Differential 9-cis-Retinoic Acid-dependent Transcriptional Activation by Murine Retinoid X Receptor (RXR) and RXR ROLE OF CELL TYPE AND RXR DOMAINS

The 9-cis-retinoic acid (9cRA)-inducible enhancer of the rat cellular retinol-binding protein type II gene (CRBP II) was shown to be differentially regulated by the murine retinoid X receptor α (RXRα) as compared with RXRβ. Transient transfection assays performed in NIH 3T3 fibroblast cells demonstrated that RXRα yielded a high level of 9cRA-dependent transcription of a reporter gene linked to the CRBP II enhancer, when compared with RXRβ. This effect was cell type-dependent, since both receptors elicited comparable transcriptional activation of the same reporter in P19 embryonal carcinoma cells. To further explore the structural determinants responsible for the differences between these two receptors, a series of chimeric receptor constructs were made. Co-transfection assays utilizing these chimeras demonstrated that both the N terminus and the hinge region connecting the DNA binding domain with the ligand binding domain of RXRα were responsible for the high level of 9cRA-dependent transcription observed in NIH 3T3 cells. Furthermore, the hinge region of RXRα was shown to be necessary to repress, in the absence of hormone, the transcriptional activation function located in the N-terminal domain of RXRα. These results stress the importance of functional links between different RXR domains and suggest an RXR subtype and cell type-dependent specificity in the control of the 9cRA response.

The 9-cis-retinoic acid (9cRA)-inducible enhancer of the rat cellular retinol-binding protein type II gene (CRBP II) was shown to be differentially regulated by the murine retinoid X receptor ␣ (RXR␣) as compared with RXR␤. Transient transfection assays performed in NIH 3T3 fibroblast cells demonstrated that RXR␣ yielded a high level of 9cRA-dependent transcription of a reporter gene linked to the CRBP II enhancer, when compared with RXR␤. This effect was cell type-dependent, since both receptors elicited comparable transcriptional activation of the same reporter in P19 embryonal carcinoma cells. To further explore the structural determinants responsible for the differences between these two receptors, a series of chimeric receptor constructs were made. Co-transfection assays utilizing these chimeras demonstrated that both the N terminus and the hinge region connecting the DNA binding domain with the ligand binding domain of RXR␣ were responsible for the high level of 9cRA-dependent transcription observed in NIH 3T3 cells. Furthermore, the hinge region of RXR␣ was shown to be necessary to repress, in the absence of hormone, the transcriptional activation function located in the N-terminal domain of RXR␣. These results stress the importance of functional links between different RXR domains and suggest an RXR subtype and cell type-dependent specificity in the control of the 9cRA response.
The molecular mechanisms by which members of the steroid and thyroid hormone receptors superfamily regulate the expression of target genes are still unclear. RXRs, 1 in particular, exert pleiotropic functions due to their ability to heterodimerize with a variety of other receptors within this family, including retinoic acid receptors, thyroid hormone, and vitamin D receptors. Thus, RXRs are assumed to play a central role in modulating the cellular response to multiple signaling pathways (1)(2)(3)(4)(5)(6)(7). In addition, RXR homodimers can transactivate RXRresponsive elements (RXREs) such as that found within the rat CRBP II gene (8 -10).
A further element of complexity is added by the presence of multiple subtypes of the receptors (␣, ␤, and ␥) that show dissimilar transcriptional activities on the same response element (11). In addition, murine RXR receptor subtypes have been shown to be differentially expressed during development (see Refs. 12-14 for reviews). However, the structural feature(s) involved in subtype-specific gene regulation are unknown.
Nuclear hormone receptors have a modular structure that consists of at least five regions (denoted A-E in Fig. 2; for reviews see Refs. [13][14][15]. The most evolutionary conserved regions are the DNA binding domain (DBD, region C) and the ligand binding domain (region E). The C region directly contacts their cognate response elements, while the E domain is responsible for ligand binding, transcriptional activation function (AF2), and homo-and heterodimerization (1-7, 13-15, 19). The D domain (hinge region) exhibits much less sequence similarity among receptor subtypes and has recently been shown to play a role in the specificity and polarity of binding of receptor heterodimers to DNA response elements (14 -18, 24, 25). The N-terminal or A/B domain appears to be partially responsible for specific differences among retinoic acid receptor subtypes (11,19). This region contains a transactivation domain function (AF-1), which can synergize with AF-2, located in the ligand binding domain (reviewed in Ref. 14).
In this report we demonstrate that RXR␣ and RXR␤ differentially transactivate the RXRE from the rat CRBP II promoter in NIH 3T3 fibroblast cells but similarly in P19 embryonal carcinoma cells. We have tested a series of chimeric receptors to elucidate the domains responsible for the differences in transcriptional activation. Our results showed that both the A/B and D domains are involved in dictating the output of the cell-dependent transcriptional response. Hence, these two regions cooperate with each other and potentially with other cell-specific factors to maintain the specificity of the response, thus providing the functional elements necessary for eliciting subtype-specific and cell type-specific transcriptional regulation.

EXPERIMENTAL PROCEDURES
Plasmids-The reporter construct CRBPII(2)tk-luc, containing two copies of the CRBP II RXRE in front of the thymidine kinase promoter driving the luciferase coding sequence, was described previously (9). RSVH-2RIIBP (7), expressing the mouse RXR␤, was utilized to generate the chimeras as described below. The RXR␣ clone was a generous gift from Drs. Ron Evans and David Mangelsdorf (12).
Construction of Chimeras-Five-part chimeric receptors were created from the original cDNAs coding for RXR␣ and RXR␤ (7,12). The sites used were either existing restriction sites or sites created by PCR introducing silent mutations. All sites and regions were amplified by * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. § Present address: Genetic Therapy, Inc., Gaithersburg, MD 20878. ¶ Both authors contributed equally to this manuscript. ‡ ‡ To whom correspondence should be addressed. Tel.: 301-496-0944; Fax: 301-402-0387. 1 The abbreviations used are: RXR, retinoid X receptor; RXRE, RXRresponsive element; CRBP II, cellular retinol-binding protein type II; DBD, DNA binding domain; PCR, polymerase chain reaction; bp, base pair. PCR using Pfu polymerase (Stratagene). Each chimeric exchange was made in regions of high conservation, and no novel amino acids were introduced. PCR fragments were first cloned into pCRscript (Stratagene) and sequenced utilizing an ABI DNA sequencer. Subcloning was then performed in pSP72 to characterize the constructs by in vitro transcription-translation (6). Following confirmation, the full-length chimeric genes were cleaved from the pSP72 construct by digestion with HindIII/HpaI and cloned into the eukaryotic expression vector RSVneo cleaved with HindIII/HpaI, resulting in replacement of the neo sequence with that of the receptors. When it was necessary to clone into the HincII site, the HpaI site in pSP72 was eliminated by insertion of an oligonucleotide containing SmaI, DraI, and SnaB1 restriction sites, and one of them was utilized to clone in the HpaI site of RSVneo (6). For the RXR␣ receptor the restriction sites used to make the five-part chimeras were a HindIII site (artificially created) at base pair (bp) 85 of the original cDNA, a PvuII site at bp 560, a HincII site at bp 738, an XbaI site at bp 818, a BamHI site at bp 964, and a HpaI site at bp 1532. The chimeric parts were named ␣1, ␣2, ␣3, ␣4, and ␣5 (see Fig. 2). The correct PvuII site for RXR␣ was selected by partial digestion. RXR⌬␣␣[␣␣␣] was made by deleting the region between the HindIII site and PvuII site of the RXR␣ construct and replacing it with the AUG start codon and 12 nucleotides upstream from position 315 of RXR␣. For the RXR␤ receptor, the corresponding sites used were the HindIII site created by PCR immediately upstream of the native NcoI site at bp 155, a PvuII site at bp 400, a HincII site at bp 578, an XbaI site at bp 652, a BamHI site at bp 808, and the HpaI site at bp 1360. The chimeric parts were named ␤1, ␤2, ␤3, ␤4, and ␤5 (see Fig. 2).
Transfections-Transfections into NIH 3T3 cells were performed as described (20). Briefly, for NIH 3T3 cells 10 g of reporter plasmid was electroporated into 4.4 ϫ 10 6 cells with 12.5 g of native or chimeric receptor plasmid. Following the transfections, the cells were cultured for 60 h with 1 M 9cRA (a gift from Dr. Arthur Levin at Hoffmann-La Roche). P19 cells were transfected by using LipofectAMINE (Life Technologies, Inc.) as described (22). Following an overnight incubation with the DNA/LipofectAMINE mixture, 9cRA (1 M) was added, and the cells were incubated for an additional 24 h in the presence of the hormone. After lysing the cells, luciferase activity assays were performed as described (21) and normalized by utilizing a ␤-galactosidase expression vector included in the transfection. Results represent the average of a minimum of three independent transfections performed in duplicate.

RESULTS AND DISCUSSION
Several reports have suggested that there may be RXR subtype-specific differences in 9cRA-dependent transcriptional activation on particular response elements (11,12,19). We compared the transcriptional activation mediated by RXR␣ and RXR␤ in transient transfection assays in NIH 3T3 cells cotransfected with either receptor and a reporter construct containing the luciferase reporter gene linked to the CRBP II RXRE and the thymidine kinase basal promoter (Fig. 1). In the presence of 9cRA, RXR␣ was 20-fold more efficient in transactivating the CRBP II response element than RXR␤. Western immunoblot analysis performed after transfection of RXR␣ and RXR␤ constructs containing C termini tagged with flag epitope demonstrated that the difference in transcriptional activation could not be attributed simply to differences in the levels of expression of these receptors (data not shown). We also examined the transcriptional activation in P19 cells. Interestingly, both RXR␣ and RXR␤ showed very similar levels of transcriptional activation in P19 cells (Fig. 1).
In order to better understand the structural requirements underlying the differences in transcriptional activation observed in NIH 3T3 cells, we constructed a series of chimeras between these two receptors (Fig. 2). Each receptor was subdivided into five parts. Parts 1, 2, 3, 4, and 5 correspond to the A/B, C, N and C termini of D, and the E domain, respectively. The expected molecular masses for the chimeric receptors were demonstrated following separation of in vitro transcriptiontranslation products by SDS-polyacrylamide gel electrophoresis (data not shown). Finally, the functional integrity of each receptor was confirmed by transfection of P19 embryonal carcinoma cells. Each chimera in P19 cells showed comparable levels of transcriptional activation, differing less than 2-fold, in the presence of 9cRA on the CRBP II response element (Fig. 3).
Next, we tested each chimera in NIH 3T3 cells to identify the functional determinants responsible for the differences between the two receptors (Fig. 4). As shown in Fig. 1, RXR␣ was approximately 20-fold more active than RXR␤ under identical conditions of transfection (compare lane 1 with 2 in Fig. 4A). This difference was independent of part 2 (the C domain) since the exchange of the RXR␣ part 2 with the corresponding part 2 of RXR␤ (RXR␣␤[␣␣␣]) was as efficient as the RXR␣ (compare lane 1 with 15 in Fig. 4A). Similarly, the RXR␤ part 2 substituted for the RXR␣ part 2 in the context of the RXR␤ receptor (RXR␤␣[␤␤␤]) (compare lane 2 with 9). This result was expected, since the two receptors are more than 95% homologous in the C domain corresponding to DBD ( Fig. 2A). However, it was not possible to completely rule out any involvement of part 2, since RXR␤␤[␣␣␣] gave lower transactivation than RXR␤␣[␣␣␣] (compare lane 13 with 14), suggesting that the C domain may contribute to optimal folding.
In agreement with a previous report (23), our results point out that the A/B domain of RXR␣ is absolutely required for achieving the high level of transcriptional activation observed with RXR␣␣[␣␣␣], since, when partially deleted (RXR⌬␣␣[␣␣␣]), a 4-fold decrease in transcriptional activation was observed (Fig. 4A, compare lane 1 with 3). In addition, the A/B domain (part 1) of RXR␤ could not functionally substitute for that of RXR␣ (RXR␤␣[␣␣␣]) (compare lane 1 with 13). These results confirm an earlier report (19) 2 with 14). Taken together, these results imply that both the A/B domain and the C terminus of RXR␣ are necessary for high transcriptional activation of the CRBP II RXRE in NIH 3T3 cells.
Next, we defined the functional part of the RXR␣ C terminus involved in the observed effects by testing chimeras that exchanged both the D (parts 3 and 4) and E (part 5) domains. . Taking into account that the E domains are fully exchangeable, these results strongly imply that the D and A/B domains function in concert to achieve optimal transcriptional activation in NIH 3T3 cells. Based on these results, we further dissected the D domain to identify the portion responsible for this effect. Region D includes the A box and ends at the N terminus of the E domain (Fig. 2, B and C). The A box is a seven-residue region that was originally shown to be critical for binding of the orphan receptor NGFI-B to its response element (14 -18, 24). The resolution of the three-dimensional structure of the RXR DBD has shown that the A box is included in a helical region beginning after the second zinc finger of the DBD (25). This helix includes also the so-called T box that forms a dimerization interface critical in the binding of homo-and heterodimeric DBDs to some response elements (14 -18, 24, 25).
New chimeras, which exchanged the N terminus, including the A box (part 3), or the C-terminal region (part 4) of the D domain (Fig. 2C), were tested for their transcriptional activation. When part 4 of RXR␤ was used to replace the corresponding portion in RXR␣ (RXR␣␣[␣␤␣]), less than 2-fold reduction in overall reporter activity was observed (compare lanes 1 and 4 in Fig. 4A). However, a more dramatic effect was seen when part 3 of RXR␤ was substituted for the analogous region of RXR␣ (RXR␣␣ [␤␣␣]). Although the transactivation in the presence of hormone was reduced less than 2-fold (Fig. 4A, compare  lanes 1 and 5), a 10-fold increase in transactivation in the absence of hormone was detected (Fig. 4B, compare lanes 1 and  Fig. 4B). In addition, when the E domain (part 5) of the ␣ receptor was substituted by the E domain of the ␤ receptor (RXR␣␣[␣␣␤]), an increase of basal activity, although weaker than the chimeras described above, was observed (compare lane 7 with 1, Fig. 4). Nevertheless, the hormone-independent enhancement of basal activity observed in these chimeras demonstrates the functional interaction of the D domain and to some degree the E domain with the A/B domain. Therefore, regions 3 in the D domain and the E domain must be from RXR␣ to achieve repression of the N-terminal AF-1 in the absence of hormone (Fig. 4A, lane 1).
The importance of the RXR␣ part 3 for 9cRA-dependent maximal activity in NIH 3T3 cells, in the absence of an increase in basal activity, was also demonstrated using chimeras containing the RXR␤ A/B domain. When construct 11 (RXR␤␣[␤␣␤]) was tested in NIH 3T3 cells, an 8-fold 9cRA-dependent induction was observed (Fig. 4A, lane 11). However, when part 3 of this construct was replaced with the corresponding part from RXR␣ (RXR␤␣[␣␣␤]), a further increase of the reporter activity (up to 70-fold 9cRA-dependent induction) was observed (lane 12 compared with lane 11). Although the levels of reporter activity observed in the presence of RXR␤␣[␣␣␤] is much lower than the levels observed with RXR␣␣[␣␣␤], the latter containing the A/B domain of RXR␣, it demonstrates again the importance of part 3.
The fact that the D domain of RXR␣ is involved in both the inhibition of AF1 in the absence of hormone and in hormonedependent inducibility adds to the growing list of regulatory functions attributed to this region. Other than the T and A boxes, immediately 3Ј of the A box is a 70-amino acid region, which has been shown to be responsible for the divergent transcriptional activities between the RAR␣, -␤, and -␥ subtypes in response to various retinoids (25).
The data presented here also support the conclusion that there is an AF1 present in RXR␣ but not in RXR␤ (11,19). Interestingly, Chen and Privalsky (23) have provided evidence that RXR␣ binds to the CRBP II RXRE as a tetramer, while the A/B domain of RXR␤ is involved in inhibiting the tetrameric binding of RXR␤ to the CRBP II response element. Differences in DNA binding properties between the receptors have been invoked as the cause for the different transactivation levels observed in SL-2 cells (23).
Our data indicate that, in addition to the intrinsic DNA binding properties of the receptors, interactions with a cellspecific factor(s) or cell type-specific post-translational modifications may be involved, given the differences in response mediated by RXR␣ and RXR␤ observed in P19 and NIH 3T3 cells. Finally, it should be emphasized that, although this study addresses the activities of RXR homodimers, a differential role of RXR subtypes and cell type-specific differences in transcriptional activation might also be found in the regulation of gene expression mediated by RXR heterodimerization with other nuclear receptors.