Ectopic Expression of Nonliganded Retinoic Acid Receptor (cid:1) Abrogates AP-1 Activity by Selective Degradation of c-Jun in Cervical Carcinoma Cells* □ S

Expression of the nuclear retinoic acid receptor (cid:1) 2 (RAR (cid:1) 2) gene is often disturbed in cervical carcinoma cells. One important mechanism by which RAR (cid:1) 2 can exert growth inhibitory function is based on its ability to repress the AP-1 transcription factor in a ligand-de-pendent manner. Because less is known about the biological effects of RAR (cid:1) in the absence of ligand, the corresponding cDNA was stably introduced into HPV18-positive HeLa cervical carcinoma cells. In the present study we describe a novel mechanism by which AP-1 becomes inactivated. Constitutive expression of nonliganded RAR (cid:1) abrogated both AP-1 binding affinity and activity by a selective degradation of the c-Jun protein as major dimerization partner, without substitution by other members of the Jun family. Blockage of the proteasomal pathway completely rescued c-Jun and reconstituted the AP-1 function. Moreover, HeLa RAR (cid:1) 2 clones treated either with tumor necrosis factor- (cid:2) or transfected with a constitutive active upstream mito-gen-activated protein kinase (MEKK1 (cid:3) ) also resulted in c-Jun phosphorylation and restoration of AP-1 affinity and functionality similar to that found in nontransfected parental HeLa cells. These data revealed an important cross-talk between trans -repression of AP-1 and nonliganded RAR (cid:1) in human papillomavirus-positive cells. Because AP-1 activity was not irreversibly disturbed, but could be switched on through activation of the Jun N-terminal kinase pathway, a model for the transient activation of AP-1 even in the presence of RAR (cid:1) as repressor is suggested.

Retinoids are regulators of multiple physiological processes ranging from embryogenesis to metabolism and are effective in the medication of several disorders including cancer (1). The clinical efficacy of retinoids mainly depends on their ability to modulate cellular growth, differentiation, and apoptosis in premalignant and malignant cells. Retinoids are currently used as preventive and therapeutic drugs against a whole variety of human malignancies such as breast, lung, ovary, liver, leukemia, prostate, and cancer of the cervix uteri (2). The biological effects of retinoids are mediated by nuclear retinoid receptors that belong to two subfamilies: retinoid acid receptors (RARs) 1 and retinoid X receptors (RXRs). Both of them include three different isotypes (␣, ␤, and ␥) that are encoded by distinct genes (1,2). The RARs bind both all-trans-retinoic acid (atRA) and 9-cisretinoic acid, whereas RXRs only bind the latter. In the presence of the ligand, RXRs can either form homodimers or heterodimers with RARs, which in turn interact with retinoic acid-response elements (RAREs) or RXREs of atRA or 9-cisretinoic acid-sensitive genes (3,4). The existence of distinct RAREs and RXREs within promoter regions indicates that RAR homodimers and RAR-RXR heterodimers traverse different pathways upon retinoid acid addition (5). Positive regulation of gene expression by retinoid receptors requires their interaction with adaptor proteins with intrinsic histone acetyltransferase activity (p300 or CREB-binding protein), providing a direct functional link with the core transcriptional machinery and the modulation of the nucleosomal structure (6).
In addition to their positive regulatory function, retinoid receptors also negatively affect gene expression. In the absence of ligands, RAR/RXRs recruit histone deacetylases, resulting in gene silencing via chromatin condensation (7,8). On the other hand, RAR-RXR heterodimers can also repress genes in a liganddependent manner (9). This important trans-repressive function is mediated by inhibition of the AP-1 (activator protein 1) (10). The transcription factor AP-1, which consists of a complex of homo-or heterodimers of the Jun/Fos family members, is a central regulatory key element, playing not only a fundamental role in transcriptional regulation of human papillomaviruses (11,12) but also exhibiting enhanced activity during cell proliferation and tumor progression (13,14). AP-1 activity is mainly determined by its composition as a dimer, by its affinity to a responsive element within a particular promoter, and by post-translational modification through mitogen-activated protein (MAP) kinases (15). In this way, AP-1 is controlled by a * 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. network of superimposed protein kinases, modulating its activity through extracellular stimuli such as growth factors, cytokines, and tumor promoters (15). Depending on the cell system, different negative regulatory mechanisms for RAR/RXR on AP-1 have been proposed: (a) direct interaction with Jun/Fos family members (16), (b) disruption of Jun-Fos dimerization (17), (c) competition with AP-1 to recruit transcriptional coactivators such as p300 or CPB (18), or (d) inhibition of the c-JunN-terminalkinase(JNK),whichpreventsphosphorylationdependent activation of c-Jun (19).
Many reports concluded that in particular RAR␤ is the most potent RAR involved in suppression of tumor-related phenotypes (20). RAR␤ is not expressed in many malignant cells either because of epigenetic modifications such as promoter hypermethylation and chromatin condensation or by loss and structural rearrangements of the RAR␤ locus on chromosome 3 (21)(22)(23). Nonetheless, in some other cell systems, RAR␤ can still be induced by pharmacological doses of retinoids (24). Houle et al. (25) first demonstrated that transfection of RAR␤2 into tumor-derived cells from epidermis was sufficient to diminish their in vitro growth capacities and in vivo tumorigenicity. Expression of RAR␤2 in other systems reduces anchorage independence (24,26) and promotes antiproliferative constraints upon retinoic acid addition (24,26,27). Notably, some of the tumor-suppressive functions exhibited by RAR␤ have been explained by the ability to trans-repress the AP-1 activity (20). Hence, to oppose unregulated cellular growth, it is of considerable therapeutic interest to unravel the mechanism by which AP-1 activity is down-regulated.
Retinoids are also therapeutically used in the treatment of human papillomavirus (HPV)-induced diseases such as cervical cancer (2). In this context, it has been demonstrated that AP-1 and its composition play a fundamental role in determining the tumorigenic phenotype of HPV16-and HPV18-positive cells (28,29). This is intriguing with respect to previous experiments showing that, in contrast to primary human keratinocytes (30), cervical carcinoma cells lack constitutive RAR␤ expression (22, 30 -32). In other words, the absence of RAR␤ obviously provides a selective advantage during multistep progression to cervical cancer, because the antagonism between the nuclear receptor and AP-1 activity is relieved. To study the effect of ligand-independent RAR␤ expression on the transcription factor AP-1 in molecular terms, HPV18-positive HeLa cervical carcinoma cells were stably transfected with a plasmid constitutively expressing RAR␤ under the control of the ␤-actin promoter (31). In the present study we demonstrate that AP-1 binding affinity and activity is abrogated by a post-transcriptional mechanism where the c-Jun protein as the major dimerization partner is selectively degraded and is not substituted in the AP-1 complex by other Jun-binding partners. These data uncover a cross-talk between AP-1 and the nonliganded RAR␤ in cervical cancer cells and describe a mechanism by which AP-1 becomes trans-repressed.

EXPERIMENTAL PROCEDURES
Cell Lines and Treatment-The parental cervical carcinoma cell line HeLa and the stably transfected HeLa RAR␤ clones (31) were maintained in Dulbecco's modified Eagle's medium (Sigma) supplemented with 10% fetal calf serum (Invitrogen), 1% penicillin/streptomycin (Sigma). To keep a constant selection pressure, the clones were grown in a media containing 750 g/ml G-418 (Invitrogen). Clones ␤1 and ␤2 correspond to clones 8 and 59/6 as described recently (31). For retinoic treatment, cells were incubated with 10 M all-trans-retinoic acid (atRA, Sigma), dissolved in dimethyl sulfoxide (Me 2 SO) as a 10 Ϫ2 M stock solution, and kept at Ϫ20°C. Four hours before atRA treatment, normal culture media were replaced with media containing 10% charcoal-stripped serum in order to reduce the retinoid levels. Control cultures received media containing 0.1% (v/v) Me 2 SO alone. The cells were exposed to atRA or Me 2 SO for 72 h without change of culture media. MG132 (Calbiochem) was dissolved in Me 2 SO at a concentration of 20 mM. Semi-confluent cells were treated with 20 M of proteasome inhibitor MG132 for 8 h. For cytokine treatment, the cells were incubated with 500 units/ml TNF-␣ (Strathmann Biotech GmbH, Hannover, Germany) as described in the figure legends.
Transient Transfections-To monitor AP-1 activity, 2.5 ϫ 10 6 cells were plated on 60-cm 2 dishes and co-transfected with 1.5 g of a luciferase reporter driven by a TPA-responsive element (TRE-Luciferase, Promega). atRA responsiveness was measured with a luciferase construct carrying a retinoic acid-response element (RARE-Luciferase) (32). Transfection was performed using an "Effectene Transfection Reagent" (Qiagen, Hilden, Germany) according to the manufacturer's instructions. Transfection efficiency was adjusted by co-transfecting 0.5 g of an SV40-controlled ␤-galactosidase gene. One day after transfection, cells were treated with MG132 or with atRA using Me 2 SO as control as indicated in the figure legends. Cells were harvested after 8 h. Luciferase activity was measured in a luminometer (Berthold, Germany), and the data were normalized in relation to the co-expressed ␤-galactosidase activity. MEKK1⌬ is a constitutive active truncated form of the MEK kinase 1 (mitogen-activated/extracellular signal-regulated kinase kinase 1, MEKK1) which lacks parts of the N-terminal regulatory domain (33,34). pcDNA3-HA.POH1 is a hemagglutinintagged cDNA encoding the human Pad1 homologue (POH1) (35). Transfections were carried out with 2 g of an expression vector, containing the cDNAs under the control of the cytomegalovirus promoter/enhancer. Cells were harvested after 48 h.
Electrophoretic Mobility Shift Assay-Oligonucleotides were generated in an Applied Biosystems synthesizer (Foster City, CA) by using phosphoramitide chemistry and further purified by high pressure liquid chromatography. For electrophoretic mobility shift assays (EMSA), oligonucleotides for AP-1 consensus 5Ј-CGCTTGATGACTCAGCCG-GAA-3Ј derived from the human collagenase promoter (36) and Oct-1 consensus sequences 5Ј-TGTCGAATGCAAATCACTAGAA-3Ј derived from the immunoglobulin light chain enhancer (37) were used. For EMSA, the annealed oligonucleotides were labeled with [␥-32 P]ATP (Amersham Biosciences, 3000 Ci/mmol) with T4 polynucleotide kinase and gel-purified from a 15% polyacrylamide gel. Nuclear extracts were prepared using the method of Schreiber et al. (38) with the only modification that N,N-(L-3-trans-carboxyoxirane-2-carbonyl)-L-leucyl-agmatine (E64) and 4-(2-aminoethyl)-benzolsulfonyl fluoride ("Pefabloc SC"), 1 mM NaF, and 0.2 mM Na 3 VO 4 were included as protease and phosphatase inhibitors in concentrations suggested by the manufacturer (Roche Applied Science). Protein concentration was determined by the Bradford method (Bio-Rad) by using defined amounts of bovine serum albumin as standard. The binding was performed in a 20-l reaction volume containing 10% glycerol, 12 mM HEPES, pH 7.9, 4 mM Tris-HCl, pH 7.9, 60 mM KCl, 1 mM EDTA, 1 mM dithiothreitol, 0.6 mg/ml bovine serum albumin, 2.0 g of poly(dI-dC), and 2 g of nuclear extract. After 5 min, 10,000 cpm of the [␥-32 P]ATP 5Ј-end-labeled double-stranded oligonucleotide probe was added, and the incubation was continued for an additional 30 min at room temperature as described previously (28,29). The sequence specificity of the binding was routinely controlled in competition experiments by the addition of a 100-fold molar excess of either unlabeled homologous or heterologous oligonucleotides. For monitoring AP-1 composition in supershift assays, 2 g of a monoclonal antibody directed against the c-Jun ("TransCruz TM " supershift reagent), recognizing the phosphorylated form of c-Jun (epitope corresponding to amino acids 56 -69 within the N-terminal domain), was added, and the reaction was further incubated for 1 h at 4°C. DNAprotein complexes were resolved on 5.5% nondenaturing polyacrylamide gels (29:1 cross-linking ratio), dried, and exposed overnight to Fuji medical x-ray films. Retarded bands were quantified with an Amersham Biosciences PhosphorImager using the "ImageQuant" program as software.
Assay for JNK Activity-JNK activity was assayed using the "stressactivated protein kinase/JNK assay kit" (Cell Signaling Technology, Frankfurt/Main, Germany) following the manufacturer's instructions. Cells treated with TNF-␣ were harvested, and JNK was pulled down from the extracts (250 g of total protein) by using 2 g of GST-c-Jun fusion protein beads. The pellets were incubated at 30°C for 30 min with 100 M ATP. Phosphorylation of c-Jun was visualized after SDS-PAGE and immunoblotting by using a phosphorylation-specific antibody. The controls represent nontreated cells.

RESULTS
Ectopic RAR␤ Expression in HeLa Cells-Primary human cervical epithelial cells regularly express high basal levels of the RAR␤ gene, whereas RAR␤ expression is either absent or strongly diminished in cervical cancer (30). HPV18-positive HeLa cells, stably transfected with the human RAR␤2 cDNA under the control of the actin promoter (31), were analyzed in order to investigate the biological consequences of ligand-independent RAR␤2 receptor expression. After Southern blot analysis, two independent clones (referred as "clones ␤1 and ␤2") were selected, each carrying a single copy of the transfected cDNA, integrated at different genomic loci (data not shown). To verify ectopic RAR␤2 expression in these cells, semiquantitative RT-PCR was performed. GAPDH was used as internal control to confirm that equal amounts of cDNA template were supplied for the PCR. Fig. 1A shows that HeLa cells, which were cultured in charcoal-stripped serum, completely lacked endogenous RAR␤ expression. However, when the cells were incubated with pharmacological doses of atRA for 3 days, the RAR␤ gene could be induced. Similar expression was reached in the HeLa RAR␤2 clones without atRA treatment, both on RNA and protein levels. Consistent with previous results (20), ectopic RAR␤2 did not change other RARs, as confirmed when RAR␣ expression was monitored. Functionality of the RAR␤ receptor was assessed by transient transfection assays by using a RARE-Luciferase reporter construct. Both parental HeLa cells and the RAR␤ clones revealed significant induction of luciferase activity after atRA addition, clearly demonstrating that ectopically expressed RAR␤ receptor can be further activated in a ligand-dependent manner (Fig. 1B).
RAR␤ Expression Reduces AP-1 by Selective Down-regulation of c-Jun at Post-transcriptional Levels-Although it is well documented that liganded RAR␤ exerts a negative regulatory effect on AP-1 (9,43,44), little is known about the transrepressing function of RARs in the absence of atRA. This is of particular interest in the context of HPV-induced cancer, where AP-1 plays a major role both in viral transcriptional regulation and in proliferation of cervical cancer cells (11,12,28,29). Considering that both RAR␤ clones showed reduced steadystate levels of endogenous HPV18 E6/E7 expression that corresponded to slower growth rates when compared with parental HeLa cells (data not shown), we reasoned that constitutive expression of nonliganded RAR␤ might act via this pathway, namely by negatively affecting AP-1 activity. To test this notion, we first examined the AP-1 content by EMSAs. Incubating nuclear extracts obtained from RAR␤ clones with 32 P-labeled AP-1 oligonucleotides, a 60 -80% reduction of AP-1 binding could be discerned. To demonstrate that suppression of AP-1 binding was a selective process, EMSAs with 32 P-labeled Oct-1-specific oligonucleotides were carried out. In fact, under conditions where AP-1 affinity was reduced, Oct-1 binding was not affected (Fig. 2A). These data clearly demonstrate that the ligand-independent decrease of AP-1 was a selective process, which did not impair the binding of transcription factors in general. Moreover, to exclude a nonphysiological squelching effect, we have also analyzed a clone, where RAR␤ was highly expressed due to multiple insertions of the transfected DNA. Even under these conditions, identical results were obtained (see Supplemental Material). This argues against the notion that the mere overexpression acts in a nonspecific, dominantnegative manner.
When parental HeLa cells were treated for 3 days with pharmacological doses of atRA, ligand-dependent trans-repression led to a reduction in their AP-1 content (about 50%). Addition of atRA to the RAR␤ clones further diminished AP-1 binding (Fig. 2B), clearly showing a two-step mechanism of AP-1 suppression (see below).
Because AP-1 is a dimeric protein, we monitored the expression of different subfamily members of AP-1. If ligandindependent RAR␤-mediated AP-1 reduction was the result of a mechanism regulated at the transcriptional level, quantitative changes of the mRNAs encoding different AP-1 subfamily members should be expected. To test this prediction, semiquantitative RT-PCRs were performed. Whereas fra-1, junB, and junD were not significantly altered in their transcription rate, the steady-state level of c-fos and c-jun RNA was even increased (Fig. 2C, left panel). Conversely, by examining corresponding nuclear extracts by Western blot analyses, c-Jun was almost completely absent, whereas the amounts of other AP-1 family members such as JunB, JunD, or ATF-2 were not quantitatively affected (Fig. 2C, right  panel). Only c-Fos, whose intracellular half-life is controlled by c-Jun (45), was reduced to levels comparable with atRAtreated parental HeLa cells. These data demonstrate that decreased AP-1 binding in EMSAs within HeLa RAR␤ clones ( Fig. 2A) was a post-translational process, which was mainly mediated by a selective degradation of c-Jun without substitution of JunB and JunD as DNA-binding partners. Although AP-1 binding was also reduced upon atRA addition in parental HeLa cells, c-Jun was not diminished, which clearly shows that ligand-dependent and -independent AP-1 trans-repression act through different pathways.
JNK Activation and AP-1 Reconstitution in RAR␤ Clones-It has been suggested that nuclear receptors can trans-repress AP-1 by blocking JNK phosphorylation (19). Another possibility is that JNK becomes activated, but constitutive RAR␤ expression leads to an increase of the prototypic MAP kinase phosphatase MKP-1, which in turn de-phosphorylates activated JNK (47). This targets c-Jun for degradation due to its physical association with inactivated JNK (48). To discriminate between these two options, we first determined whether TNF-␣, known to activate JNK (15,49), can induce c-Jun phosphorylation when supplemented exogenously as a substrate. As shown in functional pull-down assays with extracts from parental and RAR␤-expressing HeLa cells, added GST-c-Jun fusion protein became equally phosphorylated in the presence of ATP when cells were treated with TNF-␣. Re-incubation of the same filter with a JNK antibody confirmed that identical amounts of JNK were co-immunoprecipitated (Fig.  3A). Inspecting the expression levels of the dual specificity phosphatase under same experimental conditions, the amount of MKP-1 was not quantitatively changed (Fig. 3B). These data provide evidence that at least in response to pro-inflammatory cytokine stimuli (TNF-␣), irreversible inactivation of JNK did not account for the absence of c-Jun in RAR␤-expressing HeLa cells.
Moreover, to demonstrate that JNK activation was not exceptional for TNF-␣, we transiently transfected a constitutive

AP-1 Suppression by Ligand-independent RAR␤ Expression
active mutant of the mitogen-activated protein kinase kinase 1 (MEKK1⌬) (34). MEKK1⌬ lacks the N-terminal regulatory domain and is described to augment the transcriptional potential of c-Jun by activation of JNK (33). To analyze the effect of MEKK1⌬ overexpression on AP-1 reconstitution, EMSAs were performed. By utilizing Oct-1 as internal reference, Fig. 4A demonstrates that transfection of MEKK1⌬ in RAR␤-expressing clones reconstituted AP-1 binding to levels comparable with nontransfected parental HeLa cells. Addition of c-Jun-specific antibodies resulted in a slower mobility band confirming that AP-1 re-binding could be attributed to enhanced c-Jun incorporation (Fig. 4B). As revealed by supershift EMSAs, constitutive MEKK1⌬ expression also had a stimulatory effect on AP-1 in parental HeLa cells (Fig. 4B). Enhanced c-jun mRNA formation (Fig. 4C) can be explained by the assembly of accumulated c-Jun with pre-existing ATF-2 proteins (Fig. 2B), which in turn positively regulate c-jun transcription through binding within its promoter (50). Transfection of MEKK1⌬ was not affecting the steady-state level of receptor expression excluding the possibility that AP-1 reconstitution was because of a transient decrease of RAR␤ (Fig. 4D).
Treatment with Proteasome Inhibitor MG132 Reconstitutes AP-1 Binding through c-Jun Accumulation and de Novo Synthesis-Because the half-life of many cell cycle regulatory proteins and transcription factors is controlled by the 26 S proteasome (51), the fate of c-Jun was monitored in the presence of MG132, an inhibitor of the ubiquitin-dependent proteasome pathway. For this purpose, both parental HeLa cells and the RAR␤ cells were treated for 8 h in the presence of 20 M MG132, and the corresponding extracts were examined by Western blot analyses (Fig. 5A). Although there was no significant change in the steady-state level of ATF-2, known to be regulated by proteasomal degradation (51), c-Jun became strongly up-regulated in all cell lines. Because MG132 was also inducing activation of the JNK (52-54), we studied the effect of MG132 on the phosphorylation of c-Jun and ATF-2, two known substrates for JNK (55). In fact, accumulation of c-Jun was accompanied by its phosphorylation, as detected when a phosphorylation-specific antibody was applied (Fig. 5A). This was consistent with the notion that c-Jun is no longer a target for ubiquitination and subsequent degradation when phosphorylated (48,56). Moreover, although only described for RAR␣, RAR␥, and RXR␣, RAR␤ half-life might be also regulated by the ubiquitin/proteasome pathway (57,58). However, as far as our experimental conditions are concerned, the amount of ectopically expressed RAR␤ was not quantitatively altered. Whether the appearance of the slower migrating band after MG132 application represents an accumulation of phosphorylated RAR␤ as described for RAR␥ and RXR␣ (59) remains to be elucidated. Similar to the situation already shown for MEKK1⌬ transfection (Fig. 4C), elevated c-jun transcription can be attributed to the positive autoregulatory mechanism at the c-jun promoter, mediated by binding of phosphorylated c-jun/ATF-2 heterodimers (60). Taken together, these data provide direct evidence that the absence of c-Jun occurred through a transcription-independent mechanism, where the protein can be rescued by blocking the ubiquitin-proteasomal pathway.
The effect of c-Jun accumulation on AP-1 binding could be followed by EMSA using 32 P-labeled AP-1 oligonucleotides (Fig. 6A). AP-1 binding increased both in HeLa and HeLa RAR␤ clones between 70 and 150%. The ubiquitous transcription factor Oct-1, whose binding was not affected after MG132 treatment, again served as internal control for selectivity. Moreover, as outlined above, JNK can phosphorylate c-Jun at residues Ser-63 and Ser-73 (15), which in turn leads to protein stabilization by reducing ubiquitination and subsequent proteasomal degradation (56). Addition of Ser-63/73 phospho-specific antibodies confirmed that the phosphorylated form of c-Jun was in fact assembled into the AP-1 complex, because more than 50% of the signal was retarded as the slower mobility band (Fig. 6B). Finally, in order to corroborate that accumulated c-Jun protein resulted in a reconstitution of a functional AP-1 complex, cells were transfected with a luciferase reporter gene driven by a TPA-responsive element (TRE-Luc) and treated for 8 h in the presence of 20 M MG132. Fig. 6C revealed that there was a 5-7-fold induction in comparison with nontreated controls, clearly demonstrating that AP-1 reconstitution monitored by EMSA also resulted in a functionally active transcription complex.
Transient POH1 Overexpression Leads to c-Jun Stabilization in HeLa RAR-␤ Clones-To substantiate further the role of the proteasome on the half-life control of c-Jun in RAR-␤-expressing cells, we transiently expressed a cDNA encoding the human homologue of the yeast pad1 ϩ gene, conferring pleiotropic drug resistance through AP-1 (35). POH1 is part of the lid region of the proteasome and is capable of increasing the amount of c-Jun in COS7 cells (61). To test whether this protein provides a functional link between RAR-␤ expression and c-Jun stability in our model system, transient transfections assays were performed. Most intriguingly, as shown by "supershift" EMSAs in Fig. 7A, ectopic POH1 expression significantly enhanced c-Jun incorporation within the AP-1 complex, both in parental HeLa cells and in the RAR-␤ clones. Oct-1 binding (Fig. 7B) again A, EMSAs with 32 P-labeled oligonucleotides specific for AP-1 or Oct-1. B, supershift analysis for phosphorylated c-Jun. Ϫ, nontreated cells; ϩ, MG132 treatment for 8 h. C, AP-1 activity in RAR␤ HeLa cells after MG132 treatment. Cells were co-transfected with TRE-Luc and SV40 ␤-galactosidase reporter constructs. One day after transfection, cells were split in two identical plates. 38 h post-transfection, one plate was treated with 20 M MG132 for 8 h. Cell extracts were assayed for luciferase activity, whereby each value was normalized against ␤-galactosidase activity and compared with the untreated control to define the fold induction. Nontreated controls for each clone were arbitrarily set as 1. Data represent the results from three independent experiments.
showed that increased c-Jun association with its cognate oligonucleotide was selective. One can therefore conclude that constitutive expression of RAR␤ trans-represses AP-1 in a ligandindependent manner and triggers a selective degradation of c-Jun via the proteosome. DISCUSSION In order to complement our understanding by which mechanisms retinoic receptors can trans-repress AP-1 in a ligandindependent manner, we used HPV18-positive HeLa cells, where the RAR␤ gene was ectopically expressed under the control of a ␤-actin promoter (31). Constitutive RAR␤ expression, in the absence of ligand (atRA), avoids a functional overlap with other retinoic receptors, because RAR␤ transcription is specifically triggered by RAR␣, a major co-regulator of RAR␤ mRNA induction (30).
As presented in Fig. 1, the cervical carcinoma cell line HeLa cells did not show any detectable RAR␤ either at the RNA or at the protein level. In contrast, stable transfection of HeLa cells reconstitutes RAR␤ transcription to an intensity originally detected in nontransfected cells 3 days after treatment with pharmacological doses of atRA. Constitutive transcription of RAR␤ had no influence on the expression of other RARs, as verified by examining RAR␣ (Fig. 1A). Ectopic expression of the RAR␤ gene encoded a functional protein, which could be confirmed when RARE-responsive reporter constructs were transfected. Basal activity of the clones was almost the same as in parental HeLa cells stimulated with atRA for 4 h (Fig. 1B). Whether the weak inducibility in parental HeLa cells was because of the delayed kinetics of endogenous RAR␤ expression (32) or whether the gene harbors in addition any point mutations (e.g. within the transcriptional activation function at the N-terminal region) (62) remains to be elucidated.
RAR␤-positive HeLa cells were negatively affected in their proliferation rate, which correlated with enhanced expression of growth inhibitory proteins such as p53, the cyclin-dependent kinase inhibitors p21 CIP1 and p27 KIP1 , and reduced HPV18 E6/E7 transcription rates (data not shown). Because of this latter property, we focused our interest on the transcription factor AP-1, which has been shown previously to play a pivotal role both in maintaining HPV expression (12,63,64) and in the determination of the in vivo phenotype (28). By monitoring AP-1 within the RAR␤-expressing cells by EMSAs, a strong reduction of its binding affinity to a cognate recognition site could be noted ( Fig. 2A). Decreased affinity was specific for AP-1, because the transcription factor Oct-1 was not affected. Additional treatment with atRA further reduced binding (Fig.  2B), indicating that the ligand-dependent suppression of AP-1 was still active in these clones (see below). By inspecting the steady-state levels of the corresponding Jun/Fos/ATF family members, an almost complete absence of c-Jun could be observed. Although c-Fos was also reduced to levels obtained in parental HeLa cells 3 days after atRA treatment, the lack of c-Jun was selective and independent of its own transcription (Fig. 2C). This result was consistent with the preceding EMSA data, clearly demonstrating that c-Jun represents the major AP-1 dimerization partner in HPV18positive cervical cancer cells (28,29). The reason why c-jun transcription was found to be enhanced even in the absence of c-Jun was at the first sight contradictory, because it is thought that c-jun expression is mediated through binding of c-Jun/ATF-2 dimers to cis-regulatory sequences within its own promoter (65). However, our data are consistent with previous studies where also other ATF2 dimerization protein(s) can substitute c-Jun for maintaining c-jun transcription (50,66).
Depending on the cell system, trans-repression of AP-1 by ligand-loaded RARs can potentially be mediated through the MAPK pathway by negative interference with JNK phosphorylation (19). Signal transduction is profoundly controlled by both strength and duration of MAPK activation. It is therefore also conceivable that constitutive RAR␤ expression affects the amount of dual specificity phosphatases (47), in particular the MAPK phosphatase MKP-1, which consecutively de-phosphorylates the tyrosine and threonine residues within the activation motif of JNK (67). Abrogation of JNK activity in turn both impedes the activity and intracellular availability of downstream target proteins such as c-Jun (68). To exclude the possibility that JNK is functionally disturbed due to constitutive RAR␤ expression or residual amounts of atRA within charcoalstripped serum, cells were treated with TNF-␣, a known inducer of MAPK (69). JNK activity was subsequently assayed in pull-down experiments by using cellular extracts after exogenous supplementation of c-Jun as a substrate. As shown for parental and RAR␤-expressing HeLa cells, added GST-c-Jun fusion proteins became equally phosphorylated after TNF-␣ treatment (Fig. 3A), clearly indicating that JNK activity upon cytokine treatment was not affected. Although it has been reported recently (70) that retinoids inhibit activation of serum-induced phosphorylation of JNK through a post-translational increase of MKP-1, constitutive RAR␤ expression was also not changing the steady-state level of MKP-1 under the same experimental conditions (Fig. 3B). These data were consistent with the notion that induction of MKP-1 is apparently only specific for particular RAR subtypes, because in contrast to RAR␣ and RAR␥, treatment with the RAR␤-selective agonist CD2314 is not elevating MKP-1 expression, even at higher concentrations (71).
The capability to phosphorylate c-Jun was not exceptional for proinflammatory cytokines such as TNF-␣ but could also be mediated through a transient overexpression of a constitutive active mutant of the mitogen-activated protein kinase kinase kinase 1 (MEKK1). MEKK1 represents a 196-kDa serine-threonine kinase, which is turned on in response to a variety of stimuli, including cytokines, growth factors, or anticancer drugs (72). A preferential substrate of MEKK1 is MKK4, which consecutively activates JNK (73). To substantiate further that JNK was not impaired in HeLa RAR␤ clones, an expression plasmid encoding a cytomegalovirus-driven MEKK1⌬ cDNA was transfected. EMSAs revealed that the amount of AP-1 binding to its cognate cis-regulatory binding site could be completely reconstituted (Fig. 4A). Pull-down kinase assays confirmed that activated JNK was responsible for this effect (data not shown). Supplementation of specific antibodies raised against the phosphorylated form of c-Jun resulted in a strong supershift, clearly corroborating that restoration of AP-1 binding was in fact due to phospho-c-Jun incorporation (Fig. 4B). Note that the amount of nuclear receptor expression remained unchanged after MEKK1⌬ cDNA transfection, excluding the possibility that transiently reduced levels of RAR␤ were accounting for enhanced AP-1 binding (Fig. 4D).
The intracellular availability of proteins is an important post-translational control mechanism to modulate their activity (74). Although the in vitro half-life of c-Jun seems also to be controlled by a certain class of cytoplasmic cysteine proteases called calpains, there is only limited significance of these enzymes on c-Jun degradation in living cells (75). c-Jun stability is mostly regulated through physical interaction with JNK at the ␦ domain (amino acids 30 -57), the docking site for targeting c-Jun for ubiquitination and proteolytic degradation through the multisubunit proteasomal complex (76). Considering Fig. 5A, one can conclude that constitutive expression of nonliganded RAR␤ seems to trigger a selective degradation of c-Jun via the proteasomal pathway, because the amount of protein can be completely restored by short term treatment with the peptidyl aldehyde MG132.
Moreover, JNK-directed c-Jun ubiquitination is only restricted to the unphosphorylated form of the substrate (56,77). Phosphorylation of c-Jun on Ser-73 by JNK, activated either by inflammatory cytokines (Fig. 3A) or by MG132 (Fig. 5A), acting in addition to its inhibitory function on the proteasomal machinery also as MAP kinase inducer (54), protects c-Jun from degradation resulting in a prolonged half-life. Although it has been reported, at least under in vitro conditions, that depletion of c-Jun increases the stability of ATF-2 as a heterodimerization partner by reducing proteasomal proteolysis (51), there was no indication for such a cross-talk in our experimental model system. Neither parental HeLa cells nor RAR␤-expressing clones showed quantitative differences even after 8 h of MG132 treatment. However, similar to c-Jun, ATF-2 became phosphorylated upon MG132 treatment (Fig. 5A). This explains enhanced c-jun mRNA synthesis (Fig. 5B) because c-Jun/ATF-2 heterodimers bind to related cis-regulatory sequences within the c-jun promoter, which in turn stimulates transcription via an autoregulatory loop (50,65). Stabilization of c-Jun resulted not only in its mere intracellular accumulation but also in its assembly into the AP-1 complex as detected EMSA (Fig. 5A). Incorporation of phosphorylated c-Jun could be finally verified in EMSA supershifts after addition of antibodies recognizing phosphorylation within the N-terminal domain (Fig. 5B). Accumulated c-Jun protein resulted in a functional AP-1 complex, because TPA-responsive element luciferase reporter constructs showed significant induction after MG132 addition (Fig. 5C).
Recently it has been reported that POH1 (human Pad1 homologue), a regulatory subunit of the proteasome, can selectively stabilize c-Jun (35,61). Although the endogenous steadystate level of POH1 was not inherently altered in all cell lines (not shown), it is likely that RAR-␤ or a target protein interacts with POH1 to inhibit its function. To prove the assumption whether POH1 overexpression can impede RAR␤-mediated c-Jun degradation, the corresponding cDNA was transfected both in parental HeLa cells and the RAR␤ clones. As demonstrated in Fig. 7A, transient overexpression of POH1 significantly increased the amount of c-Jun within the AP-1 complex.
The stability of other transcription factors such as Oct-1 was not affected (Fig. 7B). POH1 is a component of the 26 S proteasome and therefore may act as a de-ubiquitinating protein, selectively preventing c-Jun degradation (35). How this exactly functions and where RAR-␤ interferes with this process await further elucidation. Nevertheless, the mechanism described here has far reaching implications. It suggests a model where AP-1 activity can be modulated even in the presence of RAR␤ as a potential negative regulator (see Fig. 8 for schematic overview). As shown previously (30), RAR␤ is constitutively expressed in cervical keratinocytes. Hence, in nontreated cells, expression of nonliganded RAR␤ accelerates proteasomal degradation of c-Jun (Fig. 2). Conversely, either after cytokine stimulation (Fig.  3A) or after MAP kinase pathway activation via transfection of an active upstream MAP kinase (MEKK1⌬) (Fig. 4), cells still have the possibility to revert this effect, because activation of JNK is not inevitably perturbed (Fig. 3, A and B), at least not in the absence of atRA (19). In that way, c-Jun becomes phosphorylated and stabilized (Figs. 3A and 4B), which in turn increases transcription of its own mRNA (Figs. 4B and 5B). The more protein is available for phosphorylation by active JNK, the better the RAR␤ inhibitory function can be circumvented. This provides the possibility to limit the duration and magnitude of AP-1 activation, because JNK activation is known to be temporary (67). Notably, RAR␤ expression increases with serial passage in senescent cells, which coincides with AP-1 reduction (78,79). It will be an interesting task in further studies to investigate whether an analogous cross-talk is responsible for this effect. Moreover, a recent study on normal, premalignant, and malignant human laryngeal tissues has shown that during laryngeal tumorigenesis, RAR-␤ and AP-1 were inversely regulated. Progressive up-regulation of AP-1 was accompanied by a suppression of RAR-␤, indicating that these two regulatory proteins also play an important role in laryngeal carcinogenesis (46). Phosphorylation (P-c-Jun) stabilizes c-Jun (1AP-1) which in turn acts together with ATF-2 on its promoter to stimulate its own transcription (c-jun gene induction). The more protein is available for phosphorylation by active JNK, the better the RAR␤ inhibitory function can be circumvented. This provides the possibility to limit the duration and magnitude of AP-1 activation, because JNK is only transiently activated (JNK OFF).