Overexpressed Activated Retinoid X Receptors Can Mediate Growth Inhibitory Effects of Retinoids in Human Carcinoma Cells*

Retinoic acid receptors (RARs) and retinoid X receptors (RXRs) mediate the effects of retinoids on gene expression by binding to response elements in retinoid-sensitive genes. RAR- but not RXR-selective retinoids were found in many previous studies to suppress the growth of various cells, implicating RXR-RAR in these effects. Using a co-expression vector for identifying cells that expressed retinoid receptors transiently and 5′-bromo-2′-deoxyuridine incorporation for labeling DNA-synthesizing cells, we found that RXR-selective retinoids inhibited DNA synthesis in squamous carcinoma 1483 cells transfected with RXRα but not with RARs. Ligand-induced transcription of the reporter luciferase gene via the activation of RXR-RXR but not RXR-RAR correlated with growth suppression. Studies with RXRα deletion mutants indicated that the DNA binding and the ligand binding domains are essential for mediating growth inhibition. A point mutation in the ligand binding domain (L430F) that decreased RXRα homodimerization compromised its growth inhibitory function. Further, RXRα mutant (F313A), which functions as a constitutively active receptor, inhibited DNA synthesis in the absence of ligand. These results demonstrate that RXR homodimer activation leads to growth inhibition and suggest that transfection of RXRα and treatment with RXR-selective retinoids or the transfection of constitutively activated RXRα mutant alone may have a therapeutic potential.

Retinoic acid receptors (RARs) 1 and retinoid X receptors (RXRs) are nuclear retinoid receptors, which belong to the superfamily of steroid hormone receptors (reviewed in Refs. [1][2][3]. Like other members of this superfamily, retinoid receptors have a conserved structure of domains A to F from the N terminus to the C terminus of the molecule, respectively. These domains have distinct functions and can act independently (1)(2)(3). The N terminus of the receptors (domains A and B) contains an autonomously functioning region called activation function 1 (AF-1), which is involved in ligand-independent transcriptional transactivation and is not well conserved among receptors. Domain C, a DNA binding domain composed of two class II zinc-binding motifs, is highly conserved. Domain D (hinge) is involved in ligand-induced functional changes and in the binding of receptors to co-repressors. Domains E/F, which are moderately conserved among the receptors, are thought to be involved in ligand-binding, ligand-dependent transactivation function 2 (AF-2) and dimerization (3). RARs bind both natural retinoids all-trans-retinoic acid (ATRA) and 9-cis-retinoic acid (9-cis-RA), whereas the RXRs bind only 9-cis-RA (1)(2)(3). Some synthetic retinoids can bind selectively to either RARs or RXRs or their ␣, ␤, or ␥ subtypes (4 -6).
Retinoic acid and retinoid X response elements (RARE and RXRE, respectively) are commonly composed of two half sites, each of which provides a binding site for one of the receptor molecules in a dimer. Each of the half sites is a conserved hexanucleotide DNA sequence, 5Ј-PuG(G/T)TCA-3Ј, and the two form direct repeats (DR) separated by one to five nucleotides (1)(2)(3). Both the relative orientation and spacing of the half sites is important for receptor recognition and for the subsequent activation or repression of the expression of target genes (12)(13)(14)(15). For example, in the presence of ligand, RXR-RAR heterodimers bind to and activate transcription on response elements consisting of two direct repeats separated by five nucleotides (DR-5), whereas RXR-RAR heterodimers bind direct repeats spaced by one nucleotide (DR-1) to elicit constitutive repression of gene transcription. In the presence of 9-cis-RA the RXR-RXR homodimer can activate gene transcription from the DR-1 response element (8). The function of retinoid receptors is also regulated by co-activators and co-repressors that distinguish among different conformations of dimer-DNA complexes induced by ligand binding and dimerization and consequently modulate positively or negatively the expression of target genes by retinoid receptors (3, 16 -18).
Retinoids were found to be effective inhibitors of cancer development in animal models of carcinogenesis and to suppress premalignant lesions and the development of second pri-* This work was supported by United States Public Health Service Grants P01 CA52051 (to R. L. and W. K. H.) and P01 CA51993 (to M. I. D.) and M. D. Anderson Cancer Center Core Grant P30 CA16672 from the National Cancer Institute, National Institutes of Health. 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.
ʈ To whom correspondence should be addressed: mary cancers (19). It is thought that some of these activities are the result of the ability of retinoids to inhibit the growth, induce apoptosis, and enhance the differentiation of various tumor cell types (20 -22). The potential of retinoids in cancer chemoprevention and treatment has spurred efforts to design novel retinoids with improved efficacy. One of the strategies in this effort was to synthesize retinoids that bind selectively individual retinoid receptor types. Among these novel retinoids, some RAR-selective analogues were found to be very effective in suppressing the growth of various tumor cells, whereas most of the RXR-selective analogues were reported to be ineffective in the same cell types (6,23,24). However, in some cells, RXR-selective analogs were found to enhance the effects of RAR-selective retinoids, a result that suggests that the RXR-RAR but not the RXR-RXR dimer mediates growth inhibition (26,27). Therefore, RXR-selective analogues seem to be less promising as single agents in chemoprevention and treatment of cancer.
The expression of certain RARs (e.g. RAR␤) is decreased in various tumor cell lines, and retinoid responsiveness could be restored by stably transfecting receptor expression vectors (28 -30). However, the isolation of stable transfectants expressing genes whose products exert growth-inhibitory effects may be difficult and, even if successful, may yield cells that were already partially or completely resistant to the inhibitory effect of this gene product. This problem can perhaps be solved by an inducible promoter or by transient expression of the gene. Whereas the transient transfection approach can provide results much faster than the stable transfection approach, its use is counterindicated when the efficiency of transfection is low. This limitation can be overcome by a method that allows the identification of the cells that express the transfected gene. A single cell proliferation assay has been developed to study the growth-inhibitory function of a candidate gene in a transient expression assay (31). This assay is based on the pMark vector engineered to co-express a nuclear retinoid receptor and a cell surface antigen (e.g., CD7), which permits identification of transfected cells bearing the antigen using a fluorescently labeled anti-CD7 antibody (31). Growth inhibition is then determined by labeling DNA-synthesizing nuclei with 5Ј-bromo-2Јdeoxyuridine (BrdU) followed by staining with anti-BrdU antibodies labeled with a different fluorophore (31). We used the pMark vector to express four nuclear retinoid receptors, RAR␣1, RAR␤2, RAR␥1, and RXR␣, in head and neck squamous carcinoma (HNSCC) 1483 cells and then determined the effects of the receptor expression on DNA synthesis in the cells grown in the absence or presence of different receptor-selective retinoids. After we found RXR␣ to be the most effective among these receptors in suppressing DNA synthesis, we focused on characterizing the effect of RXR␣ expression and the role of the different receptor domains on the growth and retinoid responsiveness of HNSCC 1483 cells.

EXPERIMENTAL PROCEDURES
Cell Culture and Retinoid Treatment-HNSCC 1483 cell line (32) was maintained in Dulbecco's modified Eagle's essential medium containing 10% fetal bovine serum. Cells were incubated at 37°C in humidified 5% CO 2 :95% air. Cells were detached by repeated pipetting after a brief incubation with 2 mM EDTA and 0.25% trypsin in a calcium-free and magnesium-free phosphate-buffered saline (PBS), pH 7. selective retinoids were synthesized as described previously (4,5,8,23,26). The structure and potency of these retinoids in transactivating reporter genes were published previously (summarized in Ref. 6). All retinoids were dissolved in Me 2 SO at a concentration of 10 Ϫ2 M and stored briefly in the dark at Ϫ20°C under N 2 . Stock solutions were diluted to the appropriate final concentrations in growth medium. Control cultures received the same amount of Me 2 SO alone.
Construction of Plasmids-Plasmids containing human cDNAs for RAR␣1, RAR␤2, RAR␥1, and RXR␣ were obtained from Dr. Magnus Pfahl (Sidney Kimmel Cancer Center, San Diego, CA). The cDNAs were released from pBluescript vectors by digesting with BamHI and HindIII or NotI and HindIII restriction enzymes and blunted with the Klenow fragment of DNA polymerase I and then inserted into the SmaI site of the plasmid pMarkCD7⌬5 (Ref. 31, produced by the Genetics Institute, Cambridge, MA), which was obtained from Dr. Jonathan Kurie (Department of Head and Neck/Thoracic Medical Oncology, The University of Texas M.D. Anderson Cancer Center, Houston, TX). The orientation of the inserted cDNAs was identified by restriction enzyme digestion. The resulting expression vectors were named pMarkRAR␣, pMarkRAR␤, pMarkRAR␥, and pMarkRXR␣, respectively. RXR␣ deletion mutants (shown schematically in Fig. 4A) were prepared in the plasmid pBSRXR␣, and the modified RXR␣ cDNAs were released and inserted into pMarkCD7⌬5. For example, the mutant RXR␣⌬A was prepared by deleting part of the 5Ј end of wild-type RXR␣ cDNA with HindIII and SmaI, and then the vector was blunted by T4 DNA polymerase and self-ligated. Mutant RXR␣⌬D was prepared by deleting the fragment between the first and the third NcoI sites (nucleotides  in wild-type RXR␣ cDNA. Mutant RXR␣⌬F was prepared by deleting the 3Ј end of RXR␣ from the first StuI site (nucleotide 402). RXR␣ point mutants are shown schematically in Fig.  5. pMarkRXR␣ F313A, a homologue of mouse RXR␣ F318A (36), was constructed using the QuickChange site-directed mutagenesis kit (Stratagene) according to the manufacturer's protocol. The primers used for generation of this mutation were as follows: sense primer, 5Ј-GCTGC TCATC GCCTC CGCCT CCCAC CGCTC CATCG C-3Ј; antisense, 5Ј-GCGAT GGAGC GGTGG GAGGC GGAGG CGATG AGCAG C-3Ј. pMarkRXR␣ L418F, pMarkRXR␣ L430F, and pMarkRXR␣ K431Q were prepared as follows: pBSRXR␣ L418F, pBSRXR␣ L430F, and pBSRXR␣ K431Q (37) (provided by Dr. Xiao-kun Zhang, Burnham Cancer Institute, La Jolla, CA) were digested with HindIII and NotI. The released fragments were blunted by Klenow and then ligated into the SmaI site of pMarkCD7⌬5. The success of obtaining the desired deletions and point mutations was assessed by sequence analysis.
Single Cell Proliferation Assay-The assay was performed as described by Frangioni et al. (31) with modifications. HNSCC 1483 cells were seeded at a concentration of 10 5 cells per well in 6-well plates. After 18 -24 h, cells were transfected with various pMark vectors using LipofectAMINE (Life Technologies, Inc.) following the manufacturer's instructions. Each well received 1 g of plasmid DNA and 6 l of LipofectAMINE. After 6 h, the transfection solution was removed by aspiration, and the cells were refed with medium containing 10% delipidized serum and the indicated concentration of retinoids or Me 2 SO control. After 36 h, the cells were incubated for 8. Transient Transfection and Luciferase Assays-Cells were seeded at a concentration of 1.5 ϫ 10 5 cells per well in 6-well plates. After overnight culture, cells in each well were transfected with 2 g of DNA including 1.5 g of reporter plasmid, 0.1 g of pCH110, and 0.4 g of pMark or pMark-derived expression vectors using 6 l of Lipo-fectAMINE (Life Technologies, Inc.) using the manufacturer's procedure. The reporter plasmids included the (RARE) 3 -tk-LUC, which contains three direct repeats of DR-5 RARE from the P2 promoter region of the human RAR␤2 gene (from Ϫ59 to Ϫ33 base pairs) (35) connected to the minimal herpes simplex virus thymidine kinase promoter and a luciferase cDNA and the RXRE-tk-LUC, which contains five tandem repeats of a 35-base pair sequence (DR-1) from the promoter of the mouse CRBP-II gene (Ϫ605 to Ϫ639) (38) inserted immediately upstream of tk-luciferase in a reporter plasmid (both provided by Dr. Richard A. Heyman, Ligand Pharmaceuticals, San Diego, CA). The pCH110, a ␤-galactosidase expression vector (Amersham Pharmacia Biotech), was used as the internal control for transfection efficiency. After a 6-h exposure to the transfection mixture, cells were treated in medium containing 10% fetal bovine serum and various concentrations of retinoid or Me 2 SO alone for 20 h and then harvested for measurement of ␤-galactosidase activity and luciferase activity using a luciferase assay system and protocol from Promega (Technical Manual TM033, Madison, WI) and a Lumat luminometer. Triplicate wells were used for each experimental group. Relative luciferase activity was normalized to the ␤-galactosidase activity to account for transfection efficiency.  8 -14). Gel shift and supershift analyses were performed as described under "Experimental Procedures." The name and the receptor specificity of the antibodies used are indicated above each lane. Shifted band and supershifted bands I, II, and III corresponding to RAR␤, RXR (antibody does not distinguish different isotypes), and RAR␥, respectively, are indicated by arrows on left. Although RAR␣ was not supershifted in the 1483 nuclear extract, the antibody AB9␣ was capable of supershifting RAR␣ in extracts of another HNSCC (data not shown). B, 9-cis-RA activates DR-5 RARE but not DR-1 RXRE. Cells were transfected with luciferase reporter constructs containing either the DR-5 or DR-1 response element and with a ␤-galactosidase expression vector as described under "Experimental Procedures." The cells were then treated with either 0.01% Me 2 SO (DMSO) or 1 M 9-cis-RA in Me 2 SO for 20 h. The cells were harvested and analyzed for luciferase and ␤-galactosidase activities. Data are presented as the activity of luciferase normalized to that of ␤-galactosidase, which served as a control for transfection efficiency. erodimer pathway functioned in the 1483 cells. Indeed, gel shift and supershift analyses performed with the RAR␤ DR-5 RARE (Fig. 1A) revealed that after 9-cis-RA treatment, RAR␤ (lane 11), RAR␥ (lanes 12 and 13), and RXR (lane 14) proteins complexed with DR-5. Complexes formed with DR-5 by proteins in nuclear extracts from untreated cells also contained RAR␥ (lanes 5 and 6) and RXR␣ (lane 7), but the level of RAR␤ was apparently too low to be observed in the supershift analysis (Fig. 1A, lane 4). A functional assay for endogenous receptors was performed by transient transfection of reporter plasmids containing either DR-5 RARE or DR-1 RXRE. 9-cis-RA at 1 M increased the transcription of luciferase via DR-5 RARE by about 10-fold but failed to enhance the transcriptional activity via DR-1 RXRE (Fig. 1B). These findings suggest that the endogenous RXR-RAR heterodimer is functional and can be activated by 9-cis-RA in the 1483 cells, whereas the endogenous RXR-RXR homodimer is not.

Endogenous Retinoid Receptors in HNSCC 1483 Cells Can
The Expression of an Exogenous RXR␣ Increases the Activation of RXR Homodimer and Suppresses DNA Synthesis in 1483 Cells after Treatment with 9-cis-RA-The formation of RXR-RXR homodimers requires high RXR levels (10,11). Thus, one reason for the lack of activation of the RXR homodimer pathway in the 1483 cells could be a low level of endogenous RXR. Therefore, we transiently transfected the 1483 cells with pMarkCD7⌬5 (pMark in short) vectors harboring RXR␣ as well as RAR␣1, RAR␤2, and RAR␥1 for comparison. We found that only cells transfected with RXR␣ exhibited increased growth inhibition after treatment with 1 M 9-cisRA, whereas the other transfected receptors failed to increase suppression of DNA synthesis by the same retinoid relative to cells transfected with pMark vector only (Fig. 2A). These results suggested that an increase in RXR homodimers but not RXR-RAR heterodimers contributes to growth inhibition by 9-cis-RA.
Further support for this conclusion has come from analysis of transcription activation experiments. Transfection of pMarkRXR␣ followed by treatment with 1 M 9-cis-RA induced luciferase transcription 10-fold via DR-1 RXRE as compared with almost no induction in pMark vector only transfected cells (Fig. 2B). Lower concentrations of 9-cis-RA (e.g. 0.1 or 10 nM) failed to activate DR-1 in RXR␣-transfected cells. A similar dose dependence was found for the effects of 9-cis-RA on the BrdU labeling index in transiently transfected cells. On treatment with 1 M 9-cis-RA, DNA synthesis decreased by 70% in cells transfected with RXR␣, whereas treatment with 0.1 or 10 nM 9-cis-RA had no effect (Fig. 2C). Thus, the transactivation of RXRE and inhibition of DNA synthesis seemed to be associated. The exogenous RXR␣ expression also increased luciferase transcription through DR-5 RARE, presumably by stimulating the RXR-RAR pathway; however, this increase was by less than 50% (Fig. 2B).

The Ability of RXR-selective Retinoids to Inhibit the Growth of 1483 Cells Is Related to Their Ability to Activate Transcription of DR-1 RXRE in RXR␣-transfected 1483 Cells-Because
9-cis-RA can bind both RARs and RXRs and activate transcription of both the RXR-RAR and the RXR-RXR pathways, we used RAR-selective TTNPB and several RXR-selective retinoids to determine whether they exert distinct effects on transcription of reporter genes and growth in 1483 cells transfected with pMarkRXR␣ or pMark vector alone. TTNPB, which binds to RARs with 10-fold higher affinity than ATRA but does not bind to RXRs (40), increased reporter gene transcription from DR-5 but not DR-1 in pMarkRXR␣-transfected cells (compare panels A and B in Fig. 3). 9-cis-RA increased DR-5 activity in both vector only-and RXR␣-transfected cells but strongly induced DR-1 activity only in RXR␣-transfected cells. RXR-selective SR11217, SR11203, SR11234, and SR11246 activated DR-1 9 -12-fold in RXR␣-transfected cells (Fig. 3B) but only increased DR-5 activity less than 2-fold (Fig. 3A). These results indicate that the RXR-selective retinoids activated the RXR-RXR pathway in RXR␣-transfected cells. RAR-selective TT-

FIG. 2. Transient expression of exogenous RXR␣ but not of RAR␣1, RAR␤2, or RAR␥1 in 1483 cells leads to growth suppression in response to 9-cis-RA and is associated with DR-1 RXRE activation.
A, cells were transiently transfected with pMark vector alone or pMark vector harboring RAR␣1, RAR␤2, RAR␥1, or RXR␣ and then treated with Me 2 SO (DMSO) or 1 M 9-cis-RA and labeled with BrdU. The DNA labeling index was then analyzed using a fluorescence microscope to determine the proportion of transfected cells with BrdUlabeled nuclei, as described under "Experimental Procedures." B, transcriptional activation by 9-cis-RA of reporter constructs containing luciferase via DR-1 RXRE or DR-5 RARE in cells transfected with empty pMark vector or pMark vector harboring RXR␣. Cells were cotransfected with either pMark vector or pMarkRXR␣ vector, the DR-1 or DR-5 reporter construct, and ␤-galactosidase. After transfection, cells were treated with Me 2 SO alone or different 9-cis-RA concentrations. The cells were harvested and analyzed for luciferase activity as described under "Experimental Procedures." C, dose-dependent effects of 9-cis-RA on DNA synthesis in 1483 cells transfected transiently with pMark vector alone or with pMarkRXR␣. Transiently transfected cells were treated with Me 2 SO or different 9-cis-RA concentrations and analyzed for inhibition of DNA synthesis as in A.
NPB failed to suppress DNA synthesis in RXR␣-transfected cells, whereas the RXR-selective retinoids inhibited DNA synthesis by 50 -80%, and their potency seemed to be related to their efficacy in DR-1 activation (Fig. 3C). These findings suggest that activating RXR-RAR pathway alone is not sufficient to attenuate cell growth, whereas activation of DR-1 (presumably by RXR homodimers) in RXR␣-transfected cells can inhibit growth.
RXR␣ Deletion Mutants Differ in Their Ability to Mediate Inhibition of DNA Synthesis after Treatment with 9-cis-RA-To investigate the importance of various regions of RXR␣ in inhibition of DNA synthesis by 9-cis-RA, we prepared deletion mutants of RXR␣ in the pMark vector (Fig. 4A). Their activities were compared with those of wild-type RXR␣ in transfected 9-cis-RA-treated cells. A mutant with deletion of part of domain A (RXR␣⌬A) was as effective as wild-type receptor in activating transcription from DR-5 but 35% less effective in activating transcription from DR-1 (Fig. 4, B1 and B2). The RXR␣⌬A mutant was less potent than the wild-type receptor in mediating suppression of DNA synthesis (54 and 72% inhibition for mutant and wild type, respectively) (Fig. 4B3). RXR␣ mutants with a deletion of a large part of the A/B domain and the DNA binding domain (RXR␣⌬D) or a 61-amino acid deletion at the C terminus (RXR␣⌬F) had decreased ability to activate DR-5 (the reason why the activity is not decreased to zero may be due to the presence of active endogenous RXR␣ (39)) and no ability to activate DR-1 (Fig. 4, B1 and B2). These mutants also lost the ability to suppress DNA synthesis in the presence of 9-cis-RA (Fig. 4B3). These findings, therefore, demonstrate a relationship between ability of RXR␣ to transactivate DR-1 and to mediate growth inhibition.
RXR␣ Mutants with a Reduced Ligand-induced Homodimerization Activity Decreased Growth Inhibition by 9-cis-RA-The amino acid 415-435 region in the human RXR␣ C terminus was critical for both homo-and heterodimerization as indicated by specific point mutations (37). To determine whether such mutations affect the ability of the receptor to mediate the growth inhibitory effects of 9-cis-RA in human HNSCC 1483 cells, we inserted each of three mutants, L418F, L430Q, and K431Q (Fig. 5A), into the pMark vector and analyzed their functions after transfection into the 1483 cells. Mutant L430F had lost all of the DR-1 activation potential (Fig. 5B2) and all of the ability to mediate inhibition of DNA synthesis by 9-cis-RA (Fig. 5B3) but still possessed about 50% of the DR-5 activation potential of wild-type RXR␣ (Fig. 5B1). Because the DR-5 activity in L430F-transfected cells was lower than in the pMark vector alone transfected cells, it seems that L430F interfered with the function of the endogenous RXR-RAR-mediated transcription via DR-5. The RXR␣ mutant L418F retained approximately 80 and 90% of DR-1 and DR-5 activation potential of wild-type RXR, respectively, and 70% of growth inhibition. The activities of the RXR␣ mutant K431Q were almost identical to those of the wild-type RXR␣. Thus, the findings with the mutants indicate that growth inhibition by RXR␣ can be modified when its DR-1 activation by homodimer formation is compromised. To a lesser extent, this is also true for modulation of DR-5 transactivation by the RXR␣ mutants.
The RXR␣ Mutant F313A Activates Transcription from DR-1 and Suppresses DNA Synthesis in 1483 Cells in a Ligandindependent Fashion-9-cis-RA was required for transcrip-tional activation via DR-1 and for growth inhibition by RXR␣ and its active mutants L418F and K431Q. A mouse RXR␣ mutation (F318A) was reported to produce a constitutively active RXR by mimicking ligand-induced conformational changes. This mutant was able to activate DR-1 in the absence of ligand in Cos-1 cells (36). We prepared the human homolog F313A (Fig. 5A) and cloned it into the pMark vector to determine its effects on 1483 cells. Transient transfection of pMark RXR␣F313A into 1483 cells increased transcriptional activity via the DR-5 in the absence of ligand by 6-fold relative to cells transfected with vector only (Fig. 5B1). Ligand-independent DR-5 activation was 66% of that observed in vector only transfected cells treated with 9-cis-RA. Thus, the mutant RXR␣ F313A can act as a functional partner for endogenous RARs in the absence of an RXR ligand. DR-5 reporter gene activation in 9-cis-RA-treated F313A-transfected cells was 1.7 times higher than in the absence of ligand. This increase can be due to the additive effect of the endogenous RXR␣ and the exogenous mutant F313A (Fig. 5B1). The activation of the DR-1 reporter gene in cells transfected with F313A was similar (about 10fold) in the absence or presence of 9-cis-RA and similar to that of transfected wild-type RXR␣ in the presence of 9-cis-RA (Fig.  5B2). These results suggest that F313A forms constitutively active homodimers in 1483 cells. Transient expression of F313A in the absence or presence of ligand resulted in about 70% suppression of DNA synthesis (Fig. 5B3), which was similar to that obtained in cells transfected with exogenous wildtype RXR␣ and treated with 9-cis-RA. These results suggest that RXR␣F313A homodimers can inhibit growth by mimicking the function of liganded wild-type RXR␣ homodimers. DISCUSSION Several studies have shown that RAR-selective retinoids effectively suppress tumor cell growth or induce cell differentiation, whereas RXR analogs are less effective (6,23,24). These findings suggested that the RXR-RAR pathway mediates the effects of retinoids in these cell systems, whereas RXR-RXR homodimers do not play a role. There are, however, reports that RXR-selective retinoids can mediate induction of certain genes (41)(42)(43). Some of these genes and other genes regulated by RXR signaling may be involved in growth control.
We found that retinoid signaling in the untransfected HNSCC 1483 cells is mediated primarily by the RXR-RAR heterodimer pathway as indicated by the activation by 9-cis-RA of DR-5 RARE but not DR-1 RXRE (Fig. 1B). Indeed, a DR-5 regulated gene, RAR␤ (35), was induced by 9-cis-RA in the 1483 cells as indicated by the appearance of a supershifted band using anti-RAR␤ antibodies after 9-cis-RA treatment of 1483 cells (Fig. 1B). This finding was also supported by our recent observation that both ATRA and 9-cis-RA induce RAR␤ mRNA and protein in the 1483 cells, 2 a process known to be mediated by activation of a DR-5 RARE by RXR-RAR heterodimers (35).
We found that 9-cis-RA activated transcription of reporter gene driven by a promoter containing the DR-5 but not the DR-1 response element (Fig. 1A). The lack of RXRE activation observed in other cells types, for example keratinocytes (44), may be due to low endogenous RXR-RXR levels and transrepression by RXR-RAR heterodimers that bind to the DR-1 with a higher affinity than RXR-RXR homodimers do but fail to activate this response element (16,44). Therefore, the relative abundance of RXR and RAR may determine whether the RXR-RXR signaling pathway can be activated by 9-cis-RA (8,9,14,15). One way to increase RXR␣ levels is to introduce an exogenous receptor expression vector by transfection. Indeed, overexpression of RXR␣ in human keratinocytes enabled RXRselective SR11237 to activate the RXRE via RXR-RXR homodimers (44). Unfortunately, such an approach has pitfalls. Selection of the transfectants during a prolonged subculture and cloning may result in an altered phenotype particularly if RXR␣ overexpression causes growth suppression. The selection process produces cells refractory to growth inhibition by RXR␣ or cells that express low levels of receptor. To avoid such complications, we employed the co-expression vector pMark (31) to enable the simultaneous transient transfection of RXR␣ and analysis of growth inhibition of the transfected cells by a single cell assay based on BrdU incorporation into replicating DNA.
The transient expression of exogenous wild-type RXR␣ in 1483 cells resulted in 9-cis-RA-mediated activation of DR-1 as indicated by enhancement of the luciferase reporter gene transcription ( Fig. 2A). The increased RXR␣ levels could have led to the formation of RXR homotetramers, which having a higher affinity for DR-1 than RXR-RAR may have displaced RXR-RAR heterodimers from the DR-1 RXRE (11). In the presence of 9-cis-RA, the DNA-bound tetramers may have dissociated to dimers (11) that activated DR-1-mediated transcription of the reporter. The expression of exogenous RXR␣ also modestly increased DR-5 activation by 9-cis-RA, possibly by increasing functional RXR-RAR heterodimer levels.
The DNA synthesis in 1483 cells was only minimally suppressed by 9-cis-RA despite the ability of this panagonist ligand (binding and activating both RARs and RXRs) to activate the DR-5. The activation of the DR-5 in the absence of an increase in DR-1 activity was observed in cells treated with RAR-selective TTNPB. This retinoid failed to suppress DNA synthesis (Fig. 3) and activate the DR-1. These results suggest that DR-5 activation is not sufficient to mediate growth inhibition in the parental 1483 cells and their RXR␣ transfectants. HNSCC 1483 cells expressing exogenous RXR␣ showed an increase in the activation of both DR-5 and DR-1 response elements by 1 M 9-cis-RA and also showed a decrease in DNA synthesis (Figs. 2 and 3). The four RXR-selective retinoids showed a tight correlation between their ability to activate DR-1 and to inhibit DNA synthesis in RXR␣-transfected cells. The relationship between DR-1 activation and DNA synthesis suppression was further supported by the findings using RXR␣ mutants. Specifically, deletions in RXR␣ (e.g. RXR␣⌬D and RXR␣⌬F) and the point mutation L430F abolished DR-1 activation and suppressed inhibition of DNA synthesis by 9-cis-RA, although some ability to activate DR-5 was retained (Figs. 4 and 5). Only a few reports have shown that transfection of RXR␣ enhanced the retinoid response. For example, RXR␣ overexpression in HL-60 leukemia cells increased apoptosis, whereas transfection of RAR␣ mediated induction of differentiation (45).
The mutation F318A in the ligand-binding pocket of mouse RXR␣ was found to cause the receptor to assume a conformation similar to that of agonist-bound wild-type RXR␣ and exhibit constitutive activation of the DR-1 (as a homodimer) and, to a lesser extent, a DR-5 reporter (as a heterodimer with RARs) in the absence of an RXR ligand (36). Although transcriptional regulation by this mutant has been characterized extensively, its biological activity has not been documented. Our homologous human RXR␣ mutant F313A was able to activate DR-1 effectively and DR-5 less effectively in the absence of ligand (Fig. 5). Interestingly, we found that expression of the mutant F313A inhibited DNA synthesis in 1483 cells in the absence of ligand. Thus, a constitutive activator of DR-1 is also a constitutive inhibitor of DNA synthesis. Constitutively active retinoid receptors including RXR␣ were produced previously by fusing the acidic activation domain of the herpes simplex viral protein VP16 to the C terminus of individual retinoid receptors (46); however, the biological consequences of the expression of this chimeric RXR␣ have not been demonstrated. A more re-cent report described the fusion of the VP16 transactivation domain (88 amino acids from the C terminus) to the N terminus of RAR␣ or the C terminus of RXR␣ and the expression of these constitutively active chimeric proteins by transfection of recombinant adenoviral vectors (47). Although effective in activating DR-1 RXRE, the constitutively active RXR␣VP16 failed to induce differentiation of HL-60 leukemia cells and NTera-2 teratocarcinoma cells, whereas VP16RAR␣ promoted the ligand-independent differentiation of both (47). These results are consistent with the observations that RXR-RAR pathway is dominant in mediating differentiation programs. The authors suggested that the chimeric constitutive receptors may be used as tumor suppressor genes for genetically based treatment of retinoid-responsive cancers (47). However, their data did not support the similar use of the RXR-specific signaling pathway. In contrast, we found that overexpression of RXR␣F313A inhibited DNA synthesis. Our novel observation suggests that the RXR-RXR pathway may mediate growth inhibition that is possibly unrelated to induction of differentiation and that RXR overexpression can also be used to design new gene therapies.
It is not clear how overexpression of RXR␣ followed by 9-cis-RA treatment or expression of the constitutively active F313A causes inhibition of DNA synthesis in 1483 cells. Presumably, the overexpressed RXR␣ forms tetramers or dimers that are activated by RXR-selective ligands or by an activating mutation (i.e. F313A) to increase the transcription of genes that possess DR-1 RXREs, which are involved in regulation of cell growth by controlling DNA synthesis. In addition, overexpression of RXR␣ may lead to the formation of tetramers that could displace RXR-RAR heterodimers from natural DR-1 RXREs and thereby relieve the silencing of gene transcription. The identification of genes that are regulated by RXR homodimers may lead to the understanding of the mechanism by which RXR homodimers suppress growth. To date, only a few genes were shown to be regulated by RXR-selective retinoids, presumably via RXR binding to natural RXREs. Such genes include growth hormone in pituitary cells (41), ␣-fetoprotein in hepatocytes (42), and cholesterol 7␣-hydroxylase in HepG2 cells (43). Future studies will be necessary to determine whether RXR homodimers can directly regulate any genes involved in cell growth.
In conclusion, the results of several experimental approaches indicate that activation of the DR-1 RXRE is associated with ability of 1483 cells expressing exogenous RXR␣ to respond to 9-cis-RA with diminished DNA synthesis. Similar effects were also observed in four of five other human head and neck and lung cancer cell lines (data not shown). Thus, it may be possible to develop several therapeutic strategies based on our findings (for example, treatment with RXR-selective retinoids that activate endogenous RXRs to form RXR-RXR homodimers rather than heterodimers). Such an approach may be possible because the conformational changes in RXR required for homodimerization and heterodimerization can be separately modified (36,37). Such novel retinoids could retain the minimal side effects characteristic of the currently available RXR-selective retinoids compared with the more deleterious RAR-selective retinoids (25,48). Another strategy could be to activate the RXR-RXR pathway using gene therapy based on transfer of the RXR␣ gene followed by treatment with an RXR-selective retinoid or gene transfer of the constitutively activated RXR␣F313A without the need to treat with a retinoid. the RXR␣ point mutants, Dr. Jonathan Kurie for the pMark vector, and Dr. John Clifford for instructive comments on the manuscript.