Mild mitochondrial uncoupling induces 3T3-L1 adipocyte de-differentiation by a PPARγ-independent mechanism, whereas TNFα-induced de-differentiation is PPARγ dependent

Obesity-related increase in body fat is a very common cause of insulin resistance and a major risk factor for the development of type 2 diabetes and metabolic syndrome (Lewis et al., 2002; Savage et al., 2007). Recent studies have demonstrated that a loss of just 5-10% of the body mass would be sufficient to improve dyslipidemia, hypertension and insulin resistance (Jones et al., 2007). Therefore, numerous areas of research have been developed to understand the metabolism of adipose tissue and to look for new strategies to fight obesity. However, although numerous studies have been published on mechanisms leading to adipogenesis (Fajas, 2003; Rajkumar et al., 1999; Rosen, 2002; Wu et al., 1995) and adipocyte differentiation (Gaillard et al., 1991; Gregoire, 2001; Rosen and MacDougald, 2006; Wabitsch et al., 1996), little data exist on mechanisms that could be potentially involved in adipocyte `de-differentiation'. Mature adipocyte de-differentiation has been defined as the acquisition of a more primitive phenotype and gain of cell proliferative ability (Matsumoto et al., 2008) as well as an elegant strategy to limit or reduce the adipose tissue mass and triglyceride (TG) content in adipocytes (Harper et al., 2001; Nawrocki and Scherer, 2005). In this study, we will refer to de-differentiation as a partial loss of TG content and downregulation of adipogenic markers and effectors.

The de-differentiation program has been mainly studied for TNFα, a pro-inflammatory cytokine secreted by macrophages and white adipocytes. Among the major mechanisms that have been reported to explain TG loss in adipocyte treated with this cytokine, the stimulation of lipolysis (Green et al., 1994), a process that involves the hyperphosphorylation of perilipin A mediated by both p44/p42-activated mitogen activated protein kinases (MAPK) (Ryden et al., 2004) and protein kinase A (PKA) (Zhang et al., 2002), seems to be one of the most important. Beside, TNFα suppresses the expression of many genes involved in lipid and/or glucose metabolism such as Glut4, HSL (hormone-sensitive lipase), LCFACoAS (long-chain fatty acyl-CoA synthase), Acrp30 (adipocyte complement-related protein of 30 kDa) and those coding for transcription factors such as CCAAT/enhancer binding protein-alpha (C/EBPα), retinoid X receptor-alpha (RXRα), and peroxisome proliferator-activated receptor gamma (PPARγ) (Ruan et al., 2002). The transcriptional activity of PPARγ is also targetted by this cytokine, probably mediated by the activation of other transcription factors such as NFκB, c-Jun and ATF 2 (Ruan et al., 2002; Tesz et al., 2007) (Cawthorn and Sethi, 2007).

However, specific mitochondrial dysfunction as a result of uncoupling of oxidative phosphorylation has also been strongly linked with lipid metabolism. In fact, it is known that downregulation of fatty acid (FA) synthesis can be observed in both brown and white adipose tissue of transgenic aP2-Ucp1 mice (Rossmeisl et al., 2000). In general, this phenotype suggests that the main function of respiratory uncoupling and uncoupling proteins (UCPs) in white adipocytes might be the attenuation of lipid accumulation, mainly due to a decrease in FA synthesis, increase of free fatty acid (FFA) β-oxidation, impaired noradrenaline-stimulated lipolysis and modulation of hormonal control of lipid metabolism as well as the stimulation of mitochondrial biogenesis, UCP-2 over-expression and Gαi abundance modification (Rossmeisl et al., 2004). Additionally, the forced expression of UCP-1 in 3T3-L1 preadipocytes limits the differentiation efficiency (Si et al., 2007) by increasing protein stability and decreasing expression of genes encoding proteins involved in energy-consuming processes (Senocak et al., 2007). Although there is good evidence in favour of a major role for mitochondrial uncoupling in lipid metabolism leading to a significant decrease in TG content of adipocytes, the impact on adipogenic markers and effectors is still poorly understood.

In this study, we first analyzed modifications in the expression of genes encoding adipogenic markers and effectors during 3T3-L1 adipocyte de-differentiation showing that expression of genes encoding adipogenic transcription factors such as PPARγ and C/EBPs or enzymes involved in lipid and/or glucose metabolism are differentially expressed in carbonyl cyanide (p-trifluoromethoxy)-phenylhydrazone (FCCP)-treated cells. In addition, a decrease in PPARγ DNA-binding activity in FCCP-treated adipocytes correlated with a reduction in the protein abundance. Furthermore, although rosiglitazone and 9-cis retinoic acid (PPARγ and RXR ligands, respectively) prevented TG decrease in TNFα-incubated cells, they did not affect the TG content in FCCP-treated adipocytes. Moreover, the activation of PPARγ and RXR by these ligands had no effect on several PPARγ target genes in FCCP-treated cells. In addition, although a downregulation of lipid synthesis is observed in both conditions, FFA β-oxidation is only strongly increased in TNFα-treated cells but not in FCCP-incubated cells. Finally, although lipolysis seems to be stimulated in FCCP-treated cells, this effect might be mainly mediated by a decrease in the abundance of perilipin A protein at the lipid droplet surface as both HSL and adipose triglyceride lipase (ATGL) are downregulated. Altogether, these results show that although mitochondrial uncoupling induces a decrease in TG in 3T3-L1 cells that is comparable with the effect of TNFα, molecular mechanisms and cell signalling leading to this TG loss seem to be different.

To study the effects of mitochondrial dysfunction induced by uncoupling, 3T3-L1-differentiated adipocytes were exposed chronically to FCCP. First, the effect of low concentration FCCP (0.5 μM) on the mitochondrial membrane potential (Δψm) was examined using the TMRE probe and FACS analysis (Fig. 1A). Whereas at 10 μM, the uncoupler clearly reduced the Δψm within 15 min, at 0.5 μM FCCP it only triggered a slight decrease. Second, the effect of mitochondrial uncoupling on the amount of intracellular TG was examined. Oil red O (ORO) staining of cell monolayers and spectrophotometry showed that FCCP causes a reduction in the TG content in adipocytes in a time- and concentration-dependent manner (Fig. S1 in supplementary material). The decrease in lipid content was comparable with the TG loss observed in adipocytes challenged with 10 ng/ml TNFα (Fig. 1B). In these conditions, for 3- and 6-day treatments, we found a statistically significant TG loss of 29% and 39% or 38% and 48% for FCCP- and TNFα-treated cells, respectively. This loss of TG is reflected in a decrease in the number and size of lipid vesicles as visualized by ORO staining and phase-contrast microscopy (Fig. 1C). In order to confirm these data we monitored the release of glycerol (Fig. 1D) and the TG content (Fig. 1E) using an enzymatic assay. Interestingly, as shown in Fig. 1D, glycerol release was enhanced in adipocytes treated with either FCCP or TNFα even though isoproterenol, an agonist of both α- and β-adrenergic receptors used as a positive control, is clearly the strongest inducer of lipolysis.

The specificity of the uncoupling effect on the reduction in TG content in adipocytes triggered by FCCP was confirmed by testing the effect of 2,4 dinitrophenol (DNP), another mitochondrial uncoupler, used at a low concentration (50 μM). We found that this molecule also caused a decrease in TG content (Fig. S2 in supplementary material).

We next analyzed and quantitatively compared modifications in gene expression during adipocyte de-differentiation triggered by these molecules. We took advantage of a low-density cDNA microarray to analyze gene expression of key molecular markers and effectors of adipogenesis (Vankoningsloo et al., 2006). Differentiated cells were thus treated for 24, 72 or 144 hours with 0.5 μM FCCP, 10 ng/ml TNFα or 10 μM isoproterenol and gene expression was analyzed in triplicate for each experimental condition.

We are aware that the concentrations used in our study for TNFα and isoproterenol – although based on previous studies (Gauthier et al., 2008; Jones et al., 2005; Kim et al., 2005; Park et al., 2008; Haemmerle et al., 2006; Souza et al., 1998) – are too high and thus supra-physiological. However, these two molecules were used as positive controls to compare the effects triggered by FCCP, and the concentrations were selected to maximize expected biological responses while minimizing toxic effects.

Among the differentially expressed genes identified in FCCP-treated adipocytes, 10 genes were significantly downregulated after a 24-hour treatment (Table S1; Fig. S3A in supplementary material). Many of these genes have a major role in lipid metabolism such as CL (ATP-citrate lyase), CPT-2 (carnitine palmitoyl transferase-2), DHAPAT (dihydroxyacetone-phosphate acetyltransferase), FAS (fatty acid synthase), GPAT (mitochondrial glycerol-3-phosphate acyltransferase), HSL and caveolin-1B (Cohen et al., 2004). Interestingly, we also found that the gene for C/EBPα, a major transcription factor known to maintain the differentiated phenotype of adipocytes (Rosen, 2005; Rosen and MacDougald, 2006), is repressed in cells with uncoupled mitochondria whereas expression of the PPARγ gene, another key factor in adipocyte differentiation that is mainly active during the mid and latest periods of differentiation (Cowherd et al., 1999), was not modified. For this essential transcriptional regulator in adipogenesis, there was a weak downregulation only after 72 hours of incubation.

The downregulation of DHAPAT, GPAT and FAS, was sustained during the whole de-differentiation program (Fig. S3A-C in supplementary material). This downregulation was even stronger after 72 and 144 hours of FCCP treatment, and was accompanied by a repression of specific genes involved in the synthesis of monounsaturated fatty acids such as Scd 2 and Scd 1 (Chu et al., 2006; Miyazaki et al., 2005). In addition, genes involved in FFA β-oxidation, such as CPT-2 and MCAD as well as genes involved in lipolysis such as HSL and perilipin A, are also characterized, respectively, by a sustained or transient downregulation. The expression of other genes encoding adipogenic transcription factors such as C/EBPδ or SREBP1 was also decreased at the beginning of the de-differentiation program.

Among the few upregulated genes identified, CHOP-10, encoding an endogenous dominant negative mutant of C/EBPβ (Darlington et al., 1998), showed an important over-expression within the first 72 hours of mitochondrial uncoupling. Indeed, in these conditions, a specific 4- and 2.6-fold upregulation was found after 24 and 72 hours of treatment, respectively. This upregulation was confirmed by real-time RT-PCR and reflected by an increase in the abundance of the protein in the nucleus (data not shown).

Interestingly, mitochondrial uncoupling is also able to modify the expression of genes encoding specific adipokines, because a 3-day treatment induces a transient downregulation for adipsin and resistin and an important upregulation for PAI-1.

Although the expression level of some genes could not be quantified and are reported qualitatively, they clearly displayed changes in expression. For instance, it is worth emphasizing that genes encoding other important adipokines such as Acrp30 and leptin (Rosen and Spiegelman, 2006) or genes involved in glucose uptake (such as Glut 4) (Kanzaki, 2006) or glyceroneogenesis (PEPCK 1) (Tontonoz et al., 1995) were transiently downregulated in response to FCCP (Table S1 in supplementary material). These data were confirmed for a 3-day treatment by real-time RT-PCR for Acrp30, Glut 4 and PEPCK 1 (Fig. S9A in supplementary material).

The observed effects of FCCP on gene expression are obviously dependent on the mitochondrial uncoupling as they can be mimicked and reproduced by the presence of another mitochondrial uncoupler, DNP (50 μM) (Table S2 in supplementary material).

After 24 hours, all the genes involved in lipid metabolism that were significantly downregulated in the FCCP-treated cells, were also found to be downregulated in cells incubated with TNFα (Table S3 and Fig. S4A in supplementary material). Furthermore, for these genes the downregulation was sustained during the whole de-differentiation program. A profile of transient downregulation was found for Scd 1 and Scd 2 as also observed in FCCP-treated cells. Beside this set of genes, we can observe that TNFα also induces the downregulation of other important genes involved in lipid/glucose metabolism such as UCP2, caveolin-2, Glut 4, LPL1, FABP4 and perilipin A. Interestingly, the expression of PPARγ was significantly downregulated after 3 days and was even lower after 6 days. However, CHOP-10 transcript and the abundance of the protein in the nucleus, was not modified in TNFα-treated cells. CHOP-10 upregulation, found in FCCP-incubated adipocytes, is thus a stress-specific process, also triggered by a mild mitochondrial uncoupling (Fontanier-Razzaq et al., 1999). Time-course profiles of gene expression for representative genes are illustrated in Fig. 2A-D. Surprisingly, the expression of Pref-1, a repressor of adipogenesis (Smas and Sul, 1993) was downregulated in cells treated with TNFα for 72 and 144 hours, while no modification was found in cells responding to the mitochondrial uncoupler. Furthermore, Acrp 30, Glut 4, leptin and PEPCK1 were also downregulated in cells incubated with TNFα (Table S3 in supplementary material). Validation of these data was performed by real-time RT-PCR for Acrp30, Glut 4 and PEPCK1 in cells treated for 3 days with the cytokine (Fig. S9A in supplementary material).

It is worth mentioning that the effect of isoproterenol on gene expression in differentiated adipoytes was totally different to the effect of FCCP or TNFα (Table S4 and Fig. S5 in supplementary material).

As we showed that transcripts for PPARγ and C/EBPα are differentially modified in `de-differentiating' adipocytes, we next decided to analyze the activity status of these transcription factors. Their activity was determined by the DNA-binding capacity of the proteins to a specific consensus sequence using a colorimetric assay (Active Motif). We found that after 24 hours of treatment, the amount of bound PPARγ was significantly decreased in cells with uncoupled mitochondria when compared with differentiated cells and remained low after 72 hours. In TNFα-treated cells, the decrease in the PPARγ DNA-binding activity was even stronger whereas isoproterenol triggered no reduction (Fig. 2E). A similar profile was obtained for C/EBPα DNA-binding activity (Fig. 2F).

Furthermore, in adipocytes two PPARγ isoforms exist (PPARγ1 and PPARγ2) differing only in their amino-terminal A/B domain: PPARγ2 contains an additional 30 amino acids and is specifically expressed in adipocytes, being the essential regulator of adipogenesis (Rosen and Spiegelman, 2001). As the DNA-binding activity of PPARγ might be dependent on the abundance of the protein or the phosphorylation status of residues in the DNA-binding domain (Burns and Vanden Heuvel, 2007; Hauser et al., 2000), we performed western blot analysis on nuclear extract from differentiated cells incubated with or without FCCP or TNFα (Fig. 2G, Fig. 3C). PPARγ appears as a double band; the upper and the lower band corresponding to PPARγ2 and PPARγ1 isoforms, respectively. The results obtained show a decrease in the nuclear abundance of both PPARγ isoforms in cells incubated with either FCCP or TNFα for 24 hours, which was even stronger after 72 hours. Based on these results, it seems that the decrease in DNA-binding activity of PPARγ found in FCCP- and TNFα-treated cells results from a decrease in the abundance of the protein. Nevertheless, as mentioned before, several different mechanisms are involved in the decrease in PPARγ DNA-binding activity observed in response to TNFα. One of these mechanisms is the phosphorylation of PPARγ by the extracellular signal-regulated kinase (ERK1/2) MAPK pathway (Cawthorn and Sethi, 2007). However, western blot analysis of the phosphorylated form of ERK1/2 showed an activation of ERK1/2 in cells incubated with TNFα whereas no effect was observed in FCCP-treated cells (data not shown).

In order to test the effect of mitochondria uncoupling on the transcriptional activity of PPARγ, differentiated cells were transiently transfected with a PPARγ driven luciferase reporter construct. As observed in Fig. 3A, the luciferase activity is significantly decreased in adipocyte treated with FCCP for 24 and 72 hours. A strong decrease in PPARγ activity can also be observed in TNFα-treated cells.

Based on these results, we next decided to determine whether this reduced activity could be overridden by the action of PPARγ and RXR ligands. A luciferase assay was thus performed, in transfected cells incubated for 24 and 72 hours with TNFα or FCCP in the presence or absence of ligands for both PPARγ (rosiglitazone) and RXR (9-cis retinoic acid) transcription factors (Fig. 3A). Surprisingly, whereas ligands in TNFα-treated cells induced a 2.3-fold increase in PPARγ activity compared with differentiated cells and a 5.1-fold increase when compared with TNFα-treated cells alone, the increase in PPAR activity in FCCP-treated cells triggered by the ligands was more modest, with an increase of 1.3-fold compared with differentiated cells and a 2.5-fold increase when compared with FCCP-treated cells. Similar results were also obtained after 72 hours of incubation.

Based on the previous results, we next wanted to determine whether or not PPARγ plays a role in the de-differentiation of FCCP-treated cells. These ligands were added, at the same concentrations, during cell treatment in the presence of FCCP or TNFα. At the end of the incubations, cells were stained with ORO and quantitative analyses showed that, after 6 days, loss of neutral lipids induced by TNFα could be totally prevented in the presence of both ligands, but the inhibitory effect was not observed in FCCP-incubated cells (Fig. 3B; Fig. S6 in supplementary material).

We next analyzed gene expression profiles in adipocytes incubated for 3 days with either FCCP or TNFα in the presence or absence of rosiglitazone and 9-cis retinoic acid.

Interestingly, the gene expression profile was not significantly modified in cells incubated with FCCP in the presence of ligands when compared with the gene expression pattern found for cells treated with FCCP alone (Table S5 in supplementary material). However, the PPARγ target genes that are downregulated in TNFα-treated adipocytes are dramatically upregulated in cells incubated with TNFα in the presence of ligands (Fig. S7 and Fig. S8 in supplementary material). In these conditions, we found that the expression of several PPARγ-target genes was modified in the presence of ligands when compared with modifications found for TNFα-treated cells (Fig. S7B in supplementary material), whereas very few genes were affected by the presence of ligands in adipocytes alone (Fig. S7A in supplementary material).

We next decided to look for the abundance of PPARγ by western blot analysis in adipocytes incubated with TNFα or FCCP in the presence or absence of ligands (Fig. 3C). Interestingly, the abundance of both isoforms was strongly decreased in both TNFα- and FCCP-treated cells exposed to ligands for 24 or 72 hours when compared with TNFα and FCCP-treated cells alone. The stronger reduction occurred in the PPARγ2 isoform. This decrease in PPARγ in cells exposed to ligands has already been observed and described as a feedback mechanism for balancing the transcriptional activity of PPARγ in response to activating ligands that trigger the degradation of the protein by proteasome 26S (Hauser et al., 2000). In our experimental conditions after 72 hours the degradation of the protein seems to be stronger for FCCP-treated cells incubated with both ligands.

Based on the modification observed in the expression of several genes involved in metabolic pathways, such as FAS, GPAT, MCAD, CPT-2, DHAPAT and CL, by mitochondrial uncoupling we decided to compare the relative rates of FFA β-oxidation and lipid synthesis. For the FFA β-oxidation assay, cells were prepared as described in the Materials and Methods. There was no significant difference in the relative rates of FFA β-oxidation (¹⁴CO₂ release) between FCCP-treated cells and differentiated cells at both times (24 and 72 hours), but a huge increase occurred in adipocytes treated with TNFα (Fig. 3D).

Regarding the lipid synthesis assay, [³H]acetate was added as described in the Materials and Methods. We found that the incorporation of [³H]acetate into lipids in adipocytes treated with TNFα or FCCP was significantly decreased when compared with differentiated cells at 24 and 72 hours (Fig. 3E).

These data suggest that the TG reduction in cells, triggered by FCCP, most probably reflects – at least partly – a downregulation of lipid synthesis rather than an upregulation of FFA β-oxidation despite an acute lipolysis (Fig. 1D).

The lipases HSL and ATGL (adipose triglyceride lipase) as well as the protein perilipin have been described as three major effectors in lipolysis (Granneman et al., 2007). Knowing that the transcript abundance levels of HSL and perilipin A are downregulated in both TNFα- and FCCP-treated cells, real time RT-PCR was performed for ATGL in order to determine the relative abundance of this transcript (Fig. 4A). After 24 and 72 hours, TNFα and FCCP exert a negative regulation on the transcription of this gene whereas isoproterenol triggers an increase.

We next evaluated the effect of TNFα, FCCP and isoproterenol on the expression of these genes at the protein level using western blot analysis (Fig. 4B). A decrease in the abundance of HSL and ATGL was observed in the cells incubated with TNFα or FCCP after 72 hours, but was unchanged in cells incubated with isoproterenol. For perilipin A we observed a slight decrease in the abundance in the cells incubated with TNFα and FCCP after 24 and 72 hours, whereas no modification was found in isoproterenol-treated cells. Although these observations are in agreement with the transcript abundance profile obtained for a 3-day treatment in the presence of TNFα or FCCP, they cannot explain the stimulation of the lipolysis induced by these molecules. It is known that beside its abundance, the phosphorylation status of HSL at Ser563 and Ser660 plays a major role in the activation of the enzyme (Kraemer and Shen, 2002). Therefore, we decided to determine its phosphorylation status. The abundance of the phosphorylated form of the HSL was monitored for short times as previous studies showed that the activation of PKA in adipocytes treated with β-adrenergic hormones is very rapid (Carmen and Victor, 2006; Granneman et al., 2007). As shown in Fig. 4C, the phosphorylation status of HSL at both Ser563 and Ser660 was not modified by the FCCP treatment, whereas TNFα caused only a transient phosphorylation of Ser660. By contrast, an important increase in the phosphorylation status of both serine residues could be observed in cells stimulated by isoproterenol.

Lipolysis can be modulated by the accessibility of lipases to TG, a process known to be regulated by perilipin (Granneman et al., 2007). In order to determinate whether or not the localization of HSL and perilipin A changed in cells engaged in the de-differentiation program, immunostaining for these proteins and confocal microscopy observations were performed. We could not observe any significant modification in the distribution of HSL in the cytoplasm in response to FCCP or TNFα, indicating that the translocation of this lipase to the TG vesicles is not as strong as in cells treated with isoproterenol (Fig. 4D). However, we demonstrated that TNFα, FCCP and isoproterenol reduced perilipin A abundance at the surface of lipid droplets when compared with untreated adipocytes (Fig. 4E). These results strongly suggest that the reduction of perilipin A levels at the surface of lipid droplets could be an important factor in the stimulation of lipolysis in FCCP-treated adipocytes.

In this work, we have clearly shown that mild mitochondrial uncoupling, triggered by FCCP used at a low concentration, in differentiated adipocytes is able to reduce the intracellular TG content after several days. Mild mitochondrial uncoupling achieved in our experimental conditions results in a slight decrease in mitochondrial membrane potential (Fig. 1A) and triggers glucose uptake but does not modify the ATP content when compared with differentiated cells, even if more glucose is consumed for glycolysis (data not shown). We also show that the cytokine TNFα induces similar effects. However, mechanisms leading to the FCCP-induced reduction in TG are clearly different than the PPARγ-dependent mechanisms already reported for adipocytes treated with TNFα (Xing et al., 1997; Cawthorn and Sethi, 2007).

It is unlikely that the effects observed in FCCP-treated adipocytes result from undifferentiated adipoblast that might have (re)colonized the culture plate after the death and detachment of differentiated adipocytes in response to either TNFα or FCCP. Indeed, although these treatments trigger apoptotic markers, such as caspase-3-like activation and nuclear DNA fragmentation (data not shown), we were unable to show any increase in cell cycle proliferation index in adipocytes treated for 3 or 6 days with either FCCP or TNFα, as determined by a cell proliferation assay based on [³H]thymidine incorporation (Fig. S11 in supplementary material).

We also showed that in response to mitochondrial uncoupling, the decrease in endogenous TG vesicles found in adipocytes is accompanied by the hydrolysis of TG into glycerol and FFA. However, it is unlikely that lipolysis completely explains the reduction of TG in adipocytes incubated with FCCP or TNFα, as adipocytes challenged with isoproterenol displayed a stronger glycerol release while the intracellular lipid content was maintained at a higher level than in cells challenged with FCCP (Fig. 1D,E). We thus hypothesized that gene expression modifications and/or changes in the activity of adipogenic transcription factors might explain TG loss in FCCP-treated cells.

Indeed, we found that even if the relative transcript abundance of PPARγ is only slightly modified in adipocytes treated with FCCP, at the protein level, a 24 hour-treatment with FCCP triggers a decrease in PPARγ abundance, as well as a diminution in its DNA-binding activity. A decrease in its transcriptional activity was also found in FCCP- or TNFα-treated cells, which confirms the diminution of the protein function (Fig. 3A). A decrease in expression (Fig. S3 and S4 in supplementary material) and DNA-binding activity (Fig. 2F) of C/EBPα was also found in both FCCP- and TNFα-treated cells, probably enhancing the repression of specific adipogenic markers.

Also, FCCP treatment lead to the downregulation of several genes involved in FA synthesis, such as CL, FAS, Scd 1 and Scd 2, and TG synthesis such as DHAPAT. Moreover, whereas the expression of CPT-1, which encodes the rate-limiting enzyme controlling mitochondrial β-oxidation, is not modified (Nada et al., 1995), the expression of CPT-2 and MCAD was constantly downregulated in both TNFα- and FCCP-treated cells, but only FFA β-oxidation is found to be stimulated in FCCP-treated cells (Fig. 3D). It is therefore probable that one of the mechanisms leading to the decrease in TG content in response to FCCP, results from the downregulation of a highly ATP-consuming process such as lipid synthesis (Fig. 3E), rather than from an upregulation of FFA β-oxidation.

It might seem surprising that FFA β-oxidation is not modified in the presence of low concentration in FCCP, as it is known that mitochondrial uncoupling leads to stimulation of FFA β-oxidation in several models (Li et al., 1999; Rossmeisl et al., 2002). This apparent discrepancy might be explained by the fact that Rossmeils and collaborators demonstrated that over-expression of UCP1 from the aP2 gene promoter in mice induces mitochondrial biogenesis in unilocular adipocytes. This mitochondrial biogenesis could increase (and be necessary for) the oxidation of FFA observed in these studies (Rossmeisl et al., 2002). However, in our experimental conditions we do not see any modification in the abundance of markers of mitochondrial biogenesis such as the transcription factors NRF-1, mtTFA/Tfam and the co-activator PGC-1α, when adipocytes are treated with 0.5 μM FCCP (our unpublished data). This discrepancy could also be explained by the use of a low concentration of FCCP, as Rossmeils and collaborators found that the concentration of the uncoupler 2,4 dinitrophenol (DNP) had to be over 200 μM to increase FFA β-oxidation (Rossmeisl et al., 2000).

Our results are also in total agreement with results recently obtained by Si et al. (Si et al., 2007) using 3T3-L1 cells that ectopically expressed UCP1. These authors show that oxygen consumption and FFA β-oxidation were minimally affected by mitochondrial uncoupling triggered by UCP-1 overexpression.

By contrast, several PPARγ target genes such as Cav1B, encoding a protein known to modulate lipid vesicle formation and lipolysis (Cohen et al., 2004) as well as HSL and ATGL (Dalen et al., 2004; Deng et al., 2006) known to play an important role in lipolysis, were found to be downregulated by the FCCP treatment. Moreover, a direct correlation with the abundance of their corresponding proteins, specifically for HSL, ATGL were also found in FCCP-treated adipocytes. We also found that the phosphorylation status of the principal PKA-stimulated HSL serines has unchanged in FCCP-treated cells (Fig. 4). Taken together these data, might suggest that lipolysis is compromised in FCCP-treated cells. However, the final balance on lipolysis stimulated in FCCP-treated cells is complex and difficult to establish as it should be noted that a reduction in the expression of perilipin A as well as a modification in the abundance of the protein at the lipid droplet surface, observed during FCCP treatment, might be sufficient to increase lipolysis (Deng et al., 2006).

In addition, several studies have demonstrated that TNFα reduces the expression of genes involved in preventing lipolysis such as Gαi, perilipin, Cav1B, ATGL (Kralisch et al., 2005; Ryden et al., 2004; Souza et al., 2003; Zhang et al., 2002) as well as increasing lipolysis in adipocytes by activating several signalling pathways such as ERK1/2, JNK, PKA, IKK (Cawthorn and Sethi, 2007; Ryden et al., 2002; Souza et al., 2003). More particularly, MEK1/2 has been directly involved in the activation of PKA and subsequent phosphorylation of HSL. These data correlated well with the increased glycerol release observed in TNFα-treated cells and suggest that mitochondrial uncoupling in 3T3-L1 adipocytes could also activate similar signaling pathways to induce lipolysis. However, we clearly observed that inhibition of MEK1/2 (Green et al., 1994; Zhang et al., 2002) or PKA, using specific inhibitors, did not prevent the loss of TG in adipocytes incubated with FCCP, whereas it significantly prevented TG reduction in TNFα-treated cells (Fig. S10A,B in supplementary material). This data was supported by the absence of modification in the phosphorylation and abundance status of the protein ERK1/2 observed in FCCP-treated cells (data not shown).

It is well known that thiazolidinedione (TZD), an antidiabetic drug with a high-affinity for PPARγ (Lehmann et al., 1995), activates gene expression in a coordinated manner, leading to an increase in net lipid partitioning into adipocytes (Olefsky, 2000; Schoonjans et al., 1996). In addition, it is now known that rosiglitazone can regulate the signalling pathways of FFA β-oxidation, TCA cycle, oxidative phosphorylation and FA catabolism in mature 3T3-L1 adipocytes (Wang et al., 2007). We thus wanted to address the role of the reduction in activity of this transcription factor in the TG loss in adipocytes incubated with the mitochondrial uncoupler. We found that rosiglitazone and 9-cis retinoic acid, are able to simulate PPARγ activity in both TNFα- and FCCP-treated cells and thus to compensate for the decrease in the activity of the transcription factor induced by both molecules. Nonetheless, the effect of these ligands is more pronounced in TNFα-treated cells inducing most of the genes involved in lipid/energy metabolism that are downregulated in adipocytes incubated with TNFα alone and protecting against the decrease in adipocyte TG in this condition.

This difference in the effect of the activation of the transcription factor could be explained, at least partly, by the dramatic decrease in the abundance of PPARγ protein in cells treated with FCCP in the presence of ligands, as shown by western blot analysis. Another explanation could be that, the differences in gene expression observed in response to the presence of ligands in adipocytes treated with TNFα or FCCP could also result from different regulators and/or transcription factors that in addition to PPARγ bind promoters of these genes. For example, in agreement with the literature (Ruan et al., 2002) we found that TNFα activated NFκB in adipocytes, whereas no effect was observed in FCCP-treated cells (data not shown). As NFκB has been shown to antagonize the transcriptional activity of PPARγ and that troglitazone (another PPARγ ligand) can inhibit the transcriptional activity of NFkB (Ruan et al., 2003), one can speculate that the inhibitory effect of NFκB on PPARγ activity might be relieved in the presence of PPARγ ligands.

Moreover, PPARγ is most probably not responsible for the TG loss observed in adipocytes subjected to mitochondrial uncoupling, because when FCCP-treated adipocytes were incubated with the ligands, TG loss was not prevented. It is thus possible that other transcription factors such as CHOP-10 (gadd153), C/EBPα and/or SREBP-1 that are also affected by mitochondrial uncoupling might be responsible for the de-differentiation of adipocytes facing a mitochondrial uncoupling.

In conclusion, our data clearly identify differentially expressed genes, transcription factor activities and lipid metabolic pathways, showing that signaling pathways and key factors responsible for the de-differentiation of adipocytes treated with FCCP might be totally different from those affected by TNFα treatment. Our data also suggest that a mild mitochondrial uncoupling could offer new clinical outlooks and potential targets for therapies required for fighting obesity and diabetes. However, further experiments are still necessary to assess the putative effect of mitochondrial uncoupling on other adipose cell lines or biopsies as well as in vivo approaches to evaluate the advantages, some of which have already been described in recent studies (such as decrease in ROS production, regulation of food intake, protection against cell death) (Maassen et al., 2008; Sullivan et al., 2004; Yamada et al., 2006) and disadvantages (such as FFAs release) (Frayn et al., 2008) of limiting TG accumulation in adipocytes.

Mouse 3T3-L1 preadipocytes from ATCC, were differentiated as previously described (Vankoningsloo et al., 2006). Cell differentiation lasted for 10 or 12 days to obtain more than 90% of differentiated cells, as estimated by phenotype analysis and Oil red O (ORO) staining. Cells were next incubated for an extra 1, 3 or 6 days in the presence of 10 ng/ml recombinant human TNFα (R&D Systems), 0.5 μM carbonyl cyanide p-(trifluoromethoxy)phenylhydrazone (FCCP; Sigma), or 10 μM isoproterenol hydrochloride (Sigma) diluted in DHG-L1 containing 10% fetal calf serum (FCS) (regular medium). Control differentiated adipocytes were maintained in DHG-L1 containing 10% FCS without insulin. During the de-differentiation program, media were replaced every 24 hours. When indicated, rosiglitazone (Alexis) or 9-cis retinoic acid (Biomol) were added to the de-differentiation program. TG accumulation in cells was monitored by ORO staining, as described previously (Vankoningsloo et al., 2005).

Differentiated adipocytes were trypsinized and permeabilized with 800 μg/ml digitonin as described previously (Baumruk et al., 1999). After permeabilization, cells were incubated in KCl buffer with 0.5 μM TMRE (tetramethylrhodamine, ethyl ester, perchlorate; Molecular Probes) with or without 10 ng/ml TNFα, 0.5 μM FCCP or 10 μM FCCP (as positive control) for 15 minutes on ice before fluorescence-activated cell sorting (FACS) analysis (Becton Dickinson).

Differentiated cells were incubated for 1, 3 or 6 days in regular medium with or without 10 ng/ml TNFα, 0.5 μM FCCP or 10 μM isoproterenol. Glycerol release was measured in the 24-hour-old conditioned medium as an index of lipolysis using the enzymatic INT kit (Sigma) according to the manufacturer's instructions. TG content was determined after lipid extraction in methanol–chloroform as previously described (Vankoningsloo et al., 2006) and quantified with the same INT kit. Results were then normalized for protein content, determined by the BCA method (Pierce).

Nuclear protein extracts in high salt buffer were prepared as previously described (Chen et al., 1996) from undifferentiated cells or cells differentiated for 12 days and then incubated for an extra 24 or 72 hours with or without 10 ng/ml TNFα, 0.5 μM FCCP or 10 μM isoproterenol. The DNA-binding activity of C/EBPα and PPARγ transcription factors was estimated using the TransAM ELISA assay (Active Motif) according to the manufacturer's instructions.

Sample preparation and gene expression analysis, using a low-density cDNA microarray (containing 89 capture probes), were performed as previously described (Vankoningsloo et al., 2006). For this assay, first pre-adipocytes were left undifferentiated or differentiated for 12 days. The adipocytes were then left untreated or treated for an extra 1, 3 or 6 days with 10 ng/ml TNFα, 0.5 μM FCCP or 10 μM isoproterenol, while preadipocytes were incubated with just regular medium for the same times. When indicated, differentiated cells were incubated for 3 days with 10 ng/ml TNFα, 0.5 μM FCCP with or without 10 μM rosiglitazone and 10 μM 9-cis retinoic acid.

For statistical analysis in respect of differentiated cells and cells treated either with FCCP, TNFα or isoproterenol, ratios obtained in the first analysis were transformed to log₂ and then analyzed using the `TIGR Multiexperiment Viewer microarray' data analysis (MeV 4.0) free software (www.tm4.org/mev.htlm). One-way ANOVA (ANOVA I) was first performed followed by SAM (Significance Analysis of Microarrays) (Tusher et al., 2001). A cut-off criterion for differentially expressed genes was set for a fold-change of >1.5 and a delta-threshold value was chosen. Gene expression data that could not be determined (qualitative and non detected values) were not considered in the analyses. For gene name, function and GenBank accession number see Table S6 in supplementary material.

Reverse transcription (Invitrogen) and amplification reaction assay was performed according to the manufacturer's protocol using an ABI 7900 HT thermal cycler (Applied Biosystem). Sense and antisense primers for Acrp30, Glut4, perilipin, B3AR, ATGL and TBP (used for normalization) were designed using the Primer Express 1.5 software (Applied Biosystem; Fig. S9B in supplementary material).

Adipocytes differentiated in 6-well plates, were transfected by electroporation with a luciferase reporter construct, driven by a synthetic promoter containing three PPAR-responsive element sites (tk-PPRE3x-Luc), and an expression vector encoding β-galactosidase using the Cell Line Nucleofector Kit L (Amaxa) according to the manufacturer's instructions specific for adipocytes. The next day, cells were incubated with or without 10 ng/ml TNFα or 0.5 μM FCCP and with or without 10 μM rosiglitazone and 10 μM 9-cis retinoic acid for 24 or 72 hours. Cells were then harvested and luciferase activity was determined in the cell lysate using a Reporter Assay System from Promega. Results were normalized against β-galactosidase activity.

Nuclear and cleared cell lysates from preadipocytes, differentiated adipocytes and adipocytes incubated for different times with 10 ng/ml TNFα, 0.5 μM FCCP or 10 μM isoproterenol in the presence or not of 10 μM rosiglitazone and 10 μM 9-cis retinoic acid and western blot analysis were prepared as previously described (Vankoningsloo et al., 2005). Primary antibodies: anti-PPARγ (81B8), anti-ATGL (30A4) and anti-phospho HSL (Ser565 and Ser660; Cell Signaling), anti-HSL total and anti-perilipin (generous gifts from Constantine Londos, National Institute of Health, MD and Andrew Greenberg, Tufts University, Boston, MA, respectively).

After trypsinization of differentiated adipocytes, cells were seeded at 600,000 cells/well on coverslips in 12-well plates and incubated or not with 10 ng/ml TNFα, 0.5 μM FCCP or 10 μM isoproterenol for 3 days. Immunofluorescence staining and fluorescence confocal microscopy observation (TCS confocal microscope Leica) were performed as previous described (Vankoningsloo et al., 2005). Primary antibodies anti-HSL and anti-perilipin A as well as the Alexa-Fluor-568-conjugated secondary antibody were diluted 100 times and nuclei were visualized by TOPRO-3 staining.

Rates of lipid synthesis and fatty acid β-oxidation were assessed in cells differentiated and undifferentiated cells incubated with 10 ng/ml TNFα or 0.5 μM FCCP for another 1 or 3 days using radioisotope-labelled tracers. The lipid assay was based on the method of Lin and collaborators (Lin et al., 2001) with some modifications. Briefly, after 3 days of incubation (with or without de-differentiating molecules), adipocytes and de-differentiated adipocytes were incubated with [³H]acetate (2 μCi/ml; Perkin Elmer; used as a metabolic precursor of lipids) for the last 24 hours of de-differentiation. Lipids were extracted using the chloroform–methanol method as described above, and lipid radioactivity was counted in an aliquot with a scintillation fluid (Lipoluma; Lumac). Fatty acid β-oxidation assay was assessed by the release of ¹⁴CO₂ after [¹⁴C]oleate uptake (1 μCi/ml; Perkin Elmer) as described previously (Muoio et al., 1999).

Data from at least three independent experiments were analyzed by one-way ANOVA and the Holm-Sidak method. Group means were considered to be statistically significant with P<0.05 or less.