GR SUMOylation and formation of an SUMO-SMRT/NCoR1-HDAC3 repressing complex is mandatory for GC-induced IR nGRE-mediated transrepression

Significance

Glucocorticoids (GCs), acting through binding to the GC receptor (GR), are peripheral effectors of circadian and stress-related homeostatic functions fundamental for survival throughout vertebrate life span. They are widely used to combat inflammatory and allergic disorders, and their therapeutic effects have been mainly ascribed to their capacity to suppress the production of proinflammatory cytokines. The present study unveils, at the molecular level, the mechanisms that underlie the GC-induced GR direct transrepression function mediated by the evolutionary conserved inverted repeated negative response element. This knowledge paves the way to the elucidation of the functions of the GR at the submolecular levels and to the future educated design and screening of drugs, which could be devoid of undesirable debilitating effects on prolonged GC therapy.

Abstract

Unique among the nuclear receptor superfamily, the glucocorticoid (GC) receptor (GR) can exert three distinct transcriptional regulatory functions on binding of a single natural (cortisol in human and corticosterone in mice) and synthetic [e.g., dexamethasone (Dex)] hormone. The molecular mechanisms underlying GC-induced positive GC response element [(+)GRE]-mediated activation of transcription are partially understood. In contrast, these mechanisms remain elusive for GC-induced evolutionary conserved inverted repeated negative GC response element (IR nGRE)-mediated direct transrepression and for tethered indirect transrepression that is mediated by DNA-bound NF-κB/activator protein 1 (AP1)/STAT3 activators and instrumental in GC-induced anti-inflammatory activity. We demonstrate here that SUMOylation of lysine K293 (mouse K310) located within an evolutionary conserved sequence in the human GR N-terminal domain allows the formation of a GR-small ubiquitin-related modifiers (SUMOs)-NCoR1/SMRT-HDAC3 repressing complex mandatory for GC-induced IR nGRE-mediated direct repression in vitro, but does not affect transactivation. Importantly, these results were validated in vivo: in K310R mutant mice and in mice ablated selectively for nuclear receptor corepressor 1 (NCoR1)/silencing mediator for retinoid or thyroid-hormone receptors (SMRT) corepressors in skin keratinocytes, Dex-induced direct repression and the formation of repressing complexes on IR nGREs were impaired, whereas transactivation was unaffected. In mice selectively ablated for histone deacetylase 3 (HDAC3) in skin keratinocytes, GC-induced direct repression, but not bindings of GR and of corepressors NCoR1/SMRT, was abolished, indicating that HDAC3 is instrumental in IR nGRE-mediated repression. Moreover, we demonstrate that the binding of HDAC3 to IR nGREs in vivo is mediated through interaction with SMRT/NCoR1. We also show that the GR ligand binding domain (LBD) is not required for SMRT-mediated repression, which can be mediated by a LBD-truncated GR, whereas it is mandatory for NCoR1-mediated repression through an interaction with K579 in the LBD.

Glucocorticoids (GCs) hormones, the function of which is transduced by a single receptor, the GR, have pleiotropic effects on almost all aspects of physiology. Their anti-inflammatory and immunosuppressive properties were demonstrated more than 60 years ago (1, 2). Since then, and despite their multiple side effects, GCs have been increasingly used in the treatment of numerous inflammatory conditions, such as rheumatoid arthritis and allergic disorders. The GR regulates the expression of target genes either by transcriptional transactivation through binding to GC response elements (GREs) (3) or by transrepression (1, 2). To initiate “tethered” indirect transrepression, the GR is thought to physically interact with DNA-bound transcriptional activators (e.g., NF-κB, activator protein 1 (AP1), STAT3) and to repress their downstream target genes (4). Interestingly, it became widely accepted that most GC anti-inflammatory effects can be ascribed to tethered transrepression, whereas transactivation was responsible for many undesirable side effects (1). This led to searches for dissociated ligands that would preferentially induce tethered transrepression and be devoid of transactivation activity. Such a ligand was found to exhibit the expected dissociation profile in vitro, but its administration in vivo did not confirm this dissociation (1, 5). Our own discovery of a GC-induced direct transrepression activity mediated via direct GR binding to evolutionary conserved inverted repeated negative response elements (IR nGREs) indicated that this GC analog failed because it had kept the latter activity (6). Thus, a search for improved dissociated anti-inflammatory compounds should aim at finding GR agonists that would repress gene expression through tethered indirect repression while lacking IR nGRE-mediated transrepression and (+)GRE-mediated transactivation activities. However, although the molecular mechanisms involved in (+)GRE-mediated transactivation have been deciphered, those underlying tethered indirect transrepression and IR nGRE-mediated transrepression are still poorly understood, thus precluding an educated design and a differential screening of compounds that would selectively exert the anti-inflammatory activities of GCs. We report here an analysis of the molecular mechanisms involved in direct transrepression, which demonstrates at the molecular level the role played by SUMOylation of the GR in this transrepression.

Results

A Conserved Sequence (AA283-295) Within the GR N-Terminal Domain Is Required for IR nGRE-Mediated Transrepression, Whereas the Ligand Binding Domain Is Dispensable.

Human GR cDNAs for full-length (FL) GR, GR isoforms (2), and GR truncated in the N-terminal domain (NTD) were expressed in the pcDNA3 vector (Fig. 1A and Fig. S1A). Their transcriptional activities were determined in Cos-1 cells (low in GR content) cotransfected with pGL3-nGRE or pGL3-(+)GRE luciferase reporters (6) (Fig. S1 B and D–F). From GR FL to GR 280, the pGL3-nGRE activity decreased by ∼50% upon dexamethasone (Dex) treatment, whereas no repression was observed for GR 336 (the GRα-D3 isoform; Fig. 1A) and beyond (Fig. S1G). Of note, transcripts of the IR2 nGRE-containing stimulated by retinoic acid 13 (STRA13) gene in Cos-1 cells exhibited similar profiles of Dex-induced repression (Fig. S1H), and as expected (6), this repression was relieved by the GC antagonist RU486 (RU) (Fig. S1H). ChIP assays revealed the binding of GR, GR 90, GR 190, and GR 280, but not of GR 336 (the human GRα-D3 isoform), to the IR1 nGRE of the pGL3-nGRE reporter (Fig. S1I).

Fig. 1.

GR binding to an IR nGRE in vitro and in vivo requires a discrete sequence located in the NTD. (A) GR NTD and LBD deletion mutants. (B) Quantitative (q)RT-PCR for IR nGRE-containing genes in GRwt and GRα-D3 MEFs. Vehicle, Dex (0.5 µM), and RU (3 µM) treatments were for 6 h. (C) qPCR of ChIP assays performed with GRwt and GRα-D3 MEFs treated with vehicle, Dex (1 µM), and RU (6 µM) for 1 h, showing the association of GR, nuclear receptor corepressor 1, and silencing mediator for retinoid or thyroid-hormone receptors (SMRT) on IR nGREs in the promoter region of genes as indicated. (D) Schematic representation for pGL3-17mer Gal4 reporter, pSG5-Gal4, and pSG5-GR NTD-Gal4 plasmids. (E) Luciferase assays of Cos-1 cells transfected with pGL3-17mer Gal4 reporter and pSG5-Gal4 alone or fused to GR NTD or its mutants as indicated (see D). (F) As in C, but performed with Cos-1 cells transfected with pGL3-17mer Gal4 reporter and pSG5-Gal4 constructs (see D), showing the association of indicated proteins on the Gal4 RE sequence, using specific antibodies. (G) As in E, but transfected with pGL3-nGRE reporter and GR mutants. Vehicle or Dex (0.5 µM) treatment was for 6 h. (H and I) As in F but transfected with pGL3-nGRE reporter and GR mutants, showing the binding of GR, NCoR1, and SMRT on an IR nGRE. Cells were treated with vehicle or Dex (1 µM) for 1 h. (J) Alignment of the 283–295 AA sequence in human GR with homologous GR sequences in vertebrates (Left) and predicted SUMOylation sites in the human and mouse GR sequence (Right). Values are mean ± SEM. *P < 0.05, **P < 0.01.

Fig. S1.

GR binding to an IR nGRE in vitro requires a discrete sequence located in the NTD. (A) Immunoblots of WCE from Cos-1 cells transfected with GR and NTD mutants, NT, not transfected; arrow, nonspecific protein. (B) Schematic representation of pGL3 reporter. (C) Immunoblots of NE from Cos-1 cells transfected with GR or GR ABCD, treated with vehicle, Dex, or RU486 (0.5 µM) for 1 h. Sp1 factor was a control for nuclear localization. (D and G) Luciferase assays of Cos-1 cells transfected with pGL3 luciferase reporters and GR NTD mutants. Vehicle or Dex (0.5 µM) treatment was for 6 h. (E and H) qRT-PCR for SGK1 and STRA13 transcripts in transfected Cos-1 cells. Vehicle, Dex (0.5 µM) and RU (3 µM) treatments were for 6 h. (F and I) qPCR of ChIP assays performed with Cos-1 cells transfected with pGL3 reporters and GR NTD mutants, using GR antibody to show GR binding on a (+)GRE or an IR nGRE. Cells were treated with vehicle or Dex (1 µM) for 1 h. IgG control for IP was ∼0.02% of input. (J) Luciferase assays of Cos-1 cells transfected with pGL3-nGRE reporter, GR ABCD, or GR K579A and indicated siRNA. Vehicle or Dex (0.5 µM) treatment was for 6 h. (K and L) qPCR analyses of ChIP assays performed with Cos-1 cells transfected as in J, showing the binding of indicated proteins to IR1 nGRE, using specific antibodies. Cells were treated with vehicle or Dex (1 µM) for 1 h. Values are mean ± SEM. *P < 0.05.

To demonstrate that the mouse GR NTD was similarly mandatory for repression in vivo, we removed 352 AAs from the N terminus to generate a mutant expressing selectively the mouse GRα-D3 (GR 353) isoform similar to the human GR 336 (Fig. S2 and SI Materials and Methods). As 95% of GRα-D3 homozygote embryos died before or at birth, mouse embryonic fibroblasts (MEFs) were derived from 13-day embryos. In these MEFs, the GRwt, but not GRα-D3, repressed transcription from IR nGRE-containing genes (Fig. 1B), and interestingly, ChIP assays showed that GRwt, but not GRα-D3, was associated with SMRT and/or NCoR1 corepressors on IR nGREs of these genes (Fig. 1C; note the effect of RU486). Together, these in vitro and in vivo results demonstrated that the GR NTD region was instrumental in GR binding to an IR nGRE.

Fig. S2.

Generation of GR mutant mice selectively expressing GRα-D3 isoform. (A) Strategy of the insertion of LoxP sites into the first exon of murine GR genomic locus. (B) Schematic representation of conditional deletion of 352 AAs from the GR NTD. The primers indicated are used for GR transcript assessment. (C) Immunoblot analyses for GR expression using liver samples from E13 WT, GR^F/F, and GR^F/+ embryos. GAPDH was used as a loading control. (D) Immunoblot analyses for GR expression using WT (lane 3) and GRα-D3 (lane 2) mice liver samples. Lysates from Cos-1 cell transfected with either GR FL (lane 4) or GR 336 (lane 1) were used as controls. Arrows indicated the bands for GR WT, GRα-D3, and nonspecific protein. (E) qRT-PCR for transcripts of GR expression using liver samples from WT, GR^F/F and GR^F/+ embryos, or WT and GRα-D3 mice; primers as indicated in B. Values are mean ± SEM.

To determine whether the GR NTD could exert a repressing function on its own, we used our previous construct pSG5-GR(AB)-GAL (7) [hereafter called pSG5-GR(NTD)-Gal4], and made a luciferase reporter vector by replacing the IR1 nGRE of the pGL3-nGRE vector with the 17-mer Gal4 DNA response element (17mer Gal4 RE), thereby creating the pGL3-17mer Gal4 luciferase reporter (Fig. 1D). Transactivation by the yeast transactivator Gal4 was repressed when the GR NTD was fused to Gal4(1-147) (Fig. 1 D and E). ChIP assays revealed the binding of GR NTD and SMRT (but not of NCoR1) to 17-mer Gal4 RE upon transfection of the pSG5-GR(NTD)-Gal4(1-147) vector (Fig. 1F). These data indicated that the GR NTD on its own can exert a repressing function that involves the SMRT corepressor.

The nuclear ligand binding domain (LBD)-deleted GR (GR ABCD) (Fig. 1A and Fig. S1C) is a constitutive activator of transcription (8) (Fig. S3 C and D). Similarly, GR ABCD repressed the expression of the IR1 nGRE reporter in the absence of Dex (Fig. 1G), while ChIP assays revealed its IR1 nGRE binding (Fig. 1H). Interestingly, these assays showed that both NCoR1 and SMRT could bind together with the GR on an IR1 nGRE, whereas only SMRT was recruited by GR ABCD (Fig. 1I). As expected, RU addition did not affect the GR ABCD repressing activity and IR nGRE binding (Fig. 2C). Of note, the crystal structure of a GR LBD-NCoR1 complex has revealed a strong hydrogen bond interaction between NCoR1 and the GR K579 residue (9). However, mutation of K579 did not affect IR1 nGRE-mediated transrepression (Fig. 1G), nor binding of GR to an IR1 nGRE (Fig. 1H), whereas the binding of NCoR1, but not of SMRT, was markedly reduced (Fig. 1I), indicating that GR K579 is required for NCoR1 binding to the GR. That the interaction between NCoR1 and GR K579 could be instrumental in IR nGRE-mediated transrepression was suggested by transfecting the GR K579A mutant in Cos-1 cells in which SMRT expression was knocked down with siRNA (Fig. S1 J–L, and see below).

Fig. S3.

GR SUMOylation at K293 in the GR NTD is crucial for Dex-induced IR nGRE-mediated direct transrepression. (A and C) Luciferase assays of Cos-1 cells transfected with pGL3-nGRE or pGL3-(+)GRE reporters and GR or GR ABCD bearing mutations in SUMOylation sites, treated with vehicle or Dex (0.5 µM) for 6 h. (B) SUMOylation assays performed in Cos-1 cells transfected with GR and SUMO1, treated with or without 1 µM Dex for 1 h. WCEs were immunoprecipitated with a GR antibody and washed five times in RIPA buffer. Eluates were electrophoresed and immunoblotted against GR antibody. (D and E) qPCR analyses of ChIP assays performed with Cos-1 cells transfected with pGL3-(+)GRE reporter and GR mutants as indicated, using specific antibodies to show the bindings of GR, NCoR1, SMRT, and SUMOs on a (+)nGRE. Cells were treated with vehicle or Dex (1 µM) for 1 h. (F) qRT-PCR for transcripts from IR nGRE-containing genes TSLP and STRA13 in A549 cells transfected with or without DAXX expression vector, and treated with vehicle, 0.5 µM Dex, and 3 µM RU for 6 h. (G) As in C, but using Cos-1 cells transfected with GR ± DAXX, treated with vehicle, Dex (1 µM), and RU (6 µM) for 1 h, showing the binding of indicated proteins on the IR nGREs of the TSLP and STRA13 genes. Values are mean ± SEM.

Fig. 2.

GR SUMOylation at K293 in the GR NTD is crucial for Dex-induced IR nGRE-mediated direct transrepression through interaction between SUMOs and corepressors NCoR1 and SMRT. (A) qRT-PCR for STRA13 and serum and glucocorticoid-regulated kinase 1 (SGK1) transcripts in Cos-1 cells transfected with GR bearing mutations in SUMOylation sites, treated with vehicle, Dex (0.5 µM), and RU (3 µM) for 6 h. (B and C) qPCR of ChIP assays using Cos-1 cells transfected with pGL3-nGRE reporter and GR expression vectors, treated with vehicle, Dex (1 µM), or RU (1 or 6 µM for the cotreatment) for 1 h, showing the binding of GR, NCoR1, SMRT, SUMO1, and SUMO2/3 to an IR nGRE sequence. (D) As in B, but showing the association of indicated proteins on the IR1 nGRE of the TSLP gene. (E) As in D, but using WT mouse dorsal epidermis topically treated with vehicle, Dex (6 nmol/cm²) and RU (36 nmol/cm²) for 6 h. (F) As in B, but on the IR1 nGRE present in exon 6 of the GR gene. (G) As in A, but for IR nGRE and (+)GRE-containing genes from ear extracts of GRwt and GR K310R mutant mice, treated with vehicle or Dex (6 nmol/cm²) for 18 h. (H) As in E, but with dorsal epidermis from WT and GR K310R mutant mice topically treated with Dex, showing the association of GR, SUMO1, and SMRT on indicated DBSs. (I) As in D, but using Cos-1 cells transfected with GR, GR S226A/S404A, or GR 5SA, treated as indicated. (J) As in D, but using A549 cells treated as indicated; 25 µM JNK inhibitor II or/and 20 mM LiCl was added into the medium 30 min before Dex treatment. (K) As in I, but transfected with GR ABCD or its mutants. (L) Sequence alignment of DAXX SIM with putative human and mouse NCoR1 and SMRT SIM. (M) GST-pull down of ³⁵S-labeled C-terminal moieties of NCoR1 and SMRT proteins by GST or GST-SUMO1 protein. Input, 10% of ³⁵S-labeled proteins used for the binding assay. Values are mean ± SEM. *P < 0.05, **P < 0.01.

Deletions (Fig. 1A) to map the GR sequence involved in the NTD repressing function showed that GR 280-527 exhibited the same repressing activity and binding to IR1 nGRE as GR ABCD, while both were abolished in GR 336-527 (Fig. 1 G and H). As expected, SMRT, but not NCoR1, was bound to GR 280-527, and no corepressor was bound to GR 336-527 (Fig. 1I). Importantly, analysis of the evolutionary conservation of GR NTD revealed a vertebrate highly conserved sequence: FIELCTPGVIKQE (AA283-295) (Fig. 1J, Left), the deletion of which in GR ABCD totally abolished the IR1 nGRE-mediated transrepression (Fig. 1 G and H). Furthermore, this sequence was mandatory for the repressing function of the GR NTD, as its deletion in the GR (NTD)-Gal4(1-147) protein (Fig. 1D) fully reversed the GR NTD-induced repression (Fig. 1E), while SMRT binding to GR NTD was abolished (ChIP assays; Fig. 1F).

SUMOylation of Human GR at K293 Is Crucial for IR nGRE-Mediated Transrepression.

The reports of Pascual et al. (10) and Leuenberger et al. (11) showing that SUMOylation of the LBDs of peroxisome proliferator-activated receptor (PPAR)γ and PPARα were associated with transcriptional repression, prompted us to test whether the SUMOylation site, which was previously identified (12) in the NTD of human GR at position K293 within the AA283-295 conserved sequence (Fig. 1J), could be involved in Dex-induced IR nGRE-mediated repression. Interestingly, the K293R mutation, but not those of the additional GR SUMOylation sites at position K277 and K703 (12), abolished Dex-induced IR1 nGRE-mediated repression (Fig. 2A and Fig. S3A, Left). ChIP assays revealed that, upon Dex treatment of transfected Cos-1 cells, GR K293R and GR ABCD K293R mutations drastically reduced GR binding to the IR1 nGRE of the pGL3 reporter, and importantly, no repressing complex containing SMRT and NCoR1 could be assembled with these mutants (Fig. 2 B and C). Small ubiquitin-related modifier (SUMO)1 and SUMO2/3 were associated upon Dex treatment with GR, GR K277R, and GR K703R, but not with GR K293R, and, as expected, the association of SUMOs with GR ABCD was not ligand dependent (Fig. 2C). The same SUMO/corepressor repressing complex was associated on the IR1 nGRE of the Cos-1 cell endogenous thymic stromal lymphopoietin (TSLP) gene (Fig. 2D). Importantly, the formation of this GR complex was prevented by addition of RU486, which is known to promote GR translocation to the nucleus (Fig. S1C and Fig. 2 C and D), indicating that, on its own, SUMOylation of the GR is GC dependent. Furthermore, upon topical Dex treatment of the WT epidermis, the same repressing complex was found on the TSLP IR1 nGRE (Fig. 2E; note the inhibitory effect of RU486), thus establishing that SUMOs are associated with Dex-induced IR nGRE-mediated transrepression complexes in vivo. Similar data were obtained for the IR1 nGRE recently found in exon 6 of the human GR (13). Indeed, upon transfection of WT GR, but not of GR K293R, the same Dex-induced SUMO/corepressor repressing complex was associated on the IR1 nGRE present in exon 6 of the Cos-1 cell endogenenous GR gene (Fig. 2F). Interestingly, introduction of the K293R SUMOylation site mutation in the GR(NTD)-Gal(1-147) fusion protein (Fig. 1D) confirmed that SMRT binding to the GR NTD was dependent on SUMOylation of GR K293 (Fig. 1F) and, consequently, that this SUMOylation was instrumental in the repressing function of the GR NTD on its own in vitro (Fig. 1 E and F), in keeping with the data of Holmstrom et al. (14).

To unequivocally demonstrate that GR SUMOylation is crucial for IR nGRE-mediated repression in the mouse in vivo, we engineered a mouse SUMOylation mutant bearing a K-R mutation at GR position K310 (homologous to human K293). Examining, in the ear skin of this SUMOylation mutant, the effect of a topical Dex treatment on the expression of IR nGRE-containing genes revealed in all cases, including for the GR gene, a significant decrease in repression compared with WT (Fig. 2G), whereas no repression was exerted by GRα-D3 (Fig. 1B). Accordingly, ChIP assays carried out on mouse epidermis showed that GR, SUMO1, and the corepressor SMRT were associated with IR nGREs present in genes (including the GR) from WT mice, whereas this association was decreased in GR K310R mutants (Fig. 2H), but not abolished as in the GRα-D3 MEFs (Fig. 1C), suggesting that SUMOylation at position K294 (K277 in human) could contribute to the GC-induced IR nGRE-mediated repression activity of the GR in vivo, but not in vitro.

Interestingly, in all cases the integrity of the SUMOylation site (K293 in human, K310 in mice) is mandatory for GR binding to an IR nGRE and assembly of a repressing complex. However, most of intracellular GR is not SUMOylated (Fig. S3B), which suggests that the binding of GR to an IR nGRE may precede its SUMOylation.

Phosphorylation of the NTD Facilitates the SUMOylation of the GR, but Not of GR ABCD.

As it has been reported that phosphorylation of the NTD could enhance SUMOylation of the GR (15), we mutated (serine to alanine) five known phosphorylation sites in the human GR NTD: S134A, S203A, S211A, S226A, and S404A (2). Interestingly, upon GR transfection in Cos-1 cells under limiting Dex concentration (10 nM), the bindings of GR 5SA (all five S mutated), as well as those of SUMOs and SMRT/NCoR1 corepressors to the IR1 nGRE of the Cos-1 cell endogenous TSLP gene, were significantly decreased compared with GR WT (Fig. 2I). Similarly, mutations of the phosphorylation sites for both the c-Jun N-terminal kinase (JNK for S226) and the glycogen synthase kinase-3β (GSK-3β for S404) revealed that, at limiting Dex concentration (10 nM), these sites played a major role in the stimulation of GR NTD SUMOylation (Fig. 2I). Accordingly, when using the JNK inhibitor II and a GSK-3β inhibitor (LiCl) to depress GR phosphorylation in A549 cells, ChIP assays indicated that, at 10 nM Dex, these inhibitors decreased the binding of GR and SUMOs to the TSLP IR1 nGRE (relative to no addition of inhibitors; Fig. 2J). Most notably, IR nGRE-mediated direct repression was not altered by mutation of the NTD phosphorylation sites present in GR ABCD that lacks the LBD (Fig. 2K).

NCoR1 and/or SMRT Corepressors Are Mandatory for GR Binding to IR nGREs in Vivo.

As Lin et al. (16) reported that the DAXX repressor possesses a SUMO-interacting motif (SIM), we investigated whether DAXX could be involved in IR nGRE-mediated Dex-induced repression. Upon cotransfection of GR and DAXX into Cos-1 cells, ChIP assays did not provide any evidence supporting a DAXX association with IR nGREs present in TSLP and STRA13 endogenous genes, and their RNA transcripts were not affected in A549 cells on DAXX transfection (Fig. S3 F and G). However, alignment of the C-terminal AAs sequences of human and mouse NCoR1 and SMRT with the DAXX SIM revealed a conserved sequence LSDSD/E (Fig. 2L). Upon its deletion, GST-pull down assays (SI Materials and Methods) showed that NCoR1 and SMRT lost their interaction with SUMO1, demonstrating that this sequence was instrumental in these interactions (Fig. 2M).

As selective knock down of either SMRT or NCoR1 expression with siRNA in A549 cells was not fully effective at abolishing Dex-induced IR nGRE-mediated transrepression (6), we prepared NCoR1^−/− and SMRT^−/− MEFs lacking either NCoR1 or SMRT expression (Fig. S4A). Transcript analyses of endogenous IR nGRE-containing genes showed that, for most of them, Dex-induced repression was not affected in NCoR1^−/− or SMRT^−/− MEFs (Fig. 3A and Fig. S4B), but no Dex-induced repression was observed in NCoR1^−/− MEFs for the peptidyl-prolyl isomerase-like 5 (PPIL5) gene and in SMRT^−/− MEFs for upstream transcription factor 1 (USF1) and insulin receptor (INSR) genes (Fig. 3A and Fig. S4B). ChIP assays showed that, upon Dex treatment, the binding of GR together with those of NCoR1 and/or SMRT to IR nGRE elements could be detected in WT MEFs (Fig. 3B). However, these NCoR1 and SMRT bindings were gene dependent as (i) in one instance, one of the two corepressors was bound in WT MEFs and, in its absence, was replaced by the other one (corepressor preferential binding, as in the case of IL6 and STRA13), whereas (ii) in another instance, there was no such substitution (corepressor-specific binding, as in the case of USF1, PPIL5, and INSR) and no GR binding to IR nGRE elements could be detected, and (iii) in a third instance, both corepressors were concomitantly found bound in WT MEFs [no preferential binding, as in the case of TSLP and tumor necrosis factor receptor superfamily member 19 (TNFRSF19)] (Fig. 3B and Fig. S4C). However, in this latter case, the absence of either one of the two corepressors did not abolish the repression of the TSLP and TNFRSF19 genes (Fig. 3A and Fig. S4B), indicating that the simultaneous binding of both NCoR1 and SMRT to their nGREs was not required for repression.

Fig. S4.

NCoR1 and/or SMRT corepressors and HDAC3 are required for Dex-induced IR nGRE-mediated repression in vivo. (A) qRT-PCR for SMRT and NCoR1 transcripts in MEFs derived from WT, NCoR1^−/−, or SMRT^−/− mouse embryos. (B) As in A, but for IR nGRE-containing genes in MEFs, treated with vehicle, Dex (0.5 µM), and RU (3 µM) for 6 h. (C) qPCR analyses of ChIP assays performed with WT, SMRT^−/−, or NCoR1^−/− MEFs, treated with vehicle, Dex (1 µM), and RU (6 µM) for 1 h, showing GR and corepressors bindings to the indicated IR nGRE regions. (D) As in A, but from ear epidermis in WT, SMRT^ep−/−, and/or NCoR1^ep−/− mutant mice. (E) As in D but for (+)GRE-containing genes from [SMRT/NCoR1]^ep−/− mutant mice. Mouse ears were treated with vehicle, Dex (6 nmol/cm²), and RU (36 nmol/cm²) for 18 h. (F) As in C, but using dorsal epidermis from [SMRT/NCoR1]^ep−/− mutant mice, treated as in E for 6 h showing the binding of indicated proteins to the [(+)GRE]2x region of FKBP5 gene. (G) As in D, but for HDAC3 transcripts from mouse ear epidermis in WT and HDAC3^ep−/− mutant mice. Values are mean ± SEM.

Fig. 3.

NCoR1 and/or SMRT corepressors and HDAC3 are required for Dex-induced IR nGRE-mediated transrepression in vivo. (A) qRT-PCR for IR nGRE-containing genes in MEFs derived from WT, NCoR1^−/−, or SMRT^−/− mouse embryos, treated with vehicle, Dex (0.5 µM), and RU (3 µM) for 6 h. (B) qPCR analyses of ChIP assays performed with WT, SMRT^−/−, or NCoR1^−/− MEFs, treated with vehicle, Dex (1 µM), and RU (6 µM) for 1 h, showing GR and corepressors bindings to the indicated IR nGRE regions. (C) As in A, but using mouse ear epidermis in WT, SMRT^ep−/− and/or NCoR1^ep−/− mutant mice. Mouse ears were treated with vehicle, Dex (6 nmol/cm²), and RU (36 nmol/cm²) for 18 h. (D) As in B, but using dorsal epidermis of SMRT^ep−/− and/or NCoR1^ep−/− mutant mice, treated as in C for 6 h. (E) As in C, but in WT and HDAC3^ep−/− mutant mice. (F) As in D, but in WT and HDAC3^ep−/− mutant mice. Values are mean ± SEM.

To extend these observations in vivo, we then generated keratinocyte-selective KO of NCoR1 and/or SMRT in mouse epidermis, which resulted in decreased expression of NCoR1 (70–80%) and SMRT (higher than 90%) (Fig. S4D; these single and double mutant mice are hereafter designated as NCoR1^ep−/−, SMRT^ep−/− and [SMRT/NCoR1]^ep−/−). Dex-induced repression of three IR nGRE-containing genes, STRA13, keratin 14 (K14), and keratin 5 (K5), was abolished in [SMRT/NCoR1]^ep−/− mice, but not in SMRT^ep−/− or NCoR1^ep−/− single mutants (Fig. 3C). ChIP assays using WT dorsal epidermis showed that upon Dex treatment, GR was associated on IR nGRE sequences with either both SMRT and NCoR1 (for K14 and K5) or NCoR1 alone (for STRA13) (Fig. 3D). Neither GR binding nor corepressor recruitment was observed in the [SMRT/NCoR1]^ep−/− mouse epidermis, whereas GR binding and corepressor substitution was observed in repressing complexes for the STRA13 IR1 nGRE in NCoR1^ep−/− cells and for the K14 IR1 nGRE in SMRT^ep−/− cells (note in this case the significant increase in NCoR1 binding). Thus, depending on the identity of the nGRE-containing gene, one of the two corepressors (NCoR1 or SMRT) is indispensable for both GR binding to an IR nGRE sequence and direct transrepresssion.

Histone Deacetylase 3 Is Essential for Dex-Induced IR nGRE-Mediated Transrepression in Vivo.

The association of histone deacetylase 3 (HDAC3) with SMRT has been previously demonstrated in vitro (17), and we reported that HDAC3 is found together with SMRT/NCoR1 in complexes bound on IR nGREs in vivo (6) (Fig. 2E). To examine whether HDAC3 was required for Dex-induced IR nGRE-mediated repression, we selectively ablated HDAC3 in mouse epidermal keratinocytes (Fig. S4G). No Dex-induced repression was observed in HDAC3^ep−/− mice for K14, K5, and STRA13 IR nGRE-containing genes (Fig. 3E). ChIP assays using mouse epidermis showed that the binding of GR, as well as those of corepressors SMRT/NcoR1 to IR nGREs, was not affected in HDAC3^ep−/− mice (Fig. 3F). Interestingly, selective mutations of NCoR1 and SMRT abolished HDAC3 binding to IR nGREs, indicating that this binding is mediated by SMRT/NCoR1 (Fig. 3D).

The Repressing Complexes Containing the SUMOylated GR and NCoR1/SMRT Corepressors Do Not Repress (+)GRE-Mediated Transactivation and Do Not Bind to a (+)GRE.

Dex-induced (+)GRE-mediated transactivation of 3 genes [regulated in development and DNA damage responses 1 (REDD1), MURF and FK506 binding protein 5 (FKBP5)] was not increased in vivo in epidermal keratinocytes of [SMRT/NCoR1]^ep−/− mice (Fig. S4E), and ChIP assays did not reveal any SMRT/NCoR1 binding on the [(+)GRE]2× DNA binding site (DBS) region of the FKBP5 gene (14) (Fig. S4F). Of note, using the mouse SUMOylation mutant GR K310R in vivo, we did not find, upon Dex treatment, any increase in the expression of the single (+)GRE REDD1 gene, nor of that of the [(+)GRE]2× DBS FKBP5 gene, whereas the Dex-induced repression was alleviated in IR nGRE-containing genes (Fig. 2G). ChIP assays demonstrated that GR from dorsal keratinocytes of both WT and GR K310R mutant mice did bind efficiently to the [(+)GRE]2× DBS of the FKBP5 gene, but in both cases, no binding of SUMO1 and SMRT could be detected (Fig. 2H). Accordingly, upon GR transfection into Cos-1 cells, ChIP assays revealed that activating complexes [GR, steroid receptor coactivator (SRC)2/transcriptional mediator/intermediary factor 2 (TIF2), SRC3] were present on (+)GRE regions of endogenous genes (Fig. S5A), whereas repressing complexes (GR, SUMOs, NCoR1, SMRT) were found only on IR nGRE regions of endogenous genes (Fig. S5B).

Fig. S5.

SUMOylation of the GR did not affect (+)GRE-mediated transactivation. (A) qPCR analyses of ChIP assays performed with Cos-1 cells transfected with GR or GR K293R, treated with vehicle, Dex (1 µM), and RU (6 µM) for 1 h, showing the association of indicated proteins on the (+)GRE region of genes as indicated. (B) As in A, but for IR nGRE regions.

To conclusively demonstrate that the inability of the SUMOylated GR repressing complex to inhibit (+)GRE-mediated transactivation reflects its inefficiency to bind to a (+)GRE, a vector containing both a single (+)GRE and an IR1 nGRE DBS separated by ∼600 bp was transfected in Cos-1 cells (Fig. 4A). GR, SUMOs, and NCoR1/SMRT were bound on the IR1 nGRE, whereas non-SUMOylated GR and the coactivators SRC2(TIF2) and SRC3 were assembled on the (+)GRE (Fig. 4B), thus indicating that the repressing SUMOylated GR was bound to the IR1 nGRE, but not to the (+)GRE, whereas the activating non-SUMOylated GR was bound only to the (+)GRE. Furthermore, using the same vector, but containing a [(+)GRE]3× DBS instead of a single (+)GRE DBS, ChIP assays did not reveal any decrease of GR, SUMOs, and NCoR1/SMRT bound to the IR1 nGRE (Fig. 4B).

Fig. 4.

SUMOylation of the GR did not affect (+)GRE-mediated transactivation. (A) Schematic representation for pGL3 luciferase reporters containing either a single (+)GRE or a [(+)GRE]3× DBS together with an IR nGRE DBS. (B) qPCR analyses of ChIP assays performed with Cos-1 cells transfected with GR and luciferase reporters (as in A), treated with vehicle, Dex (1 µM), and RU (6 µM) for 1 h, showing the binding of GR, cofactors, and SUMOs on indicated regions.

The GR DNA Binding Domain Is Differentially Involved in (+)GRE-Mediated Transactivation and IR nGRE-Mediated Transrepression.

It is known that the GR DNA binding domain (DBD) interacts with (+)GRE and IR nGRE sequences (18, 19), and that deletion of the GR DBD (ΔDBD) or either one of the two zinc fingers (ΔZF1/ΔZF2) abolished these interactions (Fig. 5A and Fig. S6 A–C). Interestingly, among the three AA residues (K442, V443, and R447) that were reported to directly interact with (+)GRE and IR nGRE bases in crystal structures (Fig. 5C) (18, 19), we found that only K442 was indispensable for GR binding to an IR nGRE, whereas mutation of all three residues affected GR binding to a (+)GRE (Fig. 5 A and B and Fig. S6 B and C). Interestingly, a triple AA mutation (TM) within the GR first zinc finger into the corresponding estrogen receptor α (ERα) AAs (G439, S440, and V443 into E, G, and A, respectively) (20) abolished its interaction with (+)GRE, but not with IR nGRE (Fig. 5 A and B and Fig. S6 A–C). The double mutation G439E/S440G could only partially affect GC-induced (+)GRE-mediated transactivation, and had no effect on GC-induced IR nGRE-mediated repression (Fig. 5 A and B and Fig. S6 A–C).

Fig. 5.

The GR DBD is differently involved in (+)GRE-mediated transactivation and IR nGRE-mediated transrepression. (A) Relative RNA transcripts (normalized to GR full length, taken as 100%) for GR mutants as indicated (Fig. S6 A and B). (B) As in A, but for the binding of GR mutants to DBS of indicated genes (Fig. S6 A and C). (C) Representation of interactions between GR DBD AAs and TSLP IR1 nGRE bases and between GR DBD AAs and consensus (+)GRE bases. (D) qPCR analyses of ChIP assays performed with A549 cells, treated with vehicle, IL6 (10 ng/mL), and Dex (1 µM) for 1 h, showing the binding of GR, STAT3, and corepressors to the IR0 nGRE and SBE of the SOCS3 gene. Values are mean ± SEM.

Fig. S6.

The GR DBD is differently involved in (+)GRE-mediated transactivation and IR nGRE-mediated transrepression. (A) Location of point mutations within the zinc fingers (ZF1 and ZF2) of human GR. Residues that prevent the GR from interacting with (+)GRE and IR nGRE are indicated by black dot and red star, respectively. (B) qRT-PCR for transcripts from the SGK1 and STRA13 genes in Cos-1 cells transfected with GR or GR DBD mutants, treated with vehicle, 0.5 µM Dex, and 3 µM RU for 6 h. (C) qPCR analyses of ChIP assays performed with Cos-1 cells transfected with GR or GR DBD mutants, treated with vehicle or 1µM Dex for 1 h, showing the binding of GR to the SGK1 (+)GRE and TSLP IR1 nGRE, using GR antibody. The IgG control for i.p. (∼0.02% of input) is not shown. (D) IR nGRE sequences present in the indicated genes. Values are mean ± SEM. **P < 0.01.

IR nGREs Are Not STAT3 Binding Elements (SBEs).

Langlais et al. (21) speculated that IR nGREs could be STAT3 binding elements (SBEs). To investigate this possibility, we analyzed in A549 cells the binding of GR to the SBE and IR nGRE DBSs present in the regulatory region of the suppressor of cytokine signaling 3 (SOCS3) gene promoter. Upon Dex treatment, GR, SMRT, and NCoR1 corepressors were associated on the SOCS3 IR0 nGRE (Fig. 5D, Left), whereas the association of STAT3 to the SOCS3 SBE was observed upon IL6 treatment on its own, and the tethered association of GR and SMRT with STAT3 bound to the SBE site was detected only after further Dex cotreatment (Fig. 5D, Right). Thus, the STAT3 binding element and the IR nGRE are distinct DBSs, which are selectively and independently bound by STAT3 and the GR, respectively.

SI Materials and Methods

Materials.

GR rabbit polyclonal antibody and control rabbit IgG, SMRT, NCoR1, and TIF2/SRC2 antibodies were generated in-house. All in-house generated antibodies were tested by ELISA, immunoprecipitation, Western blotting, and absence of reaction in null mutant mice to ensure specificity. HDAC3 (SC-11417X), SUMO1 (sc-5308), SUMO2/3 (sc-26969), SRC3 (sc-9119x), and DAXX (sc-7152) antibodies were from Santa Cruz Biotechnology. Stat3 (#9132) rabbit polyclonal antibody was from Cell Signaling. HDAC2 (ab7029) rabbit polyclonal antibody was from Abcam. Dexamethasone (Dex), RU 486 (RU), and LiCl were from Sigma Aldrich. JNK inhibitor II was from Calbiochem.

Bioinformatics Analysis.

For identification of the evolutionary conserved region within the GR NTD, GR protein sequences for human (Homo), frog (Xenopus), rat (Rattus), cock (Gallus), zebrafish (Danio), and salmon (Salmo) were obtained from the National Center for Biotechnology Information (NCBI) (www.ncbi.nlm.nih.gov/protein/). Sequence alignment was performed with Jalview online software (29).

Mice.

WT BALB/cByJ and C57BL/6J 8- to 10-wk-old mice were from Charles River Laboratories. NCoR1 floxed (NCoR1^fl/fl), SMRT floxed (SMRT^fl/fl), and GR K310R mutant mice were generated at the IGBMC/ICS. Floxed HDAC3 (HDAC3^fl/fl) mice were a gift from Mitchell A. Lazar, Division of Endocrinology, Diabetes, and Metabolism, Department of Medicine, and The Institute for Diabetes, Obesity, and Metabolism, Perelman School of Medicine at the University of Pennsylvania, Smilow Translational Research Center, Philadelphia. Epidermal keratinocyte-specific KO mutant of SMRT (SMRT^ep−/−), NCoR1 (NCoR1^ep−/−), and HDAC3 (HDAC3^ep−/−) were generated by initially crossing floxed female mice with K14-Cre-ERT² male mice (30) and subsequent tamoxifen IP injection for 5 d. SMRT^−/+ and NCoR1^−/+ heterozygous mice were generated by crossing floxed female mice with CMV-Cre males. Genotyping was performed by PCR on genomic DNA isolated from mouse tails. SMRT^−/− and NCoR1^−/− MEF cells were isolated from homozygous embryos (E13) derived from heterozygous crosses.

To generate mutant mice expressing exclusively the GRα-D3 isoform in a given tissue, we first inserted two LoxP sites into the first exon of the murine GR (NM_008173) (Fig. S2 A and B). The upstream LoxP1 was inserted into the noncoding part of the first exon, and the second LoxP2 was inserted into the same exon between the initiation codons of GRα isoforms D2 and D3, in such a manner that 16 codons were introduced in the GR coding sequence (Fig. S2A). This modified region of exon 1 was then introduced in the GR gene of ES cells to generate blastocysts and subsequently floxed GR^fl/+ heterozygous mice. However, crosses between GR^fl/+ mice did not result in any GR^fl/fl offspring. No GR protein could be detected in liver samples collected from GR^fl/fl E13 embryos, whereas it was expressed in both WT and GR^fl/+ embryos (Fig. S2C). To determine whether GR transcription was occurring in floxed GR^fl/fl mice, we then looked for transcripts using primers corresponding to regions located before (FR1), after (FR3), or across the second LoxP site (LoxP2; Fig. S2B). Transcript analyses using E13 liver samples from WT, GR^fl/+, and GR^fl/fl embryos showed that transcription of the GR did take place over the LoxP2 sequence in GR^fl/fl embryos (Fig. S2E, Left). Moreover, similar qRT-PCR searches of transcripts of the last exons encoding the LBD showed that the whole floxed GR was transcribed, indicating that the insertion of LoxP sites into the murine GR first exon was preventing its translation, thereby resulting in GR-null mice. Therefore, to obtain GRα-D3 mice, GR^fl/+ mice were bred with the Rosa26-deleter Cre mouse line to create GR/GRα-D3 heterozygous mice, which when bred yielded GRα-D3 homozygous mice. Much like GR-null mice (31), GRα-D3 mice died before or shortly after birth from respiratory failure. However, less than 5% of the GRα-D3 embryos survived after birth. Analyses of protein extracts from ears of GRα-D3 survivors showed that they selectively expressed the GRα-D3 isoform (Fig. S2D).

All primers used for genotyping are available on request.

Mouse Epidermis Isolation.

Epidermis preparation from dorsal skin was as previously described (6). For RNA isolation from ear epidermis, mouse ears were cut and put on ice for 10 min before splitting the ventral and dorsal halves with fine forceps. Both halves were then incubated in 2.5 mg/mL dispase (Gibco) in PBS for 1 h at 37 °C. Epidermal sheets were recovered by using a forceps, directly put in TRI reagent (Molecular Research Center) at room temperature, and used for RNA isolation as per the manufacturer’s instruction.

For ChIP assays, the inner side of dorsal skin was scraped off fat and floated over a 0.8% trypsin solution [0.8% trypsin (wt/vol) dissolved in PBS] for 30 min at 37 °C. Epidermal sheets were recovered in cell culture medium [Eagle's minimum essential medium (EMEM) without calcium + 10% (vol/vol) FCS], incubated at 37 °C for 15 min, and filtered through a 70-mm nylon cell strainer (BD Falcon). Epidermal cells were then cross-linked for 10 min in 1% formaldehyde and further processed as previously described (6).

Luciferase Reporter Vectors and Expression Vectors.

pGL3-nGRE or (+)GRE luc reporters were previously described (6). Double-stranded (+)GRE or [(+)GRE]3× was inserted into the pGL3-IR1 nGRE luc reporter (without SV40 enhancer) using the NotI site to create the pGL3-(+)GRE or [(+)GRE]3×-IR1 nGRE luc reporter.

GR NTD and LBD deletion fragments, as well as the DAXX coding sequence (CDS), were PCR amplified from an A549 cell cDNA mixture using primers with flanking KpnI and BamHI sites at the 5′ and 3′ ends, repectively. PCR products were digested with KpnI and BamHI and ligated into pcDNA3.1 vector (Invitrogen) digested with KpnI and BamH1.

All clones were screened by PCR and restriction digestion and confirmed by sequencing. Primers are available on request.

Site-Directed Mutagenesis.

Site-directed mutagenesis was performed with the QuikChange II XL Site-Directed Mutagenesis Kit (Agilent). Primers are available on request.

Cell Culture Conditions for Luciferase Assays, qPCR RNA Determination, and ChIP Assays.

Cos-1 monkey kidney fibroblast-like cells (CCL-70; ATCC) were maintained in DMEM (1 g/L glucose) medium containing 5% (vol/vol) FCS and gentamycin. A549 human lung epithelial cells (CCL-185; ATCC) were maintained in DMEM/HAM F12 (1:1) medium containing 10% (vol/vol) FCS and gentamycin. Cells were transfected using Fugene 6 reagent, as instructed (Roche).

For luciferase reporter assays, Cos-1 cells seeded on 24-well tissue culture plates overnight at 60% confluency were transfected with 100 ng pCMV β galactosidase, 200 ng pGL3 reporter vectors, and 500 ng expression vectors as indicated into each well. Six hours after transfection, medium was changed, and cells were maintained in medium containing charcoal-treated FCS. Twenty-four hours after transfection, the indicated chemicals were added for 6 h. The luciferase assay was carried out as instructed (Promega). Normalized values are reported as the mean ± SEM; each value originates from at least three individual transfections with assays performed in triplicate.

For qPCR RNA determination, Cos-1 or A549 cells seeded on six-well tissue culture plates overnight at 60% confluency were transfected with 2 µg expression vectors into each well. Six hours after transfection, the medium was changed, and cells were maintained for 36 h in medium containing charcoal-treated FCS, before being treated as described in the legends and figures for 6 h. RNA was isolated using TRI reagent (MRC) following the manufacturer’s protocol. RT-PCR using SuperScript II Reverse Transcriptase kit (Invitrogen) and qPCR using SYBRgreen reagent (Roche) were performed according to the manufacturer’s instruction.

For the ChIP assay, Cos-1 cells seeded on a 15-cm tissue culture dish at 60% confluency were transfected with 12 µg expression vector with or without 6 µg pGL3 reporter. Six hours after transfection, the medium was changed, and cells were maintained for 36 h in medium containing charcoal-treated FCS and then treated as described in the legends and figures. Confluent A549 cells seeded on 15-cm tissue culture dishes were maintained in medium containing charcoal-treated FCS and treated as described above. Cells were treated with formaldehyde and further processed for ChIP assays. Primers are available on request.

siRNA Treatment.

Control siRNA (sc-37007) was from Santa Cruz Biotechnology; ON-TARGETplus SMARTpool siRNAs against SMRT (l-020145-01-0050) and NCoR1 (l-003518-00-0050) were from Dharmacon. siRNAs were transfected into cells using Lipofectamine 2000 (Invitrogen) reagent following the manufacturer’s instructions and maintained in charcoal-treated FCS. Forty-two hours after transfection, cells were treated as described for 6 h, followed by the luciferase assay, RNA isolation, and qRT-PCR. For ChIP assays, 70 h after transfection, cells were treated for 1 h as described, followed by formaldehyde cross-linking.

SUMOylation Assay.

Cos-1 cells seeded on a 100-mm tissue dish overnight at 60% confluency were transfected with 6 µg of each of the indicated expression vectors. Six hours after transfection, medium was changed, and cells were maintained in medium containing charcoal-treated FCS. Thirty-six hours later, 1 µM Dex was added to the medium for 1 h; 300 µg whole cell extracts or total nuclear extracts was immunoprecipitated with GR antibody and washed five times in RIPA lysis buffer containing 2 mM EDTA, 20 mM Tris⋅HCl, pH 8, 150 mM NaCl, 1% Nonidet P-40, and 10% (vol/vol) glycerol. Immunoprecipitates were resolved by SDS/PAGE and immunoblotted against GR antibody.

Immunoblot Analyses.

Immunoblots from cell or tissue extracts were performed following standard SDS/PAGE procedures. Primary antibodies were as described in Materials and Materials. Proteins were visualized through enhanced chemiluminiscence (Pierce).

Pull-Down Assay of NCoR1/SMRT with GST-SUMO1.

C-terminal moieties of NCoR1 and SMRT fragments were amplified from an A549 cell cDNA mixture using primers with flanking BamHI and EcoRI sites at 5′ and 3′ ends, respectively, and ligated into the pcDNA3.1 vector. C-terminal NCoR1 or SMRT proteins were then in vitro translated with TNT Quick Coupled Transcription/Translation Systems (Promega). The SUMO1 coding sequence was amplified from the same A549 cell cDNA mixture using primers with flanking BamHI and EcoRI sites at the 5′ and 3′ ends, respectively, and ligated into the pGEX-2T vector (GE Healthcare). Ten micrograms purified GST or GST-SUMO1 fusion proteins was incubated with indicated in vitro-synthesized [³⁵S]methionine-labeled NCoR1 or SMRT proteins for binding assay, as described in ref. 32. Primers are available on request.

Discussion

SUMOylation of the GR NTD and the Subsequent Formation of an NCoR1/SMRT/HDAC3 Repressing Complex Is Mandatory for GC-Induced IR nGRE-Mediated Direct Transrepression, but Does Not Affect (+)GRE-Mediated Transactivation.

We demonstrate here that SUMOylation of human GR at NTD position K293 (K310 in the mouse) is mandatory for IR nGRE-mediated transrepression (Fig. 2 A–D). In vitro studies and ChIP assays have revealed that GR SUMOylation allows the formation of a GC-induced NCoR1/SMRT-containing repressing complex, in which SMRT and NCoR1 are bound to SUMO through their SUMO-interacting motifs (Fig. 2 L and M). Most notably, using GR K310R SUMOylation mutant mice, we unequivocally establish that this site is essential for efficient GC-induced IR nGRE-mediated repression in vivo, whereas SUMOylation at K294 (K277 in human) may also contribute, albeit to a lesser extent (Fig. 2 G and H). In keeping with this conclusion, MEFs expressing selectively the GRα-D3 isoform, which lacks both SUMOylation sites, could not assemble a repressing complex and did not exert any GC-induced IR nGRE-mediated repression (Fig. 1 B and C). In this respect, we note that Gross et al. (22) found that in U2OS cells expressing the GRα-D isoform only, Dex did not repress the expression of the IR nGRE-containing Bcl-xL and Survivin/BIRC5 genes (Fig. S6D) (6). Along the same lines, Paakinaho et al. (23) reported that mutation of all three GR SUMOylation sites in HEK293 cells prevents GC-induced repression of CCND2 and ZIC2 genes that we found to contain IR nGREs (Fig. 2G and Fig. S6D).

Importantly, using either mice in which SMRT and/or NCoR1 were selectively ablated in epidermal keratinocytes (Fig. 3 C and D) or SMRT^−/− and NCoR1^−/− MEFs (Fig. 3 A and B), we demonstrate that both SMRT and NCoR1 are instrumental in IR nGRE-mediated repression in vivo. Intriguingly, depending on the expressed gene identity, the repressing complex contains both SMRT and NCoR1 repressors or only one of them, and in the latter case, its mutation could lead or not to its replacement by the other one (Fig. 3 B and D). Moreover, the presence of NCoR1 in the repressing complex requires an interaction with K579 in the LBD (Fig. 1I). Whether SMRT may also interact with the LBD is unknown, but it is noteworthy that an active repressing complex containing SMRT, but not NCoR1, can be assembled on the LBD-deleted GR ABCD (Fig. 1 G–I). Structural studies on purified repressing complexes are required to establish whether SMRT could similarly bind to the GR LBD, and also to elucidate how the association of SMRT/NCoR1 results in a stable GR binding to IR nGREs, in marked contrast with the weak in vitro interaction found between the isolated GR DBD and an IR nGRE DBS (19). Most notably, the latter study revealed an interaction between GR DBD K442 and the TSLP IR1 nGRE (Fig. 5C), which is remarkably in keeping with our own data showing that, in transfected cells, mutation of GR K442 prevents the GC-induced interaction between the GR and the TSLP IR1 nGRE (Fig. 5B). Interestingly, the GR DBD is not specifically required for Dex-induced direct repression, because Gal4 transactivation was repressed when the GR NTD was fused to the yeast Gal4 DBD (1-147) (Fig. 1 D and E), indicating that the GR NTD on its own could assemble a repressing complex.

Of note, and as expected (24), the binding of HDAC3 to IR nGREs in vivo is mediated through interaction with SMRT/NCoR1 (Fig. 3D). Using mice in which HDAC3 was selectively ablated in epidermal keratinocytes, we demonstrated that it is essential for Dex-induced IR nGRE-mediated repression; however, its ablation did not affect the binding of GR, nor that of NCoR1/SMRT on IR nGREs (Fig. 3 E and F). Our results also indicated that although HDAC2 could be found in repressing complexes bound on the IR nGRE regions, it was not mandatary for repression (Fig. 3 E and F).

Previous transfection studies have suggested that SUMOylated nuclear receptors, including GR, could control negatively the expression of luciferase reporter genes containing multiple, but not single DNA binding element (14, 25, 26). However, we found that the (+)GRE-mediated transactivation of genes containing 1 or 2 (+)GRE DBS was not increased in vivo in epidermal keratinocytes of [NCoR1/SMRT]^ep−/− mice compared with WT mice (Fig. S4 E and F). Similar in vivo results were obtained using GR K310R SUMOylation mouse mutants (Fig. 2 G and H). Furthermore, upon transfecting Cos-1 cells with a vector containing both a (+)GRE DBS and an IR1 nGRE DBS separated by 600 bp, ChIP assays showed that repressing complexes were selectively assembled on the IR nGRE DBS, whereas only activating complexes were associated with the (+)GRE DBS (Fig. 4). We conclude that the SUMOylated GR cannot assemble a repressing complex on a (+)GRE.

Two Main GC-Regulated Functions Are Exerted by a Unique GR Under the Control of a Single GC Ligand.

As the LBD-truncated GR ABCD is a constitutive activator of transcription, it was proposed that for GR binding to a (+)GRE, the only requirement is a GC-dependent conformational modification of the LBD, which on its own unmasks the DBD (8). Similarly, our present results indicated that the binding of GR to an IR nGRE requires a GC ligand, which, by binding to the LBD, allows the unmasking of not only the DBD, but also of the NTD SUMOylation site and, consequently, the formation of a SMRT/NCoR1-repressing complex on the IR nGRE. In this respect, note that the GR ABCD on its own, but not its SUMOylation mutant GR ABCD K293R, is readily SUMOylated to assemble a repressing complex, which efficiently binds to an IR nGRE to repress transcription (Fig. 2 B and C). Note also that under similar conditions, the (+)GRE-mediated activation of transcription by GR ABCD is unaffected by the K293R mutation (Fig. S3 C–E).

Clearly, in addition to its well-established role in the translocation of unliganded GR from the cytoplasm to the nucleus, the binding of a GC to the GR appears to unveil the DBDs used for binding to (+)GRE and IR nGRE DBSs. In the case of (+)GREs, this unmasking allows on its own an efficient binding of the GR DBD to a (+)GRE, which appears to be incompatible with SUMOylation of the NTD. In contrast, the much weaker binding of the GR DBD to an IR nGRE (19) requires the subsequent SUMOylation of the NTD and the subsequent formation of a repressing SMRT/NCoR1 complex, to stably anchor the GR on the IR nGRE. Interestingly, our results indicate that such unmaskings of the NTD SUMOylation site could involve not only a ligand-dependent conformational modification of the LBD, which is prevented by RU486, but also the phosphorylation of sites located in the NTD. Indeed, upon mutation of five such phosphorylation sites, as well as upon addition of selective inhibitors of two cognate kinases (JNK and GSK-3β), ChIP assays have revealed that, under limiting Dex concentration (10 nM), there was a decrease in the binding of SUMOylated GR on an IR nGRE (Fig. 2 I and J). Importantly, the loss of NTD phosphorylation did not affect the binding of SUMOylated GR ABCD on the IR nGRE (Fig. 2K). It appears therefore that SUMOylation of the GR NTD could be controlled, not only by GCs, but also by site-specific phosphorylation mediated by protein kinases, which, interestingly, have been implicated in inflammation and homeostasis (27, 28).

In conclusion, for two of the three GR main functions (transactivation and direct transrepression), there is an initial common step consisting in the binding of a GC ligand, which in both cases, due to a GC-induced conformational modification of the LBD, leads to the unmasking of the NTD/DBD (ABCD) domains. Subsequently, the unveiled GR may bind efficiently to a (+)GRE and loosely to an IR nGRE (19). It is then and only then that, in the latter case, SUMOylation and assembly of a repressing complex may occur, thereby strengthening the binding of the GR to an IR nGRE DBS. It is noteworthy that with such operational mechanisms, there might be no need for the existence of a predestined nuclear pool of SUMOylated GR, as SUMOylation may occur only upon weak, but selective, binding of the GR to IR nGREs, which would be in keeping with the observation that only a small fraction of the cellular GR is actually SUMOylated (Fig. S3B).

Materials and Methods

Mice.

GRα-D3, GR K310R, NCoR1, SMRT, and HDAC3 mutant mice are described in SI Materials and Methods. Breeding, maintenance, and experimental manipulation of mice were approved by the Animal Care and Use Committee of the Institut de Génétique et de Biologie Moléculaire et Cellulaire (IGBMC)/Institut Clinique de la Souris (ICS).

The ChIP assay, luciferase assay, siRNA treatment, RNA isolation, and quantitative PCR (qPCR) analyses were performed as in ref. 6. Primers used are available on request.

Statistics.

Data are represented as mean ± SEM of at least three independent experiments and were analyzed by Microsoft Excel statistics software using the Student t test. P < 0.05 was considered significant.