Rudolf Magnus Institute of Neuroscience, Department of Neuroscience and Pharmacology, University Medical Center Utrecht, Utrecht, The Netherlands
Posttranslational Modifications of PPAR-γ: Fine-tuning the Metabolic Master Regulator
Version of Record online: 6 SEP 2012
2009 North American Association for the Study of Obesity (NAASO)
Volume 17, Issue 2, pages 213–219, February 2009
How to Cite
van Beekum, O., Fleskens, V. and Kalkhoven, E. (2009), Posttranslational Modifications of PPAR-γ: Fine-tuning the Metabolic Master Regulator. Obesity, 17: 213–219. doi: 10.1038/oby.2008.473
- Issue online: 6 SEP 2012
- Version of Record online: 6 SEP 2012
- Received March 11, 2008; Accepted August 03, 2008
Peroxisome proliferator-activated receptors (PPARs) are members of the nuclear receptor family of ligand-dependent transcription factors, which comprise a subgroup of three closely homologous genes, PPAR-α (NR1C1), PPAR-β/δ (NR1C2), and PPAR-γ (NR1C3) (refs. 1,2). All PPARs bind to PPAR responsive elements in the promoter regions of target genes as obligate heterodimers with retinoic acid X receptors. Similar to other nuclear receptors, PPARs consist of distinct functional domains including an N-terminal transactivation domain (AF1), a highly conserved DNA-binding domain (DBD), and a C-terminal ligand-binding domain (LBD) that contains a ligand-dependent transactivation function (AF2; Figure 1a). Ligand binding stabilizes the active conformation of the LBD, thereby serving as a “molecular switch” between activation and repression functions of the receptor (3). On the promoters of some target genes, unliganded PPARs recruit corepressors such as N-CoR and SMRT, which are part of multiprotein complexes containing histone deacetylase activity that repress gene transcription (4). Upon ligand binding, these corepressor complexes are exchanged for activating co-activator complexes, including the SRC1/CBP and TRAP/DRIP/ARC complexes (5). Besides ligand-mediated regulation, the transcriptional activity of nuclear receptors can also be regulated via the ligand-independent AF1 domain. In common with other NRs (6), the AF1 domains of the different PPAR isotypes probably lack a stable secondary structure in aqueous solution, only to adopt a stable conformation upon interaction with other proteins.
Although all PPAR nuclear receptors play a part in lipid homeostasis and energy metabolism, the different PPAR proteins exhibit different physiological roles due to (i) distinct expression patterns, (ii) specific activation by different ligands, and (iii) intrinsic functional differences between the different isotype proteins.
Alternative splicing and differential promoter usage results in two PPAR-γ isoforms, PPAR-γ1 and PPAR-γ2, with the latter harboring a 30-residue extension at its N-terminus (1,2). PPAR-γ1 is expressed in several tissues, including the lower intestine, macrophages, and white adipose tissue (WAT), whereas PPAR-γ2 expression is almost exclusively restricted to WAT. The endogenous ligands for PPAR-γ are not firmly established, although some natural compounds, such as polyunsaturated fatty acids, prostaglandin J2 derivatives (15-deoxy-δ12,14-PGJ2), and nitro-oleic acid, have been shown to be able to activate PPAR-γ (7,8). Synthetic PPAR-γ agonists include the thiazolidinediones (TZDs), which ameliorate insulin resistance and lower blood glucose levels in patients with type 2 diabetes (9). A huge amount of data indicates that PPAR-γ is one of the key players in the differentiation of fibroblast-like mesenchymal stem cells into adipocytes, a process known as adipogenesis. For example, PPAR-γ+/– mice lack most adipose tissue (10,11,12), whereas in vitro differentiation of fibroblasts into mature adipocytes can be induced by the introduction of PPAR-γ (13). In addition, PPAR-γ is also essential for the maintenance of adipose tissue, because conditional knock-out of the Pparg gene resulted in reduced in vivo survival of mature adipocytes (14). Besides its role in adipocyte differentiation and maintenance, PPAR-γ directly regulates the expression of a number of genes involved in net lipid partitioning into mature adipocytes, underscoring the importance of PPAR-γ in glucose and lipid homeostasis. Compelling genetic evidence for this view comes from human familial partial lipodystrophy subtype 3 (MIM 604367) patients, harboring heterozygous mutations in the PPARG gene, as they are characterized by aberrant fat distribution and metabolic disturbances, including insulin resistance and dyslipidemia (15). Interestingly, PPAR-γ was recently shown to be involved in the inhibition of osteoblast differentiation and osteoclastogenesis (16), indicating that novel biological functions of this transcription factor remain to be identified beyond its role in lipid and glucose homeostasis.
Although the three PPAR isotypes display a high degree in primary amino acid sequence homology in their LBD and DBD, both the N-terminal AF1 region and the hinge region are less well conserved between isotypes. The idea that these regions are therefore likely to contribute to the isotype-specific responses was recently supported experimentally. Using chimeric PPAR-γ-PPAR-β/δ proteins, the AF1 region of PPAR-γ was shown to be essential for adipogenesis (17,18). Furthermore, gene expression profiling revealed that the AF1 regions of the different PPAR family members are the main determinants of isotype-selective gene expression (18). It seems plausible that, in analogy with the AF2 domain, the activity of the AF1 region is dictated by the set of proteins with which this domain interacts. Only two AF1-interacting proteins have been identified so far which display isotype-specific interactions: PPAR-γ co-activator 2/SCAN domain protein 1 (PGC-2/SDP1), a PPAR-γ-specific co-activator, which promotes adipogenesis (17), and Tip60, an adipogenic acetyltransferase, which stimulates the activity of PPAR-γ but not PPAR-α or PPAR-β/δ ((19) and O. van Beekum and E. Kalkhoven, unpublished data). Interestingly, the AF1 region of PPAR-γ is also subject to various posttranslational modifications (PTMs), and it seems likely that such PTMs regulate interactions with coregulators, and/or vice versa, ultimately controlling the transcriptional output of this nuclear receptor. In this review, we will therefore summarize the various PTMs reported for PPAR-γ, evaluate the (possible) effects on coregulator interactions, and discuss the (potential) relevance for in vivo PPAR-γ signaling.
Phosphorylation of PPAR-γ
As the first PTM reported for PPAR-γ2, phosphorylation has been studied extensively over the years. Phosphorylation of PPAR-γ2 was mapped to serine 112 (S112; S82 in PPAR-γ1), a conserved mitogen-activated protein kinase (MAPK) consensus site within the AF1 region, which is the only phosphorylation site identified in PPAR-γ so far (Figure 1b). Activation of the MAPKs extracellular signal-regulated kinase 1/2 by growth factors (epidermal growth factor, platelet-derived growth factor, transforming growth factor-β, insulin) or the prostaglandin PGF2α was indeed found to result in increased phosphorylation of S112 (refs. 20,21,22,23,24,25). In addition, the activation of MAPKs, c-Jun N-terminal kinase 1/2 and p38, by stress (UV, anisomysin) also resulted in hyperphosphorylation of S112 (refs. 25,26; Figure 1c). Phosphorylation of PPAR-γ by these treatments resulted in decreased transcriptional activity in reporter assays, whereas the mutation of the phosphorylation site by changing S112 into alanine led to increased transcriptional activity (20,22,25,26,27). Furthermore, several laboratories have shown that the overexpression of PPAR-γ S112A in mouse fibroblasts resulted in increased adipogenesis compared with cells overexpressing wild-type PPAR-γ (20,23,25,28,29,30). Growth factor- or stress-induced phosphorylation of S112 of PPAR-γ, therefore, in general, represses its transcriptional and adipogenic functions. Treatment of cells with insulin, however, has been associated with reduced (20) or increased PPAR-γ activity (24). This discrepancy may be explained by the modulation of the signaling cascade through adapter molecules such as Dok1 (Figure 1c; (31)). Although Ras/MAPK activation by insulin normally represses PPAR-γ activity (20), high levels of Dok1 in adipose tissue, as induced by excessive caloric intake, were found to counteract this repression (31).
Phosphorylation of PPAR-γ represses its activity probably through multiple molecular mechanisms. Adams et al. showed that the phosphorylation affected both ligand-independent and ligand-dependent transcription, based on reporter assays using full-length PPAR-γ proteins (25). In addition, their experiments with fusions of the AF1 region to a heterologous DBD (Gal4DBD) also revealed reduced transcriptional activity of the S112A mutant, suggesting either the recruitment of a repressor protein to the AF1 region or the release of an activator in a phosphorylation-dependent manner (25). Subsequent studies showed that PPAR-γ phosphorylation affected ligand binding, and hence reduced transcriptional activity, indicative for AF1-LBD interdomain communication (28). Finally, phosphorylation of S112 has been associated with a repressive PTM named sumoylation, as will be discussed later.
In contrast to growth factor- or stress-induced phosphorylation of S112, which results in reduced transcriptional activity, modification of the same residue by the cyclin-dependent kinases, cdk7 and cdk9, was recently found to increase PPAR-γ activity (Figure 1d; (32,33)). The cdk7 kinase is a subunit of the basal transcription factor general transcription factor II H complex, which plays important roles in DNA repair and transcription (34). Disruption of the general transcription factor II H complex, for example, by mutations in the xeroderma pigmentosum complementation group D (XPD) subunit as observed in xeroderma pigmentosum patients, results in predisposition for carcinogenesis caused by increased photosensitivity. PPAR-γ is a bona fide substrate for the cdk7 kinase, because (i) cdk9 could directly phosphorylate PPAR-γ on S112 in vitro; (ii) PPAR-γ was found to be hypophosphorylated in XPD-deficient mice, which lack a functional cdk7-containing general transcription factor II H complex; and (iii) the reintroduction of XPD proteins into these cells restored PPAR-γ phosphorylation levels. However, in contrast to previous studies showing that the phosphorylation of S112 was inhibitory (20,22,25,26,27), cdk7-mediated phosphorylation stimulated PPAR-γ transcriptional activity, as PPAR-γ target gene expression levels were lower in XPD-deficient mice when compared with wild-type mice (32). It should be noted that this effect was specific for brown adipose tissue (BAT), because XPD-deficient mice displayed increased PPAR-γ target gene expression in WAT, especially in the absence of exogenously added ligand (32). Together with the hypoplasia of adipose tissue observed in XPD patients and XPD-deficient mice, these findings suggest that the phosphorylation of PPAR-γ may stimulate its transcriptional activity in certain tissues (e.g., BAT) or under certain conditions. In agreement with this hypothesis, Iankova et al. recently reported that cdk9, which together with cyclin T1/2 forms the positive transcription elongation factor b complex (35), can phosphorylate S112 of PPAR-γ (Figure 1d; (33)). The p55 isoform of cdk9, which is strongly upregulated during adipogenesis, was found to interact with the AF1 region of PPAR-γ and phosphorylate the protein on S112 in vitro (33). Furthermore, the overexpression of cdk9 enhanced PPAR-γ-mediated transcription and stimulated adipocyte differentiation (33). Taken together, these findings indicate that the net result of PPAR-γ phosphorylation may be either inhibition or stimulation of transcriptional activity, depending on the cellular background and kinases involved (Figure 1c,d).
Sumoylation of PPAR-γ
Recently, several research groups have reported a PTM that adds another layer of modulating PPAR activity: sumoylation (36,37,38,39). The covalent attachment of small ubiquitin-like modifier (SUMO) peptides (SUMO-1, −2, −3 in mammals) to lysine residues involves an activating enzyme (SAE1/SAE2), a conjugating enzyme (Ubc9), and an E3 ligase (e.g., PIAS1) (ref. 40). This modification occurs on consensus sumoylation motifs ψKXE/D in the substrate proteins, where ψ represents a large hydrophobic residue followed by a lysine, which is the SUMO acceptor site, and X may be any residue. Mutation of either the lysine residue or the acidic residue at position +2 has been shown to ablate sumoylation on these sites (40). Ligation of SUMO peptides, which are ∼100 amino acids long, is linked to various cellular processes, including nuclear-cytoplasmatic transport, apoptosis, and transcriptional regulation (40). Sumoylation of transcriptional regulators mostly correlates with the inhibition of transcription (40). Two functional sumoylation sites have been identified for PPAR-γ, lysine 107 in the AF1 region and lysine 395 in the AF2 region (lysine 77 and 365 in PPAR-γ1, respectively). Conjugation of SUMO-1 or −2 to lysine 107 by the E3 SUMO ligases PIAS1 or PIASxβ modulates PPAR-γ activity in a negative manner, because (i) mutation of K107 itself or distortion of the sumoylation consensus motif by mutating glutamic acid at position 109 into alanine (E109A) increased PPAR-γ activity, and (ii) the overexpression of a dominant-negative form of the SUMO E2-ligase Ubc9 had the same effect. These data indicate that sumoylation of the AF1 domain induces active repression. Although K107 sumoylation has been reported to decrease the stability of the PPAR-γ protein (37), this finding has not been supported by other studies (36,38,39). The exact molecular mechanism behind sumoylation-mediated repression of PPAR-γ therefore remains to be established. Possibly, the repressive effect involves specific binding of a repressor complex, because (i) the deletion of the AF1 region results in increased transcriptional activity (13,19), and (ii) Gal4DBD-AF1 fusion proteins, which rule out DNA- and ligand-binding effects, show increased transcriptional activity upon mutation of K107 (ref. 36).
Interestingly, K107 sumoylation is to some extent linked to S112 phosphorylation. K107 and S112 are part of a so-called sumoyl-phospho switch (41) or phosphorylation-dependent sumoylation motif (42), a conserved motif present in at least 80 different proteins including several transcription factors and co-activator s. The consensus site consists of the following motif: ψKxExxSP, in which ψ is a hydrophobic residue, K is the sumo acceptor lysine, x is any amino acid, and SP forms a part of the downstream phosphorylation site. Between the different PPAR members, this motif is only found in the AF1 region of PPAR-γ and conserved from fish to humans, as shown in Figure 1b. Mutation of the serine 112 to alanine, which ablates phosphorylation, significantly diminished lysine 107 sumoylation while the opposite held true for the PPAR-γ2 phospho-mimic S112D, which shows increased sumoylation (36). A similar interplay between phosphorylation and sumoylation was recently reported for the nuclear receptors estrogen-related receptor-α and estrogen-related receptor-γ (43), which also contain a phospho-sumoyl switch (Figure 1b). Several models can be drawn as to how the interplay between phosphorylation and sumoylation may result in transcriptional repression (Figure 2a). First, phosphorylation of S112 may create in a docking site for a specific SUMO E3 ligase, which in turn modifies lysine 107, resulting in SUMO-specific recruitment of a repressor complex (Model 1; Figure 2a). Alternatively, the combined modification of S112 and K107 may generate a docking site for a repressor complex (Model 2; Figure 2a). Repressors that are recruited to gene promoters in a sumoylation-dependent manner include the transcriptional repressor protein Daxx (44), the DEAD-box protein DP103/Ddx20 (ref. 45), and the Mi-2 repressor complex (46), but further research is required to establish whether any of these proteins or another yet to be identified repressor protein (complex) is involved in SUMO-mediated repression of PPAR-γ activity.
In contrast to lysine 107 sumoylation, conjugation of SUMO-1 to lysine K395 is not involved in the regulation of direct PPAR-γ target genes, but in the transrepression of inflammatory genes by PPAR-γ in macrophages, like the inos gene (47). Treatment with ligand results in sumoylation of K395, which in turn targets PPAR-γ2 to NCoR corepressor complexes that are bound to NFκB target genes before activation (Figure 2b). These NCoR complexes are no longer cleared from the promoter once PPAR-γ2 is bound, resulting in sustained repression. A similar mechanism was recently reported for the transrepression of NFκB target genes by the nuclear receptors LXRα and LXRβ, although in this case SUMO-2 or SUMO-3 and not SUMO-1 was conjugated to the receptor (48).
Ubiquitination of PPAR-γ
Ubiquitination is the covalent attachment of ubiquitin, a 76-amino-acid peptide to lysine residues in the substrate protein. Although conjugation of ubiquitin chains linked through lysine 48 (polyubiquitination) primarily labels proteins for proteasomal degradation, attachment of single ubiquitin molecules, or with lysine 63-linked chains (monoubiquitination) is linked to many different cellular processes (49). Similar to sumoylation, ubiquitination requires an activating enzyme (E1), a conjugating enzyme (E2), and an E3 ubiquitin ligase (HECT or RING protein). A growing body of evidence indicates that the RNA polymerase II machinery and the ubiquitin-proteasome system are intimately linked (50). Groundbreaking studies were performed by the Gannon and O'Malley laboratories, who showed that the proteasomal degradation of transcription factors is an essential step in the regulation of its target genes, possibly by enabling the sequential formation of protein complexes at the promoter region (51,52). The PPAR-γ protein has a short half-life (t½=2 h) (53,54) and was found to be polyubiquitinated and degraded by the proteasome (55,56). PPAR-γ2 ubiquitination and degradation are strongly linked to ligand binding and activation, as TZDs accelerated these processes, whereas the mutation of glutamic acid 499 to glutamine (E499Q) abrogated both activation and degradation (53,55,56). The exposure of adipocytes to the cytokine interferon-γ, as may occur upon infiltration of interferon-γ-producing lymphocytes into adipose tissue (57), was also found to increase ubiquitination and degradation of PPAR-γ (53,56), indicating that phosphorylation is not the only PTM that can be regulated by external stimuli. Interestingly, as described earlier for sumoylation, ubiquitination also appears to be regulated by phosphorylation. The hypophosphorylated form of PPAR-γ, which often displays increased transcriptional activity (as described earlier), was found to be degraded more rapidly than the phosphorylated protein (55,56), supporting a direct link between protein degradation and transcriptional activity (50). It is currently unknown how the phosphorylation status of PPAR-γ may regulate its ubiquitination and subsequent degradation. It should be noted that no ubiquitination acceptor lysines have been identified for the PPAR-γ protein yet. Current progress in mass spectrometry, however, holds the promise that ubiquitination sites will actually be identified in the near future (58).
Physiological function of PPAR-γ posttranscriptional modifications
With a few exceptions, the in vivo relevance of PTMs of PPAR-γ has not been addressed experimentally. In an elegant study, Rangwala et al. showed that homozygous PPAR-γ S112A knock-in mice are no longer prone to develop insulin resistance when fed a high-fat diet (59). Although the S112A mutation was previously shown to turn PPAR-γ into a more potent inducer of adipogenesis in fibroblasts (20,23,25,28,29,30), the S112A alleles did not produce or exacerbate obesity in mice, because no significant differences in weight or percentage body fat were detected between mutant and wild-type mice. It was suggested that the underlying mechanism for the protection from obesity-induced insulin resistance lies in decreased adipocyte size and secreted adipocytokines (59).
Interestingly, heterozygous mutation of proline 113 (P113; often referred to as P115), which, similar to the S112A mutation results in a nonphosphorylatable form of PPAR-γ, was reported as a rare genetic event in the German population (29,60). In contrast to the S112A mice, four individuals with this mutation were all severely obese (BMI 37.9–47.3), with three out of four patients having developed type 2 diabetes (based on low fasting insulin levels) (29). The PPAR-γ P113Q mutant was a more potent inducer of adipogenesis compared with the wild-type protein (29), in accordance with the characterization of the nonphosphorylatable S112A protein by several laboratories (20,23,25,28,29,30). At first sight, the data from S112A mice and P113Q patients appear to be in clear disagreement with each other, suggesting fundamental differences between mice and humans, as well as between in vitro and in vivo studies. However, this interpretation is probably premature. First, comparing the mouse and human data is complicated by the fact that Rangwala et al. characterized homozygous S112A mice (59), whereas human P113Q mutations are heterozygous (29,60). Second, after the first report on human P113Q mutations, Blüher and Paschke subsequently described a fifth German P113Q case who was clearly phenotypically different: in contrast to the previous four cases this individual (BMI 28.5) displayed high fasting insulin levels and profound insulin resistance (60). These findings indicate that the heterozygous P113Q mutation is not sufficient to cause obesity (and diabetes) in humans, but that other genetic and/or environmental factors are required. The low frequency of the P113Q mutation, however, hampers experimental testing of this hypothesis. Whereas Ristow et al. found four P113Q patients in a cohort of 358 individuals (29), and Blüher and Paschke reported one case among 48 obese individuals (60), and no carriers were found in two German cohorts (85 and 67 obese individuals; (61,62)) or a French cohort of 626 obese individuals (63). Additional studies on homozygous and heterozygous S112A and P113Q mice in different genetic backgrounds and under different dietary regimes may therefore be a better approach to solve these issues.
Clearly, the physiological relevance of PTMs of PPAR-γ is understudied at present. So far, only the in vivo effects of mutating S112 in PPAR-γ were studied (59), which has given interesting new insights in PPAR-γ action. It will therefore be essential for our understanding of PPAR-γ action to extend such approaches to other posttranslationally modified residues.
Since Hu et al. reported in 1996 that PPAR-γ was a phosphoprotein (20), numerous studies have followed showing that PPAR-γ is subject to many different PTMs, which determine its transcriptional output (Figure 3). Although the modifying enzymes and substrate amino acids were identified in most cases, it is often still largely unknown how these modifications alter the function of the PPAR-γ protein. Two molecular mechanisms, which are not mutually exclusive, are most likely to occur. First, a PTM could predominantly have an allosteric, intramolecular effect, as was shown for serine 112 phosphorylation of PPAR-γ that reduces ligand binding (28). Alternatively, the effect of a PTM on PPAR-γ function could be mainly intermolecular, i.e., allowing either the recruitment of novel interacting proteins or the release of proteins that were bound to the unmodified receptor. This docking of so-called effector proteins to modified forms of PPAR-γ represents an attractive molecular control mechanism, but (most of) the effector proteins remain to be identified.
Second, little attention has so far been paid to the fact that all the PTMs described here for PPARs can probably undergo demodification by specific cellular enzymes. So far, no enzymes have been identified which can deubiquitinate or dephosphorylate (phosphatases) PPAR-γ, although these factors may play an extremely important role in balancing the transcriptional output.
Another area that is clearly understudied and needs further exploration is whether a given PTM alters the expression of all PPAR-γ target genes or just a specific subset of target genes. One example may be the PPAR-γ-mediated repression of the GLUT4 gene, which requires an intact S112 residue (64), but it is not clear how general this theme is. Recently emerging technologies such as ChIP on-chip (65) and ChIPseq (66) may help to address the intriguing link between PTMs and target gene specificity.
Finally, a great deal more needs to be done to investigate potential cross talk between the different PTMs (Figure 3). Such cross talk may be positive, as shown for the so-called phospho-sumoyl switch motif (41) or phosphorylation-dependent sumoylation motif (42) in PPAR-γ2, whereby phosphorylation of S112 precedes sumoylation of K107 (ref. 36). Negative cross talk has been described for ubiquitination, which is inhibited by S112 phosphorylation through an unknown mechanism (55,56). A different type of negative cross talk, in which two modifying enzymes compete for the same substrate residue, has been described for many proteins (67,68). In the case of PPAR-γ, this may occur on lysine 107, which has been shown to be sumoylated (36,38,39), and possibly also acetylated (O. van Beekum and E. Kalkhoven, unpublished data). Cross talk between PTMs poses the possibility of integrating multiple signaling pathways: the combination of different extracellular and intracellular signals could lead to a specific pattern of PPAR-γ modifications, which is “read” by the effector proteins, ultimately dictating the transcriptional output (Figure 3).
Obesity and type 2 diabetes are ever increasing problems in industrialized countries, affecting millions of people worldwide. The physiological roles of PPAR-γ in lipid and glucose metabolism have quickly led to the therapeutic use of synthetic PPAR-γ ligands of the TZD class for the treatment of insulin resistance. Treatment of diabetic patients with synthetic PPAR-γ ligands of the TZD class has however been linked to adverse side effects such as undesired weight gain, fluid retention, peripheral edema, and potential increased risk of cardiac failure (69). These adverse side effects may be due to the use of high doses of full PPAR-γ agonists, suggesting that “activation in moderation” may be a more sensible approach (2). This may be achieved through the use of compounds displaying partial agonism, so-called selective PPAR modulators (70). Alternatively, modulating the PTMs on PPAR-γ may present a novel approach in this respect (Figure 3). Therefore, increasing our knowledge of PPAR PTMs, the modifying and demodifying enzymes involved, the effector proteins, the specific transcriptional output, and the cellular and in vivo effects can be of great value for the development of novel pharmaceutical approaches to fight the various illnesses in which the PPAR-γ protein plays a role.
We thank our colleagues A. Koppen and E.H. Jeninga for critically reading the manuscript and S. Mandrup and A.K. Bugge (University of Southern Denmark, Odense, Denmark) for helpful discussions. We apologize to those whose original work could only be cited indirectly due to space limitations.
The authors declared no conflict of interest.
- 65Genome-wide approaches for identification of nuclear receptor target genes. Nucl Recept Signal 2006; 4: e018., , .