The Transcriptional Coactivator Peroxisome Proliferator–Activated Receptor (PPAR)γ Coactivator-1α and the Nuclear Receptor PPARα Control the Expression of Glycerol Kinase and Metabolism Genes Independently of PPARγ Activation in Human White Adipocytes

Subcutaneous abdominal WAT was obtained from female subjects undergoing plastic surgery in agreement with French laws on biomedical research. Human adipocytes in primary culture were differentiated as described previously (11,13). Adenoviruses produced at the Gene Therapy laboratory of Nantes contained, in tandem, the green fluorescent protein (GFP) gene and the PGC-1α cDNA downstream of separate cytomegalovirus promoters. An adenovirus containing only the GFP gene was used as control. Cells were infected at day 13 with the PGC-1α or control adenovirus at a multiplicity of infection of 200 for 6 h unless otherwise indicated. The next day, cells were treated as indicated in text and figure legends with the following drugs at 1 μmol/l: rosiglitazone (BRL49653; Smith Kline and French), GW7647 (Sigma), and 9-cis-retinoic acid (Sigma). Cells were harvested after 48 h of treatment for mRNA assays and 96 h of treatment for protein, enzymatic activity, and glycerol incorporation assays unless otherwise indicated. PPAR agonists and retinoic acid were prepared in DMSO and then diluted in medium. Vehicle (DMSO at 0.1%) was added to control cells.

Cultures of human subcutaneous adipose tissue explants were performed as described previously (14). After 24 h of treatment with rosiglitazone (1 μmol/l) or vehicle (DMSO), tissue pieces were collected for mRNA analysis.

We analyzed the transcriptional program of human adipocytes overexpressing PGC-1α and/or treated with rosiglitazone plus 9-cis-retinoic acid. Microarrays exhibiting 41,805 cDNA spots representing 41,126 IMAGE clones and 24,474 Unique Putative Genes were from Stanford University (15). Between 30,313 and 31,453 spots were recovered from each array slide. To assess differentially expressed genes, we used the Varmixt method that allows a control of the heterogeneity of variance (16). This procedure yielded 1,746 clones with q value <0.05 representing 1,622 genes that were common to all of the three treatments when compared with control cells. Hierarchical agglomerative cluster analysis was performed using the Euclidian distance as a measure of similarity and the Ward's method as the agglomeration criteria. The dendrogram obtained was cut in six groups. To analyze the function of the differentially expressed genes, we used the Protein Analysis Through Evolutionary Relationships (PANTHER) system database with focus on the biological process ontology. To test for the enrichment of sets of genes in expression data, we used the statistical method Gene Set Enrichment Analysis (GSEA) together with the functional database dedicated to microarrays from Stanford.

Reverse transcription and real-time quantitative PCR (Applied Biosystems) was performed using SYBR green chemistry (Supplementary Table 1, which is available in the online appendix at http://dx.doi.org/10.2337/db06-1465). Some mRNAs were quantified using pre-made gene expression assays (Applied Biosystems). 18S ribosomal RNA was used as control to normalize gene expression.

For PGC-1α Western blot analysis, total proteins from human adipocytes were separated on SDS-PAGE, transferred onto nitrocellulose membrane, and probed with a polyclonal anti-human PGC-1α (Η-300) antibody (Santa Cruz Biotechnology). To determine metallothionein quantities in adipocytes, cytoplasmic proteins from human adipocytes were prepared in Tris-HCl (20 mmol/l, pH 8) buffer. Metallothionein measurement was realized by a radioimmunoassay (17). GyK activity and glycerol incorporation into triglycerides were assessed according to Noel et al. (18) and Guan et al. (12), respectively.

Transfection experiments were performed in 3T3-F442A preadipocytes. Transient transfections were performed by LipofectAMINE 2000 reagent (Invitrogen). A GyK promoter fragment from −2007 to +44 was amplified by PCR from human genomic DNA (Roche Diagnostics) using Platinum high fidelity DNA polymerase (Invitrogen). The PCR product was cloned into the pGL3 basic (Promega). Point mutations of PPAR responsive element (PPRE) present on the GyK promoter were achieved by site-directed mutagenesis (Stratagene). Three copies of the PPRE of the human GyK promoter (AAAGTATGTGAAAGTTCAAACGT) were cloned in front of the thymidine kinase promoter in the pGL3 basic plasmid. The reporter plasmid was cotransfected with expression vectors for PGC-1α, PPARα, PPARγ, and RXRα. A β-galactosidase–expressing plasmid was used for transfection normalization. The experiments were performed in triplicates. Forty-eight hours after the transfection, luciferase activities were analyzed using the Luciferase Reporter Assay System (Promega).

Chromatin immunoprecipitation (ChIP) experiments were performed according to Annicotte et al. (19). Immunoprecipitation was carried out using IgG antibodies, affinity-purified anti-PPARα polyclonal antibody (Santa Cruz), or affinity-purified anti-acetyl histone H4 (Cell Signaling). Purified DNA was amplified using GyK promoter–specific primers and quantified by real-time PCR using SYBR Green chemistry (Supplementary Table 1). Primers targeting a genomic region localized in the human UCP2 gene were used to control for nonspecific binding.

Studies with mice followed the INSERM and Louis Bugnard Institute Animal Care Facility guidelines. Wild-type or PPARα-null mice were treated by daily gavage with Wy14643 (30 mg · kg⁻¹ · day⁻¹) (Sigma) or vehicle (3% carboxymethylcellulose) for 7 days. In vivo adenovirus injection (10¹⁰ infectious particles) were performed in epididymal fat pads of male B6D2/F1 mice. A fat pad was injected with PGC-1α adenovirus and the contralateral fat pad with GFP adenovirus. Tissues were collected after 3 days. For cold exposure experiments, animals were maintained at 4°C for 48 h.

Data are expressed as means ± SE. The data were compared using the nonparametric Wilcoxon's test.

The microarray data for this study are fully accessible at Gene Expression Omnibus with accession number GSE5184.

To gain insight in the respective action of PGC-1α and PPARγ/RXRα in human fat cells, microarray experiments were performed on primary cultures of human subcutaneous differentiated adipocytes transduced with an adenovirus encoding the human form of PGC-1α, treated with rosiglitazone and retinoic acid, or received a combination of both treatments. Only quiescent mature adipocytes were transduced by the adenovirus. This method permits to avoid the effect of continuous PGC-1α expression during adipogenesis. Transduction with the PGC-1α adenovirus led to an increase in PGC-1α mRNA and protein expression, whereas a 30% decrease in PGC-1β mRNA level was observed (Fig. 1A and B). Using a novel statistical method for the differential analysis of gene expression (16) and cluster analysis, we defined six mutually exclusive groups of genes (Supplementary Fig. 1; Supplementary Table 2). PGC-1α alone induced the expression of 125 genes. Very few genes (11) were downregulated in PGC-1α overexpressing adipocytes. These data show that PGC-1α acts almost exclusively as a coactivator. Rosiglitazone and retinoic acid up- and downregulated 245 and 331 genes, respectively.

To determine which biological processes were affected by the treatments, we used the PANTHER ontological classification system. As expected, the treatment with rosiglitazone and retinoic acid induced changes in the expression of genes involved in lipid metabolism (Table 1). Less expected, a significant number of genes involved in protein metabolism were upregulated. Regarding biological processes induced by PGC-1α overexpression, several families of genes expressed in mitochondria were identified. The regulation of sets of mitochondrial genes by PANTHER was confirmed using GSEA (Supplementary Table 3; Supplementary Fig. 2). These data reveal that a large number of genes involved in mitochondrial energy metabolism are regulated by PGC-1α independently of PPARγ/RXRα activation.

We confirmed microarray data by determination of mRNA levels using quantitative RT-PCR (Table 2; Fig. 1C–H). An excellent concordance was observed between the two techniques for gene targets of PPARγ/RXRα and PGC-1α. Known target genes of PGC-1α involved in mitochondrial energy metabolism, such as isocitrate dehydrogenase 3 and solute carrier family 25 member 4 (SLC25A4) (adenine nucleotide translocase 1), were upregulated in PGC-1α–overexpressing cells but not in cells treated only with rosiglitazone and retinoic acid (Fig. 1C and D). The regulation of new target genes, such as parvalbumin, a cytosolic calcium ion binding protein involved in muscle relaxation, p53-induced gene 3, which codes for an enzyme involved in cellular responses to oxidative stress, and microsomal glutathione S-transferase 2, an enzyme involved in leukotriene formation, was also confirmed (Fig. 1E and F; Table 2). Several metallothionein isoforms (MT1B, -F, -G, -H, -K, -L, and -X and MT2A) were induced by PGC-1α overexpression in microarray experiments (Supplementary Table 2). Metallothioneins are low–molecular weight cysteine-rich proteins that have the ability to bind and sequestrate heavy metal ions. In humans, metallothionein genes are clustered within the q13 region on chromosome 16 and include at least 11 MT1 members. MT1F and MT1X mRNA upregulation was confirmed by quantitative RT-PCR (Fig. 1G; Table 2). Moreover, total metallothionein protein level was shown to be induced by PGC-1α expression (Fig. 1H).

GyK gene was among the genes regulated by PGC-1α but not by rosiglitazone and retinoic acid. These data were surprising because previous studies have shown an induction of GyK by rosiglitazone in murine 3T3-L1 adipocytes (12). Therefore, we analyzed GyK expression and activity in several human models. The induction by PGC-1α and lack of effect of rosiglitazone was confirmed by quantitative RT-PCR in human differentiated adipocytes (Fig. 2A). The PPARγ agonist was also devoid of effect in human adipose tissue explants composed of mature adipocytes (Fig. 2B). Furthermore, GyK activity was increased only in human adipocytes overexpressing PGC-1α (Fig. 2C). Induction of PGC-1α by thiazolidinediones and recruitment of the coactivator to the GyK gene has been proposed as the model explaining GyK upregulation by rosiglitazone in 3T3-L1 adipocytes (20). PGC-1α mRNA expression was not regulated by rosiglitazone either in differentiated human adipocytes or in explants of adipose tissue (Fig. 2D and E).

An induction of GyK activity, which is normally very low in white adipocytes, suggests a role in the production of glycerol-3-phosphate, which allows the creation of a futile cycle of triglyceride hydrolysis and resynthesis (12). However, other enzymes may participate in the production of glycerol-3-phosphate. The cytoplasmic and mitochondrial forms of glycerol-3-phosphate dehydrogenase and PEPCK were not regulated by PGC-1α (Supplementary Fig. 3).

PPARα was one of PGC-1α targets (Fig. 3A). To determine an implication of PPARα in GyK gene regulation, we treated human adipocytes overexpressing or not PGC-1α with a selective agonist of PPARα, GW7647. Treatment by GW7647, adenofection of PGC-1α, and the association of both induced GyK gene expression and activity (Fig. 3B and C). The induction of GyK was accompanied by an increase of glycerol incorporation into triglycerides after PPARα activation and PGC-1α overexpression (Fig. 3D). The concomitant time course of PPARα and GyK mRNA induction in PGC-1α–expressing adipocytes was coherent with a role of PPARα in the regulation of GyK expression (Fig. 3E and F).

We then investigated whether overexpression of PGC-1α could induce GyK and PPARα expression in vivo. The PGC-1α or GFP adenoviruses were injected in epididymal fat pads of mice. This fat pad was selected because it contains very few brown fat cells. The increase in PGC-1α mRNA expression was accompanied by a comparable increase of GyK and PPARα mRNA levels (Fig. 4A). A twofold induction of PGC-1α expression was observed in epididymal white fat pad exposed to cold for 48 h (Fig. 4B). The increase in PGC-1α mRNA expression was associated with a comparable induction of PPARα and GyK gene expression. We investigated the involvement of PPARα in the regulation of GyK expression in WAT of PPARα-null mice. Wild-type and PPARα^−/− mice were treated with the PPARα agonist (WY 14643). An induction of GyK mRNA was observed in control mice treated with WY14643 but not in PPARα^−/− mice (Fig. 4C). As previously reported, the WAT expression of GyK was higher in PPARα^−/− mice than in wild-type mice (21).

A functional PPRE has been identified in the murine GyK promoter (20). This PPRE is perfectly conserved in the human promoter (Fig. 5A). To investigate the role of PPARα in the transcription of the GyK gene, we transfected a human GyK promoter construct linked to the luciferase gene with expression vectors for PPARα, RXRα, and PGC-1α into 3T3-F442A cells. Cells were also treated with the PPARα agonist, GW7647. PGC-1α and the heterodimer PPARα/RXRα induced an increase in luciferase activity (Fig. 5B). A slightly greater induction was observed in cells expressing both PGC-1α and PPARα/RXRα and treated with GW7647. Mutation of the PPRE abolished induction by PPARα/RXRα in the presence of GW7647. An induction was observed with PGC-1α, indicating possible interaction of the coactivator with other transcription factors (Fig. 5C). To get a direct demonstration of the involvement of the GyK PPRE in PPARα-mediated transactivation and to compare PPARα and PPARγ effects, PPRE was cloned upstream of a minimal thymidine kinase promoter. Transfection into 3T3-F442A cells revealed a robust induction by PPARα/RXRα, whereas PPARγ/RXRα showed poor transactivation capacity (Fig. 5D). Moreover, PGC-1α had little effect on luciferase activity, suggesting that PPARα is able to transactivate the human GyK promoter independently of PGC-1α. ChIP experiments were performed to investigate whether PPARα binds the Gyk PPRE (Fig. 5E). Recruitment of endogenous PPARα to the GyK PPRE was observed in human adipocytes overexpressing PGC-1α. The concomitant presence of acetylated histone H4 in the GyK PPRE suggests that the promoter is transcriptionally active when PPARα is bound. Binding of PPARα on an unrelated genomic region was not enhanced in PGC-1α–expressing adipocytes.

Alterations in fatty acid metabolism and mitochondrial dysfunction are thought to be critical in the genesis of insulin resistance (22). The transcriptional coactivator PGC-1α is pivotal in these pathways because it controls genes involved in fatty acid oxidation and mitochondrial respiratory chain (6). In this report, we performed an exhaustive analysis of PGC-1α and PPARγ/RXRα target genes in human white fat cells. We identified genes belonging to various pathways that were regulated independently by PGC-1α or PPARγ/RXRα. Among these genes was GyK, a key enzyme in fatty acid esterification. A model is proposed to recapitulate how PGC-1α and PPARα trigger GyK expression and metabolic function (Fig. 6). In human fat cells, PPARα is induced by PGC-1α, and its activation leads to an increase in GyK expression and glycerol incorporation into triglycerides. In vivo data in the mouse support the role of PPARα in the regulation of GyK. Studies of the human GyK promoter show that the nuclear receptor acts through a PPRE located at −1.7 kb. Interestingly, GyK expression data in human adipocytes and promoter studies reveal that PGC-1α is dispensable for PPARα-mediated transactivation. Moreover, in vitro and in vivo data suggest that other transcription factors are involved in PGC-1α–mediated upregulation of GyK.

Gene clustering analysis and classification according to functional annotations show that numerous pathways are regulated by PGC-1α independently of the activation of the PPARγ/RXRα heterodimer. Induction of PGC-1α expression in human adipocytes induces mitochondrial genes, as observed in skeletal muscles, hepatocytes, and 3T3-F442A preadipose cells (5,6,23). Analysis of biological processes shows that enzymes of the citric acid cycle and proteins of the respiratory chain are upregulated in human adipocytes. Among the gene sets selected by GSEA analysis, we found genes involved in reactive oxygen species (ROS) metabolism. PGC-1α could protect cells against an increase in ROS production resulting from its capacity to induce mitochondrial gene expression and thus a higher mitochondrial activity. Several isoforms of metallothionein were upregulated, and metallothionein protein quantity was increased by PGC-1α. MT1 is induced in BAT by cold and catecholamines, two conditions that induce PGC-1α (24). Accordingly, we found that MT1 gene expression was increased during cold exposure in mouse WAT (data not shown). Besides a role in metal metabolism, metallothionein may protect cells against oxidative stress (25). Because ROS have been recently proposed to play a causal role in multiple forms of insulin resistance (26), PGC-1α could contribute to the maintenance of insulin sensitivity through induction of protective genes, such as metallothionein. Induction of metallothionein genes may also be beneficial for the control of energy balance as MT1- and MT2-null mice develop obesity (27).

The predominant pathway for fatty acid esterification to form triglycerides in adipocytes starts with glycerol-3-phosphate. GyK catalyzes direct conversion of glycerol to glycerol-3-phosphate. By comparison with BAT, GyK activity in human WAT is low (28). In human white fat cells, we showed that PGC-1α strongly induces GyK gene expression and activity. The increase in glycerol incorporation into triglycerides after GyK upregulation allows a direct recycling of glycerol resulting from a complete hydrolysis of triglycerides (12). The existence of a triglyceride/fatty acid cycle has been demonstrated in physiological and pathological situations in humans (29). Therefore, in human white adipocytes expressing PGC-1α, a futile cycle induced by Gyk and the stimulation of fat oxidation could limit the release of fatty acids by WAT and participate in the beneficial modification of metabolism (Fig. 6) (4,11).

Unlike what has been reported in mouse adipocytes (12), we did not find any stimulation of GyK expression by rosiglitazone in human primary adipocytes and adipose tissue explants. The lack of induction of GyK mRNA by rosiglitazone has been confirmed in human adipose tissue explants (30). No change in Gyk mRNA expression has been observed in WAT of type 2 diabetic patients treated with thiazolidinediones (31,32). Fasting glycerol output from WAT was not decreased during a rosiglitazone treatment as would be expected if Gyk activity was upregulated (31). Moreover, elegant in vivo experiments using heavy water showed that, under rosiglitazone treatment, GyK is unlikely to contribute to triglyceride synthesis in adipose tissue (33). The lack of induction in human fat cells is probably related to the absence of PGC-1α upregulation because this mechanism has been proposed as essential for the induction of GyK by rosiglitazone in 3T3-L1 adipocytes (20). Next, we searched to identify which nuclear receptor cooperates with PGC-1α to regulate GyK expression. We found that PGC-1α upregulates PPARα expression in human adipocytes. Moreover, a selective agonist of PPARα increased GyK expression and activity in our model. Furthermore, promoter-reporter gene assays showed that the nuclear receptor activates human GyK transcription, and ChIP experiments revealed that PPARα was directly recruited to the GyK promoter in PGC-1α–expressing human adipocytes. Treatment of human white fat cells with forskolin, an adenylylase activator, upregulates PGC-1α and PPARα mRNA expression (34). We observed that forskolin increased GyK mRNA level in human adipocytes (data not shown). In situations of chronic activation of white fat cells by catecholamines, high intracellular cAMP levels could lead to a stimulation of lipolysis and an induction of PGC-1α, PPARα, and GyK expression promoting fatty acid esterification (Fig. 6).

The data raise the interesting possibility of a role for PPARα in regulating white fat metabolism. PPARα deficiency in mice results in disturbances of triglyceride and cholesterol metabolism in WAT with alterations in lipogenic gene expression (35). Besides the well-established systemic effect of PPARα agonists on energy expenditure and fat oxidation, it has recently been suggested that activation of the nuclear receptor in WAT could prevent adipocyte hypertrophy (36). In line with this concept, PPARα is essential for the lipopenic effect of hyperleptinemia on WAT (37). Furthermore, PPARα mediates the action of a β₃-adrenergic receptor agonist on WAT gene expression and remodeling (21). These data on mouse WAT shall however be interpreted with caution because important species differences exist between mouse and human white fat cell metabolism (38). However, our data show the direct involvement of PPARα in the control of human adipocyte gene expression. This study paves the way for future work aiming at characterizing the biological pathways modulated by PPARα agonists in human fat.

This work has received support from INSERM; the Programme National de Recherche sur le Diabète; the Agence National de la Recherche program on Cardiovascular Disease, Diabetes, and Obesity (FAIR project); the Ministerio de Ciencia y Tecnología and Feder (SAF2005-00671); the European Commission FP6 Integrated Projects Exgenesis and MolPAGE (contracts LSHM-CT-2004-005272 and LSH-2003-1.1.3-1); and the project “Hepatic and adipose tissue and functions in the metabolic syndrome” (HEPADIP; see http://www.hepadip.org/), which is supported by the European Commission as an Integrated Project under the 6th Framework Programme (contract LSHM-CT-2005-018734).

We are grateful to Jean José Maoret (Institut Fédératif de Recherche 31), Frédéric Lasserre (Institut National de la Recherche Agronomique), Arnaud Polizzi (Institut National de la Recherche Agronomique), Audrey Sicard (INSERM U858), Audrey Sambeat (INSERM U858), and Carine Valle (INSERM U858) for skillful technical assistance. We thank the Vector Core of the University Hospital of Nantes, supported by the Association Française contre les Myopathies, for the production of adenovirus vectors.