A Tissue-Specific Coactivator of Steroid Receptors, Identified in a Functional Genetic Screen

ABSTRACT Steroid receptors mediate responses to lipophilic hormones in a tissue- and ligand-specific manner. To identify nonreceptor proteins that confer specificity or regulate steroid signaling, we screened a human cDNA library in a steroid-responsive yeast strain. One of the identified cDNAs, isolated in the screen as ligand effect modulator 6, showed no homology to yeast or Caenorhabditis elegansproteins but high similarity to the recently described mouse coactivator PGC-1 and was accordingly termed hPGC-1. The hPGC-1 DNA encodes a nuclear protein that is expressed in a tissue-specific manner and carries novel motifs for transcriptional regulators. The expression of hPGC-1 in mammalian cells enhanced potently the transcriptional response to several steroids in a receptor-specific manner. hPGC-1-mediated enhancement required the receptor hormone-binding domain and was dependent on agonist ligands. Functional analysis of hPGC-1 revealed two domains that interact with steroid receptors in a hormone-dependent manner, a potent transcriptional activation function, and a putative dimerization domain. Our findings suggest a regulatory function for hPGC-1 as a tissue-specific coactivator for a subset of nuclear receptors.

Steroid hormones play important roles in development, growth, glucose and mineral homeostasis, stress responses, sexual differentiation, and reproduction. The effects of steroids are mediated by intracellular receptors that, together with receptors for thyroid hormones, retinoids, vitamin D, and other small lipophilic molecules, belong to the superfamily of nuclear receptors (2,43). Binding of hormone to these receptors triggers a conformational change that leads to the release of associated proteins, such as molecular chaperones or corepressors, recruitment of new proteins, such as coactivators, binding to specific DNA sites termed hormone response elements (HREs), and regulation of transcription from promoters in the vicinity of the HREs (reviewed in references 2, 47, 61, and 75).
The major determinant of the ability of a cell to respond to a specific steroid hormone is the presence of the cognate receptor. In addition, nonreceptor proteins contribute to the cellular response by enabling or regulating distinct steps in the hormone response pathway. For example, membrane proteins that regulate the transport of hormone across the plasma membrane modify the availability of hormone to the intracellular receptors (32,63); chaperones such as Hsp90 and p23 interact with steroid receptors in the absence of hormone and support a receptor conformation competent for hormone binding (reviewed in reference 61); DNA-binding proteins such as HMG-1 enhance the ability of steroid receptors to bind DNA (3,57,80); and the chromatin-remodeling SWI-SNF complex enables steroid receptors to regulate transcription (6,14,49,78). Finally, several nuclear proteins are recruited by the receptors in a hormone-dependent manner and are thought to mediate their transcriptional regulatory activity (reviewed in references 13, 47, and 75). Among them, corepressors such as SMRT and NcoR bind to receptors in the absence of hormone or the presence of an antagonist ligand, connect the receptors to histone deacetylases, and promote the silencing of neighboring promoters (9,21,23,35,51). In contrast, coactivators (e.g., SRC-1 [also called NcoA1], TIF2 [the human homologue of mouse GRIP1, also known as NcoA2], pCIP [also known as AIB1, ACTR, and RAC3], CREB-binding protein [CBP], p300, pCAF, and others) are recruited by hormone-activated receptors and enhance transcription (13,69,75). At least some of these coactivators are acetyltransferases that can modify histones or other target proteins and may increase promoter access to DNA-binding proteins and the transcriptional machinery (1,8,56,68,77).
The identification of nonreceptor proteins that participate in hormone signaling provides insights not just into the mechanism of receptor-mediated transcription but also into ways in which the response to hormone can be regulated. Responses to steroid hormones are often tissue specific and sensitive to other signaling pathways. For example, the activation of protein kinase A modulates glucocorticoid responsiveness in a cell type-dependent manner (48,55). Proteins that regulate the hormone response may confer tissue specificity to ubiquitously expressed receptors, if present in only some cell types, or integrate signaling information, if targeted by other signaling pathways. However, among the known nonreceptor proteins, few of them are expressed in a tissue-specific manner (e.g., ACTR) (8) or respond to other signals (e.g., CBP) (7,41,52).
Functional genetic strategies in easily manipulatable systems are powerful tools for identifying modulators that act at any step of a regulatory circuit. Although steroid receptors are not naturally present in the yeast Saccharomyces cerevisiae, the basic machinery of this organism is permissive to their function (46,65). We have taken advantage of the ability of steroid receptors to mediate hormone-dependent transcription in yeast to identify proteins that can regulate hormone responsiveness. In previous genetic screens, we isolated yeast mutants and identified genes (LEM1 to LEM4, for ligand effect modulator) that negatively regulate responses to hormones (31; R. Sitcheran, R. Emter, A. Kralli, and K. R. Yamamoto, submitted for publication). In this study, we use a genetic scheme to identify directly mammalian proteins that enhance responses to glucocorticoid hormones. In principle, the screen can reveal both conserved proteins, whose yeast homologues are limiting or do not optimally interact with the mammalian receptor, and were treated with Klenow polymerase and appropriate deoxynucleoside triphosphates and/or mung bean nuclease so as to preserve the correct reading frame and ligated. Similarly, internal deletions were made by digesting pBS/HA-hPGC-1 with NheI and EcoRI for hPGC-1⌬1 (⌬91-186), EcoRI and AgeI for hPGC-1⌬2L (⌬189-482), EcoRI and StuI for hPGC-1⌬2 (⌬189-293), and NheI and StuI for hPGC-1⌬3 (⌬91-293), treating the DNA ends with Klenow polymerase and/or mung bean nuclease, and ligating. All junctions were sequenced.
The full-length or variant HA-hPGC-1 constructs were then subcloned as BamHI-NotI fragments into the pcDNA3 vector (Invitrogen) downstream of the cytomegalovirus enhancer, for mammalian expression; as NcoI-SalI fragments into the pACT2 vector (Clontech) downstream of and in frame to the Gal4 activation domain (AD) for the expression of AD-hPGC-1 fusions in yeast; and as NcoI-SalI fragments into the pAS2-1 vector (Clontech) downstream of and in frame to the Gal4 DNA-binding domain (DBD), for the expression of Gal4-hPGC-1 fusions in yeast. For mammalian expression of the chimeric protein Gal4-hPGC-1, the NdeI-HindIII fragment of pBS/HA-hPGC-1 (lacking the HA epitope sequence and the last three codons of hPGC-1) was subcloned into vector pA4.7 (gift of Patrick Matthias) downstream of and in frame to the Gal4 DBD and upstream of and in frame to a C-terminal HA epitope sequence.
Northern analysis. A human multiple-tissue Northern blot (Clontech) was hybridized with either a radioactively labeled 0.66-kb NdeI-PstI fragment of hPGC-1 cDNA or a ␤-actin-specific probe (Clontech), as recommended by the manufacturer, and exposed on X-ray film.
Cell culture, transient transfections, and reporter gene assays. COS7 cells were cultured in Dulbecco's modified Eagle's medium supplemented with 9% charcoal-stripped fetal bovine serum. For experiments with the AR or the ER, media lacking phenol red were used. Cells from subconfluent 10-cm plates were diluted 8-to 10-fold and seeded in six-well plates 5 h prior to transfection by calcium phosphate precipitation (25). All transfections included 0.2 g of p6RlacZ (59) to normalize for transfection efficiency. Expression vectors for receptors or other DNA-binding transcription factors were added at either 2 g (AR and MR), 1 g (GR, ER, p50, p65, and Gal4-hPGC-1), or 50 ng (Gal4-VP16) of DNA. Reporter plasmids were added at either 2 g (for MR, p50, and p65) or 1 g (for GR, ER, AR, Gal4-VP16, and Gal4-hPGC-1) of DNA. For the coactivators, 1 g of hPGC-1 (or equivalent hPGC-1 variant) or 2 g of SRC-1e expression plasmids was used, unless indicated otherwise (1 g of SRC-1e DNA gave weaker enhancement of steroid receptor activity than 2 g). After overnight exposure to the DNA-calcium phosphate precipitate, cells were washed with phosphate-buffered saline (PBS) lacking calcium and magnesium and incubated for an additional 24 h in fresh media containing either hormone or vehicle (0.1% ethanol). Cells were then washed with PBS and lysed in 100 l of reporter lysis buffer (Promega). Lysates were cleared by centrifugation and assayed for luciferase activity with the Luciferase Assay System (Promega) and ␤-gal activity with the substrate chlorophenol red-␤-D-galactopyranoside (25). Luciferase values normalized to the ␤-gal activity in the extracts are referred to as luciferase units.
GFP detection and immunofluorescence. COS7 cells were seeded onto 18-mm coverslips and transiently transfected with hPGC-1 expression vectors by use of either the calcium phosphate precipitation method (for green fluorescent protein [GFP]-hPGC-1) or Effectene reagent (for HA-hPGC-1 constructs; used as recommended by the manufacturer, Qiagen). After 48 h, cells were washed with PBS, fixed for 15 min with 3% paraformaldehyde at room temperature, and permeabilized for 10 min with PBS containing 0.1% Triton X-100. Cells were then incubated for 10 min with 1% bovine serum albumin in PBS to block nonspecific binding. For immunodetection, a monoclonal antibody against the HA epitope (HA-11; BabCo) was used at a 1:1,000 dilution in PBS containing 0.5% bovine serum albumin, followed by a rhodamine-conjugated goat antimouse antibody (Jackson Immunoresearch Laboratories) at 1:100 in PBS.
Nucleotide sequence accession number. The hPGC-1 sequence has been submitted to the GenBank database under accession no. AF186379.

Mammalian modulators of glucocorticoid signaling can be functionally isolated in yeast.
To identify proteins that enhance the cellular response to hormone, we exploited a yeast strain whose growth is dependent on glucocorticoid signaling. For this, the rat GR was expressed in yeast carrying two reporters under the control of GREs: the endogenous HIS3 and the bacterial ␤-gal-encoding gene (Fig. 1A). In the absence of hormone, the lack of HIS3 expression precludes growth in selective media. The addition of hormone induces HIS3 and the ␤-gal gene and restores growth. At suboptimal hormone concentrations insufficient to support growth, the expression of mammalian proteins that enhance the hormone response will activate HIS3, enable growth in selective media, and induce ␤-gal production (Fig. 1A).
Yeast was transformed with a human hepatoma-derived (HepG2) cDNA library and selected for growth in media containing either 4 M dexamethasone or 0.5 M corticosterone, i.e., hormone concentrations that did not allow growth of the parental strain. Fifteen yeast transformants grew under selec-tive conditions and showed an increased ␤-gal response to hormone, in a manner dependent on the presence of the mammalian cDNA library plasmid (see Materials and Methods). Rescue of the plasmids and sequencing of the inserts revealed three types of cDNAs that enhanced the response to hormone in yeast (Fig. 1B). The first cDNA encoded the molecular chaperone p23, a known component of the glucocorticoid aporeceptor complex (61), thereby establishing that the screen can yield factors involved in mammalian glucocorticoid signaling. The other two cDNAs encoded novel proteins that we initially named LEM5 and LEM6. This study focuses on LEM6. The sequence of the LEM6 HepG2 library isolate (LEM6 H in Fig.  1B) predicted the expression of a truncated protein. We therefore isolated additional LEM6 clones from human liver cDNA and constructed a full-length yeast expression plasmid (see Materials and Methods). As shown in Fig. 1B, full-length LEM6 also enhanced the receptor-mediated response to hormone, albeit to a lesser extent.
LEM6 is a tissue-specific nuclear protein and the human homologue of murine PGC-1. The full-length LEM6 cDNA sequence predicted a 798-aa protein that shares high identity (95%) with the mouse coactivator PGC-1 (62), suggesting that it is the human homologue of PGC-1. We therefore refer to LEM6 as hPGC-1. Like its mouse homologue, hPGC-1 is characterized by an N-terminal region rich in acidic amino acids (26.4% of aa 1 to 140), followed by a putative nuclear receptor interaction motif (LKKLL) at residues 144 to 148 and a predominantly basic stretch of amino acids at residues 168 to 207 ( Fig. 2A). At the C terminus, there are two serine-and arginine (SR)-rich stretches (aa 566 to 599 and aa 621 to 631) and a putative RNA-binding domain (aa 677 to 753). Putative nuclear localization signals at residues 326 to 333, 627 to 633, and 651 to 667 predicted a nuclear protein. Indeed, fusion of hPGC-1 to GFP revealed exclusively nuclear fluorescence in transiently transfected mammalian cells (Fig. 2B).
Northern blot analysis of human tissues indicated that hPGC-1 is expressed in a tissue-specific manner. We detected at least two hPGC-1 mRNA transcripts (ϳ7.5 and 6 kb), predominantly in heart, skeletal muscle, kidney, and liver ( Fig.  2C). Upon longer exposures, low levels of hPGC-1 expression could be seen first in brain and lung and then in small intestine, colon, and thymus. No expression was detectable in spleen, placenta, and peripheral blood leukocytes.
hPGC-1 is a potent activator of receptor-mediated transcription in mammalian cells. We isolated hPGC-1 based on its ability to enhance receptor-mediated transcription in yeast cells. To determine whether it displays similar activity in mammalian cells, we transiently transfected COS7 fibroblasts with hPGC-1 and GR expression plasmids and a GR-responsive luciferase reporter. In the absence of exogenous hPGC-1, the addition of hormones such as corticosterone and dexamethasone resulted in 50-to 60-fold induction of luciferase expression (Fig. 3A). Expression of hPGC-1 enhanced the hormoneinduced transcription by another 20-to 60-fold, in a dosedependent manner with respect to the amount of hPGC-1 plasmid transfected (Fig. 3A). Similar enhancement profiles were seen for cells treated with corticosterone, dexamethasone, or deoxycorticosterone ( Fig. 3A and data not shown). Enhancement of transcription was strictly hormone dependent, as hPGC-1 expression did not increase luciferase expression in the absence of hormone (Fig. 3B).
The hormone requirement for the hPGC-1 effect on the glucocorticoid response could reflect the need for an agonistinduced conformation of the receptor or merely the translocation of the receptor to the nucleus and its binding to DNA. To distinguish between these possibilities, we tested whether FIG. 1. A yeast screen can identify mammalian proteins that enhance steroid hormone signaling. (A) Schematic illustration of the yeast strain and strategy. The yeast strain YNK441 expresses GR constitutively and has two reporter genes, encoding His3 and ␤-gal, under the control of GREs. The expression of mammalian proteins that increase the response to hormone enables the activation of the pathway in the presence of low hormone concentrations, thereby allowing for growth and ␤-gal expression in selective media lacking histidine (Ϫhis) and containing the His3 inhibitor 3-aminotriazole (ϩ3AT). The starshaped forms indicate some of the possible interaction points for the mammalian modulators. (B) Mammalian p23, LEM5, and LEM6 (hPGC-1) enhance the response to corticosterone. Yeast strain YNK441 carrying a vector alone (control) or expression plasmids for the indicated cDNAs was incubated overnight with 0, 1, or 10 M corticosterone and assayed for ␤-gal activity. Data represent the mean Ϯ standard deviation of results from six independent yeast transformants. LEMG H , HepG2 library isolate; LEM6/hPGC-1, full-length cDNA.
hPGC-1 could enhance the transcriptional response to RU486. RU486 is an antagonist that displays partial agonist activity, depending on cell context, and induces a conformational change of GR distinct from that induced by pure agonists (19,54). Treatment of transfected COS7 cells with RU486 led to the nuclear translocation of GR (data not shown) and a fourto fivefold activation of luciferase reporter expression (Fig.  3B). However, this modest induction was not further increased by hPGC-1, suggesting that hPGC-1 enhancement of the glucocorticoid response requires a specific, agonist-induced conformation of the receptor.
Specificity of hPGC-1 coactivator function. We next examined whether hPGC-1 could enhance hormone responsiveness for other steroid receptors. As shown in Fig. 4A, hPGC-1 expression enhanced the hormone-dependent transcription mediated by ER and MR strongly (by ϳ13-and ϳ38-fold, respectively) and that by AR only weakly (ϳ2-fold). The very weak effect of hPGC-1 on AR was confirmed in three different promoter contexts, the MMTV long terminal repeat response element (MMTV in Fig. 4A), the tyrosine aminotransferase HREs, and the probasin promoter (maximal enhancement, twofold; data not shown). AR activity was responsive to the effects of another coactivator, SRC-1e (27,58), indicating that the lack of enhancement by hPGC-1 was not due to a general inability for increased transcription (Fig. 4A). Rather, a comparison of the coactivation conferred by SRC-1e and hPGC-1 on the different receptors and promoters tested suggests that the two coactivators act in a receptor-and promoter contextdependent manner. hPGC-1 was weaker than or equal to SRC-1e in certain contexts (AR on MMTV and TAT3; GR on MMTV) and considerably stronger in others (ER; MR and GR on TAT3) ( Fig. 4A and data not shown).
To address whether hPGC-1 is a general activator of transcription, we tested its effect on two steroid-independent activators of transcription. The expression of the NF-B subunits p50 and p65 (RelA) in COS7 cells resulted in a 40-fold induction of a luciferase reporter under the control of three NF-B sites. hPGC-1 coexpression did not affect this induction (Fig.  4B). Similarly, hPGC-1 expression had no effect on the transcriptional activity of the chimeric activator Gal4-VP16 (Fig.  4B), suggesting that hPGC-1 is not a general coactivator of transcription.
hPGC-1 acts via the LBD of GR. Steroid receptors are modular proteins consisting of an N-terminal domain that carries a strong transcriptional activation function (AF1), a central domain that binds DNA, and a C-terminal domain (LBD) that binds hormone and carries a hormone-dependent transcriptional activation function (AF2). hPGC-1 could increase the response to hormone by interacting with a particular receptor domain and enhancing one or more of its functions, e.g., binding to DNA or activation of transcription. To delineate the affected receptor function, we measured the activity of truncated or chimeric variants of GR in the absence or presence of hPGC-1. As shown earlier, hPGC-1 stimulated efficiently the hormone-dependent induction mediated by full-length GR (N795 in Fig. 5). A C-terminal truncation of the receptor that removes the LBD and gives rise to a constitutively active transcription factor (18) eliminated the ability of hPGC-1 to enhance transcription (N525 in Fig. 5). In contrast, the receptor variant 407C, which lacks the N-terminal region but contains the DBD and the LBD, was still responsive to hPGC-1 (Fig. 5). Finally, the hormone-dependent activity of the chimeric Gal4-LBD protein, which contains the receptor LBD fused to the heterologous DBD of Gal4, was enhanced strongly by hPGC-1 (Fig. 5). The Gal4 DBD itself was unaffected (Fig. 4B). In conclusion, the LBD of the receptor is both essential and sufficient for a functional interaction with hPGC-1, suggesting that hPGC-1 is a coactivator of AF2, the hormone-dependent transcriptional activation function of the LBD.
The N terminus of hPGC-1 carries a potent transcriptional AD. One of the hallmarks of transcriptional coactivators is that their recruitment to a specific promoter, either by proteinprotein interactions with DNA-binding transcription factors such as GR or artificially by direct fusion to a DBD, activates transcription. This notion implies that coactivators contain transcriptional ADs. To determine whether hPGC-1 carries such a domain, we examined the ability of the chimeric Gal4-hPGC-1 protein, in which the coactivator is fused to the Gal4 DBD, to activate reporters under the control of Gal4-binding sites. As shown in Fig. 6, Gal4-hPGC-1 was a potent activator of transcription in both yeast and mammalian cells, indicating that hPGC-1 indeed harbors a transcriptional activation function.
To map the AD in hPGC-1, we tested the ability of truncated versions of hPGC-1 fused to Gal4 to activate transcription. Removal of the hPGC-1 C terminus did not reduce activity, suggesting that the AD is within the first 293 aa (N408 and N293 in Fig. 6B). Indeed, a deletion of the first 188 aa (189C) or 293 aa (294C) eliminated the ability to activate transcription ( Fig. 6B and data not shown). Most of the activation function was in fact lost when just the first 90 aa were removed (91C). Since these constructs were expressed and functional in other assays (e.g., see Fig. 9 and data not shown), we concluded that the hPGC-1 AD lies in the N-terminal region, coinciding with the acidic amino acid stretch ( Fig. 2A). We were unable to examine Gal4 fusions to smaller parts of the hPGC-1 N terminus, such as aa 1 to 90 or aa 1 to 186, because they were toxic. The toxicity of these constructs is consistent with a potent AD in this region, as similar toxicity has been observed with other strong transcriptional activators (15,16). hPGC-1 and GR interact physically in a hormone-dependent manner. The ability of hPGC-1 to enhance receptormediated transcription could be the result of its direct, hormone-dependent interaction with the receptor, leading to the recruitment of its strong transcriptional AD. To determine if the receptor and hPGC-1 interact physically, we used the yeast two-hybrid system, where the interaction between two proteins, one fused to a DBD and another fused to an AD, leads to the expression of ␤-gal (12). Yeast carrying a Gal4-responsive ␤-gal reporter was transformed with two vectors: one expressing the Gal4-LBD, shown above to be sufficient for the functional interaction with hPGC-1 (Fig. 5), and another expressing the Gal4 AD, either alone or fused to hPGC-1 (AD and AD-hPGC-1, respectively). As the transcriptional activation function of the receptor LBD is weak in yeast, the Gal4-LBD chimera was unable by itself to activate the ␤-gal reporter, even at high hormone concentrations ( Fig. 7 and data not shown). In contrast, coexpression of Gal4-LBD and AD-hPGC-1 caused a strong induction of ␤-gal activity in the presence but not in the absence of hormone (Fig. 7), suggesting that the GR LBD and hPGC-1 interact physically in a hormone-dependent manner.
To identify the domain(s) of hPGC-1 that mediates the interaction with the receptor, we tested fusions of the Gal4 AD to different parts of hPGC-1 in the two-hybrid assay. Deletion of large parts of the C terminus of hPGC-1 (as in N408 and N293) did not affect the interaction, suggesting that the interaction domain is within the first 293 aa (Fig. 7). Indeed, a deletion in this region (294C) gave rise to an otherwise functional protein (see Fig. 9) that did not interact with the receptor (Fig. 7). Further analysis of this region indicated the presence of two domains in hPGC-1 that can interact with the receptor; they were termed NID1 (aa 91 to 186) and NID2 (aa 189 to 293) (for nuclear receptor interaction domains [NIDs] 1 and 2, respectively). Constructs that have just one of the two NIDs, either NID1 (hPGC-1⌬2 and the minimal domain 91/ 186) or NID2 (189C, hPGC-1⌬1, and the minimal construct 189/293) showed a hormone-dependent interaction with the receptor (Fig. 7). In general, interactions mediated by NID1 were stronger (Ͼ250 ␤-gal units) than those mediated by NID2 (35 to 85 ␤-gal units). Deletion of both domains, as in hPGC-1⌬3, resulted in an hPGC-1 variant that was functional in other assays but could no longer interact with the receptor (Fig. 7 and data not shown).
The interactions between hPGC-1 and GR were dependent on the presence of corticosterone. To test whether the hormone dependence reflected a hormone-induced conformational change that enabled the interaction, we again tested the effect of the antagonist-partial agonist RU486. RU486 was unable to promote the interaction of GR with hPGC-1 (Fig. 7). Moreover, RU486 could compete with corticosterone for the hormone-induced interaction, i.e., acted as an antagonist (data not shown), demonstrating that both RU486 and corticosterone were able to bind the receptor but that only the full agonist corticosterone could promote a conformation competent to interact with hPGC-1.
The hPGC-1 AD and NIDs are essential for the coactivation of GR. The functions encoded by hPGC-1 suggest that the mechanism by which it enhances the response to hormone involves first its recruitment by hormone-activated steroid receptors via the identified NIDs and second the enhancement of transcription via the identified AD. If this is the case, we would expect these domains to be essential for the hPGC-1 coactivator function. To test this notion, we assayed the ability of truncation and deletion variants of hPGC-1 to enhance hormone-dependent, receptor-mediated transcription in mammalian cells.
First, we addressed the importance of the AD. As shown before, full-length hPGC-1 enhanced strongly (35-fold) the transcriptional response to hormone (Fig. 8A). An hPGC-1 variant that lacked most of the AD (91C) could no longer enhance GR activity, indicating that the AD is essential for the coactivation function (Fig. 8A). The 91C protein was expressed at levels comparable to those of full-length hPGC-1, as deter- mined by Western blot analysis (data not shown), and properly localized to the nucleus (Fig. 8C). A similar lack of activity was also seen with hPGC-1 constructs that lacked larger parts of the N terminus, such as 189C (Fig. 8A). Next, we determined the role of the two NIDs. Variants of hPGC-1 that lacked just NID1 (hPGC-1⌬1) or a region that includes NID2 (hPGC-1⌬2L) displayed reduced activity but were still able to enhance transcription by 14-to 18-fold (Fig. 8A). In contrast, deletion of both NIDs (hPGC-1⌬3) eliminated hPGC-1 activity (Fig.  8A). Again, constructs hPGC-1⌬1, hPGC-1⌬2L, and hPGC-1⌬3 expressed proteins that were properly localized to the nucleus, suggesting that the decrease in activity was due to the deletion of the interaction domains and not to changes in protein stability or localization (Fig. 8C and data not shown). In conclusion, our results indicate that both the AD and the ability to interact physically with the receptor are essential for the hPGC-1 coactivation function. While either NID in the context of full-length hPGC-1 is sufficient, the presence of both domains renders the coactivator more effective. Finally, the AD and the interaction with the receptor are sufficient for the coactivation of GR. As shown in Fig. 8A, an hPGC-1 construct that includes the AD and the stronger NID, NID1 (N186), enhanced transcription by 15-fold, suggesting that the first 186 aa form a core domain that, although compromised, can potentiate GR-mediated transcription.
The C terminus of hPGC-1 contributes to coactivator function. The C terminus of hPGC-1 contains two recognizable features, an SR-rich region and a putative RNA-binding motif. To address a possible role of these domains in coactivation, we tested the ability of hPGC-1 variants that lack the C terminus to enhance receptor-mediated transcription. hPGC-1 constructs with different C-terminal truncations, such as N293 and N408, displayed reduced activity but were still good coactivators of transcription (Fig. 8B). Interestingly, deletion of the C terminus in hPGC-1 variants that already lacked NID1 (N482⌬1, N408⌬1, and N293⌬1) reduced hPGC-1 activity to almost the background level (Fig. 8B), suggesting that in the absence of the C terminus, the AD and the weaker NID, NID2, are not sufficient to enhance receptor activity. In conclusion, the hPGC-1 C terminus, although not essential, contributes to the coactivation function.
What might the role of the C terminus be? Motif prediction programs suggest that aa 635 to 670 may adopt a coiled-coil conformation (42). We therefore speculated that the C terminus could mediate homophilic interactions between hPGC-1 molecules. Oligomerization could stabilize hPGC-1 interactions with either steroid receptor dimers or other proteins. To test this notion, we used the two-hybrid system. Yeast carrying a Gal4-responsive ␤-gal reporter was transformed with vectors expressing the C terminus of hPGC-1 fused to the Gal4 DBD (Gal4-294C) as bait and hPGC-1 variants fused to the Gal4 AD as prey. As shown in Fig. 9, coexpression of Gal4-294C and AD-hPGC-1 induced ␤-gal activity, indicating that hPGC-1 can indeed interact with itself. This interaction was dependent on the C terminus, since it was detected with hPGC-1 proteins lacking the N terminus, such as 294C, but was lost in hPGC-1 variants N408 and N293, which lack the C-terminal region (Fig. 9).

DISCUSSION
In this work, we present a genetic scheme for the identification of mammalian proteins that regulate steroid hormone responses, based on their function in yeast. Using this approach, we have identified hPGC-1 as a protein that enhances glucocorticoid signaling. We show that hPGC-1 is a nuclear protein, expressed in a tissue-specific manner, and a potent coactivator of selective steroid hormone receptors.
The conservation of cellular pathways and molecular mechanisms from yeast to mammals has enabled genetic studies of higher eukaryotic processes in the simple, unicellular yeast. Hence, genes from several species have been cloned in yeasts by complementation of yeast mutants (33,37,39,40,66,74) or in gain-of-function screens (44,76). Since mammalian steroid receptors were first expressed in S. cerevisiae (46,65), functional genetic screens have revealed yeast proteins participating in steroid signaling and led to the study of some of the conserved mammalian counterparts (6,24,29,31,32,45). In parallel, the development and application of yeast two-hybrid screens has identified a number of proteins that interact physically with nuclear receptors (36). The approach that we present here, the functional screening of mammalian cDNA libraries in yeast engineered to reconstitute a mammalian pathway, should identify proteins regulating any step in the pathway, independent of a physical interaction or subcellular localization. We show that this approach can yield proteins of interest. Of the three proteins identified, the molecular chaperone p23 is a known component of steroid aporeceptor com-FIG. 6. hPGC-1 carries a transcriptional AD. (A) Gal4-hPGC-1 activates transcription in mammalian cells. COS7 cells transfected with the luciferase reporter plasmid pGK-1 and an expression vector for either the Gal4 DBD (aa 1 to 147 of Gal4) or the Gal4-hPGC-1 chimera were assayed for luciferase activity. Data are expressed relative to the activity seen with the Gal4 DBD alone (Gal4), which was set equal to 1, and are the mean Ϯ standard deviation of six transfections from three independent experiments. (B) The hPGC-1 AD is in the N terminus. Yeast carrying a Gal4-responsive ␤-gal reporter and expression vectors for fusions of Gal4 (aa 1 to 147) to full-length or truncated hPGC-1 was assayed for ␤-gal activity. hPGC-1, aa 1 to 798; N408, aa 1 to 408; N293, aa 1 to 293; 91C, aa 91 to 798; 189C, aa 189 to 798. Data are the mean Ϯ standard deviation ␤-gal activity of six or more independent yeast transformants.
plexes. It interacts indirectly, via Hsp90, with the cytoplasmic, hormone-free GR and is thought to stabilize hormone-binding-competent receptors (11,61). LEM5 is a novel, ubiquitous nuclear protein with a putative yeast homologue (A. Kaul and A. Kralli, unpublished data). In contrast, hPGC-1 (LEM6) is expressed in a tissue-specific manner and has no yeast or nematode homologue, suggesting that it could provide a regulatory function specific for vertebrates. Given our interest in proteins that confer specificity and sensitivity to lipophilic hormone signaling, we focused our studies on hPGC-1.
We show here that the nuclear protein hPGC-1 is a potent, bona fide coactivator of steroid receptors. It interacts with the hormone-activated form of GR, contains a transcriptional AD, and enhances strongly the transcriptional response mediated by GR, MR, and ER in mammalian cells. The high sequence identity of hPGC-1 with murine PGC-1 (95%) suggests that the two proteins are functional homologues. Murine PGC-1, isolated as a protein that interacts with the nuclear receptor PPAR␥, has been shown to enhance transcription by PPAR␥ and TR and to be expressed in a tissue-specific manner (62). Taken together, these findings imply a wide role for hPGC-1 and mouse PGC-1 as tissue-specific regulators of nuclear receptor signaling.
Steroid receptors harbor two transcriptional ADs: the hormone-independent AF1 at the N terminus and the hormonedependent AF2 at the C-terminal LBD. hPGC-1 enhanced the transcriptional activity of the LBD but not of the N-terminal part of the receptor, suggesting that it operates via AF2. In support of this notion, hPGC-1 enhanced the response to agonists that activate AF2 but not the response to partial agonists or antagonists, such as RU486, that do not induce proper LBD folding. Furthermore, one of the domains of hPGC-1 that interacts with the receptor LBD carries an LXXLL sequence, a motif that mediates the interaction of several proteins with the AF2 surfaces of receptors (20,70). Altogether, these data suggest that hPGC-1 interacts with and enhances the activity of AF2.
Two adjacent domains in hPGC-1, NID1 and NID2, mediate the hormone-dependent interaction with the LBD of GR. NID1 (aa 91 to 186), as mentioned before, contains an LXXLL sequence. Secondary structure predictions for NID1 suggest that the LXXLL sequence and its flanking regions adopt a helical conformation, consistent with it being a nuclear receptor interaction surface. NID2 (aa 186 to 293) does not contain a consensus LXXLL motif but has a short sequence, LLKYL TTNDD, that resembles the third interaction box of the coactivator TIF2 (underlined amino acids are identical) and is also predicted to be part of an alpha helix. Future experiments will address the role of these motifs in the interactions mediated by NID1 and NID2. Finally, two lines of evidence suggest that NID1 is the major site of interaction with GR. First, the yeast two-hybrid assays show a stronger interaction of the receptor LBD with NID1 than with NID2 (Fig. 7). Second, hPGC-1 variants that lack NID1 display weaker GR coactivation than those that lack NID2, particularly when the C terminus of hPGC-1 is absent (compare N293⌬1 and N186 in Fig. 8).
Murine PGC-1 has also been reported to have two distinct interaction domains for PPAR␥. One of them is in the Nterminal 180 aa and could be the same as NID1 of hPGC-1 (73). The second seems to be distinct from hPGC-1 NID2, since it is in a different region (aa 292 to 338 of murine PGC-1) and mediates a hormone-independent interaction with the DBD of PPAR␥ (62). These findings raise the interesting possibility that coactivators such as PGC-1 may have one interaction site (e.g., NID1) that directs them to a family of transcription factors, such as the nuclear receptors, and auxiliary sites (e.g., hPGC-1 NID2) that can discriminate among individual receptors within the large family. The first site could play a role in the strength of the interaction, and the second one could play a role in the specificity of the interaction. Interestingly, the coactivator GRIP1 was recently shown to have a C-terminal auxiliary domain that selectively enhances interactions with the LBDs of specific receptors (22).
Recruitment by the hormone-activated receptor would bring hPGC-1 to the vicinity of promoters under hormone regulation. What are the downstream effectors of hPGC-1? The presence of a strong transcriptional AD indicates that hPGC-1 makes direct contacts via its AD with either the basal transcription machinery or other transcription cofactors. The acidic nature of the hPGC-1 AD suggests contacts with targets of other acidic domains, e.g., TFIIA (30), histone acetyltransferase complexes (71), or the Mediator complex (also known as ARC and Srb) (5,50). An LLXXLXXXL sequence at aa 88 to 96, similar to the CBP interaction motif in the p160 (NCoA) family of coactivators (70), implies possible contacts with CBP. Since Gal4-hPGC-1 is a potent transcriptional activator in both yeast and mammalian cells, it is possible that hPGC-1 can interact with more than one of these effectors, thereby activating multiple steps in transcription initiation and elongation.
Efficient coactivation of GR requires the C terminus of hPGC-1. A possible reason for this is the ability of this region to mediate homophilic hPGC-1 interactions. A "dimerization" surface could stabilize the binding of an hPGC-1 dimer to a receptor dimer, in a complex where each NID1 contacts one receptor AF2. Dimerization could also help NID2 to compensate for the loss of NID1. Deletion of the dimerization function might lead to two alternative complexes: (i) two hPGC-1 monomers and a receptor dimer, where the stronger NID (NID1) interacts with each AF2, or (ii) an hPGC-1 monomer and a receptor dimer, where NID1 and NID2 each interact with one receptor AF2. The latter case is similar to what has been observed for the crystal structure of the PPAR␥ LBD with an SRC-1 fragment that has two NIDs; a single SRC-1 molecule contacts via two LLXXLL motifs the two AF2 sur- with the GR expression vector p6RGR, the GR-responsive luciferase reporter plasmid pTAT3Luc, and either the control vector pcDNA3 or hPGC-1 variants as indicated, treated with 50 nM corticosterone for 24 h, and assayed for luciferase activity. Results are expressed as fold enhancement by hPGC-1, with activity in the presence of hormone and just the control vector pcDNA3 set equal to 1. Data are the mean Ϯ standard deviation of 6 to 12 values from at least three independent experiments. hPGC-1 variants are named as in Fig. 6 and 7; hPGC-1 ⌬2L carries a deletion of aa 189 to 482. N88, aa 1 to 88; N186, aa 1 to 186. (C) hPGC-1 variants are expressed and present in the nucleus. COS7 cells transfected with the indicated hPGC-1 constructs were analyzed for the expression and localization of the HA epitope-tagged protein by fluorescence microscopy. Note that cells transfected with hPGC-1 variants that lack the putative nuclear localization signals (N293 and N293⌬1) show some cytoplasmic staining but still have comparable levels of hPGC-1 in the nucleus. faces of the receptor dimer (53). In either case, a deletion of NID1 in the absence of the C terminus may leave the weaker NID (NID2) unable to mediate a stable interaction with the receptor dimer. This could explain why hPGC-1⌬1, which has both NID2 and the dimerization interface, but not N482⌬1, which has just NID2, can enhance GR activity (Fig. 8). A general role for dimerization is also consistent with the observations that receptor dimerization is important for other receptor-coactivator interactions (27) and that two molecules of the coactivator TIF2 bind cooperatively to a nuclear receptor heterodimer (38).
The C terminus of hPGC-1 also carries novel motifs for transcriptional coactivators, such as the SR-rich sequences and the putative RNA-binding domain. Although such features are often associated with posttranscriptional RNA-processing regulators, they have been found recently in other proteins that interact with transcription factors. The family of CTD-associated SR-like proteins (CASPs) interacts with the C-terminal domain of RNA polymerase II and has been suggested to couple transcription to pre-mRNA processing (10,79). An RNA-binding protein, TLS, has been purified as a thyroid receptor-interacting protein (60). Although speculative at the moment, interesting roles for these domains can be proposed. First, they may bind an RNA cofactor (e.g., the steroid receptor RNA activator SRA [34]) that could play a structural role in interactions with other transcription factors or regulate hPGC-1 activity. More interestingly, they may enable coactivators such as hPGC-1 to regulate steps other than transcription initiation and elongation, e.g., pre-mRNA processing. Steroid receptors could recruit hPGC-1 and deliver it to the splicing machinery, possibly via RNA polymerase II, thereby providing hormone-regulated, gene-specific RNA splicing.
The identification of hPGC-1 establishes one more member of an already large group of coactivators for steroids and other nuclear receptors: the p160 family (represented by SRC-1, TIF2, and pCIP]), the histone acetyltransferases CBP and pCAF, and the multisubunit Mediator (also called SMCC, DRIP, TRAP, ARC, CRSP, and Srb) (reviewed in references 13, 28, 69, and 75). One of the challenges in transcriptional regulation by nuclear receptors is to understand how specificity (such as tissue, receptor, signal, or promoter specificity) and sensitivity to regulation (i.e., sensing and adapting to changes in cell state) are established. The existence of multiple coactivators could confer these properties, if the coactivators were expressed tissue specifically, showed selectivity for different receptors, affected different limiting steps in the control of gene expression, and were subject to regulation. hPGC-1 and murine PGC-1 display many of these properties. First, they are expressed tissue specifically. The predominant expression in heart, skeletal muscle, kidney, liver, and brown fat overlaps the expression of GR, MR, TR, and PPAR␥ and correlates with the sensitivity of these receptors to PGC-1. Second, a comparison of hPGC-1 and SRC-1e highlights receptor selectivity. The efficiency of the two coactivators is reversed on different receptors, SRC-1e being more potent with AR and hPGC-1 being stronger with ER, GR, and MR. Third, the effects of hPGC-1 and murine PGC-1 may depend on promoter context. Enhancement of GR activity by hPGC-1 but not by SRC1e was stronger at the TAT3 reporter than at the MMTV reporter. Similarly, Wu et al. reported that murine PGC-1 coactivation is promoter dependent, affecting the transcription of selective PPAR␥-targeted genes in muscle (73). Finally, hPGC-1 and murine PGC-1 are good candidates for integrating other signaling pathways. Exposure of mice to cold and activation of ␤-adrenergic receptors lead to increased expression of murine PGC-1 (4,62). The presence of putative phosphorylation sites suggests additional levels of regulation. We have observed that the ability of hPGC-1 to enhance GR activity is cell type dependent (data not shown). Understanding the regulatory mechanisms and the specificity of coactivator function will be the next challenge.