CREB activation induced by mitochondrial dysfunction triggers triglyceride accumulation in 3T3-L1 preadipocytes

The large variety of metabolic disorders related to bioenergetical stress underlines the essential role of mitochondrial activity in cellular physiology. More particularly, several lines of evidence now show that defects in β-cells insulin secretion as well as systemic insulin resistance in type 2 diabetes could be attributable to mitochondrial dysfunction (Lowell and Shulman, 2005). In human muscle cells, the major effectors involved in insulin resistance include adipokines (Lazar, 2005) and fatty acids, which cause a direct inhibition of insulin-stimulated glucose transport activity through a decrease in phosphoinositide 3-kinase (PI 3-kinase) signaling (Dresner et al., 1999). Furthermore, impairment of mitochondrial activity associated with ageing, which could result from alterations in mitochondrial DNA (mtDNA), also leads to triglyceride (TG) accumulation in muscles and liver of healthy and lean elderly people (Petersen et al., 2003). Reduced mitochondrial activity and intramyocytic accumulation of TG were also found in the young and insulin-resistant offspring of parents with type 2 diabetes. Insulin-resistant patients usually have a lower ratio of type 1 muscle fibers to the more glycolytic type 2 muscle fibers, an observation that could be explained by the reduced expression of members of the peroxisome proliferator-activated receptor γ (PPARγ)-coactivator-1 (PGC-1) family and the downregulation of their target genes involved in mitochondrial biogenesis (Wu et al., 1999; Mootha et al., 2003). Hypoxic conditions that inhibit mitochondrial respiration also lead to TG accumulation in cardiomyocytes (Huss et al., 2001). In addition, we recently showed that, the impairment of mitochondrial activity by inhibitors of respiratory complexes also triggers TG accumulation in murine 3T3-L1 preadipocytes, resulting from a decrease in fatty acid β-oxidation and an increase in glucose uptake enhancing glycerol 3-phosphate synthesis (Vankoningsloo et al., 2005).

Changes in cellular metabolism due to a loss of mitochondrial oxidative capacity lead to the activation of cell signaling pathways and modifications in the expression of many nuclear genes. This process, known as mitochondria-nucleus retrograde communication, has mainly been studied in yeast (Liao et al., 1991; Butow and Avadhani, 2004) and more recently in mammalian cells depleted in mitochondrial DNA (mtDNA) (Amuthan et al., 2002; Biswas et al., 1999; Biswas et al., 2005; Arnould et al., 2002). For example, we have recently shown that depletion of mtDNA or inhibition of mitochondrial respiration activates the cAMP-response-element binding protein (CREB) by phosphorylation of Ser133 mediated through a Ca²⁺/calmodulin-dependent kinase IV (CaMK IV) pathway (Arnould et al., 2002).

CREB is a transcription factor with pleiotropic effects that has already been reported to play a role in the control of memory (Scott et al., 2002), cell proliferation (Della Fazia et al., 1997), and in glucose and lipid metabolism (Zhou et al., 2004; Reusch et al., 2000; Herzig et al., 2003). Previous studies showed that this transcription factor is also an early regulator of adipocyte differentiation because it is activated by phosphorylation of Ser133 in the presence of adipogenic inducers, such as glucocorticoids, cyclic AMP analogues and insulin-like growth factor-1 (IGF-1), or high concentrations of insulin (MacDougald and Lane, 1995). Furthermore, ectopic expression of the chimeric and constitutively active VP16-CREB is sufficient to trigger adipogenesis, whereas overexpression of a dominant negative form of CREB (K-CREB) inhibits the adipogenic program (Reusch et al., 2000) and leads to apoptosis of mature adipocytes (Reusch and Klemm, 2002). These data suggest that CREB acts as proadipogenic and survival factor. It has also been reported that the positive effect of activated CREB on adipogenesis is mediated by the overexpression of the CCAAT/enhancer-binding protein β (C/EBPβ) gene, a key transcription factor in the differentiation program that contains dual cAMP-response element (CRE)-like cis regulatory-elements in its promoter (Zhang et al., 2004). Whereas defects in adipocyte metabolism induced by mitochondrial dysfunction might influence muscle and liver metabolism because it has been associated with lipodystrophy and impairment of fatty acid β- oxidation in these tissues (Petersen et al., 2002), preadipocyte response to mitochondrial dysfunction is still poorly understood.

Here, we show that CREB is activated in 3T3-L1 preadipocytes when the cells are incubated with antimycin A (AA), an inhibitor of the complex III in the mitochondrial electron transporter chain that triggers the accumulation of cytosolic TG in these cells (Vankoningsloo et al., 2005). The fact that CREB can be activated by the inhibition of mitochondrial activity in several cell lines and acts as a survival and differentiating factor in adipocytes and preadipocytes, respectively, led us to hypothesize that CREB also plays a role in the accumulation of triglycerides in preadipocytes with impaired mitochondrial activity.

We thus developped and used a low-density DNA microarray to study gene expression profiles of major adipogenic markers that are potentially responsible for TG accumulation in AA-treated preadipocytes. These results were compared with differentially expressed genes obtained for differentiating 3T3-L1 cells in the presence of a standard hormone cocktail. The DNA microarray used in this study allows gene expression profiling for 89 genes related to adipogenesis and lipid metabolism. These markers have been carefully selected, based on the literature reporting on differentiation-specific gene expression during 3T3-L1 adipogenesis (Burton et al., 2004; Guo and Liao, 2000; Kratchmarova et al., 2002). We also evidenced that, inhibition of CREB expression with small interfering RNA (siRNA) and reduction of its activity by overexpression of two dominant negative mutants (K-CREB and M1-CREB) diminishes the TG accumulation induced by AA in 3T3-L1 preadipocytes. Finally, combining RNA interference (RNAi) and microarray technology, we identified several CREB-target genes that are differentially regulated when 3T3-L1 cells are incubated with AA. Taken together, these results not only extend the role of CREB in adipocyte biology and lipid metabolism but also highlight the 3T3-L1 preadipocyte response to mitochondrial dysfunction leading to TG accumulation, which might not only impair adipocyte metabolism but also the physiology of insulin-dependent tissues.

Differentiation of 3T3-L1 preadipocytes into adipocytes is easily triggered by a standard adipogenic cocktail comprising insulin, a cAMP-elevating agent and dexamethasone (MacDougald and Lane, 1995). The main morphological characteristic of adipogenesis is the progressive storage of large amounts of cytosolic TG, whereas TG that accumulate in 3T3-L1 cells incubated with 10 nM AA form small but numerous vesicles, as revealed after the staining of neutral lipids with Oil Red O (Fig. 1A) (Vankoningsloo et al., 2005). Quantitative analysis showed that TG accumulation in 3T3-L1 cells treated for 8 days with AA depends on the concentration of this metabolic inhibitor (Fig. 1B). A similar phenotype was observed for 3T3-L1 cells incubated 16 days with chloramphenicol (a well-known inhibitor of mitochondrial protein synthesis), although it took longer for TG to accumulate (Fig. 1C).

We previously reported a constitutively enhanced CREB activity in mtDNA-depleted L929 and rho0 143B cells, as well as phosphorylation of the CREB Ser133 in several cell lines that had been incubated with oxidative phosphorylation inhibitors (Arnould et al., 2002). Here, we extended these data by showing that preadipocytes also respond to the impairment of mitochondrial activity by the activation of this transcription factor. Indeed, using several approaches, we showed that CREB is phosphorylated on Ser133 and localizes in the nucleus of 3T3-L1 cells incubated with AA (Fig. 2A-B). Western blot analysis revealed that CREB phosphorylation is increased after 6 hours of treatment with 10 nM AA and remains sustained for at least 72 hours, whereas the adipogenic cocktail increases levels of Ser133-phosphorylated CREB (pCREB) for 24 to 72 hours. Increases in levels of pCREB are not observed after 8 days of treatment. The abundance of total CREB is similar in all tested conditions (Fig. 2A). These results clearly indicate that the phosphorylation of CREB on Ser133 is enhanced in 3T3-L1 cells treated with AA. These results were also confirmed by immunofluorescence experiments and confocal microscopy, which show that pCREB mainly accumulates in the nucleus of cells incubated for 24 hours with the mitochondrial inhibitor or the adipogenic cocktail (Fig. 2B). An enhanced phosphorylation of Ser133 strongly suggests that CREB is bound to DNA and transcriptionally active under these conditions. Therefore, we next used a colorimetric assay allowing the quantification of CREB and also pCREB bound to a synthetic oligonucleotide containing a CRE-site. We observed that, whereas the total amount of CREB binding to DNA remained unchanged, the relative amount of bound pCREB was increased (1.5- to three-fold increase after 24 hours) by both AA and the adipogenic cocktail (Fig. 2C). The resulting effect of pCREB on its ability to transactivate gene expression in these conditions was demonstrated in transient transfection experiments with a CREB-sensitive luciferase-reporter construct driven by the authentic α-inhibin promoter, which contains four CRE sites (Pei et al., 1991; Fig. 2D). As expected, the transcriptional activity of CREB was significantly increased in cells incubated for 24 hours with 10 nM AA or the adipogenic cocktail (threefold or fourfold increase, respectively).

CREB is suspected to play a role in TG metabolism because it binds to the promoter of adipogenic marker genes, such as C/EBPβ, during 3T3-L1 cell differentiation (Zhang et al., 2004). To delineate a potential role for CREB in the AA-induced TG accumulation, cells were transiently transfected with plasmids encoding either dominant negative CREB mutants (K-CREB and M1-CREB) or enhanced green fluorescent protein (EGFP) as a negative control, and then incubated for 8 days with 10 nM AA followed by staining for TG and spectrophotometric quantification. We found that AA-induced TG accumulation is significantly reduced in cells that overexpress either K-CREB or M1-CREB dominant negative mutants (43.1 % or 44.8 %, respectively; data not shown). To circumvent the rather low transfection efficiency of these cells with the Superfect reagent (ranging from 20 to 45 % as determined by a β-galactosidase reporter construct), which might minimize the inhibitory effect of the dominant negative forms, we next used a specific siRNA to silence CREB gene expression.

Before testing the effect that CREB silencing has on accumulation of TG in 3T3-L1 cells incubated for 8 days with 10 nM AA (Fig. 3), we verified by confocal microscopy that FITC-labeled siRNA had been efficiently introduced into 3T3-L1 preadipocytes, and estimated that transfection efficiency was at least 90% (Fig. 3A). Out of the three different siRNAs tested, only one inhibited CREB expression efficiently by more than 80% (data not shown). Using western blot analysis, we then checked that this CREB-specific siRNA reduces the abundance of CREB protein in a concentration-dependent manner (Fig. 3B). The maximal inhibitory effect was observed 24 to 48 hours post-transfection, whereas CREB knock-down was relieved 96 hours after transfection (data not shown), a result in agreement with the transient effect of siRNA in mammalian cells. The expression of TATA-box-binding protein (TBP) was monitored by western blot analysis and was found unchanged in these conditions, suggesting that siRNA targeting CREB did not repress global protein synthesis by an `off-target' effect. To test the potential role of CREB on TG that accumulate in AA-treated 3T3-L1 cells, 100 nM siRNA were transfected and cells were then incubated for 8 days with 10 nM AA. In these conditions, we observed that delivery of siRNA into 3T3-L1 cells reduces the AA-induced TG accumulation by 75% (Fig. 3C-D). These results show that CREB activation induced by AA is required for accumulation of TG - as observed in response to mitochondrial inhibition - and suggest that CREB-mediated transcription at the early stage of the mitochondria-impairing treatment is an important mechanism, leading to the observed phenotype.

The accumulation of TG in preadipocytes leading to the adipocyte phenotype is controlled by numerous genes of several classes. Products of these genes either drive differentiation (mainly C/EBPβ and PPARγ) or maintain a differentiated state (C/EBPα). We thus hypothesized that TG accumulation in 3T3-L1 preadipocytes with impaired mitochondrial activity is also caused by alterations in gene expression. We then compared differences and similarities in gene expression induced by the adipogenic cocktail or AA. To investigate the expression of genes that are differentially regulated by AA, a low-density DNA microarray was developed. This microarray allows gene expression analysis for a set of key genes related to adipogenesis and lipid metabolism (supplementary material Table S1 and Fig. S1). 3T3-L1 cells were incubated for 2, 4, 6 or 8 days with the adipogenic cocktail or with 10 nM AA. Cells were harvested and total RNA was extracted and reverse transcribed by including biotinylated nucleotides. The resulting cDNAs were hybridized on the microarray and stained with a cyanin-3-conjugated anti-biotin antibody. Three independent experiments were performed for each incubation time and each sample was hybridized to three submicroarrays, represented by the probes spotted in triplicate. The array data were subjected to a simple algorithm (Materials and Methods) to set a lower threshold and to normalize the data to internal-standard controls and house keeping genes. Then, average ratios and their standard deviations (s.d.) were calculated. We ended up with a category of gene transcripts that were not detected owing to their low abundance, detected with no modification, or quantitatively or qualitatively up- or down-regulated. Genes were assigned to the latter category when their fluorescent signal was either saturated or less than 2.5 times that of the local background in one of the experimental conditions.

Gene expression data for differentiating 3T3-L1 cells are presented in Table 1. The first list identified 36 genes found to be significantly upregulated during the time course of in vitro adipogenesis. Many of these include genes involved in lipid metabolism (CPT-2, DHAPAT, FABP4/aP2, FAS, GPAT, HSL, LPL1, MCAD, SCD1, SCD2), in the transcriptional control of adipogenesis (C/EBPα, PPARγ, SREBP-1) and in cell-cycle arrest (p18), as well as genes encoding adipokines (Acrp30, AGT, resistin). Among the upregulated genes during preadipocyte differentiation, several sets of genes (clusters) display various kinetic profiles: genes are either significantly induced after 2 days (AAAT, AGT, Cav-1, Coll VIα2, CPT-2, FAS, GPDcyt, Gsn, Hp, MMP-2, PPARγ, resistin, RXRα, SCD1, SREBP-1, UCP-2), begin to be upregulated after 4 days (Acrp30, adipsin, Cav-2, C/EBPα, Clic4, cyclin D3, DHAPAT, FABP4/aP2, GPAT, HSD, HSL, HSP 60, LPL-1, MCAD, p18, PPARδ, SCD2) or are only found to be differentially regulated after 6 days of treatment (CL, Plin, VEGF-A). The expression of the genes upregulated most is sustained during the whole adipogenic program, but some genes also display transient overexpression profiles as found for Gsn, MMP-2 and RXRα. As already reported, 3T3-L1 cell differentiation is also characterized by the downregulation of adipogenesis repressors such as Pref-1 and the transcription factors GATA2 and GATA3 (Lee et al., 2003; Tong et al., 2005). Other downregulated genes in differentiating adipocytes include PEDF, SDF1, SDF2, Smad3, TF and VEGF-C. No changes were observed for C/EBPβ, a well-known early marker of adipogenesis (Darlington et al., 1998). However, due to inaccurate probe selectivity, C/EBPβ-derived cDNA can cross-hybridize with the probe for C/EBPδ (data not shown). Therefore, we analyzed the expression profile for these two transcription factors in differentiating 3T3-L1 cells by real time PCR (RT-PCR) and found that C/EBPβ and C/EBPδ are transiently upregulated during the first 2 to 4 days of adipogenesis (Fig. 4). Several gene transcripts were not detected on the microarray (ATR II, CaMK IV, CPT-1 M, Gyk, IL-6, iNOS, leptin, renin, RXRγ) probably because low abundance makes their detection impossible below the sensitivity threshold of the microarray technique. Indeed, when analyzed by RT-PCR assays, leptin was found to be upregulated in differentiated 3T3-L1 adipocytes, whereas IL-6 was shown to be downregulated (Fig. 4).

3T3-L1 fibroblasts, differentiated with a standard proadipogenic cocktail, are a commonly used model to study differentiation for which a lot of experimental data on gene expression can be found in the literature (Rosen and Spiegelman, 2000; Guo and Liao, 2000; Burton et al., 2002; Burton et al., 2004; Ross et al., 2002). The fact that our data are in good agreement with previous studies that report changes in gene expression during adipogenesis in vitro, already validates the DNA microarray developed in this study. To further validate our data, we also performed SYBR Green quantitative RT-PCR assays for selected genes. Values obtained for a set of three genes upregulated in response to the adipogenic cocktail were confirmed for different incubation times (Fig. 5A). For these genes we found a very good correlation of the relative transcript-abundance-data obtained by DNA microarray and by RT-PCR.

We next analyzed the effect of AA-induced mitochondrial dysfunction on gene expression in 3T3-L1 cells (Table 2). Few genes were found to be continuously upregulated in cells responding to the inhibition of mitochondrial activity. These genes included those encoding cytokines (PAI-1, TGF-β1), stress proteins (HSP 60), transcription factors (CHOP-10, GATA-3) and proteins involved in energy and lipid metabolism (GPDmit, FABP4/aP2, SCD1). Some genes were transiently upregulated in these conditions (adipsin, ATR I, Cav-1, cyclin D3, GLUT-4, Gyk). We also found that AA reduces the expression of many genes, such as ADD1, AGT, ASP, C/EBPα, Coll VIα2, CPT-1 L, FAS, GPDcyt, LPL1, MCAD, NFATc2, NFATc4, p110α, p18, Pref-1, resistin, SDF1, Smad3, Stat6 and VEGF-C. Relative transcript abundance for FAS and Clic4, determined by RT-PCR at day 4 of AA treatment, is similar to the values obtained with the microarray analysis on the same samples (Fig. 5A). We also observed that AA increases the abundance of mitochondrial glycerol-3-phosphate dehydrogenase (GPDmit) protein in 3T3-L1 cells (Fig. 5B), in accordance to upregulation of the GPDmit transcript (as determined with the microarray).

The expression status of genes that were up- or down-regulated by both AA and the adipogenic cocktail was compared after 2, 4, 6 and 8 days of treatment (Fig. 6). Although it is known that many stresses, including energetic stresses, trigger a shut-down of global transcription and protein synthesis (Buttgereit and Brand, 1995; Wieser and Krumschnabel, 2001), most of these genes were not known to be differentially regulated in response to a mitochondrial dysfunction. Several patterns of gene expression were identified and some of them are illustrated in Fig. 7. In the first pattern, expression of the gene was elevated at day 2 and then declined (Fig. 7A). This group included adipsin, ATR I, Cav-1, CPT-2, cyclin D3, GLUT-4 and Gyk. In the second pattern, gene expression was increased at day 2 and sustained during the whole programme, as for GPDmit (Fig. 7B). In the third group, gene expression was elevated at day 4 and maintained at high levels, as observed for CHOP-10, GATA-3, HSP60, SCD-1 and TGF-β1 (Fig. 7C). The last profile included genes found to be downregulated from day 4 (AGT, C/EBPα, CPT-1 L, MCAD, p18, Stat6, VEGF-C) or day 6 (ADD1, CL, NFATc2, NFATc4, p110α, Pref-1, resistin, SCD2, SDF1, SDF2, Smad3) of AA treatment (Fig. 7D).

The role of CREB in the modifications of gene expression induced by AA was further investigated using the powerful combination of RNAi (to silence CREB) and microarray technology. This analysis was performed because CREB is an important transcription factor activated by mitochondrial dysfunction (Arnould et al., 2002), a primary regulator of adipogenesis leading to TG accumulation (Reusch et al., 2000) and contributes to TG accumulation in AA-treated cells (Figs 3 and 4). 3T3-L1 cells were evenly divided into two groups. One group was transfected with double-stranded RNA olignucleotides directed against CREB transcripts, whereas the other group was treated as a control with the siRNA delivery reagent JetSI™ alone. Previous studies, targeted at the NF-κB, HTLV-I tax and HIV-1 reverse transcriptase genes showed that unrelated siRNAs do not impair gene expression that is specifically altered by TNFα, and give equivalent results to the transfection reagent alone (Zhou et al., 2003). Thus, cells transfected or not with siRNA silencing CREB were incubated for 4 days with 10 nM AA and processed for hybridization (Table 3). We first observed that the JetSI reagent does not significantly modify the expression of most genes analyzed by the microarray, when compared with the profiles obtained for two independent RNA extractions prepared from 3T3-L1 cells that had been treated for 4 days with 10 nM AA (Table 3, AA1 and AA2 columns). We also found that, out of the 89 genes analyzed, the transcript abundance for eight genes (CHOP-10, Clic4, GPDmit, HSP 60, PAI-1, TF, VEGF-A and SCD1) was significantly decreased by CREB-specific siRNA, suggesting a role for CREB in the control of the transcription of these genes in response to AA. On the opposite, the differential expression of some genes, such as upregulation of FABP4 and downregulation of p18, is not affected by the inhibition of CREB expression in AA-treated cells.

Numerous experimental data and observations reported from physiopathological situations and experimental models now clearly support a direct link between mitochondrial dysfunction and metabolic disorders, therefore addressing the essential role of mitochondrial-activity impairment in alterations in lipid metabolism (Lowell and Shulman, 2005; Petersen et al., 2003; Petersen et al., 2004). The myoclonic epilepsy with ragged red fibers (MERRF) syndrome that is often caused by numerous point mutations (G611A, A3243G, A8344G, G8361A, G12147A) in mtDNA, which affect genes encoding different mitochondrial tRNAs (Mancuso et al., 2004; Mongini et al., 2002; Shoffner et al., 1990; Rossmanith et al., 2003; Melone et al., 2004), is also to some extent associated with lipid-storage disorders and TG accumulation in muscles (Munoz-Malaga et al., 2000; Naumann et al., 1997). It has also been shown that most of multiple symmetrical lipomatosis (MSL) patients also display mitochondrial dysfunction (Naumann et al., 1997; Berkovic et al., 1991), lipomas that contain atypical multivacuolar white adipocytes (Zancanaro et al., 1990) as well as ragged red fibers and TG accumulation in muscles (Klopstock et al., 1997; Munoz-Malaga et al., 2000). The role of mitochondria in TG metabolism is also strengthened by their implication in the lipodystrophy syndrome resulting from anti-retroviral therapies that combine drugs known to inhibit mitochondrial DNA polymerase γ and mitochondrial processing protease (Kakuda, 2000; Brinkman et al., 1999). Cytosolic TG accumulation is also observed in cardiomyocytes incubated under hypoxia (Huss et al., 2001) and in preadipocytes incubated with mitochondrial respiration inhibitors (Vankoningsloo et al., 2005). Both conditions lead to the downregulation of CPT-1 M (carnitine palmitoyltransferase-1, muscle isoform) and a subsequent decrease in mitochondrial fatty acid β- oxidation. These data emphasize the link between mitochondrial dysfunction and abnormalities in TG storage. However, even if TG accumulation in response to mitochondrial alterations is of interest, because it can modify cell sensitivity to agonists such as insulin (Lowell and Shulman, 2005), the mechanisms leading to cellular TG accumulation in these conditions remain largely unknown.

Cells with mitochondrial dysfunction provide an adaptative response mainly characterized by activation of glycolysis and a decrease in cell proliferation. In addition, depending on cell type and model, several signaling pathways participate to the so-called `retrograde communication' between mitochondria and the nucleus, allowing the cells to change the activity status of several transcription factors such as nuclear factor of activated T cells (NFAT), nuclear factor-κB (NF-κB) and CREB leading to modifications in gene expression (Liu and Butow, 1999; Biswas et al., 1999; Butow and Avadhani, 2004; Arnould et al., 2002). For example, we have previously shown that CREB is activated by phosphorylation on Ser133 in the cell lines L929 and 143B, with impaired mitochondrial activity induced either by mtDNA depletion or inhibitors of the oxidative phosphorylation such as AA, oligomycin or FCCP (Arnould et al., 2002). A constitutive activation of CREB was also found in cybrid cells, with the A8344G mutation in the mitochondrial genome described to be responsible for the MERRF syndrome (Arnould et al., 2002). Interestingly, CREB has also been reported to be phosphorylated on Ser133 in PC12 cells incubated under hypoxia, another condition known to inhibit mitochondrial respiration (Beitner-Johnson and Millhorn, 1998; Beitner-Johnson et al., 2000).

CREB is an ubiquitous transcription factor that regulates numerous cellular functions such as cell survival, proliferation and differentiation, as well as glucose and lipid metabolism (Reusch and Klemm, 2002; Reusch et al., 2000; Della Fazia et al., 1997; Zhou et al., 2004; Herzig et al., 2003). Here we clearly show that CREB is phosphorylated on Ser133 and transcriptionally more active in 3T3-L1 preadipocytes when mitochondrial activity is inhibited by AA, a complex-III inhibitor. The kinase that phosphorylates CREB in AA-treated 3T3-L1 cells remains to be identified but CaMK IV is a good candidate because it is the effector in mtDNA-depleted cells (Arnould et al., 2002). However, several other kinases such as PKA (protein kinase A), PKB/Akt, CaMK I/II or ribosomal S6 kinase 1/2 (RSK1/2) could also play a role because they phosphorylate CREB in response to various stimuli (Johannessen et al., 2004; Shaywitz and Greenberg, 1999). A decrease in PP1 or PP2A (protein phosphatase 1 or 2A, respectively) activity can also be considered because these enzymes dephosphorylate CREB (Hagiwara et al., 1992; Wadzinski et al., 1993).

CREB activation that is induced by impaired mitochondrial activity in different cell types, suggests an important role for this transcription factor in the adaptative cell response to energetical stress. Although the molecular mechanisms are still unidentified, we clearly show that CREB is involved in the cytosolic accumulation of TG induced by a prolonged (several days) exposure of 3T3-L1 preadipocytes to AA. Indeed, the amount of TG accumulated in cells treated for 8 days with AA is reduced by overexpression of the dominant negative forms of either K-CREB or M1-CREB (by about 40%) and the silencing of CREB expression in transiently transfected cells with a specific siRNA (by 75%). The weaker inhibitory effect of dominant negative mutants might be caused by a lower transfection efficiency of plasmids (ranging from 20 to 40% as determined in subconfluent cells transfected with a plasmid encoding a pCMV-β-galactosidase as a reporter gene) compared with that of siRNA. However, a lesser inhibition of TG accumulation, observed in cells that overexpress dominant negative CREB mutants, might also be owing to a brief inhibition of CREB activity, caused by dominant negative mutants, whereas a longer-lasting effect is obtained in the presence of siRNA.

CREB silencing was shown to take place during the first 48 hours post-transfection with siRNA but was transient because CREB expression is recovered after 96 hours (data not shown). Since TG accumulated significantly after 7 to 8 days of treatment (Vankoningsloo et al., 2005), our data suggest that the rapid CREB activation induced by AA triggers a cascade of events, leading to TG accumulation later on. CREB activation in 3T3-L1 cells responding to a pro-adipogenic cocktail was also detected very early in differentiation, before any TG started to accumulate (Reusch et al., 2000). We thus identified CREB as a key-effector in cytosolic accumulation of TG, induced by AA in 3T3-L1 cells.

To better understand the metabolic origin of TG accumulation in 3T3-L1 preadipocytes with impaired mitochondrial activity, we developed and used a low-density DNA microarray that allowed simultaneous gene-expression analysis for numerous adipogenic markers. We validated the data obtained in the array experiments by several means. First, using the 3T3-L1 preadipocyte cell line was advantageous because it has been used extensively to investigate differentiation-dependent gene expression in adipocytes (Rosen and Spiegelman, 2000; Guo and Liao, 2000; Burton et al., 2002; Burton et al., 2004; Ross et al., 2002). Since these previous studies have led to the identification of several genes that are differentially expressed during differentiation, any new approach should also successfully pick up these genes and, in doing so, serve as a compelling control method. Indeed, we found that many adipogenic markers were upregulated during adipogenesis, e.g. several transcription factors (C/EBPα, PPARγ, RXRα, SREBP-1), enzymes involved in fatty acid and TG metabolism (DHAPAT, FABP4/aP2, FAS, GPAT, SCD1, SCD2), and adipokines (AGT, resistin) (MacDougald and Lane, 1995; Gregoire et al., 1998; Burton et al., 2004; Hajra et al., 2000; Song et al., 2002). We also observed the downregulation of anti-adipogenic markers, such as Pref-1, an autocrine and/or paracrine inhibitor of adipogenesis, and the transcription factors GATA-2 and GATA-3. These genes have been reported to be only highly expressed in preadipocytes (Lee et al., 2003; Tong et al., 2005). Second, we used RT-PCR to confirm the data obtained in the microarrays for FAS, Clic4 and CPT-2. We observed in this study that, several proteins known to be overexpressed by differentiating 3T3-L1 cells (Welsh et al., 2004) are also upregulated at the transcript level, e.g. Acrp30 and GPDcyt. Third, by using a proteomic approach, Kratchmarova et al found that several proteins are secreted by differentiating 3T3-L1 cells (Kratchmarova et al., 2002); we found that mRNA levels of some of these proteins, such as Acrp30, adipsin, Gsn, Hp, resistin and MMP-2, were also upregulated. By contrast, we observed that PEDF, known to be secreted mainly by undifferentiated preadipocytes (Kratchmarova et al., 2002), is downregulated during 3T3-L1 adipogenesis.

To highlight similarities and differences of gene expression profiles in differentiating cells and also in preadipocytes with impaired mitochondrial activity (cellular circumstances that in both cases lead to the accumulation of TG) we next analyzed changes in gene expression in preadipocytes incubated with AA for different times and compared these gene expression profiles with those observed during the differentiation of 3T3-L1 cells. For example, we found hat both AA and the adipogenic cocktail downregulate Pref-1 and upregulate some adipogenic genes involved in fatty-acid and sterol metabolism, such as FABP4/aP2, SCD1 and HSD. We found that many genes that are upregulated during in vitro adipogenesis (such as those encoding the transcription factors C/EBPα, PPARδ, PPARγ, RXRα and SREBP-1, the lipid-metabolizing enzymes DHAPAT, FAS, HSL and LPL1, or the adipokines AGT and resistin) are either unaffected or downregulated by AA. Furthermore, AA also induces the transcription of anti-adipogenic transcription factors, such as GATA-3 and CHOP-10 (an endogenous dominant negative protein that heterodimerizes with C/EBP family members and represses C/EBP-dependent transcription) (Ron and Habener, 1992). CHOP-10 overexpression in AA-treated 3T3-F442A cells has been previously reported and shown to prevent adipogenesis (Carriere et al., 2004). All together, these data suggest that AA does not induce the expression of classic adipogenic markers. It is therefore likely that the mechanisms leading to TG accumulation in response to a mitochondrial dysfunction are not the same as those described for cell differentiation into adipocytes, a conclusion that also emerged from our previous report (Vankoningsloo et al., 2005).

Transcripts for Gyk (glycerol kinase) were not detected during adipogenesis, probably because of their very low abundance. However, Gyk mRNA levels raise to detectable levels in cells incubated with AA, and Gyk is transiently overexpressed under this condition. Since glycerol kinase converts glycerol into glycerol-3-phosphate, a direct precursor of TG (Lin, 1977), it might also contribute to the accumulation of TG in 3T3-L1 cells treated with AA. Moreover, the downregulation of genes, such as CPT-1, that encode enzymes of the mitochondrial fatty acid β-oxidation, controlling the mitochondrial entry of free fatty acids, and MCAD, a fatty acyl dehydrogenase, is in agreement with our previous findings, showing that AA-induced TG accumulation could result, at least partly, from a decrease in fatty acid β-oxidation (Vankoningsloo et al., 2005). Interestingly, MCAD transcription is controlled by PPAR transcription factors (Gulick et al., 1994), and we have recently shown that the activity of PPARγ is decreased in cells treated with AA (Vankoningsloo et al., 2005). Cav-1 has recently been shown to associate with intracellular lipid droplets and to modulate, in combination with perilipin, both lipolysis and vesicle formation (Cohen et al., 2004). In this study, we showed that the Cav-1 gene is upregulated during 3T3-L1 cell adipogenesis and also (although to a lesser extent) during AA treatment. Therefore, these differencies in Cav-1 expression might also contribute to the fact that TG vesicles accumulating in 3T3-L1 cells incubated with AA are smaller compared with those in differentiating 3T3-L1 cells.

By using siRNA to disrupt CREB expression, we found that the transcription of several genes (CHOP-10, Clic4, GPDmit, HSP 60, PAI-1, TF, SCD1 and VEGF-A) might depend on AA-induced CREB activation. Furthermore, it is interesting that some of these genes contain between one and six putative CRE sites in their promoters [determined by in silico promoter analysis with the Data Base for Transcriptional Start Sites (http://dbtss.hgc.jp) and TF Search (http://molsun1.cbrc.aist.go.jp/research/db/TFSEARCH.html)]. However, in silico identification of potential consensus sequences for transcriptional regulators does not reveal any information concerning their biological relevance. To test the functionality of the CRE motifs, gene expression analysis using the DNA array was performed in 3T3-L1 cells that had been incubated for 48 hours with 1 mM dibutyryl cyclic AMP (db-cAMP), a component of the adipogenic cocktail and a well-known activator of the PKA-CREB pathway (Boissel et al., 2004; Chio et al., 2004). We first tested whether db-cAMP can activate the CREB-responsive luciferase reporter gene. A 2.8-fold increase was found after 48 hours of treatment (data not shown). Under these conditions, a significant upregulation of candidate genes was only found for VEGF-A and SCD1 (1.8-fold increase). Although the discrepancy of these two sets of data is unknown we suggest that, either not all CRE sites are functional or they require the contribution of other regulators as partners in the promoter to be fully active. These regulator sets might be different in db-cAMP- and AA-treated cells. It has been reported that, in liver, the cAMP-sensivity of several genes, such as tyrosine aminotransferase (TAT) or PEPCK depends on the cooperative action of CREB and HNF4α or CREB and C/EBPs, respectively (Nitsch et al., 1993; Roesler, 2000). A similar cooperation between CREB and HNF4α has also been shown for the transcriptional control of CPT-1 L (Louet et al., 2002). These data suggest that the molecular mechanisms involved in the cAMP-mediated induction of target genes is more complex than the simple presence of a CRE motif requiring a cooperation between several factors. In addition, as for the expression of VEGF-A (a gene for which in silico analysis failed to reveal any potential CRE sites in the promoter, despite the fact that it seems to depend on CREB expression because it is downregulated when CREB is silenced by specific siRNA), one can not exclude that the CREB-dependence of gene expression is an indirect effect.

Overexpression of GPDmit protein in AA-treated 3T3-L1 cells has been verified and GPDmit was considered a potential candidate linking CREB activation to TG accumulation in response to mitochondrial dysfunction. Indeed, GPDmit knockout mice display reduced adiposity and body weight, suggesting that this enzyme is involved in the control of TG synthesis and/or storage (Brown et al., 2002). However, cells transfected - either before or after a 5-day treatment with AA - with an efficient GPDmit-specific siRNA that inhibits GPDmit expression at the protein and transcript level still accumulate the same amount of TG (as determined by Oil Red O staining) in response to an 8-day treatment with AA, when compared with control cells (data not shown). These results suggest that GPDmit is not involved in the lipid deposition in cells with impaired mitochondrial activity.

CREB has also been described to control hepatic lipid metabolism by indirectly repressing PPARγ and by inducing PGC-1α (Herzig et al., 2003). PGC-1α is a co-activator of numerous transcriptional regulators and is involved in several cell functions such as mitochondrial biogenesis (Scarpulla, 2002), gluconeogenesis (Herzig et al., 2001), and lipid catabolism because it controls the expression of genes encoding mitochondrial fatty-acid-oxidation enzymes (Vega et al., 2000). Although PGC-1α expression could not be analyzed using the microrray, we monitored its expression in RT-PCR and western blots, and found that PGC-1α mRNA and protein levels remain stable - and even slightly decreased in cells incubated for 24 or 48 hours with 10 nM AA or the adipogenic cocktail (data not shown), two conditions that activate CREB in 3T3-L1 cells. The absence of CREB-dependent PGC-1α regulation might seem surprising. However, various transcription factors have been identified to control PGC-1 expression in tissues other than liver and among them, ATF-2 in brown adipocytes (Cao et al., 2004), MEF-2 in muscle tissue (Czubryt et al., 2003), and orphan nuclear receptor ERRγ in brown adipocytes (Wang et al., 2005). Furthermore, in cells transfected with a smart-pool of four siRNAs specific for PGC-1α [which reduce the expression of the gene by more than 50% (data not shown)] either before or during the treatment with AA, no modification was found in the accumulation of TG in 3T3-L1 preadipocytes treated with the metabolic inhibitor (data not shown).

Finally, the SCD1 gene might also be an interesting candidate to explain TG accumulation in preadipocytes with impaired mitochondrial activity. Indeed, this gene encodes the enzyme that catalyzes the Δ⁹-cis desaturation of fatty acids (e.g. palmitoyl-CoA and stearoyl-CoA to palmitoleoyl-CoA and oleoyl-CoA, respectively), altogether representing up to 60% of the fatty acids esterified into TGs in differentiated cells (Kasturi and Joshi, 1982). SCD1 is upregulated in AA-treated 3T3-L1 cells, is downregulated following the inhibition of CREB expression and contains a cAMP-dependent regulatory element in its promoter sequence (position -285 to -278 relative to the start of transcription) (Ntambi et al., 1988). Recently, Miyazaki et al. showed that a lipogenic diet failed to induce hepatic triglyceride synthesis in SCD1^(-/-) mice, despite the induction of FAS and GPAT (Miyazaki et al., 2001). The authors suggested that SCD1 activity would thus help to produce substrates within the vicinity of GPAT to aid in the efficient esterification of glycerol 3-phosphate for TG synthesis. Although the role of this enzyme in TG accumulation has been clearly established (Ntambi et al., 2002), its regulation and potential implication as an effector in the accumulation of TG in response to mitochondrial dysfunction is awaiting future studies in cells with modified SCD1 gene expression.

In conclusion, although the molecular mechanism(s) and target gene(s) by which activated CREB, a factor involved in many cellular processes and in the regulation of several hundred genes, leads to TG deposition in preadipocytes with impaired mitochondrial activity still await discovery, our study clearly extends the role of the ubiquitous CREB transcription factor. We bring some evidence for its contribution to lipid metabolism alterations in preadipocytes with mitochondrial dysfunction. We also show that the mechanisms leading to TG accumulation in response to mitochondrial dysfunction are different than those described for lipid accumulation observed during cell differentiation, which is triggered by a standard hormon cocktail; we identified new potential targets involved in the acquisition of the phenotype. All together, these data contribute to a better molecular understanding of the mechanisms leading to TG accumulation in response to mitochondrial dysfunction in various pathologies such as MSL or lipodystrophy syndromes.

3T3-L1 fibroblasts, purchased from the American Type Culture Collection (ATCC), were grown to confluence in Dulbecco's modified Eagle's high glucose (DHG) medium containing 4.5 g/l glucose (Invitrogen) and 10% fetal calf serum (FCS, GibcoBRL). Differentiation of 3T3-L1 cells was initiated at confluence (day 0) by addition of medium containing an adipogenic cocktail, DHG-L1 medium (DHG containing 1.5 g/l NaHCO₃), supplemented with 10% FCS, 5 μg/ml insulin (Sigma), 300 μM dibutyryl cyclic AMP (db-cAMP, Sigma) and 1 μM dexamethasone (Sigma). After 2 days, cells were transferred to adipocyte growth medium (DHG-L1 containing 10% FCS and 5 μg/ml insulin) and re-fed every two days. To induce a prolonged mitochondrial inhibition, confluent cells (day 0) were incubated in DHG-L1 supplemented with 10% FCS and AA) or chloramphenicol (Sigma). When needed, medium was replaced every other day with DHG-L1 containing the above supplements at the same concentration. TG accumulation in cells was monitored by Oil Red O staining as described previously (Vankoningsloo et al., 2005). Briefly, cell monolayers were washed with PBS and then fixed for 2 minutes with 0.5 ml 3.7% paraformaldehyde (Sigma) in PBS. Oil red O (Sigma) was added for 30 minutes at room temperature and cells were washed twice with PBS. TGs were visualized by light phase contrast microscopy and quantitative determination was obtained by measuring the absorbance of cell monolayers at 490 nm in a spectrophotometer (Ultramark, Biorad).

3T3-L1 cells cultured in 75 cm² flasks or in 12-well plates (both Corning) were rinsed with PBS and lysed in respectively 1 ml or 150 μl of cold lysis buffer (20 mM Tris pH 7.4, 150 mM NaCl, 1 mM EDTA, 1% Triton-X100) containing protease inhibitors (Roche) and phosphatase inhibitors (1 mM Na₃VO₄, 5 mM NaF, 10 mM p-nitrophenylphosphate, 10 mM β-glycerophosphate). Clear cell lysates were prepared and protein contents was determined by the BCA method (Pierce).

Nuclear protein extractions in high-salt buffer were prepared as previously described (Chen et al., 1996). Briefly, cells seeded in 75 cm² flasks were incubated with 10 nM AA or the pro-adipogenic cocktail for 24 hours. At the end of the incubations, 3T3-L1 cells were incubated on ice for 3 minutes with 10 ml cold hypotonic buffer (HB; 20 mM HEPES, 5 mM NaF, 1 mM Na₂MoO₄, 0.1 mM EDTA) and harvested in 500 μl HB containing 0.2% NP-40 (Sigma), a protease inhibitor cocktail and phosphatase inhibitors. Cell lysates were centrifuged 30 seconds at 13,000 rpm in a tabletop centrifuge and sedimented nuclei were resuspended in 50 μl HB containing 20% glycerol, and protease and phosphatase inhibitors. Extraction was performed for 30 minutes at 4°C by addition of 100 μl HB containing 20% glycerol, 0.8 M NaCl, and protease and phosphatase inhibitors.Aliquots were frozen at -70°C and protein concentrations were determined according to Bradford (Bradford, 1976).

Samples corresponding to 20 μg or 40 μg of protein were prepared in Laemmli SDS loading buffer, resolved on 10% SDS-PAGE and transferred to PVDF membranes (Millipore). For detection of phosphorylated CREB (pCREB), membranes were blocked for 3 hours in TBS-T (20 mM Tris pH 7.4, 150 mM NaCl, 0.1% Tween-20) containing 5% bovine serum albumin (BSA, Sigma) and incubated for 16 hours (4°C) with a rabbit antibody that recognizes Ser133 pCREB (Upstate) at a 1:1000 dilution. For detection of CREB, GPDmit, PGC-1α, TBP and PARP, membranes were blocked for 3 hours in TBS-T containing 5% dry milk (Gloria) and incubated for 1 hour with either an anti-CREB (Rockland) rabbit antibody at a 1:1000 dilution, an anti-GPDmit rabbit antibody at a 1:4000 dilution, an anti-PGC-1α rabbit antibody at a 1:1000 dilution, an anti-TBP (Santa Cruz) rabbit antibody at a 1:1000 dilution or an anti-PARP (Pharmingen) mouse antibody at a 1:2000 dilution. The blots were washed and proteins were visualized with a horseradish peroxidase (HRP)-conjugated anti-rabbit or anti-mouse IgG antibody (Dako) and enhanced chemiluminescence (ECL) system (Pierce). Equal protein loading was checked by the immunodetection of TBP or PARP.

To detect the DNA-binding activity of a transcription factor we used the TransAM ELISA kit (Active Motif) according to the manufacturer's recommendations. The ELISA DNA-binding assays are based on multi-well plates coated with an oligonucleotide containing the consensus binding site of the transcription factor of interest. The presence of the DNA-bound transcription factor is then detected by specific antibodies and revealed by colorimetry. The specificity, selectivity and high reproducibility of these assays have been previously demonstrated for nuclear factor-κB (NF-κB) (Renard et al., 2001), hypoxia-inducible factor-1 (HIF-1) (Mottet et al., 2002) and peroxisome proliferator-activated receptor γ (PPARγ) (Vankoningsloo et al., 2005).

Briefly, 10 μg of nuclear proteins were incubated for 2 hours in a 96-well plate coated with a double-stranded oligonucleotide containing the consensus CRE site (TGACGTCA). Total CREB bound to DNA was detected with a rabbit antibody raised against CREB (Rockland) and presence of the phosphorylated form was determined with an anti-Ser133-pCREB rabbit antibody (Upstate). Colorimetric reaction was then performed with a HRP-conjugated anti-rabbit IgG antibody and absorbance was measured at 450 nm in a spectrophotometer (Biorad).

3T3-L1 cells were seeded at 40,000 cells/well on coverslips (Assistent) in 24-well plates; 3 days later cells were incubated or not for 24 hours with the adipogenic cocktail or with 10 nM AA. Cells were washed once with PBS, fixed and permeabilized with methanol-acetone (1:4, v/v) at -20°C for 10 minutes. Cells were then washed twice with PBS containing 1% BSA (Sigma) and incubated for 90 minutes at room temperature with a specific rabbit antibody raised against Ser133pCREB (Upstate) used at a 1:100 dilution in 1% BSA-PBS. Cells were washed twice in PBS with 1% BSA, incubated for 60 minutes with an Alexa-Fluor-568-conjugated anti-rabbit IgG antibody (Molecular Probes) at a 1:500 dilution in 1% BSA-PBS and processed for fluorescence confocal microscopy (TCS confocal microscope Leica).

3T3-L1 cells, seeded at 80,000 cells/well in 12-well plates 24 hours before transfection, were transiently co-transfected for 6 hours with 0.75 μg of a luciferase reporter construct driven by the α-inhibin promoter and 0.25 μg of an expression vector encoding β-galactosidase (Invitrogen) using SuperFect (5 μl/μg DNA) (Qiagen). The next day, cells were incubated for 24 hours with the adipogenic cocktail or 10 nM AA, or for 48 hours with 1 mM db-cAMP diluted in DHG-L1 medium. Cells were then harvested, luciferase activity was measured in cleared lysates using the commercial Reporter Assay System (Promega) and results were normalized against β-galactosidase activity.

To determine the putative role of CREB in the accumulation of TG, cells were seeded in 24-well plates and transiently transfected for 6 hours with 1 μg/well of plasmids encoding M1-CREB and K-CREB, or EGFP (Clontech) as a negative control, using SuperFect (5 μl/μg DNA). The next day, cells were induced to differentiate with the adipogenic cocktail or incubated with 10 nM of AA for 8 days before neutral lipid content was determined by Oil Red O staining.

We developed an `ADIPOCHIP', a low-density DNA array allowing gene expression analysis of 89 murine genes related to adipocyte differentiation in collaboration with Eppendorf (Germany) (see supplementary material Table S1 for the list of genes, and Fig. S1 for the array design). Results using reliable and validated arrays developed by Eppendorf were reported elsewhere (de Longueville et al., 2002; de Longueville et al., 2003; de Magalhaes et al., 2004; Debacq-Chainiaux et al., 2005). The method is based on a system with two arrays (a control and a test) on a glass slide and three identical sub-arrays (triplicate spots) per array. Except for C/EBPβ and C/EBPδ, no cross-hybridization was detected. The reliability of hybridizations and experimental data was evaluated using several positive and negative hybridization controls as well as detection controls spotted on the microarray.

3T3-L1 cells cultured in 75 cm² flasks were incubated for 2, 4, 6 or 8 days with 10 nM AA, and the adipogenic cocktail, or for 2 days with 1 mM db-cAMP. In some experiments, cells were transfected with 100 nM CREB-specific siRNA before being incubated for 4 days with 10 nM AA. At the end of the incubations, total RNA was extracted with the Total RNAgents extraction kit (Promega), quality was checked with a bioanalyzer (Agilent Technologies) and 20 μg were used for reverse transcription in the presence of biotin-11-dCTP (Perkin-Elmer) and Superscript II reverse transcriptase (Invitrogen), as described previously (de Longueville et al., 2002). Three synthetic poly(A)-tailed RNA standards (Eppendorf) were added at 10 ng, 1 ng or 0.1 ng per reaction) into the purified RNA to quantify the experimental variation introduced during labeling and analysis. For the kinetic profiles of gene expression, three independent experiments were performed in triplicate, providing hybridization on nine arrays that were carried out as described by the manufacturer and reported previously (de Longueville et al., 2002). Detection was performed with a cyanin 3-conjugated IgG anti-biotin (Jackson ImmunoResearch Laboratories). Fluorescence of hybridized arrays was scanned with the Packard ScanArray (Perkin-Elmer) at a resolution of 10 μm. To maximize the dynamic range of detection, the same arrays were scanned with different photomultiplier gains to quantify both the high-copy and low-copy expression of genes. The scanned 16-bit images were imported into ImaGene 4.1 software (BioDiscovery) to quantify signal intensities. The fluorescence intensity of each DNA spot (average intensity of each pixel present within the spot) was calculated by subtracting local mean background. A signal was only accepted when the average intensity after background subtraction was at least 2.5× higher than the local background around the spot. Intensity values of triplicate fluorescent signals were averaged and used to calculate the intensity ratio of reference and test.

The data were normalized in two steps. First, a correction was applied using a factor calculated from the intensity ratios of internal standards in the reference and test samples. The presence of the internal standard probes at different locations of the array allowed quantification of the local background and evaluation of the array homogeneity that is taken into account in the normalization. Furthermore, to consider the purity and quality of the mRNA, a second normalization step was performed based on the average of fluorescence intensities measured for a set of 9-16 housekeeping genes. The variance of the normalized set of housekeeping genes was used to generate an estimate of expected variance, leading to a predicted confidence interval for testing the significance of the ratios obtained. Ratios outside the 95% confidence interval were considered to be statistically significant, as determined by ANOVA (de Longueville et al., 2002; de Magalhaes et al., 2004).

After various treatments, total RNA was extracted using the Total RNAgent extraction kit (Promega). mRNA contained in 5 μg total RNA was reverse transcribed using SuperScript II reverse transcriptase (Invitrogen) according to the manufacturer's instructions. Forward and reverse primers, for C/EBPβ (FP: GGTTTCGGGACTTGATGCAA, RP: GCAGGAACATCTTTAAGGTGATTACTC); C/EBPδ (FP: CCGCCCGAATCGCTAGT, RP: GCAGTCCAGTGCCCAAGCT); Clic4 (FP: AGAGCCCACAGCAAGCATTCT, RP: ATCAGCCGCATGGAGACATC); CPT-2 (FP: CCTGATGGCTTTGGCATTG, RP: GGGCATTGCGTCCTGAGTA); FAS (FP: GTGAAGAAGTGTCTGGACTGTGTCAT, RP: TCGCTCACGTGCAGTTTAATTG); GPDmit (FP: GCAGCTGATGAGCGCAGTT, RP: TCCAAGTTCTCCTCGGCAGTT); IL-6 (FP: CCTAGTGCGTTATGCCTAAGCA, RP: TCGTAGAGAACAACATAAGTCAGATACCT); leptin (FP: GATCCCACGTGCCACAGTCT, RP: GGAACAAGCCATAGTGCAAGGT); PGC-1α (FP: CGGATTGCCCTCATTTGATG, RP: GAGGAAGGACTGGCCTCGTT) and TBP (FP: CAGTTACAGGTGGCAGCATGA, RP: TAGTGCTGCAGGGTGATTTCAG) were designed using the Primer Express 1.5 software (Applied Biosystem). Amplification reaction assays contained 1× SYBR Green PCR Mastermix (Applied Biosystem) and primers (Eurogentech) at the optimal concentrations. A hot start at 95° C for 5 minutes was followed by 40 cycles at 95°C for 15 seconds each, terminated at 65°C for 1 minute, using an ABI PRISM 7000 SDS thermal cycler (Applied Biosystem). TBP was used as the reference gene for normalization and relative mRNA steady-state level quantification. Melting curves were generated after amplification and data were analyzed using the thermalcycler software. Each sample was tested in duplicate.

siRNA transfection experiments were performed with double-stranded RNA designed and synthetized by Eurogentec. Out of three siRNA tested, a CREB-specific siRNA sense-orientation strand with the following sequence (5′-UACAGCUGGCUAACAAUGGdTdT-3′) was selected. To investigate the potential contribution of GPDmit and PGC-1α, some experiments were performed with a GPDmit-specific siRNA (Eurogentec: 5′-UCAGCUCCGUUGCCUAUCAdTdT-3′) or with a smart-pool of four specific siRNA for PGC-1α (Dharmacon). Cells were transfected with the siRNA delivery reagent JetSI™ (Eurogentec) at 3 μl/μg of siRNA according to the manufacturer's instructions. Transfection efficiency in cells plated on coverslips was determined with fluorescein isothiocyanate (FITC)-labeled siRNA and evaluated by cell counting using a confocal microscope (Leica) to be 90-95% after 24 and 48 hours.

Efficiency of RNA interference on CREB, GPDmit and PGC-1α expression was determined by either western blot analysis or by RT-PCR with specific primers. 3T3-L1 cells were seeded in 6-well plates at 200,000 cells/well 24 hours before being transfected for 4 hours by using JetSI with 10, 20, 50 or 100 nM CREB siRNA, 100 nM GPDmit siRNA, or 100 nM siRNA PGC-1α. Medium was replaced and gene silencing was verified 48 hours post-transfection. The effect of the disrupting the expression of CREB, GPDmit or PGC-1α by siRNA on AA-induced TG accumulation was analyzed as followed: 3T3-L1 cells (70-80% confluent) seeded in 24-well plates were treated or not for 5 days with 10 nM AA and then transiently transfected with 100 nM siRNA or an equivalent amount of JetSI alone. After transfection, cells were incubated for 3 and 8 days with or without 10 nM AA, and TGs were stained with Oil Red O. To identify genes potentially regulated by CREB that might play a role in the cellular accumulation of TG in response to AA, CREB expression was disrupted with siRNA. 3T3-L1 cells were plated in 75 cm² flasks at 50% confluence 3 days before transfection in the presence of 100 nM siRNA or were incubated with JetSI. After 24 hours, cells were incubated for 4 days with 10 nM AA before extraction of total RNA, reverse transcription of mRNA, biotin-labeling of cDNA and hybridization on microarrays.

T.A. is a research associate of the FNRS (Fonds National de la Recherche Scientifique, Belgium). S.V. is a recipient of a doctoral fellowship from Fonds pour la Recherche dans l'Industrie et l'Agriculture (FRIA). We are very grateful to Michael E. Greenberg (Harvard Medical School, Boston, MA, USA) for the constructs encoding dominant negative forms of CREB (K-CREB and M1-CREB). We also want to thank Joachim M. Weitzel (Universitätsklinikum, Hamburg-Eppendorf, Germany) for the anti- GPDmit antibody and Daniel P. Kelly (Center for Cardiovascular Research, Washington University School of Medicine, St Louis, MI, USA) for the PGC-1α antibody. This text presents results of the Belgian Programme on Interuniversity Poles of Attraction (PAI5/02) initiated by the Belgian State, Prime Minister's Office Science Policy Programming. We also thank the French speaking community governement for funding through an ARC (Action de Recherche Concertée). The scientific responsibility is assumed by the authors.