Bcl3, an IκB Protein, as a Novel Transcription Coactivator of the Retinoid X Receptor*

We have recently shown that the IκB protein IκBβ interacted with the retinoid X receptor (RXR) and inhibited the 9-cis-retinoic acid (RA)-dependent transactivations (Na, S.-Y., Kim, H.-J., Lee, S.-K., Choi, H.-S., Na, D. S., Lee, M.-O., Chung, M., Moore, D. D., and Lee, J. W. (1998) J. Biol. Chem. 6, 3212–3215). Herein, we show that a distinct IκB protein Bcl3 also interacts with RXR, as shown in the yeast two-hybrid tests and glutathioneS-transferase pull-down assays. The Bcl3 interaction involved two distinct subregions of RXR, i.e. constitutive interactions of the N-terminal ABC domains and 9-cis-RA-dependent interactions of the C-terminal DEF domains. In contrast to IκBβ, Bcl3 did not interact with the AF2 domain of RXR. Bcl3 specifically interacted with the general transcription factors TFIIB, TBP, and TFIIA but not with TFIIEα in the GST pull-down assays. TBP and TFIIA, however, were not able to interact with IκBβ. Accordingly, Bcl3 coactivated the 9-cis-RA-induced transactivations of RXR, in contrast to the inhibitory actions of IκBβ. In addition, coexpression of SRC-1 but not p300 further stimulated the Bcl3-mediated enhancement of the 9-cis-RA-induced transactivations of RXR. These results suggest that distinct IκB proteins differentially modulate the 9-cis-RA-induced transactivations of RXR in vivo.

The nuclear receptor superfamily is a group of transcriptional regulatory proteins linked by a series of conserved structure and function (for a review see Ref. 1). These include six regions of the primary structure, commonly referred to as regions A through F. The N-terminal half of the receptor contains the A/B region, which is highly variable in sequence and length. Region C, the DNA-binding domain, harbors two type II zinc fingers and is highly conserved, whereas region D, also called the hinge domain, is highly variable in length and sequence. The C-terminal half of the receptor contains the E and F regions, which harbor the ligand-binding domain (LBD) 1 and the C-terminal activation function 2 (AF2), respectively. The superfamily includes receptors for a variety of small hydrophobic ligands such as steroids, T3, and retinoids, as well as a large number of related proteins that do not have known ligands, referred to as orphan nuclear receptors (for a review see Ref. 2). The receptor proteins are direct regulators of transcription that function by binding to specific DNA sequences named hormone response elements in promoters of target genes. Transcriptional activation of nuclear receptors involves at least two separate processes: derepression and activation (2). Repression is mediated in part by interaction of unliganded receptors with corepressors such as N-CoR (3) and SMRT (4). However, ligand binding triggers dissociation of these corepressors and concomitant recruitment of coactivators. These putative coactivators include RIP-140/RIP160 (5,6), ERAP-140/ERAP-160 (7), TIF1 (8), TRIP1 (9), ARA70 (10), and CBP/p300 (11)(12)(13), as well as a group of highly related proteins, SRC-1 (11,14), AIB1 (15), TIF2 (16), RAC3 (17), ACTR (18), TRAM-1 (19), p/CIP (20), and xSRC-3 (21). Functional analysis of nuclear receptors has shown that there are two major activation domains. The Nterminal domain (AF1) contains a ligand-independent activation function, whereas the extreme C-terminal region (AF2) of the LBD exhibits ligand-dependent transactivation (1). Recent x-ray crystallographic studies of the LBD of nuclear receptors revealed that the ligand binding induces a major conformational change in the AF2 region (22)(23)(24), suggesting that this region may play a critical role in mediating transactivation by a ligand-dependent interaction with coactivators. These coactivators are postulated to function to transmit the signal of ligand-induced conformational change to the basal transcription machinery. As expected, many coactivators fail to interact with AF2 mutants of nuclear receptors (5,8,16).
The transcription factor NFB is important for the inducible expression of a wide variety of cellular and viral genes (reviewed in Refs. 25 and 26). NFB is composed of homo-and heterodimeric complexes of members of the Rel/NFB family of polypeptides. In vertebrates, this family comprises p50, p65 (RelA), c-Rel, p52, and RelB. These proteins share a 300-amino acid region, known as the Rel homology domain, which binds to DNA and mediates homo-and heterodimerization. This domain also is target of the IB inhibitors, which include IB␣, IB␤, IB␥, Bcl3, p105, and p100 (27). In the majority of cells, NFB exists in an inactive form in the cytoplasm, bound to the inhibitory IB proteins. Treatment of cells with various inducers results in the degradation of IB proteins. The bound NFB is released and translocates to the nucleus, where it activates appropriate target genes. IB␣ is degraded in response to all of the known inducers of NFB, whereas IB␤ is degraded only when cells are stimulated with inducers such as lipopolysaccharide and interleukin-1, which cause persistent activation of NFB (28). Following degradation of the initial pool of IB␤ in response to lipopolysaccharide or interleukin-1, newly synthesized IB␤ accumulates in the nucleus as an unphosphorylated protein that forms a stable complex with NFB and prevents it from binding to newly synthesized IB␣ (29,30). Bcl3 is an unusal IB protein in that it can not only inhibit nuclear NFB complexes but can bind to p50 and p52 dimers on DNA and provide the complexes with transactivating activity (31,32).
Recently, we have shown that IB␤ specifically interacted with the retinoid X receptor (RXR) and inhibited its 9-cis-RAdependent transactivation in lipopolysaccharide-treated cells (33). These results led us to examine whether another IB molecule Bcl3 is also capable of functionally interacting with RXR. Herein, we show that Bcl3 indeed interacts with two subregions of RXR. In contrast to IB␤, however, Bcl3 did not interact with the AF2 domain of RXR. Furthermore, Bcl3 and IB␤ showed different interaction profiles with general transcription factors. Accordingly, Bcl3 coactivated the 9-cis-RAinduced transactivations of RXR in cotransfections with CV1 cells, alone, or in synergy with SRC-1, in contrast to the inhibitory actions of IB␤. From these results, we propose that distinct IB proteins may differentially modulate the 9-cis-RAdependent transactivations of RXR in vivo.
Yeast Two-hybrid Test-For the yeast two-hybrid tests, plasmids encoding LexA fusions and B42 fusions were cotransformed into Saccharomyces cerevisiae EGY48 strain containing the LacZ reporter plasmid, SH/18-34 (34). Plate and liquid assays of LacZ expression were carried out as described (35). Similar results were obtained in more than two similar experiments.
GST Pull-down Assays-The GST fusions or GST alone were expressed in Escherichia coli, bound to glutathione-Sepharose-4B beads (Amersham Pharmacia Biotech), and incubated with labeled proteins expressed by in vitro translation using the TNT-coupled transcriptiontranslation system, with conditions as described by the manufacturer (Promega, Madison, WI). Specifically bound proteins were eluted from beads with 40 mM reduced glutathione in 50 mM Tris (pH 8.0) and analyzed by SDS-polyacrylamide gel electrophoresis and autoradiography, as described (37).
Cell Culture and Transfections-CV1 cells were grown in 24-well plates with medium supplemented with 10% charcoal-stripped serum for 24 h and transfected with 100 ng of the LacZ expression vector pRSV-␤-gal and 100 ng of a reporter gene TREpal-LUC, along with 10 ng of RXR␣ and increasing amounts (10 to 200 ng) of Bcl3, SRC-1, or p300 expression vectors. For control experiments, Gal4-LUC and Gal4-VP16 (36) replaced TREpal-LUC and RXR, respectively. Total amounts of expression vectors were kept constant by adding decreasing amounts of the CDM8 expression vector. After 12 h, cells were washed and refed with Dulbecco's modified Eagle's medium containing 10% charcoalstripped fetal bovine serum either in the presence or absence of 10 Ϫ7 M 9-cis-RA. Cells were harvested 24 h later, and luciferase activity was assayed as described (38), and the results were normalized to the LacZ expression. Similar results were obtained in more than two similar experiments.

The 9-cis-RA-stimulated Interactions of Bcl3 and RXR Do
Not Involve the AF2 Domain-In the yeast two-hybrid tests (34), the full-length IB␤ and Bcl3 interacted with the LBD of RXR in a 9-cis-RA-stimulated manner (Ref. 33 and Table I) but not with the LBD of glucocorticoid receptor (data not shown). A series of deletion mutants were constructed to localize the interaction interfaces (Fig. 1). As shown in Table I, B42 fusion to the LBD of RXR (B42/RXR-LBD) interacted with IkB␤⌬1, IB␤⌬2, IkB␤⌬3, and IkB␤⌬5 but not with IkB␤⌬4, localizing the RXR interaction interface to the IB␤ amino acids 253-312. Similarly, B42/RXR-LBD interacted with Bcl3⌬2 and Bcl3⌬3 but not with Bcl3⌬1, localizing the RXR interaction interface to the Bcl3 amino acids 289 to the C terminus. The IB␤-RXR interactions involved the AF2 domain of RXR, as demonstrated by loss of the 9-cis-RA dependence in interactions of IB␤ and its deletional mutants with B42/RXR-LBD⌬AF2, a B42 fusion to the previously described mutant RXR-LBD that lacks the AF2 function (39) ( Table I). These results are consistent with our previous report (33), in which IB␤ was shown to interact with the AF2 domain of TR. In contrast, however, B42/RXR-LBD⌬AF2 still retained the 9-cis-RA-dependent interactions with Bcl3, Bcl3⌬2, and Bcl3⌬3, demonstrating that the Bcl3-RXR interaction did not involve the AF2 domain, distinct from the IB␤ results (33). In addition, LexA fusions to IB␤, IB␤⌬4, Bcl3, and Bcl3⌬2 were transcriptional activators of the LacZ reporter gene controlled by upstream LexA-binding sites (34). Thus, the autonomous transactivation domains of IB␤ and Bcl3 in yeast were mapped to the N-terminal 90 amino acids of IB␤ and the Bcl3 amino acids 156 -289, respectively (Table I and Fig. 1).
To further characterize these interactions in vitro, various GST fusion proteins were expressed, purified, and tested for interaction with an in vitro translated Bcl3. These include a full-length RXR, the LBDs of wild type RXR and deletion mutant for the AF2 domain (RXR-LBD and RXR-LBD⌬AF2), the N-terminal ABC domains of RXR (RXR-ABC), and full-length TRs, wild type or point-mutated for the AF2 domain (TR and TR459, respectively) (33,39). As shown in Fig. 2, Bcl3 constitutively interacted with either the full-length RXR or RXR-ABC. In agreement with the yeast results, however, week basal interactions of Bcl3 with RXR-LBD or RXR-LBD⌬AF2 were The indicated B42 and LexA plasmids were transformed into a yeast strain containing an appropriate LacZ reporter gene. At least six separate transformants from each transformation were transferred to indicator plates containing 5-bromo-4-chloro-3-indolyl-␤-D-galactopyranoside, and reproducible results were obtained using colonies from a separate transformation. ϩϩϩ, strongly blue colonies after 2 days of incubation and strong interaction; ϩϩ, light blue colonies after 2 days of incubation and intermediate interaction; ϩ, light blue colonies after more than 2 days of incubation and weak interaction; Ϫ, white colonies and no interaction.
significantly stimulated by the presence of 1 M 9-cis-RA. These results suggest that RXR contains at least two interaction interfaces with Bcl3: a constitutive interaction interface at the N-terminal region of RXR and a 9-cis-RA-dependent interaction interface at the LBD that doesn't involve the AF2 domain. Similarly, Bcl3 also interacted with TR and TR459, and addition of 1 M T3 further stimulated these interactions. These results are in marked contrast to the previously described IB␤ results (33), in which the AF2 mutants TR⌬ and TR459 showed relatively weak, T3-independent interactions with IB␤. These results, along with the yeast results, confirm that IB␤ and Bcl3 exploit distinct regions of RXR to interact with. In particular, IB␤ utilizes the intact AF2 domain of RXR, whereas Bcl3 exploit yet unspecified non-AF2 region of the RXR-LBD for the 9-cis-RA-dependent interactions. Different Interactions of Bcl3 and IB␤ with General Transcription Factors-IB␤ showed autonomous transcriptional activities in yeast and mammalian cells (Table I and data not shown). Bcl3 was also shown to have two separate domains that are able to cooperate to confer transactivation functions to otherwise inactive p50 homodimers (31). In our hands, Bcl3 showed autonomous transactivation functions in yeast (Table  I), whereas both LexA and Gal4 fusions to Bcl3 efficiently repressed the LexA-LUC and Gal4-LUC reporter gene expressions, respectively (data not shown). Thus, we examined whether IB␤ and Bcl3 are capable of directly interacting with known general transcription factors by using the GST pulldown assays. As expected, the p50-p65 interactions, along with the p50-Bcl3 and p65-IB␤ interactions, were readily detected in this assay system (Fig. 3). In addition, p50 interacted with TFIIB but not with TFIIE␣, as described previously (40,41), whereas p65 interacted with TBP and TFIIA. The latter results further extend previous findings (40 -42) in which p65 was shown to interact with TFIIB and TBP. Bcl3 also interacted with TFIIB, TBP, and TFIIA but not with TFIIE␣ (Fig. 3). In contrast, TBP and TFIIA were not able to interact with IB␤, attesting to differences between IB␤ and Bcl3 in communicating with the RNA polymerase II preinitiation complex. These results, along with the differences of IB␤ and Bcl3 in recognizing RXR (Table I and Fig. 2), also provide a mechanistic basis for the differential regulations of the RXR transactivations by IB␤ and Bcl3 (see the following cotransfection results).
Bcl3 Stimulates the 9-cis-RA-induced Transactivations of RXR-To assess the functional consequences of these interactions, Bcl3 was cotransfected into CV1 cells along with an RXR expression vector and a reporter construct controlled by TREpal, which is transactivated by RXR-RXR homodimers as well as various receptor heterodimers (43). Increasing amounts of cotransfected Bcl3 enhanced the 9-cis-RA-induced transcription of RXR (without significantly affecting the basal level) in a dose-dependent manner, with cotransfection of 50 ng of Bcl3 increasing transcriptional activities approximately 4.5-fold (Fig. 4A). In contrast, cotransfection of Bcl3 affected neither the transcriptional activity of Gal4-VP16, as assessed using the Gal4-LUC reporter construct (36) (data not shown), nor the LacZ expression of the transfection indicator construct pRSV-␤-gal in the presence or absence of 9-cis-RA (data not shown).
In an effort to dissect the mechanistic role of Bcl3, we investi-  Table I  gated whether two known transcription coactivators of nuclear receptors, SRC-1 (11,14) and CBP/p300 (11,12,44,45) were capable of cooperating with Bcl3 in this coactivation. Coexpression of CV1 cells with SRC-1 or p300 stimulated the 9-cis-RAdependent transactivations of RXR in a dose-dependent manner, as reported (11,12,14,44,45). Interestingly, coexpression of SRC-1 significantly stimulated the Bcl3-enhanced 9-cis-RAdependent transactivations of RXR in a dose-dependent manner, with cotransfection of 50 ng of Bcl3 and 100 ng of SRC-1 increasing the 9-cis-RA-dependent transcriptional activities approximately 9-fold (Fig. 4B). However, the addition of increasing amounts of p300 did not significantly alter the 9-cis-RA-induced level of transcription obtained by Bcl3 alone or Bcl3 plus SRC-1 (Fig. 4B, compare the results with Bcl3 alone or Bcl3 plus SRC-1, in either the presence or absence of p300). These results clearly demonstrate that Bcl3 is capable of cooperating with SRC-1 but not with p300 to coactivate the RXRmediated transactivations.
In an effort to explain the inability of p300 to cooperate with Bcl3, we examined whether Bcl3 and p300 compete to bind to RXR (Fig. 4C). GST-RXR specifically retained radiolabeled Bcl3, which was competed away by increasing amounts of nonlabeled Bcl3, as expected. In contrast, the radiolabeled RXR bound by CBP1, the N-terminal region of CBP harboring the receptor interaction domain, was not competed by non-labeled Bcl3. These results indicate that Bcl3 and CBP/p300 do not mutually exclude each other in binding to receptors. DISCUSSION Cross-communications between distinct signaling pathways that lead to combinatorial controls are becoming a common theme in the area of transcriptional regulations and could involve a complex array of different mechanisms. The mutual antagonism between nuclear receptors and Jun/Fos (referred to as AP-1), for example, has been suggested to involve direct protein-protein interactions between nuclear receptors and Jun/Fos (for a review, see Ref. 46), as well as a rather indirect competition for a limiting amount of essential coactivators CBP/p300 (11)(12)(13). These factors have been shown to be essential for the activation of transcription by a large number of regulated transcription factors (thus, appropriately named as "integrators"), including nuclear receptors (11, 12, 44, 45), AP-1 (47,48), CREB (49 -51), bHLH factors (52), and STATs (53,54). In addition, nuclear receptor-mediated blockage of the induction of the Jun N-terminal kinase (i.e. JNK) signaling cascade was recently suggested to be responsible for this antagonism (55).
The mutually antagonistic interactions have been described for the NFB component p65 and a subset of nuclear receptors including glucocorticoid receptor (56 -59), estrogen receptor (60,61), and progesterone receptor (62). In addition, glucocorticoids, at least in certain cell types, have been shown to increase the synthesis of IB␣, which should then sequester NFB in an inactive cytoplasmic form (63). Recently, CBP and p300 were shown to coactivate the NFB component p65 (64,65), whereas SRC-1, originally identified as a transcription coactivator of nuclear receptors, was found to interact with p50 and coactivate the NFB transactivations (66). Thus, competition for a limiting amount of CBP and SRC-1 may represent yet another key mechanism to explain the mutual antagonisms of nuclear receptors and NFB. Adding more twists to the theme, an IB molecule IB␤ was recently shown to directly interact with RXR and function as its transcriptional corepressor (33), whereas another IB molecule Bcl3 functions as a coactivator of the nuclear receptor RXR, as described in this report. These results attest to a common theme, in which an increasing number of transcription regulators will turn out to participate in regulation of other transcription factors that were previously thought to be unrelated with each other, revealing more diverse regulatory circuits in the eukaryotic transcriptional controls.
The basis for the unexpected IBs-mediated cross-talks between nuclear receptors and NFB derives from the fact that we have discovered the cryptic interaction interfaces of IB␤ and Bcl3 with nuclear receptors. These interfaces were localized to the IB␤ amino acids 253-312 and the Bcl3 amino acids from position 289 to the C terminus (Table I). IB␤ contains six ankyrin repeats that constitute the interaction interface with the Rel homology domain of NFB, whereas Bcl3 contains seven ankyrin repeats (Fig. 1). Among these, ankyrin repeats 1, 5, and 6 of IB␤ and ankyrin repeat 3 of Bcl3 contain a single amino acid motif LXXLL, recently shown to mediate the ligand dependent interaction of the AF2 transactivation domain of the FIG. 4. Effects of Bcl3 cotransfection on the transcriptional activities of RXR. A and B, CV1 cells were transfected with 100 ng of LacZ expression vector and increasing amount of Bcl3, SRC-1, and p300 expression vectors along with a reporter gene TREpal-LUC and 10 ng of RXR␣, as indicated. Closed boxes indicate no hormone added; hatched boxes indicate 0.1 M 9-cis-RA added in the transfection media. Normalized luciferase expressions from triplicate samples are presented relative to the LacZ expressions, and the standard deviations are less than 5%. C, Bcl3 and RXR labeled with [ 35 S]methionine by in vitro translation were incubated with glutathione beads containing GST alone or GST fusions to RXR and CBP1. Beads were washed, and specifically bound material was eluted with reduced glutathione and resolved by SDSpolyacrylamide gel electrophoresis. Only the results in the presence of 9-cis-RA are shown, as the results without ligand added were similar (data not shown).
receptors with transcription cofactors such as RIP-140, SRC-1, and CBP (20,67). However, point mutations of these motifs were not able to disrupt the interactions of RXR with IB␤ and Bcl-3. 2 Consistent with these results, the Bcl3-RXR interaction did not involve the AF2 domain (Table I and Fig. 2).
IB␤ and Bcl3 showed autonomous transcriptional activities in yeast (Table I). The autonomous transactivation functions of IB␤ and Bcl-3 in yeast were mapped to the N-terminal 90 amino acids of IB␤ and the Bcl-3 amino acids 156 -289 (Table  I and Fig. 1). These results are in contrast to a previous finding (31), in which both the N-terminal and the C-terminal domains of Bcl-3 were shown to cooperate to coactivate otherwise inactive p50 homodimers in mammalian cells. The reason for this discrepancy is not currently known. However, it is likely to have reflected some fundamental differences between yeast and mammalian transcription machinery. In mammalian cells, IB␤ also showed autonomous transactivation functions, whereas both LexA and Gal4 fusions to Bcl3 efficiently repressed the LexA-LUC and Gal4-LUC reporter gene expressions (data not shown), attesting to the differences between IB␤ and Bcl3 to communicate with the RNA polymerase II preinitiation complex. Consistent with this, Bcl3 interacted with TFIIB, TBP, and TFIIA, whereas TBP and TFIIA did not interact with IB␤. Possible interactions of IB␤ with other general transcription factors are currently under investigation. These results, along with the differences of IB␤ and Bcl3 in recognizing RXR (Table I and Fig. 2), may also provide mechanistic basis for the differential regulations of the RXR transactivations by IB␤ and Bcl3 (Ref. 33 and Fig. 4). Finally, it was intriguing that Bcl3 cooperated with SRC-1 but not with p300 to coactivate the RXR transactivations (Fig. 4B). It is notable that SRC-1 was shown to form a complex with CBP/ p300 (45). Thus, Bcl3 may have overlapped functions with CBP/p300, or alternatively, its function may involve selective recruitment of SRC-1. Accordingly, it will be interesting to examine whether Bcl3 can exist as a single complex with these factors (i.e. SRC-1 and CBP/p300) or form a different kind of activation complex, particularly with respect to IB␤. Interestingly, Bcl3 and CBP did not exclude each other from binding to RXR (Fig. 4C), consistent with the former possibility. In particular, the inability of CBP/p300 to synergize with Bcl3 to coactivate the RXR transactivations may reflect enough expressions of CBP in the CV1 and HeLa cells we used. Thus, in certain cell types, we may be able to observe synergistic coactivation of the RXR transactivations by Bcl3 and CBP.
In conclusion, we have shown that distinct IB molecules are capable of differentially modulating the RXR transactivations. The IB␤ results are consistent with the fact that lipopolysaccharide and interleukin-1, mediators of the IB␤ actions, are pro-inflammatory (68), whereas retinoids are anti-inflammatory (69). The biological significance for the Bcl3-RXR interactions is not currently known. It is notable, however, that Bcl3 was cloned by virtue of its recurrent translocations in a subset of B cell chronic lymphocytic leukemias (70,71) and its expressions are cell type-specific, most prominent in spleen, liver, and lung (32). Thus, Bcl3 is likely to significantly affect the RXR functions in these tissues, and RXR may also play critical roles in the potential tumorigenesis mediated by Bcl3.