known to participate in adipocyte differentiation and growth (31,(37)(38)(39)(40), were down-regulated in PPAR␥ ϩ/Ϫ WAT. This explains the resistance to growth hormone action and the smaller adipocyte size already observed by us and others in the PPAR␥ ϩ/Ϫ WAT (15,31,67). In addition to the IGF-1 pathway, genes belonging to the insulin signaling pathway, which is an important activator of the glycolytic and lipogenic pathways in WAT, were also down-regulated.
Unexpected at first sight, but in agreement with a downregulation of the insulin pathway, basal glucose uptake and gly-colysis were decreased in the mutant WAT. Glycolysis is thought to play three major roles in WAT by (i) participating in the de novo synthesis of fatty acids, (ii) promoting their storage as fat, and finally (iii) producing ATP for adipocyte survival. First, the acetyl-CoA and NADPH synthesized by the glycolytic and pentose phosphate pathways, respectively, were decreased in the PPAR␥ ϩ/Ϫ WAT (Fig. 6). As an immediate consequence, de novo fatty acid synthesis was decreased, an effect reinforced by the down-regulation of enzymes involved in this process. Second, storage was altered, since glycolysis and glyceroneogenesis were affected in the PPAR␥ ϩ/Ϫ WAT. Third, glucose is required for adipocyte survival, and ATP production is an important feature of this process, deficiency of which affects the integrity of the cell. We observed a significant decrease of the ATP levels in the WAT of the mutant mice. This result is in agreement with previous data from cell culture experiments, where decreased ATP concentration in 3T3L1 or primary adipocytes interfered with insulin signaling and lipolysis. Glycolysis and FA oxidation are the major ATP producers, but no increase of ␤-oxidation was observed in the mutant WAT to compensate for the energy deficit. Moreover, genes involved in the lipolytic activity as well as in the glycerol release were downregulated in the PPAR␥ ϩ/Ϫ WAT. Thus, PPAR␥ ϩ/Ϫ WAT suffers from a generalized energy shortage probably due to the ATP crisis in the adipocytes, which we link to a deregulation of the glycolytic pathway. We have shown previously that tissuespecific ablation of PPAR␥ in adipose tissue causes the death, within a few days of the PPAR␥-deficient adipocytes (16). We observed a necrosis rather than apoptosis, which triggered an inflammatory response in the affected adipose tissue (16). Usually, necrosis is caused by a metabolic disruption and ATP depletion, whereas apoptosis requires ATP without a clearly defined point of no return (68). The phenotype of the PPAR␥ ϩ/Ϫ adipocytes suggests that what caused the death of PPAR␥ Ϫ/Ϫ adipocytes was in fact a metabolic breakdown associated with a loss of ATP.
To our knowledge and for the first time, the findings reported herein link PPAR␥ expression levels with the regulation of the systemic metabolic rate and physical activity. Although deletion of just one PPAR␥ allele decreased the metabolic rate and physical activity of the mice, we recorded no effect on the relative amounts of carbohydrates versus lipids used as fuel molecules. This observation is in agreement with the absence of deregulation in the expression of enzymes involved in FA oxidation. Evidence that this systemic effect is PPAR␥-specific comes from the Pioglitazone treatment of the PPAR␥ ϩ/Ϫ mice, which alleviated the deleterious effect of heterozygocity. Thus, endogenous ligands of PPAR␥ are most probably not abundant or not efficient enough to compensate for the lower levels of PPAR␥ protein by significantly increasing its activity. This observation underscores the importance of the combined effects of receptor expression level and ligand availability to reach the optimal tuning of PPAR␥ activity and, in turn, of its target genes, especially in WAT. We speculate that the reduced level of both metabolic rate and physical activity are part of a protective survival strategy in conditions of energy shortage. When faced with an energy crisis, animals employ various behavioral and physiological responses to reduce metabolism, which prolongs the period of time during which energy reserves can cover metabolic needs. Such behavioral responses can include a reduction in metabolic rate and spontaneous locomotor activity (69). It is possible that the energy crisis in WAT is signaled to the whole body by a reduction in the circulating adiponectin levels. Interestingly, some of the physiological characteristics of the PPAR␥ Ϫ/Ϫ mice resemble those of PPAR␣ Ϫ/Ϫ animals during fasting, such as hypoketonemia and reduced activity (32,70).
Food intake itself cannot explain the metabolic and activity phenotype, since the PPAR␥ ϩ/Ϫ mice ate the same amount of food and had a similar weight as their WT counterparts. Therefore, the fate of the energy spared due to the metabolic and activity down-regulation remains to be elucidated.
Collectively, the data obtained by taking advantage of PPAR␥ ϩ/Ϫ animals fed an SD showed that PPAR␥ activity in WAT is not only important for fat storage, as previously showed. In fact, in the absence of any nutritional challenge, PPAR␥ controls IGF-1 and insulin signaling with effects on energy production via the regulation of the glycolytic and lipolytic pathways. Most importantly, deregulation of these functions in WAT most likely influences the whole body energy balance with its impacts on metabolic rate and physical activity, which appear to adapt to the available energy.