Epigenetic Silencing of Cell Adhesion Molecule 1 in Different Cancer Progenitor Cells of Transgenic c-Myc and c-Raf Mouse Lung Tumors

Understanding molecular mechanisms underlying lung cancer is a prerequisite toward treatment. To enable mechanistic investigations into the epigenetic regulation of the tumor suppressor gene cell adhesion molecule 1 (Cadm1) in lung cancer progenitor cells, we developed 10 cell lines from single, spontaneously transformed lung tumor cells isolated from c-Myc and c-Raf double-transgenic mice. Specifically, we investigated Cadm1 promoter hypermethylation, which was significantly induced in transgenic transformed cells. Analysis of 69 CpGs displayed differential methylation pattern between and within progenitor cell lines, and the degree of methylation correlated well with transcriptional repression. Indeed, restoration of Cadm1 gene expression was achieved by treatment with the experimental demethylating drug 5-aza-2′-deoxycytidine. Furthermore, methylation of core CpGs in the binding sites of Sp1, Sp3, and zinc finger 5 along the promoter region of Cadm1 abrogated DNA-protein binding. Treatment with mithramycin A, an inhibitor of Sp1 or Sp3 binding, resulted in reduction of Cadm1 gene expression, therefore suggesting a potential role of Sp1/Sp3 in Cadm1 regulation. Identifying molecular rules for the epigenetic control of tumor suppressor genes enables mechanistic insights into lung cancer growth and opportunities for novel therapies. [Cancer Res 2008;68(18):7587–96]

Lung cancer is a leading cause of death, but molecular mechanisms underlying the disease are largely unknown. Thus, mouse models for human lung cancer have been developed in recent years to understand disease mechanisms (see reviews, refs. 1, 2). These models include mouse strains with spontaneous or carcinogen-induced tumors, as well as transgenic and knockout mice, in which lung tumors arise due to engineered genetic modifications. There is evidence for the proto-oncogenes c-Myc and c-Raf to play a role in human lung malignancies (3, 4), and transgenic mouse lines for these proto-oncogenes have been established to study tumorigenesis of adenocarcinomas derived from alveolar epithelium (5–7). Transgenic c-Myc mice developed within 10 to 14 months bronchioloalveolar carcinomas and papillary adenocarcinomas, whereas transgenic c-Raf mice exhibited atypical adenomatous hyperplasia as early as 3 months, postnatally. In the past, proteome mapping of c-Raf tumors has been reported (8).

Here, we developed 10 cell lines from single, spontaneously transformed lung tumor cells isolated from mice double-transgenic for the proto-oncogenes c-Myc and c-Raf. These lines would enable mechanistic investigations into the regulation of the gene encoding cell adhesion molecule 1 (Cadm1; Mouse Genome Informatics) in lung cancer progenitor cells. Cadm1 is the mouse orthologue of the human gene CADM1 (TSLC1, IGSF4; for other synonyms, see Human Gene Nomenclature;¹

ref. 9). CADM1 displays tumor suppressor activity in human cancer, particularly non–small cell lung cancer (NSCLC; ref. 10). Indeed, loss or reduction of CADM1 protein expression was observed in human lung adenocarcinomas and correlated with poor prognosis of patients, thus indicating clinical significance (11–13). Nonetheless, this relationship may not be true for all types of lung tumors. After analysis of 47 primary lung adenocarcinoma of patients, only ∼40% had decreased CADM1 protein expression (12). Specifically, CADM1 was similar to healthy tissue in 16 bronchioloalveolar carcinomas (BAC) and 12 adenocarcinomas other than BAC, but a further study on CADM1 protein in 93 lung adenocarcinomas evidenced loss of expression in 65% of the cases (13). This loss of expression correlated significantly with the non-BAC component and proliferative activity of the tumors. Lung adenocarcinomas with high CADM1 expression showed better prognosis than those without expression. When CADM1 protein was reexpressed in an NSCLC line A549, induction of apoptosis and inhibition of tumor growth were observed, suggesting the potential of CADM1 for gene therapy (14).

In general, the CADM1 gene encodes a 442-aa immunoglobin superfamily cell adhesion molecule and participates in cell-cell interactions (15). The CADM1 protein is a transmembrane glycoprotein consisting of an extracellular domain containing three immunoglobin-like C2-type domains, the transmembrane domain, and the cytoplasmic domain (10, 15). In vivo tumorigenesis in nude mice showed that deletion of the cytoplasmic domain abrogates the tumor suppressor activity of Cadm1 (16), and epigenetic silencing through promoter hypermethylation of CADM1 was observed in lung cancer (10, 17, 18).

Notably, epigenetic silencing of tumor suppressor genes is a frequent phenomenon in cancer (for instance, see reviews; refs. 19, 20), but the epigenetic mechanisms are still unknown. To gain more insights into the epigenetic control of Cadm1 resulting in transcriptional repression, we have undertaken a comprehensive analysis of methylated CpGs along the promoter region of Cadm1. We also determined whether methylation of specific CpGs within putative binding sites of transcription factors in the promoter region of Cadm1 would lead to abrogation of binding and confirmed results by electrophoretic mobility shift assay (EMSA) supershift functional assays. Identifying molecular rules for the epigenetic control of tumor suppressor genes in lung cancer progenitor cells of transgenic c-Myc and c-Raf mouse lung tumors enables mechanistic insights into lung cancer growth and opportunities for novel therapies.

Lung tumors and lung cancer cell lines. We analyzed 6 lung tumors and 10 cancer cell lines from c-Myc and c-Raf transgenic mice. The development of the c-Myc and c-Raf double-transgenic mice and the histopathology of lung tumors resulting from the overexpression of these proto-oncogenes form part of a related study on global gene expression analysis.²

Cells were isolated from lung tumors of 8-month-old transgenic mice using our protocol for the isolation and primary culture of rodent respiratory epithelial cells (21). With this protocol, by day 4, mRNA for the pulmonary surfactant-associated protein C (Sp-C), which is specifically activated in alveolar epithelium, can be detected in nontransgenic cells. Normal lungs from three nontransgenic mice were used as control.

Gene expression analysis. Total RNA was isolated from frozen mouse lung tissues or cell lines with RNeasy Mini kit, and reverse transcription–PCR (RT-PCR) was undertaken with Omniscript RT kit (Qiagen). Semiquantitative RT-PCR was performed on T3 thermocyclers (Biometra), whereas quantitative RT-PCR on Light Cycler (Roche) using Thermostart Taq polymerase (ABgene).

Typically, 25 to 50 ng of cDNA were used for template. Reaction components and cycling variables are according to standard procedure. For both semiquantitative and quantitative RT-PCR, the primers used for Cadm1 consisted of forward primer 5′-caactatccctcctcccaca-3′ and reverse primer 5′-tagctgtgtctgcgtctgct-3′.

5-Aza-2′-deoxycytidine treatment of cells. Cells (3 × 10⁶) were seeded in T25 cell culture flasks containing 5 mL of DMEM with 10% FCS, 2× l-glutamine, and 2× penicillin/streptomycin. Cells were cultured 48 h, treated with fresh 2 μmol/L 5-aza-2′-deoxycytidine (5-aza-dC; Sigma) dissolved in medium for 3 d, and allowed to recover for 2 d.

Prediction of promoter region and DNA methylation analysis. To predict for the promoter region of mouse Cadm1 and to identify CpG island in this region, we used the 2,000 nt (−2,000 bp) of the translation start site ATG. We used publicly available prediction programs, such as Promoter 2.0,³

Promoter Scan,⁴ and CpG island searcher.⁵ Genomic DNA was isolated in tissue samples and cell lines using Nucleo Spin Tissue (Macherey-Nagel). Bisulfite treatment was undertaken on 1 μg genomic DNA using CpGenome DNA Modification kit (Chemicon International) or EpiTect Bisulfite kit (Qiagen) using manufacturer's instructions, but treatment at 65°C instead of 60°C. Primers for methylation assays (Supplementary Table S1) were designed with MethPrimer.⁶ PCR fragments were directly sequenced using BigDyeTerminator v3.1 kit and ABI 3100 Genetic Analyzer (Applied Biosystems), or PCR fragments were cloned using TOPOTA Cloning kit (Invitrogen) before sequencing. Sequences were analyzed using SeqMan (Lasergene 7.0). Differential methylation was determined in at least five clones.

Isolation of protein extracts, Western blot analysis, and EMSA. Procedures for the isolation of protein extracts, Western blot analysis, and EMSA have been described previously (22). Protein extracts were isolated from mouse lung tissues or cell lines. For binding experiments, 2 to 10 μg of nuclear proteins were used. Sp1 antibody (Sc-59×) and Sp3 antibody (Sc-644×; Santa Cruz Biotechnology) were used for the supershift experiments. Oligonucleotides were purchased from MWG Biotech and summarized in Supplementary Table S2. EMSA probes were generated by annealing complementary oligonucleotides using standard procedures. For Western blot analysis, total (100 μg) or nuclear (30 μg) protein extracts isolated from mouse lung tissues or cell lines were used. Sp1 and Sp3 antibodies were described above, whereas Cadm1 antibody was purchased from Abnova (Igsf4, H00023705-M01, clone 2A5).

Mithramycin A treatment. Cells (1 × 10⁶) were seeded in T75 cell culture flasks containing 5 mL of DMEM with 10% FCS, 2× l-glutamine, and 2× penicillin/streptomycin for 5 d (∼90% confluence), treated with 500 ng/mL mithramycin A for 24 h. Cells were pelleted for isolation of nuclear extracts and RNA.

Characterization of lung cancer cell lines from c-Myc and c-Raf double-transgenic mice. To enable mechanistic studies, we isolated spontaneously transformed tumor cells from mice double-transgenic for c-Myc and c-Raf (see Fig. 1A and B). After 4 to 7 days of culture, cells from the transgenic mice exhibited the typical cuboidal phenotype (Fig. 1B). By day 10, some cells began to attain different morphologic characteristics (Fig. 1B,), indicating individual phenotypes of tumor cells. From single cuboidal cells, 10 cell lines could be established. Double-transgenicity for c-Raf and c-Myc in the cell lines were confirmed by PCR detection of the transgenes (Fig. 1C). The 10 cell lines no longer expressed the mRNA for surfactant proteins (data not shown), which tend to disappear in culture by day 7 (23). Nonetheless, transcripts of Vim and Eno2, which are markers of epithelial tumors, were strongly expressed (Fig. 1D).

Because genetic alterations are known to occur in cancer cells, we determined by sequencing the presence of any hotspots mutations in Kras and Trp53 two genes implicated in lung cancer, particularly in smokers (24). We did not detect sequence alterations in 6 lung tumors and 10 cell lines from the c-Myc and c-Raf double-transgenic mice. Furthermore, sequencing of the entire Cadm1 gene amplified from these lung tumors and cell lines did not reveal any mutations.

Cadm1 is repressed in lung tumors and cancer cell lines from c-Myc and c-Raf double-transgenic mice. We then examined the gene expression of Cadm1 in three normal lungs from nontransgenic mice, as well as 6 lung tumors and 10 cancer cell lines from c-Myc and c-Raf double-transgenic mice. Cadm1 expression was markedly reduced in tumors and cell lines compared with the three normal lungs from nontransgenic mice (Fig. 2A). In 7 of 10 cell lines, Cadm1 expression was basically absent. This loss of Cadm1 gene expression correlated with corresponding protein expression, which was also reduced in lung tumors and absent in all cancer cell lines except one (Fig. 2B). For the cell line B3, which displayed the highest Cadm1 expression among the cell lines, a faint protein expression was observed (data not shown). The Cadm1 repression pointed to a possible epigenetic regulation, especially because two cell lines responded to treatment with the demethylating agent 5-aza-dC, thereby restoring Cadm1 expression (Fig. 2C).

For comparison with Cadm1, we also investigated the gene expression of Dal1 (differentially expressed in adenocarcinoma of the lung), a gene encoding an actin-binding protein and with which Cadm1 directly associates (25). Dal1 expression was reduced in the tumors than in controls, but much elevated in the cell lines (Fig. 2D). This relationship remained essentially true even when the cell lines were cultured at 50% confluency. Nevertheless, Dal1 expression seemed to be less at 50%, especially with βD10, which showed the lowest among the cell lines (data not shown). Indeed, whereas Cadm1 expression decreased to nonexistent in the cell lines, Dal1 expression remarkably increased (Fig. 2A versus Fig. 2D). Interestingly though, in the two cell lines where Cadm1 was restored after 5-aza-dC treatment, Dal1 increased correspondingly (Fig. 2C).

Aberrant hypermethylation in the promoter region of Cadm1 correlates with loss of gene expression. The absence of Cadm1 mutations in the lung tumors and cell lines and restoration of the Cadm1 expression through 5-aza-dC treatment prompted us to investigate the epigenetic silencing of this gene through CpG methylation of the promoter region. In the past, CpG methylation of the Cadm1 promoter has been investigated in rat and human genomes, but not in mouse. Initially, we annotated the promoter region of mouse Cadm1 and used the 2,000 nt (−2,000 bp) upstream of the translation start site ATG of Cadm1 to identify CpG islands in this region. We used several programs to predict the CpG island (see Materials and Methods); essentially, the same results were obtained. For instance, a CpG island was predicted starting −808 ATG (i.e., length, 809 bp; %GC, 60; ObsCpG/ExpCpG, 0.846). In human CADM1, the predicted TATA Box is located −374 to −371 from ATG (see ref. 17) and six CpGs (−458, −440, −433, −397, −394, −381) are routinely analyzed for hypermethylation in this gene. Using this information from human CADM1, we performed alignment using human and mouse sequences to determine homologous regions (Fig. 3A). There is very high similarity between human and mouse sequences, especially within the 200 nt before translation start site.

Most techniques used in identifying methylated CpGs are based on bisulfite treatment of genomic DNA, which converts cytosine to uracil, but methylated cytosines remain unaltered in this process. During PCR amplification, uracil will be converted to thymidine, which can then be detected by sequencing. PCR is performed using a pair of primers not containing CpGs, thereby amplifying both methylated and unmethylated fragments (bisulfite sequencing) or a pair of primers that specifically amplify only methylated fragments (MSP-PCR). To search for methylated CpGs associated with loss of Cadm1 expression, especially those which may abrogate binding of transcription factors, we developed five assays covering a total of 69 CpGs, including five CpGs in the coding region (Fig. 3B).

Assay 1 (6 CpGs, −944 to −837) and assay 2 (10 CpGs, −682 to −531) are based on bisulfite sequencing, whereas assay 3 (14 CpGs, −456 to −341), assay 4 (27 CpGs, −396 to −180), and assay 5 (37 CpGs, −302 to +41) are based on MSP-PCR and sequencing. Due to high frequency of CpGs, no PCR primers for bisulfite sequencing could be obtained to investigate methylation status nearer the translation start site. The three MSP-PCR assays overlap each other. After bisulfite treatment and direct sequencing of the PCR product, a CpG site was scored methylated if the C (methylated) allele was ∼20% of the T (not methylated) allele. But for assay 4, PCR products were cloned before sequencing due to high CG content, causing difficulties in obtaining clear direct sequencing electropherograms. Examples and results of the five different assays are shown in Fig. 4A–C and summarized in Supplementary Table S3.

Overall, we observed methylation in 69 CpGs, including five CpGs after the translation start site (see Fig. 4A and C). Methylation pattern in seven cell lines lacking or with minimal Cadm1 gene expression was denser than those lines with gene expression (Fig. 4C). Furthermore, after cloning and sequencing of clones, differential methylation pattern was observed between and within cell lines (Fig. 4B and C). The degree of methylation as determined by methylation index (MI), which is the number of methylated CpGs over total CpGs, correlated highly with the loss of gene expression (Fig. 5). Regardless of assay, this correlation was clearly observed in cell lines than in tumors, which may be caused by a mixture of normal and tumor cells.

Methylated CpGs within the core binding sequence of transcription factors lead to abrogation of binding. Methylation of CpGs in the promoter region of genes can lead to abrogation of binding of transcription factors. Using Transfac,⁷

we analyzed for putative binding sites of transcription factors in the promoter region of Cadm1, in particular those containing CpGs within the predicted core sequence. Such analysis revealed three Sp1, an Sp3, and a zinc finger 5 (Zf5) binding sites (Fig. 2B and Supplementary Fig. S1). The Sp1 binding sites would affect CpGs −224 (Sp1-3), −211 (Sp1-2), and −164 (Sp1-1), whereas Sp3 that of CpG −437. The Zf5 binding site would involve CpGs −192, −190, and −188. Methylation of these CpGs sites was observed especially in cell lines with markedly reduced Cadm1 expression. For Sp1 site, CpG −164 (Sp1-1) was more heavily methylated than CpGs −224 (Sp1-3) and −211 (Sp1-2; Supplementary Fig. S1).

We designed oligonucleotide probes and carried out EMSA on Sp1 using nuclear extracts from the cell line A2C12, in which Cadm1 was not expressed. The EMSA probes consisted of the following: wild type, a probe containing two mutated nucleotides in the core sequence, a probe with methylated CpG in the core sequence, and a probe with two methylated CpGs outside the core sequence (see Supplementary Table S2). Compared with the wild-type probe, binding was dramatically reduced in the mutated probe, as well as in the probe with methylated CpG in the core sequence (Fig. 6A). Binding was not abolished in the probe where methylation occurred outside the core sequence. Similar results were obtained from nuclear extracts isolated from normal lung tissues, as well as from the cell line A2C12 that was treated with 5-aza-dC which were used as controls (data not shown). The Sp1 antibody could supershift the bound protein and detect a protein of an expected size of 106 kDa in the nuclear extracts of the cell line A2C12 after Western blot analysis (Fig. 6A).

Similar experiments were conducted on Sp3, using three oligonucleotide probes consisting of a wild type, a probe containing two mutated nucleotides in the core sequence, and a probe in which the CpG in the core sequence was methylated (see Supplementary Table S2; Fig. 6B). Binding was observed in the wild-type probe, and bound protein was supershifted with the Sp3 antibody. In the mutated probe, no binding was observed, whereas in the methylated probe, binding was observed, but the bound protein was not supershifted by the Sp3 antibody. After Western blot analysis, the Sp3 antibody detected a protein of an expected size of 100 kDa in the nuclear extracts of the cell line A2C12 (Fig. 6B).

It is of considerable importance that mithramycin A, which is an inhibitor of Sp1 or Sp3 binding, could block the expression of genes regulated by these transcription factors (26). On the other hand, mithramycin A may also function as a demethylating agent on lung cancer cells (27). To determine such effects on the regulation of Cadm1, we treated a cell line (B3) that still expressed Cadm1, as well as a cell line (A2C12) with no Cadm1 expression, both with 500 ng/mL mithramycin A for 24 h (Fig. 6C). We observed a decrease of Cadm1 gene expression in the cell line B3. In contrast, there was no restoration of gene expression in A2C12.

To determine methylation of the putative binding site of Zf5 in the promoter region of Cadm1, we used three oligonucleotide probes for EMSA as follows: wild type, a probe with all three CpGs in the core sequence that were methylated, and a probe in which the nucleotide C in the three CpGs of the core sequence were mutated to T (see Supplementary Table S2; Fig. 6D). With the wild-type probe, binding was observed with 10 μg of nuclear extracts whether from normal lung tissues or from the cell line A2C12, treated or untreated with 5-aza-dC. This binding was not observed in the methylated probe. In the mutated probe, binding was also observed, but this was different from the wild-type and the methylated probe. Because no Zf5 antibody is available commercially, supershift experiments and Western blot analysis could not be undertaken.

Recent developments in the field of “epigenetics,” i.e., heritable changes in gene expression without accompanying alterations in the DNA sequence, show a much wider contribution of epigenetic alterations in the development of cancer (see reviews; refs. 19, 20, 28). Epigenetic changes, which include DNA methylation, histone modifications, and noncoding RNAs, can disrupt the normal stem-cell or progenitor-cell program leading to disease (19, 29). This epigenetic progenitor model of cancer puts forward that cancer first arises as an epigenetic alteration of stem/progenitor cells within a given tissue, which is mediated by aberrant regulation of tumor progenitor genes. Furthermore, tumor cell heterogeneity is due in part to epigenetic variation in progenitor cells, and epigenetic plasticity together with genetic lesions drives tumor progression. Thus, in line with this view, epigenetically disrupted stem/progenitor cells might be a crucial target for cancer risk assessment and chemoprevention.

We thus investigated the epigenetic control of Cadm1 in 10 different cancer progenitor cells of transgenic c-Myc and c-Raf mouse lung tumors. These cell lines were established from single, spontaneously transformed cells. We also analyzed six lung tumors from c-Myc and c-Raf double-transgenic mice, as well as three normal lungs from nontransgenic mice as controls. As compared with the normal lungs, Cadm1 gene expression was markedly reduced or absent in lung tumors and cell lines, and this result corresponded with loss of protein expression. Treatment with the demethylating agent 5-aza-dC restored Cadm1 expression in two cell lines, suggesting epigenetic regulation. Using five different assays, we investigated the methylation status of 69 CpGs, including five CpGs after the translation start site. Overall, methylation pattern in seven cell lines lacking or with minimal Cadm1 gene expression was denser than those lines with gene expression (see Fig. 4). Furthermore, after cloning and sequencing of clones, differential methylation pattern was observed between and within cell lines. The degree of methylation correlated highly with the loss of gene expression, and this correlation was clearly observed in the cell lines regardless of assay used. Thus, we show transcriptional repression and promoter hypermethylation of Cadm1 in lung cancer. Investigations into the different cancer progenitor cells of transgenic c-Myc and c-Raf mouse lung tumors bring mechanistic insights into the epigenetic control of Cadm1.

Promoter hypermethylation of CADM1 (TSLC1, IGSF4) that is associated with loss of gene expression is known in various cancer in humans, including NSCLC (10, 17, 18). Analysis of methylation of CADM1 promoter region in primary NSCLCs showed methylation in 21 of 48 tumors (17). Furthermore, investigation of 103 primary NSCLCs for association of CADM1 methylation with tobacco smoking and with the clinical characteristics of tumors found methylation in 45% cases (18). Methylation was observed in all histologic subtypes of NSCLC, including adenocarcinoma, squamous cell carcinoma, adenosquamous carcinoma, and large cell carcinoma. In rats, the expression of the Cadm1 gene and its methylation pattern in lung adenocarcinomas induced by N-nitrosobis (2-hydroxypropyl) amine (BHP) has been studied as well (30). Cadm1 was highly methylated in four lung adenocarcinomas, but unmethylated in two normal lung tissues, suggesting aberrant Cadm1 methylation may be involved in BHP-induced development of lung adenocarcinomas in rats.

Results from methylation studies may be influenced by the number and location of analyzed CpGs sites, as well as the assay used. To obtain a comprehensive map of methylation in the promoter region of Cadm1, we used five assays, in which four could be directly sequenced giving an average level of methylation of a CpG site in the mixture of amplified fragments. Bisulfite sequencing, when feasible, is the most reliable method. Amplification with methylation-specific primers, which relies heavily on primer design, may lead to false negatives or positives and could be biased unless followed by sequencing (see Supplementary Table S3). This can be exemplified in assay 5 to interrogate 37 CpGs (−302 to +41). This assay was carried out using a pair of methylation-specific primers, in which three CpGs (−302, −298, −296) are contained in the forward primer and two CpGs in the reverse primer (+38, +41). We were able to amplify fragments in all samples, whether they are normal lungs, tumors, or cell lines. But after sequencing, in contrast to the cell lines, no other methylated CpGs were observed in the normal lungs and tumors, except those CpGs contained in the primers (see Supplementary Fig. S1). The degree of methylation obtained with this assay for CpGs −302 to +41 seems to correlate highly with the loss of Cadm1 expression (see Fig. 4C). Similar finding was obtained in the homologous region in human CADM1. After methylation mapping of CADM1 in human cervical cancer, a 180-bp unmethylated core immediately upstream of the ATG site was found to correlate most strongly with the active expression of this gene (31).

CADM1 directly associates with DAL1, a gene encoding an actin-binding protein likewise implicated in the suppression of growth in lung cancer cells. DAL1 anchors CADM1 to the actin cytoskeleton; thus, loss of DAL1 expression could lead to decreased cell adhesion (25). Inactivation of DAL1 due to promoter hypermethylation has been observed in human lung cancer as well (32–34). We, therefore, also determined Dal1 expression in our study samples. We found that Dal1 expression was reduced in c-Myc and c-Raf mouse lung tumors than in normal lungs, but surprisingly, much more elevated in the cell lines. Indeed, whereas Cadm1 expression decreased to nonexistent in the cell lines, Dal1 expression remarkably increased. A previous study generated mice deficient for Dal1 gene to elucidate its function in development and tumorigenesis (35). Dal1 null mice developed normally and were fertile. Rates of cellular proliferation and apoptosis in brain, mammary, and lung tissues from the 4.1B/Dal-1 null mice were indistinguishable from those seen with wild-type mice. Aging studies indicate that these mice did not have a propensity to develop tumors. These findings indicate that the Dal1 gene is not required for normal development and that Dal1 may not function as a tumor suppressor gene.

DNA methylation in the promoter of certain genes is associated with transcriptional silencing. Such methylation can affect gene expression by interfering with transcription factor binding or by recruiting histone deacetylases through methyl-DNA–binding proteins. Indeed, many of these transcription factors bind to sequences containing CpGs, and binding can be hindered by methylation. Our analysis of potential binding sites of transcription factors in the promoter region of Cadm1 revealed those of Sp1, Sp3, and Zf5 in which CpGs are present in the core site. The transcription factors Sp1 and Sp3 are ubiquitously expressed and bind to GC-rich promoter region. Several studies show methylation-dependent disruption of Sp1 or Sp3 binding leading to reduced expression of targeted gene (36–38). This disruption may even involve methylated CpGs outside the consensus binding site, such as in p21(Cip1) promoter (36). In p21(Cip1) promoter, by using an electrophoretic mobility shift assay, methylation within the consensus Sp1-binding site did not reduce Sp1/Sp3 binding, but methylation outside of the consensus Sp1 element induced a significant decrease in Sp1/Sp3 binding.

We determined whether methylation of potential Sp1 or Sp3 binding sites in the promoter region of Cadm1 could affect binding. Our results on Sp1 showed that binding was dramatically reduced in the mutated EMSA probe, as well as in the probe with methylated CpG in the core sequence compared with the wild-type probe. However, binding was not abolished in the probe wherein methylation occurred outside the core sequence. With Sp3, mutation in the core consensus site led to no binding, but in the methylated probe, binding was observed, but the bound protein was no longer supershifted by the Sp3 antibody. These findings suggest that Sp1 or Sp3 may be involved in the transcriptional regulation of Cadm1 in mouse and that promoter hypermethylation can lead to abrogation of binding. Furthermore, we analyzed the effects of mithramycin A, a drug that modifies GC-rich regions of the DNA and blocks Sp1 or Sp3 binding on the expression of Cadm1. Our treatment of a cell line (B3) that still expressed Cadm1 with 500 ng/mL mithramycin A for 24 h led to a decrease of gene expression, thus further documenting the importance of Sp1 and Sp3 in causing transcriptional activation. Further studies are required to determine quantitatively the role of Sp1 and Sp3 in the transcriptional regulation of Cadm1.

Additionally, Zf5 is a ubiquitously expressed protein that contains five COOH terminal zinc fingers and a conserved NH₂ terminal ZiN/POZ domain. Originally cloned as a transcriptional repressor from the mouse c-Myc promoter, Zf5 can act as an activator, as well as a repressor of transcription (39, 40). The POZ domain contributes to the repressor activity, whereas the active function results from the DNA-binding ability of the zinc finger domain. We also investigated whether methylation of the putative binding site of Zf5 in the promoter region of Cadm1 could affect binding. Compared with the wild-type probe, binding was not observed in the methylated probe. This suggests a possible role of Zf5 in the regulation of Cadm1, but further studies on its activation are necessary, although a previous report suggested a direct relationship between intact Sp1 sites and Zf5 in promoter activation (39). Indeed, activation of the HIV-1 long terminal repeat by Zf5 depends on Zf5 binding to DNA, three intact Sp1 sites, and the Zf5 ZiN/POZ domain.

In conclusion, we investigated the consequences of c-Myc and c-Raf overexpression on the epigenetic regulation of the tumor suppressor gene Cadm1. Indeed, tumor progenitor cells isolated from lung tumors can be used in determining the molecular basis for response and resistance to drug therapies or to even predict which patients are likely to respond to treatment. Unlike studies with DNA isolated from entire tumors, where results are rather difficult to interpret due to contamination of normal cells, the use of spontaneously transformed cancer cell lineages enable mechanistic investigations of Cadm1 gene expression. Notably, the cell lines exhibited heterogeneity of DNA methylation pattern resulting in differential repression of Cadm1 expression and response to 5-aza-dC, a demethylating drug in clinical development for the treatment of leukemia.

A further point of consideration is the plasticity seen in the lung tumor cell lines. Upon treatment with 5-aza-dC and despite repeated treatments at different dose levels (e.g., 10 μmol/L) and prolonged treatment regiments, Cadm1 expression was restored in two cell lines only, whereas the remaining eight lines from the same tumors did not respond to this treatment. This shows epigenetic plasticity in response to treatment with 5-aza-dC. Knowledge of the precise epigenetic control of tumor suppressor genes is crucial for their controlled reactivation in lung cancer cells by epigenetic drugs, which represents a new class of therapeutic strategy to hinder tumor growth and subsequent progression.

No potential conflicts of interest were disclosed.

Grant support: Ministry of Science and Culture, Lower Saxony, Germany grant 25A.5-7251-99-3/00.

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.

We thank Annika Roskowetz, Wiebke Goettmann, Antje Schulmeyer, Stefanie Marschke, and Andreas Hiemisch for the technical support; Jessica Decke for her contribution in the establishment of the cell lines; Roman Halter for the mouse lung tissues; and Susanne Reymann for support in Transfac analysis.