|Home | About | Journals | Submit | Contact Us | Français|
Obese patients have chronic, low-grade inflammation that predisposes to type 2 diabetes and results, in part, from dysregulated visceral white adipose tissue (WAT) functions. The specific signaling pathways underlying WAT dysregulation, however, remain unclear. Here we report that the PPARγ signaling pathway operates differently in the visceral WAT of lean and obese mice. PPARγ in visceral, but not subcutaneous, WAT from obese mice displayed increased sensitivity to activation by its agonist rosiglitazone. This increased sensitivity correlated with increased expression of the gene encoding the ubiquitin hydrolase/ligase ubiquitin carboxyterminal esterase L1 (UCH-L1) and with increased degradation of the PPARγ heterodimerization partner retinoid X receptor α (RXRα), but not RXRβ, in visceral WAT from obese humans and mice. Interestingly, increased UCH-L1 expression and RXRα proteasomal degradation was induced in vitro by conditions mimicking hypoxia, a condition that occurs in obese visceral WAT. Finally, PPARγ-RXRβ heterodimers, but not PPARγ-RXRα complexes, were able to efficiently dismiss the transcriptional corepressor silencing mediator for retinoid and thyroid hormone receptors (SMRT) upon agonist binding. Increasing the RXRα/RXRβ ratio resulted in increased PPARγ responsiveness following agonist stimulation. Thus, the selective proteasomal degradation of RXRα initiated by UCH-L1 upregulation modulates the relative affinity of PPARγ heterodimers for SMRT and their responsiveness to PPARγ agonists, ultimately activating the PPARγ-controlled gene network in visceral WAT of obese animals and humans.
From a clinical perspective, visceral obesity predisposes to an increased incidence of type 2 diabetes mellitus (T2DM) and associated cardiovascular diseases (1, 2). The visceral white adipose tissue (visWAT; i.e., epididymal WAT) depot is believed to contribute to the low-grade, chronic inflammatory state that occurs in obese patients and animals and favors the progression toward T2DM. This feature stems from the specific functional properties of adipocytes from this WAT depot, which are highly sensitive to β-adrenergic stimulation and relatively resistant to the antilipolytic effects of insulin compared with subcutaneous adipocytes (3). Indeed, although subcutaneous WAT (scWAT; i.e., inguinal WAT) is predominantly, but not exclusively, a lipid storage tissue exhibiting a high adipocyte plasticity, visWAT also triggers complex endocrine regulations by releasing FFAs, hormones, and cytokines that reach the liver through the portal vein (reviewed in ref. 4). How visWAT functions are affected upon disease progression is unknown, but metabolic challenges increase the release of proinflammatory cytokines and decrease that of insulin-sensitizing adipokines by visWAT.
Results from recent clinical trials (ADOPT, DREAM, and PROACTIVE; ref. 5) indicate that the insulin-sensitizing thiazolidinediones (TZDs) are highly efficient in maintaining glycemic control, and may exert pancreas-sparing and vascular-protective effects in T2DM patients. TZDs are synthetic agonists for PPARγ (also known as NR1C3), a member of the nuclear receptor (NR) superfamily. PPARγ is a key regulator of adipocyte differentiation and lipid storage, thereby exerting major effects on energy homeostasis (6). Gene ablation studies have confirmed the major role of adipocyte PPARγ (adPPARγ) in mediating the insulin-sensitizing effect of TZDs in obese mice (7–12). Ligand-activated adPPARγ has a positive effect on glucose homeostasis by favoring scWAT expansion and FFA redistribution to this fat depot (13). Removing FFA from other tissues, including skeletal muscle and visWAT, prevents the so-called lipotoxic effect, which causes insulin resistance and hence translates into a greater insulin sensitivity. TZDs also act directly on visWAT functions by regulating the expression of adipokines (14) and several key metabolic genes (15). Quite intriguingly, however, TZDs exert neither detectable insulin-sensitizing effects nor significant metabolic effects in lean individuals and mice, which suggests that the PPARγ pathway operates differently in normal and pathological conditions (16–18).
PPARγ activates target gene transcription by forming obligate heterodimers with the retinoid X receptor (RXR) isotypes — RXRα, RXRβ, and RXRγ — onto PPAR-responsive elements (PPREs). PPREs are found in genes controlling key steps in lipid and glucose metabolism, such as the adipose-specific fatty acid binding protein (aP2), phosphoenolpyruvate carboxykinase (PEPCK), or lipoprotein lipase (LPL). Of note, RXRα also plays a critical role in adipogenesis in vivo (19). Agonist-mediated activation of PPARγ induces structural transitions occurring in the ligand-binding domain (LBD) of this receptor, creating a hydrophobic groove and a charge clamp that binds LXXLL motifs found in most coactivators, such as the p160-related coactivator family (SRC1–SRC3; ref. 20), the integrator complex CBP/p300 (21), components of the mediator complex (22), and the metabolically regulated coactivator PGC-1α (23, 24). Agonist-dependent coactivator recruitment to PPARγ is concomitant to corepressor dismissal, and both the NR corepressor (NCoR) and the silencing mediator for retinoid and thyroid hormone receptors (SMRT) have been shown to affect PPARγ-controlled cellular processes in distinct cellular backgrounds (25–27).
PPARγ transcriptional activity is therefore dependent on its sequential association with multiprotein complexes. It is thus likely that processes controlling protein stability and degradation also influence the overall activity of the PPARγ complex and may possibly affect its function under conditions of obesity and T2DM. Indeed, components of the ubiquitin-proteasome system (UPS) have been shown to be involved in PPARγ-mediated transactivation (28). Although a few studies documented dysregulated expression of several UPS components in cardiovascular diseases (29, 30), a potential contribution of the UPS in the pathogenesis of T2DM and obesity remains unexplored. In the present study, we investigated the relationship among metabolic states, UPS component expression, and the PPARγ signaling pathway. We report that the expression of the ubiquitin hydrolase/ligase ubiquitin carboxyterminal esterase L1 (UCH-L1) was specifically and strongly upregulated in visWAT from obese patients and mice. Mimicking hypoxia in vitro also upregulated UCH-L1 expression. This enzyme promotes the selective breakdown of RXRα, which correlates with an increased response of PPARγ to a synthetic agonist in vivo. Moreover, decreasing the RXRα/RXRβ ratio in vitro increased PPARγ transcriptional activity. The molecular basis of this phenomenon was found to be a stable, ligand-insensitive tethering of SMRT to PPARγ-RXRα heterodimers.
Because previous studies reported an unexpected insensitivity of lean mice or humans to TZD treatment, we investigated whether TZDs have a differential efficiency in WAT of normal lean (OB/OB) versus pathological (ob/ob) mice. Treatment with rosiglitazone (RSG; 3 mg/kg/d) lowered plasma glucose, insulinemia, and, to a lesser extent, triglycerides in 10-week-old male ob/ob mice, but had no effect in OB/OB littermates (Supplemental Figure 1; supplemental material available online with this article; doi: 10.1172/JCI38606DS1). This suggests that activation of PPARγ could induce distinct biological responses depending on the energetic status. To assess this at the molecular level, adipose tissue gene expression profiles were examined by oligonucleotide microarray analysis of visWAT mRNAs from OB/OB and ob/ob mice treated or not with RSG (Figure (Figure1A).1A). Although these tissues exhibited similar PPARγ protein expression levels (Figure (Figure1B),1B), RSG upregulated (>1.5-fold, P < 0.05) 32 genes in OB/OB visWAT, whereas 153 genes were upregulated in ob/ob visWAT, a substantial fraction of the latter being involved in metabolic control. Of these, only 4 genes were upregulated in both OB/OB and ob/ob visWAT, which suggests that PPARγ regulates distinct transcriptional networks in normal and pathological tissues. This was confirmed by a Gene Ontology functional classification. Importantly, upregulated genes in ob/ob visWAT included known direct target genes for PPARγ (e.g., CIDEA, UCP1, pyruvate carboxylase, and malic enzyme), which were not markedly upregulated in visWAT from OB/OB animals. We thus compared the ability of RSG to induce a subset of PPRE-driven adipocyte target genes in scWAT or visWAT from ob/ob mice and OB/OB littermates by real-time PCR. Whereas mRNA levels of the PPARγ target genes adiponectin (Adpn), glycerol kinase (GyK), aP2, Glut4, Pparg, and PEPCK were upregulated to a similar extent in scWAT from OB/OB and ob/ob mice, they exhibited a much stronger responsiveness to RSG in visWAT from ob/ob mice (Figure (Figure1C),1C), in line with the microarray data. These comparative results thus demonstrate that PPARγ target genes are more sensitive to agonist-mediated activation in obese than in lean visWAT. In sharp contrast, such an activity shift was not observed in scWAT.
PPARγ transcriptional activity depends on its association with a number of cofactors, including RXRs. We thus reasoned that the increased PPARγ transcriptional activity in obese visWAT might result from altered expression of a PPARγ cofactor. Preliminary characterization of the mRNA expression levels of several PPARγ primary cofactors did not reveal substantial differences between lean and obese visWAT, with the exception of PGC1, whose expression decreased by approximately 3-fold in obese visWAT (data not shown). We then monitored Rxra, Rxrb, and Rxrg mRNA levels by quantitative PCR (QPCR) in mouse scWAT and visWAT (Figure (Figure2,2, A and B). Rxr mRNA levels were similar in WAT of OB/OB and ob/ob mice fed a regular diet (LFD) or high-fat diet (HFD). However, RXRα activity has previously been shown to be regulated by protein degradation in several cell types (31–35). We thus investigated whether RXRα polypeptide stability is affected in visWAT and scWAT of OB/OB and ob/ob mice. Immunohistochemical and Western blotting analyses revealed that RXRα was expressed in adipocytes and preadipocytes of both visWAT and scWAT from OB/OB mice (Figure (Figure2C).2C). We thus quantified the expression level of RXRα and RXRβ polypeptides, the highest expressed in OB/OB and ob/ob mouse visWAT and scWAT (Figure (Figure2D).2D). Western blot analysis of mouse WAT extracts for RXRα and RXRβ revealed that RXRα expression was strongly diminished in ob/ob visWAT, whereas RXRβ was expressed, although less abundantly so, at levels similar between OB/OB and ob/ob WAT (Figure (Figure2,2, D–F). This altered expression was specific for visWAT; scWAT from OB/OB and ob/ob mice displayed similar RXRα and RXRβ levels. To rule out the possibility that RXRα downregulation is specific to the ob/ob genetic background, RXRα and RXRβ content was also quantified in WAT from mice fed either LFD or HFD. Interestingly, Rxrα, but not Rxrβ, protein levels were also specifically decreased in visWAT from HFD-fed mice (Figure (Figure2,2, D–F). Similarly, we investigated whether RXR polypeptide stability is affected in human scWAT and visWAT biopsies from patients with different levels of obesity and diabetes (Figure (Figure2,2, G and H). RXRA, RXRB, and RXRG mRNA levels did not differ between scWAT and visWAT biopsies from normal, obese, obese glucose intolerant, and obese diabetic individuals. Western blot analysis of WAT extracts revealed that RXRα, but not RXRβ, was much less abundant in visWAT from obese diabetic subjects than in visWAT from lean subjects (Figure (Figure2,2, I–K). In contrast, RXRA and RXRB expression were not different in scWAT from lean and obese diabetic individuals (Figure (Figure2,2, I–K). Of note, PPARγ expression was comparable in all tissues (Figure (Figure2,2, I and L). Collectively, these findings show that the RXRα protein is specifically degraded in visWAT from obese humans and mice.
Because RXRA mRNA levels were not altered, we postulated that the severe decrease in RXRα protein might stem from dysregulated expression of components of the UPS during metabolic disease progression. Therefore, UPS component expression was monitored in scWAT and visWAT from lean, obese, obese glucose intolerant, and obese diabetic individuals. Although the expression of components constituting the canonical 26S proteasome was not markedly altered in any of these WAT depots (data not shown), the expression of UCH-L1, an ubiquitin esterase/ligase enzyme, increased in visWAT, but not in scWAT, with progressing stages of the disease (Figure (Figure3,3, A and B). In contrast, the expression of other UPS components, such as USP22 and WWP2 (known to regulate the stability of transcription factors such as NFκ-B or RNA polymerase 2), was not substantially modified. Because the dysregulated UCH-L1 expression was depot specific in humans, we investigated whether a similar phenomenon occurs in mouse WAT tissues by comparing the expression of Uch-L1, Usp22, and Wwp2 in WAT from OB/OB, ob/ob, LFD-fed, and HFD-fed mice (Figure (Figure3C).3C). Strikingly, a strong upregulation of Uch-L1 mRNA was observed in visWAT tissues of ob/ob and HFD-fed mice, whereas Usp22 and Wwp2 exhibited less pronounced upregulation (5-fold versus 2.5-fold). In contrast, the expression of these genes was similar in scWAT of OB/OB, ob/ob, and HFD-fed mice. This upregulation was also observed at the protein level (Figure (Figure3D).3D). These data demonstrate that UCH-L1 is strongly and specifically upregulated in visWAT from metabolically challenged humans or mice.
The correlation between UCH-L1 upregulation and the strongly reduced RXRα polypeptide steady-state levels in obesity led us to speculate that UCH-L1 could be critical in controlling RXRα stability. We therefore transfected 3T3-L1 preadipocytes with expression vectors coding for RXRα or RXRβ, with or without an expression vector coding for UCH-L1 (Figure (Figure4A).4A). Incubation of RXR-transfected cells with the proteasome inhibitor MG132 moderately increased the amount of RXRα, but did not influence RXRβ protein level. A combination of ammonium chloride and leupeptin (NH4Cl/Leu), which inhibits the lysosomal degradation pathway, did not modify RXRα and RXRβ protein levels in RXR-transfected cells. Remarkably, UCH-L1 overexpression in RXR-transfected cells induced the breakdown of the RXRα polypeptide, whereas RXRβ stability was unaffected. UCH-L1–induced RXRα breakdown was prevented by MG132, but not by NH4Cl/Leu, which indicates that UCH-L1 promotes RXRα breakdown through proteasomal degradation. RXR ubiquitinylation was then examined in 3T3-L1 preadipocytes transfected with expression vectors coding for RXR and HA-tagged ubiquitin. Immunoprecipitation of RXR followed by Western blot detection of HA-tagged ubiquitin (Figure (Figure4B),4B), or immunoprecipitation of HA-tagged proteins followed by Western blot detection of RXR (Supplemental Figure 2), showed that only RXRα was intensively conjugated to HA-ubiquitin and that UCH-L1 overexpression markedly increased ubiquitinylation of RXRα, but not RXRβ. Furthermore, an in vitro ubiquitinylation assay using purified RXRs as substrates showed that RXRα was ubiquitinated in an ATP-dependent manner (Figure (Figure4C).4C). Ubiquitin-conjugated RXRα was stabilized in the presence of ubiquitin aldehyde, a general inhibitor of ubiquitin hydrolases (36), or of MG132. In sharp contrast, RXRβ was not detectably conjugated to ubiquitin in similar conditions (Figure (Figure4B4B and Supplemental Figure 2). Taken together, these data demonstrate that RXRβ is refractory to proteasome-mediated breakdown. Hypoxia is a feature of obese WAT (reviewed in ref. 37), characterized by induced gene expression and protein stabilization of hypoxia-inducible transcription factors (HIFs). Accordingly, Hif1a mRNA was upregulated in visWAT from ob/ob and HFD-fed mice (Figure (Figure4D),4D), whereas Hif1b and Hif2b expression was selectively increased in visWAT from HFD-fed mice. Hif2a expression was unaffected in both types of WAT. Because HIF1α upregulation in obese WAT correlated with increased Uch-L1 expression (Figure (Figure3C),3C), and given the occurrence of several HIF-responsive elements in the promoter of the mouse and human UCH-L1 genes, we tested whether cobalt chloride (CoCl2), a hypoxia-mimicking compound, also upregulated UCH-L1 in differentiated 3T3-L1 adipocytes. CoCl2 caused concomitant upregulation of Hif1a and Uch-L1 mRNAs (Figure (Figure4E)4E) and dose-dependent degradation of the RXRα protein (Figure (Figure4F).4F). Furthermore, both MG132 and LDN-54777, a specific inhibitor of UCH-L1 (38), blocked the CoCl2-induced RXRα degradation in a dose-dependent manner (Figure (Figure4G).4G). Pulse-chase labeling of 3T3-L1 adipocyte proteins showed that the decay of RXRα was linear, with a t1/2 of about 8 hours (Figure (Figure4H).4H). CoCl2 treatment decreased the t1/2 of RXRα to 4 hours, whereas LDN-54777 clearly prevented the CoCl2-induced RXRα breakdown (t1/2, 7 hours). Taken together, these data argue for a role of UCH-L1 in the UPS-mediated RXRα degradation process, in which hypoxia might play a role.
In light of the above results, we investigated whether an altered RXRα/RXRβ ratio might directly alter PPARγ responsiveness to agonist challenge in a cell-autonomous manner. Rxrα was strongly expressed in 3T3-L1 differentiated adipocytes, whereas Rxrβ and Rxrγ were expressed at much lower levels (Figure (Figure5A).5A). Pparγ, as well as Pparα, were also expressed in differentiated 3T3-L1 cells. RSG treatment readily activated aP2, GyK, and Adpn gene expression (Supplemental Figure 3). ChIP assays using anti-RXR or anti-PPAR antibodies (Figure (Figure5B)5B) failed to detect Rxrγ on the aP2 and Adpn PPREs, in agreement with its low expression level, whereas comparable occupancy by both Rxrα and Rxrβ was detected on these promoters. Pparγ, but not Pparα, also bound to these promoters, and the density of Pparγ and RXRs on these promoters was similar in the absence and presence of RSG (Figure (Figure5B).5B). Identical results were obtained for the GyK promoter (data not shown). Thus, simultaneous expression of RXRα and RXRβ generates PPARγ-driven promoters on which both RXR isotypes can be loaded.
To further assess the role of the RXR isotype on aP2, Adpn, and GyK gene responsiveness to RSG, we generated Rxrα-depleted 3T3-L1 adipocytes by siRNA-mediated knockdown, which induced a specific Rxra mRNA decrease of 70% (Supplemental Figure 4). RXRα knockdown resulted in a more pronounced induction of the PPARγ target genes by RSG (aP2, 9.0- versus 3.6-fold; Adpn, 4.6- versus 2.6-fold; GyK, 6.3- versus 4.4-fold; Figure Figure5C).5C). Conversely, 3T3-L1 CAR cells, which stably overexpress the adenovirus receptor (39), were differentiated into adipocytes and transduced at day 6 with adenoviral particles encoding GFP as a negative control, with Rxrα, or with Rxrβ, allowing for strong overexpression of each RXR isotype (Supplemental Figure 5). Interestingly, RSG treatment caused stronger induction of aP2, Adpn, GyK, and Pparg in RXRβ-overexpressing cells compared with GFP-expressing cells (Figure (Figure5D).5D). Inversely, forced expression of Rxrα attenuated the induction of Pparγ target genes by RSG (Figure (Figure5D). 5D).
We used QPCR analysis to investigate whether decreased Rxrα expression causes a derepression of PPARγ target genes in other cellular backgrounds that exhibit varying RXRα/RXRβ ratios (Figure (Figure5E).5E). 3T3-L1 preadipocytes displayed a Rxra/Rxrb mRNA ratio of 10:1, much like the HepG2 hepatocarcinoma and C2C12 myeloblastic cell lines (10:1 and 4:1, respectively). In contrast, the β-insulinoma cell line Min6 expressed comparable levels of Rxra and Rxrb mRNAs, at a 1:1 ratio. Assessment of the transcriptional activity of Pparγ in these different cell types using the J6 tk-Luc reporter gene showed that a much higher maximal transcriptional activity was reached in Min6 cells than in 3T3-L1, C2C12, and HepG2 cells in response to RSG (Figure (Figure5F).5F). Overexpressing Rxrα in Min6 cells significantly blunted the response to RSG, whereas overexpression of Rxrβ potentiated this response (Figure (Figure5G).5G). A similar pattern was obtained in COS cells and on another PPRE-driven reporter gene, ApoA2 tk-Luc, which demonstrates that the RXRα-repressive function was dependent neither on the cell type nor on the response element (Figure (Figure5H).5H). We then used a system in which the P box, located in the DNA-binding domain (DBD) of either Rxrα or Rxrβ, was mutated to confer a high affinity for a glucocorticoid-responsive element (GRE) half site. The RXR binding polarity can thus be imposed when using a chimeric direct repeat 1 response element PPRE in which the GRE half site is located in either 3′ or 5′. This system was only functional when RXR was bound on the 3′ half site of the chimeric PPRE (ref. 40 and Figure Figure5I).5I). In this configuration, Pparγ was more sensitive to a moderate 30-nM concentration of RSG when Rxrβ was expressed (Figure (Figure5I). 5I).
Taken together, these data indicate that (a) PPARγ can bind as a dimer with RXRα or RXRβ to PPRE-driven promoters in adipocytes; (b) this interaction is not ligand sensitive; and (c) decreased RXRα expression correlates in vitro and in vivo to an increased responsiveness to RSG. RXRα thus acts as a repressor of PPARγ responsiveness to agonist challenge in murine adipocytes.
Collectively, these results show that PPARγ-RXRα heterodimers display lower sensitivity to agonist challenge than do PPARγ-RXRβ dimers. To test whether this might be the result of increased interaction with corepressors, the influence of the RXR isotype on the binding of PPARγ to SMRT or NCoR was further characterized. A yeast 2-hybrid assay was first performed to assess the interaction of PPARγ fused to the Gal4 DBD, with the NR interaction domain (NRID) of either SMRT (aa 2,061–2,472) or NCoR (aa 1,906–2,313) fused to the NF-κB activation domain. SMRT interacted more efficiently with the PPARγ LBD than did NCoR, whereas interaction of the PPARγ LBD with the coactivator Med1 (also known as TRAP220 or DRIP205) was negligible (Figure (Figure6A).6A). A similar 2-hybrid assay in mammalian HeLa cells using the PPARγ LBD fused to the Gal4 DBD and NRIDs fused to the VP16 activation domain (VP16-AD) revealed a similar pattern of interaction (Figure (Figure6B),6B), demonstrating that the preferential interaction of SMRT with PPARγ is an intrinsic property not affected by the cellular background. SMRT was therefore selected as the representative corepressor in further experiments.
A glutathione S-transferase–pulldown (GST-pulldown) assay was then carried out to assess the interaction between full-length PPARγ and SMRT. Using an immobilized GST-SMRT fusion protein and radiolabeled PPARγ, we detected a strong interaction between the proteins in this system (Figure (Figure6C).6C). RSG (Figure (Figure6C)6C) and 2 other PPARγ agonists, pioglitazone and troglitazone (Supplemental Figure 6A), were unable to promote significant release of SMRT from monomeric PPARγ. In similar conditions, RSG promoted the recruitment of Med1, GRIP1 (also known as TIF-2), and PGC-1α to the PPARγ LBD (Supplemental Figure 6B), showing that the PPARγ LBD undergoes appropriate structural transitions. Thus, the PPARγ-SMRT interaction was not sensitive to PPARγ agonist binding in this setting. To assess whether a stable interaction with SMRT also occurred on PPARγ target gene promoters, ChIP assays were performed in 3T3-L1 adipocytes. 3T3-L1 cells express several coactivators, as well as NCoR and SMRT, whose mRNA expression levels did not vary significantly during the differentiation process (Supplemental Figure 7). In differentiated, unstimulated 3T3-L1 adipocytes, SMRT was clearly detected on each promoter, whereas NCoR displayed barely detectable binding (Figure (Figure6D).6D). Upon RSG treatment, SMRT was partially dismissed from both promoters, which indicates that, in the context of an endogenous functional promoter, agonist-activated PPARγ is able to at least partially release SMRT. To confirm these findings, we monitored the association of 2 SMRT-associated proteins, HDAC3 and SIRT1. Similar to SMRT, both HDAC3 and SIRT1 were partially dismissed from the aP2 and Adpn promoters (Figure (Figure6E).6E). Thus a partial, agonist-induced dismissal of corepressor complex correlated with mixed RXRα/RXRβ occupancy at PPARγ-regulated target genes (Figure (Figure5B). 5B).
We hypothesized that the observed increase in PPARγ-RXRβ responsiveness to agonist could result from a specific feature of the RXRα-SMRT interaction. This hypothesis was tested using a 2-hybrid assay in NIH-3T3 cells, which expresses neither PPARγ nor C/EBPα, but can be fully differentiated into adipocytes upon ectopic expression of these 2 transcription factors (ref. 41 and Figure Figure7A).7A). A Gal4-SMRT NRID fusion protein was overexpressed together with a PPARγ LBD–VP16-AD fusion protein in the presence or absence of RXRα or RXRβ. The transcriptional activity of the system was monitored with a UAS tk-Luc reporter gene, whose activity was predicted to decline upon dissociation of the PPARγ-VP16 protein. As a control experiment, we expressed a human RAR-VP16 (hRAR-VP16) fusion protein together with the Gal4-SMRT NRID. This system displayed clear agonist-dependent decreased transcriptional activity (all trans retinoic acid; atRA), reflecting the dissociation of the hRAR-VP16 fusion protein from the Gal4-SMRT bait upon agonist binding (Figure (Figure7A).7A). In similar conditions, the PPARγ-VP16–Gal4-SMRT interaction was not sensitive to increasing concentrations of RSG. Concomitant overexpression of RXRα, RXRβ, or RXRγ increased the basal level of interaction of PPARγ with SMRT. However, RSG decreased system activity only in the presence of RXRβ, which indicates decreased SMRT association with the PPARγ-RXRβ heterodimer. Taken together, these data suggest that the PPARγ-SMRT interaction is disrupted by RSG only when RXRβ is integrated in the ternary SMRT-PPARγ-RXR complex.
We concluded from ChIP and 2-hybrid data that the PPARγ-RXRα interaction with SMRT was ligand insensitive. To further validate this hypothesis in the context of a DNA-bound PPARγ-RXR heterodimer, we used a modified 2-hybrid assay in NIH 3T3 cells. The J6 tk-Luc reporter gene was cotransfected with PPARγ and RXR expression vectors, together with an expression plasmid coding either for VP16-AD as a control or for a SMRT–VP16-AD fusion protein (Figure (Figure7B).7B). Preliminary experiments showed that detected responses necessitated a PPRE sequence (data not shown). We first used the RARα-RXRα heterodimer as a control system. The basal activity of this dimer strongly increased in the presence of VP16-SMRT, indicative of a hRARα-RXRα–SMRT interaction. A saturating concentration of atRA (1 μM) yielded luciferase activity similar to that observed in control conditions, showing that SMRT was fully dissociated from the RXR-RAR complex upon agonist binding (Figure (Figure7B).7B). When transposed to the PPARγ-RXRα system, VP16-SMRT expression also strongly increased the basal activity of the reporter gene (5-fold induction) compared with VP16-AD alone. However, in contrast to the RXRα-RARα dimer, higher transcriptional activity of the PPARγ-RXRα dimer compared with control conditions was maintained even at a saturating RSG concentration. This additive induction was a result of the cumulative effect of the AF2- and VP16-mediated transcriptional activation domains, strongly suggesting that agonist treatment does not induce the dismissal of VP16-SMRT from the PPARγ-RXRα heterodimer. RXRγ exhibited behavior similar to that of RXRα (Figure (Figure7C).7C). In sharp contrast, RXRβ overexpression generated a system in which saturating RSG concentrations prevented SMRT-VP16 from further activating the reporter gene activity, showing that the liganded PPARγ-RXRβ dimer cannot bind the SMRT moiety. We conclude from the above data that only the PPARγ-RXRβ dimer fully releases SMRT upon agonist binding.
The transcriptional control of programs regulating metabolic homeostasis can be substantially altered by manipulating the activity of transcription factors belonging to the NR superfamily. Among the NRs, PPARγ plays an important role as a central regulator of lipid and glucose metabolism and is important for maintaining whole-body insulin sensitivity (42). However, how its function may be modulated upon disease progression is unknown. Our results provide what we believe to be a novel paradigm on how a transcriptional regulator may adjust its transcriptional activity in the face of metabolic challenges.
The inverse correlation in visWAT from obese mice and humans between UCH-L1 expression and RXRα polypeptide stability highlights a regulatory mechanism of the PPARγ signaling pathway. UCH-L1 exhibits ubiquitin C-terminal hydrolase activity, which, by maintaining a sufficient pool of cellular ubiquitin, sustains protein degradation. Additional roles, such as ubiquitin ligation and stabilization of monoubiquitin, have previously been described for UCH-L1 (reviewed in ref. 43). UCH-L1 substrates are unknown, but its involvement in regulating NF-κB activity in vascular cells (44) and its varied gene expression in rat pancreatic islets exposed to high glucose concentrations (45) hint at a role in metabolic control. Although we found increased UCH-L1 expression at various stages of disease progression in humans, the mechanism leading hereto is unknown. Hypoxia, which affects obese WAT (37), may be a possible link between pathological states and increased UCH-L1 expression. In addition, hypoxia has been reported to promote RXRα breakdown in cardiac myocytes (46). In ob/ob and HFD visWAT, we observed increased expression of Hif1a, and the UCH-L1 promoter contains several HIF-responsive elements. In agreement with these observations, CoCl2, an agent mimicking partially hypoxia through HIF1α induction and stabilization, was found to increase UCH-L1 expression in 3T3-L1 adipocytes and to promote RXRα protein breakdown in a UCH-L1–dependent manner. RXRα is a known target of the UPS in several cell types (31–35), and our present data showed that RXRα was a substrate for in vitro ubiquitin conjugation, whereas RXRβ was unable to undergo this posttranslational modification. However, recombinant UCH-L1 neither modified the ubiquitin conjugation rate in the acellular system nor engaged direct protein-protein interaction with RXRα in vitro, as assayed by GST-pulldown assays (data not shown), which suggests that this enzyme promotes RXRα breakdown in an indirect manner. Combined with the fact that visWAT from ob/ob mice is more sensitive to TZD treatment than is visWAT from OB/OB mice, our data raised the possibility that an UCH-L1–dependent RXR expression isotype switch could control PPARγ transcriptional activity.
As with all other ligand-regulated NRs, PPARγ exerts its transcriptional activity through the recruitment of coactivators, and PPARγ has been documented to bind in vitro to a variety of primary coactivators. PPARγ transcriptional activity is also modulated by corepressor recruitment, and we showed here that PPARγ-containing heterodimers preferentially recruited SMRT over NCoR. Such a preference has been reported by some in a variety of cell-free and cellular systems (25, 40, 47, 48), but not confirmed by others (26, 49). Although RNA interference studies showed that SMRT repressed PPARγ transcriptional activity (data not shown), we found that the physical interaction of PPARγ with SMRT was not affected by agonist binding. Thus, assays involving solely PPARγ lack a critical component required for the ligand-regulated association of SMRT to PPARγ, in line with a previous report (50). As an obligate heterodimerization partner conferring specific DNA binding activity to PPARγ, RXR is a major regulator of PPARγ activity. Mammalian 2-hybrid assays established that RXR association to PPARγ allowed for a reversible association of SMRT to the PPARγ-RXR heterodimer in an isotype-dependent manner. Only RXRβ generated a complex able to dismiss SMRT in an agonist-dependent manner, providing a molecular basis for the higher inducibility of PPARγ in RXRβ-enriched cellular backgrounds. This was also true in a mouse adipocyte background, in which RXRα and RXRβ normally associate to the PPRE of PPARγ target genes. The interaction of PPARγ, as well as that of RXRs, with the PPRE of the endogenous aP2, Adpn, and GyK gene promoters was constitutive and ligand insensitive in differentiated 3T3-L1 adipocytes, in agreement with previous data (26). The mixed composition of PPRE-bound heterodimers, in light of our interaction data, led us to predict that SMRT would be only partially released upon agonist challenge. This was indeed the case, confirmed by the partial release of SMRT-interacting proteins HDAC3 and Sirt1. In agreement with the repressive role of RXRα, siRNA-mediated decrease of RXRα expression led to a higher sensitivity to RSG; conversely, its overexpression in 3T3-L1 adipocytes and other cellular backgrounds invariably blunted responsiveness to RSG. A differential affinity of NRs for distinct corepressors has only been documented for NRs expressed as monomers. For example, restricted SMRT recruitment has been observed for the atRA receptors RARα, RARβ, and RARγ (51, 52), and differences are due to the poorly conserved C-terminal F domain (53). However, since neither RXRs nor PPARγ have an F domain, other structural determinants must regulate the differential affinity of RXRα- versus RXRβ-containing heterodimers for SMRT and are yet to be identified.
It has been postulated that the tissue-specific recruitment of nuclear coactivators by specific agonists influences PPARγ transcriptional activity, a mechanism by which side effects triggered in TZD target tissues, such as kidney, bone, or even adipose tissues, could be circumvented (e.g., refs. 54–56). Our data suggest that PPARγ operates differently in normal and pathological tissues as a result of heterodimerization with distinct RXR isotypes. Whether this phenomenon also impinges on the ability of PPARγ to recruit a specific subset of coactivators is unknown, but nevertheless underlines the need for a careful selection of screening procedures aimed at identifying novel PPARγ synthetic ligands. It also leaves open the possibility to design ligands favoring — or not — heterodimerization with RXRα, thereby predictably moderating the transcriptional response to synthetic PPARγ ligands. The need for moderated PPARγ activation has already been underlined by (a) the protective effect of the transcriptionally altered Pro12Ala PPARγ mutant against the development of insulin resistance and T2DM; (b) the improved insulin sensitivity of heterozygous PPARγ mice fed a HFD; and (c) the beneficial effects on obesity and insulin resistance upon treatment with PPARγ antagonists (57).
In summary, our studies elucidated an unexpected mechanism by which the UPS regulates PPARγ transcriptional activity in pathological states through selective degradation of RXRα. Modifying PPARγ transcriptional activity in the face of metabolic challenges may be a mechanism by which cells adapt to novel conditions through transcriptional reprogramming. Whether RXRα- and RXRβ-containing heterodimers control an overlapping or a specific transcriptional program is under investigation. A better understanding of the as-yet putative specific roles of RXR isotypes and of their regulation in pathological conditions may help define new therapeutical strategies to treat T2DM and associated comorbidities.
Isobutylmethylxanthine (IBMX), insulin, dexamethasone, and atRA were obtained from Sigma-Aldrich. LDN-57444 was purchased from Calbiochem. RSG was synthesized at the chemical facilities at Institut de Recherches Servier.
pSG5-hRAR, pSG5-hRXR, pSG5-hRXR, pSG5-hRXR, pSG5-hPPARγ (DR5)3-tk Luc, VP16-hRAR (PPRE)6-tk Luc, and Gal4-tk Luc were described elsewhere (58–61). VP16-PPARγ was provided by B.M. Forman (City of Hope National Medical Center, Duarte, California, USA; ref. 62). Gal4-hSMRT2117–2357 and VP16-hSMRT2117–2357 were gifts from A.N. Hollenberg (Beth Israel Deaconess Medical Center, Boston, Massachusetts, USA; ref. 25). pCMX-VP16-NCoR1585–2453 was provided by R. Renkawitz and M. Schulz (Institute for Genetics, Giessen, Germany; ref. 63). The pCMV-based human UCH-L1, USP22, and WWP2 expression vectors were purchased from OriGene.
NIH 3T3 cells were grown in DMEM (Glutamax-1, high glucose; Invitrogen) supplemented with 10% FCS (BioWhittaker) and 100 U/ml penicillin and 100 μg/ml streptomycin (Invitrogen). Transient transfection experiments were performed using Lipofectamine 2000 (Invitrogen). Luciferase assays were performed with the dual luciferase assay system (Promega) according to the manufacturer’s guidelines.
Adipocyte differentiation was induced as follows: 3T3-L1 preadipocytes were cultured in DMEM supplemented with 10% FCS until they reached confluence. The medium was then changed, and cells were allowed to grow for 2 additional days. Adipocyte differentiation was then induced by replacing the culture medium by DMEM supplemented with 10% FCS, 0.5 mM IBMX, 1 μM dexamethasone, and 10 μg/ml insulin. 2 days later, the medium was replaced with DMEM supplemented with 10% FCS and 10 μg/ml insulin for 2 additional days. Cells were then refed at 48-hour intervals with DMEM supplemented with 10% FCS only.
Differentiated 3T3-L1 cells were starved in methionine/cysteine-free DMEM supplemented with glutamine and 10% dialyzed FCS for 4 hours. [35S] methionine and cysteine (250Ci, EasyTag EXPRESS35S Protein Labeling Mix; Perkin-Elmer) was added to the culture medium for 4 hours. At 2 hours prior to the end of the pulse, 500 μM CoCl2 and/or 10 μM LDN-57444 was added to the medium. Radioactive media was removed and substituted for regular medium supplemented with 2 mM cysteine and methionine, CoCl2, and/or LDN-57444. At the indicated times, approximately 107 cells were collected in ice-cold 1× PBS, centrifuged, and lysed into 500 μL RIPA buffer (50 mM Tris-HCl, pH 7.4; 150 mM NaCl; 1% NP-40; 0.5% deoxycholate; 0.1% SDS; 10 μg/ml leupeptin; 10 μg/ml pepstatin; 5 μg/ml aprotinin; 5 mM DTT; 5 mM PMSF; and 1 mM benzamidine). The homogenate was spun down, and the supernatant was precleared with a protein A/protein G sepharose mix. The supernatant was incubated overnight at 4°C with 5 μg anti-RXRα antibody (ΔN-197, sc-774; Santa Cruz Biotechnology Inc.). Complexes were precipitated with a protein A/protein G mix, washed, and analyzed by 8% SDS-PAGE.
In vitro ubiquitinylation assays were carried out using a S100 HeLa cell extract and reagents according to the manufacturer’s instructions (Enzo Life Sciences).
3T3-L1 preadipocytes were transfected with expression vectors encoding HA-tagged ubiquitin (MT123-Ub HA; refs. 64, 65), RXRα, or RXRβ. At 24 hours after transfection, whole cell extracts were prepared in RIPA buffer supplemented with 10 μM MG132, and HA-tagged proteins were immunoprecipitated using an anti-HA tag antibody (ab9110; Abcam) or anti-RXR antibodies (see below). Ubiquitinylated RXR was detected by Western blotting using either anti-RXR or anti-HA tag antibodies as indicated in the figure legends.
The anti-RXRα antibody used in immunochemistry experiments was from Perseus (PP-K8508-00). WAT was embedded in paraffin and sectioned (7 μm). Sections were deparaffinized and rehydrated. After brief heating, the endogenous peroxidase activity was quenched, and sections were then incubated with the anti-RXR antibody (1:200 dilution). Sections were subsequently incubated with biotinylated goat anti-mouse antibody, then streptavidin-HRP (Vectastain). Antigenic complexes were detected with 3,3′-diaminobenzidine, and sections were mounted in Mowiol (Kuraray). Images were collected with a Leica microscope and a camera coupled to the Leica IM500 image Manager software with ×10 magnification.
Primary antibodies used in Western blotting experiments were directed against RXRα (D-20, sc-553), RXRβ (C-20, sc-831), RXRγ (Y-20, sc-555), PPARγ (H-100, sc-7196; all from Santa Cruz Biotechnology Inc.), and UCH-L1 (3525S; Cell Signaling Technology). Secondary antibodies coupled to HRP were from Sigma-Aldrich. Immune complexes were detected using the ECL+ system from Amersham/GE Healthcare.
Yeast 2-hybrid experiments, as described by Carmona et al. (54), were carried out at Phenex GmbH.
ChIP assays were carried out as described previously (58, 66). Primer sequences used in ChIP assays were described elsewhere (26), except for the Adpn promoter region (forward, 5′-CCATGCCTGCAGTCCATCTA-3′; reverse, 5′-GCTTCTGTCAAGCCATCCTGT-3′). The antibody to SMRT (PA1-844A) was from Affinity Bioreagents, whereas anti-NCoR (sc-1609), anti-PPARα (sc-9000), anti-PPARγ (sc-7196), anti-RXRα (sc-553), anti-RXRβ (sc-831), and anti-RXRγ (sc-555) were purchased from Santa Cruz Biotechnology Inc.
GST-pulldown experiments were performed as described previously (67). Data were acquired on a Storm 860 phosphorimager, and band intensities were quantified using ImageQuant TL software.
Total RNA was prepared using the RNeasy minikit (Qiagen). Total RNA was isolated from mouse WAT or from 3T3-L1 cells using the RNAeasy Lipid Tissue kit (Qiagen). Purified RNA was adjusted to 1 g/l, and its integrity was assessed by gel electrophoresis or on an Agilent Bioanalyzer. RNA was purified from visWAT from 3 mice per group (OB/OB, OB/OB plus RSG, ob/ob, and ob/ob plus RSG) and hybridized to Affymetrix 430 2.0 arrays after cDNA labeling. Data were analyzed using the GeneSpring GX software (Agilent). Normalization was performed using the RMA algorithm, followed by a Benjamini-Hochberg false discovery rate statistical analysis. Genes that were significantly upregulated or repressed by more than 1.5-fold were then classified by a Gene Ontology functional classification. For RT-QPCR analysis of transcripts, reverse transcription was performed using random hexamers as recommended by the manufacturer (Promega). cDNAs were analyzed by PCR amplification using the TaqMan PCR master mix (Applied Biosystems) and a mix of actin primers and appropriate FAM probes. Absolute quantification of RXR mRNAs were determined by generating standard curves with known amounts of cloned RXR cDNAs. Actin, aP2, GyK, and Adpn primers and all other probes were purchased from Applied Biosystems (Assay on Demand). PCR (40 cycles) and data analysis were carried out using an ABI Prism 7500 (Perkin-Elmer).
Specific siRNA duplexes targeting RXRα, SMRT, and nonspecific siRNA controls were synthesized by Santa Cruz Biotechnology Inc. siRNAs were transfected using the DeliverX Plus siRNA transfection kit (Panomics) according to the manufacturer’s guidelines.
Mouse RNA and tissue samples were provided by M. Brun and A. Géant (Institut de Recherches Servier, Suresnes, France) and H. Duez (INSERM U545, Lille, France). OB/OB and ob/ob mice (8–10 weeks old) were purchased from Charles River. Mice were housed in a temperature-controlled room (22°C–24°C), with a relative humidity of 36%–80% and 12-hour light/dark cycles. Mice were fed ad libitum with free access to filtered tap water (0.22-μm filter) and received irradiated pelleted laboratory chow (A03-10; UAR) throughout the study, supplemented or not with RSG to achieve a 3-mg/kg/d intake. Mice were euthanized by cervical dislocation, and visWAT and scWAT were dissected, weighed, and immediately stored in liquid N2. All procedures were validated by the ethical committee of the Institut Pasteur de Lille and carried out in accordance with European Union (EEC, no. 07430) and French ethical guidelines.
Cells were prepared from WAT of obese or control mice as previously described (68), with minor modifications. Briefly, tissue was digested at 37°C in 1× PBS containing 0.2% BSA and 2 mg/ml collagenase (type II collagenase; Sigma-Aldrich). After filtration of the homogenate through 25-μm filters, mature adipocytes were separated from pellets (stromal vascular fraction) by centrifugation at 600 g for 10 minutes. After red cell lysis in 140 mM NH4Cl and centrifugation, stromal vascular fraction cells were pelleted and resuspended in 1× PBS. Cells were seeded at 10,000 cells/cm2 in DMEM-F12 supplemented with 10% newborn calf serum. Extensive washes were performed 12 hours later, and whole cell lysates were prepared 48 hours later.
Human tissue samples were provided by M.-F. Six and the Centre d’Investigations Cliniques (C.H.R.U. Lille). Human WAT samples were collected from patients undergoing abdominal surgery by laparoscopy or coelioscopy after informed consent was obtained. All procedures were approved by the C.H.R.U. Lille ethical committee and were compliant to the French National Ethics Committee guidelines. Tissue samples from female patients — aged 35–59 years and not receiving any oral antidiabetic treatment — were removed within the first 30 minutes of the surgical procedure and immediately frozen in liquid N2. Visceral fat was removed from the great omentum, and the subcutaneous fat was taken in the vicinity of the laparotomy incision. Based on biochemical and morphological parameters, patients were classified as lean (BMI, <25; fasting glucose [FG], 6 mM; oral glucose tolerance test [OGTT], <7.8 mM); obese (BMI, >35; FG, <6 mM; OGTT, <7.8 mM); obese glucose intolerant (BMI, >35; FG, between 6 and 7 mM, OGTT, between 7.8 and 11.1 mM) and obese diabetic (BMI, >35; FG, >7 mM; OGTT, >11.1 mM). Glycemia was assayed in the OGTT test 120 minutes after the glucose load.
Data are mean ± SEM. Calculations were carried out using Prism software (GraphPAD Inc.). QPCR, Western blot, and transient transfection experiments were analyzed with the 2-tailed Student’s t test. Statistical significance of differences between pairs of groups in animal studies was assessed using ANOVA followed by Tukey analysis. A P value less than 0.05 was considered significant.
The authors acknowledge the skillful technical assistance of M. André, B. Derudas, C. Brand, A. Lucas, and M. Ploton. We are indebted to M.-F. Six and the Centre d’Investigations Cliniques for human tissue samples; M. Brun, A. Géant, and H. Duez for mouse RNA and tissue samples; and J. Brozek (Genfit S.A.) for help with statistical analysis. This work was supported by research grants from Institut de Recherche Servier (IdRS), Région Nord-Pas de Calais/FEDER, and Fondation Coeur et Artères. A. Guédin, A. Langlois, B. Lefebvre, and Y. Benomar were supported by funds from IdRS.
Conflict of interest: The authors have declared that no conflict of interest exists.
Citation for this article: J Clin Invest. 2010;120(5):1454–1468. doi:10.1172/JCI38606.
Bruno Lefebvre’s present address is: INSERM U859, Faculté de Médecine de Lille-Pôle Recherche, Lille, France.
Luc Pénicaud’s present address is: Centre des Sciences du Goût et de l’Alimentation, UMR 6265 CNRS, Dijon, France.