DNA Methyltransferase 1 Knock Down Induces Gene Expression by a Mechanism Independent of DNA Methylation and Histone Deacetylation*

DNA methyltransferase 1 (DNMT1) catalyzes the post-replication methylation of DNA and is responsible for maintaining the DNA methylation pattern during cell division. A long list of data supports a role for DNMT1 in cellular transformation and inhibitors of DNMT1 were shown to have antitumorigenic effects. It was long believed that DNMT1 promoted tumorigenesis by maintaining the hypermethylated and silenced state of tumor suppressor genes. We have previously shown that DNMT1 knock down by either antisense oligonucleotides directed at DNMT1 or expressed antisense activates a number of genes involved in stress response and cell cycle arrest by a DNA methylation-independent mechanism. In this report we demonstrate that antisense knock down of DNMT1 in human lung carcinoma A549 and embryonal kidney HEK293 cells induces gene expression by a mechanism that does not involve either of the known epigenomic mechanisms, DNA methylation, histone acetylation, or histone methylation. The mechanism of activation of the cell cycle inhibitor p21 and apoptosis inducer BIK by DNMT1 inhibition is independent of the mechanism of activation of the same genes by histone deacetylase inhibition. We determine whether DNMT1 knock down activates one of the nodal transcription activation pathways in the cell and demonstrate that DNMT1 activates Sp1 response elements. This activation of Sp1 response does not involve an increase in either Sp1 or Sp3 protein levels in the cell or the occupancy of the Sp1 elements with these proteins. The methylation-independent regulation of Sp1 elements by DNMT1 unravels a novel function for DNMT1 in gene regulation. DNA methylation was believed to be a mechanism for suppression of CG-rich Sp1-bearing promoters. Our data suggest a fundamentally different and surprising role for DNMT1 regulation of CG-rich genes by a mechanism independent of DNA methylation and histone acetylation. The implications of our data on the biological roles of DNMT1 and the therapeutic potential of DNMT1 inhibitors as anticancer agents are discussed.

DNA methyltransferase 1 (DNMT1) catalyzes the postreplication methylation of DNA and is responsible for maintaining the DNA methylation pattern during cell division. A long list of data supports a role for DNMT1 in cellular transformation and inhibitors of DNMT1 were shown to have antitumorigenic effects. It was long believed that DNMT1 promoted tumorigenesis by maintaining the hypermethylated and silenced state of tumor suppressor genes. We have previously shown that DNMT1 knock down by either antisense oligonucleotides directed at DNMT1 or expressed antisense activates a number of genes involved in stress response and cell cycle arrest by a DNA methylation-independent mechanism. In this report we demonstrate that antisense knock down of DNMT1 in human lung carcinoma A549 and embryonal kidney HEK293 cells induces gene expression by a mechanism that does not involve either of the known epigenomic mechanisms, DNA methylation, histone acetylation, or histone methylation. The mechanism of activation of the cell cycle inhibitor p21 and apoptosis inducer BIK by DNMT1 inhibition is independent of the mechanism of activation of the same genes by histone deacetylase inhibition. We determine whether DNMT1 knock down activates one of the nodal transcription activation pathways in the cell and demonstrate that DNMT1 activates Sp1 response elements. This activation of Sp1 response does not involve an increase in either Sp1 or Sp3 protein levels in the cell or the occupancy of the Sp1 elements with these proteins. The methylation-independent regulation of Sp1 elements by DNMT1 unravels a novel function for DNMT1 in gene regulation. DNA methylation was believed to be a mechanism for suppression of CG-rich Sp1-bearing promoters. Our data suggest a fundamentally different and surprising role for DNMT1 regulation of CG-rich genes by a mechanism independent of DNA methylation and histone acetylation. The implications of our data on the biological roles of DNMT1 and the therapeutic potential of DNMT1 inhibitors as anticancer agents are discussed.
DNA modification by methylation of cytosines residing at the dinucleotide sequence CG plays an important role in epigenomic programming of gene expression (1). Not all CGs are methylated, and the pattern of distribution of methylated and unmethylated CGs is cell type-specific (2). DNA methylation in regulatory regions of genes plays a role in silencing genes either by directly inhibiting the interaction of transcription factors with their regulatory sequences (3,4) or by attracting methylated DNA-binding proteins, which in turn recruit histone deacetylases and histone methyltransferases, resulting in an inactive chromatin structure (5,6). DNA methylation is catalyzed by DNA methyltransferases DNMTs, 1 which transfer the methyl moiety from the methyl donor S-adenosylmethionine to 5th position on the cytosine ring (7). DNMT1 is responsible for maintaining the DNA methylation pattern during embryonal development and cell division (8,9). DNMT1 deregulation was proposed to play a critical role in cellular transformation (10). Forced expression of DNMT1 was shown to transform NIH 3T3 cells (11), DNMT1Ϫ/Ϫ knockouts are resistant to colorectal tumorigenesis (12), and antisense knock down of DNMT1 reverses tumorigenesis in vitro (13,14) and in vivo (15). The mechanisms through which DNMT1 causes cellular transformation and through which inhibition of DNMT1 reverses cellular transformation are unknown (16). The most obvious mechanism is that aberrant expression of DNMT1 causes methylation and silencing of tumor suppressor genes (17,18). This hypothesis is supported by numerous documentations of methylated tumor suppressor genes in tumors (19). In accordance with this hypothesis, knock down of DNMT1 by either antisense or siRNA results in demethylation and activation of tumor suppressor genes such as p16 (14,20).
However, it was surprisingly previously shown that knock down of DNMT1 results in induction of the unmethylated tumor suppressor gene p21 by a mechanism that does not involve DNA methylation (21). More recently it was shown that DNMT1 interacts with histone deacetylases (HDACs) 1 (22) and 2 (23) as well as histone methyltransferase SUV39H1 (24), suggesting that DNMT1 silences gene expression by recruiting chromatin-modifying enzymes. It was also shown that ectopic expression of DNMT1 could suppress exogenous genes bearing E2F1 sites by recruiting Rb⅐E2F1⅐HDAC1 complex (25). However, it is not clear whether DNMT1 regulates endogenous genes by these mechanisms.
We have previously shown that DNMT1 expression is regulated with the cell cycle (8,26,27) and that antisense knock down of DNMT1 results in an intra-S-phase arrest of DNA replication (28). This intra-S-phase arrest requires knock down of the DNMT1 protein rather than inhibition of DNA methyl-ation, suggesting that DNMT1 protein plays an important role in the regulatory circuitry independent of its methyltransferase activity (29). We and others have previously shown that antisense ODNs directed at DNMT1 exhibit sequence-specific down-regulation of DNMT1 and that they could be utilized to delineate the immediate consequences of knock down of the DNMT1 protein on gene expression and cellular regulation (14,20,28,30,31). In difference from agents such as 5-azacytidine that inhibit the catalytic activity of DNMT1 and, thus, mainly measure the DNA methylation-dependent roles of this protein (32), antisense knock down of DNMT1 causes an immediate arrest of DNA replication (28), resulting in limited passive demethylation. This allows us to study mainly DNA methylation-independent regulatory functions of DNMT1. We have previously shown that a number of stress response-related genes are activated by DNMT1 knock down (28). In this report we dissect the transcription factor pathway activated by DNMT1 knock down and show that surprisingly DNMT1 regulates gene expression by a new mechanism that does not involve either DNA methylation or chromatin modification. We show that DNMT1 knock down and histone deacetylation inhibition activate gene expression by independent mechanisms. These results have important implications on understanding the regulatory roles of DNMT1 in the cell cycle and cellular transformation, as well as on future approaches to targeting DNMT1 in anticancer therapy.

MATERIALS AND METHODS
Cell Culture, Antisense Oligonucleotides, and TSA Treatment-A549 cells, a human non-small cell lung carcinoma-derived cell line (33) (ATCC:CCL 185), were grown in Dulbecco's modified Eagle's medium (low glucose) supplemented with 10% fetal calf serum and 2 mM glutamine. 18 -24 h before treatment, cells were plated at a density of 3 ϫ 10 5 cells/100-mm tissue culture dish, or 4 ϫ 10 4 cells/well in a six-well plate in the absence of antibiotics. The phosphorothioate oligodeoxynucleotides used in this study were MG88 (human DNMT1 antisense) and its mismatch control MG208, which has a 6-bp difference from MG88 (14). Oligonucleotides were transfected into cells at a concentration of 120 nM with 3.75 g/ml Lipofectin (Invitrogen) in serum-free Opti-MEM (Invitrogen). The oligonucleotide-containing Opti-MEM was removed from the cells and replaced with regular growth media after 4 h. The treatment was repeated after 24 h. The cells were harvested 12, 18, 24, 30, 36, 42, or 48 h after the first treatment. For TSA treatment, cells were grown in regular culture media in the presence of 1 M of trichostatin A (TSA). The cells were harvested 2, 6, or 12 h after the initiation of treatment. HEK293 cells, a human adenovirus type 5-transformed human embryonal kidney cell line (ATCC, CRL 1573), were grown in Dulbecco's modified Eagle's medium (high glucose) supplemented with 10% fetal calf serum and 2 mM glutamine.
Western Blot Analysis-50 g of nuclear protein was fractionated on a 5% SDS-polyacrylamide gel, transferred to polyvinylidene difluoride membrane, and reacted with the polyclonal anti-DNMT1 antibody (New England Biolabs) at a dilution of 1:2,000 in the presence of 0.05% Tween and 5% milk, and it was then reacted with anti-rabbit IgG (Sigma) at a dilution of 1:5,000 in the presence of 0.05% Tween and 5% milk. As a control for protein loading, ␤-actin protein levels were examined in the same nuclear extracts. 50 g of protein was fractionated on a 10% SDS-polyacrylamide gel, transferred to nitrocellulose membrane, and reacted with the monoclonal anti-␤-actin antibody (Sigma, A5316) at a dilution of 1:5000 in the presence of 0.05% Tween and 5% milk, and it was then reacted with anti-mouse IgG (Jackson Immunochemicals) at a dilution of 1:20,000 in the presence of 0.05% Tween.
Bisulfite Mapping of the BIK Promoter-A549 cells were treated with 120 nM MG88 or MG208 for 24 h. Bisulfite mapping was performed as described previously with minor modifications (34). 50 ng of sodium bisulfite-treated DNA samples were subjected to PCR amplification using the first set of primers (sense, 5Ј-GTAAAAAAAGTTAGATTT-GTGG-3Ј; antisense, 5Ј-CTCACCTCCTCTAAATACC-3Ј). PCR products were used as templates for the nested PCR using the second set of primers (sense, 5Ј-AGGGATTGGGGGAGGAG-3Ј; antisense, 5Ј-CAAC-TACTCACACCTCAAC-3Ј). The PCR products of the second reaction were subcloned into a TA cloning vector (Invitrogen), and the clones were sequenced using the T7 Sequencing kit (Amersham Biosciences).
Reporter Constructs and Luciferase Assay-The calcium phosphate precipitation method was used to transiently co-transfect HEK293 cells, plated in six-well dishes, with 2 g of luciferase reporter constructs and 5 g of either control pcDNA3.1 vector (Invitrogen) or a DNMT1 antisense (bp 396 -5066) subcloned into pcDNA3.1 vector (as DNMT1). The reporter constructs that were used contained either full p21 promoter (Ϫ2326 bp) or four different 5Ј deletions (Ϫ1481, Ϫ883, Ϫ291, and Ϫ94) upstream from the luciferase gene (35). In addition, Sp1-responsive reporters were constructed using basic luciferase reporter containing only TATA box and initiator sequence (pGL2 TϩI). An oligonucleotide containing consensus Sp1 binding site corresponding to the bases Ϫ71 to Ϫ86 of the p21 promoter was synthesized and inserted either once (1 ϫ Sp1) or four times (4 ϫ Sp1) into the pGL2 TϩI vector (36). 48 h after transfection, the cells were lysed and luciferase activity was assayed using the Luciferase Assay System (Promega) according to the manufacturer's instructions.
A549 cells plated in six-well dishes were transfected using 3.75 g/ml Lipofectin (Invitrogen) and 0.5 g of the reporter vectors containing different cis-acting enhancer elements upstream from the firefly luciferase gene. All the reporters, except for the 4 ϫ Sp1, belong to the Mercury Pathway Profiling System (Clontech). 24 h after transfection, the cells were either treated with 120 nM MG88 and MG208 for 24 h or with 1 M TSA for 6 h. The luciferase activity was assayed using the Luciferase Assay System (Promega).
The full p21 promoter-luciferase reporter construct (Ϫ2326 p21) was used to create a construct with 2-bp mutations (CC3 GA) within the Sp1 site corresponding to bases between Ϫ71 and Ϫ86 of the p21 promoter (Ϫ2326 p21 CC3 GA). A QuikChange XLII site-directed mutagenesis Kit (Stratagene) was used with the following mutagenic primers: 5Ј-GCGCGGGTCCCGGATCCTTGAGGCGGG-3Ј and 5Ј-CCCGC-CTCAAGGATCCGGGACCCGCGC-3Ј. The amplification cycles were as follows: 98°C 3 min, 18 cycles (98°C 50 s, 68°C 17 min), 68°C 7 min. A549 cells plated in six-well dishes were transfected using 3.75 g/ml Lipofectin (Invitrogen) and 0.5 g of either wild type or Sp1-mutated full p21 promoter construct. 24 h after transfection, the cells were treated with 120 nM MG88 and MG208 for 24 h, and the luciferase activity was assayed using the Luciferase Assay System (Promega).
To assess the effect that Sp1 mutation might have on Sp1 occupancy of the p21 promoter in living cells, the wild type (Ϫ2326 p21) and Sp1-mutated (Ϫ2326 p21 CC3 GA) p21 promoter luciferase reporter construct, were transiently transfected into A549 cells using 3.75 g/ml Lipofectin (Invitrogen) and 1.5 g/10 cm plate of the reporter construct. 48 h after transfection, the cells were cross-linked, and the chromatin was immunoprecipitated using 10 g of Sp1 antibody (Upstate Biotech-nology, #07-124) as described above. The precipitated wild type and Sp1-mutated reporter constructs were PCR-amplified using the same primers (sense, 5Ј-ACCAACGCAGGCGAGGGACT-3Ј; antisense, 5Ј-CGCCGGGCCTTTCTTTATG-3Ј). The 5Ј primer corresponds to the bases between Ϫ256 and Ϫ236 of the p21 promoter, whereas the 3Ј primer corresponds to the luciferase sequence. Thus our primers will selectively amplify the exogenous p21 promoter luciferase reporter and not the endogenous wild type p21 promoter. The FailSafe PCR system (Epicenter) was used with the pre-mix D, 0.5 M of each primer, 0.5 l of FailSafe polymerase, and 2 l of ChIP product with the following program: 95°C 5 min, one cycle (95°C, 60°C, and 72°C for 30 s each), one cycle (95°C, 58°C, and 72°C for 30 s each), 23 cycles (95°C, 56°C, and 72°C 30 s each), 72°C 5 min. 20 l of the PCR products was run on a 1.4% gel and visualized by ethidium bromide staining. The intensity of the bands was quantified by densitometric analysis using MCID software (Imaging Research Inc.). The intensities of the p21-amplified bands immunoprecipitated with the anti-Sp1 antibody were normalized to the intensities of the bands amplified from the input material.
Electrophoretic Mobility Shift Assays-Complementary oligonucleotides carrying either the Sp1 binding site corresponding to bases Ϫ71 and Ϫ86 of the wild type p21 promoter (5Ј-GGTCCCGCCTCCT-TGA-3Ј and 5Ј-TCAAGGAGGCGGGACC-3Ј) or to a mutated Sp1 site containing 2-bp mismatch (CC3 GA) (5Ј-GGTCCCGGATCCTTGA-3Ј and 5Ј-TCAAGGATCCGGGACC-3Ј) were synthesized and annealed. 20 ng of the annealed oligonucleotides was end-labeled with [␥-32 P]ATP and T4 polynucleotide kinase (MBI) and gel-purified from a 5% non-denaturing polyacrylamide gel. To measure the binding of either the Sp1 site or its mutated version to Sp1 protein 1 ϫ 10 5 counts of the labeled oligonucleotides were used in a gel shift reaction containing 400 ng of the purified human recombinant Sp1 protein (Promega, catalog number E6391), 4% glycerol, 1 mM MgCl 2 , 0.5 mM EDTA, 0.5 mM dithiothreitol, 50 mM NaCl, 10 mM Tris-HCl (pH 7.5), and 0.05 mg/ml poly(dI-dC)/(dI-dC), and the reaction was incubated for 20 min at 23°C. To demonstrate the presence of Sp1 in the DNA⅐protein complex, the Sp1⅐DNA reaction mixtures were preincubated with anti-Sp1 antibody (Santa Cruz Biotechnology, #sc-59) for 10 min before the addition of the labeled oligonucleotides. The complexes were resolved on a 5% non-denaturing polyacrylamide gel, and the gel was dried and analyzed by autoradiography.

DNMT1
Antisense and HDAC Inhibitor TSA Induce p21 and BIK Gene Expression; p21 and BIK Are Regulated by HDAC Activity-DNMT1 knock down was previously shown using a microarray gene expression analysis to induce expression of a cluster of genes involved in stress response (28). The kinetics of induction indicated that some of these genes were induced by a methylation-independent mechanism. All of the methylationindependent pathways of gene repression by DNMT1 that have been proposed to date are based on the ability of DNMT1 to interact with and possibly recruit histone deacetylases HDAC1 and HDAC2 (22, 25) (23) and histone methyltransferase SUV39H1 to promoters (24). However, the data supporting these interactions are based on ectopic expression experiments, and there is no evidence as of yet that an endogenous unmethylated gene is regulated by DNMT1-HDAC1 interaction through a histone acetylation-dependent mechanism. We therefore tested the hypothesis that DNMT1 knock down induces gene expression by a histone acetylation-dependent mechanism.
We focused on two genes that were demonstrated to be induced by DNMT1 knock down, the cell cycle inhibitor p21 and apoptosis inducer, Bcl-2 interacting killer, BIK. We have previously shown that the induction of p21 is methylationindependent (21), and the kinetics of BIK induction suggested that it is also induced by a methylation-independent mechanism (28).
We first tested the hypothesis that p21 and BIK genes are regulated by the state of histone acetylation by taking advantage of the general HDAC inhibitor trichostatin A (TSA). A549 cells were treated with 120 nM of either antisense to DNMT1 MG88, or its mismatch control MG208, for 12-42 h as well as with 1 M TSA for 2-12 h. We show that both DNMT1 knock down and HDAC inhibition by TSA induce these genes to a comparable degree (4-to 5-fold). MG88 induces almost complete knock down of DNMT1 mRNA and protein 24 h after treatment (Fig. 1, A and B) at which point we observe peak induction of p21 and BIK mRNA (Fig. 1, B and C). TSA causes peak induction somewhat earlier, between 6 and 12 h (Fig. 2) as previously demonstrated (37). The slight delay in induction of BIK and p21 by DNMT1 knock down reflects the different mechanisms of action of these two inhibitors. TSA inhibits the catalytic activity of HDACs, whereas MG88 inhibits de novo synthesis of DNMT1 mRNA and protein. The delay in induction probably reflects the turnover rate of DNMT1 protein. The induction of p21 and BIK with TSA illustrates that these genes are down-regulated by histone deacetylation in their basal state, and it is therefore possible that knock down of DNMT1 affects the state of histone acetylation of these genes. We now address the question of whether induction of these genes by DNMT1 knock down and histone deacetylation inhibition act through identical or independent pathways.
Knock Down of DNMT1 Does Not Cause Demethylation of BIK Promoter and Induces Its Expression by a Methylationindependent Pathway-CG methylation is believed to be an on-off switch in gene expression. Hypermethylation of promoter regions is associated with compact chromatin inaccessi- ble to the transcriptional machinery, which results in a stable suppression of gene expression (1). Hence, DNA methylation is unlikely to play a role in the regulation of p21 and BIK genes, because they are expressed even in the untreated cells ( Figs. 1  and 2). In accordance with this hypothesis, we have previously shown that the regulatory region of the p21 promoter upstream of transcription start site is unmethylated in A549 cells (21).
To rule out the possibility that induction of the BIK expression by DNMT1 knock down involves demethylation of its promoter, we examined the state of methylation of the BIK promoter after treatment with either MG88 or MG208. BIK gene bears a CG island around the transcriptional start site, which renders it susceptible to control by methylation (Fig. 3A). We performed bisulfite mapping of the proximal part of this CG island and found that it is predominantly unmethylated, except for the sporadically methylated CG sites in both MG88-and MG208-treated A549 cells (Fig. 3B). These results suggest that DNMT1 knock down does not result in demethylation of the BIK promoter. Thus, similar to p21 (16,21), BIK seems to be induced by a methylation-independent mechanism.
Histone Acetylation of Promoters of p21 and BIK Is Induced by TSA Treatment but Remains Unchanged after DNMT1 Antisense Knock down; DNMT1 and HDAC Control p21 and BIK Expression by Independent Pathways-If DNMT1 knock down induces p21 and BIK through either a histone deacetylase-dependent pathway (22) or a histone methyltransferase-depend-ent pathway (24), as predicted by the published protein-protein interactions of DNMT1, MG88 treatment should cause a change in either the state of acetylation or methylation of histones associated with these genes. To test this hypothesis we performed chromatin immunoprecipitations (ChIP) with antibodies against acetylated H3 or acetylated H4 histones, followed by PCR amplification of the p21 and BIK promoters. As shown in Fig. 4, there was no detected difference in histone acetylation at either 24 or 48 h after MG88 treatment. We then tested whether the knock down of DNMT1 reduces K9 H3 histone methylation. We performed ChIP analysis on A549 cell treated with 120 nM MG88 or MG208 for 48 h, using antidimethyl K9 H3 antibody. We did not detect any changes in histone methylation at either p21 or BIK promoters (Fig. 4B), which suggested that the effects of DNMT1 knock down are not mediated through the disruption of DNMT1⅐SUV39H1 complex. To verify that histones associated with these regions are subject to a change, we tested whether HDAC inhibitor TSA, which produced comparable induction of p21 and BIK mRNA (Fig. 2), induced a change in acetylation of histones associated with these regions. As shown in Fig. 4C  The primers used to amplify the bisulfited DNA are shown as arrows on the top of the promoter. B, A549 cells were treated with 120 nM MG88 or MG208 for 24 h. DNA was isolated and treated with sodium bisulfite, which converts all unmethylated cytosines into thymidines while methylated cytosines remain intact. The specific methylation pattern is revealed by PCR amplification, subcloning, and sequencing of the region of interest. The methylation pattern of 19 CG sites in the BIK promoter is shown (10 clones for each treatment). Open and closed circles indicate unmethylated and methylated sites, respectively. treatment for 6 h causes a 5-7 fold increase in the state of acetylation of both histones H3 and H4 associated with promoter regions of p21 and BIK. We conclude that the state of acetylation of histones associated with these regions is regulated by HDACs, but DNMT1 knock down does not affect either HDAC or histone methyltransferase activity on these histones.  4. Histone acetylation of promoters of p21 and BIK is induced by TSA treatment but remains unchanged after DNMT1 antisense knock down. A, the map of p21 and BIK promoters is shown, indicating the primers and the sizes of regions that they amplify. The analysis of promoter sequences for Sp1 binding sites (gray circles) was done using the SignalScan program of the BioBase web site (www.gene-regulation-.com). B, A549 cells were treated with either 120 nM MG88 or MG208 for 24 or 48 h. Chromatin was isolated and immunoprecipitated with antibodies specific for acetylated histones H4 and H3, as well as dimethyl Lys-9 of histone H3. Associated DNA was analyzed by PCR using primers that amplify 255-and 307-bp regions of p21 and BIK promoters, respectively. Triplicates are shown for each experiment. The regions of p21 and BIK rich in Sp1 elements (324 and 307 bp, respectively) were tested for Sp1 and Sp3 binding in Fig. 7. C, similar chromatin immunoprecipitations were performed on control cells or cells treated with 1 M TSA for 6 h. Anti-acetylated H4 and H3 antibodies were used. Triplicates are shown for each experiment. D, the relative occupancy of the promoters with each of the studied histone states is determined by densitometric quantification of PCR products and normalization to the input DNA. The rate of induction following MG88 and TSA treatment is shown as a percentage of MG208 and control cells, respectively. The error bars are standard deviations of the triplicates. There was no PCR amplification in no antibody control immunoprecipitations (data not shown). tin, we tested whether this induction is mediated by the activation of any of the principal transcription activation pathways. We first used p21 promoter deletion constructs and identified a minimal promoter (94 bp) responsive to reduction of DNMT1 levels. We transiently co-transfected HEK cells with either control vector or vector expressing full-length antisense to DNMT1 and five p21 deletion constructs. We identified that the 94-bp region of the p21 promoter is equally induced with DNMT1 antisense as the entire 2.3 kbp of p21 promoter. The analysis of this region for transcription factor binding sites revealed that this region is rich in Sp1 and E2F consensus sites (Fig. 5A). Previous studies using mutation analysis of this region identified the Sp1 site corresponding to bases between Ϫ71 and Ϫ86 of the p21 promoter to be essential for its activation by histone deacetylation inhibition (37). It was also shown that the luciferase reporter containing four tandem copies of this Sp1 site is as capable in mediating transforming growth factor-␤ effects on p21 gene expression, as is the luciferase reporter containing the full p21 promoter (36). We tested whether this 4 ϫ Sp1 reporter construct is induced by DNMT1 antisense and found that this Sp1 site is also sufficient to mediate the effects of DNMT1 inhibition (Fig. 5B). This data demonstrates that both antisense ODN (Fig. 6) and expressed antisense to DNMT1 (Fig. 5) induce 4 ϫ Sp1 and p21 promoter activity suggesting that the effects observed in this report with ODN are a true consequence of DNMT1 knock down rather than a nonspecific effect of the ODNs. In addition, this exper-iment shows that DNMT1 knock down can induce gene expression in a different cell line than A549, the human embryonal kidney, HEK293 cell line (Fig. 5).

HDAC inhibitors and
Mutation of the Sp1 Site Corresponding to Bases between Ϫ71 and Ϫ86 of the p21 Promoter Abolishes Induction of Expression by DNMT1 Knock Down-The data presented in the previous section illustrated that the Sp1 site at Ϫ71 to Ϫ86 of the p21 promoter could mediate induction of transcription in response to DNMT1 knock down. To test whether this Sp1 site is required for induction of p21 within the context of the full p21 promoter, we generated a construct containing CC3 GA mutation of this site within the full-length p21 promoter luciferase reporter construct (Ϫ2326 CC3 GA). We transfected A549 cells with either the wild type (Ϫ2326 p21) or the Sp1mutated (Ϫ2326 p21 CC3 GA) reporter construct and then treated the cells with either 120 nM MG88 or MG208 for 24 h. The construct bearing the mutated Sp1 site showed a marked reduction in luciferase expression suggesting that this site is important for the basal activity of the p21 promoter. In addition, this Sp1 site is essential for the effects of DNMT1 knock down by MG88, because its mutation abolishes the induction of expression observed with the wild type p21 promoter construct (Fig. 6B). To confirm that CC3 GA mutation precludes Sp1 binding, we performed gel shift assays using either wild type or mutated synthetic oligonucleotides corresponding to the bases between Ϫ71 and Ϫ86 of the p21 promoter. Purified recombinant human Sp1 protein formed a DNA⅐protein complex exclu-  6. Mutation of the Sp1 site corresponding to bases between ؊71 and ؊86 of the p21 promoter precludes Sp1 binding and abolishes the activation produced by DNMT1 knock down. A, the physical maps of wild type (Ϫ2326 p21) and Sp1 mutated (Ϫ2326 p21 CC3 GA) full p21 promoter luciferase reporter construct are shown. The sequence of the p21 promoter Sp1 site used for the gel shifts (C) is indicated, with the boldface letters representing the Sp1 recognition element and the underlined letters indicating the mutated bases. The arrows indicate the primers used for ChIP (D). B, A549 cells were transfected with 0.5 g of either wild type (Ϫ2326 p21) or Sp1-mutated (Ϫ2326 p21 CC3 GA) full p21 promoter luciferase reporter construct. 24 h post-transfection, the cells were treated with 120 nM MG88 and MG208 for 24 h, and the luciferase activity was measured and normalized as described under "Materials and Methods." C, 400 ng of recombinant human Sp1 protein (hrSp1) was incubated with end-labeled oligonucleotides corresponding to bases between Ϫ71 and Ϫ86 of either a wild type or CC3 GA mutated p21 promoter. A supershift was generated by incubating the reaction with 1 g of anti-Sp1 antibody. The positions of the Sp1⅐DNA complex and the anti-Sp1-Sp1-DNA supershift are indicated with arrows. D, A549 cells were transfected with 1.5 g of either Ϫ2326 p21 or Ϫ2326 p21 CC3 GA construct. Chromatin was isolated and immunoprecipitated with anti-Sp1 antibody, followed by PCR amplification of the 256 bp of the proximal p21 promoter of the transfected luciferase reporters. The relative occupancy of the promoters with Sp1 is determined by densitometric quantification of PCR products in the anti-Sp1 immunoprecipitate and normalization to the input DNA. The normalized Sp1 occupancy of the wild type construct (Ϫ2326 p21) and the mutated construct (Ϫ2326 p21 CC3 GA) is represented in arbitrary units. The error bars are S.E. values of the triplicates. There was no PCR amplification in no antibody control immunoprecipitations (data not shown). sively with the wild type Sp1 oligonucleotide but not with the CC3 GA mutated Sp1 oligonucleotide (Fig. 6C). To determine that Sp1 is present in the complex, we demonstrated that an anti-Sp1 antibody supershifted the complex (Fig. 6C). To test whether this mutation precludes the binding of Sp1 to the p21 promoter in living cells, we transfected A549 cells with either wild type (Ϫ2326 p21) or Sp1-mutated (Ϫ2326 CC3 GA p21) full p21 promoter reporter construct and performed a ChIP assay using anti-Sp1 antibody 48 h after transfection. The binding of the endogenous Sp1 protein to the Sp1-mutated construct was markedly reduced compared with the wild type p21 promoter construct (Fig. 6D). Some minute Sp1 binding to the Ϫ2326 CC3 GA construct was detected suggesting that other putative Sp1 elements found in the p21 promoter might also mediate Sp1 protein binding in A549 cells. However, our results suggest that the occupancy of the p21 promoter with Sp1 is mostly determined by the Sp1 site at Ϫ71 to Ϫ86.
DNMT1 Antisense and TSA Affect Different Transcription Activator Pathways-Our previously published differential microarray gene expression analysis revealed that DNMT1 knock down results in induction of multiple genes, including many stress response genes (28). The downstream response of a cell to different physiological and toxic signals involves activation of distinct transcription factors, which in turn activate genes bearing their cognate response elements. To test which nodal transcription factor pathway in the cell is responsive to DNMT1 knock down, we utilized the Mercury Pathway Profiling System (Clontech). This system consists of a set of reporter constructs containing distinct enhancer elements upstream from the firefly luciferase gene. The description of the enhancer elements used in our study is shown in Fig. 7A. To further determine whether DNMT1 knock down acts through an HDAC-independent pathway, we compared the pathway profile induced by DNMT1 knock down to the profile elicited by HDAC inhibition. 24 h after transient transfection of A549 cells with each of these reporter vectors, we treated cells with 120 nM MG88 or MG208 for 24 h, or with 1 M TSA for 6 h. We then determined the luciferase activity in these cells and found that the profiles of signal transduction pathway activation by MG88 and TSA were different. HDAC inhibition resulted in a general induction of 9 out of the 11 pathways tested, which is consistent with its general effects on histone acetylation. On the other hand out of 11 pathways tested, 4 were induced by MG88 treatment by at least 2-fold (Fig. 7B). These different profiles of responses to the two treatments are consistent with our results showing that MG88 induces unmethylated promoters independent of histone acetylation (Fig. 3). Two of the response elements activated by DNMT1 knock down, NF-B and AP-1, were expected, because our gene array analysis revealed the induction of transcription factors binding to these elements, NF-B and c-Jun, respectively (28). The induction of the heat shock elements (HSEs) was strong, and this is also in accordance with our gene array data showing that DNMT1 knock down induces stress response genes (28). However, one transcription factor that consistently stood out as the highest responder to DNMT1 knock down was Sp1. Based on our analysis of p21 and BIK promoters, which revealed the presence of multiple Sp1 elements (Fig. 4A), and based on our deletion analysis of the p21 promoter, we conclude that these elements are primarily responsible for mediating the effects of DNMT1 knock down. Interestingly, the Sp1 family of proteins contains zinc finger motifs that primarily bind to the CG-rich cis-elements, and it is well established that CG-rich promoters are silenced by DNA methylation. One excellent example of this silencing is the tumor suppressor p16. It has also been shown that binding of the methylated DNA-binding protein MeCP2 to methylated CG-rich promoters inhibits transactivation by Sp1 (38). Also, methyl-binding domain protein 1 can repress transactivation of a methylated promoter by Sp1 through its interactions with methyl-binding domain protein 1-containing chromatin-associated factor (39). Methylation of CGs adjacent to Sp1/Sp3 binding elements inhibits Sp1 binding and the activity of the p21 promoter (40). It is therefore important to note that all the reporters used in this profiling experiment were unmethylated, because they were all amplified in Escherichia coli, which does not express a CG DNA methyltransferase. We have also previously demonstrated that plasmids do not undergo de novo methylation in the cell (21). Thus, our assay detects exclusively the methylation-independent effects of DNMT1 knock down. The fact that DNMT1 regulates Sp1 responsiveness in the absence of DNA methylation might explain how DNMT1 regulates the expression of multiple housekeeping genes rich in unmethylated CG sites.
MG88 Activates Sp1 Firing by a Mechanism, Which Is Independent of Histone Acetylation-It has been previously demonstrated that induction of p21 by TSA requires the presence of FIG. 7. Profile of the transcription factor pathways affected by TSA and DNMT1 antisense knock down. A549 cells were transfected with 0.5 g of luciferase reporter vectors containing cis-acting enhancer elements upstream from the TATA box and the firefly luciferase gene. The description of the enhancer elements is shown in A. 24 h after the transfection of the luciferase reporters, the cells were treated with either 120 nM MG88 and MG208 for 24 h (B) or with 1 M TSA for 6 h (C). Luciferase activity was measured and normalized as described under "Materials and Methods." The induction of luciferase activity in MG88-and TSA-treated cells is shown as a percentage of MG208treated or control cells, respectively. Error bars are standard deviations of triplicates. Sp1 response elements (37). We examined whether the induction of 4 ϫ Sp1 reporter vector by DNMT1 knock down is acetylation-dependent or -independent, like the induction of endogenous p21 and BIK genes (Fig. 4). A549 cells were transiently transfected with 4 ϫ Sp1 construct, and the cells were treated with either MG88 or MG208 for 24 h. The cells were then subjected to a ChIP assay using antibodies against acetylated H3 or acetylated H4 histones, followed by PCR amplification of the region containing four tandem Sp1 copies of this luciferase reporter. The results indicate that MG88 treatment does not cause a change in acetylation of either H3 or H4 histones associated with the transiently transfected 4 ϫ Sp1 reporter construct, similar to the endogenous p21 and BIK promoters (Figs. 4 and 8). On the other hand, HDAC inhibitor TSA, which induces 4 ϫ Sp1 reporter construct to the similar extent as the MG88 treatment (6-and 8-fold, respectively, Fig.  7), produces a marked induction of acetylation of both histones H3 and H4 associated with this reporter (Fig. 8). Thus, Sp1 element firing could be induced by either inhibition of histone deacetylation or by knock down of DNMT1 through independent mechanisms.
Knock down of DNMT1 Does Not Increase the Occupancy of Either p21 or BIK or the 4 ϫ Sp1 Reporter Gene Promoter with Sp1/Sp3-One possible explanation for the increase in Sp1 promoter firing is that DNMT1 knock down results in an increase in either Sp1 mRNA transcription or Sp1 binding activity. The microarray gene expression analysis did not reveal an induction of any of the Sp1-binding proteins upon MG88 treatment (28). It was previously reported that Sp1 binding activity is increased after treating cells with a DNA methylation inhibitor 5-aza-2Ј-deoxycytidine (DAC) (41). We performed gel shift assays but did not detect any increase in either Sp1 or Sp3 binding to Sp1 recognition elements upon MG88 treatment (data not shown).
We then tested whether Sp1 or Sp3 bind to p21, BIK, and the 4 ϫ Sp1 promoter, which is expected if DNMT1 knock down activates these promoters through Sp1 elements. Because it was previously shown that Sp3 binding might negatively affect promoters regulated by Sp1 (42,43), we determined the occupancy of the p21 promoter with both transcription factors. ChIP assays using either anti-Sp1 or anti-Sp3 antibodies was performed on either MG88-or MG208-treated A549 cells, followed by PCR amplification of p21, BIK, and 4 ϫ Sp1 promoters. The results in Fig. 7B show that both Sp1 and Sp3 interact with these promoters in A549 cells, which is consistent with either Sp1 or Sp3 mediating the effects of DNMT1 knock down. However, DNMT1 knock down does not change the occupancy of these promoters by either Sp1 or Sp3. In summary, DNMT1 knock down induces the promoters containing Sp1 recognition elements by a mechanism that does not involve increasing the occupancy of these promoters by Sp1 or Sp3. DNMT1 regulates CG-rich Sp1-containing promoters by a mechanism that does not involve DNA methylation.
DNMT1 Knock Down Induces Heat Shock 70-kDa Protein 2 by a Histone Acetylation-independent Mechanism-Our pathway profiling revealed that MG88 and TSA induce the firing of a promoter bearing the heat shock element (HSE) (Fig. 7). HSEs are found within the promoters of heat shock proteins (HSPs), and they bind heat shock factors resulting in gene activation (44). HSPs are a family of molecular chaperons that are involved in response to heat shock and a range of other cellular insults and might be involved in the epigenomic stress response, which is launched by DNMT1 knock down (28). To validate that the induction of firing of HSEs by DNMT1 knock down applies also to an endogenous HSP gene, we determined whether the gene encoding the nodal heat shock 70-kDa pro-tein 2 (HSPA2) is induced by this treatment. The results presented in Fig. 9A illustrate that DNMT1 knock down induces HSPA2 3-fold. Induction of HSPA2 by DNMT1 knock down was also observed in the microarray gene expression analysis (data not shown).
We then addressed the question whether the acetylationindependent effect that DNMT1 knock down has on Sp1 elements is specific to these elements, or whether it applies to other nodal transcriptional regulatory pathways such as the heat shock response. We first demonstrated that the HSPA2 gene is partially suppressed by histone deacetylation in our system, as indicated by its induction with the HDAC inhibitor TSA (Fig. 9B). It has been previously shown that TSA induces Hsp70 gene expression in Drosophila (45). As expected, ChIP analysis of the HSPA2 promoter on A549 cells treated with 1 M TSA for 6 h resulted in an increase in the state of acetylation of both H3 and H4 histones associated with this promoter. However, as is the case with Sp1 elements, DNMT1 knock down did not change the state of acetylation of HSPA2 pro-  Fig. 4A). For the amplification of 4 ϫ Sp1 luciferase reporter construct, the primers were designed within the pGL2 TϩI vector on both sides of tandem Sp1 copies (Fig. 5B). The enrichment in specific chromatin is calculated after densitometric quantification of PCR products and normalization to the input DNA. The induction in MG88-and TSA-treated cells is shown as a percentage of MG208-treated and control cells, respectively. The error bars are standard deviations of the triplicates. There was no PCR amplification in any of the antibody control immunoprecipitations (data not shown). moter (Fig. 10B). To test whether these effects on histone acetylation are carried through the gene downstream of the enhancer elements, we analyzed a region of the second exon of HSPA2 gene. We found that, unlike TSA treatment, which produces an increase in acetylation of both histones H3 and H4, DNMT1 knock down does not cause any change in histone acetylation associated with exon 2 (Fig. 10C).
In summary, DNMT1 knock down results in induction of two nodal transcriptional regulatory pathways in the cell by a mechanism that is surprisingly independent of both DNA methylation and histone acetylation. We propose a new function of DNMT1 that is independent of both its methyltransferase activity and its histone deacetylase-recruiting activity. DISCUSSION This report addresses the question of whether DNMT1 regulates gene expression by a DNA methylation independent pathway and whether this regulation is mediated by the state of chromatin modification through histone acetylation and methylation. We took advantage of well characterized second generation DNMT1 antisense ODNs (MG88) and their mismatch controls (MG208). We first show that our ODNs achieve a complete knock down of DNMT1 protein 24 h after initiation of treatment. In the same time frame, DNMT1 knock down induces the expression of critical cell-cycle regulatory genes such as p21 and BIK. The regulation of expression of either p21 or BIK does not involve a change in their DNA methylation state (Fig. 3.) and previous data (21). Moreover, DNMT1 knock down results in activation of transiently transfected unmethylated reporter genes directed by tandem copies of distinct transcription factor recognition elements. Because it was previously shown that DNMT1 interacts with histone-modifying enzymes (22,24), we tested the hypothesis that DNMT1 knock down effects are mediated by a change in histone acetylation. Both BIK and p21 are partially suppressed by HDACs in our system, because the inhibition of HDACs by TSA increases the state of acetylation of their promoters and induces their expression. Nevertheless, we demonstrate that DNMT1 regulates the expression of these genes without changing their state of acetylation (Fig. 4). Our data therefore unravel a third novel mechanism of regulation of gene expression by DNMT1 that does not involve either DNA methylation or histone modification, the two most established fundamental principles of epigenomic regulation.
Knock down of DNMT1 results in induction of many genes, which includes a class of genes encoding stress response proteins (28) suggesting that DNMT1 acts on a common transcriptional regulatory pathway. Our analysis of the p21 promoter revealed that the Sp1 element corresponding to the bases between Ϫ71 and Ϫ86 mediates the effects of DNMT1 knock down, because four copies of this site (4 ϫ Sp1) were sufficient to mediate the induction of the luciferase reporter produced by DNMT1 knock down (Fig. 5B). In addition, 2-bp mutation (CC3 GA) of this Sp1 site in the full-length p21 promoter abolished Sp1 binding and the induction by DNMT1 knock down. p21 promoter responds to both antisense ODN as well as a plasmid expressed DNMT1 antisense mRNA, demonstrating that this induction is not a consequence of nonspecific effects of the modified ODNs (Figs. 5 and 6). We further tested a number of basic cis-acting elements that are known to respond to major regulatory pathways in the cell. Among these, E2F elements were previously shown to play an important role in regulating p21 gene expression (40 -43). Our data exclude the possibility that DNMT1 acts on a pathway triggering E2Fs, because the E2F recognition element was unresponsive. This is surprising, because DNMT1 was shown to suppress exogenous E2F-responsive elements by recruiting HDACs into E2F1⅐Rb⅐DNMT1 complex (25). Our data, showing that histone deacetylation is not changed by DNMT1 knock down, further support our conclusion that E2F elements are not involved. Sp1 therefore appears to be the main transcription regulatory element triggered by DNMT1 knock down.
Induction of Sp1 response by DNMT1 knock down is not mediated by increased Sp1 expression or increased occupancy of the p21 and BIK promoters with either Sp1 or Sp3 (Fig. 8). DNMT1 must therefore be acting either directly or indirectly on proteins interacting with Sp1 or Sp3 and modulating their trans-activation activity. Identifying these proteins will re-quire extensive future experiments. Sp1 protein contains zinc finger motifs that primarily bind to the GC-rich cis-elements that are widely distributed in the promoters, enhancers, and locus-control regions of housekeeping genes and some cellspecific genes (46). It was generally accepted that these genes escape regulation by DNMT1, because their promoters are especially enriched in unmethylated CG sites and are usually not methylated. The fact that DNMT1 regulates Sp1 responsiveness in the absence of DNA methylation might explain how DNMT1 regulates the expression of multiple house keeping cell cycle regulatory genes (28). Whereas ectopic DNA methylation is known to suppress Sp1 responsive genes by attracting methylated DNA-binding proteins that suppress Sp1 activation (38 -40), our data suggest an entirely distinctive mechanism through which DNMT1 suppresses Sp1-responsive genes.
The observation that DNMT1 regulates Sp1-responsive genes suggests that it may play a role in the regulation of the cell cycle. Sp1 elements are found in genes that perform contradictory cellular roles such as, thymidine kinase, which FIG. 10. Histone acetylation of the promoter and exon regions of HSPA2 gene is induced by TSA treatment but remains unchanged after DNMT1 antisense knock down. A, the structure of the HSPA2 gene is shown, indicating the primers and the sizes of regions that they amplify. B and C, A549 cells were treated with 120 nM of either antisense MG88 or its mismatch control MG208 for 24 h, or with 1 M TSA or control cells for 6 h. Chromatin was isolated and immunoprecipitated with antiacetylated H3 and H4 antibodies. Associated DNA was used to amplify the promoter (B) and exon 2 (C) regions of HSPA2. The enrichment in specific chromatin is calculated after densitometric quantification of PCR products and normalization to the input DNA. The induction in MG88-and TSA-treated cells is shown as a percentage of MG208 treated and control cells, respectively. The error bars are standard deviations of the triplicates. There was no PCR amplification in any of the antibody control immunoprecipitations (data not shown).
stimulates the cell cycle, and p21, which causes cell cycle arrest (47)(48)(49). Because these two classes of genes are expressed at different phases of the cell cycle, there must be regulatory factors that switch their responsiveness to Sp1. Such factors are responsible for coordinating cell arrest and cell division during the different stages of the cell cycle. It stands to reason that DNMT1 acts on a regulatory pathway, which inhibits Sp1 activation of a distinct subclass of Sp1regulated genes involved in cell cycle arrest. DNMT1 is therefore proposed to play a role in orchestrating the cell program toward DNA synthesis and cell division. Such a mechanism possibly evolved to coordinate DNMT1 expression and the inheritance of the DNA methylation pattern with the cell cycle. Unraveling of the Sp1 regulatory pathway that DNMT1 is acting upon is therefore extremely important for our understanding of the cell cycle regulation.
The fact that DNMT1 can act on unmethylated elements found in most housekeeping genes obliges us to revisit the biological roles of DNMT1 and its possible roles in cancer (29,50). The common thinking is that DNA methylation is the main function of DNMT1 and that its downstream effectors are methylated genes. Although it was previously shown that DNMT1 could interact with HDACs, it was still believed that the main role of such an interaction is to enhance epigenomic silencing through a mechanism that involves both histone deacetylation and DNA methylation. Our data suggests that DNMT1 acts on genes independent of these common mechanisms of epigenomic regulation, DNA methylation, and histone acetylation.
If DNMT1 functions through methylation-independent mechanisms, we must be cautious in our interpretation of the effects of DNMT1 inhibition, knock out, and knock down data. It is possible that many of the effects seen in either normal or cancer cells should be attributed to methylation-independent roles of DNMT1 rather than the loss of DNA methylation. This has obvious therapeutic implications. There are currently several attempts to develop DNA methylation inhibitors as anticancer agents. However, inhibition of DNA methylation has been also shown to induce genes that promote metastasis (51)(52)(53). Hypomethylation was also shown to promote T cell lymphoma in mice hypomorphic for DNMT1 (54). However, if DNMT1 is involved in transformation through methylation-independent mechanisms, then it stands to reason that these functions of DNMT1 should be targeted. Such an approach would circumvent the unwanted effects of inhibition of DNA methylation (29,50). The fact that DNMT1 and histone deacetylase inhibitors act on Sp1-bearing sequences by an independent mechanism suggests that a combination of DNMT1 knock down and histone deacetylase inhibition would have an additive effect on genes involved in the cell cycle arrest. It was previously shown that DNMT1 inhibitors and histone deacetylase inhibitors have synergistic effects on activation of methylated genes. Our data suggest that a combination of TSA and DNMT1 inhibitors might be advantageous for inducing the DNA-methylation independent effects of DNMT1 knock down as well.