Phosphorylation-Dependent Sumoylation Regulates Estrogen-Related Receptor-α and -γ Transcriptional Activity through a Synergy Control Motif

Abstract

Interplay between different posttranslational modifications of transcription factors is an important mechanism to achieve an integrated regulation of gene expression. For the estrogen-related receptors (ERRs) α and γ, regulation by posttranslational modifications is still poorly documented. Here we show that transcriptional repression associated with the ERR amino-terminal domains is mediated through sumoylation at a conserved phospho-sumoyl switch, ψKxEPxSP, that exists within a larger synergy control motif. Arginine substitution of the sumoylatable lysine residue or alanine substitution of a nearby phosphorylatable serine residue (serine 19 in ERRα) increased the transcriptional activity of both ERRα and -γ. In addition, phospho-mimetic substitution of the serine residue with aspartate restored the sumoylation and transcriptional repression activity. The increased transcriptional activity of the sumoylation-deficient mutants was more pronounced in the presence of multiple adjacent ERR response elements. We also identified protein inhibitor of activated signal transducer and activator of transcription y as an interacting partner and a small ubiquitin-related modifier E3 ligase for ERRα. Importantly, analysis with a phospho-specific antibody revealed that sumoylation of ERRα in mouse liver requires phosphorylation of serine 19. Taken together, these results show that the interplay of phosphorylation and sumoylation in the amino-terminal domain provides an additional mechanism to regulate the transcriptional activity of ERRα and -γ.

NUCLEAR HORMONE RECEPTORS (NRs) play essential roles in the regulation of a wide array of developmental and physiological pathways. NRs are regulated by specific ligands with the exception of orphan members, for which no known natural ligands have been identified to date (1). The estrogen-related receptors (ERRs) α and β (NR3B1 and NR3B2) were the first orphan NRs identified based on their high level of sequence identity with estrogen receptor (ER)α (NR3A1) (2). The ERR subfamily also contains a third isoform (ERRγ, NR3B3) (3, 4), and all three proteins possess the typical functional domains of NRs. The amino acid sequences of the three ERR isoforms are highly similar, sharing as expected the highest identity in the DNA-binding domain and ligand-binding domain (LBD). However, unlike most related NRs, the three ERR isoforms also share considerable amino acid sequence similarity in their respective amino-terminal domain (NTD), suggesting that the NTD may influence the transcriptional acitivity of the three ERRs by common mechanisms. Although the ERRs have no known natural ligand, the three receptors can be activated in a ligand-independent manner in the presence of coactivator proteins, most notably by members of the steroid receptor coactivator (SRC) and the peroxisome proliferator-activated receptor (PPAR)-γ coactivator 1 (PGC-1) families (4, 5, 6, 7). Indeed, the elucidation of the crystal structures of ERRα LBD bound to a PGC-1α peptide, as well as that of ERRγ LBD bound to a SRC-1 or receptor-interacting protein-140 peptides, have shown that the two ERRs assume the conformation of ligand-activated NRs in the apparent absence of a ligand, again suggesting that the presence of an agonist ligand may not be an obligatory requirement for the activation of the receptors (8, 9, 10). Thus, posttranslational modifications could play a major role in the control of ERR transcriptional activity. Although ERRα has been shown to be a phosphoprotein (11, 12, 13), the current knowledge about ERR posttranslational modifications is still very limited.

Phosphorylation of NRs, as well as their coactivators, is a well-documented mechanism involved in the control of their activities (reviewed in Refs. 14, 15, 16). Similarly, sumoylation, the process of conjugating the small ubiquitin-related modifier (SUMO) protein, has been reported for NRs and coregulators, namely the androgen (17), glucocorticoid (18), progesterone (19, 20), estrogen (21), and mineralocorticoid receptors (22, 23) as well as SF-1 (24, 25), PPARγ (26, 27, 28), liver receptor homolog 1 (29), liver X receptors (28), Tr2 (30), the SRC coactivators (19, 31, 32), the histone acetyltransferase p300 (33), and the nuclear receptor corepressor 1 (34). The exact function of sumoylation is still unknown although a growing body of evidence now supports the role of SUMO proteins in the negative regulation of transcription, mainly through corepressor recruitment or clearance-related mechanisms (27, 28, 35, 36, 37, 38).

SUMO proteins are conjugated to a lysine residue within the core consensus site ψKxE, where ψ represents a hydrophobic residue and x is any residue. Sumoylation is carried out by a set of enzymes that are distinct from those acting on the ubiquitin pathway (39, 40) and consist of a SUMO-activating heterodimeric complex consisting of Aos1 and Uba2 (E1), the single E2-type conjugating enzyme UBC9, and E3-like proteins, which serve to increase the affinity between UBC9 and the substrates by bringing them in a close proximity to UBC9 with a catalytically favorable orientation. In vitro, however, the E3-like activity is not necessary for sumoylation to occur, because E1 and UBC9 are sufficient to induce sumoylation. Three types of SUMO E3 ligases have been described: RanBP2 (a component of nuclear pore complex), the Polycomb protein Pc2, and members of the protein inhibitor of activated signal transducer and activator of transcription (PIAS) family (reviewed in Ref. 41).

A subset of consensus SUMO conjugation motifs has recently been extended to include ψKxExxSP, establishing new sumoylation sites that are phosphorylation dependent and thus referred to as phosphorylation-dependent sumoylation motifs (PDSMs) or phospho-sumoyl switches (42, 43, 44). In addition, this sequence also corresponds to the synergy control (SC) motif Px _(0–3)[I/V]K[Q/T/S/L/E/P]Ex _(0–3)P, a protein determinant that was identified before the sumoylation consensus and initially proposed to modulate higher-order interactions among transcription factors, including NRs and their coregulators (45, 46). The efficiency of the transcriptional repression exerted by sumoylation of transcription factors within SC motifs has been proposed to depend on the number of consecutive DNA response elements present in the target promoter (47).

Here we present evidence that ERRα and -γ are sumoylated within their respective NTDs and that this modification is induced by phosphorylation of a functional PDSM. Our results show that sumoylation negatively affects ERRα and -γ transcriptional activity without altering subcellular localization, DNA binding properties, or interactions with the coactivator PGC-1α. The PDSM within the NTD of the ERRs also controls synergy in the presence of multiple ERR response element (ERRE) and consequently, sumoylation-deficient receptor variants are more potent activators of transcription on the polymorphic ESRRA promoter containing multiple copies of the ERRE. We have also demonstrated that PIASy interacts with and possess an E3-ligase activity toward ERRα. Using a phosphorylation-specific antibody, we found that phosphorylation of serine 19 is required for sumoylation of endogenous ERRα in mouse liver. Thus, the interplay of phosphorylation and sumoylation at the SC motif provides a novel mechanism to regulate the transcriptional activities of ERRα and -γ.

RESULTS

Sumoylation of ERRα and -γ

Scanning of the amino acid sequence of the ERR isoforms led to the identification of two consensus attachment sites for SUMO proteins (lysines 14 and 403) in ERRα and three (lysines 40, 360, and 439) in ERRγ (Fig. 1A). Interestingly, the NTD sumoylation sites (lysine 14 in ERRα and 40 in ERRγ) were found to be embedded within a PDSM motif (42, 43, 48) (Fig. 1B). To determine whether the two ERR isoforms are targets of sumoylation, we first cotransfected human embryonic kidney (HEK)293 cells with either a myc-ERRα- or a Flag-ERRγ-tagged expression vector along with an hemagglutinin (HA)-SUMO2 plasmid. Immunoprecipitation with either anti-myc or anti-Flag antibodies followed by immunoblotting using an anti-HA polyclonal antibody suggested that both ERRα (Fig. 1C) and -γ (Fig. 1D) are modified by SUMO2.

Fig. 1.

ERRα and -γ Contain Consensus Sites for Sumoylation and Are Modified by SUMO2

A, Schematic representation of the domain organization of ERRα and -γ along with their putative sumoylation sites. B, Sequence alignment of the ERRα NTD showing the conservation of the N-terminal sumoylation site in different species, and among the human ERRα, -β, and -γ isoforms. ψ, hydrophobic residue; h, human; m, mouse; r, rat; d, Drosophila. The correspondence of this sumoylation site with the PDSM and the SC motif is also depicted. C, HEK293 cells were transfected with expression plasmids for HA-SUMO2 and myc-ERRα as specified. Whole-cell lysates were prepared and used for immunoprecipitation with anti-myc or -Flag antibody, followed by Western blotting with anti-HA, -ERRα, or -Flag antibody as indicated. Ab, Antibody; i, input. D, Same as panel C except that Flag-ERRγ was expressed and analyzed. The asterisk denotes a nonspecific band. DBD, DNA-binding domain; WB, Western blot.

Identification of Sumoylation Sites

To identify the sumoylation sites in ERRα and -γ, potential target lysines were mutated to arginines, and the point mutants were subjected to in vitro sumoylation assays with recombinant SUMO1 and SUMO3 proteins. As shown in Fig. 2, A and B, respectively, only the ERRα K14R and ERRγ K40R mutants displayed significantly decreased levels of sumoylation whereas the remaining point mutants (ERRα K403R, ERRγ K360R, and K439R) were sumoylated to a level similar to the wild-type receptors. In the absence of the recombinant E1 activating enzyme, the sumoylated forms were totally absent. The absence of one band in the K403R mutant in comparison to wild-type ERRα suggests residual sumoylation of lysine 403 (Fig. 2A).

Fig. 2.

The NTDs of ERRα and -γ Harbor the Main SUMO Attachment Sites

A, ERRα and its mutants were translated and labeled with ³⁵S in vitro and then subjected to in vitro reconstituted sumoylation assays with (+) or without (−) the E1 recombinant activating enzyme along with recombinant SUMO1 or SUMO3. The single asterisk (*) denotes a missing band for the ERRα K403R mutant. B, Same as panel A except that ERRγ and its mutants were analyzed. C, HEK293 cells were transfected with expression plasmids for ERRα and its mutants, along with constructs for SUMO2GG and UBC9 (SUMO +; lanes 1, 3, 5, and 7) or for SUMO2ΔGG and UBC9(C93S) (SUMO Δ; lanes 2, 4, 6, and 8). Whole-cell lysates (80 μg) were prepared for Western blot analysis with the indicated antibodies. D, Same as panel C except that ERRγ and its mutants were analyzed along with constructs for SUMO2GG and UBC9 (SUMO +; lanes 1, 3, 5, 7, and 9) or for SUMO2ΔGG and UBC9(C93S) (SUMO Δ; lanes 2, 4, 6, 8, and 10). Sumoylated ERRα (S-ERRα) and ERRγ (S-ERRγ) are labeled with open arrowheads, whereas the nonsumoylated forms are marked by solid arrowheads. The double asterisks (**) represent a nonspecific band.

Endogenous SUMO proteins are subjected to a maturation step before they can be conjugated to the acceptor protein. The extreme carboxy-terminal end is cleaved by SUMO-specific proteases to expose a diglycine motif necessary for conjugation, and the removal of the diglycine motif prevents the conjugation (49). To determine whether the same site would be subjected to sumoylation in cells, the KR point mutants of ERRα and -γ were transfected in HEK293 cells, along with an HA-SUMO2GG-activated form and UBC9 or with HA-SUMO2ΔGG and UBC9-C93S dominant-negative forms, and 80 μg of extracts was subjected to Western blot analysis using an anti-ERRα or anti-Flag M2 antibody. As shown in Fig. 2, C and D, slower migrating bands were present when the activated form of SUMO2 and UBC9 were introduced in the HEK293 cells. Consistent with this, the bands were absent when the dominant-negative forms were used, demonstrating that the slower migrating bands were the sumoylated forms of the receptors. Moreover, the slower migrating band was significantly decreased for ERRα K14R (Fig. 2C) and absent for ERRγ K40R mutants (Fig. 2D), confirming that the NTD of both ERR isoforms is the main SUMO attachment site. The residual sumoylation in the ERRα K14R lane may suggest a possible modification of lysine 403 (Fig. 2C).

Increased Transcriptional Activity of Sumoylation-Deficient Mutants of ERRα and -γ Requires Multiple DNA Response Elements

To assess whether sumoylation of the two ERR isoforms affects their transcriptional activity, we next transfected HeLa cells with either ERR wild-type forms and NTD KR mutants in the presence or the absence of the coactivator PGC-1α together with the reporter construct 3xESRRApromoter-luciferase (LUC). This reporter is driven by the promoter of the gene encoding ERRα (ESRRA) which is a known target of both ERRα (50) and -γ (51). As shown in Fig. 3A, in the presence of PGC-1α, the mutant ERRα K14R displays a greater transcriptional activity than its wild-type counterpart at both 50 and 100 ng. In Fig. 3B, ERRα induced only a modest transcriptional response in the absence of PGC-1α. However, a more significant induction of basal LUC activity was observed with the ERRα K14R mutant. As previously reported (50, 52), introduction of PGC-1α in HeLa cells stimulated the basal activity of the ESRRA promoter due to the presence of endogenous ERRs in these cells but also stimulated the activity of exogenous ERRα. In the presence of PGC-1α, the K14R mutant displayed a marked increase in transcriptional activity as compared with the wild-type receptor. Similar results were obtained with ERRγ and the K40R mutant (Fig. 3C). The increased activity of NTD KR mutants was not caused by differences in protein expression, because wild-type and mutant constructs were expressed at similar levels for both ERRα and γ (Fig. 3, B and C, insets). To avoid interference by endogenous ERRα in the transfection assay, we performed the same experiment using ERRα-null mouse embryonic fibroblasts (MEFs) immortalized with simian virus 40 large T antigen [ERRα-null MEFs-T]. In these cells, introduction of PGC-1α had only a minor effect on the basal activity of the reporter construct (Fig. 3D) and both wild-type ERRα and the K14R mutant failed to display basal transcriptional activity on ESRRA-driven reporter. However, both proteins showed a strong transcriptional response to the presence of PGC-1α, and the K14R mutant displayed much higher transcriptional activity than the wild-type receptor. To rule out any other interference by endogenous ERR isoforms on the ESRRA-driven reporter, we also assessed the activity of Gal4 DNA-binding domain-ERRα and -K14R mutant fusion proteins on a two-copy upstream activating sequence (UAS)-tk-LUC reporter (Fig. 3E). In this context, the sumoylation-deficient ERRα K14R mutant displayed a synergistic response to the presence of PGC-1α.

Fig. 3.

Arginine Substitution at the Major Sumoylation Site of ERRα and -γ Increases Their Transcriptional Activity

A, HeLa cells were transfected with 50 or 100 ng of the indicated ERRα expression plasmid, along with the expression construct pCDNA3.1-HA-PGC-1α (250 or 500 ng, respectively) on the 3xESRRApromoter-LUC reporter (250 ng). The LUC activity has been adjusted relative to the CMVβGAL internal control. Results are presented in relative LUC units (RLU). B and C, HeLa cells were transfected with 100 ng of the indicated ERR expression plasmid and 500 ng of PGC-1α expression plasmid or pCDNA-HA along with the 3xESRRApromoter-LUC reporter. Results are presented in fold activation relative to the control condition (vector). Whole-cell lysates were analyzed by Western blot with anti-ERRα or anti-Flag antibody to determine the expression levels of the different ERR mutants. D, Same as panel B except that ERRα-null MEFs-T were used. E, COS-1 cells were transfected with 50 ng of the expression plasmid for Gal4-ERRα or Gal4-ERRα K14R mutant on a two-copy UAS-tk-LUC reporter plasmid in the presence of 250 ng of the PGC-1α or its control vector. Whole-cell lysates were analyzed by Western blot with the anti-ERRα antibody to examine the expression levels of the different ERR mutants. F, COS-1 cells were transfected with 100 ng of the indicated expression plasmid for ERRα on the synthetic 1xERRE-TK-LUC or the 3xERRE-TK-LUC reporter plasmid in the presence of 500 ng of the PGC-1α expression plasmid. G, ERRα-null MEFs-T were transfected with 100 ng of the indicated ERR expression plasmids in the presence of the PGC-1α expression plasmid or its vector. The reporter activities of the ΔESRRA-LUC, 2xESRRA-LUC, and 3xESRRApromoter-LUC reporter plasmids were compared. H, Same as panel E except that the activity of the Flag-ERRγ or Flag-ERRγ K40R expression plasmids was analyzed. I, ³⁵S-labeled in vitro translated ERRα or -γ and their respective NTD mutants were subjected to pull-down analysis with bacterially expressed GST-PGC-1α (amino acids 1–250) protein. V, Vector; WT, wild-type; KR, N-terminal ERRα K14R or ERRγ K40R mutant.

As depicted in Fig. 1C, the consensus motif for sumoylation in both ERRα and -γ also overlaps with an SC motif responsible for regulation of SC in the presence of multiple response elements (45). We therefore wanted to determine whether the SC mechanism was regulating the transcriptional activity of ERRα. As shown in Fig. 3F, the greater transcriptional activity of the ERRα K14R mutant was observable solely on the 3xERRE-TK-LUC reporter constructs, whereas on the reporter bearing only a single copy of the ERRE, the transcriptional activity was lower and similar for both ERRα and ERRα K14R. Related to this observation, we have previously demonstrated that a naturally occurring polymorphism within the ESRRA promoter changes the number of consecutive ERREs present in the distal region of the promoter from 1 to 4 (50). We next tested whether changing the sumoylation status of the ERRs affects the regulation of the polymorphic ESRRA promoter. Results presented in Fig. 3, F and G, show that, in the presence of PGC-1α, the KR NTD mutants were more potent on the reporter construct driven by the ESRRA promoter containing three copies of the ERR response elements than the reporter containing only two copies of the element. We next tested whether the mutations of lysine 14 and 40 to arginines could, by themselves, change the interaction between the coactivator PGC-1α and the receptor proteins. As shown in Fig. 3I, all four proteins interacted with PGC-1α to a similar level when assayed by glutathione-S-transferase (GST)-pulldowns using equal amounts of immobilized GST and GST-PGC-1α/1–250. Taken together, these results demonstrate that sumoylation of ERRα and -γ represses the transcriptional activity of the receptors. Therefore, the regulatory effect of ERR sumoylation would be of greater importance for the individuals expressing promoters bearing multiple copies of the ERRE.

Sumoylation Site Mutants Display Wild-Type Nuclear Localization and DNA Binding Properties

Sumoylation has been shown to affect subcellular localization of transcription factors, an effect that has been associated with repression due to sequestration within nuclear bodies (37, 53). To verify whether sumoylation of ERRα affects its cellular localization, we transfected HeLa cells with green fluorescent protein (GFP) constructs for ERRα (Fig. 4A, top row) and the K14R mutant (Fig. 4A, middle row). The nuclear localization remained unchanged between the ERRα wild-type and K14R mutant (Fig. 4A, first column). Furthermore, no targeting to nuclear bodies was observed for both ERRα variants as shown by the absence of colocalization between the HA-SUMO2-induced nuclear bodies (Fig. 4A, second and third columns). Sumoylation is also known to modulate the DNA binding properties of certain transcription factors (36, 37, 53). Thus, we assessed whether sumoylation affects the DNA binding properties of ERRα. Using nuclear extracts from HEK293 cells transfected with ERRα wild-type, K14R, or K403R mutants, we observed no change in the DNA binding pattern or in supershift generated by an ERRα antibody or a purified GST-PGC-1α/1–250 fusion protein (Fig. 4B). We also compared the DNA binding pattern of in vitro sumoylated ERRα, with both SUMO1 and SUMO3 (Fig. 4C). We observed that sumoylated ERRα binding to DNA was as efficient as that of the K14R mutant. The same experiment was then reproduced with the addition of GST-PGC-1α in the DNA binding reaction, and the result showed that the sumoylated form of ERRα was similarly supershifted by PGC-1α on DNA (Fig. 4D). Because the SC mechanism is dependent on the presence of more than one copy of the response element, we assessed the DNA binding properties of nonsumoylated and sumoylated ERRα on a tandem element probe. We observed no difference in the binding pattern (Fig. 4E) or the intensity of binding (Fig. 4F) with increasing amounts of the nonsumoylated and sumoylated ERRα.

Fig. 4.

ERRα Subcellular Localization, DNA Binding, and Interaction with PGC-1α Are Not Affected by the K14R Mutation

A, HeLa cells were transfected with the expression plasmid for GFP-tagged ERRα (top row) or GFP-tagged ERRα K14R mutant (middle row) along with the HA-SUMO2 expression plasmid. The empty GFP expression vector (bottom panel) was used as control. The corresponding GFP fluorescent images (green, first column) and anti-HA antibody staining images (red, second column) were merged together (merged, third column), with the last column representing the 4′6-diamidino-2-phenylindole staining (DAPI) (blue). B, Nuclear extracts of HEK293 cells transfected with the control vector (V) or the indicated expression plasmid for ERRα or its mutants were subjected to EMSA with the consensus ERRE probe. Ab, Anti-ERRα antibody; M, ERRα monomer; D, ERRα dimer; SS, supershift; SSAb, antibody supershift; P, GST-PGC-1α 1–250 purified protein; SSPGC-1, supershift with GST-PGC1α (1–250) purified protein. C, ³⁵S-labeled in vitro translated ERRα and K14R proteins were sumoylated in vitro with (+) or without (−) the recombinant activating E1 enzyme in the presence of recombinant SUMO1 or SUMO3 and subjected to EMSA on the ERRE consensus probe. SUMO-D, Dimer of sumoylated ERRα. Sumoylated ERRα (S-ERRα) complexes are represented by open arrowheads, and ERRα complexes are represented by solid arrowheads. D, ³⁵S-labeled in vitro translated ERRα and K14R proteins were sumoylated in vitro with (+) or without (−) the E1 enzyme with recombinant SUMO1 and were subjected to EMSA in the presence or absence of GST-PGC-1α/1–250 fusion protein. E, Increasing amounts of in vitro translated ERRα protein were sumoylated in vitro using recombinant SUMO1 with (+) or without (−) the E1 enzyme and subjected to EMSA on a ³²P-labeled tandem ERRE probe. F, Graphic representation of the total binding intensities for each lane (all bands) of the tandem probe EMSA gel (in panel E) quantified by phosphor imager.

Phosphorylation of ERRα on Serine 19 Is Essential for Sumoylation of Lysine 14

Interplay with phosphorylation events on a target protein has been shown to modulate sumoylation (54, 55, 56). Moreover, a PDSM has recently been proposed to be present in sumoylated proteins and to constitute a general mechanism for regulating sumoylation (42, 43, 44). As described above, the NTD sumoylation sites of ERRs possess this consensus motif (Fig. 1B). We then assessed the effect of ERRα S19A and S19D mutations (Fig. 5, A and B) as well as the ERRγ S45A and S45D mutations (Fig. 5C) on the transactivation properties of the receptors. Using the 3xESRRApromoter-LUC reporter construct in HeLa cells (Fig. 5, A and C) or the two-copy UAS-tk-LUC reporter in COS-1 cells (Fig. 5B), we observed that the ERRα S19A and ERRγ S45A mutants displayed increased transcriptional activity compared with their wild-type counterparts. In contrast, the phospho-mimetic mutants ERRα S19D and ERRγ S45D had similar transcriptional activity as the wild-type receptors. This result is in agreement with the sumoylation level of wild-type and mutant ERRs. HEK293 cells were transfected with the ERR point mutants along with an HA-SUMO2GG-activated form and UBC9 or with HA-SUMO2ΔGG and UBC9-C93S dominant-negative forms, and 80 μg of extracts were subjected to Western blot analysis using a anti-ERRα or to an anti-FlagM2 agarose beads immunoprecipitation and Western analysis with anti-flag antibody. We observed that the sumoylation capacity was impaired in a similar manner as the KR mutants and after alanine mutations of serine 19 for ERRα (Fig. 5D) and of serine 45 for ERRγ (Fig. 5E). Conversely, the phospho-mimetic mutations of these serines to aspartic acid residues restored the capability for sumoylation for both ERRα (Fig. 5D) and -γ (Fig. 5E). Taken together, these results support a phosphorylation-dependent sumoylation mechanism of the ERRα and -γ NTDs.

Fig. 5.

Phosphomimetic Mutants Display Elevated Sumoylation and Reduced Transcriptional Activity

A, HeLa cells were transfected with 100 ng of the expression plasmid for ERRα or the ERRα S19A and ERRα S19D mutants in the presence of 500 ng of the PGC-1α expression construct or its control vector on the 3xESRRApromoter-LUC reporter (250 ng). The LUC activity has been adjusted relative to the CMVβGAL internal control. Results are presented in fold activation relative to the control condition (vector). Whole-cell lysates were analyzed by Western blot with anti-ERRα antibody to examine the expression levels of the different ERR mutants. B, Same as panel A except that COS-1 cells were transfected with Gal4-tagged ERRα expression construct or the indicated mutant on a two-copy UAS-TK-LUC reporter. C, The ERRγ, S45A, or S45D mutant expression plasmids were transfected as in panel A. Whole-cell lysates were analyzed by Western blot with anti-Flag antibody to compare the expression levels of the different ERR mutants. D, The ERRα indicated expression plasmids were transfected as in A. E, HEK293 cells were transfected as in Fig. 2C with the expression plasmid for ERRα or the indicated mutant, and sumoylation levels were determined by Western blot analysis of whole-cell lysate (80 μg) with anti-ERRα. F, Same as in Fig. 2D with the indicated ERRγ expression plasmid except that sumoylation levels were determined by immunoprecipitation (250 μg) of whole-cell lysate on anti-flagM2 agarose followed by Western blot analysis with anti-Flag antibody. IP, Immunoprecipitation; V, vector; WB, Western blot.

PIASy Interacts with and Induces Sumoylation of ERRα

We next sought to identify the E3 ligase that generates ERRα sumoylation. Of potential SUMO E3 ligase candidates tested, we observed that PIASy was the most effective in promoting the sumoylation of ERRα in cultured cells (Fig. 6A). HEK293 cells were cotransfected with pCMV5-Flag-PIASy and myc-ERRα wild-type or K14R mutant expression plasmid. The coexpression of Flag-PIASy with ERRα resulted in covalent modification of myc-ERRα with endogenous SUMO. Considering that in vitro, both SUMO1 and SUMO3 were able to modify the ERRs, the endogenous SUMO modifier in this setting could be either one of the SUMO isoforms. After immunoprecipitation on FlagM2 agarose, PIASy was shown to interact with both the nonsumoylated and the sumoylated form of ERRα (Fig. 6B). In a similar manner, PIASy also interacted with an apparent similar affinity with the ERRα sumoylation-deficient mutant K14R (Fig. 6B). The cotransfection of flag-PIASy along with the ERRα and phospho-sumoyl switch mutants markedly stimulated the sumoylation of ERRα. In agreement with the role associated with serine 19, the S19A mutant showed no modified form in the presence of the ligase, whereas the S19D mutation restored the modification (Fig. 6C). Considering the high potential for ERRα multisite phosphorylation events suggested by large-scale studies of HeLa cells and mouse liver nuclear phosphoproteins (57, 58), mutants of the adjacent serine 22 (S22A and S22D) were also tested for sumoylation levels and showed no difference in comparison with the wild-type receptor (Fig. 6C). Interestingly, the transcriptional activity of the mutants (Figs. 5A and 6D) shows good correlation with the levels of sumoylation observed in Fig. 6C.

Fig. 6.

PIASy Enhances ERRα Sumoylation in a Phosphorylation-Dependent Manner

A, HEK293 cells were transfected with the expression constructs for Flag-PIASxα, xβ, 3, or y along with the ERRα expression plasmid. Whole-cell lysates (80 μg) were subjected to Western blot analysis with the indicated antibody. Sumoylated ERRα (S-ERRα) is indicated by open arrowheads and ERRα by solid arrowheads. B, HEK293 cells were transfected with the expression plasmid for Flag-PIASy along with the ERRα or ERRα K14R expression plasmids. Whole-cell lysates (250 μg) were subjected to immunoprecipitation on anti-FlagM2 agarose beads followed by Western blot analysis with the indicated antibody. C, HEK293 cells were transfected with the expression plasmids for Flag-PIASy along with the indicated ERRα expression plasmid. Whole-cell lysates (80 μg) were subjected to Western blot analysis with the indicated antibody. D, ERRα-null MEFs-T were transfected with the 3xESRRApromoter-LUC reporter plasmid and the indicated ERRα mutant expression plasmid in the presence or absence of PGC-1α construct. IP, Immunoprecipitation; RLU, relative LUC units; WT, wild-type.

Phosphorylation-Dependent Sumoylation of ERRα in Vivo

We next wanted to determine whether the ERRα phospho-sumoyl switch was functional in vivo. Therefore, we generated a custom-made rabbit antiserum directed against the phosphorylated S19 of ERRα. The relative affinity of the anti-ERRα and the anti-ERRα pS19 antisera was assessed by dot blot against the immobilized phosphorylated synthetic peptide antigen and its nonphosphorylated homolog. We observed that the anti-ERRα recognized this particular epitope very weakly in comparison with the anti-ERRα pS19 antisera. Also, the anti-ERRα pS19 antisera, although detecting slightly the nonphosphorylated antigen, displayed a strong preference for the immobilized phospho-peptide (data not shown).

Mouse liver extracts prepared in a phosphorylation- and sumoylation-preserving buffer were subjected to parallel immunoprecipitations using the anti-ERRα as well as the anti-ERRα pS19 antisera along with corresponding preimmune serum. Subsequent Western blot analysis with the anti-ERRα antiserum confirmed the high levels of ERRα protein in mouse liver and the effective immunoprecipitation of the nonsumoylated form of ERRα by the anti-ERRα antiserum and in a weaker manner, by the anti-ERRα pS19 antibody (Fig. 7A). In both immunoprecipitation conditions (Fig. 7A, left side with anti-ERRα and right side with anti-ERRα pS19), the higher band corresponding to sumoylated ERRα could not be detected by the anti-ERRα. This is because the anti-ERRα displays a weaker recognition of the serine 19 epitope involved in the present mechanism (Fig. 7A). To overcome this, a commercially available anti-ERRα antibody recognizing the C terminus of hERRα (amino acids 339–364) was used to validate the identity of the different bands. Western blot analysis with this antibody revealed a weak higher band after immunoprecipitation with the anti-ERRα antibody (Fig. 7B, left side). Interestingly, the higher band corresponding to sumoylated ERRα was more abundant when the immunoprecipitation was performed using the anti-ERRα pS19 antibody (Fig. 7B, right side).

Fig. 7.

Coupled Phosphorylation and Sumoylation of ERRα in Mouse Liver

A–D, 2 mg of mouse liver total extracts was subjected to immunoprecipitation with the anti-ERRα antibody (αERRα) or the anti-ERRα phospho serine 19 (αpS19) antisera followed by Western blot analysis with anti-ERRα (A), anti-ERRα (B; Upstate Biotechnology, catalog no. 07–662), anti-ERRα pS19 (C) or anti-SUMO2 (D). Sumoylated ERRα (S-ERRα) is represented by open arrowheads and ERRα by filled arrowheads. I, Input; P, preimmune serum; Ab, antibody; N-S, nonspecific; LC, IgG light chain; HC, IgG heavy chain.

Remarkably, subsequent Western blot analyses with anti-ERRα pS19 antiserum (Fig. 7C) and anti-SUMO2 (Fig. 7D) antibodies revealed only a SUMO2-modified form of ERRα when immunoprecipitation was performed with the phospho-specific antisera and not with the anti-ERRα. Also, in agreement with the sumoylation-promoting role of serine 19, Fig. 7C (right side) demonstrates that the pool of serine 19-phosphorylated ERRα is completely sumoylated, as depicted by the shift in molecular weight observed. The relative abundance of the sumoylated vs. the nonsumoylated species is usually low. The detection of endogenously sumoylated proteins is therefore difficult. The anti-ERRα pS19 antiserum also slightly detected the lower ERRα band in the input lanes (Fig. 7C). The reason for this is the slight recognition of the nonphosphorylated ERRα, in agreement with the much lower affinity for the nonphosphorylated peptide (data not shown). However, after enrichment for the serine 19-phosphorylated ERRα by immunoprecipitation with the anti-ERRα pS19, the only band detected is the sumoylated ERRα.

Furthermore, the higher molecular weight band observed in Fig. 7, B and C (right side), comigrates with the one observed when Western blot analysis was performed with the anti-SUMO2 antibody (Fig. 7D, right side). Taken together these results not only indicate that ERRα is sumoylated in vivo but demonstrate the requirement of serine 19 phosphorylation for efficient sumoylation of the NTD.

DISCUSSION

In this report, we identified ERRα and -γ as bona fide targets of sumoylation. We showed that the NTDs of ERRα and -γ contain a phosphorylation-dependent sumoylation consensus motif ψKxExxSP, also known as PDSM (42, 43, 44), that is conserved from Drosophila to humans and is also present in the ERRβ isoform. Moreover, the PDSM is embedded within a SC motif (Fig. 1B), an extended motif determinant proposed to regulate higher-order interactions among transcription factors (45). Our study demonstrated that the three overlapping motifs are functional, placing the ERR isoforms in a unique category within the superfamily of NRs.

Sumoylation is generally associated with transcriptional repression. In agreement with this observation, we have shown that mutation of the main SUMO acceptor sites in the ERRα and -γ NTDs increases their transcriptional activities. Importantly, the greater transcriptional activity of the NTD KR mutants can be observed only in the presence of PGC-1α. This reinforces the notion that ERRα is a major conduit of PGC-1α activity (5, 6, 50, 59, 60, 61). In fact, the presence of a PGC-1 coactivator family member appears to be crucial for ERRα activation and has been proposed to act as a protein ligand for the ERRs (62). The apparent dependency on PGC-1α for the effect observed in this study mirrors the importance of this cofactor in ERR function. Furthermore, because PGC-1α activates ERR function via the LBD (63), our results suggest that intramolecular interactions between the NTD and the LBD may play an important role in controlling ERR transcriptional activity.

SUMO conjugation has been shown to negatively regulate the activity of transcription factors through the regulation of different molecular properties of the target protein. These effects seem to be specific to each target protein and vary from reduction of DNA binding to alteration of protein stability or sequestration to subnuclear bodies (35, 36, 37). For the ERRs, it seems that properties such as localization, coactivator recruitment, DNA binding, and protein stability are not affected by sumoylation. Therefore, these may not account for the repressive effect of sumoylation suggested by the increased transcriptional activity of the mutated ERR proteins (Figs. 3 and 5). Instead, we showed that sumoylation regulates ERR transcriptional activity via a SC mechanism. Indeed, the increased transcriptional activity of the ERR NTD sumoylation-deficient mutants on the ESRRA promoter containing three copies of the ERR response element suggests that sumoylation of the SC motif could have a direct impact on the expression of ERRα itself. Thus, sumoylation of ERRα may be an important component for the fine tuning of the autoregulatory loop regulating ERRα expression in the presence of PGC-1α. This regulatory mechanism is also likely to influence the expression of other ERR target genes that harbor multiple ERREs in their promoter/regulatory regions.

In the absence of a known natural ligand, the regulation of ERRα and -γ transcriptional activities by posttranslational modifications becomes of crucial importance. To our knowledge, the only identified phosphorylation site within the ERRs so far is threonine 124 of ERRα, which lies within a consensus PKCδ phosphorylation site (12). In addition, ERRα has also been shown to be phosphorylated after epidermal growth factor treatment in MCF-7 cells (64) and hyperphosphorylated in BT-474 cells, a human breast cancer cell line overexpressing the oncogene ErbB2 (11). Moreover, a large-scale characterization of HeLa cell nuclear phosphoproteins confirmed the phosphorylation status of the endogenous ERRα, identifying serine 19 as one of multiple phosphorylated residues within the NTD by tandem mass spectroscopy (57). Using phosphorylation-mimicking mutants of ERRα and -γ, we have provided evidence for the importance of the phosphorylation status of serines 19 and 45 for sumoylation of the ERRα lysine 14 and ERRγ lysine 40, respectively (Fig. 6). The negative charges created by the phosphorylation events close to the SUMO acceptor site can be compared with the recently identified negatively charged amino acid-dependent sumoylation motif (NDSM). The NDSM mechanism relies on the presence of negatively charged amino acids in close proximity to the sumoylation site to enhance the sumoylation. The ERR isoforms also possess such an acidic patch, although it is located outside of the limit of the identified NDSM (48). The main difference with the ERRs occurs in the possibilities offered by a regulated phosphorylation event as opposed to the fixed enhancement provided by the presence of a negatively charged glutamate residue.

Our data, together with other recent reports showing phosphorylation-dependent sumoylation of MEF2 family members (43, 65, 66, 67), HSF-1 (54), and PPARγ (55) strengthen this model as a common signaling-dependent regulating mechanism for sumoylation of transcription factors and demonstrate the functionality of the phospho-sumoyl switch in vivo. Furthermore, the observation that PPARγ is the only member of the NR family sharing this PDSM with the three ERR isoforms also suggests a potential link between phosphorylation-dependent sumoylation and metabolic control by NRs. As both ERRα and -γ have emerged as essential regulators of energy metabolism (13, 52, 59, 60, 68, 69, 70, 71, 72, 73, 74, 75), the identification of the signaling pathways regulating ERR sumoylation and the study of how phosphorylation-dependent sumoylation specifically affects the expression of metabolic genes will be of importance for our understanding of pathologies such as cardiovascular diseases, diabetes, and obesity.

MATERIALS AND METHODS

Plasmids and Constructs

The ΔESRRApromoter-LUC, 2xESRRApromoter-LUC, and the 3xESRRApromoter-LUC reporter constructs as well as the pCMX-ERRα expression plasmid have been previously described (50). The expression vectors for wild-type SUMO2 and wild-type UBC9 were previously described (49, 76). The shorter forms of SUMO2, one terminated with the diglycine motif (SUMO2GG) acting as a constitutively activated protein as well as one nonconjugatable form terminated before the diglycine motif (SUMO2ΔGG) acting as a dominant negative, were generated by PCR and the purified products were cloned into pcDNA3.1 derivatives (Invitrogen, Carlsbad, CA). Point mutants of ERRα (K14R, K403R, S19A, S19D, S22A, S22D), ERRγ (K40R, K360R, K439R, S45A, S45D), and UBC9 (C93S) were made by site-directed mutagenesis. The DNA fragments were sequenced and subcloned into pCMX-myc for ERRα constructs, pCMX-Flag for ERRγ constructs, and pcDNA-HA for UBC9 (C93S). The wild-type and mutant versions of ERRα and -γ were also subcloned into pCMX-Gal4 and pEGFP-C1 (CLONTECH Laboratories, Inc., Palo Alto, CA). The expression vector pcDNA3/HA-hPGC-1α was provided by A. Kralli (77). The plasmids pGEX2T-PGC-1α/1–250 and pCMV5-Flag-PIASy, -PIASxα, -PIASxβ, and -PIAS3 were described previously (64, 78, 79).

Cell Culture and Transient Transfections

COS-1, HeLa, and HEK293 cells were cultured in DMEM (Invitrogen) supplemented with 10% fetal bovine serum. Cells were plated in 12-well or 10-cm plates 16–18 h before transfection with Fugene 6 (Roche Diagnostics, Mannheim, Germany). MEFs were freshly isolated as previously described (80) using ERRα-null mice embryos (81) and were immortalized by inducing a stable expression of the simian virus 40 large T antigen (ERRα-null MEF-T).

Reporter Gene Assays

Cells were harvested on ice 48 h after transfection for determination of LUC and β-galactosidase activities. Each transfection was performed in duplicate at least three times.

Fluorescence Microscopy

HeLa cells were plated on glass coverslips 16 h before transfection with GFP and HA-SUMO2 expression constructs for 48 h. Cells were washed three times with PBS, fixed, and permeabilized on ice with 2% paraformaldehyde-0.2% Triton X-100 in PBS for 15 min. The cells were then washed three times in PBS, quenched with 50 mm NH₄Cl in 1× PBS for 10 min, rinsed twice with PBS, and blocked with 5% BSA in PBS for 2 h at room temperature. After a 30-min incubation at room temperature with α-HA monoclonal antibody (Roche), the cells were washed three times with 1× PBS and then incubated with α-mouse Alexa555 (Molecular Probes, Inc., Eugene, OR) for 30 min at room temperature and rinsed three times with PBS. The cells were then subjected to staining with 4′6-diamidino-2-phenylindole (Sigma Chemical Co., St. Louis, MO) rinsed again, and mounted on glass slides with Immu-Mount (Thermo Fisher Scientific, Inc., Waltham, MA). Cells were analyzed under a Zeiss epifluorescence confocal microscope (Carl Zeiss, Thornwood, NY).

GST Pull-Down Assay

Equal amounts of bacterially expressed GST or GST-PGC-1α/1–250 protein containing the NR interaction motifs immobilized on glutathione sepharose beads were combined with 10 μl of ³⁵S-labeled ERRα as well as ERRγ and NTD KR mutant proteins produced with the TNT T7 coupled reticulocyte lysate system (Promega Corp., Madison, WI) in 500 μl of GST binding buffer (20 mm Tris, pH 7.5; 100 mm KCl; 0.1 mm EDTA; 0.05% Nonidet P-40; 10% glycerol; 1 mg/ml BSA; 1 mm phenylmethylsulfonyl fluoride; protease inhibitor tablet complete mini (Roche) for 1 h at 4 C. The beads were washed five times with cold binding buffer, and the immobilized proteins were eluted by boiling in 2× sample buffer. The eluted proteins were resolved on SDS-PAGE, and the fixed and dried gels were visualized by autoradiography.

In Vitro Sumoylation Assay

[³⁵S]-myc-ERRα, K14R, and K403R and [³⁵S]-flag-ERRγ, K40R, K360R, and K439R proteins were produced using the TNT T7-coupled reticulocyte lysate system (Promega) and subjected to in vitro sumoylation reactions with E1 and E2 purified enzymes along with SUMO1 or SUMO3 purified proteins (LAE BIO, Rockville, MD). Briefly, 2 μl of [³⁵S]-myc-ERRα, [³⁵S]-flag-ERRγ, and KR variant in in vitro translation reactions were combined with 150 ng of purified human SAE1/SAE2 (E1), 1 μg of purified human UBC9, and 1 μg of purified human SUMO1 or SUMO3 proteins, and then incubated in a sumoylation buffer containing 50 mm Tris-HCl (pH 7.5), 5 mm MgCl₂, 1 mm dithiothreitol, and 2.5 mm ATP at 37 C for 1 h. The control reaction was performed under the same conditions but the purified E1 enzyme was omitted to prevent sumoylation from occurring. Reactions were stopped by boiling in reducing sodium dodecyl sulfate sample buffer for separation by SDS-PAGE and detection by autoradiography.

In Cells and in Vivo Sumoylation Assay

HEK293 cells were transfected with the specified ERRα or -γ constructs along with the activated HA-SUMO2GG and UBC9 to favor the sumoylation of targets or with the dominant-negative forms HA-SUMO2ΔGG and UBC9 (C93S) to inhibit sumoylation. About 48 h after transfection, cells were lysed in buffer S (15 mm Tris-HCl, pH 6.7; 0.5% sodium dodecyl sulfate; 3% glycerol, 0.8× PBS; 4% Nonidet P-40; 0.1% β-mercaptoethanol) containing 25 mm N-ethylmaleimide, 20 mm iodoacetamide, 1 mm phenylmethylsulfonyl fluoride, 1 mm sodium orthovanadate, 2.5 mm sodium pyrophosphate, 50 mm sodium fluoride, and 1× complete miniprotease inhibitor tablet (Roche). The lysates were sonicated at power 5 for 15 sec using a VirSonic 100 (VirTis, Gardiner, NY) sonicator. Protein concentration was determined by Bradford assay and 80 μg of lysates were used for Western blot analysis or 250 μg of lysates were used for immunoprecipitation with a monoclonal anti-myc antibody (Roche) for ERRα constructs and an anti-flag M2 resin (Sigma) for ERRγ constructs. For endogenous ERRα sumoylation, 2-month-old male C57/BL6 mouse liver extract was prepared in buffer K (20 mm phosphate buffer, pH 7; 150 mm NaCl; 0.1% Nonidet P-40; 5 mm EDTA) containing 25 mm N-ethylmaleimide, 20 mm iodoacetamide, 1 mm phenylmethylsulfonyl fluoride, 1 mm sodium orthovanadate, 2.5 mm sodium pyrophosphate, 50 mm sodium fluoride, 25 mm β-glycerol phosphate, and 1× complete miniprotease inhibitor tablet (Roche). Briefly, mouse livers were homogenized in buffer K for 5 sec with a polytron-type homogenizer. The crude lysate was incubated for 1 h at 4 C with rotation followed by centrifugation at 13,000 rpm for 5 min at 4 C. The supernatant was then quantified by Bradford assay and 2 mg of whole-cell liver extract was used for immunoprecipitation with a previously described anti-hERRα polyclonal antiserum raised in our laboratory against the whole N terminus (first 74 amino acids) (73) or with an anti-ERRα pS19 (described below). The anti-SUMO2/3 antibody (no. AB3876) was obtained from Chemicon International (Temecula, CA).

EMSA

EMSAs were performed as previously described (82) using the consensus ERR response element (ERRE) probe (5′-TCGACGCTTTCAAGGTCATATCCG-3′) and a tandem probe containing two ERRE elements from the ERRα promoter endogenous sequence (5′-CCGTGACCTTCATTCGGTCACCGCAGTGACCTTCAT-3′). The ERRE sequences are underlined. HEK293 cells were transfected with 10 μg of pCMX-myc-ERRα, K14R, or K403R expression vectors for 48 h after which nuclear extracts were prepared as previously described (83). About 2 μg of extract was used per EMSA reaction. For PGC-1α supershift experiments, 2 μg of nuclear extracts was mixed with 2 μg of purified GST-PGC-1α/1–250. EMSA with in vitro sumoylated proteins was performed as described above for in vitro sumoylation after which the complete reaction (20 μl) or increasing amounts of the reaction (5, 10, 20 μl) were used for the EMSA reaction.

Generation of the pS19 Phospho-Specific Antibody

The anti-ERRα pS19 rabbit antiserum was custom generated by Chemicon International against the phosphorylated Ser19 peptide (CPLYIKAEPApSPD) conjugated to keyhole limpet hemocyanin. To assess specificity of the antisera for the phosphorylated epitope, a dot blot analysis was performed by spotting 0.2 to 1 μg of synthetic phosphorylated and the correponding nonphosphorylated peptides on nitrocellulose membranes followed by Western blot analysis using the previously described general anti-ERRα antiserum and the new anti-ERRα pS19 phospho-antiserum.

Statistics

One-way ANOVA followed by Bonferonni post-test analysis were performed using GraphPad InStat software (GraphPad Software, Inc., San Diego, CA). Where indicated, *** is P < 0.0001 and ** is P < 0.001.

NURSA Molecule Pages:

• Coregulators: ARIP3 | PGC-1 | PIAS3 | PIAS4;

• Nuclear Receptors: ERRα | ERRγ.