Retinoic Acid Receptor β Silences Human Papillomavirus-18 Oncogene Expression by Induction of de Novo Methylation and Heterochromatinization of the Viral Control Region*

Retinoic acid receptor β2 (RARβ2) is often down-regulated during the multistep process to cervical cancer. In that way, its inhibitory function on the transcription factor AP-1, indispensable to maintain human papillomavirus (HPV) gene expression is relieved. Using HPV-18 positive HeLa cells as a model system, we show that ectopic expression of RARβ2 is able to down-regulate HPV-18 transcription by selectively abrogating the binding of AP-1 to the viral regulatory region in a ligand-independent manner. This resulted in down-regulation of the viral mRNAs at the level of initiation of transcription. Decreased oncogene expression was accompanied by a re-induction of cell cycle inhibitory proteins such as p53, p21CIP1, and p27KIP as well as by a cessation of cellular growth. Reduced transcriptional activity as a consequence of AP-1 reduction by selective c-Jun degradation apparently targets the HPV-18 regulatory region for epigenetic modification such as de novo methylation and nucleosomal condensation. This mechanism is otherwise counterbalanced by active and abundant viral transcription in malignant cells, because RARβ2 itself becomes inactivated during cervical carcinogenesis. Hence, our study shows that the temporal co-existence of a potential repressor and viral oncoproteins is mutually exclusive and provides evidence of a cross-talk between a nuclear receptor, AP-1, and the epigenetic machinery.

Cancer is a multistep process that is characterized by accumulation of various cell-damaging events such as chromosomal instabilities, inactivation of cell cycle regulatory proteins, and epigenetic modifications (1,2). In the case of cervical cancer, the second leading malignancy in women worldwide, certain types of human papillomaviruses (HPV) 2 have been identified as etiological agents being involved both in initiation and maintenance of the transformed phenotype (3). The transcription of the viral oncogenes E6 and E7 is controlled by the viral upstream regulatory region (URR), harboring the tissue-specific enhancer and promoter elements (4). Here, beside other cellular transcription factors, activator protein-1 (AP-1) is the main regulator determining the efficiency of expression and in turn the intracellular net amount of the viral oncoproteins E6/E7 (5).
As shown for many malignant cells, including HPV-positive cervical cancer cells, the retinoic acid receptor (RAR) ␤2 gene (RAR␤2), the most potent RAR involved in suppression of tumor-related phenotypes (6,7), is epigenetically silenced via de novo methylation and chromatin remodeling (8). In other words, the absence of RAR␤ provides a selective advantage during multistep progression to cervical cancer, because the gene is obviously not compatible with unscheduled cell proliferation (9,10). Indeed, re-introduction of RAR␤ into malignant cells interferes with anchorage-independent growth (6,11) and diminishes tumor formation upon heterotransplantation into immunocompromised animals (12). Together with other isotypes of retinoid acid receptors (␣ and ␥) or retinoid X receptors (RXR), which are encoded by different genes (13), the biological and therapeutic outcome of retinoids (e.g. all-trans-retinoic acid or 9-cis-retinoic acid) is either dependent on RAR homodimers or heterodimer formation with RXRs, which in turn bind to retinoic acid response elements or retinoid X response elements within their respective target promoters (14).
Both RARs and RXRs possess trans-repressive function on the AP-1 in a ligand-dependent manner. AP-1 is a key element in a regulatory network, playing not only a fundamental role in transcriptional regulation of various HPVs (15,16), but also in cell proliferation and tumor induction (5,17). In the presence of ligands, RARs or RAR/RXR can negatively affect AP-1 either by a direct interaction with Jun/Fos family members (18) or by disrupting Jun-Fos dimerization (19). Retinoic receptors can also compete with the recruitment of transcriptional co-activators (20) or inhibit the c-JUN N-terminal kinase (JNK), in turn preventing phosphorylation-dependent activation of c-JUN (21).
Moreover, as previously reported, we have unraveled a mechanism that explains how RAR␤ can also diminish AP-1 activity even in the absence of any ligand. Ectopic re-expression of RAR␤ abrogates AP-1 binding by a post-transcriptional mechanism where c-JUN is selectively degraded and not substituted in the AP-1 complex by other Jun family members. This uncovers a cross-talk between AP-1 and non-liganded RAR␤ and explains how AP-1 is turned off in the absence of an exogenous stimulus, but can be activated by the mitogen-activated protein (MAP) kinase pathway upon stimulation (22).
To follow up this functional interplay, the present study was aimed to investigate further consequences of ectopic RAR␤ expression in cervical carcinoma cells. Evidence is provided that AP-1 is the only factor involved in the transcriptional regulation of the HPV-18 URR that is quantitatively affected. Reexpression of RAR␤ was accompanied by a strong transcriptional down-regulation of the viral E6/E7 oncogenes and by an induction of cell cycle inhibitory proteins. Continuous availability of transcription factors can protect CpG islands from DNA methyltransferase activity (23). Conversely, deficiency of activating proteins (e.g. AP-1), followed by a reduced transcriptional activity, enhances the probability that regulatory regions become targets for de novo DNA methylation. Such a scenario seems to be the case after ectopic RAR␤ expression and subsequent AP-1 reduction, where HPV-18 E6/E7 transcription becomes down-regulated. Monitoring epigenetic modifications of the viral URR, our results establish a link between RAR␤-mediated AP-1 trans-repression, de novo DNA methylation, and transcriptional silencing of integrated copies of HPV-18 in cervical cancer cells.

EXPERIMENTAL PROCEDURES
Cell Lines and Treatments-The parental cervical carcinoma cell line HeLa and the stably transfected HeLa RAR␤ clones (12) were maintained in Dulbecco's modified Eagle's medium (Sigma), supplemented with 10% fetal calf serum (Invitrogen), 1% penicillin/streptomycin (Sigma). To keep the selection pressure, the clones were grown in a media containing 750 g/ml G-418 (Invitrogen). For retinoic acid treatment, cells were incubated with 10 M all-trans-retinoic acid (Sigma) as previously described (22). Decitabine (5-aza-2Ј-deoxycytidine, Calbiochem) treatment was carried out for 4 days, adding 1 M final concentration to the medium. Decitabine was diluted in 0.005% (v/v) of acetic acid. Culture medium was replaced daily. Control cells received medium containing 0.005% (v/v) acetic acid.
Transient Transfections-2.5 ϫ 10 6 cells were plated on 60-cm 2 dishes and transfected with 2 g of a cytomegalovirusdriven MEKK1⌬ cDNA. MEKK1⌬ is a constitutive active truncated form of the MEK kinase-1 (mitogen-activated/extracellular signal-regulated kinase kinase, MEKK1), which lacks parts of the N-terminal regulatory domain (24). Transfection was performed using "Effectene Transfection Reagent" (Qiagen) according to the manufacturer's instructions. Cells were harvested 48 h after transfection.
DNA Analysis and Southern Blotting-Total DNA was extracted according to standard methods. 5 g was digested with HindIII (5 units/g for 3 h), electrophoresed in a 1% agarose gel, blotted overnight on nylon membranes (Gene Screen, PerkinElmer Life Sciences), and hybridized with a radioactive labeled HPV-18 DNA probe (see below).
RNA Analysis, Northern Blotting, and RT-PCR-RNA was isolated with the "Absolutely RNA RT-PCR miniprep kit" (Stratagene) according to the manufacturer's instructions. Total RNA was separated on 1% agarose gels in the presence of ethidium bromide under non-denaturating conditions (5), blotted on nylon membranes (GeneScreen, PerkinElmer Life Sciences) and hybridized with radioactive labeled probes (25). Bands were quantified with a GE Healthcare PhosphorImager using the ImageQuant program as software. cDNA was obtained from 1 g of RNA using random primers (Roche) and SuperScript II reverse transcriptase (Invitrogen) in a volume of 20 l, following the manufacturerЈs recommendations. One l of the RT was amplified for 35 cycles consisting of 30 s at 94°C, 45 s at the corresponding annealing temperature, and 30 s at 72°C, with a final extension of 10 min. The following primers were used: for matrix metalloproteinase-1 (MMP-1), 5Ј-TGC-TGAAACCCTGAAGGTG-3Ј and 5Ј-CTGCTTGACCCTCA-GAGACC-3Ј (annealing temperature 55°C); for GAPDH (glyceraldehyde-3-phosphate dehydrogenase) 5Ј-TGGATATTGT-TGCCATCAATGACC-3Ј and 5Ј-GATGGCATGGACTG-TGGTCATG-3Ј (26) (annealing temperature 65°C). PCR products were analyzed in 2% agarose gels.
Restriction Enzyme Accessibility Assay on Cross-linked and Native Chromatin-1 ϫ 10 7 cells were fixed with 1% formaldehyde and cross-linking was stopped by adding 0.125 M glycine as described below (see "ChIP assays"). After centrifugation, 1 ϫ 10 6 nuclei were resuspended in 1ϫ MfeI restriction enzyme buffer (New England Biolabs) supplemented with 0.3% SDS and incubated 1 h at 37°C to remove uncross-linked proteins. 1.8% Triton X-100 was added for 10 min to sequester SDS. Finally, 1ϫ bovine serum albumin and 200 units of restriction enzyme (MfeI) were added and the digestion was incubated overnight at 37°C. Next day, proteins were digested by 100 g/ml proteinase K at 37°C and the DNA was uncross-linked by 0.3 M NaCl at 65°C. Digested DNA was purified by phenol/chloroform extraction and ethanol precipitation. For native nuclei isola-tion, 1 ϫ 10 7 cells were seeded. Next day, cells were washed in phosphate-buffered serum supplemented with protease and phosphatase inhibitors (5,29), scrapped, and resuspended in 500 l of TBS buffer (150 mM NaCl, 100 mM Tris-HCl, pH 7.5, 3 mM CaCl 2 , 2 mM MgCl 2 ) supplemented with proteases and phosphatases inhibitor (5,29). Cell lysis was done by adding 1% Tween 20 for 1 h at 4°C. Nuclei were collected by centrifugation for 15 min at 5000 ϫ g at 4°C. Subsequently, 1 ϫ 10 6 nuclei were resuspended in 1ϫ restriction enzyme buffer, 1ϫ bovine serum albumin, and 200 units of the restriction enzyme. The digestion was incubated at 37°C overnight, the DNA was isolated as described above and PCR amplified (30) (see below "PCR Analysis of Immunoprecipitated Chromatin"). As control for MfeI functionality, 10 g of genomic DNA was digested (5 units/g of DNA) for 2 h at 37°C.
ChIP Assays-1 ϫ 10 7 cells were cross-linked with 1% formaldehyde, stopped by 0.125 M glycine and briefly washed with phosphate-buffered serum, trypsinized 20 min at 37°C, and scrapped off. Cell pellets were resuspended in 1 ml of cell lysis buffer (10 mM Tris-HCl, pH 8, 10 mM NaCl, 0.2% Nonidet P-40) supplemented with proteases and phosphatases inhibitors (5, 29). 1.5 ϫ 10 6 nuclei per ChIP were enzymatic sheared (Active Motif enzymatic shearing kit) according to the manufacturer's instructions. After shearing, an aliquot of chromatin (equivalent to 25 g of DNA) was removed and stored at Ϫ20°C. Chromatin fractions were pre-cleared adding 40 l of protein A-agarose blocked with 40 g of salmon sperm during 2 h at 4°C on a rotating plate. The suspensions were then centrifuged to discard nonspecifically bound chromatin fragments. Supernatants (equivalent to 25 g of DNA) were incubated with 1.5 g of RNA polymerase II antibody (Abcam, ab-5131-50, lot 37712) or 2 g of rabbit IgG antibody (Santa Cruz, sc-2027, lot C1605). Volumes were adjusted to 500 l using ChIP buffer (16.7 mM Tris, pH 8, 167 mM NaCl, 1.2 mM EDTA, pH 8, 1.1% Triton X-100, and 0.01% SDS) supplemented with proteases and phosphatases inhibitors. Next day, ChIPs were incubated with 40 l of blocked protein A-agarose beads for an additional period of 3 h. The immunocomplex containing chromatin fragments/␣-RNA pol II/protein A-agarose was recovered by centrifugation at 13,000 ϫ g for 2 min and washed once with buffer 1 (500 mM NaCl, 50 mM Tris, pH 8, 5 mM EDTA, pH 8 and 0.5% Nonidet P-40), twice with buffer 2 (150 mM NaCl, 20 mM Tris, pH 8, 1% Triton X-100, and 0.1% SDS), twice with buffer 3 (250 mM LiCl, 10 mM Tris, pH 8, 1 mM EDTA, pH 8, 0.5% Nonidet P-40, and 0.5% sodium deoxycholate), and finally twice with TE (10 mM Tris, pH 8, and 5 mM EDTA, pH 8). The immunoselected chromatin was eluted from protein A-agarose by adding 250 l of elution buffer (1% SDS, 100 mM NaHCO 3 ). The eluted chromatin was diluted 1:1 in a uncross-linking buffer (320 mM NaCl, 80 mM Tris, pH 8, 20 mM EDTA, pH 8, and 100 g/ml proteinase K) and incubated overnight at 65°C. DNA samples were purified and used for semiquantitative PCR. We routinely tested the specificity and linearity of chromatin immunoprecipitation by semiquantitative PCR. Utilizing RNA polymerase II antibodies for ChIP analyses of the GAPDH promoter (euchromatin) in comparison with the MYO D coding region (facultative heterochromatin), the following primers were used: GAPDH promoter, 5Ј-CCCAACTTTCCCGCCTCTC-3Ј and 5Ј-CAGCCG-CCTGGTTCAACTG-3Ј (31); MYO D coding region, 5Ј-CCGC-TTTCCTTAACCACAAA-3Ј and 5Ј-CGGGTCGTCATAG-AAGTCGT-3Ј. For double ChIP experiments, chromatin immunoprecipitation was first carried out with RNA pol II antibodies and the trapped protein-DNA complexes were eluted in 50 l of buffer (1% SDS, 100 mM NaHCO 3 ). The eluted chromatin was diluted 1:10 in ChIP buffer and 20 g of P-c-Jun antibodies (sc-822X, lot E0704, Santa Cruz Biotechnology) were added. The second precipitation was carried out overnight at 4°C and immunocomplexes were pulled down and washed as described above.

Constitutive Expression of Non-liganded RAR␤ in HeLa Cells-
Reconstitution of RAR␤ expression under the control of the ␤-actin promoter in cervical carcinoma cells (Fig. 1A, right) resulted in a selective reduction of viral transcription (Fig. 1A, left) and a strong cessation of cellular growth (data not shown). Consistent with the decrease of HPV-18 on the RNA level, there was also a decline of the corresponding viral oncoprotein E7 (Fig. 1A, right). Viral down-regulation in the absence of any ligand occurred to the same extent as detected in non-transfected HeLa cells, incubated with pharmacological doses of alltrans-retinoic acid for 3 days. Note that viral transcription was not completely suppressed, because at least a certain threshold amount of oncogene expression is required to maintain cell proliferation (32). Moreover, RAR␤ restoration was also not altering the HPV-18 copy number due to the inherent karyotype instability of many cancer cells (33), because Southern blotting revealed no rearrangement or loss of viral genomes of the clones when compared with parental HeLa cells (see supplementary Fig. S1).
To examine the consequence of constitutive RAR␤ expression on cell cycle regulatory proteins, Western blot analyses were performed. The results depicted in Fig. 1B show not only an increase of p53 and the cyclin-dependent inhibitors p21 CIP1  590 -907). D, semiquantitative PCR analysis of the RNA pol II ChIP assay. Inp., input; IgG, DNA immunoprecipitated with immunoglobulin G, nonspecific antibody; Pol II, DNA immunoprecipitated with a phosphorylation specific polymerase II antibody. E, same as D, semiquantitative PCR analysis of ChIP assays using the GAPDH promoter (euchromatin) and the MyoD coding region (heterochromatin) as target for RNA pol II antibodies. Each sample was titrated using increasing amounts of recovered DNA to show the linearity of the PCR. PCR products were separated in 2% agarose gels. The sizes of the PCR fragments are indicated. and p27 KIP , but also down-regulation of the mid-G 1 cyclin D1. In contrast, no quantitative changes of the cyclin-dependent kinase-2 (CDK-2) could be discerned.
Decline of HPV-18-specific mRNA can be either due to labilization and shortening of the cytoplasmic half-life or by direct suppression of viral URR-directed transcriptional activity (34). To discriminate between these possibilities and to quantify the actual transcription rate at the HPV-18 integration locus, chromatin immunoprecipitation against transcriptional active RNA polymerase II were carried out (RNA-Pol-ChIP assays) (35). Semi-quantitative PCR with three different sets of primers covering the coding region of the E6/E7 oncogenes (Fig. 1C) revealed that viral transcription was diminished up to 80% (as determined by quantitative PCR analysis, data not shown) in comparison with non-transfected cells (Fig. 1D). To demonstrate the selectivity of this mechanism, two additional genes either indicative for eu-or heterochromatin were examined. GAPDH represents a housekeeping gene that is constitutively expressed, whereas myogenic factor MYO D is only active during muscle differentiation, but not in epithelial cells (36). As depicted in Fig. 1E, GAPDH shows equal loading of RNA pol II in all cell lines, whereas the heterochromatized MYO D gene cannot be immunoprecipitated by pol II antibodies, because of absence of transcriptional activity. These data indicate that reconstitution of RAR␤, even in the absence of all-trans-retinoic acid, leads to a selective suppression of HPV-18 E6/E7 oncogene expression at the level of initiation of transcription.
AP-1 Binding at the HPV-18 URR Is Selectively Reduced-Because the steady-state levels of endogenous HPV-18 E6/E7 mRNA was downregulated in RAR␤ clones, we reasoned that constitutive expression of non-liganded RAR␤ was negatively affecting AP-1 activity, the major factor in URR-directed transcription (15,37). This was mainly deduced from our previous finding showing that RAR␤ abrogates AP-1 binding by selective degradation of c-JUN (22), the major dimerization partner with FOS family members in cervical carcinoma cells (5,29) (see also Fig. 2D). To test this notion in the context of HPV regulation, we examined the two AP-1 binding sites within the URR in comparison with an AP-1 site in the human metallothionein IIA promoter (hMT IIA) (27) by EMSAs. Fig. 2A demonstrates that AP-1 binding in control cells to the cognate HPV-18 enhancer region (referred as "E") and the hMT IIA promoter (referred as "C") was similar in parental HeLa cells, whereas AP-1 binding within the viral promoter region (referred as "P") was even enhanced. However, inspecting the RAR␤ clones under the same experimental conditions, binding was drastically reduced to levels usually only obtained after treatment of non-transfected cells with all-trans-retinoic acid (70 -80% reduction, as determined by PhosphorImager analysis, data not shown). To confirm that decreased AP-1 binding at the HPV-18-URR was a selective process, EMSAs with other 32 P-labeled URR-derived oligonucleotides were carried out. In fact, under conditions where AP-1 was reduced, neither the binding of enhancer binding factor Oct-1/KRF-1 nor Sp1 binding at the viral promoter was affected (Fig. 2, B and C). Depletion of c-JUN binding on native chromatin could be also demonstrated by RNA pol II/c-JUN double ChIP assays, where in contrast to parental HeLa cells, no HPV-18 E6-specific PCR fragment was obtained (Fig.  2E). The selectivity of this assay was confirmed using the GAPDH gene as control, because the corresponding promoter lacks functional AP-1 sites. Here, in contrast to the previous assay (see Fig. 1E), no PCR fragment could be amplified. These results demonstrate that decreased AP-1 binding at the HPV-18-URR was a selective process, without impairing the binding of other transcription factors involved in URR-directed E6/E7 expression.  ing to the enhancer and promoter region of HPV-18 URR after MAP kinase activation showed that MEKK1⌬ expression can completely restore AP-1 binding to the HPV-18 URR enhancer and promoter without affecting the binding of other transcription factors such as Oct-1 (Fig. 3A). Because the integrity of both AP-1 elements within the URR is absolutely required for efficient transcription (15,16), we therefore anticipated that restoration of AP-1 upon MEKK1⌬ delivery should be also accompanied by a reconstitution of E6/E7 expression. However, as outlined in Fig. 3B, this was not the case. Instead, as provided for parental HeLa cells, endogeneous HPV-18 transcription was even further diminished. To resolve this apparent contradiction, supershift EMSAs after MEKK1⌬ transfection were performed. Fig. 3C shows that not only c-JUN was reconstituted and incorporated into the AP-1 complex (lower panel), but also FRA-1 (middle panel), an AP-1 family member, known to be involved in HPV negative regulation (5,29). In contrast, c-FOS was only slightly affected (upper panel). To ensure the biological activity of reconstituted AP-1, we used the matrix metalloproteinase-1 (MMP-1) as reference gene, whose expression is both dependent on AP-1 functionality and on the availability of FRA-1 (38,39). Whereas HPV-18 E6/E7 transcription (Fig. 3B) remained down-regulated even after reconstituted AP-1 binding at the HPV-18 URR (Fig. 3, A and C), a significant increase of MMP-1 could be visualized (Fig. 3D). Although the extent of MMP-1 transcription differed in comparison with parental HeLa cells, these data indicate that in contrast to cel-lular genes, viral gene expression in RAR␤ cells was refractory against AP-1 reconstitution.

De Novo DNA Methylation and Heterochromatinization Around the AP-1 Enhancer Binding Site of HPV-18 in RAR␤2 Expressing HeLa
Cells-The lack of E6/E7 up-regulation suggests that not AP-1 per se, but a cis-regulatory mechanism may account for the suppressive effect of constitutive RAR␤ expression. In fact, decreased transcriptional activity due to the absence of a trans-activator may lead to epigenetic modification, which comprises histone modification and spreading of de novo DNA methylation at the respective promoter sequences (40,41). Because HPV can be also targeted by the host methylation machinery (42) and URR-directed transcription is negatively affected by in vitro DNA methylation (43,44), we monitored the CpG sequences (embedded within HhaI recognition sites) nearby the AP-1 binding site within the enhancer region. Using the appearance of a 254-bp PCR product (after lack of DNA cleavage with the methylation-sensitive restriction enzyme HhaI) as indicator for the presence of methylation, Fig. 4A shows that RAR␤ expressing HeLa cells exhibit de novo DNA methylation within this region. In contrast, HeLa parental cells were not (or not entirely) methylated because no PCR fragment could be amplified. The presence of a 300-bp product covering the E6 region confirmed that equal amounts of DNA were used for the PCR.
To ascertain the selectivity of this process, the non-coding 3Ј-variable tandem region (VTR) of the Ha-Ras protooncogene was analyzed. The VTR of Ha-ras is GC-rich and known to be targeted by DNA methylation (45). As depicted in Fig. 4B, the HpaII sites within the VTR were protected in all cells, giving rise to a 292-bp PCR fragment. This indicates that there were apparently no differences in methylation in general between parental HeLa cells or RAR␤ expressing clones. Notably, treatment of cells with decitabine did not induce demethylation of the Ha-Ras 3Ј-VTR, which is consistent with previous reports showing that certain genes were refractory to demethylation (46,47). In contrast, when RAR␤ expressing clones were treated under the same experimental conditions, the 254-bp fragment could no longer be amplified, indicating that demethylation of the HPV-18 enhancer was successful (Fig. 4A).
DNA methylation of a regulatory region may have principally two consequences: (i) interference with the binding of transcription factors (48) and/or (ii) being a step of a complex network of subsequent chromatin modifications, finally leading to a close chromatin structure and reduced accessibility of tran- scription factors to their corresponding binding sites (42). To analyze whether de novo DNA methylation of the HPV-18 URR block AP-1 binding simply due to sterical hindrance, oligonucleotides were in vitro methylated by HhaI methylase. Using nuclear extracts from parental HeLa cells, EMSAs showed that the same binding efficiency could be discerned (data not shown). Moreover, to examine whether AP-1 differs in its affinity to methylated versus non-methylated DNA, cross-competition experiments with increasing molar excess of non-labeled oligos were carried out. Fig. 4C reveals that AP-1 binds with the same affinity suggesting that methylation per se was not sufficient to impair AP-1 binding.
To check whether de novo methylation within the URR was associated with a condensed nucleoprotein structure (49,50), rendering the region more resistant to exogenously added restriction endonucleases (43,51), isotonic nuclei were digested with MfeI. Fig. 4D shows a schematic overview of the MfeI restriction sites relative to the enhancer AP-1 binding site and the position of the primers used in this assay. Whereas the URR of HeLa RAR␤ clones exhibited much less accessibility to MfeI, as indicated by the appearance of a 254-bp amplification product, chromatin of HeLa parental cells was highly susceptible under the same experimental conditions. Here, no protected fragment could be amplified. The same results can be obtained when chromatin was digested with MfeI after cross-linking with formaldehyde (see supplementary Fig.  S2). Control PCRs of an adjacent E6 coding region confirmed that comparable DNA inputs was used for each reaction. Because DNA methylation is mostly linked with nucleosomal condensation (52,53), these results indicate that the regulatory region of most HPV-18 copies in the RAR␤ expressing cells were in a heterochromatic state.
Induction of E6/E7 Expression Requires Demethylation Prior to AP-1 Reconstitution-To test whether demethylation of the URR can restore HPV-18 E6/E7 gene expression, HeLa RAR␤ clones were treated for 4 days with the demethylating agent decitabine. However, even though the viral URR was demethylated (Fig. 4A), no quantitative changes of the HPV-18 E6/E7 mRNA steady-state levels could be observed (Fig. 5A, compare  lanes 5, 6 and lanes 9, 10, respectively). This was further corroborated by RNA polymerase II ChIP assays (Fig. 5B), also demonstrating an absence of augmented transcriptional activity covering the E7 coding region. Although decitabine itself can potentially induce the MAP kinase pathway (54), accumulation of c-JUN was apparently not sufficient to increase AP-1 to threshold levels to enhance viral gene expression (data not shown). Conversely, AP-1 binding in RAR␤ clones could be readily induced by serum addition (Fig. 5C), but HPV-18 transcription was not elevated (Fig. 5, D and E). Only when cells were preincubated with decitabine and subsequently stimu- lated by serum addition, could oncogene expression be restored (Fig. 5A, compare lanes 5, 8 and lanes 9, 12, respectively). To prove that combined treatment had a transcriptional effect, ChIP assays revealed increased RNA polymerase II activity spanning the coding region (Fig. 5F). In contrast, neither decitabine nor serum addition alone has a significant effect on RNA polymerase II loading (Fig. 5, B and E), which is consistent with the absence of HPV-18 mRNA induction (see Fig. 5, A and D). These data indicate that restored expression of RAR␤2 not only reduces the amount of AP-1 (Fig. 2), but also leads to a cisinhibitory effect on the viral upstream regulatory region by inducing de novo DNA methylation (Fig. 4A) and heterochromatinization (Fig. 4D).

DISCUSSION
The present study set out to determine the biological outcome of RAR␤2 restoration in cervical carcinoma cells, known to become down-regulated during cervical carcinogenesis (9,10,55). Re-expression of physiological levels of non-liganded RAR␤2 in HPV-18-positive HeLa cervical carcinoma cells (Fig. 1A, right) induced not only growth retardation in vitro and in vivo (12), but was also accompanied by a strong down-regulation of the E6/E7 mRNA at the level of transcription (Fig. 1, A and D). Reminiscent to the repressive function of the HPV E2 regulatory protein, predominantly lost after HPV integration into the host genome (56,57), ectopic expression of RAR␤2 resulted in an up-regulation of p53 and the cyclin-dependent kinase inhibitors (CKI) p21 CIP1 and p27 KIP as well as suppression of cyclin D1 (Fig. 1B).
Similar to the reduction of HPV-18 URR-directed transcription, cyclin D1 down-regulation is consistent with the finding that c-JUN is selectively degraded upon RAR␤2 restoration (see Fig. 2D). Cyclin D1 contains two AP-1 binding sites within its promoter and c-JUN has been found to induce its transcription (58). Referring to p53 and p21 CIP1 , two mechanisms may account for their increase: notably, c-JUN can also act as the negative regulator of p53 via a conserved atypical AP-1 site within the p53 promoter and in turn with the activation of its downstream target gene (59). In favor with this notion is the observation that c-jun Ϫ/Ϫ knockout cells actually contain higher basal levels of p53 than their parental counterparts (58). In addition, reduction in the E6 expression (Fig. 1A) rescues p53 from E6-AP-mediated proteasomal degradation, therefore leading to its increase on protein level (60). p27 KIP up-regulation has also been demonstrated after RAR␤ re-expression in lung cancer cells or after neuroblastoma differentiation (61,62), but the exact mechanism is still elusive.
Regarding host regulatory proteins that bind to the viral URR, the transcription factor AP-1 plays an essential role in determining both the efficiency of HPV expression (4,16,63) and the phenotype of HPV-positive cells in nude mice (5,29). Monitoring AP-1 in comparison with other transcription factors within the viral enhancer and promoter region, decreased binding was only detectable for AP-1, whereas Oct-1/KRF-1 and Sp-1, each participating in the transcriptional regulation of the E6/E7 promoter (4), were not affected (Fig. 2, A-C). This is in accord with the fact that ectopic re-expression leads to a Inp., input; IgG, DNA immunoprecipitated with immunoglobulin G, nonspecific antibody; Pol II, DNA immunoprecipitated with a phosphorylation specific polymerase II antibody. The size of the E7-specific PCR fragment is indicated. C, EMSA showing AP-1 binding reconstitution after serum addition in RAR␤ expressing cells (clone ␤2). D, Northern blot, absence of HPV-18 induction after serum addition in parental HeLa or RAR␤ expressing cells. Order of treatments, see A for details. E, semiquantitative PCR analysis of RNA pol II ChIP assays after 6 h serum induction. The size of the E7-specific PCR fragment is indicated. F, semiquantitative PCR analysis of the RNA pol II ChIP assays after decitabine treatment for 4 days and subsequent fetal calf serum addition for 6 h. PCR products were separated in 2% agarose gels. The sizes of the PCR fragments are indicated.
selective depletion of c-JUN (see Fig. 2D), whereas AP-1 binding can be restored when its proteasomal degradation is prevented (22).
The ability to phosphorylate and stabilize c-JUN cannot only be mediated by proinflammatory cytokines (64) but also by transient overexpression of a constitutive active mutant of the MEKK1⌬. A preferential substrate of MEKK1 is MKK4, in turn activating the JNK pathway (65). We have previously shown that JNK was not impaired in HeLa RAR␤ clones, because the MEKK1⌬ mutant was also able to rescue c-JUN from degradation (22). EMSAs revealed that the amount of AP-1 binding to its two cognate cis-regulatory sites within the HPV-18 URR could be completely restored under these conditions (Fig. 3A). However, AP-1 reconstitution had no positive functional effect on HPV-18 E6/E7 transcription, but even more reduces viral gene expression in parental HeLa cells (Fig. 3B). This discrepancy is apparently due to the fact that MEKK1⌬ transfection not only increased c-JUN incorporation into the AP-1 complex, but also FRA-1 (Fig. 3C), known to be involved in viral negative regulation (5,29). Nonetheless, these data suggest that features other than AP-1 availability may regulate the response of HPV-18 URR after MAP kinase activation.
We therefore examined potential epigenetic modifications of the HPV-18 URR, monitoring the CpG sequences nearby the AP-1 binding site within the tissue specific enhancer (43). These sequences coincide with cleavage sites of methylation sensitive restriction enzyme HhaI (Fig. 4A), which can be used as a tool to screen for cleavage protection followed by PCR in the case of de novo DNA methylation. Notably, hypermethylation of integrated HPV genomes can be considered as part of the multistep process to cervical cancer (43,66,67), particularly when tandem integration of unit-length HPV DNA occurred. De novo methylation obviously selects against the expression of those copies encoding the viral E2 repressor protein, known to negatively interfere with URR-directed E6/E7 transcription (44). For instance, such a situation is found in the cervical carcinoma cell line CaSki, where most of the 600 unit-length copies of HPV-16 were extensively hypermethylated (43,68). Although HeLa cells harbor between 30 and 50 genome equivalents of HPV-18, integrated at different chromosomes (69,70), none of them express the E2 repressor and consequently the URR was not found to be methylated (Fig. 4A). In contrast, RAR␤ expressing cells revealed that CpG sites around the AP-1 binding site within the viral enhancer were de novo methylated (Fig. 4A), forming a distinct nucleosomal structure with reduced accessibility against exogenously added nucleases (Fig.  4D, see also supplementary Fig. S2). Alterations in chromatin can therefore account for the inability of restored AP-1 binding to its cognate cis-regulatory sequences within the URR (Fig.  3A), because in vitro methylation of corresponding oligonucleotides did not prevent binding when EMSAs were performed (Fig. 4C). This argues against a simple steric hindrance of methylated DNA as described for E2 protein binding to the URR (44).
The induction/maintenance of gene silencing and heterochromatin formation is mediated by a concerted action between DNA methyltransferases, methyl-CpG-binding proteins, and accessory co-repressors (e.g. histone deacetylases) (for review, see Ref. 71). Whereas it have been reported that both de novo and maintenance DNA methyltransferases in conjunction with methyl-CpG-binding proteins can be up-regulated in certain cancers (72,73), no increase of the respective genes could be noticed in RAR␤ expressing HeLa cells (data not shown). Moreover, because E7 was suppressed on the transcriptional level (Fig. 1A), an enhanced methyltransferase activity due to physical interaction between E7 and DNA methyltransferase 1 (74) can be excluded.
Removal of methyl residues after decitabine treatment had no beneficial effect on the transcription of HPV-18 oncogenes (Fig. 5, A and B), which can be explained by a still reduced AP-1 binding upon RAR␤ expression ( Fig. 2A). Conversely, mere accumulation of AP-1 after serum stimulation (Fig. 5C) also did not induce viral transcription due to an impaired chromatin access to the URR, as revealed by the reduced susceptibility to exogenously added nucleases (Fig. 4D) and the lack of RNA polymerase II loading at the viral integration locus (Fig. 5E). However, when demethylation was followed by serum addition, AP-1 binding was reconstituted and an increase in the oncogene expression could be achieved (Fig. 5, A and F).
In summary, we propose a model where HPV-18 gene silencing upon ectopic re-expression of RAR␤ is first triggering c-JUN down-regulation (22). This leads to a reduced transcriptional activity, which in turn labels the viral URR for subsequent epigenetic modification. De novo DNA methylation and following heterochromatinization (or vice versa) is then no longer counterbalanced by a functional availability of AP-1 and therefore an efficient viral enhanceosome cannot be built up (75). Hence, not only the binding strength but also the concentration of AP-1 seems to be a critical determinant. Experimental evidence that support this mode of regulation derives from artificially assembled episomes where binding of the lac repressor on operator sites prevents de novo methylation in cells overexpressing DNA methyltransferases 3a (23) (for review, see Ref. 71). Whether RAR␤, similar to the orphan receptor germ cell nuclear factor (GCNF) (76), is directly involved in the recruitment of DNA methyltransferases to the URR remains to be elucidated.