Interaction between the Retinoid X Receptor and Transcription Factor IIB Is Ligand-dependent in Vivo *

The retinoid X receptor (RXR) influences gene activation through heterodimeric and homodimeric association with DNA and associates with TATA binding protein, TAF110, and cAMP response element-binding protein-binding protein; yet the molecular mechanisms responsible for gene activation by RXRs remain incompletely defined. Since the general transcription factor IIB (TFIIB) is a common target of sequence-specific transcriptional activators, we suspected that RXR might regulate target genes via an interaction with TFIIB. Co-immunoprecipitation, far Western analysis, and glutathioneS-transferase binding studies indicated that murine RXRβ (mRXRβ) was capable of binding to human TFIIB in vitro. Functional analysis with a dual-hybrid yeast system and cotransfection assays revealed the interaction of mRXRβ with TFIIB to be ligand-dependent in vivo. Truncation experiments mapped the essential binding regions to the carboxyl region of mRXRβ (amino acids (aa) 254–389) and two regions in the carboxyl region of TFIIB (aa 178–201 and aa 238–271). Furthermore, the Δ390–410 mRXRβ mutant bound to TFIIB in vitro but was not active in the dual-hybrid yeast system, suggesting that the extreme carboxyl region of RXR was required for in vivo interaction with TFIIB. These data indicate that interaction of mRXRβ with TFIIB is specific, direct, and ligand-dependent in vivo and suggest that gene activation by RXR involves TFIIB.

Nuclear hormone receptors are ligand-dependent transcription factors that regulate numerous cellular functions, including growth, development, differentiation, reproduction, and metabolism (Refs. 1-3 and references therein). Ligand binding to nuclear hormone receptors leads to assembly of a functional preinitiation complex (PIC) 1 and subsequent RNA transcription in vitro (4,5). Activated transcription of RNA polymerase II (RNAPII) genes requires binding of transcription factor IID (TFIID) to the TATA element, which is facilitated by TFIIA, and stepwise assembly of the following transcription factors: TFIIB, RNAPII/TFIIF, TFIIE, and TFIIH (reviewed in Refs. 6 and 7). TFIID includes the TATA binding protein (TBP) associated factors (TAFs) (Ref. 7, and references therein). Interaction of nuclear receptors with the basal transcription complex may occur directly (8 -14) or in association with TAFs (15) or other cofactors (Refs. 16 -25; reviewed in Refs. 26 and 27). Ligand binding to the receptor results in recruitment of factors that facilitate assembly of an activated complex (28 -30) or, conversely, the release of repressor molecules, such as SMRT (16) or N-CoR (17,18) that inhibit transcription (16 -18, 31-32). A significant advance was the demonstration that TFIIB interacts with the estrogen receptor (10), progesterone receptor (10), chicken ovalbumin upstream promoter-transcription factor (10), thyroid hormone receptor (TR) (11,31), and vitamin D receptor (VDR) (12)(13)(14). The interaction of nuclear hormone receptors with TFIIB is significant since recruitment of TFIIB to the PIC by transcriptional activators is believed to be a critical, rate-limiting step for PIC assembly (7,33,34) and TFIIB has been suggested to play a critical role in gene activation by nuclear hormone receptors (12,31,32,35,36).
The specific molecular steps governing the temporal and spatial specificity of ligand-dependent gene activation by RXRs remain to be precisely defined. RXRs have been shown to interact with transcriptional co-factors including TAF II 30 (15), TBP (58), CBP (29,30), and TAF II 110 (58,66,67), as well as Trip1 (24), SUG1 (24), and TIF1 (25). Studies in yeast revealed that RXRs could associate with either TBP or TAF II 110 (58,67), but these associations could be altered in the presence of a specific RXR agonist (67), suggesting that conformational states of the RXR ligand-binding domain (LBD) could determine the ability of the receptor to bind specific transcription factors in vivo. In a cell-free system, addition of RXR to the PIC reproducibly enhanced TFIIB⅐TBP complex formation (27); however, a RXR⅐TFIIB⅐TBP complex was not observed in native gel electrophoresis, suggesting that in vitro the complex might be unstable. Since RXRs have been shown to bind proteins in the transcription complex, it is possible that RXRs might actively augment binding of receptor proteins to the transcription complex (i.e. not as a silent partner), perhaps in a holocomplex with CBP (29,30). In support of this mechanism, several proteins interact in a ligand-dependent manner with the carboxyl (AF-2 or c) region of nuclear hormone receptors that heterodimerize with RXRs (21)(22)(23)(24)(25). Also against an indirect or silent partner role for RXR is the observation that a dominant negative RXR mutant (e.g. ⌬AF-2 or c) could abrogate the ability of a heterodimer to augment gene activation (64,66,69,70,71); since RXRs associate weakly, or not at all with SMRT (16) or N-CoR (17,18), trans-repression is not easily explained. Similarly, an indirect mechanism of action does not explain the ability of RXR to function as a homodimer (68). Furthermore, studies have specifically revealed the ability of liganded RXR to alter the response of a heterodimer complex associated with DNA (63,72). These observations suggest that RXRs may actively participate in assembly of the transcription complex.
In an effort to clarify some of the molecular mechanisms of gene activation by RXRs, we have tested the interaction of the general transcription factor TFIIB with mRXR␤. Using several experimental approaches, we observed the association of mRXR␤ with TFIIB to be ligand-dependent, specific, and direct. Furthermore, interaction required the carboxyl region of mRXR␤ and two distinct regions within TFIIB. These data support the conclusion that RXR may contribute directly to the recruitment of essential transcription factors involved in liganddependent gene activation.
Co-immunoprecipitation Assays-Antibody-protein complexes were precipitated using 20 l of pre-equilibrated protein A-Sepharose beads as described, with slight modifications (75), and then washed extensively. Bound products were eluted in sample buffer and analyzed by 12% SDS-PAGE and autoradiography.
Glutathione-Sepharose Binding Assays-20 l of glutathione-Sepharose beads coated with bacterially expressed GST fusion proteins were mixed with equal amounts of 35 S-labeled in vitro translated RXR or TFIIB and luciferase, added in the presence or absence of 1 M 9-cis-RA, and rocked at 4°C for 1 h. Beads were collected, washed extensively in HBB (40 mM Hepes, 75 mM KCl, 0.5 mM EDTA, 5 mM MgCl 2 , 1 mM DTT, 1 mM 4 -2(aminoethyl)benzenesulfonyl fluoride hydrochloride, 0.5 mg/ml bovine serum albumin, and 0.05% Nonidet P-40) containing 1 M 9-cis-RA or vehicle (Me 2 SO). Bound proteins were eluted in sample buffer and analyzed by 10% SDS-PAGE and autoradiography. Input lanes show one-tenth of loaded lysate.
For precipitation of TFIIB from nuclear extracts, 25 g of Namalwa human B lymphocyte nuclear extract was prepared as described (40) and incubated with 10 l of coated beads at 4°C for 1.5 h in 20 mM Hepes, 20% glycerol, 0.2 mM EDTA, 50 mM KCl, 0.5% bovine serum albumin, 0.5 mM phenylmethylsulfonyl fluoride. Beads were collected and washed extensively in 20 mM Hepes, 20% glycerol, 0.2 mM EDTA, 100 mM KCl, 0.5 mM phenylmethylsulfonyl fluoride, and 0.05% Triton X. Bound proteins were eluted by boiling and loaded onto 12% SDS-PAGE, blotted to nitrocellulose, and examined using immunoblot analysis.
Partial Proteolysis Assays-35 S-Labeled TFIIB was incubated with 7 g of Sf9 nuclear proteins prepared from cells infected with wild-type baculovirus (control) or recombinant mRXR␤ baculovirus. In other experiments, labeled TFIIB was added to a 20-l bed volume of Sepharose beads coated with bacterially expressed GST, or GST-RXR, fusion proteins. After binding for 2 h at 4°C, trypsin (50 or 500 ng) was added for 5 min at 30°C and the reaction stopped with 1 ϫ SDS, and products were resolved by 12-18% SDS-PAGE and subjected to autoradiography.

RESULTS
Interaction of RXR with TFIIB Is Direct-Since other nuclear hormone receptors have been shown to interact with TFIIB (10 -14), and TFIIB is a common target of sequence-specific transcriptional activators (6, 7), we tested whether RXR could precipitate TFIIB from proteins extracted from mammalian cell nuclei. Sepharose beads were coated with either bacterially expressed GST-RXR (RXR) fusion proteins or control proteins (GST), incubated with nuclear proteins, and washed, and immunoblots of bound proteins were performed using antibodies directed against TFIIB. As shown (Fig. 1A), TFIIB could be specifically precipitated from extracts prepared from mammalian cell nuclei using GST-RXR-coated beads. In control experiments, beads that lacked the RXR fusion protein failed to bind TFIIB, thus supporting the specificity of the interaction. TFIIB precipitated from extracts of nuclear proteins co-migrated with endogenous TFIIB detected in immunoblot analysis of nuclear proteins and recombinant TFIIB (Fig. 1A), thus suggesting that RXR could precipitate TFIIB from mammalian cell extracts (see below).
To examine the specificity of the in vitro interaction between RXR and TFIIB, co-immunoprecipitation experiments (75) were performed using in vitro transcribed and translated proteins with antibodies directed against either TFIIB or mRXR␤ (40). As shown in Fig. 1B (lane 3), 35 S-labeled in vitro transcribed and translated RXR and TFIIB were both precipitated by anti-TFIIB antibody, consistent with an interaction between the two proteins. The anti-TFIIB antibody precipitated labeled mRXR␤ in the presence of TFIIB (lanes 3 and 5) but not in the absence of TFIIB (lane 6). Conversely, anti-RXR antibody precipitated labeled TFIIB, but only in the presence of RXR. The assay was specific since neither the anti-TFIIB nor anti-RXR antibodies precipitated a control protein, in vitro translated luciferase (lanes 2, 4, 5, and 8); and conversely, a control antibody (lane 1) did not precipitate labeled RXR, TFIIB, or luciferase.
A possibility not addressed by co-immunoprecipitation experiments, or GST "pull down" experiments with nuclear proteins, was that an endogenous factor(s) present in reticulocyte lysates (or mammalian nuclear extracts) functioned as a molecular bridge to mediate the interaction between RXR and TFIIB, such as CBP (29,30). To address this concern we tested for direct interaction in far Western experiments, since bridging factors would not likely contribute to interaction in far Western analysis given the exceedingly dilute concentrations of bacterially expressed, chromatographically purified, 32 P-labeled probe proteins. Furthermore, we have not observed RXRbinding proteins in the bacterial extracts, suggesting that the probe extracts themselves did not contain bridging factors (not shown). Far Western experiments (Fig. 1C) were performed using purified recombinant 32 P-labeled GST-RXR fusion protein and recombinant hTFIIB protein. As a positive control, we observed that 32 P-labeled GST-RXR protein detected RXR (i.e. RXR-RXR homodimers) in Sf9 extracts containing recombinant RXR (Fig. 1C, lane 1) but not in control Sf9 extracts (lane 2).
Consistent with a direct interaction between the two proteins, labeled RXR detected a strong band in lanes containing TFIIB (lanes 3-5), but no binding was observed to a 32 P-labeled control protein (lanes 6 -8). Specificity of RXR binding was further substantiated in a related series of experiments demonstrating FIG. 1. Interaction between RXR and TFIIB is direct. A, precipitation of TFIIB from nuclear proteins. 25 g of nuclear proteins were incubated with 10 l of Sepharose beads coated with bacterially expressed fusion proteins of either GST-mRXR␤ (RXR), GST fused to the ⌬AF-2 mRXR␤ mutant (⌬AF-2), or as a control GST alone (lane 1). Beads were washed extensively, and bound proteins (Bound) were resolved in SDS-PAGE, blotted to nitrocellulose, and probed with primary antibody directed against TFIIB, followed by horseradish peroxidase-conjugated anti-rabbit secondary antibody, as indicated under "Materials and Methods." As controls, 5 g of input nuclear proteins (NE) prepared from mammalian cells and 1, 3, 5, and 7 ng of recombinant TFIIB were simultaneously probed with antibody directed against TFIIB. Coomassie gels of bacterially expressed protein were used to approximate amounts of fusion proteins used in binding reactions. B, co-immunoprecipitation assay showing direct interaction between mRXR␤ and TFIIB. 5 l of in vitro translated 35 S-labeled lysates (rows across) programmed with mRXR␤, hTFIIB, and/or luciferase (as indicated by ϩ) were incubated with 1 l of antibody (vertical columns) directed against either TFIIB (lanes 2-6), mRXR␤ (lanes 7-9), or major histocompatibility complex H2 Ld (Con; lane 1). After incubation and washing, antibody-bound proteins were resolved in SDS-PAGE. Black arrows indicate either precipitated mRXR␤ (upper band) or TFIIB proteins (lower band). Input labeled proteins are shown for comparison (lanes 10 -12). Antibody directed against TFIIB precipitated labeled RXR but only in the presence of TFIIB (compare lanes 3 and 6). Conversely, anti-mRXR␤ antibody precipitated TFIIB but only in the presence of RXR (lanes 7 and 8). C, far Western assay. 10, 20, or 40 g of bacterial extracts containing hTFIIB (lanes 3-8) were resolved in SDS-PAGE; proteins were blotted to nitrocellulose, denatured, and renatured with serial dilutions of 6 M guanidine hydrochloride, and probed with 500,000 cpm/ml of a 32 P-labeled GST-mRXR␤ fusion protein (lanes 1-5) or 32 P-labeled GST-0 protein as a control (lanes 6 -8). As controls, 6 g of nuclear proteins were prepared from Sf9 cells infected with baculovirus expressing either mRXR␤ (positive control, lane 1) or wildtype virus lacking RXR (wt, negative control, lane 2) and probed with 32 P-labeled GST-RXR. Labeled GST-RXR (but not labeled GST-0, lanes 6 -8) detected a band in bacterial extracts containing human TFIIB (lanes 3-5). that 32 P-labeled GST-RXR detected TR, but only if the C terminus of the labeled RXR molecule was intact. 2 As expected, migration of TFIIB detected by 32 P-labeled GST-RXR in far Western experiments precisely matched migration of 35 S-labeled TFIIB in co-immunoprecipitation experiments (Fig. 1A) and Western analysis of TFIIB using anti-TFIIB antibodies (see above). Although co-migration is not conclusive proof of identity between two proteins, Fig. 1 B and C together with A provide substantial evidence that RXR and TFIIB interact in vitro, and far Western analysis revealed that additional factors were not essential for the interaction.
TFIIB Augments RXR-dependent Reporter Activity in Mammalian Cells-Since mRXR␤ could associate with TFIIB in vitro, we examined whether overexpression of both TFIIB and RXR might augment RXR-dependent reporter activity in mammalian cells. To test this possibility, transient transfections were performed in undifferentiated P-19 embryonal carcinoma cells (12) using a luciferase reporter plasmid containing two copies of an RXR response element coupled to a tk promoter (64). Addition of an expression plasmid for RXR led to a 7-fold increase in reporter activity, whereas addition of expression plasmids for both RXR and TFIIB led to a 19-fold increase; thus addition of TFIIB led to an additional 2-3-fold ligand-dependent increase in reporter activity (Fig. 2). There was no increase in basal reporter activity or ligand-dependent reporter activity when equivalent amounts of control expression vector were added, indicating that TFIIB specifically augmented RXR-mediated gene activation. In addition, the activation depended upon the presence of the RXR response element and was not observed in experiments performed using the control tk-luciferase reporter (right panel). The magnitude of activation in mammalian cells attributable to TFIIB was modest (2-3-fold), largely due to a 7-fold level of basal activation in the presence of transfected RXR alone. The increase in basal reporter activation required RXR and was element-specific, since activity of the tk-luciferase reporter (that lacked an RXRE) was not augmented by RXR alone. At the quantities of expression plasmids tested (500 ng), our results resembled levels of activation observed by others using similar approaches (12). Although Blanco et al. (12) demonstrated a sizable in vivo effect attributable to TFIIB with VDR, other studies have not indicated a strong in vivo effect in mammalian transfection systems attributable to basal transcription factors such as TBP (9) and TFIID (8). Given the modest effect observed, we concluded that additional experiments in mammalian cells would not likely elucidate the interaction between RXR and TFIIB, because endogenous expression of TFIIB, RXR, and available RXR heterodimerization partners in mammalian cells would confound interpretation of data. Similarly, we rejected the use of mammalian two-hybrid approaches, since in that system endogenous RXR heterodimerization partners have been reported to repress RXR action (76). For this reason, we turned to an in vivo system devoid of endogenous RXR expression.
Interaction between RXR and TFIIB Is Augmented by Ligand in Vivo-We examined the interaction between RXR and TFIIB using the yeast dual-hybrid system (reviewed in Ref. 77), since yeast do not contain RXRs (78), and RXRs have been shown to function in yeast (78,79). A Saccharomyces cerevisiae strain (Y190) containing an integrated lacZ reporter was transformed with expression plasmids (Fig. 3A) encoding full-length mRXR␤ fused to the GAL4 DNA binding domain and fulllength hTFIIB fused to the GAL4 activation domain (AD). Reconstitution of this two-hybrid system would permit quantitative analysis of reporter activation and not colony survival alone (77). Furthermore, the incorporation of an integrated lacZ reporter permitted paired examination of ligand effects from individual colonies, thus controlling for variability in levels of receptor expression as determined by immunoblot analysis.
Following transformation several colonies were chosen, subjected to two rounds of dual selection, and analyzed by Western analysis to confirm similar levels of fusion protein expression. Yeast colonies that expressed both TFIIB and RXR showed a 6-fold increase in reporter activity in the presence of 9-cis-RA (Fig. 3B, lane 4), but in the absence of ligand (open bars) reporter activity did not differ from RXR alone (lane 3). Reporter activity in yeast colonies that expressed only TFIIB (lane 2) showed no significant increase in reporter activity in the presence or absence of ligand. Not surprisingly, yeast that expressed mRXR␤ coupled to the GAL4-DBD did show a slight increase in reporter activity (Fig. 3B, lane 3), likely due to the interaction of RXR (see below) with transcription factors endogenously present in yeast. The magnitude of activation observed with RXR and TFIIB in our particular system resembled levels of activation reported by others for the ligand-dependent association of LBD-RXR and GAL4TAF110 in yeast (67) and other transcription factors (80). To confirm that differences in reporter activity did not reflect variation in expression of fusion proteins, immunoblot analyses were performed on extracts prepared from yeast colonies subjected to reporter studies (Fig. 3B,  lower panel). The ligand-dependent increase in reporter activity observed in the presence of both RXR and TFIIB, but not for either construct alone, is consistent with a ligand-dependent interaction between RXR and TFIIB in vivo.
Alternatively, ligand-dependent activation of the lacZ reporter could occur if TFIIB were able to squelch negative regulator (repressor) of RXR present in yeast or if TFIIB were acting through intermediary yeast protein complexes (29,81). Although far Western analysis suggested that additional proteins were not required for the interaction, it was possible that formation of a ternary complex in vivo could confound inter- pretation of the data. To address these two possibilities, we deleted the GAL4 AD from the TFIIB expression construct (Fig.  3A, TFIIB-GAL4-AD minus). If the effect of TFIIB was to interfere with a repressor that interacted with RXR, then an increase in reporter activity should occur in the presence of TFIIB whether the GAL4 AD was present or not. Yeast transformed with RXR and the TFIIB construct lacking the GAL4AD showed an increase in basal activity but no increase in ligand-dependent reporter activity (Fig. 3B, lane 6). These results strongly suggest that the ligand-dependent reporter activity measured in Fig. 3B reflects reconstitution of a TFIIB-RXR interaction rather than indirect squelching of a repressor by TFIIB.
The Carboxyl Region of TFIIB Interacts with RXR-We next sought to define the specific regions of TFIIB required for interaction with mRXR␤ via GST binding assays (11). Glutathione-Sepharose beads were coated with intact, or truncated, GST-TFIIB fusion proteins and incubated with 35 S-labeled mRXR␤. As a control for binding specificity, 35 Slabeled luciferase was added to binding reactions. Labeled RXR did not bind to beads coated with GST alone nor did labeled luciferase bind to any of the GST-TFIIB-coated beads (Fig. 4A). Beads coated with the ⌬1, ⌬2, ⌬3, ⌬5, and dN1 GST TFIIB mutant proteins bound mRXR␤ as avidly as wild-type TFIIB. In contrast, beads coated with protein corresponding to the ␣1 (⌬178 -201) and dc3 (⌬238 -316) TFIIB deletion mutants bound less avidly to RXR, suggesting that regions from amino acids 178 -201 and 238 -316 of TFIIB are required for in vitro binding to mRXR␤. Under the conditions tested, addition of 9-cis-RA only slightly augmented in vitro binding (see also Fig. 5A).
To confirm these observations, we analyzed a series of TFIIB mutants with the dual-hybrid system. As shown in Fig. 4B, the ⌬116 mutant (which encompasses regions deleted with the TFIIB dN1 and ⌬1 mutants), ⌬2, and ⌬5 TFIIB mutants showed a slight reduction (30 -40%) in ligand-dependent reporter activity. In contrast, reporter activity was reduced to basal levels in the absence of TFIIB (GAL4 AD only) or with the dc3 and ⌬3 TFIIB mutants. The slight increase in ligand-dependent reporter activity observed in the presence of the GAL4-AD construct (that lacked TFIIB) resembled levels of activation observed with addition of RXR-GAL4-DBD alone (Fig. 3B), was not dependent upon TFIIB, and was attributed to association of RXR with endogenous yeast transcription factors. The dual-hybrid approach supported the conclusion that the N terminus of TFIIB (aa 1-175) was not required for interaction with RXR, in complete agreement with the GST binding data. Also in agreement with GST binding data, deletion of the C terminus of TFIIB (aa 238 -316, dc3) was associated with reduced activation in vivo. Results with the ⌬3 mu-

FIG. 3. In vivo interaction of RXR with TFIIB. A, schematic representation of yeast expression constructs used in dual-hybrid yeast assays
showing mRXR␤ cDNA fused to GAL4 DBD (aa 1-147), hTFIIB cDNA fused to GAL4 AD (aa 768 -881), and TFIIB GAL4 AD (Ϫ) construct. hTFIIB contains two direct repeats (arrows) and basic ␣-helix region (plus signs). Numbers above constructs refer to amino acid. Expression of fusion proteins in yeast was confirmed by immunoblot analysis (see below). B, interaction between TFIIB and mRXR␤ is ligand-dependent in yeast. Upper panels, yeast containing an integrated GAL1-lacZ reporter gene were transformed with expression plasmids (as indicated by ϩ) for RXR-GAL4-DBD (lanes 3, 4, and 6), TFIIB-GAL4-AD (lanes 2 and 4), or TFIIB-GAL4 AD(Ϫ) (a construct with no activation domain, lanes 5 and 6). As a control, yeast were transformed with GAL4-AD and GAL4-DBD constructs (indicated by Ϫ). ␤-Galactosidase activity (from the lacZ gene) of at least six independent yeast colonies grown in the absence (open bar) or presence of 1 M 9-cis-RA (black bar), normalized for protein, is shown in relative light units (mean Ϯ S.E.). Lower panels, immunoblot analysis of representative yeast extracts functionally analyzed (in lanes 1-6) were probed with a mouse monoclonal anti-RXR␤ (40) antibody 13.17 (anti-RXR) or a rabbit polyclonal antibody directed against amino acids 299 -316 of TFIIB (anti-TFIIB; Santa Cruz Biotechnology, Santa Cruz, CA) and incubated with secondary antibody (anti-mouse Ig or anti-rabbit Ig, peroxidase-linked antibody, respectively). Protein-antibody complexes were detected by enhanced chemiluminescence (ECL, Amersham Corp., UK) and exposed to autoradiography. Migration of 69-and 46-kDa molecular mass standards are indicated. tant were conflicting. The ⌬3 mutant clearly bound to RXR in vitro but may not have functioned as expected in the dualhybrid system due to a 2-fold increase in binding to RNAPII noted by Ha et al. (33) for this particular TFIIB mutant; strong binding to RNAPII may have squelched interaction with RXR and interfered with reconstitution of the system. If the results of the in vivo analysis are considered, the essential region of TFIIB required for interaction with RXR in yeast appears to be narrowed to aa 238 -271 of TFIIB, since both ⌬2 and ⌬5 mutants functioned well in vivo. In sum, mutation analysis of TFIIB identified aa 178 -201 and 238 -316 to be required for in vitro binding and aa 238 -271 of TFIIB to be required for in vivo association of TFIIB with RXR in yeast.
The C Terminus of RXR Interacts with TFIIB-To map the regions of RXR required for binding to TFIIB, we used intact and mutant GST-RXR fusion proteins immobilized on Sepharose beads (Fig. 5A). Beads coated with full-length GST-RXR fusion protein specifically bound labeled TFIIB (lane 2). No binding was observed to beads coated with GST alone. Also, minimal binding to TFIIB was observed with a ligand-binding domain RXR mutant (⌬LBD). Furthermore, an N-terminal deletion mutant (⌬AF-1) and an AF-2 deletion mutant of RXR (⌬390 -410) also bound labeled TFIIB, although the ⌬AF-1 mutant showed reduced binding in the absence of ligand. These data suggest that amino acids 254 -389 of RXR are required for in vitro binding to TFIIB.
Next, we tested the RXR deletion mutants using the dualhybrid system. To minimize variability in TFIIB expression between yeast strains, the respective RXR mutant plasmids were transformed into one parent yeast strain expressing satisfactory levels of TFIIB GAL4-AD, thus generating five distinct combinations, of which six individual yeast colonies of each combination were analyzed. Both full-length RXR (wild type) and the deletion construct lacking the DNA binding domain of RXR (⌬DBD) showed preservation of a ligand-dependent interaction between TFIIB and RXR (fold inductions of 14.5 and 29.4, respectively, Fig. 5B). In agreement with in vitro binding studies, the ⌬LBD mutant showed no ligand-dependent reporter activity and only a slight increase in basal reporter activity. The increase in basal reporter activity in the absence of ligand suggested relief of transcriptional repression, perhaps signifying loss of binding of a repressive factor to the C terminus of RXR (as observed for other nuclear receptors; see Refs. 16 -18). Deletion of the RXR N terminus showed a reduction in maximal levels of reporter activity but preservation of a liganddependent response (fold induction ϭ 11), consistent with the in vitro binding studies. Although the RXR mutant with deletion of the extreme C terminus (aa 390 -410) bound avidly to TFIIB in vitro (as was reported for VDR, see Ref. 12), the deletion mutant showed no increase in reporter activity, with or without ligand, in keeping with the dominant negative phenotype displayed by this mutant receptor in mammalian cells (55,56,70,71). With the sole exception of the ⌬390 -410 mutant, the dual-hybrid data are in agreement with the in vitro binding results and revealed that amino acids 254 -389 of RXR are essential for in vivo interaction with TFIIB.
In addition, we performed partial proteolysis assays (82,83) to determine whether RXR could alter proteolysis of TFIIB, since such an effect would support an in vitro association between the two molecules. Also, our mapping analysis (Fig. 5, A  and B) localized the region of TFIIB involved in binding with RXR to include a region common to interaction with VP-16 (84), and partial proteolysis assays suggested that VP-16 could induce a conformational change in TFIIB (82,83). As shown in Fig. 5C, digestion of TFIIB in the presence of baculovirus expressed recombinant RXR resulted in a pattern of cleavage FIG. 4. Analysis of regions within TFIIB required for binding to mRXR␤. A, in vitro binding analysis. In vitro translated, 35 S-labeled full-length mRXR␤ was incubated with glutathione-Sepharose beads coated with bacterially expressed GST-TFIIB fusion proteins (11), washed, and bound products resolved in SDS-PAGE and exposed to autoradiography. Proteins retained on GST-0 (bracket 1) or GST-TFIIB fusion protein (wild-type TFIIB; bracket 2) coated beads are as indicated. Alternatively, beads were coated with fusion proteins containing mutants of TFIIB: ⌬1 (⌬45-123) (bracket 3), ⌬2 (⌬118 -174) (bracket 4), ⌬3 (⌬202-269) (bracket 5), ⌬5 (⌬272-297) (bracket 6), dN1 (⌬4 -24) (bracket 7), ␣1 (⌬178 -201) (bracket 8), and dc3 (⌬238 -316) (bracket 9). Addition of 1 M 9-cis-RA is indicated by ϩ. Input proteins are one-tenth of added lysates. 35 S-Labeled luciferase was included in all reactions as a negative control. B, deletion analysis of TFIIB with wild-type RXR in the dual-hybrid system. Left panel, schematic representation of TFIIB mutants analyzed. ϩ indicates that yeast colonies had similar expression of RXR-GAL4-DBD by immunoblot analysis (data not shown). Right panel, reporter activity (␤-galactosidase) for each TFIIB mutant construct is indicated in relative light units (mean Ϯ S.E.) over background values, normalized for protein, and represents triplicate determinations from at least six independent colonies. Immunoblot analysis revealed differences in expression of the wild-type TFIIB, and TFIIB mutant fusion proteins did not account for the observed changes in reporter activity (data not shown). Far right column, fold induction is reported for each TFIIB mutant construct in the presence of ligand (1 M 9-cis-RA; black bars) versus control (vehicle; open bars) in the absence of ligand. Even in the absence of TFIIB, slight reporter activity was observed with GAL4-AD alone (GAL4 AD only). different from that observed with wild-type baculovirus control protein. An alteration in digestion pattern was also observed for TFIIB in the presence of bacterially expressed RXR but not GST control protein (not shown). These data are further support for the interaction of RXR with TFIIB and are consistent with either an induced conformational change in TFIIB or protection from proteolytic cleavage due to binding of RXR to TFIIB (82,83). DISCUSSION Several nuclear hormone receptors have been shown to interact in vitro with the general transcription factor, TFIIB (10 -14, 31), recruitment of which is thought to be a ratelimiting step in pre-initiation complex formation (7,33,34). The current view is that nuclear receptors interact with a complex of proteins involving co-activators, co-repressors, integrators, and other transcription factors (reviewed in Refs. 26, 27, and 65). In this report, we provide substantial evidence using both in vivo and in vitro methods to show that mRXR␤ may interact with hTFIIB in a direct and ligand-dependent manner to promote gene activation, an observation of significance due to the central role of RXRs in gene regulation in vivo (1,2,48,50,51,(61)(62)(63).
Specifically 132-257) of the LBD of the VDR. The results of Blanco et al. (12) contrast with the results of MacDonald et al. (13) who observed a ligand-independent interaction in vivo between the N terminus of TFIIB and the ligand-binding domain of VDR, an observation that was buttressed by site-directed mutagenesis studies (14). Interestingly, MacDonald et al. (13,14) also showed that TFIIB interacts with the extreme C terminus of the VDR, the latter region lacking in the ⌬390 -410 mRXR␤ mutant. In the case of the TR (at least in vitro) the N terminus of the hTR␤ has been shown to interact with the C terminus of TFIIB, whereas the C terminus of the LBD of hTR␤ interacts with the N terminus of TFIIB (11), as supported by experiments in a cell-free system (28). Baniahmad et al. (11) observed that two regions in the C terminus of TR (aa 168 -259 and 260 -465) interacted with TFIIB, and interaction of the latter region was not hormone-dependent. In comparison, Fondell et al. (31) showed that the C terminus (aa 213-410) of hTR␣ interacted with both TFIIB and TBP, whereas the N terminus (albeit weakly) and the C terminus of hTR␣ interacted with the C terminus of TFIIB (aa 244 -316). The findings with different TR isoforms suggest that different receptors may interact in an isoform-specific manner with TFIIB. Our inability to demonstrate involvement of the N terminus of mRXR␤ in binding to TFIIB may be specific for murine RXR␤, since mRXR␤ has not been found to possess an N-terminal amino acid region homologous to that in TR (36) that interacted with TFIIB.
Evidence also suggests that TFIIB may present different interfaces for binding to different nuclear receptors, activators, and co-activators. For instance, interaction of RXR, and some other nuclear receptors, does not involve regions of overlap with binding of TFIIB to TFIIF (RAP30) which appears to involve the first 111 aa of TFIIB (33). However, studies have mapped the binding motifs of TBP and RNAPII to regions of TFIIB (33) common to regions utilized in binding of TFIIB to RXR (Fig. 5) and other receptors (11,13,14,31,32). In addition, it has been shown that other sites in TFIIB are important for co-activator binding (6, 7, 82, 84 -87), such as for TAFII40 (aa 195-217) and VP16 (aa 165-217 and aa 269 -286) (85). That interfaces of TFIIB involved in binding to RXR differ (in part) from those observed for TR suggests that both RXR and TR might jointly contact TFIIB through distinctly different, but not mutually exclusive, binding interfaces within TFIIB. More definitive proof of this notion will require additional experiments. From the perspective of RXR, different ligands (e.g. LG100268 and LG1007540) have been shown to quantitatively and qualitatively affect interaction of RXR with TBP and TAF110 (67), suggesting that differential interaction of the receptor with basal transcription machinery could be dependent upon the conformation of the LBD (67). These findings support the idea that RXR may augment binding of the heterodimeric complex to the PIC. Thus, our results suggest that RXR may enhance the ability of TR or RAR (for instance) to recruit TFIIB to the PIC, in support of findings in cell-free transcription models (28).
It should be mentioned that MacDonald et al. (13) demonstrated that VDR interacts with TFIIB using a dual-hybrid yeast system but did not observe an interaction between hRXR␣ or mRXR␤ with TFIIB, although RXRs were not the principal focus of the study. The critical difference between the experiments of MacDonald et al. (13) and our experiments is likely that the interaction between RXR and TFIIB in yeast was entirely dependent on 9-cis-RA, a variable not reported in that study (13), and ligand has been noted to be important for binding to other transcription factors (29, 58, 67; see also below). Additionally, it is likely that other significant differences between the two model systems may account for the differences observed between results of MacDonald et al. (13) and our findings.
Our observations with TFIIB, and those of others with TBP and TAF II 110 (58,67), suggest that RXR interacts specifically with distinct components of the basal transcription machinery to augment transcription. In most instances, the in vivo mapping data agreed with in vitro findings, but the exceptions serve to elucidate distinct features of receptor function. First, the association of RXR and TFIIB was not ligand-dependent in vitro but was ligand-dependent in vivo. This observation was supported by results of Baniahmad et al. (11) and Tong et al. (28) with TR; likewise, we observed only a slight increase in binding of mRXR␤ to TFIIB in vitro in the presence of ligand. It is possible that failure of the bacterially expressed, recombinant RXR protein to assume a proper conformation or damage to (or proteolysis of) the receptor during purification affected our ability to detect ligand-dependent differences in vitro. In support of the necessity of ligand for association of RXR and TFIIB in vivo, a ligand-dependent interaction in yeast was also observed between RXR and TBP (58,67), CBP (29), and TAF110 (58,67). Overall, the necessity of ligand for in vivo interaction suggests that specific conformation is necessary in order for mRXR␤ to associate with TFIIB.
Second, we observed an interaction between the ⌬390 -410 RXR deletion mutant and TFIIB in vitro but not in vivo, suggesting that in vivo the extreme carboxyl region of RXR is required to permit interaction with TFIIB (Fig. 5B), as was observed for TBP (58). Since the ⌬390 -410 RXR deletion mutant has previously been shown to bind ligand (71), this suggests that ligand binding alone is not sufficient to promote interaction of RXR with TFIIB. Our results are consistent with dominant negative activity reported for the ⌬390 -410 mutant (55,56,66,69,70,71) and support other studies (66,69) suggesting that the mechanism involves binding of the receptor to DNA. In addition, our results suggest the mutant is unable to interact properly with TFIIB in vivo. Specifically, use of the dual-hybrid system to characterize this mutant strongly argues against other mechanisms, since DNA binding and nuclear localization of the ⌬390 -410 mutant to the promoter were ensured by the inclusion of the GAL-4-DBD in the construct. Others (9,15) have drawn similar conclusions based on observations with mutant estrogen receptor constructs and TFIIB. An alternate explanation for dominant negative activity of the ⌬390 -410 RXR deletion mutant in vivo could be that an AF-2 binding factor (e.g. CBP, TIF1, ERAP-160, or RIP 140) is required in vivo for stabilization of RXR interaction with TFIIB. In support of this alternate explanation, it has been suggested that yeast proteins, such as SUG1, interact with basal transcription factors TBP (84) and hTAF II 30 (15) as well as the AF-2 region of nuclear receptors (23,24), consistent with the notion that binding between TFIIB and RXR might be influenced by factors present in yeast. Further experiments are required to definitively distinguish between these two possibilities.
Since it had been shown that the acidic activator, VP-16, induced a conformational change in TFIIB (82,83) and bound to a similar region of TFIIB, we performed a partial proteolysis assay to test whether RXR may induce a conformational change within TFIIB. Using in vitro translated TFIIB and baculovirus/or bacterially expressed RXR proteins, we observed that RXR induced an altered proteolytic digestion pattern of TFIIB. These data suggest that one possible action of nuclear receptors may be to induce a conformational change in TFIIB which, as suggested for VP16, then exposes other binding interfaces of TFIIB to other transcription factors, possibly TAF II 40 (82, 85), TBP (9, 31, 86, 87), TFIID (8, 34, 85, 88), TFIIF, and RNAPII (89). These results raise the question of whether gene activation by nuclear hormone receptors might result not simply from a passive assembly of transcriptional proteins due to acidic, proline-rich, or glutamine-rich regions, but rather as the active alteration of protein interfaces of TFIIB which in turn present protein interfaces with affinity to other transcription factors, thus influencing assembly of the PIC.
In summary, the demonstration that RXR is capable of interacting with TFIIB in a ligand-dependent manner is consistent with the notion that RXR may contribute to recruitment of TFIIB to heterodimeric DNA-bound receptor complexes as well as homodimeric receptor complexes.