Transactivation by Retinoid X Receptor–Peroxisome Proliferator-Activated Receptor γ (PPARγ) Heterodimers: Intermolecular Synergy Requires Only the PPARγ Hormone-Dependent Activation Function

Abstract

The ability of DNA sequence-specific transcription factors to synergistically activate transcription is a common property of genes transcribed by RNA polymerase II. The present work characterizes a unique form of intermolecular transcriptional synergy between two members of the nuclear hormone receptor superfamily. Heterodimers formed between peroxisome proliferator-activated receptor γ (PPARγ), an adipocyte-enriched member of the superfamily required for adipogenesis, and retinoid X receptors (RXRs) can activate transcription in response to ligands specific for either subunit of the dimer. Simultaneous treatment with ligands specific for both PPARγ and RXR has a synergistic effect on the transactivation of reporter genes and on adipocyte differentiation in cultured cells. Mutation of the PPARγ hormone-dependent activation domain (named τc or AF-2) inhibits the ability of RXR-PPARγ heterodimers to respond to ligands specific for either subunit. In contrast, the ability of RXR- and PPARγ-specific ligands to synergize does not require the hormone-dependent activation domain of RXR. The results of in vitro and in vivo experiments indicate that binding of ligands to RXR alters the conformation of the dimerization partner, PPARγ, and modulates the activity of the heterodimer in a manner independent of the RXR hormone-dependent activation domain.

Members of the nuclear hormone receptor superfamily are ligand-dependent transcription factors that profoundly influence vertebrate development, differentiation, and homeostasis (44). The peroxisome proliferator-activated receptor γ (PPARγ), an adipocyte-enriched member of the superfamily, has been shown to play a pivotal role in the process of adipocyte differentiation. Ligands that activate PPARγ induce preadipocytes to differentiate to adipocytes in culture. Furthermore, overexpression of PPARγ in fibroblasts and myoblasts promotes their differentiation into adipocytes when PPARγ-specific ligands are administered (for a recent review, see reference 42). Similarly, PPARγ-specific ligands have been shown to induce the differentiation of human liposarcomas in vitro (64). The recent finding that a class of anti-diabetic insulin-sensitizing drugs, the thiazolodinediones, are ligands for PPARγ is additional evidence of the important role of this receptor in human disease (18, 39).

PPARγ, like many members of the nuclear hormone receptor superfamily, functions as a heterodimer with the retinoid X receptor (RXR; for a review, see reference 43). Recent work has identified two types of RXR-dependent heterodimers, nonpermissive and permissive. In nonpermissive heterodimers, such as those between RXR and retinoic acid receptors (RARs) or thyroid hormone receptors (TRs), the partner actively interferes with the ability of RXR to activate transcription in response to RXR-specific ligands. In contrast, permissive heterodimers, such as RXR-PPARγ, allow RXR signaling (for a review, see reference 37). The ability of RXR to respond to ligands when dimerized with PPARγ in vivo has been supported by recent work demonstrating that RXR-specific ligands enhance insulin sensitivity in diabetic animals (46). Thus, the RXR-PPARγ heterodimer represents a unique bifunctional transcription factor that allows integration of two independent hormonal signaling pathways by a single functional unit.

RXR and PPARγ, like many members of the superfamily, are tripartite in structure, comprised of an amino-terminal ligand-independent transactivation function, a central DNA binding domain, and a carboxy-terminal ligand binding domain (LBD). The LBD is functionally complex and, along with ligand binding, encodes domains required for dimerization, repression of transcription in the absence of ligand, and ligand-dependent activation of transcription (for a review, see reference 44). The recently published crystal structures of the RXR, RAR, TR, and estrogen receptor LBDs support a long-standing hypothesis that binding of ligand induces a significant conformational change in receptors. Upon ligand binding, the omega loop connecting helices 1 and 3 of RAR (helices 2 and 3 of RXR) appears to undergo a 180° flip. Helix 12, the receptor domain encompassing the ligand-dependent activation function (named the τc or AF-2 domain), appears to move almost 90° from a position extended away from the rest of the LBD to a position loosely packed upon the surface (7, 20, 53, 69). The functional consequence of this conformational change arises from the ability of different trans-acting factors to distinctly recognize either unliganded or liganded receptors. In the absence of ligand, many receptors interact with a family of corepressors (silencing mediator of retinoid and thyroid receptors [SMRT] and nuclear receptor corepressor [NCoR]) that actively repress transcription. Upon ligand binding, the ligand-dependent conformational change results in the release of corepressors and promotes interactions with positively acting cofactors (for a recent review, see reference 21). The essential role of the τc/AF-2 domain in receptor activity is illustrated by the observation that deletion or mutation of this domain produces receptors that bind ligand normally but fail to release corepressors or interact with coactivators (3, 9, 11, 12, 15, 25, 27, 28, 35, 50, 54, 55, 61, 62, 65, 68). Thus, the observation that RXR-PPARγ heterodimers respond to ligands binding to either subunit suggests that the τc/AF-2 domains of both receptors can independently activate transcription.

In this work we have observed that not only are RXR-PPARγ heterodimers permissive for activation by RXR- and PPARγ-specific ligands individually, but together ligands specific for each receptor can synergistically activate transcription and promote adipocyte differentiation in cultures. Surprisingly, the ability of RXR-specific ligands to act synergistically with PPARγ-specific ligands does not require the hormone-dependent activation function (τc/AF-2 domain) of RXR. The results of in vitro and in vivo experiments indicate that ligand binding to RXR influences the conformation and activity of PPARγ in a fashion independent of the RXR τc/AF-2 domain.

MATERIALS AND METHODS

Plasmids.

GAL4 DNA binding domain fusions of the receptor-interacting domains of mouse CREB binding protein (CBP) (amino acids 1 to 171) and human steroid receptor coactivator 1 (SRC-1) (amino acids 381 to 891) were made by PCR amplification of the appropriate regions followed by cloning into pCMX-GAL4 (55). Glutathione S-transferase (GST)-CBP (amino acids 1 to 352) and GST–SRC-1 (amino acids 381 to 891) were made by PCR amplification of the appropriate fragments followed by cloning into pGEX-5X-1 (Pharmacia). All PCR products were verified by DNA sequencing. To construct the RXR τc/AF-2 domain mutant in which both the methionine at position 454 and the leucine at position 455 were changed to alanine (M454A/L455A), the complete coding region of human RXRα was amplified by PCR using oligonucleotides that had the appropriate mutation introduced. Following amplification, the correct fragment was cloned into pCMX (66) and verified by DNA sequencing. To introduce the RXR τc/AF-2 domain mutant into the context of the VP16-RXRLBD vector, the LBD fragment from pCMX-RXR(M454A/L455A) was isolated by digestion with SalI and NheI and used to replace the wild-type LBD in pCMXVP16-RXRLBD (19). To construct the PPARγ τc/AF-2 domain mutant in which the leucines at positions 466 and 467 were changed to alanine (L466A/L467A), amino acids 250 to 474 of mouse PPARγ were amplified by PCR using an oligonucleotide containing the appropriate mutation. After amplification, the fragment was digested with EcoRI and NheI and used to replace the wild-type fragment in pCMX-mPPARγ (32) and verified by DNA sequencing. To express the PPARγ L466A/L467A LBD, pCMX-PPARγ (L466A/L467A) was digested with ScaI and NheI and the mutant fragment was used to replace the wild-type PPARγ LBD in pCMXHANLS-PPARγLBD (18). Expression plasmids for human RXRα, mouse PPARγ, VP16-RXRLBD, PPARγLBD, β-galactosidase, the GST-RXR bacterial expression plasmid, and the reporters PPREx3-TK-LUC and UAS_Gx4-TK-LUC have been previously described (18, 19, 32, 38, 55, 66).

Cell culture and transfection.

NIH 3T3 cells were cultured in Dulbecco modified Eagle medium (DMEM) supplemented with 10% fetal bovine serum. Prior to transfection, cells were seeded in 48-well plates (1.5 × 10⁴ cells/well) in DMEM supplemented with 10% charcoal-resin-treated fetal bovine serum. After 12 to 16 h of growth at 37°C, cells were transfected with the N-[1-(2,3)-dioleoloxy)propyl]-N,N,N-trimethylammonium methylsulfate (DOTAP) transfection reagent following the manufacturer’s instructions (Boehringer Mannheim). For each well, 12 ng of luciferase reporter, 36 ng of the appropriate expression construct, and as an internal control, 60 ng of pCMX-βgal was transfected. When necessary, the parental expression plasmid pCMX was added to ensure that equal amounts of DNA were transfected in each well. After 5 h at 37°C, the medium was removed, the cells were washed once, and 200 μl of fresh medium was added with or without the ligands described in the legend to each figure. Cells were harvested after an additional 36 h of growth at 37°C. Luciferase activity of each sample was normalized by the level of β-galactosidase activity. Each transfection was carried out in duplicate and repeated at least three times.

Electrophoretic mobility shift assays.

Receptors were produced with a T₇ Quick TNT in vitro Transcription/Translation Kit (Promega). Reactions were set up in a final volume of 20 μl of 1× binding buffer (20 mM HEPES [pH 7.5], 75 mM KCl, 2.0 mM dithiothreitol [DTT], 0.1% Nonidet P-40 [NP-40], 7.5% glycerol) with 1.0 μl (each) of the appropriate unlabeled receptors, 2.0 μg of poly(dI-dC), and 0.02 pmol of an ³²P-labeled oligonucleotide with a single peroxisome proliferator response element (PPRE) derived from the acyl-coenzyme A (acyl-CoA) oxidase promoter (GTCGACAGGGGACC AGGACA A AGGTCA CGTTCGGGAGT; boldfacing indicates the direct repeat). The receptor-specific ligands, 1.0 μM LG100268 and 1.0 μM BRL49653, were added to the appropriate samples, and heterodimers were allowed to assemble on DNA in the presence or absence of ligands for 20 min at room temperature. After this initial incubation, the crosslinker bis(sulfosuccinimidyl)suberate was added to final concentration of 1.0 mM and the reaction mixture was incubated for 20 min at room temperature. To stop the cross-linking, Tris (pH 8.0) was added to a final concentration of 100 mM and the DNA-protein complexes were resolved on 5% nondenaturing acrylamide gels. After electrophoresis, the gels were dried and exposed to film.

Far-Western blotting.

GST fusion proteins were purified as previously described (54) and resolved on sodium dodecyl sulfate (SDS)–10% acrylamide gels. After electrophoresis, the proteins were transferred to polyvinylidene fluoride membranes (Novex) for 2 h at 60 V. Following transfer, the proteins on the blot were renatured by two 30-min washes at room temperature in HB buffer (25 mM HEPES [pH 7.7], 25 mM NaCl, 5 mM MgCl₂, 1 mM DTT) containing 6.0 M guanidine HCl. After the first two washes, the guanidine HCl was diluted 1:1 with fresh HB buffer and washed for another 30 min. Dilution of the guanidine HCl (1:1) was repeated until a concentration of 0.187 M was reached. After renaturation, the blot was washed in HB buffer for 1 to 2 h at room temperature. To block nonspecific binding, the blot was washed in HB buffer containing 5% nonfat dry milk for 30 min at room temperature followed by a second 30-min wash in HB buffer plus 1.0% nonfat dry milk. In vitro-translated ³⁵S-labeled RXR and ³⁵S-labeled PPARγ were produced using a T₇ Quick TNT in vitro Transcription/Translation Kit (Promega). Equal amounts of ³⁵S-labeled receptors were used, as determined by phosphorimaging analysis of SDS-acrylamide gels. To determine the effect of ligands, receptors were preincubated for 1 h with 1.0 μM LG100268 (RXR) or 5.0 μM BRL49653 (PPARγ) before being mixed with the blots. Receptors were then incubated with the blots in 5.0 ml of H buffer (20 mM HEPES [pH 7.7], 75 mM KCl, 2.5 mM MgCl₂, 1 mM DTT, 0.1 mM EDTA, 0.05% NP-40, 1.0% nonfat dry milk) in the presence of ligands. Purified GST (5.0 μg/ml) was added to the incubation along with 5.0 μM LG100268 or 5.0 μM BRL49653 when appropriate. The blots were then incubated 14 to 16 h at 4°C. Following incubation with ³⁵S-labeled receptors, the blots were washed four times for 15 min each time at room temperature with buffer H, air dried, and exposed to film.

Protease protection.

Unlabeled or ³⁵S-labeled receptors were produced using a T₇ Quick TNT in vitro Transcription/Translation Kit (Promega). Reactions were set up in a final volume of 20 μl of 1× binding buffer (20 mM HEPES [pH 7.5], 75 mM KCl, 2.0 mM DTT, 0.1% NP-40, 7.5% glycerol) with 1.0 μl of the appropriate unlabeled or ³⁵S-labeled receptors, 2.0 μg of poly(dI-dC), and 0.1 pmol of an oligonucleotide with a single PPRE derived from the acyl-CoA oxidase promoter (GTCGACAGGGGACC AGGACA A AGGTCA CGTTCGGGAGT). For every sample, the total volume of reticulocyte lysate from the in vitro-translated receptors was 2.0 μl. The receptor-specific ligands, 1.0 μM LG100268 and 5.0 μM BRL49653, were added to the appropriate samples, and heterodimers were allowed to assemble on DNA in the presence or absence of ligands for 20 min at room temperature. After this initial incubation, 1.0 μg of modified sequencing grade trypsin (Boehringer Mannheim) was added and allowed to digest for 20 min at room temperature. The reactions were stopped by addition of an equal volume of 2× SDS gel sample buffer and immediate boiling for 3 min, and then 25% of each sample was resolved on SDS–14% acrylamide gels. After electrophoresis, the gels were fixed, treated for 20 min with Amplify (Amersham), dried, and exposed to film.

RESULTS

Synergistic activation by RXR- and PPARγ-specific ligands.

Several studies have indicated that RXR-PPARγ heterodimers activate transcription in response to both RXR- and PPARγ-specific ligands (16, 32, 46, 64). The permissive nature of RXR-PPARγ heterodimers is illustrated by the transfection experiment in Fig. 1A examining the response to receptor-specific ligands. NIH 3T3 cells were chosen as the recipient for transfection, because in the absence of cotransfected receptors these cells do not exhibit a detectable response to either RXR- or PPARγ-specific ligands. Also, in NIH 3T3 cells, the ability to detect a response to the PPARγ-specific ligand BRL49653 requires transfection of expression plasmids for both PPARγ and RXR (data not shown). The requirement for transfection of both RXR and PPARγ provides the opportunity to determine the contributions of each receptor subunit without complications arising from endogenous receptors. When expression plasmids for both RXR and PPARγ are transfected, a response to both RXR-specific (LG100268) (6) and PPARγ-specific (BRL49653) (18, 39) ligands is observed (Fig. 1A).

FIG. 1.

Synergistic activation by RXR- and PPARγ-specific ligands. (A) NIH 3T3 cells were transfected with PPREx3-TK-LUC and expression constructs for mouse PPARγ and human RXRα. After transfection, cells were cultured in the absence (None) or presence of the RXR-specific ligand LG100268 (1.0 μM), the PPARγ-specific ligand BRL49653 (5 μM), or 1.0 μM LG100268 plus 5.0 μM BRL49653 for 36 h. Activity relative to that of the PPREx3-TK-LUC reporter alone is reported. Each number above a bar indicates the fold induction relative to the activity in the absence of ligand. Note the break in the y axis. Transfections were normalized by cotransfection with a β-galactosidase expression plasmid (see Materials and Methods). (B to E). Confluent 3T3 L1 preadipocytes were cultured for 7 days in the absence (B) or presence of 100 nM BRL49653 (C), 100 nM LG100268 (D), or 100 nM BRL49653 plus 100 nM LG100268 (E). After treatment, cells were stained with oil red O to visualize lipids.

The results of Fig. 1A described above indicate that RXR-PPARγ heterodimers can respond to either RXR- or PPARγ-specific ligands. When NIH 3T3 cells transfected with RXR and PPARγ are treated with the combination of LG100268 and BRL49653, transactivation threefold greater than the sum of each ligand alone is observed, indicating that activation of the individual subunits leads to a synergistic response (Fig. 1A; note the break in the y axis). Synergy is also observed using the natural RXR ligand 9-cis-retinoic acid (data not shown). To determine if the synergy observed with receptor-specific ligands in transfection experiments holds true for endogenous RXR-PPARγ heterodimers in vivo, the ability of LG100268 and BRL49653 to differentiate 3T3 L1 preadipocytes was examined (Fig. 1B to E). Several laboratories have shown that treatment of 3T3 L1 preadipocytes with ligands that activate PPARγ induces a genetic program that leads to adipocyte differentiation (10, 18, 31, 63, 64). Preadipocytes treated with receptor-specific ligands alone or in combination for 7 days were fixed and stained with oil red O to visualize lipids. At suboptimal concentrations (100 nM), BRL49653 (Fig. 1C) or LG100268 (Fig. 1D) poorly promote adipocyte differentiation. In contrast, the combination of ligands has a synergistic effect, producing an endogenous response significantly greater than the additive effects of each ligand alone (Fig. 1E), consistent with the synergy observed in transfection experiments in Fig. 1A.

Receptor-specific ligands promote differential interactions with CBP and SRC-1.

To begin to address the mechanism of synergistic transactivation by RXR-PPARγ heterodimers, the ability of receptor-specific ligands to promote interactions with CBP, a known coactivator for the nuclear hormone receptor superfamily, was examined (for a recent review, see reference 21). Western blotting experiments indicate that CBP is present at a constant level throughout 3T3 L1 adipocyte differentiation (data not shown). Figure 2A illustrates the modified mammalian two-hybrid system used to examine interactions between RXR-PPARγ heterodimers and CBP. NIH 3T3 cells were transfected with constructs expressing a GAL4-CBP fusion (amino acids 1 to 171 of CBP encoding the receptor-interacting domain), a VP16 activation domain-RXR LBD fusion (VP16-RXRLBD), and the LBD of PPARγ (PPARγLBD). The results show that the PPARγ-specific ligand BRL49653 promotes a relatively strong interaction with CBP (8.2-fold above the activity observed in the absence of ligands [Fig. 2A]). The RXR-specific ligand LG100268 also promotes a detectable interaction with CBP, although the LG100268-dependent CBP interaction is weaker than the BRL49653-dependent interaction (3.1-fold above the activity in the absence of ligands [Fig. 2A]). Once again, the combination of ligands has a synergistic effect, giving rise to signal 30.3 times that observed in the absence of ligands and approximately 3-fold more than the sum of the individual receptor-specific ligands. When the VP16-RXRLBD fusion is omitted from the transfection, no signal is detected, indicating that a RXR-PPARγ heterodimer is required to detect a response to BRL49653 in this assay (Fig. 2B). Transfection of a VP16-PPARγ fusion alone, however, does allow a BRL49653-dependent interaction with CBP (data not shown; see Fig. 4).

FIG. 2.

RXR- and PPARγ-specific ligands synergistically promote interaction with CBP. (A) A fusion between the DNA binding domain (DBD) of GAL4 and the receptor-interacting domain of CBP (amino acids 1 to 171) was cotransfected into NIH 3T3 cells along with a construct expressing a VP16 activation domain-human RXRα ligand binding domain fusion protein (VP16-RXRLBD) and a construct expressing the LBD of mouse PPARγ (PPARγLBD). A luciferase reporter with four GAL4 binding sites (UAS_Gx4-LUC) was also included. After transfection, cells were cultured in the absence (None) or presence of the RXR-specific ligand LG100268 (1.0 μM), the PPARγ-specific ligand BRL49653 (5.0 μM), or 1.0 μM LG100268 plus 5.0 μM BRL49653 for 36 h. Activity relative to GAL4-CBP (amino acids 1 to 171) alone is reported. Panels B and C are identical to panel A except that VP16-RXRLBD (B) or PPARγLBD (C) was omitted. Note the y axis in panel A differs from that in panels B and C. Panels D to F are identical to panels A to C except that GAL4-CBP (amino acids 1 to 171) was replaced with a fusion between the DNA binding domain of GAL4 and the central receptor-interacting domain of human SRC-1 (amino acids 381 to 891). Each number above a bar indicates the fold induction relative to the activity in the absence of ligand. Transfections were normalized by cotransfection with a β-galactosidase expression plasmid (see Materials and Methods).

FIG. 4.

Inactivation of the PPARγ τc/AF-2 domain inhibits the response to RXR- and PPARγ-specific ligands. (A) NIH 3T3 cells were transfected with a reporter containing three copies of the acyl-CoA oxidase PPRE cloned upstream of the thymidine kinase-luciferase (TK-LUC) reporter (PPREx3-TK-LUC) and expression constructs for mouse PPARγ and human RXRα. Panel B is identical to panel A except that a PPARγ τc/AF-2 domain double point mutant (L466A/L467A) was used in place of the wild-type PPARγ. After transfection, cells were cultured in the absence (None) or presence of the RXR-specific ligand LG100268 (1.0 μM), the PPARγ-specific ligand BRL49653 (5.0 μM), or 1.0 μM LG100268 plus 5.0 μM BRL49653 for 36 h. Activity relative to that of the PPREx3-TK-LUC reporter alone is reported. Note the break in the y axis. (C) A fusion between the DNA binding domain of GAL4 and the receptor-interacting domain of CBP (amino acids 1 to 171) was cotransfected into NIH 3T3 cells along with constructs expressing a VP16 activation domain-human RXRα ligand binding domain (VP16-RXRLBD) fusion protein and the LBD of mouse PPARγ (PPARγLBD). Panel D is identical to panel C except that a PPARγ τc/AF-2 domain double point mutant (L466A/L467A) was used in place of the wild-type PPARγ. After transfection, cells were cultured in the absence (None) or presence of 1.0 μM LG100268, 5.0 μM BRL49653, or 1.0 μM LG100268 plus 5.0 μM BRL49653 for 36 h. Activity relative to GAL4-CBP (amino acids 1 to 171) alone is reported. Note the break in the y axis. Each number above a bar indicates the fold induction relative to the activity in the absence of ligand. Transfections were normalized by cotransfection with a β-galactosidase expression plasmid (see Materials and Methods). Western blot experiments indicate that the PPARγ mutant is expressed at a level similar to the wild-type level (data not shown).

To examine the interaction between RXR homodimers and CBP, a standard mammalian two-hybrid assay comprised of GAL4-CBP and VP16-RXRLBD was used (Fig. 2C). Addition of the RXR-specific ligand LG100268 promotes an interaction similar to that observed with RXR-PPARγ heterodimers (4.5-fold above the activity in the absence of ligands; note the different y axes for Fig. 2A and C). The relatively weak interaction between RXR and CBP is similar to that observed by other investigators using this assay (9).

A similar modified two-hybrid assay (VP16-RXRLBD plus PPARγLBD) was used to examine interactions between RXR-PPARγ heterodimers and a second coactivator, SRC-1 (for a recent review, see reference 21). As observed with CBP, addition of the RXR-specific ligand LG100268 promotes an interaction with SRC-1 (Fig. 2D). The PPARγ-specific ligand BRL49653, however, has little or no effect (Fig. 2D). In contrast to CBP, combining the two ligands is not synergistic and a signal similar to that observed with LG100268 alone is obtained (Fig. 2D). Interestingly, a relatively weak BRL49653-dependent interaction between RXR-PPARγ heterodimers and SRC-1 can be detected in CV-1 cells by the same two-hybrid assay (data not shown), suggesting that species and/or tissue-specific factors may influence receptor-coactivator interactions.

The results of the modified two-hybrid analysis of Fig. 2 suggest that treatment with RXR- and PPARγ-specific ligands favors recruitment of different coactivators. To confirm the observed differences in coactivator recruitment, electrophoretic mobility shift assays were performed with in vitro-translated receptors and recombinant GST fusion proteins (Fig. 3A and B). Incubation of GST-CBP (amino acids 1 to 352) in a standard gel shift assay with RXR-PPARγ heterodimers bound to a single ³²P-labeled PPRE results in the appearance of a slower-migrating RXR-PPARγ-CBP complex (Fig. 3A, lanes 5 and 6). The RXR-PPARγ-CBP complex is increased by treatment with BRL49653 (Fig. 3A, lanes 6 and 7; 3-fold as determined by phosphorimaging analysis of three independent experiments), while treatment with LG100268 has little or no effect (Fig. 3A, lane 8). The results of gel shift experiments comparing constitutive and ligand-stimulated CBP interactions correlate well with the results of the modified two-hybrid assay. In contrast to the modified two-hybrid analysis, however, the combination of receptor-specific ligands in the gel shift assay does not result in synergistic recruitment of CBP. The inability to detect synergy in the gel shift assay may result from the relative insensitivity of the gel shift assay, the effects of ligands on receptor dimerization, or the requirement for additional trans-acting factors that are absent from the in vitro system (see Discussion). A similar gel shift assay was used to examine interactions with SRC-1 (amino acids 381 to 891) (Fig. 3B). In agreement with the two-hybrid results, the addition of the RXR-specific ligand LG100268 promotes the appearance of a slower-migrating RXR–PPARγ–SRC-1 complex, while the PPARγ-specific ligand BRL49653 does not. The combination of receptor-specific ligands is not significantly different from LG100268 alone (Fig. 3B, lanes 2 to 5).

FIG. 3.

RXR and PPARγ prefer different coactivators. (A) In vitro-translated receptors were incubated with a ³²P-labeled PPRE oligonucleotide and 10 μg of GST (lanes 2 to 5) or GST-CBP (amino acids 1 to 352) (lanes 6 to 17). DNA-protein complexes were resolved as described in Materials and Methods. As noted above the gel, 1.0 μM BRL49653 and/or LG100268 were included (+) or not included (−). Heterodimers formed with wild-type RXR and wild-type PPARγ (lanes 1 to 9), with wild-type RXR and mutant PPARγ L466A/L467A (lanes 10 to 13), and with mutant RXR M454A/L455A and wild-type PPARγ (lanes 14 to 17) are shown. A nonspecific band derived from the reticulocyte lysate is indicated by the asterisk. (B) Interaction with SRC-1 was carried as described in panel A using wild-type receptors. GST–SRC-1 (amino acids 381 to 891) (10 μg) was included in lanes 2 to 5. (C) The interaction of PPARγ and RXRα with CBP and SRC-1 was examined by far-Western blotting. A Coomassie blue-stained gel (Stain Gel) of immobilized GST fusion proteins and far-Western blots probed with ³⁵S-labeled PPARγ or ³⁵S-labeled RXR are shown. Molecular mass markers (lane 1), GST (lanes 2, 6, 10, 14, and 21), GST–SRC-1 (amino acids 381 to 891) (lanes 3, 7, 11, 15, and 19), GST-CBP (amino acids 1 to 352) (lanes 4, 8, 12, 16, and 20), and GST-RXR (lanes 5, 9, 13, 17, and 21) were used. After electrophoresis, proteins were transferred to polyvinylidene difluoride membranes, renatured, and incubated with ³⁵S-labeled mPPARγ in the absence (−Ligand) or presence of 5 μM BRL49653 or with ³⁵S-labeled RXR in the absence (−Ligand) or presence of 5 μM LG100268.

The results of Fig. 2 and 3 indicate that the PPARγ-specific ligand BRL49653 promotes a stronger interaction between RXR-PPARγ heterodimers and CBP than the RXR-specific ligand LG100268 does. On the other hand, LG100268 promotes a stronger interaction between RXR-PPARγ heterodimers and SRC-1 than BRL49653 does. One interpretation of this result is that the relative affinities of the individual PPARγ and RXR subunits for CBP and SRC-1 are different. To test this hypothesis, far-Western blots were used to analyze direct interactions between the individual in vitro-translated subunits of the heterodimer and the receptor-interacting domain of CBP and SRC-1. Figure 3C shows that in vitro-translated ³⁵S-labeled PPARγ makes a strong interaction with immobilized CBP (lanes 8 and 12), while a weak but detectable interaction is observed with immobilized SRC-1 (lanes 7 and 11). The PPARγ-CBP interaction is stimulated approximately threefold by BRL49653 (compare lane 8 with lane 12). As expected, PPARγ has a strong interaction with immobilized RXR (lanes 9 and 13). In contrast, when in vitro-translated ³⁵S-labeled RXR is used as the probe, little or no interaction with CBP is observed (lanes 16 and 20). Nevertheless, a relatively strong LG100268-dependent interaction with SRC-1 is detected (lane 19). The ability to detect an RXR-CBP interaction in vivo (Fig. 2C) most likely arises from the increased sensitivity of the two-hybrid assay relative to that of far-Western blotting. Nevertheless, the possibility that the RXR-CBP interaction is indirect or requires additional cofactors absent from the in vitro system cannot be ruled out. Taken together, the results of Fig. 2 and 3 support the conclusion that the individual RXR and PPARγ subunits prefer different coactivators.

The RXR τc/AF-2 domain is not required for synergistic transactivation.

The ability of receptor-specific ligands to synergistically activate transcription suggests that the hormone-dependent activation functions (τc/AF-2 domain) of both RXR and PPARγ participate in transactivation. As expected from this prediction, inactivation of the PPARγ τc/AF-2 domain by mutating the leucines at positions 466 and 467 to alanine (L466A/L467A) eliminates the response to the PPARγ-specific ligand BRL49653 (Fig. 4A and B) while only producing a modest twofold decrease in hormone-independent basal activity (Fig. 4A and B, white bars). Similar results are observed when the CBP interaction is examined by the two-hybrid (Fig. 4C and D) and gel shift assays (Fig. 3A, lanes 10 to 13). In the CBP interaction assays, however, mutation of the PPARγ τc/AF-2 domain also significantly reduces the hormone-independent CBP interaction (compare the white bars in Fig. 4C and D and lane 6 with lane 10 in Fig. 3A). The requirement for the PPARγ τc/AF-2 domain in the hormone-independent interaction with CBP indicates that either an endogenous ligand is present in both NIH 3T3 cells and reticulocyte lysate or that the PPARγ τc/AF-2 domain can be at least partially active in the absence of the ligand. We have previously suggested that in the absence of ligands receptors exist in a dynamic equilibrium shifting between active and inactive conformations (55).

Interestingly, inactivation of the PPARγ τc/AF-2 domain also reduces activation by the RXR-specific ligand LG100268 by 80% (Fig. 4A and B). The synergistic recruitment of CBP in the two-hybrid assay is also eliminated (Fig. 4C and D, shaded bars). The residual response to LG100268 observed in the two-hybrid experiment most likely arises from the activity of RXR homodimers (compare Fig. 2C with Fig. 4D). The results of Figure 4 indicate that inactivation of the PPARγ τc/AF-2 domain has a negative effect on RXR signaling. Thus, mutation of the PPARγ τc/AF-2 domain transforms a heterodimer permissive for RXR signaling into a nonpermissive heterodimer (see Discussion).

To determine the contribution of the RXR τc/AF-2 domain to synergistic activation by RXR-PPARγ heterodimers, the RXR τc/AF-2 domain mutant M454A/L455A was examined. The M454A/L455A mutant eliminates the ability of RXR homodimers to respond to LG100268 bound to the PPREx3 reporter (Fig. 5A and B), the cellular retinol-binding protein type II (CRBPII) RXR response element, and as a GAL4-RXRLBD fusion (54, 56) (data not shown). This mutant also eliminates the ability to detect ligand-dependent interactions with all coactivators tested but still binds ligand with wild-type affinity (54) (data not shown). In contrast to the negative effect of the PPARγ τc/AF-2 domain mutant (Fig. 4), inactivation of the RXR τc/AF-2 domain has little or no effect on the ability of RXR-PPARγ heterodimers to respond to the PPARγ-specific ligand BRL49653 in either transactivation assays (Fig. 5C and D), two-hybrid assays (Fig. 5E and F), or gel shift assays (Fig. 3A, lanes 14 to 17). Strikingly, even when the RXR τc/AF-2 domain is inactivated, the addition of LG100268 significantly enhances the activity of BRL49653 in both transfection and two-hybrid assays. Similar results are observed when other RXR τc/AF-2 domain mutants are examined (data not shown).

FIG. 5.

Inactivation of the RXR τc/AF-2 domain still allows synergy with PPARγ-specific ligands. (A and B) NIH 3T3 cells were transfected with a reporter containing three copies of the acyl-CoA oxidase PPRE cloned upstream of the TK-LUC reporter (PPREx3-TK-LUC) and expression constructs for human RXRα (wild type) or the human RXRα τc/AF-2 domain mutant M454A/L455A. After transfection, cells were cultured in the absence (None) or presence of the RXR-specific ligand LG100268 (1.0 μM) for 36 h. Activity relative to that of the reporter alone is reported. Western blot experiments indicate that the RXR mutant is expressed at levels similar to the wild-type level (data not shown). C and D) NIH 3T3 cells were transfected with a reporter containing three copies of the acyl-CoA oxidase PPRE cloned upstream of the TK-LUC reporter (PPREx3-TK-LUC), an expression construct for mouse PPARγ, and expression constructs for human RXRα (wild type) (C) or the RXRα τc/AF-2 domain mutant M454A/L455A (D). After transfection, cells were cultured in the absence (None) or presence of 1.0 μM LG100268, 5.0 μM BRL49653, or 1.0 μM LG100268 plus 5.0 μM BRL49653 for 36 h. Activity relative to that of the PPREx3-luciferase reporter alone is reported. Note the break in the y axis. (E and F) A fusion between the DNA binding domain of GAL4 and the receptor-interacting domain of CBP (amino acids 1 to 171) was cotransfected into NIH 3T3 cells along with a construct expressing the LBD of mouse PPARγ (PPARγLBD) and constructs expressing VP16-RXRLBD (wild type) (E) or VP16-RXRLBD fusions with the M454A/L455A mutation (F). After transfection, cells were cultured in the absence (None) or presence of 1.0 μM LG100268, 5.0 μM BRL49653, or 1.0 μM LG100268 plus 5.0 μM BRL49653 for 36 h. Activity relative to that of GAL4-CBP (amino acids 1 to 171) alone is reported. Western blot experiments indicate that the RXR mutants are expressed at levels similar to the wild-type level (data not shown). Each numbers above a bar indicates the fold induction relative to the activity in the absence of ligand. Transfections were normalized by cotransfection with a β-galactosidase expression plasmid (see Materials and Methods).

Numerous studies have suggested that binding of ligand to members of the nuclear hormone receptor superfamily induces a conformational change required for receptor-mediated transactivation (for reviews, see references 57 and 72). The results of Fig. 5 indicate that binding of ligand to RXR can have a positive effect on activation by RXR-PPARγ heterodimers in the absence of the RXR hormone-dependent activation function (τc/AF-2 domain). The ability to separate a positive effect of LG100268 binding from the transactivation function of RXR suggests that not only does binding of ligand to RXR alter its own conformation but that the conformation of PPARγ must also be influenced. A large body of work has shown that binding of ligands to nuclear hormone receptors makes receptors more resistant to limited protease digestion (1, 2, 5, 20, 30, 40, 41, 45, 58, 67, 75). The more-compact structures observed by crystallography for the liganded LBDs of RAR, TR, and estrogen receptor compared to the unliganded LBD of RXR support the idea that the resistance to protease digestion observed in the presence of ligands correlates with the ability of a ligand to induce receptors to undergo a conformation change.

To test the hypothesis that LG100268 binding to RXR can affect the conformation of PPARγ, protease protection experiments were performed with dimers comprised of unlabeled RXR and [³⁵S]methionine-labeled PPARγ assembled on DNA (Fig. 6; see Materials and Methods). Consistent with the earlier work described above, addition of the PPARγ-specific ligand BRL49653 leads to increased protection of PPARγ from digestion compared to the level of digestion in the absence of ligand (Fig. 6A, compare lane 3 to lane 5). The RXR-specific ligand LG100268 also leads to increased protection of PPARγ (Fig. 6A, compare lane 3 to lane 4). Quantitation from six independent experiments indicates that LG100268 results in a 4-fold (±1.5-fold) increase in intensity of the top band of the doublet migrating at 46 kDa. This band is the tryptic fragment whose digestion is most affected by LG100268. Addition of both ligands leads to protection slightly greater than that observed with BRL49653 alone (Fig. 6A, lanes 5 and 6; quantitation of the top band of the doublet migrating at 46 kDa from six experiments indicates an average difference of 2.3-fold). Interestingly, the pattern of protection produced by the two ligands is slightly different. Addition of LG100268 results in protection of the doublet migrating at approximately 46 kDa (Fig. 6A, lane 4). In comparison, addition of BRL49653 results in significant protection of the doublet migrating just above the 30-kDa marker along with the 46-kDa doublet (Fig. 6A, lane 5). The difference in digestion patterns suggests that the two ligands may induce slightly different conformations in PPARγ.

FIG. 6.

Ligand binding to RXR alters the protease sensitivity of PPARγ. In vitro-translated receptors were incubated with an oligonucleotide containing a single PPRE and receptor-specific ligands. After this initial incubation, trypsin was added for 20 min (lanes 3 to 6), the reaction was stopped, and the ³⁵S-labeled peptides were visualized by autoradiography (see Materials and Methods). ³⁵S-labeled PPARγ plus unlabeled RXRα (wild type) (A) ³⁵S-labeled PPARγ plus unlabeled RXR M454A/L455A (τc/AF-2 domain mutant) (B), ³⁵S-labeled PPARγ plus unlabeled RXR L436S (ligand binding mutant) (C), ³⁵S-labeled PPARγ alone (D), and unlabeled PPARγ plus ³⁵S-labeled RXRα (wild type) (E) are shown. All autoradiographs were exposed for 15 h. In each panel, ¹⁴C-labeled molecular mass markers (lane 1), undigested controls (lane 2), no ligand (lane 3), 1.0 μM LG100268 (RXR specific) (lane 4), 5.0 μM BRL49653 (PPARγ specific) (lane 5), and 1.0 μM LG100268 plus 5.0 μM BRL49653 (lane 6) were run.

The results of Fig. 6A support the hypothesis that the binding of ligand to RXR influences the conformation of PPARγ. To determine whether the RXR τc/AF-2 domain is necessary for the LG100268-dependent protection of PPARγ, an experiment similar to that in Fig. 6A was performed with the RXR mutant M454A/L455A (Fig. 6B). As observed with wild-type RXR, addition of the RXR-specific ligand LG100268 increases protection of PPARγ from trypsin digestion (Fig. 6B, compare lanes 3 to lane 4). Thus, consistent with the results of Fig. 5, the ability of LG100268 to influence the conformation of PPARγ is τc/AF-2 domain independent.

To confirm the protective effect of LG100268 was mediated by binding to RXR, the RXR ligand binding mutant in which leucine at position 436 was changed to serine (L436S) was examined in protease protection experiments (Fig. 6C). The L436S mutant exhibits no detectable binding to RXR agonists in vitro and fails to activate transcription in response to LG100268 in transfection experiments (51). As expected, comparison of the 46-kDa doublet in lanes 3 and 4 of Fig. 6C demonstrates that the ability of LG100268 to protect PPARγ from digestion requires binding to RXR.

To ensure the protective effect of LG100268 on PPARγ requires RXR, a protease protection experiment was performed in the absence of RXR (Fig. 6D). In the absence of RXR, LG100268 has little or no effect on the trypsin sensitivity of PPARγ (Fig. 6D, compare lane 3 to lane 4). As expected, protection of PPARγ by BRL49653 is still observed (Fig. 6D, compare the 30-kDa doublet in lanes 3 and 5). A similar result to Fig. 6D is seen if RXR is included and the PPRE oligonucleotide is omitted from the experiment (data not shown). The requirement for DNA binding most likely arises by promoting efficient heterodimerization. Comparison of Fig. 6A with D also indicates that the doublet migrating at approximately 46 kDa in Fig. 6A is dependent on dimerization with RXR. An extensive dimerization interface between the two subunits most likely accounts for the ability of RXR to alter the protease sensitivity of PPARγ in the absence of ligands for either receptor.

Finally, increased protection of RXR upon LG100268 binding may leave a larger number of intact free RXR subunits available for dimerization with PPARγ and could provide an explanation for the ability of LG100268 to protect PPARγ. The possibility that LG100268 significantly increased the quantity of RXR was addressed by performing a protease digestion experiment where RXR was ³⁵S labeled and PPARγ was unlabeled (Fig. 6E). Compared to PPARγ, RXR is more resistant to trypsin digestion. As shown in Fig. 6E, under the conditions used, LG100268 has only a small effect on the digestion of RXR (Fig. 6E, compare lane 3 with lane 4). The results of the experiment in Fig. 6E indicate that LG100268-dependent protection of PPARγ does not result from increased quantities of RXR. Nevertheless, treatment with higher trypsin concentrations at elevated temperatures allows detection of LG100268-dependent changes in RXR protease protection (data not shown). Taken together, the results of Fig. 5 and 6 support the hypothesis that binding of ligand to RXR promotes a conformational change throughout the heterodimeric complex, ultimately influencing the conformation and activity of PPARγ.

DISCUSSION

The ability of sequence-specific transcription factors to activate transcription synergistically is a common property of genes transcribed by RNA polymerase II (for reviews, see references 8, 22, 24, 26, and 52). The results of this work describe a unique type of intermolecular transcriptional synergy. Two structurally distinct ligands binding directly to nonequivalent subunits of a single transcription factor, RXR-PPARγ heterodimers, produce a level of transcription greater than the sum of individual ligands alone. The observation that the combination of receptor-specific ligands promotes the differentiation of 3T3 L1 cells into adipocytes better than either ligand alone (Fig. 2) indicates that synergy can also be detected when a hormonal response mediated by endogenous RXR-PPARγ is examined. The abilities of ligands specific to RXR and PPARγ to promote the differentiation of liposarcomas and to increase the insulin sensitivity of diabetic animals indicate that an understanding of synergistic transcription by RXR-PPARγ heterodimers will have important implications for human disease (46, 64).

The results of protein-protein interaction studies presented in this work suggest two nonexclusive mechanisms for synergy by receptor-specific ligands. First, we have observed that the individual receptor subunits have different affinities for different cofactors. For instance, compared to RXR, PPARγ has a stronger interaction with CBP (Fig. 2 and 3), while compared to PPARγ, RXR has stronger interactions with SRC-1 and the TATA binding protein (Fig. 2 and 3; also data not shown). A similar observation has recently been made for RXR-PPARα heterodimers (17). Thus, synergistic activation of transcription may occur by each subunit of the heterodimer contacting different components of a common coactivator complex (CBP physically interacts with SRC-1 [23, 28, 59, 74]) or by each receptor recruiting different coactivator complexes. Increased recruitment of coactivators is a common model for transcriptional synergy (for reviews, see references 8, 22, 26, and 52).

A second mechanism for synergistic transactivation by RXR- and PPARγ-specific ligands arises from the observation that synergy is still detected when the RXR ligand-dependent activation function (τc/AF-2 domain) is inactivated. The results of two-hybrid analysis and protease protection experiments support the hypothesis that binding of ligand to RXR can alter or influence the conformation of PPARγ. The effect of this RXR-dependent conformational change in PPARγ manifests itself by increased recruitment of PPARγ’s “preferred coactivator” CBP. We consider CBP the preferred coactivator for PPARγ based upon the results of the protein-protein interaction experiments of Fig. 2 and 3. The coactivator preference for PPARγ in vivo, however, may also be influenced by factors such as response element sequence, promoter architecture, and coactivator concentration that are not accounted for by the assays used in this study. Nevertheless, in all assays tested, the PPARγ-specific ligand BRL49653 promotes stronger interactions between RXR-PPARγ heterodimers and CBP compared to SRC-1.

Although the combination of receptor-specific ligands promotes a synergistic interaction with CBP in vivo, synergy is not observed in the in vitro biochemical assays. Given that the level of synergy observed in the transfection experiments is small (2.5- to 3-fold greater than additive), the fixed-point in vitro assays may not have the required sensitivity to detect subtle changes. For instance, the gel shift assay requires a large excess of GST-CBP and a cross-linking reagent to detect the RXR-PPARγ-CBP complex. Also, under the gel shift conditions, the combination of receptor-specific ligands results in an approximate twofold decrease in heterodimerization (56a) similar to that observed for RXR-vitamin D receptor heterodimers (14). Any decrease in dimerization will have a negative effect on the CBP interaction detected in the gel shift assay. Nevertheless, the possibility that other trans-acting factors absent from the in vitro systems are required for synergy cannot be ruled out. It is noteworthy that CBP has been shown to interact with other coactivators such as SRC-1 and p300/CBP-associated factor (11, 28, 59, 65, 73, 74). CBP is also reported to be a component of a RNA polymerase II holoenzyme complex (29, 47, 48). The observation that in vivo coactivators are components of multimeric complexes that have enzymatic (acetyltransferase) activity (4, 11, 33, 34, 49, 60) suggests it is unlikely that synergistic transactivation observed in the transfection and two-hybrid experiments is determined simply by the strength of protein-protein interactions.

One possible trivial explanation for the ability of the two receptor-specific ligands to synergize in the two-hybrid assay (Fig. 2A) is that the PPARγ-specific ligand BRL49653 promotes a direct interaction between VP16-RXRLBD-PPARγLBD heterodimers and CBP via the PPARγLBD. Addition of the RXR-specific ligand LG100268 could lead to activation of the hormone-dependent activation function of RXR with synergy resulting from the combination of the constitutive VP16 activation domain present in the fusion protein and the ligand-dependent RXR activation function. We feel there are three reasons why such an explanation is unlikely. First, synergy is observed with full-length receptors on a typical hormone response element in the absence of the VP16 activation domain (Fig. 1A). Second, if synergy simply required two activation functions, one would expect to see BRL49653 and LG100268 synergize in the two-hybrid assay when SRC-1 is used as the bait. This result is not observed (Fig. 2D). Third, synergy is still observed when the RXR τc/AF-2 domain is inactivated by mutation (Fig. 5).

The recent crystal structures of the LBDs of RXR, RAR, and TR suggest that ligand binding to receptors results in an almost 90° movement of the τc/AF-2 domain from a position extended away from the rest of the structure to a position loosely packed upon the LBD surface. This conformational change is thought to allow receptors to achieve a structure that promotes the release of corepressors and association of coactivators (for reviews, see references 57 and 72). A key role for the τc/AF-2 domain in this ligand-dependent conformational change is the observation that deletion or mutation of this domain produces receptors that bind ligand but do not activate transcription (3, 15, 55, 61). Not only does inactivation of the τc/AF-2 domain block the interaction with coactivators, but τc/AF-2 negative receptors also exhibit stronger binding to corepressors and fail to release corepressors upon ligand binding (9, 11, 12, 27, 28, 35, 50, 65, 68). The interaction between PPARγ and the corepressors SMRT and NCoR is significantly weaker than the well-characterized interactions between corepressors and RAR or TR (76). Nevertheless, as observed with RAR and TR, inactivation of the PPARγ τc/AF-2 domain produces a dominant-negative receptor that fails to release corepressors (Fig. 4 and data not shown). In contrast to nonpermissive RXR-RAR and -TR heterodimers that always restrict RXR signaling, inactivation of the PPARγ τc/AF-2 domain changes the nature of RXR-PPARγ heterodimers from permissive for RXR signaling to nonpermissive. Likewise, mutation of the hinge region of RAR to decrease its affinity for corepressors has the opposite effect, transforming nonpermissive RXR-RAR heterodimers to permissive (35). The effect of corepressor binding on RXR signaling suggests that one major determinant of the permissive or nonpermissive nature of a particular RXR-dependent heterodimer is the affinity of the heterodimer for corepressors. Furthermore, this observation suggests that if other factors, such as corepressor concentration, posttranslational modifications, or response element sequence, can influence heterodimer-corepressor interactions (76), then the permissive or nonpermissive nature of a particular RXR-dependent heterodimer may be tissue and/or promoter specific. Nevertheless, the possibility that mutation of the PPARγ τc/AF-2 domain inhibits RXR activity by a mechanism that does not require corepressor function cannot be ruled out. Preliminary protease protection experiments, however, have failed to detect an effect of the PPARγ τc/AF-2 domain on LG100268-dependent conformational change of RXR (data not shown).

In contrast to PPARγ, point mutations that inactivate the RXR τc/AF-2 domain have little or no effect on the ability of PPARγ to respond to ligand (Fig. 3 and 5). Not surprisingly, RXR has little or no interaction with corepressors (12, 13, 27). Strikingly, however, inactivation of the RXR τc/AF-2 domain does not eliminate the ability of the RXR-specific ligand LG100268 to synergize with the PPARγ-specific ligand BRL49653. The absence of a requirement for the RXR activation function in transcriptional synergy suggests that binding of ligand to RXR induces a conformational change throughout the heterodimer that facilitates transactivation by PPARγ. For RXR, therefore, the transactivation function of the τc/AF-2 domain is not required for ligand to induce a conformational change. The possibility that when dimerized with PPARγ, a novel ligand-dependent activation surface on RXR is utilized cannot be ruled out. Nevertheless, the ability of LG100268 binding to RXR to alter the protease sensitivity of PPARγ in vitro supports the conclusion that ligand binding to RXR can alter the conformation and activity of PPARγ. The conclusion that RXR can modulate the activity of its dimerization partner is not unprecedented. We have recently described a novel RXR-specific ligand, LG100754, that activates RXR-RAR heterodimers independently of the RXR τc/AF-2 domain (36, 56). Similarly, Willy and Mangelsdorf (71) have shown that transactivation by RXR-LXR heterodimers in response to RXR agonists does not require the RXR τc/AF-2 domain. Finally, Wiebel and Gustafsson (70) have determined that the constitutive activity of the orphan receptor OR1 requires dimerization with RXR but is independent of the RXR τc/AF-2 domain. From these results, it appears likely that ligand binding to RXR results in conformational changes that are propagated through the dimer interface (helices 9 and 10) to the partner. The ability of RXR to significantly modulate the activity of its dimerization partner not only has important implications for the treatment of diabetes and other human disease but also indicates that heterodimers must be considered single functional entities that are greater than the sum of their parts.