Increased Retinoic Acid Responsiveness in Lung Carcinoma Cells that Are Nonresponsive Despite the Presence of Endogenous Retinoic Acid Receptor (RAR) β by Expression of Exogenous Retinoid Receptors Retinoid X Receptor α, RARα, and RARγ¹

Retinoids, a group of natural and synthetic vitamin A analogues,include compounds that suppress carcinogenesis in experimental animals(1, 2) and have shown promise as chemopreventive and therapeutic agents (2, 3) presumably by modulating the growth, differentiation, and apoptosis of normal, premalignant, and malignant cells (2, 3, 4, 5). Retinoids can regulate gene expression by activating nuclear retinoid receptors, which are members of the steroid hormone receptor gene superfamily and function as ligand-dependent DNA-binding transcription enhancing factors (6, 7). Each of the two subtypes of retinoid receptors,RARs⁵and RXRs, includes three isotypes designated α, β, and γ. The RARs bind both ATRA and 9-cis-RA, whereas the RXRs bind only 9-cis-RA. Each RAR and RXR subtype can be expressed in several isoforms (e.g., RARβ1, RARβ2), which differ in their NH₂-terminal domain as a result of alternative promoter usage and splicing (6). RARs can form heterodimers with RXRs, and RXRs can also form homodimers. Such dimers can bind to RAREs in the regulatory regions of certain target genes(6, 7). The RAREs consist of DRs PuG(G/T)TCA(X)nPuG(G/T)TCA with one or five intervening nucleotides (X) or closely degenerate motifs(6). Activation of transcription by RAR-RXR and RXR-RXR dimers is usually mediated via DR5 (RARE) and DR1 (RXRE), respectively(6, 7). Each receptor isotype and isoform can regulate a distinct subset of retinoid-responsive genes because deletion of individual receptors by homologous recombination resulted in loss of induction by RA of different genes (6, 7, 8, 9). However,studies in mice by knockout of single or multiple receptors have demonstrated that receptors may possess both specific and redundant functions (6). Transcription regulation by retinoid receptors is determined by interplay of cofactors with opposite effects. Corepressors bind to complexes formed between retinoid receptors and response element and suppress transcriptional activation. However, ligand binding causes corepressors to dissociate and coactivators to associate with the retinoid receptors and activate the transcriptional machinery (6, 10).

Retinoid receptors contain six domains designated A to F. The NH₂ terminus (domains A and B) includes a ligand-independent activation function (AF-1). The following C domain contains a highly conserved DNA-binding domain, which may also participate in protein-protein interaction with cofactors. The D domain is involved in ligand-induced functional change and is critical for the binding of receptor to corepressors. The E and F domains, which are moderately conserved among receptors, are involved in ligand binding and include a ligand-dependent transactivation function (AF-2) and a dimerization surface (6).

Endogenous and transfected retinoid receptors can antagonize the function (transrepression) of AP-1 (11, 12, 13), a complex comprised of dimers of members of the Jun and Fos family of DNA-binding proto-oncogenes that mediate mitogenic signals from a variety of growth factors and tumor promoters (14).

Abnormalities in the expression or function of retinoid receptors have been found in various cell types (15, 16). Decreased expression of RARα in breast cancer (17) and of RARβin lung cancer cells in vitro (18, 19, 20, 21, 22, 23, 24) and in vivo (25, 26, 27), in head and neck cancers(28, 29), and in breast cancer cells in vitro(17, 30) and in vivo (31, 32) has been proposed to result in resistance to the effects of retinoids on cell growth and differentiation and to enhance the development of certain malignancies (15, 16). Transfection of exogenous receptors into some of these cells restored responsiveness to certain effects of retinoids (16, 17, 19, 33, 34, 35, 36, 37).

Although RARβ expression decreases in lung cancer cells and tissues(18, 19, 20, 21, 22, 23, 24, 25, 26, 27), about 50% of lung tumors express this receptor(25), suggesting that the latter cancers develop despite the presence of RARβ. Likewise, certain lung cancer cell lines were found to express RARβ, yet they resisted the growth-inhibitory effects of ATRA (20, 24, 38). It is possible that the downstream steps in RARβ signaling are defective in such tissues and cells (23, 24) or that the expression or function of other retinoid receptors is abnormal (16).

In the present study, we attempted to restore responsiveness to the growth-inhibitory effects of retinoids in NSCLC H1792 adenocarcinoma cells, which express RARβ but are resistant to the growth-inhibitory effects of ATRA, by transfecting expression vectors for different receptors. We demonstrated that overexpression of RARα, RARγ, or RXRα but not RARβ resulted in the suppression of the growth of H1792 cells to different degrees, in a ligand-dependent fashion. We further demonstrated that the growth-inhibitory effects of nuclear retinoid receptors were closely correlated with their ability to antagonize AP-1 transcriptional activity.

The H1792 cell line was obtained from Dr. Adi Gazdar (University of Texas Southwestern, Dallas, TX). The Calu-1 NSCLC cell line was purchased from the American Type Culture Collection (Manassas, VA). The cells were maintained in DMEM containing 10% FBS. In some experiments,delipidized serum (free of endogenous retinoids) was used. Cells were incubated at 37°C in a humidified atmosphere of 5%CO₂:95% air.

The retinoids ATRA, 9-cis-RA, and TTNN were obtained from Dr. Werner Bollag (Hoffmann La Roche, Ltd., Basle,Switzerland). AM80 was obtained from Dr. Koichi Shudo (University of Tokyo, Tokyo, Japan). SR11217 was obtained from Dr. Marcia Dawson(Stanford Research Institute International, Menlo Park, CA). CD2314, CD2325, and CD437 were obtained from Dr. Braham Shroot (Centre Internationale de Recherche Dermatologique/Galderma, Sophia Antipolis, France). LG69 was obtained from Dr. Richard Heyman (Ligand Pharmaceuticals, San Diego, CA). The receptor selectivity of these retinoids and their transactivation potencies have been reported elsewhere (24). All of the retinoids were dissolved in DMSO at a concentration of 10 mm and stored briefly under N₂ in the dark at −20°C. The stock solutions were diluted to the desired final concentrations in growth medium. Control cultures received the same amount of DMSO as retinoid-treated cultures.

The assay was performed exactly as described by us elsewhere(37).

Plasmids containing human cDNAs for RARα1, RARβ2, RARγ1, and RXRα were obtained from Dr. Magnus Pfahl (Sidney Kimmel Cancer Center, San Diego, CA). To increase translation efficiency, we deleted the 5′ region and introduced a new Kozak sequence into the RARβ cDNA to form the RARk12β construct (37). cDNA fragments containing the entire open reading frames of the different receptors were inserted into plasmid pMARKCD7D5 (36) obtained from Dr. Jonathan Kurie (The University of Texas M. D. Anderson Cancer Center).

RXRα deletion mutants (shown schematically in Fig. 5 A)were inserted into pMARKCD7D5. Specifically, the mutant RXRαDA was prepared by deleting part of the 5′ end of wt RXRα cDNA with HindIII and SmaI. Mutant RXRαDD was prepared by deleting a fragment between nucleotides 29 and 197 in wt RXRα cDNA. Mutant RXRαDF was prepared by deleting the 3′ end of RXRα from nucleotide 402. The RXRα point mutant pMARKRXRαF313A was constructed using the QuickChange site-directed mutagenesis kit(Stratagene, San Diego, CA). The point mutant constructs pMARKRXRαL430F and pMARKRXRαK431Q were prepared from pBSRXRαL430F and pBSRXRα K431Q (39), respectively (provided by Dr. Xiao-kun Zhang; Burnham Cancer Institute, La Jolla, CA).

Cells were seeded at a concentration of 2 × 10⁵ cells/well in 6-well plates. After an overnight incubation, cells in each well were transfected with 1.2 μg of DNA (0.7 μg of luciferase reporter plasmid described below, 0.1μg of the β-galactosidase expression vector pCH110, and 0.4 μg of pMARKCD7D5 or pMARKCD7D5-derived expression vectors) in 6 μl of LipofectAMINE (Life Technologies, Inc., Gaithersburg, MD) following the manufacturer’s procedure. Two reporter plasmids were used:(a) (RARE)₃-tk-LUC, which contains three tandem repeats of DR5 RARE from the P2 promoter region (−59 to−33 bp) of the human RARβ2 gene (40) connected to the minimal herpes simplex virus tk promoter and a luciferase cDNA;and (b) RXRE-tk-LUC, which contains five tandem repeats of DR1 RXRE, a 35-bp sequence (−605 to −639) from the promoter of the mouse CRBP-II gene (41) inserted immediately upstream of tk-luciferase in a reporter plasmid (both provided by Dr. Richard A. Heyman). The reporter plasmid for analysis of AP-1 transcriptional activity (Col-AP-1-LUC) was obtained from Dr. Jonathan Kurie. The promoter of this reporter construct was derived from the matrix metalloproteinase 1 promoter (−73 to +63), which contains the AP-1 binding site TGAGTCA (42). The pCH110 β-galactosidase expression vector (Pharmacia Biotech, Piscataway, NJ) was used at 0.1μg for transfection as the internal control for transfection efficiency. After a 6-h exposure to the transfection mixture, cells were treated in medium containing 10% FBS and various concentrations of retinoid or DMSO alone for 20 h and then harvested for measurement of β-galactosidase activity and luciferase activity using a luciferase assay system and protocol from Promega (Technical Manual TM033; Promega, Madison, WI) and a Lumat luminometer. Triplicate wells were used for each experimental group. Relative luciferase activity was related to the β-galactosidase activity to normalize for transfection efficiency.

The single cell proliferation assay was performed by a modification of the method described by Frangioni et al. (36). H1792 cells were seeded at a concentration of 10⁵cells/well in 6-well plates. After 18–24 h, cells were transfected with various pMARKCD7D5 vectors using LipofectAMINE. Each well received 1 μg of plasmid DNA and 6 μl 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 DMSO control. After 36 h, the cells were incubated for 7 h with a labeling reagent containing 10μ m BrdUrd and 1μ m 5′-fluoro-2′-deoxyuridine (Amersham,Arlington Heights, IL) to incorporate BrdUrd into DNA in cells engaged in DNA synthesis. Cells were then washed three times with PBS and fixed with absolute methanol (prechilled to −20°C) for 10 min. Cells were then rehydrated with PBS and washed once in water. Chromosomal DNA was depurinated by treatment with 2 m HCl for 15 min at room temperature. The acid was neutralized by one wash with 0.1 mNa₂B₄O₇(pH 8.5), followed by a 2-min incubation in the same solution. Cells were then washed twice with 0.1% NP40 in PBS (0.1% NP40/PBS) and incubated for 1–2 h at room temperature with anti-BrdUrd mAb (IgG1;Becton Dickinson, San Jose, CA) and anti-CD7 mAb (IgG2b clone 3A1E-12H7; Sera-Lab, Sussex, United Kingdom), both of which were diluted 1:6 in 0.3% BSA and 0.1% NP40/PBS. The cells were washed five times with 0.1% NP40/PBS and incubated for 45 min at room temperature with Texas Red-conjugated goat antimouse IgG2b and FITC-conjugated goat antimouse IgG1 (Southern Biotechnology, Birmingham, AL) diluted 1:200 in 0.1% NP40/PBS. Cells were washed four times with 0.1% NP40/PBS and twice with PBS. The cells were then observed using an immunofluorescence microscope with filters for the red fluorescence of Texas Red in the cytoplasm and on the cell surface and the green fluorescence of FITC in the cell nuclei. Cells that have taken up the plasmid and synthesized DNA are stained both red and green(RG), and their nuclei appear yellow because of the overlap of the two colors. Cells that have taken up the plasmid but failed to synthesize DNA stained only red (R). The BrdUrd labeling index (the percentage of cells synthesizing DNA among the cells that have taken up the plasmid) was determined using the following formula:[(RG/(R + RG)]100. Usually, >500 cells were analyzed in several arbitrarily chosen microscopic fields.

Previously, we found that NSCLC H1792 cells express abundant amounts of RARβ mRNA compared with other NSCLC cells, in which no RARβ mRNA could be detected (24). We found no mutations in RARβtranscripts from H1792 cells after amplification by reverse transcription-PCR and sequencing (data not shown). High levels of RARγ, RXRβ, and RXRγ mRNAs were detected in H1792 cells, whereas RXRα and RARα transcripts were about 80% and 60% less abundant,respectively, in H1792 cells than in other NSCLC cells(24).

Electrophoretic mobility shift and supershift assays shown in Fig. 1 revealed two major shifted bands (Lane 1), which appeared to be specific for RARE because they could be competed by excess unlabeled DR5 RARE oligonucleotide (Fig. 1, compare Lanes 1 and 13), but not by a mutated oligonucleotide (Fig. 1, Lane 14). Complexes containing RARβ and RXRα represented the major RARE binding heterodimer, as indicated by supershift with anti-RARβ (Fig. 1, Lane 3) and anti-RXRα (Fig. 1, Lane 5) antibodies as well as by double shift using a mixture of antibodies against both RARβ and RXRα in the same reaction (Fig. 1, Lane 8). The antibodies against all RXRs produced a shifted complex that migrated differently from the complex shifted with RXRα antibody (compare Fig. 1, Lanes 5 and 6), suggesting that RXRβ and/or RXRγ may also be associated with RARβ on the RARE.

Complexes containing RARγ were also detected (Fig. 1, Lanes 4,9, and 12), but the low intensity of the RARγ band indicates that it is a minor component of the DNA-binding heterodimer. Although RXRα did not show abundant mRNA expression in Northern blotting, we detected high-level expression of RXRα protein as indicated by the intensity of the bands supershifted and double supershifted by the RXRα-specific antibody (Fig. 1, Lane 5). In contrast, antibodies specific for RARα failed to produce any supershifted band (Fig. 1, Lanes 2, 7, and 10), although we had previously found that the same antibodies could supershift RARα when incubated with nuclear extracts from head and neck squamous cell carcinoma cells and labeled RARE(37).

To determine whether exogenous retinoid receptors can restore retinoid responsiveness, we transfected H1792 cells transiently using coexpression vector pMARKCD7D5 (Ref. 36; Fig. 2,A), into which cDNAs of the different retinoid receptors were inserted. The cells were then grown without or with 1μ m ATRA and labeled with BrdUrd. After fixation and double immunostaining with anti-CD7 antibodies and anti-BrdUrd antibodies (detected with different secondary antibodies tagged with red and green fluorescent fluorophores, respectively), cells were observed under a fluorescence microscope (Fig. 2,B). The proportion of red cells with labeled nuclei within a population of about 500 cells exhibiting red fluorescence was considered to represent the mitotic index (Fig. 2,C). The mitotic indices in cell populations transfected with pMARKCD7D5 vector harboring no retinoid receptor was about 32% and was reduced only slightly by ATRA treatment. Transient expression of RARα or RXRα decreased DNA synthesis in the absence of ATRA to 24% (about 25% inhibition relative to pMARKCD7D5 vector only), whereas transfection of the other receptors did not decrease BrdUrd incorporation in the absence of ATRA. Overexpression of RARα and RXRα caused about 60% inhibition of DNA synthesis, and overexpression of RARγ caused 45% inhibition of DNA synthesis, whereas RARβ failed to enhance the response of the cells to inhibition of DNA synthesis by ATRA (Fig. 2 C).

Similar experiments with the RA-resistant NSCLC cell line Calu-1, which expresses no detectable RARβ but does express the other RARs and RXRs(24), have demonstrated that overexpression of RARα,RARβ, RARγ, and RXRα, followed by treatment with 1μ m ATRA, caused inhibition of DNA synthesis of 35.5%,40%, 10.1%, and 70%, respectively.

To determine whether the transfected receptors alter retinoid-regulated transcription in intact H1792 cells, the cells were cotransfected with pMARKCD7D5 vector bearing no receptor or individual retinoid receptors and a luciferase reporter construct containing DR5 RARE (Fig. 3,A). ATRA or 9-cis-RA treatment of cells transfected with pMARKCD7D5 vector alone activated luciferase transcription by 11- and 15-fold, respectively, via endogenous retinoid receptors (Fig. 3,A, top left panel). The RARα-selective retinoid AM80 and the RARγ-selective retinoid CD437 activated transcription by 4-fold, whereas the RARβ-selective retinoid CD2314 increased transcription by only 1.6-fold (Fig. 3,A, top left panel). In cells transfected with pMARKCD7D5 containing RARα(Fig. 3,A, top right panel), the activation of the DR5 reporter by ATRA, 9-cis-RA, and AM80 was 25-, 32-, and 13-fold, respectively; more than 2-fold higher than that in cells transfected with pMARKCD7D5 vector alone. There was no increase in the response of the cells to RARβ- and RARγ-selective retinoids. This indicated that the overexpressed RARα increased the response of the cells to retinoid signaling. Cells transfected with pMARKCD7D5 vector containing RARβ (Fig. 3,A, bottom left panel) showed a 35% increase in transcriptional activation of DR5 RARE by ATRA and 9-cis-RA compared with cells transfected with pMARKCD7D5 only but showed no increase in response to AM80, CD2314, or CD437. Treatment of these cells with several other RARβ-selective agonists such as TTNN and LG030369 also failed to show any stimulation of transcription (data not shown). These results indicate that exogenous RARβ contributes much less than RARα to the activation of transcription via RARE. Cells overexpressing RARγ (Fig. 3 A, bottom right panel) exhibited a higher transcriptional activation of DR5-driven reporter in the absence of retinoids than cells transfected with the other receptors. The activation of transcription by ATRA, 9-cis-RA, AM80, CD2314, and CD437 was 4.4-, 6.3-,1.5-, 1.0-, and 2.1-fold, respectively. Overexpression of RXRα in H1792 cells failed to increase activation of DR5 RARE by any of the RAR-selective retinoids (data not shown).

H1792 cells transfected with pMARKCD7D5 alone showed no increase in transcriptional activation of DR1 RXRE (38) on treatment with a variety of retinoids including some that bind to RXRs(e.g., 9-cis-RA, LG69, and SR11217; Fig. 3,B, top panel). This finding is surprising because the cells express high levels of RXR protein as detected by the electrophoretic mobility shift assay. Nonetheless, overexpression of exogenous RXRαvia the pMARKCD7D5 vector enhanced transcriptional activation of DR1 RXRE by ATRA, 9-cis-RA, LG69, and SR11217 by 5.0-, 6.6-,8.5-, and 5.6-fold, respectively. However, AM80, TTNN, and CD2325,which are selective for RARα, RARβ, and RARγ, respectively,showed no increase or up to a 1.6-fold increase in transcription (Fig. 3 B, bottom panel).

Nuclear retinoid receptors can antagonize AP-1 activity in a ligand-dependent fashion, and this antagonism often leads to growth inhibition (12, 14, 35, 43, 44, 45, 46, 47, 48). Therefore, we examined whether an association exists between the growth-inhibitory effects mediated by the receptors and their anti-AP-1 activity in the H1792 cells. Overexpression of each of the three RARs or RXRα had only minor effects on AP-1 activity in the absence of exogenous retinoids(Fig. 4; cells were treated with DMSO as a control). Treatment of cells transfected with pMARKCD7D5 alone using ATRA or 9-cis-RA not only failed to suppress AP-1 activity but rather increased it by 30–40% (Fig. 4, □), suggesting that the anti-AP-1 activity of retinoids via constitutively expressed retinoid receptors is aberrant in the H1792 cells. However, overexpression of RARα and RXRα and,to a lesser extent, RARγ resulted in marked suppression of AP-1 activity in H1792 cells after ATRA or 9-cis-RA treatment compared with AP-1 activity in cells transfected with pMARKCD7D5 alone(Fig. 4). 9-cis-RA was more effective than ATRA in suppressing AP-1 activity in cells overexpressing RXRα. Overexpression of RARβ resulted in a smaller decrease in AP-1 activity after retinoid treatment compared with the other transfected receptors. These results suggest a positive association between retinoid-induced growth-inhibitory effects mediated by the different retinoid receptors and their respective anti-AP-1 activity.

Because RXRα appeared to be the most potent of the four receptors in mediating both inhibition of DNA synthesis and antagonism of AP-1, we generated and used the deletion mutants shown in Fig. 5,A to determine which of the different domains of this receptor are required for the two activities. The construct RXRαDA,which was predicted to express a mutant protein with a deletion of the 71 NH₂-terminal amino acids of wt RXRα, was as effective as intact RXRα in mediating both inhibition of DNA synthesis (Fig. 5,B) and antagonism of AP-1 (Fig. 5,C). In contrast, mutant RXRαDD, which was predicted to express a protein with a deletion between amino acids 29 and 197, and mutant RXRαDF, which was predicted to express a truncated protein lacking the 61 COOH-terminal amino acids of RXRα, completely lost the growth-inhibitory effect (Fig. 5,B) and anti-AP-1 (Fig. 5 C) effect of RXRα. These findings indicated that both the DNA-binding domain and the ligand-binding/heterodimerization domain are required for the two RXRα activities.

To assess additional aspects of the expression and function of these RXRα mutants in H1792 cells, their ability to activate the transcription of DR1- and DR5-containing luciferase reporter constructs was examined. RXRα- and RXRαDA-transfected cells exhibited similar potency for activation of DR5 RARE (Fig. 6, compare B and C). However, compared with wt RXRα (Fig. 6,B), RXRαDA was able to partially (50–60%of the effect of wt RXRαt) activate transcription via DR1 after treatment of the cells with ATRA, 9-cis-RA, and the RXR-selective retinoids LG69 and SR11217 (Fig. 6 C). Mutants RXRαDD and RXRαDF showed no ability to activate DR1 (data not shown).

Although they were unable to activate DR1, RXRαDD and RXRαDF interfered with the ligand-induced transcriptional activation of DR1 by wt RXRα in cotransfection experiments. RXRαDD (Fig. 7,A) was more effective than RXRαDF (Fig. 7 B) in antagonizing RXRα. It is possible that these mutants formed abortive heterodimers with wt RXRα or competed for the same coactivators.

Several RXRα point mutations in the F domain were found to interfere with RXR-RXR and RXR-RAR dimerization (39). One such mutant, RXRαL430F, had lost the ability to act as homodimer,whereas a mutation in the adjacent codon (RXRαK431Q) retained this ability. We cloned these two point-mutated RXRαs into pMARKCD7D5 vector and analyzed their activities after transient transfection into H1792 cells and treatment with 9-cis-RA. Fig. 8 shows that the point mutation L430F, which abrogated DR1 activation as expected but retained some ability to activate DR5, resulted in a loss of ability to suppress DNA synthesis and to antagonize AP-1. In contrast, the point mutation RXRαK431Q, which retained the ability to activate DR5 as well as DR1, like wt RXRα, also exhibited the ability to both suppress DNA synthesis and antagonize AP-1 in the presence of 9-cis-RA (Fig. 8).

Mutation F318A in the ligand-binding pocket of mouse RXRα was found to cause the receptor to exhibit constitutive activation of DR1 (as a homodimer) and, to a lesser extent, a DR5 reporter (as a heterodimer with RARs) in the absence of a RXR ligand (49); namely, to act like a ligand-bound receptor. We generated a homologous mutant in human RXRα (RXRαF313A) and cloned it into pMARKCD7D5 vector. After transient transfection into H1792 cells, we found that this mutant has a constitutive transcriptional activity on DR1 but not on DR5 in the absence of 9-cis-RA (Fig. 8, top two panels,respectively). Furthermore, this mutant exhibited ligand-independent antagonism of AP-1 and inhibition of DNA synthesis (Fig. 8, bottom two panels, respectively). These results indicate that DR1 activation, antagonism of AP-1, and inhibition of DNA synthesis are strongly associated and may be causally related.

Most NSCLC cell lines that we (24) and others(22, 22) have examined previously were found to be resistant to the growth-inhibitory effects of ATRA and other retinoids. Many of these cell lines failed to express RARβ(18, 19, 20, 21, 22, 23, 24), and this aberration has been proposed to be the reason for their retinoid resistance (22). In another lung carcinoma cell line, transfection of RARβ restored sensitivity to growth inhibition by ATRA when the cells were grown in low serum concentration (50). Furthermore, knockout by homologous recombination of the RARβ gene in F9 embryonal carcinoma cells resulted in loss of growth inhibition by ATRA (51). Despite this evidence that RARβ can mediate growth inhibition, the H1792 cells were resistant to ATRA (24), although they express high levels of RARβ mRNA as confirmed and extended in the present study. The H1792 cells were found to express mRNAs for all of the RARs and for RXRα (24). Furthermore, RARβ and, to a much lesser degree, RARγ proteins were detected in complexes with DR5 RARE, presumably as heterodimers with RXRα (Fig. 1). These endogenous receptors were able to mediate transactivation of DR5 RARE(Fig. 3,A). However, there was no tight association between the ability to transactivate DR5 RARE and the ability to inhibit DNA synthesis, as seen in cells transfected with RARβ (Fig. 2,Cand Fig. 3,A, bottom left panel). The endogenous RXRs failed to mediate activation of DR1 RXRE (Fig. 3,B) in H1792 cells exposed to ATRA or 9-cis-RA,and endogenous RARs and RXRs failed to mediate AP-1 antagonistic activity in H1792 cells exposed to ATRA or 9-cis-RA (Fig. 4,□). Based on these findings, we suggest that the resistance of H1792 cells to the growth-inhibitory effects of retinoids is due to the inability of the constitutively expressed receptors to mediate DR1 activation and AP-1 antagonism.

Several previous studies have demonstrated that stable transfection of different retinoid receptors can restore response to the effects of retinoids on cell growth, differentiation, or apoptosis. RARs were able to restore responsiveness in leukemia cells (52); RARαand RARβ were able to restore responsiveness breast carcinoma cells(17); RARβ was able to restore responsiveness in breast(33), lung (50), and head and neck(37) carcinoma cells; RARγ was able to restore responsiveness in head and neck squamous carcinoma (53)and teratocarcinoma cells (54); and RARs plus RXRα were able to restore responsiveness in ovarian carcinoma cells(35).

In the present study, we have chosen the H1792 cells to determine whether transient transfection of different receptors can overcome the apparent block in response to growth inhibition by retinoids. We found that exogenous retinoid receptors can mediate growth inhibition by ATRA in the following rank order: RARα = RXRα > RARγ (Fig. 2 C). RARβ overexpression was without an effect. A similar ranking of activity of the exogenous receptors was noted when their effects on AP-1 antagonism were determined. Combined,these results suggest that the growth inhibition mediated by the transfected receptors may be associated with antagonism of AP-1.

In another RA-resistant NSCLC cell line, Calu-1, which does not express RARβ, transfection of pMARKCD7D5 vectors containing RARα, RARβ,or RXRα but not RARγ restored growth inhibition by ATRA. These results show that some receptors (e.g., RARα and RXRα)can restore ATRA responsiveness in more than one NSCLC cell line,whereas other receptors (e.g., RARβ and RARγ) may do so in some cells, but not in others. Thus, cell context may determine the ability of a transfected receptor to mediate growth inhibition.

Another interesting result was that although H1792 cells express endogenous RXRα, this receptor does not activate DR1 RXRE after the cells are treated with 9-cis-RA and other RXR-selective retinoids (Fig. 3,B, top panel). However, after transfection of exogenous RXRα, the same retinoids activated DR1 RXRE (Fig. 3 B, bottom panel). Thus, the putative increase in the level of RXRα in the transfected cells causes a qualitative change in ability to activate RXRE. The lack of RXRE activation may be due to low endogenous RXRα-RXRα levels and transrepression by more abundant RXRα-RARβ heterodimers that can bind to the DR1 with a higher affinity than RXRα-RXRα homodimers but fail to activate this response element (55, 56). One explanation for the effect of overexpression of RXRα is that the increased RXRα level leads to the formation of RXR homotetramers, which, having a higher affinity for DR1 than RXR-RAR, may displace RXR-RAR heterodimers from the DR1 RXRE(57). In the presence of RXR ligands, the DNA-bound tetramers dissociate to dimers (57) that activate DR1-mediated transcription of the reporter.

Because of the interesting effects of RXRα, we wished to gain some additional understanding of its mechanism of action. Therefore, we generated several deletion and point mutants of this receptor and compared their effects with that of wt RXRα. Studies with deletion mutants indicated that the DNA-binding and ligand-binding domains were important for growth inhibition (Fig. 5,B), DR1 RXRE activation, and AP-1 antagonism (Fig. 5,C). Some deletion mutants of RARs have been shown to possess dominant negative activity(58, 59). Our results indicate that mutants with deletions in the DNA-binding domain (RXRαDD) or ligand-binding domain(RXRαDF) can exert a dominant negative effect on RXRα, as indicated by the partial suppression of the effect of wt RXRα on the activation of DR1 RXRE (Fig. 7). The RXRα mutant with a deletion in the NH₂-terminal domain was almost as active as wt RXRα when examined for all of the activities, suggesting that the ligand-independent transactivation function residing in this domain is dispensable for all activities. The finding that point mutation L430F,which inhibits the ability of RXRα to transactivate transcription as a homodimer, but not mutation K431Q, which has intact ability to act as a homodimer (39), has lost the ability to transactivate DR1, suppress DNA synthesis, and transrepress AP-1 (Fig. 8) provides further support for the conclusion that the mechanisms underlying the growth-inhibitory effects mediated by RXRα in H1792 cells involve homodimerization, transactivation of RXRE (DR1), and transrepression of AP-1. The data point to the possibility that RXRα homodimers can mediate AP-1 transrepression. Interestingly, this conclusion is also supported by the ability of the point mutant RXRαF313A to activate DR1, inhibit DNA synthesis, and antagonize AP-1 in the absence of 9-cis-RA (Fig. 8). The effects of F313A are important because they exclude pleiotropic effects resulting from different signaling pathways that can be provoked by 9-cis-RA and point to the central role of the RXR receptor in the above-mentioned effects.

AP-1 is a complex comprised of dimers of members of the Jun and Fos family of DNA-binding proto-oncogenes that mediate mitogenic signals from a variety of growth factors and tumor promoters (14). AP-1 activity is particularly important for the progression of lung cancer cells because altered expression of members of AP-1 family was found to be an early event in human lung carcinogenesis(60). The transrepression of AP-1 by liganded retinoid receptors has been demonstrated previously both in vitro(11, 12, 13, 35, 43, 44, 45, 46, 47, 48) and in vivo(61). The retinoids that induced the repression of AP-1 activity were also found to be able to inhibit the growth of many types of cancer cells (45, 46, 47, 48). Normal HBE cells are sensitive to both the anti-AP-1 and growth-inhibitory effects of RA. However,tumorigenic HBE cells and many lung cancer cells are resistant to both effects of RA (48). Our study has extended these reports by demonstrating that several overexpressed retinoid receptors can mediate anti-AP-1 effects of retinoids. The mechanism of AP-1 transrepression is not entirely clear. No binding of liganded retinoid receptors to the AP-1 consensus DNA sequence was observed, thus excluding competition for DNA binding as a mechanism of antagonism. Several mechanisms have been suggested including binding of liganded retinoid receptors to c-Jun or c-Fos, which would thus interfere with c-Jun/c-Jun homodimerization and c-Jun/c-Fos heterodimerization and prevent the formation of AP-1 complexes capable of DNA binding(12, 13). A variation on this mechanism was suggested by the demonstration that RARs, RXRs, and c-Jun form a complex at the AP-1 site of the collagenase promoter in which c-Jun binds directly to the DNA and apparently links the retinoid receptors to the complex(62). Competition for the limited amount of the coactivator CBP/p300, which is required for the transcriptional activity of both nuclear retinoid receptors and AP-1, has also been proposed as a mechanism of mutual retinoid and AP-1 antagonism(13). Interference with the Jun NH₂-terminal kinase signaling pathway represents another mechanism by which nuclear hormone receptors can antagonize AP-1. This mechanism is based on the blockade by hormone-activated nuclear receptors of c-Jun phosphorylation on Ser⁶³/Ser⁷³, which is required to recruit the transcriptional coactivator CBP(63). A similar mechanism was observed in HBE cells, in which ATRA decreased the amount and activation of AP-1 components. ATRA inhibited Jun NH₂-terminal kinase and, to a lesser extent, extracellular signal-regulated kinase activity and also reduced c-fos mRNA (48). Recently, pretreatment of human skin with ATRA was found to inhibit UV induction of c-Jun protein and, consequently, AP-1 via a posttranscriptional mechanism because ATRA did not inhibit UV induction of c-Jun mRNA(61).

Taken together, our data suggest that the sensitivity to the growth-inhibitory effect of RA can be restored by overexpression of several exogenous nuclear retinoid receptors in lung cancer cells. One potential clinical implication is that one could combine receptor gene transfer and retinoid treatment as a strategy for therapy or prevention of lung cancer. With respect to the mechanism of growth inhibition, the data indicate that antagonism of AP-1 by the transfected receptors may be important for inhibition of DNA synthesis.