A Conserved 3′-Untranslated Element Mediates Growth Regulation of DNA Methyltransferase 1 and Inhibits Its Transforming Activity*

Ectopic expression of DNA methyltransferase 1 (DNMT1) has been proposed to play an important role in cancer.dnmt1 mRNA is undetectable in growth-arrested cells but is induced upon entrance into the S phase of the cell cycle, and until now, the mechanisms responsible for this regulation were unknown. In this report, we demonstrate that the 3′-untranslated region (3′-UTR) of the dnmt1 mRNA can confer a growth-dependent regulation on its own message as well as a heterologous β-globin mRNA. Our results indicate that a 54-nucleotide highly conserved element within the 3′-UTR is necessary and sufficient to mediate this regulation. Cell-free mRNA decay experiments demonstrate that this element increases mRNA turnover rates and does so to a greater extent in the presence of extracts prepared from arrested cells. A specific RNA-protein complex is formed with the 3′-UTR only in growth-arrested cells, and a UV cross-linking analysis revealed a 40-kDa protein (p40), the binding of which is dramatically increased in growth-arrested cells and is inversely correlated with dnmt1 mRNA levels as cells are induced into the cell cycle. Although ectopic expression of humanDNMT1 lacking the 3′-UTR can transform NIH-3T3 cells, inclusion of the 3′-UTR prevents transformation. These results support the hypothesis that deregulated expression of DNMT1with the cell cycle is important for cellular transformation.

DNA methylation is a covalent modification of DNA that can modulate gene expression and is now recognized as a major component of the epigenome (1,2). The methylation of cytosines in vertebrates occurs when they are found 5Ј to deoxyguanosine in the sequence CpG. 80% of CpGs are methylated (3), and they are distributed in a pattern that is unique in each tissue and is inversely correlated with gene expression (3).
The pattern of methylation is faithfully maintained during cell division by the enzyme DNMT1, 1 the maintenance DNA methyltransferase, (4) which catalyzes the transfer of a methyl group from S-adenosyl-methionine to the 5-position of the cytosine ring (5,6). DNMT1 has a bilateral relationship with DNA replication and cell growth (7). DNA methylation occurs concurrently with DNA replication (8), and the expression of dnmt1 is tightly coordinated with the cell cycle (9), whereas inhibition of DNMT1 inhibits DNA replication (10), leading to senescence and inhibition of cell growth (11,12). Deregulated expression of dnmt1 was previously suggested to play a causal role in cellular transformation (13,14). Several lines of evidence are in accordance with this hypothesis. First, elevated levels of dnmt1 mRNA and activity were reported in tumors and cancer cell lines (15)(16)(17). Second, ectopic expression of dnmt1 results in cellular transformation (18,19). Third, dnmt1 expression is regulated by nodal proto-oncogenic signaling pathways (20 -22). And fourth, inhibition of DNMT 1 by 5-azacytidine (a pharmacological inhibitor of DNMT1) or a reduction in DNMT1 levels by either antisense mRNA or antisense oligonucleotides reverses tumorigenesis (23)(24)(25).
The mechanism by which DNMT1 influences cellular transformation is unclear. It has been proposed that high levels of DNMT1 can lead to ectopic methylation and silencing of tumor suppressor genes (26). However, a clear correlation between general dnmt1 overexpression and tumor suppressor hypermethylation has not been established (27,28). It has recently been suggested that deregulated expression of dnmt1 during the cell cycle might be critical for the effects of DNMT1 on cell growth (29,30). Previous nuclear run-on experiments have demonstrated that although both growth-arrested and cycling cells transcribe dnmt1 mRNA at a similar rate, the mRNA is only detected in cycling cells (9). This suggests that the cell cycle regulation of dnmt1 occurs at the posttranscriptional level. It stands to reason that the mechanisms involved are linked to basic control points of the cell cycle, and potentially linked to cellular transformation. To test this hypothesis, the factors responsible for regulation of dnmt1 expression with the cell cycle must be identified. One potential element is the AU-rich 3Ј-untranslated region (3Ј-UTR) of the dnmt1 mRNA, because it is well documented that this type of element can regulate mRNA levels (31).
Regulation of mRNA stability by AU-rich elements (AREs) is an important mechanism involved in orchestrating the expression of critical genes in development (32,33) and cell cycle (34,35), early response genes involved in cellular growth such as c-myc (36) and c-fos (37), and cytokines such as granulocyte macrophage colony-stimulating factor (GM-CSF) (38) and interleukin 3 (39). These regions often modulate mRNA levels in response to a change in cell environment by growth factors (40,41), developmental factors (42), and hormones (43). Although the mechanism by which AREs act has not been fully elucidated, it appears to involve the binding of various protein factors, several of which have been well characterized, such as AUF1 (44,45) and the embryonic lethal abnormal vision family members (46,47).
In this study, we test the hypothesis that the 3Ј-UTR of the dnmt1 mRNA plays a role in regulating its levels with the cell cycle and that deregulation at this level has an effect on cellular transformation by DNMT1. We show that a 54-nucleotide highly conserved element within the 3Ј-UTR can confer a growthdependent regulation on both homologous and heterologous mRNAs in living cells and that this region can destabilize mRNA in vitro. This element interacts with a 40-kDa protein, the binding of which is inversely correlated with dnmt1 mRNA levels throughout the cell cycle. Finally, we demonstrate that the 3Ј-UTR influences the cellular changes observed upon overexpression of DNMT1 in NIH-3T3 cells, thus providing a link between dnmt1 regulation, cell cycle, and oncogenesis.

MATERIALS AND METHODS
Cell Culture and Flow Cytometry-Balb/c-3T3 cells (ATCC) were maintained as a monolayer in Dulbecco's modified Eagle's medium (Life Technologies, Inc.) containing 10% calf serum (Colorado Serum Co). To arrest cells at G 0 /G 1 , confluent Balb/c 3T3 cells were cultured in a medium containing 0.5% calf serum for 48 h. The cells were induced to enter the cell cycle by replacing the growth medium with medium containing 10% calf serum. Cells at various stages of the cell cycle were obtained by harvesting them at varying lengths of time after serum induction. To determine the percentage of cells at different stages of the cell cycle, cells were stained with propidium iodide and the DNA content was measured by flow cytometry (48).
Plasmid Construction-For the construction of pCRII-DNMT1 3Ј-UTR, human DNMT1 3Ј-UTR (5090 -5408) was amplified by reverse transcription-PCR from 1 g of total RNA prepared from human lung cancer H446 cell line (ATCC). Reverse transcription was performed using random primers (Roche Molecular Biochemicals) and Superscript reverse transcriptase as recommended by the manufacturer (Life Technologies, Inc.). For PCR amplification with Taq polymerase (CLON-TECH), 50 ng of the sense primer (5Ј-TCTGCCCTCCCGTCACCC-3Ј) and antisense primer 5Ј-GGTTTATAGGAGAGATTT-3Ј) was used to amplify the 3Ј-UTR from 5 l of reverse-transcribed cDNA in the presence of 1 mM dNTPs, and the manufacturer's amplification buffer was supplemented to 2 mM MgCl 2 . Cycling conditions were as follows: 95°C for 1 min, 50°C for 1 min, and 72°C for 1.5 min for 30 cycles. The amplified fragment was subcloned into pCR II using the TA cloning kit (Invitrogen) as recommended by the manufacturer. To generate pRBG-DNMT1 3Ј-UTR, the human DNMT1 3Ј-UTR was excised from PCRII-DNMT1 3Ј-UTR with BamHI and XbaI and ligated into a XbaI-SalI digested pRBG-GC (38). The incompatible ends were then filled using Klenow DNA polymerase (Roche Molecular Biochemicals) and the blunt ends were joined using T4 DNA ligase (MBI Fermentas).
For the construction of pRBG-DNMT1⌬ 5Ј259 and pSK-⌬5Ј259, sense and antisense oligonucleotides coding for bases 5349 -5405 bearing 5Ј XbaI and 3Ј SalI overhangs were annealed and ligated into the XbaI and SalI sites of both pRBG-GC and pBluescript SK.
For the construction of pUD-hDNMT1 and pUD-hDNMT1⌬UTR, the pUD1 vector (a gift from Dr. L. Chasin) was digested with EcoRI and HindIII (to remove the DHFR gene) and either blunt-ended or modified with the following linkers: 5Ј-AATTCTAG-3Ј and 5Ј-AGCTCTAG-3Ј to generate an XbaI site. Human DNMT1 cDNA (360 -5085) was inserted into the blunt pUD1 to give pUD-hDNMT1⌬UTR, whereas hDNMT1 cDNA (360 -5408) was inserted into the modified pUD1 to give pUD-hDNMT1.
Preparation and Analysis of RNA-Total RNA was prepared by the guanidinium isothiocyanate method (50), and mRNA levels were deter-mined by Northern blot analysis. Approximately 10 g of RNA was electrophoresed on a 1.2% denaturing agarose gel and then transferred to Hybond-Nϩ membrane (Amersham Pharmacia Biotech). Blots were probed with the indicated 32 P-labeled cDNA probes synthesized using a random priming labeling kit (Roche Molecular Biochemicals). The membranes were hybridized at 68°C for 4 -6 h in a buffer containing 0.5 M sodium phosphate, pH 6.8, 1 mM EDTA, 7% SDS, and 0.2 mg⅐ml herring sperm DNA. Following hybridization the membranes were washed twice for 10 min in a 5% SDS, 0.04 M sodium phosphate, pH 6.8, 1 mM EDTA solution and then four times for 10 min in the same solution containing 1% SDS. The following probes were used: rabbit ␤-globin (a 500-bp Xho-Bam fragment from pRBG-GC), hDNMT1 (a 1.5-kilobase pair fragment from nucleotides 601-2191 of the full-length human DNMT1 cDNA), mDNMT1 (a 1.7-kilobase pair fragment from nucleotides 844 -2544 of the full-length mouse dnmt1 cDNA), green fluorescent protein (GFP) (a 765-bp NheI-HindIII fragment from pEGFP-C1 (CLONTECH)).
The level of expression of the different mRNAs was quantified by densitometric scanning of the relevant autoradiogram. Each experiment was normalized for the amount of total RNA by hybridization with a 32 P-labeled 18-s ribosomal RNA oligonucleotide probe (51), with the exception of the adenoviral experiments, which were normalized to GFP, which is expressed by the adenoviral vector.
Stable Transfections-To create the pRBG stable lines, 5 g of the following plasmids were introduced into Balb/c 3T3 cells by DNAmediated gene transfer using the calcium phosphate protocol (52): pRBG-AT (bearing the 3Ј-UTR of human GM-CSF and a neomycin resistance gene as a selectable marker), pRBG-GC (control) (both plasmids were a kind gift of Dr. G Shaw) (38), pRBG-DNMT1 3Ј-UTR, and pRBG-DNMT1⌬ 5Ј259. G418-resistant colonies were cloned and propagated in selective medium containing 0.25 mg/ml Geneticin (Life Technologies, Inc.).
Stable lines expressing pUD-hDNMT1 and pUD-hDNMT1⌬3Ј-UTR (the pUD vector bears a tetracycline-repressible promoter) were obtained by transfecting 6 g of each construct together with 6 g of a puromycin resistance expression vector pBABEpuro (53) into TetOff NIH3T3 cells (CLONTECH) using SuperFect reagent (Qiagen) according to the manufacturer's protocol. Resistant colonies were selected in medium containing 3 g/ml puromycin (Sigma) and were then further cultured in the presence of 1 g/ml doxycycline (Sigma) to repress transcription of the transfected DNMT1. Northern blot analyses were used to verify expression of the ectopic hDNMT1 for both sets of transfectants. Growth arrest was accomplished by culturing in 0.5% serum for 48 h; to induce growth, arrested cells were cultured in 10% serumcontaining medium for 20 h. Doxycycline was removed 48 h before serum starvation to induce ectopic hDNMT1 transcription.
Growth rates were determined following plating the cells in six-well culture dishes at a density of 25,000 cells/well. Cells were then counted on six subsequent occasions throughout the span of 11 days, and phase contrast photography was performed on the last day. An average of four counts was performed for each time point.
Adenoviral Infections-Balb/c cells were grown to confluence in sixwell plates and were then transferred to 0.5% serum medium for 48 h to arrest cell growth. Cells were then infected for 4 h in serum-free medium with pAd-DNMT1, pAd-DNMT1⌬3Ј56, and pAd-DNMT1⌬3Ј-UTR at multiplicities of infection of 25, 50, and 150 respectively. 100% of cells were infected as determined by visualization of the green fluorescent protein, which is expressed by the adenoviral vector. Cells were then maintained in 0.5% medium for an additional 24 h before the medium was replaced with a medium containing either 10 or 0.5% serum. RNA was harvested after 20 h and was analyzed by a Northern blot analysis as described above.
In Vitro RNA Synthesis-To generate RNA transcripts for the gel mobility shift, UV cross-linking, and in vitro degradation assays, the following plasmids were linearized at the indicated restriction sites: pCRII-DNMT1 3Ј-UTR (BamHI), pSK-⌬5Ј259 (SalI), pSK-⌬3Ј56 (SalI), and pSK-⌬3Ј158 (SalI). In vitro transcription was then carried out in the presence of 50 Ci of [␣ 32 P]-UTP (3000Ci/mmol, PerkinElmer Life Sciences) using either T7 polymerase (for DNMT1 3Ј-UTR) or T3 polymerase (for ⌬5Ј259, ⌬3Ј56, and ⌬3Ј158). The Ambion in vitro transcription kit was used as recommended by the manufacturer. Cold competitor RNA transcripts were synthesized using the Promega Ribo-Max system according to the protocol supplied by the manufacturer.
Preparation of Whole Cell Extracts-Balb/c 3T3 cells were harvested and resuspended in extraction buffer (20 mM Tris-HCl, pH 7.5, 0.4 M KCl, 20% (v/v) glycerol, 2 mM dithiothreitol, and 1ϫ Complete TM protease inhibitor (Roche Molecular Biochemicals)). The suspension was frozen immediately at Ϫ80°C for 1 h, and the whole cell extract was isolated by centrifugation at 10000 ϫ g for 15 min. The supernatant was recovered and used for RNA binding assays. Protease K treatment was carried out for 1 h at 37°C at a final concentration of 1 g/l.
RNA Mobility Gel Shift Assays and UV Cross-linking-Reaction mixtures were incubated on ice for 1 h in a 20-l mixture containing 50 g of whole cell extract, 1,000,000 cpm of probe), 10 g of tRNA (as a nonspecific RNA competitor), and the following buffer: 10 mM Hepes, pH 7.6, 3 mM MgCl 2 , 40 mM KCl, 5% glycerol, 1 mM dithiothreitol. Following incubation, free unbound 32 P-labeled RNA was removed by treating the reaction mixture with 1.5 l of RNase T1 (1 unit/ml) and 1.5 l of RNase A (10 mg/ml) for 10 min at room temperature. Heparin 5 g/ml was added, and the reaction was incubated for an additional 10 min at room temperature. The resulting complexes were resolved on a 10% nondenaturing polyacrylamide gel. For UV cross-linking assays, the same reaction mixture was subjected to incubation on ice for 1.5 h under UV light (254 nm). Cold RNA competitors were used at a concentration of 1000ϫ. RNases were added as above, and the reaction mixture was incubated for an additional 30 min at 37°C. 5 l of 5ϫ loading buffer (1.05 M Tris-Cl, pH 6.8, 36% glycerol, 100 mg/ml SDS, 0.2% ␤-mercaptoethanol, 0.12 mg/ml bromphenol blue) were added, and the mixture was boiled for 10 min. Protein-RNA complexes were resolved on a 10% SDS-polyacrylamide gel electrophoresis gel.
In Vitro RNA Degradation Assay-RNA decay rates were assessed using the protocol described in Ref. 42 with the following modifications. Quantification of hDNMT1 mRNA Expression by Competitive Reverse Transcription-PCR-Total RNA (1 g) from TetOff NIH 3T3 stable lines was reverse-transcribed with M-MuL-V reverse transcriptase and a random primer (MBI) using the manufacturer's protocol in the presence of 25 Ci of 35 S-labeled dATP (1250 Ci/mmol) to quantify the efficiency of reverse transcription. Equal amounts of cDNA (30000 cpm as determined by [ 35 S]dATP incorporation) were subjected to DNMT1 PCR amplification using Taq polymerase (MBI) in the presence of decreasing concentrations (10 Ϫ12 -10 Ϫ16 nM) of a competitor DNA fragment. Primers 5Ј-ACCGCTTCTACTTCCTCGAGGCCTA-3Ј (sense, starting at 3479) and 5Ј-GTTGCAGTCCTCTGTGAACACTGTGG-3Ј (antisense, starting at 3813) were used to amplify the target sequence, and 5Ј-CCTCGAGGCCTAGAAACAAAGGGAAGGGCAAG-3Ј (ending at 3576) was used to create the competitor, as previously described (54) PCR conditions were as follows: 96°C for 1 min, 65°C for 1 min, and 72°C for 1 min (33 cycles). To differentiate between the endogenous mdnmt1 and transfected hDNMT1 mRNA, the PCR was transferred to Hybond-Nϩ membrane and hybridized with an oligonucleotide corresponding to bases 3595-3617 of the full-length human DNMT1 mRNA (GenBank TM accession number NM001379.1). The oligo was labeled at the 5Ј end using 20 Ci of [␥-32 P]-ATP (3000 Ci/mmol, PerkinElmer Life Sciences) and T4 polynucleotide kinase (MBI Fermantas) according to the manufacturers' protocols. The membranes were incubated for 16 h at 42°C in the following hybridization buffer: 1% SDS, 5ϫ SSC, 5ϫ Denhardt's solution (0.5 mg/ml Ficoll, 0.5 mg/ml polyvinylpyrrolidone, 0.5 mg/ml bovine serum albumin), 0.1 mg/ml herring sperm DNA, and then washed twice with 6ϫ SSC ϩ 0.1% SDS for 10 min at room temperature, twice for 10 min at 37°C, and twice for 10 min at 42°C.
To further confirm that equal amounts reverse-transcribed cDNA were being used, we amplified ␤-actin using the following primers: 5Ј-GTTGCTATCCAGGCTGTGCTA-3Ј (sense, starting at 473) and 5Ј-GCGGATGTCCACGTCACACTT-3Ј (antisense, starting at 943) (Gen-Bank TM accession number XM_004814). Touchdown PCR was used by decreasing the annealing temperature from 66 to 60°C over four cycles, and then continuing with 96°C for 30 s, 60°C for 30 s, and 72°C for 30 s (21 cycles).

Time Course of dnmt1 mRNA Increase following Serum Stimulation of Arrested Balb/c-3T3 Cells-Previous studies
have demonstrated that DNMT1 is regulated with cell growth (9,29,55,56). To correlate the levels of dnmt1 mRNA with different stages of the cell cycle in our system, Balb/c-3T3 cells were stimulated with serum for varying times following growth arrest, and concurrent FACS and Northern blot analyses were performed. As seen in Fig. 1, A and C, the majority of cells were arrested at G 0 /G 1 (70%) after serum deprivation for 48 h (t ϭ 0). Addition of serum-containing medium induced entry of the cells into the S phase of the cell cycle, which peaked 24 h poststimulation (S ϭ 51%). dnmt1 mRNA was barely detectable in arrested cells and was induced late in G 1 (8 -12 h), reaching its maximum levels at the G 1 -S boundary (20 h) (Fig.  1, B and C). dnmt1 mRNA remained elevated throughout S, and then began to decrease as cells continued into G 2 -M. Because nuclear run-on transcription experiments have previously shown that dnmt1 is transcribed at a constant rate throughout the cell cycle (9), the variation in mRNA levels must occur at the posttranscriptional level. This mode of regulation has already been reported for dnmt1 in several cases, such as in myoblast differentiation (57), F9 cell differentiation (58), and dnmt1 up-regulation by T-antigen (22). The 3Ј-UTR of the dnmt11 mRNA Contains a Highly Conserved AU-rich Element-Because the 3Ј-UTR of an mRNA is often implicated in mediating posttranscriptional regulation, we examined this region of dnmt1 for the presence of any potential mRNA regulating motifs. We first reasoned that an important regulatory element should remain conserved throughout evolution. As can be seen in Fig. 2, an alignment of the 3Ј-UTR of the dnmt1 mRNAs from human, mouse, rat, Xenopus, chicken, and zebrafish (daniorerio) identifies a 54nucleotide region that is 100% conserved between human and chicken, and a smaller region of 36 nucleotides that is conserved among all the species. Also shown are several AU-rich sequences that may act as potential RNA destabilizing motifs (Fig. 2, boxed and underlined) (59). Such a homology in a noncoding region is unusual (less than 30% would be expected without selective pressure (60)) and strongly suggests an important regulatory role for this RNA element. The 54-nucleotide sequence is unique to dnmt1, and no other example was found in the human genome by a BLAST search of the human genome data bank. Many mRNAs that are regulated posttranscriptionally contain highly conserved 3Ј-UTRs, such as c-fos, bcl2, fibronectin (61), B-Ar (62), p21 (35), and IgfII (40). As many of these regions often modulate mRNA levels in response to varying environmental and growth conditions, it is highly probable that this region of the dnmt1 mRNA is involved in its cell cycle regulation.
The DNMT1 3Ј-UTR Can Down-regulate the Stable ␤-Globin mRNA in Growth-arrested Cells-To test the hypothesis that the regulation of dnmt1 mRNA levels involves cis-acting sequences within the 3Ј-UTR, we assayed whether this region could direct growth-dependent regulation of the stable rabbit ␤-globin mRNA. Various AREs, such as those of GM-CSF (38) and type 1 plasminogen activator inhibitor (63), have been shown to destabilize rabbit ␤-globin mRNA when inserted into its 3Ј-UTR. We placed either the full 3Ј-UTR of DNMT1 (pRBG-DNMT1 3Ј-UTR) or only the 54-nucleotide conserved sequence (pRBG-DNMT1⌬ 5Ј259) in the rabbit ␤-globin gene 3Ј-UTR as indicated in Fig. 3A. We used the stable pRBG-GC construct as a negative control and the pRBG-AT construct bearing the instability element from GM-CSF as a positive control (38). These were transfected into Balb/c-3T3 cells, stable transfectants expressing the different constructs were then selected, and RNA was prepared from either growth-arrested cells or from cells that had been serum-stimulated for 20 h (time at which dnmt1 mRNA peaks; see Fig. 1). The level of globin mRNA under each condition was determined by Northern blot analysis and normalized relative to 18-s rRNA (Fig. 3, B and  C). To take into account any possible difference in mRNA levels due to a decreased promoter activity of the RBG vector in arrested versus serum-stimulated cells, the results presented in Fig. 3C were normalized relative to that of the GC stable control. Our findings demonstrate that the DNMT1 3Ј-UTR is able down-regulate the ␤-globin mRNA in arrested cells but not in serum-stimulated cells. The full 3Ј-UTR is not required, however, because the chimeric ␤-globin construct containing the 54-nucleotide element by itself (⌬5Ј259) is also able to perform the same function. As expected, the stable pRBG-GC construct does not show significant variation with the growth state of the cells, whereas pRBG-AT is constitutively destabilized. These data indicate that the conserved 54-nucleotide region within the DNMT1 3Ј-UTR is sufficient to down-regulate a heterologous mRNA in growth-arrested cells.
The 54-Nucleotide Highly Conserved Region of the 3Ј-UTR Is Required to Down-regulate the DNMT1 mRNA in Growth-arrested Cells-We then assessed whether the DNMT1 3Ј-UTR could regulate its own mRNA in a similar manner, by generating adenoviral vectors expressing human DNMT1 containing deletions within the 3Ј-UTR (Fig. 3D). The adenovirus system is advantageous because sufficient levels of transient expression of the exogenous constructs are obtained, thus avoiding the tedious process of generating stable cell lines. In addition, the adenoviral vector contains the gene for GFP, which is under control of the same promoter and poly(A) signal as the gene of interest and can serve as an ideal internal control for infection efficiency, as well as any variations in promoter activity due to changes in growth conditions. Fig. 3E displays a representative Northern blot of RNA from Balb/c cells infected with the different adenoviral constructs. We used a human DNMT1 cDNA probe to differentiate between the exogenous human DNMT1 deletions and the endogenous mouse dnmt1, and under our conditions of hybridization, the endogenous dnmt1 was undetected. Once the results were quantified and normalized to GFP (Fig. 3F), they demonstrated that only the DNMT1 construct bearing the full 3Ј-UTR (DNMT1) was down-regulated in arrested cells. By removing the 3Ј-UTR (⌬3Ј-UTR), or by simply deleting the conserved 54-nucleotide sequence (⌬3Ј56), we abolished this cell growthdependent change in DNMT1 mRNA levels. Taken together, these results and those from the previous experiment indicate that the DNMT1 3Ј-UTR can down-regulate an mRNA in either a heterologous or homologous context in a growth-dependent manner and that the conserved 54-nucleotide sequence is necessary and sufficient for this regulation.
A Specific Protein (p40) Binds the Conserved Region of the DNMT1 3Ј-UTR and Is Up-regulated in Growth-arrested Cells-The 3Ј-UTR can mediate its regulatory effects through interaction with various RNA-binding proteins, such as AUF1  (44,45), and the embryonic lethal abnormal vision proteins (46,47). Many other 3Ј-UTR mRNA binding activities have been found (64 -67). However, most of these have not been well characterized as of yet. In many instances, the expression or binding activity of the proteins themselves is modified in response to various stimuli (46,59) and is correlated with either up-or down-regulation of the associated mRNAs (68,69).
We therefore addressed the question of whether the growth regulation of dnmt1 mRNA levels involves the formation of growth phase-specific RNA-protein complexes. To answer this, we performed an RNA gel mobility shift assay using a 32 Plabeled DNMT1 3Ј-UTR RNA transcript (Fig. 4A, 3Ј-UTR) and whole cell extracts prepared from Balb/c cells that were growth-arrested and then serum-stimulated for specific lengths of time (as in Fig. 1). The RNA-protein complexes were subjected to RNase digestion to remove free unbound RNA prior to resolving of the complex on a nondenaturing polyacrylamide gel. As observed in Fig. 4B, the profile of RNA protein complexes within the 3Ј-UTR changed as passage into the cell cycle occurred. Most strikingly, a specific complex (indicated by arrow) was only observed in extracts prepared from growtharrested cells. The fact that these complexes were formed in the presence of excess of nonspecific tRNA suggests that these interactions are specific.
UV cross-linking analysis was then utilized to further study RNA-protein interactions within the 54-nucleotide element, as this region appears to be critical for dnmt1 mRNA down-regulation. The UV cross-linking method allows one to observe the binding of individual proteins to the RNA, and the molecular mass of the proteins can be approximated. As can be seen in Fig. 4C, the interaction of ⌬5Ј259 with a protein of ϳ40 kDa (which we will from now on refer to as p40) was found to occur almost exclusively in arrested cells. A competition experiment was then carried out, with cold competitors comprising the various 3Ј-UTR constructs (Fig. 4, A and D). As expected, both 3Ј-UTR and ⌬5Ј259 are effective competitors and displace p40 binding from the labeled ⌬5Ј259 probe. However, ⌬3Ј56 and ⌬3Ј158 did not compete, demonstrating that the binding of p40 is specific to the conserved 54-nucleotide element. No binding activity was observed in experiments performed with probe alone or with cell extracts (serum-stimulated and arrested) treated with protease K (Fig. 4C), confirming that the p40 band is indeed a protein.
The above findings, together with the results in Fig. 3, suggest that p40 may be involved in the mRNA down-regulation mediated by the 54-nucleotide element in arrested cells. To provide further support to this theory, we performed a UV cross-linking with extracts prepared at various time points after serum stimulation of arrested cells. A representative UV cross-linking is shown in Fig. 3F, and the average results of three quantified experiments are illustrated in Fig. 3G. We can clearly see that p40 binding was decreased between 0 and 6 h after cell growth is stimulated, and other proteins interacted with the 54-nucleotide element as cells progressed through the cell cycle. When we overlaid the time course of p40 binding with dnmt1 mRNA levels following serum stimulation, we observed an inverse correlation between p40 binding and dnmt1 mRNA levels; dnmt1 mRNA increased shortly after p40 levels fell and remained elevated while p40 binding was low. Although these results do not prove that p40 is responsible for the change in dnmt1 mRNA levels, the facts that p40 binding is highly induced in growth-arrested cells and that it is specific for the Balb/c 3T3 cells were transfected with either one of the following plasmids: pRBG-DNMT1 3Ј-UTR, pRBG-DNMT1⌬ 5Ј259, pRBG-AT, or pRBG-GC, and stable G418 resistant transfectants were isolated as described under "Materials and Methods." RNA was extracted from growth-arrested cells (-) or from arrested cells that had been incubated in serum for 20 h (ϩ). RNA was subjected to a Northern blot analysis using standard protocols and hybridization to a labeled ␤-globin probe (500-bp Xho-Bam fragment from pRBG-GC). Blots were then stripped and rehybridized with an 18-s oligonucleotide probe. C, quantification of ␤-globin mRNA expressed in the stable transfectants. The amount of globin mRNA expressed in each of the transfectants was quantified by scanning densitometry and normalized to the signal obtained following hybridization with an 18-s probe. All ␤-globin levels were then calculated as a percentage of the GC control. The experiment was repeated with similar results. D, map of the adenoviral constructs. DNMT1 deletion constructs shown were cloned into the XbaI site of pAdTrack-CMV. E, representative Northern blot of infected Balb/c cells. Arrested cells were infected with the indicated adenoviral constructs and were then either maintained in serum-deprived conditions (arrested), or stimulated with 10% serum for 20 h (ϩserum). RNA was extracted and subjected to a Northern blot analysis by hybridization with a labeled hDNMT1 cDNA probe (bp 601-2191). Blots were then stripped and rehybridized with a probe for GFP (765-bp NheI-HindIII from pEGFP-C1). F, quantification of hDNMT1 mRNA levels in the infected cells. The amount of hDNMT1 mRNA expressed in each infection was quantified and normalized to the signal obtained following hybridization with the GFP probe. The graph shows an average of three independent experiments. same region that is required for mRNA down-regulation are consistent with this hypothesis.
The Rate of mRNA Decay in Vitro Is Accelerated in the Presence of the 54-Nucleotide Region and Arrested Cell Extracts-Because mRNA posttranscriptional regulation can occur at multiple levels, we addressed the question of how the DNMT1 3Ј-UTR down-regulates mRNA levels in growth-arrested cells. We decided to investigate the role of the DNMT1 3Ј-UTR in mRNA decay for two reasons. First, the 3Ј-UTR contains several AREs, which have been shown to modulate the stability of a wide variety of mRNAs, and second, alterations in mRNA half-life involving AREs have been found to involve changes in RNA-protein interactions.
We used an in vitro or cell-free RNA degradation assay (42) to assess the decay rates of the various DNMT1 3Ј-UTR deletion constructs. This technique measures the decay of in vitro synthesized radiolabeled RNA transcripts that are incubated with extracts prepared from cells grown under various conditions. This technique has several advantages over other methods of measuring mRNA decay (31), the most important being that it avoids the in vivo use of transcriptional inhibitors such as actinomycin D and dichlororibofuranosylbenzimidazole, both of which have been found to alter the stability of certain mRNAs (70 -73) and would therefore confound any half-life measurements. In our hands, actinomycin D stabilized the dnmt1 mRNA and therefore could not be used. We also tried using stable lines expressing tetracycline-repressible DNMT1 constructs; however, all the cell lines we obtained expressed background levels of DNMT1 in the presence of the repressor. Fig. 5, A and B, shows the results of the decay experiment performed with the various 3Ј-UTR deletions in the presence of extracts prepared from arrested Balb/c cells (arrested) or in decay buffer alone (control). We observed that the entire 3Ј-UTR and the 54-nucleotide element (⌬5Ј259) decay with halflives of 35 and 60 min, respectively. However, the half-lives of ⌬3Ј158 and ⌬3Ј56, both of which lack the conserved sequence, were significantly longer (100 and 132, min respectively), suggesting that in the presence of arrested cell extracts, this element can accelerate the degradation of the DNMT1 3Ј-UTR in vitro. No significant decay was observed with any of the RNA transcripts in buffer alone. The fact that the half-life of the full 3Ј-UTR is shorter than that of the 54-nucleotide element implies that other regions upstream of this sequence may enhance its destabilizing properties.
To test whether this decay is affected by cell growth, we compared the stability of ⌬5Ј259 RNA in the presence of growth-arrested and serum-stimulated cell extracts (Fig. 5, C  and D). The half-life of the RNA increased from 58 min when it was incubated with arrested extracts to 105 min when it was incubated with growth-induced extracts. These data are consistent with the hypothesis that the likely mechanism by which ⌬5Ј259 down-regulates mRNA in arrested cells in vivo (Fig. 3) is by decreasing mRNA stability.
The 3Ј-UTR Inhibits DNMT1-induced Transformation of NIH-3T3 Cells-It has been well documented that dnmt1 is overexpressed in a variety of tumors and cancer cell lines (15)(16)(17) and that ectopic expression of DNMT1 can transform both NIH-3T3 cells (18) and human fibroblasts (19). However, only the coding region of DNMT1 was used in these latter two studies.
DNMT1 interacts with various cell cycle regulators, such as proliferating cell nuclear antigen (74) and Rb-E2F (75), and recent data suggest that the cell cycle regulation of DNMT1 is disrupted in colorectal cancer cells (76) and in estrogen recep- stimulation. An RNA gel shift assay was performed using 32 P-labeled in vitro transcribed DNMT1 3Ј-UTR and extracts prepared from cells treated as in Fig. 1A. Time indicates hours after serum stimulation of arrested cells. The complexes were resolved by nondenaturing polyacrylamide gel electrophoresis. The complex specific to arrested cells is indicated by an arrow. C, UV cross-linking assay. Cells were arrested for 48 h (-) and serum-stimulated for 20 h (ϩ). Extracts were then prepared and subjected to a UV cross-linking assay using a 32 P-labeled ⌬5Ј259 probe, and RNA-protein complexes were resolved on an SDS-polyacrylamide gel electrophoresis gel. Protease K treatments (ϩPK) were performed for 1 h at 37°C. Control indicates reaction performed in absence of cell extract. D, competition UV cross-linking. The assay was performed as above, using extracts prepared from arrested cells. Cold competitors were synthesized from templates shown in A and used in a 1000ϫ excess. E and F, time course of p40 binding. The assay was performed as stated above, using extracts prepared from cells at various times after serum stimulation. p40 levels were quantified by densitometric scanning and normalized to the total protein in each lane, which was measured by scanning the gels after staining with Coomassie Blue. A representative experiment is shown in E, and an average of three experiments is shown in F. G, correlation between p40 binding, dnmt1 mRNA levels, and cell growth. Results from Fig. 1 and the results in F were plotted on the same graph to compare the pattern of p40 binding to dnmt1 mRNA levels. In C-E, the molecular masses in kilodaltons are indicated.
tor-negative breast cancer cells (56). We therefore tested the hypothesis that the loss of this cell cycle regulation is the main reason for the transforming properties of DNMT1 and that expression of a DNMT1 construct with its full 3Ј-UTR, which should be properly regulated, would not result in changes in cellular identity.
We generated stable NIH-3T3 cell lines expressing human DNMT1 with the full 3Ј-UTR (hDNMT1) or without the 3Ј-UTR (hDNMT1⌬UTR) under the direction of a tetracycline-repressible promoter. Our first interesting observation was that the hDNMT⌬UTR clones (n ϭ 10) grew at a much faster rate than the hDNMT1 clones. Then, after several passages, many of the transfected lines exhibited signs of cell death, indicating that expression of hDNMT1⌬UTR had toxic effects. This is not surprising, because others have also observed harmful consequences of DNMT1 overexpression (19,77). To inhibit expression of the transfected constructs and prevent this effect, we cultured the transfectants in the presence of doxycycline until 48 h prior to each experiment.
We first examined whether the constructs were differentially regulated with cell growth as expected. Fig. 6A depicts the results of a competitive reverse transcription-PCR performed to quantify the levels of hDNMT1 mRNA in transfectants that were either serum-starved or serum-stimulated for 20 h. Because the PCR primers used amplify both the mouse and human DNMT1, we then hybridized our PCR with a humanspecific DNMT1 32 P-labeled oligonucleotide probe to specifically detect the hDNMT1 constructs. Similar to the results obtained with the adenovirally expressed DNMT1 and the ␤-globin chimeras (Fig. 3), no hDNMT1 mRNA was detected in arrested cells expressing the full hDNMT1 construct (Fig. 6A,  left panel). However, in cells expressing hDNMT1⌬UTR, mRNA was present at competitor concentrations of 10 Ϫ14 -10 -16nM. As expected, in growth-stimulated cells, both constructs showed a similar expression profile (Fig. 6A, right panel).
We then studied in detail the growth characteristics of the different clones with respect to cell cycle distribution, growth rates, and morphology. Fig. 6B displays a growth curve of representative hDNMT1 and hDNMT1⌬UTR expressing clones, as well as the parental NIH-3T3 cells. Although both the control cells and the hDNMT1 clone exhibited similar growth rates, which began to level off after 10 days, the hDNMT1⌬UTR clone grew at a higher rate and did not reach saturation with high density. Accordingly, a FACS analysis carried out in the presence or absence of serum (Fig. 6C) revealed that a higher percentage of hDNMT1⌬UTR cells are found in S phase both in the presence and absence of serum. The observation that the parent cell line grows slightly faster that the hDNMT1 clone could be because the hDNMT1 clone was maintained in the presence of puromycin (the selection drug), which we found slowed down cell growth. But perhaps the most striking are the morphological changes observed in the hDNMT1⌬UTR clone. As can be seen in Fig. 6D, although both the control and hDNMT1 cells formed an organized monolayer and were contact-inhibited upon reaching confluence, the hDNMT1⌬UTR clone grew in a disorganized pattern and continued to so after confluence was reached, forming large foci. Similar results were observed with six other clones. These results suggest that the cell cycle regulatory activity of the 3Ј-UTR inhibits the cellular changes observed upon ectopic expression of DNMT1, thus supporting the hypothesis that deregulation of the cell cycle-dependent expression of DNMT1 is important for cellular transformation. DISCUSSION Because hyperactivation of DNMT1 has been implicated in cellular transformation (14), understanding how this enzyme is regulated with cell growth is of obvious interest.

FIG. 5. Differential DNMT1 3-UTR decay rates in vitro.
A, in vitro decay assay using 3Ј-UTR deletions and arrested cell extract. Radiolabeled 3Ј-UTR RNA transcripts were incubated with 10 g of whole cell extract from arrested Balb/c cells (arrested) or with incubation buffer alone (control) for various lengths of time and electrophoresed on an acrylamide gel, and the signals were detected by autoradiography and quantified by densitometry. The constructs used and their sizes are indicated beside each experiment. B, half-life measurements. Experiments in A were quantified and the signal at each time point calculated as the percentage of RNA remaining relative to the RNA at time 0. The half-life was calculated as the time point when 50% of the RNA had been degraded. Squares, ⌬3Ј56; diamonds, 3Ј-UTR; circles, ⌬ 3Ј158; triangles, ⌬ 5Ј259. C, in vitro decay assay using ⌬5Ј259 and arrested or growth-stimulated extracts. Radiolabeled ⌬5Ј259 transcript was incubated with 10 g of whole cell extract from arrested or serum-stimulated (ϩserum) Balb/c cells for various lengths of time and processed as in A. D, ⌬5Ј259 half-life measurements. The experiment shown in C was quantified, and half-lives were calculated as in B. Squares, ϩserum; diamonds, -serum.
In this paper, we first confirm previous findings that the dnmt1 mRNA is regulated with the cell cycle (9,29). dnmt1 mRNA is essentially absent in arrested cells and then increases as cells are induced by serum and reaches its maximum level just before the peak of S phase; this coordination would ensure that sufficient DNMT1 protein is synthesized so as to carry out its maintenance function during replication. Second, we demonstrate that the 3Ј-UTR of DNMT1 confers a growth-dependent regulation on both its own mRNA and a heterologous stable ␤-globin mRNA. Third, we identified a highly conserved 54nucleotide element within the 3Ј-UTR that is required for mediating the down-regulation of dnmt1 mRNA in arrested cells. Thus, cis-acting sequences in the 3Ј-UTR are likely responsible for the changes in dnmt1 mRNA which occur during the cell cycle. Fourth, the 3Ј-UTR is destabilized in vitro in the presence of extracts prepared from arrested cells, suggesting that a protein that is present primarily in arrested cells targets the 3Ј-UTR for degradation. The decrease in half-life of ⌬5Ј259 cannot be attributed to the fact that it is shorter than ⌬3Ј158 and ⌬3Ј56, because the full 3Ј-UTR decays at a faster rate. One conceivable explanation for the reduced destabilizing activity of ⌬5Ј259 relative to the full 3Ј-UTR is that AU-rich instability sequences exist upstream of ⌬5Ј259. A possible model is that the 54-nucleotide element is responsible for the growth-dependent stability changes, and region(s) upstream enhance its destabilizing activity.
Whereas we were able to show that the 3Ј-UTR down-regulates dnmt1 mRNA in growth-arrested Balb/c cells, we were not able to demonstrate that this occurs by destabilization of dnmt1 mRNA in living cells. Actinomycin D is routinely used to block de novo synthesis of RNA in order to measure mRNA decay. However, we found that actinomycin D stabilizes the dnmt1 mRNA. This is not uncommon, and it has been reported for several other mRNAs (70 -73). We also could not obtain repressible DNMT1 transfectants that show zero DNMT1 expression in the repressed state, which could be used for measuring stability of mRNA in cells. However, the fact that the same element that reduces the presence of dnmt1 mRNA in arrested cells also destabilizes dnmt1 mRNA in vitro supports the conclusion that the 3Ј-UTR acts as a cell cycle-regulated destabilizing element.
Many studies have shown that mRNA posttranscriptional regulation plays an important role during development (32,33) and cell cycle progression (34,35) and exerts an important control over expression of proto-oncogenes (36,37) and cytokines (38,39). Several pathological states are associated with aberrant regulation at this level, including various immune and inflammatory diseases and several types of cancer (78). One potential mechanism by which the 3Ј-UTR may function is by the differential binding of trans-acting factors during the cell cycle; these RNA-protein complexes could then either increase or decrease mRNA levels. Studies involving several mRNAs, such as parathyroid hormone (79), GM-CSF (68), and CYP7A1 (73) have found that changes in mRNA stability are correlated with the binding of proteins within the 3Ј-UTR. In accordance with this postulated mechanism, we show, using an RNA mobility gel shift assay, that a specific RNA-protein complex is formed within the 3Ј-UTR sequence only in growtharrested cells. In addition, our UV cross-linking experiments reveal a protein of ϳ40 kDa (p40) that exhibits specific binding to the conserved 54-nucleotide element. This binding is highly induced in arrested cells and decreases at the same time as dnmt1 mRNA increases during cell cycle progression. Although our data do not demonstrate a causal link between the p40 -3Ј-UTR interaction and dnmt1 mRNA down-regulation, our results are in accordance with this hypothesis. Further purification and characterization of p40, which is currently under way in our laboratory, may provide us with a better understanding of its mechanism of action.
It has been recently suggested that the deregulation of dnmt1 with the cell cycle rather than its overall levels is im-FIG. 6. Effects of the 3-UTR on cell growth following ectopic expression of hDNMT1 in NIH-3T3 cells. A, measurement of hD-NMT1 mRNA levels by competitive PCR. 1 g of RNA from TetOff NIH 3T3 hDNMT1 or hDNMT1⌬UTR stable lines, either growth-arrested or serum-stimulated, was reverse-transcribed, and equal amounts of cDNA (as determined by [ 35 S]dATP incorporation) were subjected to DNMT1 PCR amplification in the presence of decreasing concentrations (10 Ϫ12 -10 Ϫ16 nM) of a competitor DNA fragment (upper panels). To differentiate between the endogenous mouse dnmt1 and transfected hDNMT1, the PCR was transferred to Hybond Nϩ membrane and hybridized with an oligonucleotide corresponding to bases 3595-3617 of the full-length human DNMT1 (lower panels). T, target; C, competitor. B, growth curves of the stable and parent lines. Cells were plated in six-well culture dishes at a density of 25,000 cells/well (day 0). Cells were then counted on 6 subsequent occasions throughout the span of 11 days. An average of 4 measurements was obtained for each time point. Squares: hDNMT1⌬UTR, triangles: NIH-3T3, diamonds: hDNMT1. C percentage of cells in S phase of the cell cycle. Serum-starved (-serum) or growth-stimulated (ϩserum) cells were harvested and stained with propidium iodide, and the percentage of cells in S phase was determined by flow cytometry. D, morphology of hDNMT1 transfectants. Phase contrast light microscopy at increasing magnifications on day 11 of the growth curve described in B. Top panels, NIH-3T3; middle panels, hDNMT1; bottom panels, hDNMT1⌬UTR portant for cellular transformation by dnmt1 (7,29,30). The data presented here are consistent with this hypothesis. In agreement with previous studies (18,19), we found that expression of an exogenous human DNMT1 construct comprising only the coding region, which is expressed under conditions of serum deprivation, is able to alter the growth properties of NIH 3T3 cells; the cells grow at a faster rate, and a higher percentage of cells are in S phase. They also exhibit a loss of contact inhibition and do not reach saturation even at high density. In contrast, cells expressing a DNMT1 construct that includes the 3Ј-UTR and is not expressed in serum-starved cells maintain the growth properties of the control cells.
The mechanisms responsible for cellular transformation by DNMT1 are still unresolved. However, one proposed hypothesis is that elevated levels of DNMT1 lead to the hypermethylation and inactivation of tumor suppressor genes (80). It is possible that ectopic expression of dnmt1 in arrested cells results in de novo methylation and silencing of a critical tumor suppressor gene. However, if this is the mechanism, it is not clear why such a gene is methylated only when dnmt1 is expressed in arrested cells. An alternative hypothesis is that ectopic expression of dnmt1 causes cell transformation by interfering with cell cycle regulatory circuits through DNMT1 protein-protein interactions. Because DNMT1 forms a complex with Rb and E2F (75) as well as histone deacetylase 1 and 2 (81,82), it can inhibit the expression of tumor suppressors by a mechanism that does not involve DNA methylation, as has been previously shown (83). DNMT1 can also displace p21 from proliferating cell nuclear antigen and allow replication to occur (74), and inhibition of DNMT1 has been found to inhibit DNA replication (10,11). Thus, aberrant expression of dnmt1 during the G 0 /G 1 phase may override the silencing of tumor suppressors and eliminate normal arrest signals, leading to the inappropriate entry into the cell cycle that is observed in cancer cells, as has been previously suggested (84).
Although further studies are required to confirm this hypothesis, our findings clearly illustrate the importance of the dnmt1 3Ј-UTR in regulating its expression with the state of cell growth.