The Optimal Corepressor Function of Nuclear Receptor Corepressor (NCoR) for Peroxisome Proliferator-activated Receptor γ Requires G Protein Pathway Suppressor 2*

Background: PPARγ tends to adopt an active conformation that cannot be directly bound by NCoR. Results: GPS2 binds to active PPARγ, facilitates its repression by NCoR, and is required for the optimal NCoR corepressor function for PPARγ. Conclusion: GPS2 mediates a novel NCoR repression pathway targeting active PPARγ. Significance: The GPS2-dependent pathway provides new insights into how NCoR regulates PPARγ function in vivo. Repression of peroxisome proliferator-activated receptor γ (PPARγ)-dependent transcription by the nuclear receptor corepressor (NCoR) is important for homeostatic expression of PPARγ target genes in vivo. The current model states that NCoR-mediated repression requires its direct interaction with PPARγ in the repressive conformation. Previous studies, however, have shown that DNA-bound PPARγ is incompatible with a direct, high-affinity association with NCoR because of the inherent ability of PPARγ to adopt the active conformation. Here we show that NCoR acquires the ability to repress active PPARγ-mediated transcription via G protein pathway suppressor 2 (GPS2), a component of the NCoR corepressor complex. Unlike NCoR, GPS2 can recognize and bind the active state of PPARγ. In GPS2-deficient mouse embryonic fibroblast cells, loss of GPS2 markedly reduces the corepressor function of NCoR for PPARγ, leading to constitutive activation of PPARγ target genes and spontaneous adipogenesis of the cells. GPS2, however, is dispensable for repression mediated by unliganded thyroid hormone receptor α or a PPARγ mutant unable to adopt the active conformation. This study shows that GPS2, although dispensable for the intrinsic repression function of NCoR, can mediate a novel corepressor repression pathway that allows NCoR to directly repress active PPARγ-mediated transcription, which is important for the optimal corepressor function of NCoR for PPARγ. Interestingly, GPS2-dependent repression specifically targets PPARγ but not PPARα or PPARδ. Therefore, GPS2 may serve as a unique target to manipulate PPARγ signaling in diseases.

Nuclear receptors (NRs), 3 which comprise the largest superfamily of ligand-inducible transcription factors, play important roles in homeostasis, metabolism, and development. NRs are receptors of various natural and synthetic lipophilic small molecules that can freely enter cells and bind to NRs via their ligand-binding domains (LBDs). Ligand binding induces a conformational change of NRs and alters their ability to recruit corepressors (CoRs) and coactivators (CoAs), leading to transcriptional repression or activation of target genes (1)(2)(3)(4). On the basis of sequence homology, NRs can be classified into different categories. One category includes thyroid hormone receptors (TR␣ and TR␤) and peroxisome proliferator-activated receptors (PPAR␣, ␥, and ␦), which share the ability to form heterodimers with the retinoid X receptor. TRs are receptors for thyroid hormone (T3) and play important roles in development and metabolism. PPARs bind to fatty acids and various other ligands. Among the three PPARs, PPAR␣ and PPAR␥ have been studied extensively. Although both PPAR␣ and PPAR␥ regulate lipid metabolism and have anti-inflammation functions, PPAR␣ is mainly involved in lipid utilization, whereas PPAR␥ is associated with lipid storage, adipogenesis, and insulin signaling. In addition, PPAR␥ serves as the pharmacological target of thiazolidinedione (TZD) antidiabetic drugs.
NRs can adopt two conformational states. Whereas the repressive state binds to CoRs, the active state binds to CoAs. Although it is generally true that unliganded NRs (apo-NRs) exist in the repressive conformation and liganded NRs exist in the active conformation (5), the ligand dependency for the active conformation may vary depending on the specific NRs. For example, the ability of TR␣ to adopt the active state is strictly dependent on binding to T3. Accordingly, apo-TR␣ is a constitutive repressor. On the other hand, structural studies have shown that apo-PPAR␥ can adopt the same active conformation as the agonist-bound PPAR␥-CoA complex (6,7). The intrinsic ability of PPAR␥ to assume the active state is also supported by its ability to drive activation in the absence of ligands and the reported conformational flexibility of the activation function 2 (AF2) domain (7)(8)(9).
The abilities of NRs to adopt the mutually exclusive active and repressive states are governed by the conformation of AF2, a conserved amphipathic sequence located at the C terminus of NRs. Earlier studies have mapped AF2 as a region required for the ability of NRs to activate the transcription of target genes (10). In the active state of NRs, AF2 adopts a conformation that allows it to directly contact CoAs. AF2 has a distinct conformation in the repressive NR state, which allows NRs to bind to CoRs but not CoAs. It has been shown that DNA-bound apo-PPAR␥/retinoid X receptor heterodimer cannot directly recruit NCoR or SMRT (11)(12)(13)(14). In vivo, the binding of CoAs may further reduce the ability of PPAR␥ to recruit CoRs. Underscoring the importance of AF2 in regulating PPAR␥ interactions with CoRs and CoAs, deleting AF2 or mutating its residues involved in CoA interactions allows PPAR␥ to recruit CoRs and to function as a repressor like unliganded TR␣ (11,12,15).
Despite the inability of DNA-bound PPAR␥ to mediate a strong interaction with CoRNR boxes (11,12), functional studies have shown that CoRs, in particular, NCoR, can repress PPAR␥-dependent transcription in vivo (33)(34)(35)(36). In this study, we generated GPS2 KO mice. These mice are embryonically lethal. Using GPS2 KO mouse-derived embryonic fibroblast cells (MEFs), we further studied the role of GPS2 in NCoR-and NR-mediated repression. Our results reveal a novel GPS2-dependent mechanism for PPAR␥ repression by NCoR. We show that this repression pathway, which targets the active conformation of PPAR␥, plays an important role in maintaining NCoR-mediated repression of PPAR␥. Loss of GPS2 is sufficient to cause constitutive activation of endogenous PPAR␥/ NCoR target genes in vivo and to predispose cells to adipogenesis in the absence of ectopic PPAR␥. Our results extend the view that NCoR is only capable of regulating NRs in its repressive state and show that the GPS2-enabled strategy to repress active PPAR␥ is an important gatekeeper mechanism to ensure proper repression of PPAR␥ target genes, consistent with the susceptibility of PPAR␥ to adopt the active conformation. Our results also show that the GPS2-mediated regulation is receptor type-, isoform-, and conformation-specific. Loss of GPS2 does not affect the repression mediated by apo-TR␣ or by an AF2deleted PPAR␥ mutant. Neither does it activate PPAR␣ or PPAR␦. In addition, given the early lethality of GPS2 KO mice, our work also adds GPS2 to the list of CoR-HDAC3 complex subunits required for embryonic development in mice.

EXPERIMENTAL PROCEDURES
Generation of Global GPS2 KO Mice and MEFs-The targeting vector contained a phosphoglycerate kinase (PGK) promoter-driven neomycin cassette (neo), along with the left and right arms amplified from genomic DNA of 129Sv-derived mouse cells by high-fidelity PCR (Fig. 1A). The left arm contained the GPS2 upstream region, the GPS2 promoter, part of exon 1 lacking the ATG and 3Ј section, and a GFP cassette. The right arm contained GPS2 exons 3-10 and 3Ј downstream sequences. DNA sequencing confirmed that no mutation existed in either arm. The targeting vector was electroporated into mouse ES cells to generate GPS2 ϩ/Ϫ ES cell clones via homologous recombination. Upon confirmation by Southern blot analysis (Fig. 1B), two independent ES cell clones were injected into blastocysts to generate chimeric mice from which two independent germ line-transmitted mouse lines were obtained. Both lines showed identical phenotypes.
To generate MEFs, mouse embryos were isolated under a microscope. After washing once with PBS, the embryos were minced by pipetting up and down 5 times in 50 l of trypsin in 24-well plates. The embryos were incubated at 37°C for 30 min and then resuspended in 500 l of complete DMEM (see below) by pipetting up and down for 20 times using a 1-ml pipette. The cells were passaged every 3 days until they became immortalized (ϳ3 months).
Adipogenesis-Self-immortalized MEFs were maintained in complete DMEM (DMEM containing 10% FBS) and passaged regularly to prevent overconfluency. For adipogenesis, the cells were allowed to grow to confluency. Two days later (defined as day 0), postconfluent MEFs were treated with rosiglitazone or vehicle. Oil Red O staining was performed on day 14 (37).
GST Pulldown Assays-GST pulldown assays were performed as described previously (38). Briefly, GST fusion proteins were expressed in bacteria and coupled to glutathione-Sepharose beads (GE Healthcare Life Sciences). PPAR␥ and its mutants were translated and labeled with 35 S in vitro using the TNT coupled reticulocyte lysate system (Promega). A BC200/ 0.1% Nonidet P-40 buffer containing rosiglitazone (2 M, Cayman Chemical) or vehicle was used for incubation and washing. CoRNR box peptide (PASNLGLEDIIRKALMGSFD) was dissolved in DMSO (16).
Coimmunoprecipitation Assays-Coimmunoprecipitation assays were performed as described previously (38). 293T cells were transfected with the desired plasmids using Turbofect in vitro transfection reagent. Rosiglitazone (1 M) was added to the transfected cells the next morning. On the third morning, the cells were lysed in lysis buffer (20 mM Tris (pH 7.9), 1 mM EDTA, 20% glycerol, 1 M rosiglitazone, 1ϫ protease inhibitor cocktail (Roche), 180 mM NaCl, and 0.5% Nonidet P-40). Cell extracts were incubated with anti-FLAG M2 affinity gel (Sigma) at 4°C for 3 h, followed by extensive washing with lysis buffer. The same lysis buffer was used for GST-GPS2 and derivatives to pull down endogenous PPAR␥ from MEFs treated with rosiglitazone (1 M) for 5 h before lysis of the cells.
Plasmids, Chemicals, and Antibodies for Western Blot Analysis-Mammalian and in vitro expression vectors for PPAR␥, PPAR␣, PPAR␦, NCoR, GPS2, and their derivatives have been described previously (9,22,40) or were generated by PCR and cloning techniques. The SMRT construct was kindly provided by Dr. Mitchell Lazar (University of Pennsylvania). PPAR␥ response element-and Gal4-UAS-driven reporters were made by inserting the PPAR␥ and Gal4 response elements into the polylinker of the pGL2-SV40 plasmid (Promega). Polyclonal rabbit anti-GPS2 antiserum was raised against the C-terminal 300 -327 peptide. Anti-Gal4 antibody was obtained from Santa Cruz Biotechnology (catalog no. sc-577). Anti-actin antibody was obtained from Millipore (catalog no. MAB1501). Anti-FLAG antibody was obtained from Sigma. Anti-NCoR antibody was obtained from Thermo Scientific. Anti-PPAR␥ antibody was obtained from Santa Cruz Biotechnology (catalog no. sc-7196).

Loss of GPS2 Does Not Reduce TR␣-or RD1-mediated
Repression-To understand the biological function of GPS2, we generated whole-body GPS2 KO mice (Fig. 1). The knockout removed part of exon 1 and the entire exon 2, which encode the ATG start codon and an N-terminal NCoR-interacting region of GPS2. GPS2 KO mice died prenatally around embryonic day 10 ( Fig. 1, C and D). A similar embryonic lethality has been reported for the knockout mice of NCoR, SMRT, and HDAC3 (41)(42)(43)(44). These results are consistent with an important role of GPS2 as a CoR complex subunit in development.
To study GPS2 function at the cellular level, MEFs from sibling GPS2 KO and WT embryos were isolated and immortalized by continuous passage. Western blot analysis confirmed the lack of expression of GPS2 in KO cells (Fig. 2B, inset). Next we performed Gal4-based reporter assays to test whether GPS2 plays a role in TR␣and NCoR-RD1-dependent transcriptional repression. Both Gal4-TR␣ and Gal4-RD1 strongly repressed basal transcription in KO cells (Fig. 2B, black columns). The potency of repression was not reduced in KO cells compared with WT cells (Fig. 2B). Because Gal4-TR␣ did not show a higher level of T3-induced activation in KO cells (Fig. 2B), the strong repression observed in KO cells was probably not caused by higher protein expression in these cells.
PPAR␥-dependent Transcription Is Derepressed in GPS2 KO Cells-Because NCoR can repress PPAR␥-dependent transcription and because GPS2 binds to NCoR, we asked whether GPS2 plays a role in repressing PPAR␥-dependent transcrip-tion. Compared with WT cells, the activity of PPAR␥ observed in KO cells was increased significantly both in the absence and presence of the exogenous ligand rosiglitazone (Fig. 2C). The increase was more dramatic in the absence (3-fold) than in the presence (2-fold) of rosiglitazone, possibly reflecting reduced binding of corepressors in the presence of rosiglitazone.
We excluded the possibility that the increased activity of PPAR␥ observed in KO cells was due to its differential binding to ligands in these cells (Fig. 2D). PPAR␥Q286P is a PPAR␥ mutant defective in binding to ligands (45,46). Consistent with its inability to bind to ligands, PPAR␥Q286P failed to activate transcription in WT cells both in the absence and presence of rosiglitazone (Fig. 2D, lanes 2 and 4, white columns). In KO cells, WT PPAR␥ showed increased activities as observed earlier (Fig. 2D, lanes 1 and 3). Unlike what was observed in WT cells, PPAR␥Q286P was still able to activate transcription in KO cells, and the fold increase of its activities in KO versus WT cells was similar to that observed with WT PPAR␥ in the absence of rosiglitazone (Fig. 2D, lanes 2 and 4). These results suggest that GPS2 plays a role in suppressing the constitutive activity of PPAR␥ that may arise from its ability to adopt the active conformation.
GPS2-dependent Repression Targets the Active Conformation of PPAR␥-We next sought to define the conformational state(s) of PPAR␥ susceptible to GPS2-dependent regulation. Because PPAR␥ is under equilibrium between repressive and active conformations, loss of GPS2 could possibly increase the activity of PPAR␥ as a result of derepression of the repressive PPAR␥, shifting from the repressive to the active conformation of PPAR␥, or failure to repress active PPAR␥. Because deleting AF2 will block PPAR␥ from adopting the active conformation, we first asked whether the ability of PPAR␥⌬AF2 to repress transcription was affected in GPS2 KO cells. Deleting AF2 allowed PPAR␥⌬AF2 to repress basal transcription in WT cells, as expected (Fig. 2E, white columns, lanes 3 and 4 versus lanes 1 and 2). Compared with WT cells, no significant reduction in the ability of PPAR␥⌬AF2 to repress transcription was observed in KO cells (Fig. 2E, black columns, lanes 3 and 4 versus lanes 1 and 2). This result is consistent with the lack of effect of GPS2 KO on TR␣-mediated repression, as shown earlier, and indicates that the classical CoRNR box-dependent repression pathway was intact in GPS2 KO cells.
We next asked whether loss of GPS2 would tip the balance toward the active PPAR␥ conformation. This was tested by mammalian two-hybrid assays to measure the in vivo interaction between PPAR␥ and an NCoR C-terminal fragment containing all three CoRNR boxes. PPAR␥⌬AF2, which showed comparable repression in WT and KO cells, was used as a control to set the maximal level (100%) of interaction in both cell types (Fig. 2F). As expected, PPAR␥⌬AF2 showed a stronger interaction than WT PPAR␥ in both cell types (Fig. 2F, white versus black columns). Compared with GPS2-WT cells, the GPS2 KO cells did not show a reduced interaction between WT PPAR␥ and the NCoR CoRNR boxes (Fig. 2F, cf. white columns). Together, these results showed that loss of GPS2 did not affect either the ability of the repressive PPAR␥ to repress transcription or the relative abundance of the repressive PPAR␥ that can directly recruit NCoR, therefore providing support for the idea that GPS2 facilitates the repression of the active form of PPAR␥ and that loss of its function in this context accounts for the increased PPAR␥ activity in GPS2 KO cells.
We also examined the effects of GPS2 loss on other PPAR isoforms. As shown in Fig. 2G, loss of GPS2 specifically increased PPAR␥-dependent transcription and may slightly repress PPAR␣and PPAR␦-dependent transcription. This reveals that GPS2-mediated regulation is PPAR isoform-specific.
GPS2 Is Required for NCoR to Repress Active PPAR␥-Ectopic expression of GPS2 reduced PPAR␥ activity in GPS2 KO but not WT cells (Fig. 3A). Interestingly, a weak stimulatory effect was observed in WT cells (see also Fig. 3B). These results indicate that the ability of GPS2 to repress PPAR␥ should not result from an autonomous function of GPS2. Because a majority of ectopic GPS2 may exist in the free form, we asked whether increasing its complex formation with NCoR would allow GPS2 to manifest its inhibitory function on PPAR␥ in WT cells. Indeed, although GPS2 alone failed to repress PPAR␥, as shown above, it enhanced the ability of ectopic NCoR to repress PPAR␥ (Fig. 3B, GPS2-WT). A possible explanation for the weak stimulatory effect of GPS2 alone in WT cells may be that the free GPS2 can exert a dominant negative activity against the endogenous GPS2-NCoR complex.
We also confirmed the cooperative function of GPS2 and NCoR in repressing PPAR␥ in GPS2 KO cells. In these cells, although both GPS2 and NCoR alone were able to modestly repress PPAR␥-dependent transcription, cotransfection of GPS2 and NCoR produced a stronger repression comparable with that observed in cotransfected WT cells (Fig. 3B). Given that TR␣-mediated repression was not affected by GPS2 KO, we next directly compared the ability of NCoR to repress PPAR␥ and TR␣ in GPS2 KO and WT cells. As expected, in both cell types, TR␣ showed similar levels of repression, and NCoR similarly potentiated TR␣-dependent repression in a dose-dependent manner (Fig. 3C). In contrast, and consistent with a require-ment of GPS2 for the optimal corepressor function of NCoR for PPAR␥, a dramatic reduction in the ability of NCoR to mediate dose-dependent repression of PPAR␥ was observed in KO cells (Fig. 3D, right). This occurred despite the high level of PPAR␥ activity in KO cells (Fig. 3D, left), further showing that GPS2 is an important rate-limiting factor for NCoR-mediated PPAR␥ repression. In KO cells, NCoR was still able to manifest a weak repression on PPAR␥ (Fig. 3, B and D). Because TR␣-and PPAR␥⌬AF2mediated repression was unaffected in KO cells, we hypothesized that the residual corepressor function of NCoR may reflect its ability to act on a subset of PPAR␥ in the repressive conformation. To test this, cells were treated with rosiglitazone (1 M) to saturate its binding to PPAR␥. Confirming the hypothesis, rosiglitazone completely abolished the ability of NCoR to repress PPAR␥ in KO cells (Fig. 3E, lanes 2 and 4). In WT cells, rosiglitazone reduced but did not abolish NCoR-mediated repression (Fig. 3E, lanes 1 and 3). These results demonstrate that NCoR is capable of repressing active PPAR␥-mediated transcription via a GPS2-dependent mechanism.
Given the similarity between NCoR and SMRT, we also examined the effect of GPS2 KO on the ability of SMRT to repress PPAR␥. SMRT repressed PPAR␥-dependent transcription, as expected (Fig. 3F). The repression was similarly reduced in GPS2-KO cells, as observed with NCoR. These results suggest that the GPS2-dependent mechanism is conserved between NCoR and SMRT.
GPS2, but Not NCoR, Is Able to Bind PPAR␥ in the Liganded Conformation-Because CoRNR boxes cannot bind PPAR␥ in the active conformation, we hypothesized that GPS2 directly binds to the active PPAR␥, thereby allowing its repression by NCoR. Consistent with this idea, GST pulldown assays revealed distinct abilities of NCoR and GPS2 to recognize the repressive and active conformations of PPAR␥ (Fig. 4A). Compared with GPS2, NCoR bound much more strongly to the repressive conformation of PPAR␥ (apo-PPAR␥ and PPAR␥⌬AF2 with or without rosiglitazone). The addition of rosiglitazone essentially  abolished NCoR interaction with PPAR␥. Rosiglitazone, however, did not reduce, and may slightly increase, the binding of PPAR␥ to GPS2 (Fig. 4A, densitometry data not shown). Unlike the TIF2 NR box, whose binding to PPAR␥ is strictly ligand-dependent (Fig. 4A), ligand binding is not required for GPS2 to interact with PPAR␥. Because both GPS2 and NCoR were able to bind apo-PPAR␥, we asked whether their interactions were mutually exclusive. Previous studies have mapped NR residues in the conserved hydrophobic cavity that directly contact CoRNR box and NR box motifs (16 -19). In PPAR␥, these residues include Thr-325, Lys-329, Leu-339, and Val-343 (Fig. 4B). We confirmed that mutations of T325R-K329A and L339R-V343R abolished the PPAR␥/NCoR interaction, as expected. These mutations, however, only slightly affected the PPAR␥-GPS2 interactions in the absence or presence of rosigli-tazone (Fig. 4C). To test whether the CoRNR box and GPS2 may simultaneously bind to PPAR␥, CoRNR box peptide was added to the reaction mixture. Consistent with the mutation results, although the peptide strongly inhibited the NCoR interaction with PPAR␥, it did not significantly affect the GPS2 interaction (Fig. 4D). Together, these results suggest that GPS2 targets a separate region (or surface) in PPAR␥ that is distinct from the classic docking site for CoRNR and NR motifs.
We next confirmed that GPS2, but not NCoR, was able to bind the active form of PPAR␥ in vivo. 293T cells were transfected with PPAR␥ (as a fusion to Gal4-DBD) together with FLAG-GPS2 or FLAG-NCoR. Cells were treated with rosiglitazone to ensure that all PPAR␥ was in the liganded, active conformation. PPAR␥ was detected in immunoprecipitates derived from FLAG-GPS2 but not in immunoprecipitates

F L A G -V e c t o r F L A G -G P S 2 F L A G -G P S 2 -N F L A G -V e c t o r F L A G -N C o R F L A G -G P S 2 F L A G -G P S 2 -C G S T -G P S 2 G S T I N P U T G S T -G P S 2 -N G S T -G P S 2 -C α-NCoR α-Gal4
GPS2-KO + Rosi. derived from FLAG-NCoR (Fig. 4E). No interaction was observed between FLAG-GPS2 and Gal4-DBD (data not shown). Analysis of truncated GPS2 derivatives showed that both the N-terminal region (amino acids 1-155) and C-terminal region (amino acids 105-327) of GPS2 were required for GPS2-PPAR␥ interaction (Fig. 4F). Although the N-terminal fragment of GPS2 was able to bind endogenous NCoR, as expected (22), it did not associate with PPAR␥. This confirmed that the endogenous NCoR, which was also present in FLAG-GPS2 immunoprecipitates, was not responsible for the observed PPAR␥ association with GPS2. We also confirmed the ability of GPS2 to interact with endogenous PPAR␥ in the presence of rosiglitazone. GST-GPS2, but not GST alone, was able to pull down significant levels of endogenous PPAR␥ from both KO and WT MEF cells (Fig. 4G). Consistent with the coimmunoprecipitation results, the interaction required both N-and C-terminal regions of GPS2. Interestingly, GPS2 was more capable of pulling down PPAR␥ from KO than from WT cells, consistent with the notion that a subset of PPAR␥ in WT cells was in complex with endogenous GPS2 and, therefore, unavailable for interaction with GST-GPS2.
Loss of GPS2 Activates Endogenous PPAR␥ and NCoR Target Genes-Loss of GPS2 also increased the ability of full-length PPAR␥ to drive transcription from a natural PPAR␥ response element-dependent reporter both in the absence and presence of rosiglitazone (Fig. 5A). Notably, in vector-transfected cells, rosiglitazone treatment was sufficient to increase the reporter activity in KO but not WT cells (Fig. 5A). This result raised the possibility that subconfluent GPS2-KO cells expressed functional endogenous PPAR␥ whose activity was also increased as a result of GPS2 depletion. RT-qPCR showed that, compared with WT cells, KO cells expressed higher levels of PPAR␥1. The total PPAR␥ level was also higher in KO cells despite a lower level of PPAR␥2 (Fig. 5B). These results are consistent with the notion that PPAR␥1 is the predominant PPAR␥ expressed in subconfluent cells. To explore whether loss of GPS2 activated endogenous PPAR␥ in KO cells, we asked whether PPAR␥ target genes were up-regulated in these cells. RNA-Seq studies were performed in KO, WT, and GPS2-transduced KO cells. 362 genes were up-regulated at least 2-fold in GPS2 KO cells but not in GPS2-re-expressed KO cells compared with their expression in WT cells (Fig. 5C). Gene ontology results showed that the 362 genes (also referred here as "GPS2 KO up-regulated" genes) were enriched with features that are characteristic of PPAR␥ target genes, such as lipid metabolism (gene ontology, 0006629, p ϭ 5.829E-5) and rosiglitazone response (C089730, p ϭ 2.55E-6) (see also Fig. 5F).
Loss of GPS2 Renders Immortalized MEFs Proadipogenic-To functionally demonstrate that loss of GPS2 increases PPAR␥ activity, we asked whether GPS2-KO MEFs were able to undergo spontaneous adipogenesis in the absence of ectopic PPAR␥. Consistent with the notion that self-immortalized MEFs require ectopic expression of PPAR␥ for adipogenesis (58), WT MEFs were refractory to adipogenesis either in the absence or presence of rosiglitazone (Fig. 6A). A small fraction of GPS2-KO MEFs, however, was able to differentiate into adipocytes spontaneously. The differentiation was greatly enhanced by rosiglitazone treatment, confirming that the process was PPAR␥-dependent. The ability of KO cells to undergo adipogenesis was due to the loss of GPS2 because re-expression of GPS2 completely abolished the adipogenic potential of GPS2-KO cells (Fig. 6B). A time-course experiment showed that ectopic GPS2 only slightly reduced the expression of PPAR␥1 and the initial level of total PPAR␥ but was able to completely prevent the induction of PPAR␥2, aP2, and adiponectin (Fig. 6C). While the slight reduction of PPAR␥1 level by ectopic GPS2 in KO cells was consistent with the difference between WT and KO cells we observed previously, the more dramatic difference observed between KO and WT cells may be related to the use of different cell lines or reflect a long-term cell culture effect on the cells. Nevertheless, the "all-or-none" changes in differentiation potential and the expression of adipocyte-specific PPAR␥ target genes are consistent with the conclusion that GPS2 regulates adipogenesis by regulating the transcriptional activity of PPAR␥. Loss of GPS2 renders MEFs proadipogenic because of impaired repression control of PPAR␥.

DISCUSSION
Since the discovery of GPS2 as a G protein pathway suppressor, GPS2 has emerged as a multifunctional protein. GPS2 can both act as a corepressor and act as a coactivator for various transcription factors, including NRs (25, 59 -65), consistent with the presence of distinct repression and activation domains in GPS2 (22). This study focused on the role of GPS2 as an NCoR complex subunit in repression, which is often the first step in signal-dependent transcription activation pathways.
To understand the physiological function of GPS2, we established whole-body GPS2 knockout mice. The mice died during embryonic development between E9.5 and E10.5. A similar lethality has also been reported for NCoR, SMRT, and HDAC3 knockout mice, which die at E15.5, E16.5, and E9.5, respectively (41)(42)(43). GPS2 ϩ/Ϫ mice were fertile and indistinguishable in appearance from their WT counterparts. It is possible that the level of GPS2 in cells may be in excess relative to that of CoRs. Although the precise reason for the death of GPS2 knockout mice remains to be determined, these results are consistent with an essential role of GPS2 as a CoR complex subunit in mouse embryonic development. It is tempting to speculate that the earlier death of GPS2 and HDAC3 KO mice, compared with NCoR and SMRT KO mice, could reflect the fact that GPS2 and HDAC3 are unique subunits of the NCoR/SMRT corepressor complexes, whereas NCoR and SMRT may have redundant functions that allow the embryos to survive longer.
Previous studies have shown that GPS2, TBL1/TBLR1, and RD1 interact with each other to form a stable heterotrimeric complex (22,32). These results raise the possibility that GPS2 may play a role in complex assembly, which would suggest that GPS2 contributes to RD1-and NCoR-mediated repression. Our results, however, argue against this possibility. Using GPS2-deficient MEFs, we demonstrated that GPS2 was dispen-   RNA-Seq analysis was performed in GPS2 WT, GPS2 KO, and GPS2-re-expressed KO MEF cells. The mapped reads were analyzed for differential gene expression using DESeq and GeneSpring NGS software. Compared with GPS2-WT cells, 362 genes showed at least 2-fold higher expression in GPS2 KO but not in GPS2-re-expressed KO cells. The Venn diagram identified 34 overlapping genes between the 362 genes and genes containing overlapping binding sites of PPAR␥ and NCoR, which were determined by analyzing the ChIP-Seq data sets from mouse macrophages that express PPAR␥ (47,48) using Homer (75). sable for RD1-mediated repression. Further supporting the notion that GPS2 is not required for the intrinsic repression function of NCoR, loss of GPS2 did not affect repression mediated by unliganded TR␣ or TBL1 ( Fig. 2B and data not shown). Neither did it affect the corepressor function of NCoR for TR␣ (Fig. 3C). These results point to fundamental differences between GPS2 and other stoichiometric NCoR-interacting proteins, namely, HDAC3 and TBL1/TBLR1, and suggest that GPS2 is primarily involved in targeting the NCoR complex to transcription factors. GPS2 plays a critical role in NCoR-mediated repression of PPAR␥. Our effort to dissect the conformation requirement of PPAR␥ for its regulation by GPS2 revealed that GPS2 functions at a step after PPAR␥ attaining the active conformation (Fig. 7).
Our results support the model in which, whereas the CoRNR box allows NCoR to target the repressive state of PPAR␥, GPS2 allows NCoR to target the active state of PPAR␥. In the absence of the ligand, although GPS2 and the CoRNR box may independently bind to the repressive conformation of PPAR␥ (and, possibly, apo-TR␣), CoRNR box binding should be sufficient to recruit NCoR to mediate repression. This explains why loss of GPS2 did not affect the repression by PPAR␥⌬AF2 and apo-TR␣. GPS2 may be the only component in the corepressor complex that can recognize and bind the active PPAR␥. Given that a subset of unliganded PPAR␥ has the active conformation, our model also explains why loss of GPS2 increased PPAR␥ activities both in the absence and presence of the ligand. Because we have shown that GPS2 and the CoRNR box target different regions of PPAR␥, it will be interesting to explore whether GPS2 and the CoRNR box may have cooperative functions in driving the active PPAR␥ toward a stable inactive complex with NCoR (shown by the reversed arrow line in Fig. 7). Functionally, the GPS2-endowed ability of NCoR to repress active PPAR␥ is important for cells to maintain proper control of PPAR␥ target genes, as evidenced by our finding that loss of GPS2 in MEFs was sufficient to cause activation of endogenous PPAR␥ target genes and to drive adipogenesis of MEFs. Therefore, combined functions of the CoRNR boxes and GPS2 confer on NCoR the abilities to repress diverse NRs, including TR␣ and PPAR␥.
In the presence of agonists such as rosiglitazone, the binding of agonists to PPAR␥ should abolish the CoRNR box-dependent interaction with PPAR␥. Subsequently, the increased association of CoAs with agonist-bound PPAR␥ further reduces GPS2-dependent repression, leading to commitment of activation. Nevertheless, in the presence of agonists, GPS2 could still impose an inhibition on PPAR␥, as evidenced by the increased PPAR␥ activity in GPS2-KO cells versus WT cells (Figs. 2, C and D, and 3E). These results suggest that GPS2 is not only important for maintaining the repressed state of apo-PPAR␥ but also plays a role in delimiting the maximal ligand response of PPAR␥, which may be important for its biological function.
The GPS2-dependent repression pathway explains why NCoR is still capable of repressing PPAR␥-dependent transcription despite the weak interaction between its CoRNR boxes and PPAR␥ on DNA (11)(12)(13). A role of GPS2 for NCoR to repress active PPAR␥ also provides new mechanistic insights into previous studies showing that NCoR is able to repress PPAR␥ target genes in the presence of PPAR␥ agonists. For example, in 3T3-L1 cells cultured in the presence of thiazolidinediones, NCoR depletion dramatically enhances the expression of PPAR␥ target genes and PPAR␥-dependent adipogenesis (33). A recent study also reported that adipocyte-specific knockout of NCoR increases PPAR␥ activity under high-fat diets (34), a condition that would increase the levels of endogenous PPAR␥ ligands.
Our finding that GPS2 represses PPAR␥ activity is in line with previous studies showing that reduced expression of GPS2 is associated with increased obesity in humans (66). It also explains why adipogenesis of immortalized MEFs requires overexpression of ectopic PPAR␥ (58). On the basis of our data, overexpressing PPAR␥ may be necessary for PPAR␥ to escape the interaction with and subsequent inhibition mediated by endogenous GPS2. In a related scenario, the ability of GPS2 to recognize and bind the active state of PPAR␥ also provides new insights into how GPS2 inhibits inflammation (60,66). One of the anti-inflammatory pathways involves trans-repression by ligand-associated PPAR␥, which inhibits inflammation by preventing the discharge of corepressor complexes from cytokineinducible genes. Although sumoylation of liganded PPAR␥ plays a role in its association with the corepressor complex (67), it is not known whether sumoylation is sufficient and how core-pressors discriminate different sumoylated proteins. An interaction between the GPS2 subunit of CoR complex and liganded PPAR␥ may provide the necessary specificity while ensuring the stability of the association between liganded PPAR␥ and the CoR complex.
Our results show that PPAR␥2 expression is strongly induced in post-confluent GPS2-KO cells. Interestingly, the level of PPAR␥2 in subconfluent KO cells, as judged by RT-qPCR, was lower than that in WT cells. It should be noted that PPAR␥2 expression is already very low in subconfluent cells compared with differentiated cells, consistent with the notion that its expression is adipocyte-specific. In non-differentiated cells, transcription from the PPAR␥2 promoter may be epigenetically silenced via H3K9 methylation (68). The further reduction of the PPAR␥2 level in KO cells may be related to the increased transcription of PPAR␥1 in these cells or reflect an independent "coactivator" function of GPS2 in promoting H3K9 demethylation, as reported recently (61,69). During differentiation, PPAR␥ is able to bind to its sites. This allows GPS2 to manifest its corepressor function to prevent the premature activation of PPAR␥2, explaining the lack of PPAR␥2 induction in WT cells.
In summary, this work identified, for the first time, a novel GPS2-dependent mechanism that allowed NCoR to directly repress active PPAR␥-mediated transcription and showed that this repression mechanism was important for the corepressor function of NCoR (and possibly SMRT) for PPAR␥. Interestingly, the GPS2-dependent repression appears to affect only PPAR␥ but not other PPAR isoforms. Therefore, a further understanding of how GPS2 binds and regulates PPAR␥ may lead to strategies to specifically target PPAR␥ signaling in various diseases, such as obesity and insulin resistance. In addition to PPAR␥, GPS2 also interacts with other NRs, such as liver X receptor (LXR), liver receptor homolog-1 (LRH-1), hepatocyte nuclear factor 4 (HNF4), and farnesoid X receptor (FXR) ( FIGURE 7. Role of GPS2 in NCoR-mediated repression of PPAR␥-dependent transcription. NCoR can directly regulate PPAR␥ not only in the repressive but also in the active conformation via CoRNR box-dependent and GPS2-dependent mechanisms, respectively. In the absence of agonists, a subset of PPAR␥ spontaneously adopts the active conformation that requires AF2, explaining why depleting AF2 insensitizes PPAR␥ to GPS2-dependent regulation. Agonists such as rosiglitazone increase the ability of PPAR␥ to assume the active state and to bind to CoAs. Unlike thought previously, repression occurs not only to the repressive state but also to the active state of PPAR␥. See text for details. tion with CoR complexes (64). Therefore, the GPS2-dependent regulation, as observed here, may be applicable to these NRs as well. GPS2 appears to belong to a unique class of coregulators different from the classic CoRs and CoAs that only bind to repressive and active conformations of NRs, respectively. Because ligands may not be required for GPS2 to bind NRs, it is possible that GPS2 may recognize a constitutive surface present in both repressive and active forms of NRs, including the constitutively active orphan receptors. We speculate that GPS2 may have evolved to play a general protective role against premature activation (hormone-independent) or hyperactivation (hormone-dependent) of NR target genes. A broad involvement of GPS2-NR interactions may also explain why the CoRNR box region of CoRs (70 -74), but not the entire CoR molecule, is dispensable for embryonic development in mice.