Role of DNA 5-Methylcytosine Transferase in Cell Transformation by fos

See allHide authors and affiliations

Science  15 Jan 1999:
Vol. 283, Issue 5400, pp. 387-390
DOI: 10.1126/science.283.5400.387


The Fos and Jun oncoproteins form dimeric complexes that stimulate transcription of genes containing activator protein–1 regulatory elements. We found, by representational difference analysis, that expression of DNA 5-methylcytosine transferase (dnmt1) infos-transformed cells is three times the expression in normal fibroblasts and that fos-transformed cells contain about 20 percent more 5-methylcytosine than normal fibroblasts. Transfection of the gene encoding Dnmt1 induced morphological transformation, whereas inhibition of dnmt1 expression or activity resulted in reversion of fos transformation. Inhibition of histone deacetylase, which associates with methylated DNA, also caused reversion. These results suggest that fosmay transform cells through alterations in DNA methylation and in histone deacetylation.

The fosproto-oncogene (c-fos) is the cellular homolog of the oncogene that is carried by the Finkel-Biskis-Jinkins (FBJ) murine sarcoma virus (1). Its protein product, Fos, is a nuclear phosphoprotein that forms heterodimeric complexes with Jun and activating transcription factor/cyclic adenosine 3′,5′-monophosphate (cAMP) response element–binding protein family members and regulates transcription through activator protein–1 (AP1) and cAMP-responsive elements (1). Fos can be induced rapidly and transiently in many cells by diverse extracellular stimuli (2). It is believed to function in a transcription factor network that couples extracellular stimuli to alterations in gene expression. Gene disruption studies have shown that Fos is required for differentiation of osteoclasts, for oncogenic conversion of phorbol ester–induced skin tumors, and for light-induced death of photoreceptor cells in the eye (3). It is likely that the many heterodimeric complexes containing Fos regulate distinct target genes in different physiological contexts. Continuous expression offos results in the morphological transformation of cultured rodent fibroblasts in vitro and the induction of bone tumors in mice (1, 4). Cell transformation requires the leucine-zipper and the DNA binding regions of fos (5). In transformed rat fibroblasts, the major partner of Fos is c-Jun, which was first described as the Fos-binding protein, p39 (6). Using a conditional expression system based on LacI, we demonstrated that a continuous period (∼72 hours) of c-fos expression is required for complete morphological conversion, which occurs even in quiescent cells (7). Like other oncogenic transcription factors, fos is thought to transform cells by inappropriate regulation of gene expression (7, 8).

To identify target genes of fos that are effectors of cell transformation, we used representational difference analysis (RDA) (9) to isolate mRNA species, which are present infos-transformed cells at levels that are higher than those present in normal fibroblasts. cDNA was prepared from LacIc-fos cells, which contain a fos gene regulated by the LacI activator protein [comprising LacI, a nuclear localization signal, and the VP16 activator (7)], in the presence (normal phenotype) and absence (transformed phenotype) of isopropyl-β-d-thiogalactopyranoside (IPTG). In these cells, fos is rapidly repressed in the presence of IPTG and is rapidly induced in its absence. Although several fostarget genes are induced within minutes of fos expression in these cells, transformation requires 2 to 3 days. The cDNA populations obtained from transformed and normal cells were used in RDA as tester and driver, respectively (10).

The difference products that were obtained after three rounds were analyzed by DNA sequencing and included the cathepsin L and aquaporin-1 genes (Fig. 1A) that were previously shown to be regulated by Fos or to contain AP1 regulatory elements (11, 12). In transformed cells, the candidate target genes were expressed at levels that were 2 to 8 times the levels in control cells. The RDA products also included several novel sequences and four genes: dnmt1; breast cancer candidate gene-1 (bcsc-1) (this gene is no longer thought to be a strong breast cancer candidate gene); ribosomal protein S3 (rps3); and the gene encoding pyruvate kinase, which has not been previously shown to be regulated by fos. We examined the expression levels of dnmt1, bcsc-1, and cathepsin L by Northern (RNA) blot analysis of several independently derived fos-transformed cell lines and determined the time course of expression during transformation and reversion of LacIc-fos cells (Fig. 1A). Normal rat fibroblasts express c-fos at a low basal level, but this level was increased ∼25-fold after transformation by a c-fos gene under the control of the cytomegalovirus (CMV) immediate early gene CMV promoter (CMVc-fos) (Fig. 1A). The v-fos gene in the FBJ/R construct is a highly transforming mutant (13) that is expressed at relatively low levels in transformed cells. In LacIc-fos cells, fos was expressed in the presence of IPTG at a level similar to that in 208F cells, but this level rapidly increased ∼18-fold in the absence of IPTG (Fig. 1A). The expression patterns of the selected RDA clones paralleled the expression pattern of fos (Fig. 1A). No consistent fluctuations were observed in expression of control genes (β-actin, mitochondrial 16S rRNA, and 28S and 18S rRNA). Several other genes (including those encoding aquaporin-1, pyruvate kinase, rps3, annexin IV, andrad21) were expressed preferentially infos-transformed cells. In contrast, other previously identified targets of fos [including the genes encoding ezrin, tropomyosin-1, and vascular endothelial growth factor–D (8)] were not differentially expressed in ourfos-transformed cells.

Figure 1

Analysis of genes differentially expressed in fos-transformed cells. (A) For Northern analysis, total RNA (10 μg per lane) was prepared from normal 208F cells (lane 1), CMVc-fos cells (lane 2), FBJ/R-transformed cells (lane 3), and LacIc-fos cells (lanes 4 through 11). The LacIc-fos cells were maintained in the presence of 5 mM IPTG for 8 (lane 5), 24 (lane 6), 48 (lane 7), and 72 (lane 8) hours and were maintained in the absence of IPTG for 4 (lane 9), 24 (lane 10), 48 (lane 11) and 72 (lane 4) hours. Blots were hybridized with probes specific for c-fos (1164-bp fragment of rat cDNA), cathepsin L [1100-bp polymerase chain reaction (PCR) fragment of EST clone (GenBank accession number AA617608)], bcsc-1 (219-bp PCR fragment), and dnmt1 (2138-bp fragment of mousednmt1). Probes for mitochondrial 16S rRNA (mt-16S rRNA) and β-actin were used as controls. (B) PhosphorImager quantitation of dnmt1 mRNA levels in normal and fos-transformed cells. The level ofdnmt1 expression was normalized to the level of β-actin. The data represent the average of three independent experiments. Error bars indicate ±SD. (C) Northern blot analysis ofdnmt1 expression in fos-transformed cells that were maintained in 0.5% serum. Total RNA (10 μg per lane) was prepared from LacIc-fos cells that were cultured in the presence of 5mM IPTG for 72 hours in 0.5% serum (lane 1); from LacIc-fos cells that were cultured in the absence of IPTG for 24 (lane 2), 48 (lane 3), and 72 (lane 4) hours; and from normal 208F cells in 0.5% serum (lane 5). The filter was hybridized with adnmt1 probe. (D) Analysis of dnmt1 and cathepsin L transcription in normal and fos-transformed cells by nuclear run-on. cDNA fragments (1 μg per slot, except for cDNA of 18S rRNA, which represented 0.2 μg) were hybridized with RNA probes prepared by transcription in nuclei from LacI-fos cells that were maintained for 72 hours in the absence (–IPTG) or presence of IPTG (+IPTG).

In principle, genes that contribute to fos transformation should be (i) inducible by fos, (ii) present in allfos-transformed cells, (iii) able to encode stable products, (iv) required for fos transformation, and (v) capable of inducing transformation in the absence of fos. Of the nine genes that we identified by RDA, dnmt1 was the best candidate to match these criteria. A 190–base pair (bp) fragment of the dnmt1 gene, corresponding to regions 2108 through 2297 of the mouse dnmt1 gene (GenBank accession number AF036009), was present in a composite RDA clone that also included the β-actin and bcsc-1 genes.

Dnmt1, which catalyzes the transfer of methyl groups fromS-adenosyl methionine to the C-5 position of cytosines in DNA, has been previously implicated in cancer; it is overexpressed in tumor cells, and it can induce transformation when introduced into fibroblasts (14). High levels of DNA methylation are associated with gene silencing, and progressive methylation of several genes has been observed in cells that overexpress dnmt1(15).

The 5.5-kb dnmt1 mRNA was present in CMVc-fos– and FBJ/R-transformed cells at levels that were 2 to 3 times as high as the levels in parental 208F fibroblasts (Fig. 1, A and B). Induction of c-fos expression by removal of IPTG increaseddnmt1 mRNA levels by ∼1.5-fold within 4 hours and by about threefold by 48 hours (Fig. 1, A and B). Abrogation of c-fosexpression by the addition of IPTG decreased the level ofdnmt1 mRNA. Although the increase in dnmt1expression was modest, a small increase in the steady state level ofdnmt1 can transform NIH3T3 cells (14), and there is only a threefold elevation in dnmt1 levels inras-transformed adrenocortical Y1 cells (16). The increase in dnmt1 was not simply due to the proliferative state of fos-transformed cells, because the same increase occurred during the morphological conversion of serum-deprived fibroblasts by LacIc-fos (Fig. 1C).

The increase in dnmt1 expression appears to be a consequence of an approximately threefold increase in its transcription rate as determined by nuclear run-on analyses (Fig. 1D). These data are consistent with the possibility that Fos regulates dnmt1transcription directly. In support of this hypothesis, a dominant-negative c-jun gene has been shown to reducednmt1 expression in ras-transformed cells (16), and AP1 sites have been identified in a region of the dnmt1 gene that activates transcription in aras- and jun-dependent manner (16,17).

Dnmt1 protein levels were elevated ∼10-fold in three independent fos-transformed cell lines (CMVc-fos, FBJ/R-4, and LacIc-fos cells) (Fig. 2A). In LacIc-fos cells that were maintained in the presence of IPTG for 3 days, Dnmt1 protein levels were substantially reduced, although they remained about twofold higher than those in normal fibroblasts (Fig. 2A). In contrast, no changes were observed in the levels of Ref-1, a redox/DNA repair enzyme (18). Protein extracts from fos-transformed cells stimulated a twofold greater incorporation of [3H]-methyl into poly(dI-dC:dI-dC) in comparison with normal cell extracts in vitro (Fig. 2B). Furthermore, DNA from fos-transformed cells contained ∼20% more 5-methylcytosine than DNA from normal cells (Table 1). The 5-methylcytosine content of LacIc-fos cells increased 24 hours after the withdrawal of IPTG and reached a maximum after 48 hours. For comparison, rat cells transfected with a CMVdnmt1 construct contained ∼70% more 5-methylcytosine than normal cells (Table 1). These data are consistent with the range (20 to 100%) of the increased DNA methylation content that was reported for NIH3T3 cells and human fibroblasts transformed bydnmt1 (14).

Figure 2

Expression and activity of Dnmt1 infos-transformed cells. (A) Immunoblot analysis of Dnmt1 levels. Proteins (40 μg per lane) from normal andfos-transformed cells were separated by electrophoresis on SDS-polyacrylamide gels, transferred onto filters, and treated with polyclonal antibodies to Dnmt1 or Ref1. Lanes are as follows: 208F cells (lane 1), CMVc-fos cells (lane 2), FBJ/R cells (lane 3), LacIc-fos cells plus IPTG (lane 4), LacIc-foscells with no IPTG (lane 5), CMVdnmt1 cells (lane 6), and CMVc-fos cells expressing antisense dnmt1 (lane 7). (B) Analysis of Dnmt1 activity. Protein extracts (from LacIc-fos cells maintained for 72 hours in the presence or absence of IPTG and from 208F cells transfected with CMVdnmt1) were assayed for Dnmt1 activity by the incorporation of [3H]-methyl into poly(dI-dC:dI-dC) as described (14). The data represent the average of three independent experiments. Error bars indicate ±SD.

Table 1

Effect of fos transformation on the 5-methylcytosine content of DNA. The level of methylation relative to that in normal 208F cells (100%) was determined by high-performance liquid chromatography (HPLC) analyses of genomic DNA digested to nucleosides (26). DNA (10 μg per sample) was digested with deoxyribonuclease I, phosphodiesterase, and alkaline phosphatase, and the products were separated by HPLC. The data represent the average of three independent experiments (±SD), except for data from the 0.5% serum samples, which are from two experiments. Dash, not applicable.

View this table:

Fibroblasts transfected with CMVdnmt1 exhibited a transformed cell morphology that was strikingly similar to that offos-transformed cells (Fig. 3, E and F). To determine whether Dnmt1 is required forfos-mediated transformation, we inhibited its expression infos-transformed cells with a dnmt1 antisense vector, and we blocked its activity with 5-aza-deoxycytidine (5-aza-dC), an inhibitor of dnmt1. Stable cell lines expressing antisense dnmt1 together with CMVc-fosexhibited a flat morphology and a contact inhibition phenotype, which were similar to that expressed in normal fibroblasts (Fig. 3G). Dnmt1 protein levels were substantially reduced in these cells, whereas Fos protein levels remained unchanged (Fig. 2A). Treatment offos-transformed cells with 5-aza-dC for 5 days also resulted in a flatter cell morphology and contact inhibition; however, Fos expression was also reduced (Fig. 3H). Thus, elevated levels of Dnmt1 are required for the maintenance of the transformed cell phenotype that is induced by fos.

Figure 3

Cell transformation byfos requires dnmt1 and histone deacetylation. Photomicrographs show (A) normal 208F rat fibroblasts, (B) LacIc-fos cells maintained in the absence of IPTG for 72 hours, (C) LacIc-fos cells maintained in the presence of IPTG for 72 hours, (D) CMVc-fos cells, (E) FBJ/R cells, (F) CMVdnmt1 cells, (G) CMVc-fos cells transfected with antisense dnmt1, (H) CMVc-fos cells treated with 5-aza-dC for 5 days, (I) CMVdnmt1 cells treated with 80 nM TSA, (J) LacIc-fos in the absence of IPTG plus 80 nM TSA, and (K) FBJ/R cells treated with 120 nM TSA. Scale bar, 80 μm.

DNA methylation causes repression of gene expression by promoting the condensation of chromatin. Methylated sites on DNA bind the 5-methylcytosine binding protein (MeCP2), which exists in a complex with Sin3A and histone deacetylase (HDAC) (19). This complex decreases the levels of histone acetylation, resulting in a compact chromatin structure. To investigate whether fostransformation requires histone deacetylation activity, we used the HDAC inhibitor trichostatin A (TSA) (20). Treatment of bothfos- and CMVdnmt1-transformed cells with TSA for 3 days induced a normal cell morphology (Fig. 3, J and K). These results suggest that fos transformation is mediated by elevated expression of dnmt1, which represses gene expression (possibly by the increased recruitment of HDAC to methylated DNA). TSA and other HDAC inhibitors also cause reversion ofras-transformed cells (21).

Paradoxically, many tumors that overexpress dnmt1 have hypomethylated genomic DNA (15). Furthermore, whereas hypomethylation has been suggested as contributing to colon cancer in humans (22), suppression of dnmt1 reduces the incidence of intestinal polyps in Min mice (23). These apparent contradictions can be resolved by considering the patterns of DNA methylation in tumor cells. Increased expression ofdnmt1 is associated with regional areas of hypermethylation in CpG islands that result in transcriptional repression of nearby genes. This may occur even if the overall 5-methylcytosine content of the cell is reduced (15). Hypomethylation increases the frequency of rearrangement of endogenous retroviral sequences and the incidence of mutations resulting from translocations and deletions (24). Thus, hypomethylation has been implicated as a mutator mechanism in tumor cells, whereas hypermethylation has been linked to an epigenetic mechanism that contributes to tumorigenesis by altered gene regulation. Our results imply that, in addition to activating gene expression directly, Fos can regulate gene expression indirectly through alterations in DNA methylation and histone acetylation. DNA methylation and histone deacetylation are associated with global changes in chromatin structure and the silencing of gene expression (15, 25). The requirement for increased Dnmt1 (and, presumably, increased DNA methylation) explains in part the slow time course of fostransformation and reversion in LacIc-fos cells (7). For transformation to occur, Dnmt1 may need to increase the 5-methylcytosine content of the cell beyond a critical threshold before gene expression is affected.

  • * To whom correspondence should be addressed. E-mail: fos1{at}


View Abstract

Stay Connected to Science

Navigate This Article