Reversible phosphorylation is a widespread molecular mechanism to regulate the function of cellular proteins, including transcription factors. Phosphorylation of the nuclear receptor PPARγ (peroxisome-proliferator-activated receptor γ) at two conserved serine residue (Ser112 and Ser273) results in an altered transcriptional activity of this transcription factor. So far, only a very limited number of cellular enzymatic activities has been described which can dephosphorylate nuclear receptors. In the present study we used immunoprecipitation assays coupled to tandem MS analysis to identify novel PPARγ-regulating proteins. We identified the serine/threonine phosphatase PPM1B [PP (protein phosphatase), Mg2+/Mn2+ dependent, 1B; also known as PP2Cβ] as a novel PPARγ-interacting protein. Endogenous PPM1B protein is localized in the nucleus of mature 3T3-L1 adipocytes where it can bind to PPARγ. Furthermore we show that PPM1B can directly dephosphorylate PPARγ, both in intact cells and in vitro. In addition PPM1B increases PPARγ-mediated transcription via dephosphorylation of Ser112. Finally, we show that knockdown of PPM1B in 3T3-L1 adipocytes blunts the expression of some PPARγ target genes while leaving others unaltered. These findings qualify the phosphatase PPM1B as a novel selective modulator of PPARγ activity.
The transcription factor PPARγ (peroxisome-proliferator-activated receptor γ) is a ligand-activated transcription factor of the nuclear receptor superfamily that regulates genes involved in differentiation, metabolism and immunity . Similar to other nuclear receptors, PPARγ consists of distinct functional domains including an N-terminal transactivation domain [AF (activation function)-1], a highly conserved DBD (DNA-binding domain) and a C-terminal LBD (ligand-binding domain) that contains a ligand-dependent transactivation function (AF-2). The LBD of PPARγ can accommodate a wide variety of ligands such as prostaglandins, eicosanoids and fatty acids. TZDs (thiazolidinediones), a class of anti-diabetic drugs including rosiglitazone, also function as ligands for PPARγ. Ligand binding stabilizes the active conformation of the PPARγ LBD, resulting in the release of corepressor proteins [e.g. NCoR (nuclear receptor corepressor)] and the recruitment of co-activator proteins. Ligand binding thereby serves as a ‘molecular switch’ between the activation and repression functions of the receptor. It has however become increasingly clear that PTMs (post-translational modifications) represent an additional important molecular mechanism to regulate the activity of nuclear receptors, including PPARγ [2,3]. The first PTM described for PPARγ2 was phosphorylation of serine residue 112 (Ser82 in PPARγ1) by various kinases such as ERK1 (extracellular-signal-regulated kinase 1), p38, JNK (c-Jun N-terminal kinase), CDK (cyclin-dependent kinase) 7 and CDK9 (reviewed in [2,3]). Ser112 phosphorylation can impair PPARγ transcriptional activity, and several non-mutually exclusive mechanisms have been reported: inhibition of ligand binding , stimulation of the repressive PTM SUMOylation [5,6] or reduced DNA binding owing to interaction with the PER2 (period circadian protein homologue 2) protein . It should be noted that Ser112 phosphorylation can also result in increased PPARγ activity [8,9], but the molecular mechanism(s) underlying this differential output are currently unclear. Previously a second phosphorylation site, Ser273, was identified . Phosphorylation of this residue by CDK5, a kinase which is activated in obesity, results in reduced expression of specific PPARγ target genes that have anti-obesity effects . Interestingly, treatment with PPARγ agonists leads to decreased phosphorylation at this site, an effect which may be mediated by the corepressor protein NCoR [10–12].
Protein phosphorylation events are generally reversible and many nuclear receptors are subject to phosphorylation . However, only a very limited number of cellular enzymatic activities have been described so far which can dephosphorylate nuclear receptors, including the phosphatase PP (protein phosphatase) 5 for PPARγ and the GR (glucocorticoid receptor) ([14,15]; also see the Discussion section). Previous studies using high-throughput siRNA (small interfering RNA) screening have highlighted the importance of phosphatases in many different cellular processes, including cell survival, apoptosis and cell-cycle progression (e.g. ). Protein phosphatases are defined by structurally distinct gene families which can be divided into two major classes: the protein serine/threonine phosphatase family and the PTP (protein tyrosine phosphatase) family, including both tyrosine-specific and dual-specificity phosphatases . Protein serine/threonine phosphatases are further classified into two subfamilies, PPP (phosphoPP) and PPM (metallo-dependent PP), on the basis of substrate specificity, divalent cation dependency and sensitivity to specific inhibitors. The PPP family includes PP1, PP2A and PP2B, whereas the PPM family consists of the PP2C isozymes and pyruvate dehydrogenese phosphatase. The PPM family of PPs differ from other phosphatases since they: (i) depend on Mg2+ or Mn2+ ions for their catalytic activity; (ii) function as monomers; and (iii) are insensitive to the phosphatase inhibitor okadaic acid. PPM family members have been reported to function in the regulation of different cellular processes including apoptosis, cell cycling and differentiation [17,18]. To conform to the nomenclature for the human Mg2+-dependent phosphatases, in the present paper PP2Cα and PP2Cβ will be denoted as PPM1A and PPM1B respectively. In the present study we used immunoprecipitation assays coupled with MS analysis to identify novel PPARγ-associated proteins. We describe the serine/threonine phosphatase PPM1B as a novel PPARγ-interacting protein, which can selectively modulate PPARγ-mediated transcription.
Rosiglitazone was purchased from Alexis. FuGENE®6 transfection reagent and protease inhibitor tablets (catalogue number 11697498001) were purchased from Roche Applied Biosciences. PEI (polyethyleneimine; catalogue number 23966) was purchased from Polysciences. The following antibodies were used: anti-PPARγ (catalogue number sc-7273, Santa Cruz Biotechnologies); anti-PPM1B (catalogue number A300-887A, Bethyl Laboratories); HRP (horseradish peroxidase)-conjugated anti-FLAG M2 (A8592), anti-HA (haemagglutinin; H9658) and anti-tropomyosin (T2780) (Sigma–Aldrich); HRP-conjugated anti-rabbit IgG (111035144) and HRP-conjugated anti-mouse IgG (115035146) (Jackson Immunoresearch Laboratories); anti-(PPARγ phosphoSer112) (Euromedex); and anti-(PPARγ phosphoSer112) (ab60953, Abcam). Anti-FLAG M2–agarose beads (A02220), Oil-Red-O (O-0625) and purified PPM1B (P-1743) were purchased from Sigma–Aldrich, and Lipofectamine™ 2000 was purchased from Invitrogen.
All recombinant DNA work was performed according to standard procedures . pCMV (plasmid cytomegalovirus) Renilla luciferase was purchased from Promega. All mutations were generated by QuikChange® mutagenesis (Stratagene) and verified by sequencing. The pSport PPM1B isoform 1 and PPM1A expression vectors were purchased from RZPD (clones IRAT p970B0984D and IRAT p970A1077D). The pCDNA-Gal4DBD-PPARγAF1 (amino acids 1–129) was generated by cloning a PCR fragment into the pCDNA3-Gal4DBD vector . The reporter plasmid containing the aquaporin promoter was a gift from Dr N. Maeda (Osaka University, Osaka, Japan) . All other plasmids have been described previously [22,23].
Cell culture, transient transfections and reporter assays
The HEK (human embryonic kidney)-293T cell line and the human osteosarcoma cell line U2OS and were maintained in DMEM (Dulbecco's modified Eagle's medium) Glutamax containing 10% fetal bovine serum (Gibco Life Technologies), 100 μg of penicillin/ml and 100 μg of streptomycin/ml (Gibco Life Technologies). The murine 3T3-L1 cells were cultured in the same medium, but with 10% bovine serum (Gibco Life Technologies), 100 μg of penicillin/ml and 100 μg of streptomycin/ml (Gibco Life Technologies). For differentiation, 3T3-L1 cells were grown to confluency and after 2 days incubated with culture medium containing dexamethasone (250 nM), 3-isobutyl-1-methylxanthine (500 μM) and insulin (170 nM) for 2 days. On day 3, medium was changed for culture medium supplemented with insulin (170 nM) and left for 2–5 days. Subsequently, cells were stained with Oil-Red-O, or lysed and subjected to Western blot analysis as described previously [22,23].
Reporter assays were performed in 24-well plates with 1 μg of 3xPPRE-tk-Luc or AQP7-Luc reporter construct, 2 ng of PPAR expression construct, 100 ng of PPM1B expression construct (or empty vector) and 2 ng of pCMV-Renilla (Promega) as described previously . For immunoprecipitation experiments, U2OS cells or HEK-293T cells were grown in 15 cm dishes and transiently transfected with PPARγ2 or PPM1B expression vectors (10 μg) using either FuGENE®6 or PEI transfection reagent, as described previously .
Overexpressed FLAG–PPARγ2 was isolated from 20 dishes (15 cm) of HEK-293T cells. Proteolytic digestion of proteins and LC (liquid chromatography)-tandem MS were performed exactly as described previously [23,24], except that proteins were not only subjected to trypsin, but also chymotrypsin and Arg-C, digestion. Moreover, the raw data files were processed by Mascot Distiller (version 22.214.171.124, Matrix Science) and searched against the SwissProt database (release 2012_07) using the Mascot Search Engine (version 2.3.01).
Western blot analysis
For detection of phosphorylated PPARγ in intact cells, FLAG–PPARγ2 was overexpressed in HEK-293T cells. The next day, the cells were incubated with or without rosiglitazone (2 μM) for 24 h prior to lysis in RIPA buffer [200 mM Tris/HCl (pH 8.0), 1% Triton X-100, 1% sodium deoxycholate, 0.1% SDS, 10mM EDTA and 150 mM NaCl with protease inhibitors). After immunoprecipitation with anti-FLAG M2–Sepharose beads for at least 4 h at 4°C, the samples were analyzed by Western blotting. Blots were probed with various primary antibodies and immunoreactive complexes were visualized by enhanced chemiluminescence as described previously .
In vitro dephosphorylation assay
FLAG-tagged PPARγ2 was immunopurified from transiently transfected HEK-293T cells. Purified PPM1B was incubated in phosphatase buffer [20 mM Tris (pH 7.4), 150 mM NaCl, 5mM MgCl2, 1 mM DTT (dithiothreitol), 1 mg/ml BSA, 0.1% Tween and complete protease inhibitors] together with FLAG–PPARγ2 coupled with 25 μl of FLAG beads for 30–45 min at 30°C. Phosphorylated PPARγ2 was detected by Western blot analysis using phosphoSer112-specific antibody labelling.
Immunofluorescence microscopy and PLA (proximity ligation assay)
Immunofluorescence staining was performed as described previously , using primary antibodies against PPM1B (A300-887A) and PPARγ (sc-7273). PLA detection was performed using the Duolink II kit (Olink Bioscience) according to the manufacturer's protocol. In short, differentiated 3T3-L1 cells were enabled to adhere to coverslips and subsequently washed with PBS and fixed with 4% formaldehyde for 20 min. Afterwards, the samples were permeabilized with 0.2% Triton X-100 for 5 min and then incubated for 30 min with blocking buffer (10% normal human serum in PBS). After blocking, cells were incubated overnight at 4°C with anti-PPM1B (A300-887A) and anti-PPARγ antibodies (sc-7273) in blocking buffer containing 10% normal human serum. The cells were washed three times with PBS, followed by PLA according to the manufacturer's protocol. Coverslips were mounted and analysed as for immunofluorescence microscopy.
RNA isolation and quantitative RT (real-time)-PCR
The 3T3-L1 fibroblasts were differentiated as described above. Three independent samples of total RNA were extracted at different time points using TRIzol® reagent (Invitrogen). cDNA was synthesized using the superscript first-strand synthesis system (Invitrogen) according to the manufacturer's protocol. The gene expression levels were determined by quantitative RT-PCR with the MyIq cycler (Bio-Rad Laboratories) using SYBR Green (Bio-Rad Laboratories) and normalized to TfIIb (general transcription factor IIB) expression.
The primers used were as follows: murine TFIIb forward, 5′-TCCTCCTCAGACCGCTTTT-3′ and reverse, 5′-CCTGGTTCATCATCGCTAATC-3′; murine Pparγ forward, 5′-CGCTGATGCACTGCCTATGA-3′ and reverse, 5′-AGAGGTCCACAGAGCTGATTCC-3′; murine PPM1B primer, 5′-TCAGAGTTGGATAAGCACTTGG-3′ and reverse, 5′-CATCACATGGGCAAGATCAG-3′; murine Fabp4 (fatty acid-binding protein 4) primer, 5′-CGCAGACGACAGGAAGGT-3′ and reverse, 5′-TTCCATCCCACTTCTGCAC-3′; murine Adipoq (adiponectin, C1Q and collagen domain-containing) primer, 5′-GGAACTTGTGCAGGTTGGAT-3′ and reverse, 5′-TCTCCAGGCTCTCCTTTCCT-3′; murine Lpl (lipoprotein lipase) forward, 5′-TTTGTGAAATGCCATGACAAG-3′ and reverse, 5′-TCAAACACCCAAACAAGGGTA-3′; murine Cd36 forward, 5′-TTGTACCTATACTGTGGCTAAATGAGA-3′ and reverse, 5′-TCTACCATGCCAAGGAGCTT-3′; and murine Lpin1 (lipin 1) forward, 5′-CGCCAAAGAATAACCTGGAA-3′ and reverse, 5′-TGAAGACTCGCTGTGAATGG-3′.
3T3-L1 cells were transfected with siRNA oligonucleotides as described previously  using Lipofectamine™ RNAiMax (Invitrogen) according to the manufacturer's protocol. The siRNA oligonucleotides used were siControl (D-001810-10-20, Dharmacon), siPPARγ (L-040712-00-0010, Dharmacon), siPPM1B #6 (J-040053-06-0005, Dharmacon) and siPPM1B #5 (J-040053-05-0005, Dharmacon).
PPARγ interacts with the phosphatase PPM1B (PP2Cβ)
In order to identify novel PPARγ-interacting proteins, ectopically expressed FLAG-tagged human PPARγ2 was immunopurified from HEK-293T cells and the associated proteins were identified by MS analysis . Several peptides corresponding to the serine/threonine phosphatase PPM1B (also known as PP2Cβ) were repeatedly identified in independent experiments (Figures 1A and 1B, and Supplementary Table S1 at http://www.biochemj.org/bj/451/bj4510045add.htm). At least five different splice variants of PPM1B have been reported in humans and four in mice (Figures 1A and 2A, and Supplementary Table S2 at http://www.biochemj.org/bj/451/bj4510045add.htm), all identical in their catalytic domains, but differing in the N- and C-terminal domains which are thought to be involved in localization and substrate specificity [26,27]. Three peptides (e.g. SGSTAVGVMISPK) corresponded to the N-terminal region not present in isoform 3, whereas one of the peptides detected (QLLEEMLTSYR) corresponded to a region only present in human PPM1B isoforms 1 and 3. Although we cannot formally exclude that PPARγ may interact with isoform 3, PPM1B isoform 1 is more likely to represent a genuine PPARγ-interacting protein as: (i) expression of human isoform 3 on the protein level has never been reported, whereas human isoform 1 is expressed ubiquitously on the mRNA and protein level ; and (ii) no mRNA corresponding to human isoform 3 has been described in other species (e.g. mouse; Figure 2A), whereas isoform 1 is highly conserved (Supplementary Figure S1 at http://www.biochemj.org/bj/451/bj4510045add.htm). We therefore focused on the interaction between PPARγ and PPM1B isoform 1 in our subsequent experiments. To confirm the interaction between these two proteins by independent means, HEK-293T cells were co-transfected with HA-tagged PPM1B and FLAG-tagged PPARγ, followed by immunoprecipitation. As shown in Figure 1(C), PPM1B co-immunoprecipitated with FLAG-tagged PPARγ. As controls, immunoprecipitations were performed on the lysates of cells expressing either PPM1B or PPARγ alone, or neither protein. No co-immunoprecipitations were detected in these lysates (Figure 1C). Taken together, these findings indicate that PPM1B isoform 1 is a novel PPARγ-interacting protein.
The phosphatase PPM1B is a novel PPARγ-interacting protein
PPARγ and PPM1B interact in 3T3-L1 adipocytes
Nuclear PPM1B interacts with PPARγ in adipocytes
Given the key role of PPARγ in adipocyte differentiation, function and maintenance , we examined the expression and subcellular localization of the PPARγ-interacting phosphatase PPM1B in mouse 3T3-L1 adipocytes. To address which PPM1B isoform(s) are expressed in these cells, expression of the different mouse PPM1B isoforms was analysed by RT-PCR. Whereas expression of isoforms 2–4 could not be detected, mouse PPM1B isoform1, which encodes a protein highly homologous to the human PPM1B isoform 1 protein (Supplementary Figure S1), was clearly expressed in 3T3-L1 adipocytes (Figure 2A). Next, we examined PPM1B protein expression during differentiation of 3T3-L1 cells into mature adipocytes. The expression levels of PPM1B isoform 1 increased up to day 4, whereas PPARγ expression was detectable from day 3 onwards (Figure 2B). As both cytoplasmic and nuclear localizations of PPM1B have been reported previously [27,29–31], we investigated the localization of the endogenous PPM1B protein in differentiated 3T3-L1 cells. For this immunofluorescence staining of PPM1B was combined with PPARγ staining to mark differentiating cells. PPM1B and PPARγ both displayed nuclear localization in mature 3T3-L1 adipocytes (Figure 2C). Finally, we validated the PPM1B–PPARγ interaction by performing an in situ PLA on the endogenous proteins in undifferentiated and differentiated 3T3-L1 adipocytes (see the Experimental section). Since a PLA signal can only be obtained when the proteins of interest are in extremely close proximity, this technique enables the detection of direct protein–protein interactions in cells. Association of PPM1B and PPARγ was observed specifically in differentiated cells, and localized in the nucleus (Figure 2D). Taken together, these data indicate that PPM1B is expressed in the nuclei of mature 3T3-L1 adipocytes, where it can interact with PPARγ.
PPM1B directly dephosphorylates PPARγ
Given the interaction between the phosphatase PPM1B and PPARγ (Figures 1 and 2D), and since PPM1B is expressed in mature adipocytes (Figure 2), we investigated whether PPARγ may be a direct PPM1B dephosphorylation substrate. First, FLAG–PPARγ and HA–PPM1B were transiently co-expressed in HEK-293T cells followed by FLAG immunoprecipitation. To detect PPARγ phosphorylation, a phosphoSer112-specific antibody was used. Phosphorylated PPARγ was readily detected either in absence or presence of the synthetic PPARγ ligand rosiglitazone (Figure 3A, lanes 3 and 4). Co-expression of PPM1B resulted in a significant decrease of PPARγ phosphorylation, both in the absence and presence of ligand (Figure 3A, lanes 5 and 6). As a control, a PPM1B catalytic mutant was used in which a conserved arginine residue at position 179 was changed into glycine (R179G; ). This mutant displayed reduced dephosphorylation ability towards PPARγ (Figure 3A, lanes 5 and 6). Remarkably, co-expression of either the wild-type or mutant PPM1B led to increased PPARγ levels (Figure 3A, lanes 5–8). In order to determine whether PPM1B can directly dephosphorylate PPARγ, in vitro assays were performed. FLAG–PPARγ was immunoprecipitated from transiently transfected HEK-293T cells and incubated with purified PPM1B. Phosphospecific antibodies were used to detect PPARγ dephosphorylation and, as is shown in Figure 3(B), lane 3, PPM1B functions as a direct PPARγ phosphatase. Taken together, these data indicate that PPM1B functions as a PPARγ phosphatase both in intact cells and in vitro.
PPM1B stimulates PPARγ activity
Next, we examined whether PPM1B affected PPARγ activity. For this the human osteosarcoma cell line U2OS was used, which lacks endogenous PPARγ expression and displays a robust transcriptional response upon introduction of PPARγ [22,23]. To investigate whether PPM1B predominantly affects the N-terminal AF-1 region or the C-terminal AF-2 region, the two AFs were fused individually to a heterologous DBD (Gal4DBD). Co-transfection of PPM1B isoform 1 clearly enhanced the ligand-independent activity of the Gal4DBD–AF1 protein, whereas the activity of AF-2, either in the presence or absence of ligand, was unaffected (Figure 4A). These experiments therefore indicate that PPM1B specifically targeted the AF-1 region (including Ser112), and not the AF-2 region (including Ser273).
PPM1B functions as a direct phosphatase for PPARγ, dephosphorylating Ser112 within the AF-1 region
PPM1B is a specific activator of PPARγ
Next, the effect of PPM1B on the transcriptional activity of the full-length PPARγ protein was assessed. As is shown in Figure 4(B), co-transfection of PPARγ together with PPM1B increased PPARγ activity on a synthetic 3xPPRE reporter approximately 2-fold, either in the absence or the presence of the synthetic ligand rosiglitazone. Importantly, the catalytically inactive R179G mutant of PPM1B failed to activate PPARγ. Furthermore, the phosphatase PPM1A, which is 70% identical with PPM1B, did not affect PPARγ-mediated reporter activation. Mutating the phosphorylation site at Ser112 into an alanine residue resulted in increased activity as described earlier . PPM1B was not able to further increase transcriptional activity of this S112A mutant, indicating that PPM1B specifically targets Ser112 in the AF-1 region. PPARγ-mediated activation of a more ‘natural’ promoter [AQP7 (aquaporin 7)] was also potentiated by PPM1B, but not by its catalytically inactive R179G mutant (Supplementary Figure S2 at http://www.biochemj.org/bj/451/bj4510045add.htm). From these findings we conclude that the phosphatase PPM1B can specifically modulate PPARγ transcriptional activity, most likely through dephosphorylation of Ser112 in the N-terminal AF-1 region.
PP1MB knockdown selectively impairs endogenous PPARγ target gene expression
Having established that PPM1B is able to stimulate PPARγ-mediated reported gene activity (Figure 4), we wished to address the role of this phosphatase in regulating endogenous PPARγ target genes. For this siRNA-mediated knockdown experiments were performed. Two independent oligonucleotides were used, with oligonucleotide #6 being the most efficient, whereas oligonucleotide #5 only marginally reduced PPM1B expression at the protein and mRNA levels (Figure 5A and 5C). Knockdown of PPM1B increased PPARγ phosphorylation at Ser112 (Figure 5B), in line with a direct role for PPM1B in regulating the phosphorylation status of PPARγ (Figure 3). When the expression of a number of PPARγ target genes was examined, knockdown of PPM1B expression with oligonucleotide #6 was found to blunt the expression of a number of PPARγ target genes including lpin1, lpl, adipoq and fabp4, but not Cd36 (Figure 5C). Introduction of oligonucleotide #5, which reduced PPM1B expression poorly, had no appreciable effect on PPARγ target gene expression (Figure 5C). In contrast, knockdown of PPARγ itself efficiently reduced the expression of all of the genes examined (Figure 5C). Importantly, the decrease in PPARγ target gene expression observed upon PPM1B knockdown was not caused by down-regulation of PPARγ protein or mRNA (Figures 5A and 5C). In agreement with a selective role for PPM1B in regulating PPARγ activity, knockdown of PPM1B did not significantly impair the differentiation of 3T3-L1 cells into adipocytes, as assessed by PPARγ protein expression and Oil-Red-O staining (Figures 5B and 5D), whereas knockdown of PPARγ itself efficiently blocked adipogenesis (Figure 5D). These findings indicate that the phosphatase PPM1B functions as a selective modulator of PPARγ activity in adipocytes.
PPM1B knockdown selectively impairs PPARγ target gene expression
In the present study we show that the phosphatase PPM1B is a novel PPARγ-interacting protein, mediating direct dephosphorylation of Ser112, and thereby stimulating its transcriptional activity. Knockdown of PPM1B reduced the expression of a number of PPARγ target genes (e.g. Lpin1, Lpl and Adipoq) while leaving others unaltered (e.g. Cd36). In agreement with this, PPM1B knockdown did not abolish adipogenesis. As such, the phosphatase PPM1B qualifies as a novel selective modulator of PPARγ activity. The finding that phosphorylation of Ser112 does not dictate, but rather fine-tunes, the transcriptional and adipogenic activity of PPARγ [4,32] supports this view.
Previously the disruption of PPM1B in mice was shown to lead to early pre-implantation lethality . Although PPM1B KO (knockout) ES (embryonic stem) cells were viable, the embryos died between the two- and eight-cell stage. Also knockdown of PPM1B in wild-type ES cells did not affect proliferation suggesting that PPM1B expression is specifically required during the early stages of embryogenesis and does not affect cell-cycle progression . So far, only a limited number of substrates for PPM1B have been identified, including the kinases IKKβ [IκB (inhibitor of nuclear factor κB) kinase]  and TAK1 (transforming growth factor-β-activated kinase 1) , the pro-apoptotic protein Bad , the cyclin-dependent kinases CDK2 and CDK6 , the transcription factor p53  and the CDK9 subunit of P-TEFb (positive transcription elongation factor b) . Interestingly, Iankova et al.  showed that CDK9-mediated phosphorylation of PPARγ on Ser112 results in enhanced transcriptional activity. PPM1B may therefore potentially affect PPARγ activity indirectly, by stimulating CDK9 activity. In general, dephosphorylation of Thr186 in the T-loop of CDK9 can result in either activation of this enzyme by enabling its release from an inhibitory complex, or subsequent inhibition of its activity as phosphorylated Thr186 is required for substrate binding [40,41]. As PPM1B can only efficiently dephosphorylate CDK9 in the absence of the inhibitory complex , PPM1B-mediated dephosphorylation is most likely to result in reduced CDK9 activity and ultimately reduced gene expression. As we found PPM1B to stimulate rather than inhibit reporter activity, it is most probable that PPM1B regulates PPARγ-mediated transcription directly, and not indirectly via CDK9.
Although not identified as a bona fide dephosphorylation substrate as yet, the transcription factor EKLF (erythroid Kruppel-like factor)/KLF1 (Kruppel-like factor 1) was recently reported to interact with PPM1B . Interestingly, both the wild-type and catalytically inactive PPM1B stabilized the EKLF protein, reminiscent of our findings with PPARγ (Figure 3A). Future experiments should clarify whether PPM1B can also stabilize other proteins, and unravel the molecular mechanism underlying this phenomenon.
Recently Hinds et al.  reported that the serine/threonine PP5, an enzyme mainly implicated in the response to stress and hormones , interacts with and dephosphorylates PPARγ. It should be noted that PP5 also interacts with and dephosphorylates the GR , which also plays an important role in adipocyte differentiation and function . Interestingly, PP5-KO MEFs (mouse embryonic fibroblasts) display a reduced capacity for adipocyte differentiation (lipid accumulation and lipogenesis, adipogenic genes), and introduction of PPARγ Ser112 mutant, but not the wild-type, rescued the adipogenesis defect in PP5-KO cells . Despite being both identified as phosphatases targeting Ser112 of PPARγ, PPM1B and PP5 seem to differ in their exact mode of action. First, we found that PPM1B interacts with PPARγ in the absence of ligand (Figure 2), whereas PP5–PPARγ interaction was induced by ligand (1h rosiglitazone treatment) and lost again at longer time points . Furthermore, in contrast with PPM1B which co-localized with PPARγ in the nuclei of 3T3-L1 adipocytes (Figure 2B), PP5 predominantly displayed perinuclear localization . Additional studies are therefore required to establish the relative importance of PP5 and PPM1B in regulating PPARγ phosphorylation and activity.
Although activation of PPARγ by strong agonists of the TZD class clearly inhibits insulin resistance, their use has been linked to adverse side effects such as undesired mass gain, fluid retention, peripheral oedema and potential increased risk of cardiac failure . Modulation of PPARγ PTMs could offer the possibility of more subtle therapeutic intervention and provide new ways of improving insulin sensitivity. A clear example of this is the recently identified phosphorylation site at Ser273 in PPARγ . CDK5-mediated phosphorylation of this residue is induced by various cytokines, the levels of which are commonly increased in obesity, and leads to dysregulation of a subset of genes whose expression is altered in obesity including the insulin-sensitizing adipokine, adiponectin. Interestingly, Ser273 phosphorylation is blocked in vivo and in vitro by TZDs, but also by certain anti-diabetic drugs that are weak PPARγ agonists or non-agonists [10,12]. These findings indicate that future PPARγ-based anti-diabetic drugs should be selected on the basis of inhibition of Ser273 phosphorylation rather than transcriptional agonism.
Alternatively, PPARγ-based anti-diabetic drugs could be based on Ser112 phosphorylation. Indeed, in vivo evidence suggests that inhibition of Ser112 phosphorylation can contribute to metabolic health: PPARγ S112A-knockin mice display unaltered fat content and mass compared with the wild-type animals, but are protected against diet-induced insulin resistance . Activation of the PPARγ Ser112 phosphatases PPM1B and PP5 might therefore potentially improve metabolic health. Although not much is known about upstream signalling cascades, the catalytic activity of both PPM1B and PP5 can be activated by unsaturated fatty acids [47–49]. More research is needed to explore the potential of modulators of PPARγ PTMs, like the phosphatases PPM1B and PP5 as future targets for the development of insulin-sensitizing drugs.
adiponectin, C1Q and collagen domain-containing
erythroid Kruppel-like factor
fatty acid-binding protein 4
nuclear receptor corepressor
proximity ligation assay
peroxisome-proliferator-activated receptor γ
small interfering RNA
general transcription factor IIB
Ismayil Tasdelen, Olivier van Beekum, Olena Gorbenko, Veerle Fleskens, Niels van den Broek, Arjen Koppen and Nicole Hamers performed and analysed experiments; Ismayil Tasdelen, Olivier van Beekum, Olena Gorbenko, Veerle Fleskens, Niels van den Broek, Arjen Koppen, Paul Coffer, Arjan Brenkman and Eric Kalkhoven conceived and designed experiments; Ruud Berger, Paul Coffer, Arjan Brenkman and Eric Kalkhoven supervised all experiments; Ismayil Tasdelen, Olivier van Beekum and Eric Kalkhoven wrote the paper; and all authors agreed on the final paper.
We thank Dr J. Bakema, Dr H.Th.M. Timmers and Dr Y. Gao for helpful discussions and for critical reading of the paper prior to submission.
This work was supported, in part, by the European Commission Transfog Consortium [grant number LSHC-CT-2004-503438 (to A.B.B and N.J.F. vdB)].
These authors contributed equally to this work.
Current address: Department of Medical Biophysics, University of Toronto, Toronto, Ontario, Canada M5G 2M9.