Role of DNA Methyltransferases in Regulation of Human Ribosomal RNA Gene Transcription*

We have previously demonstrated that the expression of human ribosomal RNA genes (rDNA) in normal and cancer cells is differentially regulated by methylation of the promoter CpG islands. Furthermore, we showed that the methyl CpG-binding protein MBD2 plays a selective role in the methylation-mediated block in rDNA expression. Here, we analyzed the role of three functional mammalian DNA methyltransferases (DNMTs) in regulating the rDNA promoters activity. Immunofluorescence analysis and biochemical fractionation showed that all three DNMTs (DNMT1, DNMT3A, and DNMT3B) are associated with the inactive rDNA in the nucleolus. Although DNMTs associate with both methylated and unmethylated rDNA promoters, DNMT1 preferentially associates with the methylated genes. The rDNA primary transcript level was significantly elevated in DNMT1-/- or DNMT3B-/- human colon carcinoma (HCT116) cells. Southern blot analysis demonstrated a moderate level of rDNA promoter hypomethylation in DNMT1-/- cells and a dramatic loss of rDNA promoter methylation in double knockout cells. Transient overexpression of DNMT1 or DNMT3B suppressed the luciferase expression from both methylated and unmethylated pHrD-IRES-Luc, a reporter plasmid where the rDNA promoter drives luciferase expression. DNMT1-mediated suppression of the unmethylated promoter involves de novo methylation of the promoter, whereas histone deacetylase 2 cooperates with DNMT1 to inhibit the methylated rDNA promoter. Unlike DNMT1, both the wild type and catalytically inactive DNMT3B mutant can suppress rDNA promoter irrespective of its methylation status. DNMT3B-mediated suppression of the rDNA promoter also involves histone deacetylation. Treatment of HCT116 cells with Decitabine (a DNMT inhibitor) or trichostatin A (a histone deacetylase inhibitor) up-regulated endogenous rDNA expression. These inhibitors synergistically activated methylated pHrD-IRES-Luc, whereas they exhibited additive effects on the unmethylated promoter. These results demonstrate localization of DNMTs with the inactive rDNA in the nucleolus, the specific role of DNMT1 and DNMT3B in rDNA expression and the differential regulation of rDNA expression from the methylated and unmethylated rDNA promoters.

We have previously demonstrated that the expression of human ribosomal RNA genes (rDNA) in normal and cancer cells is differentially regulated by methylation of the promoter CpG islands. Furthermore, we showed that the methyl CpG-binding protein MBD2 plays a selective role in the methylation-mediated block in rDNA expression. Here, we analyzed the role of three functional mammalian DNA methyltransferases (DNMTs) in regulating the rDNA promoters activity. Immunofluorescence analysis and biochemical fractionation showed that all three DNMTs (DNMT1, DNMT3A, and DNMT3B) are associated with the inactive rDNA in the nucleolus. Although DNMTs associate with both methylated and unmethylated rDNA promoters, DNMT1 preferentially associates with the methylated genes. The rDNA primary transcript level was significantly elevated in DNMT1 ؊/؊ or DNMT3B ؊/؊ human colon carcinoma (HCT116) cells. Southern blot analysis demonstrated a moderate level of rDNA promoter hypomethylation in DNMT1 ؊/؊ cells and a dramatic loss of rDNA promoter methylation in double knockout cells. Transient overexpression of DNMT1 or DNMT3B suppressed the luciferase expression from both methylated and unmethylated pHrD-IRES-Luc, a reporter plasmid where the rDNA promoter drives luciferase expression. DNMT1-mediated suppression of the unmethylated promoter involves de novo methylation of the promoter, whereas histone deacetylase 2 cooperates with DNMT1 to inhibit the methylated rDNA promoter. Unlike DNMT1, both the wild type and catalytically inactive DNMT3B mutant can suppress rDNA promoter irrespective of its methylation status. DNMT3B-mediated suppression of the rDNA promoter also involves histone deacetylation. Treatment of HCT116 cells with Decitabine (a DNMT inhibitor) or trichostatin A (a histone deacetylase inhibitor) up-regulated endogenous rDNA expression. These inhibitors synergistically activated methylated pHrD-IRES-Luc, whereas they exhibited additive effects on the unmethylated promoter. These results demonstrate localization of DNMTs with the inactive rDNA in the nucleolus, the specific role of DNMT1 and DNMT3B in rDNA expression and the differential regulation of rDNA expression from the methylated and unmethylated rDNA promoters.
DNA methylation, histone modifications, and chromatin remodeling mediate epigenetic regulation of gene expression (for review, see Refs. [1][2][3][4][5][6][7]. Most studies on this unique process have focused on genes transcribed by RNA polymerase II (pol II). 4 Recent studies from a few laboratories including our own have shown that epigenetic mechanisms also regulate RNA polymerase I (pol I)-directed ribosomal RNA gene (rDNA) expression (8 -10). Only a fraction of the multiple copies of these genes is, however, transcriptionally active at any given time (11). The promoters of active ribosomal RNA genes are devoid of CpG methylation and associate with acetylated histones (12)(13)(14), whereas epigenetically suppressed genes have a reverse pattern (9,15). These studies suggest that epigenetic mechanisms can modify the rDNA chromatin structure and control the ratio of active to silent genes (8).
Studies on the methylation profiles of rDNA in rodents and humans have resulted in unique observations. The human gene contains 19 CpGs in the upstream promoter element and 6 CpGs in the core promoter region, whereas the mouse and rat promoters contain only one and five CpGs, respectively (9). Unlike numerous pol II-directed genes silenced in response to methylation of short CpG regions, designated CpG islands (CGI); methylation of the single CpG located at Ϫ133 (with respect to initiation site) suppresses mouse DNA expression. Methylation at this site inhibits access of the key transcription factor UBF to the upstream control region of the mouse promoter when packaged into nucleosomes. Although human rDNA promoter methylated at a single site can significantly impede promoter activity when transfected into human cells, methylation of multiple sites in the promoter region resulted in complete inhibition of the promoter activity. This observation suggests an inverse relation between promoter activity and the density of methylation (9). Furthermore, analysis of the methylation profile of human hepatocellular carcinomas and matching normal liver tissue by bisulfite genomic sequencing showed significant hypomethylation of the rDNA promoter in tumors compared with the corresponding matching normal tissues. This is consistent with the relatively high level of ribosomal RNA (rRNA) synthetic activity of the rapidly proliferating tumor tissue (9). * This work was supported in part by National Institutes of Health Grants ES10874 and CA86978. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. 1 Both authors contributed equally to this work. 2 To whom correspondence may be addressed. Although the factors involved in the epigenetic regulation of pol II-directed genes have been well studied, such an approach has not been fully used in deciphering their role in pol I-directed ribosomal gene expression. The existence of CGI in the human rDNA promoter compared with only a few CpGs in the rodents (9), particularly in the mouse promoter (16), suggests distinct mechanism of transcriptional regulation in the two systems. Methylation at C-5 of CpG by DNA methyltransferases (DNMTs) results in recruitment of proteins, designated MBDs (methyl CpG domain-binding proteins), followed by histone modifications and association of distinct chromatin remodeling factors (17,18). We have shown specific association of one of the MBDs, namely MBD2, with the endogenous methylated human rDNA promoter and suppression of this promoter by MBD following transfection (9).
Three distinct DNMTs, namely DNMT1, -3A, and -3B encoded by different genes direct DNA methylation in mammalian cells (19,20). DNMT1 generally utilizes hemimethylated DNA as the substrate and is involved in maintenance methylation. Recent studies have associated DNMT1 with methylation of unmethylated human CGIs in cancer cells (21). DNMT3A and -3B participate predominantly in the de novo methylation of unmethylated DNA (22). All DNMTs share a common catalytic domain in the carboxyl terminus, but the NH 2 -terminal domain differs significantly between DNMT1 and DNMT3A/3B. The unique NH 2 -terminal domains of mammalian DNMTs harbor several regulatory domains that mediate both protein-protein and protein-DNA interactions (for review, see Refs. 19 and 23). Relatively large NH 2 -terminal domains of DNMTs can also mediate transcriptional repression of genes independent of their methyltransferase activity (24 -27). The NH 2 -terminal domains of DNMT3A and -3B exhibit some homology but are distinct from that of DNMT1 (19). Here, we focused on the role of DNA methyltransferases in regulating rDNA promoter activity in both the methylated as well as the unmethylated state, and the involvement of the transcriptional repressor domains of these enzymes in suppressing rDNA promoter.

Construction of Plasmids
Human rDNA-Luciferase Vector (pHrD-IRES-Luc)-Construction of plasmid pHrD-IRES-Luc has been described earlier (9). In brief, human rDNA promoter spanning Ϫ410 to ϩ81 bp (accession number K01105) with respect to the transcription initiation site was amplified from ϳ2-kb fragments of HeLa genomic DNA digested with EcoRI and cloned into pGL3-IRES (9) to generate pHrD-IRES-Luc.
Expression Vectors for Mouse DNMT1, -3A, and -3B-These plasmids were constructed in mammalian expression vector pcDNA3.1(ϩ/Ϫ) (Invitrogen) to obtain pc-DNMT1, pc-DNMT3A, and pc-DNMT3B. Briefly mouse DNMT1 cDNA in pBluescript SK(ϩ) (a generous gift from Dr. Tim Bestor) was digested with EcoRI and the resultant ϳ5.2-kb fragment was cloned into the same site in pcDNA3.1(ϩ) to generate pcD-NMT1. PCR with the vector-and insert-specific primers determined the correct orientation of DNMT1. To generate DNMT1/⌬CAT, a 1.5-kb fragment from the COOH-terminal of DNMT1 was excised and the rest of the plasmid was religated. The DNMT1/⌬NLS plasmid is described earlier (28). Similarly, mouse DNMT3A cDNA from pSX137 (a generous gift from En Li) was excised with BamHI and XbaI and cloned in the same sites of pcDNA3.1(ϩ) vector. The mouse DNMT3B was subcloned from the antisense clone (26) in pcDNA3.1(Ϫ), to generate the sense clone.

Isolation of Nucleoli
Nucleoli were isolated from HeLa cells following the protocol of Muramatsu and co-workers (29).

Western Blot Analysis
Whole cell extract (200 g) prepared from HeLa cells overexpressing different DNMTs was separated on a 7.5% SDSpolyacrylamide gel, transferred to nitrocellulose membrane, and blocked in 3% milk in Tris-buffered (pH 7.5) saline. For the detection of the overexpressed proteins the membrane was subjected to immunoblot analysis with anti-FLAG M2 antibody (Sigma) (30,31) and horseradish peroxidase-conjugated antimouse IgG as the secondary antibody. The antigen-antibody complex was detected using ECL TM kit (Amersham Biosciences), following the manufacturer's protocol. Nuclear and nucleolar fractions (250 g) prepared from HeLa cells were separated on a 7.5% SDS-polyacrylamide gel, transferred to nylon membrane and probed with antibodies against nucleolin, RNA polymerase II (both from Santa Cruz), DNMT1 (New England Biolabs), DNMT3A and -3B (antibodies raised in our laboratory).

Chromatin Immunoprecipitation
Formaldehyde cross-linked chromatin was prepared as described (32). For chromatin immunoprecipitation (ChIP) analysis, antisera against UBF (Santa Cruz), DNMT1 (New England Biolabs), DNMT3A and -3B raised in our laboratory, were used (31). The chromatin was first pre-cleared with preimmune sera coupled to protein A/G beads followed by overnight incubation with preimmune or immune sera. The immune complex was then captured by protein A beads, and washed successively with buffers as described (31,32). Immunoprecipitated DNA-protein complex was eluted, un-crosslinked, treated with RNase A and proteinase K, and purified as described (31). ChIP and the input DNA were digested with HpaII or MspI and the digests along with an equal amount of the undigested immunoprecipitated DNA were subjected to real time PCR with human rDNA promoter-specific primers (9). The results are expressed as the ratio of methylated or unmethylated DNA precipitated with the antibodies to the respective input.

DNA Isolation and Southern Blot Analysis
Genomic DNA isolation, HpaII/MspI digestion, and Southern blot hybridization were performed as described (26). For quantitative analysis of the Southern blot data, rDNA promoter was amplified from undigested and HpaII/MspI-digested genomic DNA. GAPDH promoter and albumin promoter were amplified as control for complete restriction enzyme digestion and equal input DNA, respectively. The primer sequences for rDNA and GAPDH promoters are as mentioned (9), and primer sequences for albumin promoter are as follows: hALB-PF, 5Ј-TCCATTTTCCTCTCCATCTCTGC-3Ј; hALB-PR, 5Ј-ATGTTCCCATTCCTGCTGTGGC-3Ј, with an annealing temperature of 52.5°C.

RNA Isolation and Northern Blot Analysis
Total RNA was isolated using TRIzol (Invitrogen) following the manufacturer's protocol. Total RNA (10 g) was separated on a formaldehyde-agarose (1.0%) gel and transferred to nylon membrane. The membrane was then hybridized to human random-primed [␣-32 P]dCTP-labeled pre-47 S rRNA or 18 S rRNA probe in rapid hybridization buffer (GE Healthcare) following the manufacturer's protocol.

Reverse Transcription and Real Time RT-PCR Analysis
Reverse transcription was carried out with random hexamers and Moloney murine leukemia virus reverse transcriptase from 3 g of total RNA in 20 l of total volume following the manufacturer's protocol. An aliquot of the cDNA (equivalent to 100 ng of RNA for 47 S rRNA, and 10 pg for 18 S rRNA) was used for real time PCR analysis. All real time PCR were carried out using the Mx3000 Multiplex Quantitative PCR System (Stratagene). The optimum primer concentration was 150 nM. All PCR amplifications were performed using Brilliant SYBR Green QPCR Master Mix (Stratagene) with ROX as a reference dye in a 10-l reaction volume. A standard curve for each cDNA was first generated using 10-fold serial dilutions (10 8 to 10 2 copies) of the respective cDNAs as template. The copy number of each cDNA expressed was calculated from the standard curve and normalized to that of 18 S rRNA. PCR cycling conditions were as follows: initial denaturation at 95°C for 10 min, followed by 45 cycles of 95°C for 30 s, the annealing temperature for the specific primer pair, for 30 s, and a dissociation cycle of 95°C for 60 s, 55°C for 30 s (to check the formation of primer dimer). Dissociation profile of the amplified products indicated that none of the primer pairs generated dimers. The primer sequences are the following: 47 S (product size 92 bp): 47S-F, 5Ј-CCTGTCGTCGGAGAGGTTGG-3Ј, 47S-R, 5Ј-ACCCCA-CGCCTTCCCACAC-3Ј, an annealing temperature of 60°C; 18 S rRNA (product size: 492 bp): 18 S rRNA-F, 5Ј-TCAAGAAC-GAAAGTCGGAGG-3Ј, 18 S rRNA-R, 5Ј-GGACATCTAAG-GGCATCACA-3Ј, an annealing temperature of 57°C.

DNMT1 and DNMT3B Synergistically Maintain the Methylation Profile of the Human rDNA Promoter and Regulate Its
Expression-To identify the DNMTs involved in methylation and regulation of rDNA we took advantage of the human colon cancer cell line HCT116 in which homologous recombination disrupted both alleles of DNMT1 (DNMT Ϫ/Ϫ or MT1KO cells), DNMT3B (DNMT3B Ϫ/Ϫ or 3BKO cells), or both alleles of DNMT1/DNMT3B (DKO cells, double knockout) (33). Western blot analysis confirmed the absence of the respective proteins in null cells, whereas the third functional enzyme DNMT3A was detectable in all three cell lines (Fig. 1A).
Next we addressed whether any alteration in methylation status of the rDNA promoter occurred in DNMT1 or DNMT3B null cells. HpaII (methylation sensitive) or MspI (methylation insensitive)-digested genomic DNA from the wild type and knockout HCT116 cells were subjected to Southern blot analysis with the rDNA promoter-specific probe. The probe harbors 4 HpaII/MspI sites (Fig. 1B). The DKO cell genome exhibited significantly reduced methylation compared with MT1KO or 3BKO as evident from ethidium bromide staining of HpaIIdigested DNA (Fig. 1C, compare lanes 2, 5, 8, and 11), and is consistent with the earlier report (33). Extensive hybridization of the rDNA probe to low molecular weight DNA fragments in HpaII-digested DNA in DKO cells showed dramatic hypomethylation of rDNA in these cells compared with the parental cells (Fig. 1D, lanes 2 and 11). In contrast, small but significant hypomethylation of the rDNA promoter was observed in MT1 KO cells (Fig. 1D, lanes 2 and 5), whereas 3BKO cells did not reveal any noticeable hypomethylation (Fig. 1D, lanes 2 and 8). Comparable hybridization profile of DNA digested with methylation-insensitive isoschizomeric MspI in all four cells indicate that different profiles in the HpaII digest are because of differential methylation of the promoter (Fig. 1D, lanes 3, 6, 9, and 12). Significantly reduced rDNA promoter methylation in DKO cells relative to other cell lines suggests that both DNMT1 and DNMT3B are essential to maintain its methylation profile. For quantitative analysis of this co-operative effect, rDNA promoter was PCR amplified from undigested (U), HpaII-digested (H), and MspI-digested (M) genomic DNA using rDNA specific primers (Fig. 1B). This data showed (Fig. 1F, lanes 1 and 2) that 28% of the rDNA promoter in HCT116 cells is unmethylated. DNMT1 disruption resulted in an additional 22% demethylation (Fig. 1F, compare lanes 2 and 5). DNMT3B deletion alone resulted in an additional 10% demethylation (compare lanes 2 and 8). In DKO cells 76% of rDNA promoter is demethylated, demonstrating an additional 48% demethylation (76 -28%) in these cell lines following disruption of DNMT1 and DNMT3B. To confirm complete restriction enzyme digestion, the GAPDH promoter harboring 2 HpaII/MspI sites was amplified from the same digest (9). GAPDH was amplified only from the undigested DNA from each cell line. Albumin promoter fragment that does not span any HpaII/MspI site was amplified equally well from each DNA confirming a comparable amount of DNA in PCR (Fig. 1E). This data further authenticates the Southern blot analysis showing synergism between DNMT1 and -3B in maintaining the rDNA promoter methylation profile.
We then determined whether DNMT gene disruption alone or in combination had any effect on rDNA expression in HCT116 cells. Northern blot analysis of the total RNA with a probe complimentary to the 47 S precursor rRNA, the primary transcript, revealed at least a 2-fold increase in its level in both MT1KO and 3BKO cells compared with the parental or DKO cells (Fig.  1G). Quantitative analysis of the rRNA transcript by real time RT-PCR showed a 2-and 3.4-fold increase in the primary transcript in MT1KO and 3BKO cells, respectively (Fig. 1H). A significant increase in precursor rRNA level in MT1KO cells correlated with hypomethylation of its promoter (Fig. 1, D and E). Striking elevation of the rRNA transcript in 3BKO cells suggests that DNMT3B regulates rDNA transcription by a methylation independent mechanism, because the rDNA promoter did not undergo significant hypomethylation in 3BKO cells (Fig. 1, D and E). No significant change in the 47 S rRNA level in DKO (Fig. 1H) cells despite remarkable hypomethylation of the promoter (Fig. 1, D and E) could be attributed to the detrimental effect of global hypomethylation on growth regulatory genes that significantly compromised cell growth (33). Increased rRNA levels in MT1KO and 3BKO cells implicates their role in regulating its expression in vivo.
DNMT1 and DNMT3B Localize in the Nucleolus and Associate with rDNA Promoter-Immunofluorescence analysis and biochemical fractionation demonstrated localization of DNMTs in the nucleolus ( Fig. 2A). These proteins also co-lo- The digested products were separated on a 0.8% agarose gel and stained with ethidium bromide. D, DNA was transferred to nylon membrane and probed with a 32 P-labled rDNA promoter probe. E, rDNA, GAPDH, and albumin promoters were amplified from mock-digested (U, Uncut), HpaII (H), or MspI (M) digested genomic DNA from W T HCT116, MT1KO, 3BKO, and DKO cells and resolved on 1.5% agarose gel. F, quantitative analysis of the PCR amplified rDNA products. G, total RNA isolated from the cells was resolved on a 1.0% agarose/formaldehyde gel, transferred to nylon membrane, and probed with the 91-bp pre-RNA probe. H, total RNA from the cells was treated with RNase-free DNase I, reverse transcribed, and subjected to real time PCR using 47 S rRNA-specific primers. Data presented was normalized to 18 S rRNA. RNA without reverse transcription did not generate PCR product. calized with nucleolin, a nucleolar marker, because the two stains (fluorescein isothiocyanate and TRITC) merged ( panels e, j, and o). Furthermore, double staining of cells with anti-UBF and anti-nucleolin antibodies (panels p-t) showed UBF localization at the center of the nucleoli, and peripheral nucleolin staining. Similarly, dual staining of cells with DNMT1 and UBF antibodies demonstrated distinct localization of these two proteins within the nucleolus (panels u-y). This is probably due to association of UBF primarily with the active promoter and predominant localization of DNMTs with the inactive rDNA promoter. Furthermore, biochemical fractionation coupled with Western blot analysis demonstrated the presence of DNMT1, -3A, and -3B as well as nucleolin in the purified nucleolar fraction of HeLa cells (Fig. 2B). As expected, DNMTs are also present in the nuclear fraction. The absence of RNA polymerase II in the nucleolar fraction ruled out possible contamination of this fraction with nucleoplasmic proteins (Fig. 2B). These observations support the notion that DNMTs are present in the nucleoli and associated with inactive rDNA.
Mammalian cells harbor ϳ400 copies of rDNA, some of which are methylated in somatic cells (34,35). To determine the nature of association of DNMTs with the rDNA promoter we performed the ChIP-CHOP assay. HeLa cells express all three DNMTs (28). DNA pulled down by specific antibodies as well as the input DNA was digested with HpaII, MspI, or mock digested prior to amplification and analysis (fragment shown in Fig. 1B). This amplified promoter region harbors four HpaII/ MspI sites. DNA pulled down was quantified from the standard curve generated using serial dilution of the input DNA. The HpaII-resistant PCR product reflects the level of the rDNA promoter, methylated at all four sites (M). The PCR product from the undigested DNA minus that from the HpaII-resistant product represents the level of unmethylated promoter (U). PCR product was not generated from MspI-digested DNA because it is a methylation insensitive enzyme. Real time PCR data revealed that all 4 HpaII sites of only 20% of the total input DNA are methylated in HeLa cells. DNMT1 association with the methylated promoter was 30-fold higher than the rest of the promoter (Fig. 2C). On the contrary, DNMT3B was almost equally associated with both unmethylated and methylated rDNA promoters (Fig. 2C, lanes 5 and 6). Like DNMT1, DNMT3A associate predominantly with methylated promoter (lanes 7 and 8). As a control, we performed ChIP assay with anti-UBF antibody, which showed exclusive association of UBF with a large fraction of unmethylated promoter (lanes 9 and 10), demonstrating the stringency of the ChIP assay. Immunoprecipitated chromatin in the absence of the antibody (control) showed minimal amplification of the rDNA promoter ( lanes  1-3). Thus, all three DNMTs associate with the methylated rDNA population, which confirms localization of DNMTs with the inactive rDNA at the perinucleolar heterochromatin ( Fig. 2A).
DNMT1 and DNMT3B Suppress rDNA Promoter Activity-To determine the role of DNMTs in regulating rDNA transcription, pHrD-IRES-Luc and mammalian expression vectors for FLAG-tagged DNMT1, -3A, or -3B were transiently transfected in HeLa cells (9). Western blot analysis demonstrated comparable levels of ectopic DNMT3A and -3B expression, whereas DNMT1 expression was significantly less (Ͻ50%) than that of DNMT3A/3B (Fig. 3A). Expression of DNMT1 or DNMT3B resulted in significant down-regulation of the rDNA promoter activity (50 and 60%, respectively, Fig. 3B). Under this condition, DNMT3A did not exhibit any inhibitory effect on the promoter. Similarly, the HhaI-methylated rDNA promoter was significantly inhibited by DNMT1 and DNMT3B (60 and 72%, respectively, Fig. 3B), whereas DNMT3A did not have any effect.
Methyltransferase Activity of DNMT1 Is Essential for Inhibition of Unmethylated rDNA Promoter-To decipher the mechanism of DNMT1-mediated repression of the rDNA promoter several deletion mutants of DNMT1 were generated (Fig. 4A). DNMT1 is a large protein comprising 1620 amino acids with a large NH 2 -terminal regulatory domain and relatively smaller COOH-terminal catalytic domain (Fig. 4A). The NH 2 -terminal domain harbors a region that interacts with various proteins (24,36), and different targeting motifs (37-39) (Fig. 4A). A bromo homology domain containing two BAH (bromo adjacent homology) motifs is located adjacent to the zinc-binding domain, which is probably involved in protein-protein interaction (for review, see Ref. 40). Recently we have shown that nuclear localization of DNMT1 requires both NLS and BAH domains (28). We generated the catalytic domain deletion FIGURE 2. A, all three DNMTs localize in the nucleolus. HeLa cells were stained with TRITC-tagged mouse monoclonal antibody against nucleolin and with fluorescein isothiocyanate-tagged rabbit polyclonal antibody against DNMT1, -3A, -3B, or UBF (panels a-t). In a separate set, cells were stained with fluorescein isothiocyanate-tagged rabbit polyclonal antibody against UBF and TRITC-tagged mouse monoclonal antibody against DNMT1 (panels u-y). All five sets of cells were also stained with 4Ј,6-diamidino-2-phenylindole and visualized under a fluorescence microscope. B, all three DNMTs co-fractionated with nucleolin in the nucleolar fraction: nucleolar extract and nuclear extract (nucleolus and nucleoplasm) from HeLa cells (250 g) were subjected to Western blot analysis with antibodies against nucleolin, RNA polymerase II, DNMT1, -3A, and -3B. C, DNMT1, -3B, and -3A are associated with methylated rDNA promoter. Formaldehyde cross-linked chromatin was precleared and immunoprecipitated overnight with antisera specific for DNMT1, DNMT3A, DNMT3B, UBF, or preimmune sera. The immune complexes were precipitated by protein A/G beads, washed with different buffers (detailed under "Experimental Procedures"), eluted, and un-crosslinked. DNAs pulled down by different antibodies as well as input DNA were divided into three identical fractions that were either mock digested or digested with HpaII or MspI. An aliquot of each digestion product was subjected to real time PCR with primers specific for rDNA promoter. Association of different DNMTs and UBF with the rDNA promoter was analyzed using a standard curve generated by serial dilution of the undigested input DNA. Association with methylated promoter equals HpaII signal in ChIP DNA/HpaII signal in input (1:300 dilution). Association with unmethylated promoter corresponds to the signal in undigested minus signal in HpaII-digested ChIP DNA/Input signal in undigested minus HpaII-digested DNA (1:300 dilution). U and M indicate methylated and unmethylated rDNA promoters, respectively. mutant (DNMT1/⌬CAT) where ϳ400 amino acids from the COOH-terminal catalytic domain were eliminated. In a second deletion mutant the NH 2 -terminal 350 amino acids were deleted, which removed the nuclear localization signal with proliferating cell nuclear antigen and DMAP1 interacting regions (DNMT1/⌬NLS).
HeLa cells were transfected with mock methylated rDNA promoter along with expression vectors for different DNMT1 deletion mutants. The luciferase assay revealed that only the wild type DNMT1, but not DNMT1/⌬CAT or DNMT1/⌬NLS, was able to impede activity of the mock methylated promoter (Fig. 4B). This data suggests that the catalytic activity of DNMT1 is critical for inhibiting unmethylated promoter activity. Inability of a catalytically inactive point mutant (DNMT1CS) (Fig. 4A) (28) to impede pHrD-IRES-Luc activity reinforces the notion that DNMT1 inhibits the promoter activity by catalyzing its methylation (Fig. 4B). To confirm that the exogenous promoter was indeed methylated by ectopic DNMT1, pHrD-IRES-Luc purified from the transfected cells was digested with HpaII or MspI and subjected to amplification of the rDNA promoter fragment. A 10-fold increase in the HpaII-resistant PCR product from cells expressing ectopic DNMT1 compared with that from the vector-transfected cells suggests significant methylation of the rDNA promoter by DNMT1 (Fig. 4C, lanes 2 and 5). Because MspI cleaved the plasmid irrespective of its methylation status, there was no PCR product obtained after MspI digestion. This data demonstrates that DNMT1-mediated rDNA promoter methylation is necessary to suppress the promoter activity, which is in accordance with increased rDNA transcription concomitant with its promoter hypomethylation in MT1KO cells (Fig. 1, D and E).
DNMT1 Suppresses the Activity of the Methylated rDNA Promoter in the Absence of the Catalytic Domain-We previously demonstrated significant inhibition of the rDNA promoter activity when M.HhaI-methylated CpG dinucleotides were present in the upstream control element and in the further upstream region (9). Interestingly, ectopic expression of DNMT1 further suppressed the methylated promoter (Figs. 3B and 5A). To get mechanistic insight of this process HeLa cells were co-transfected with methylated pHrD-IRES-Luc along with the expression vectors for different DNMT1 deletion mutants. The results showed that DNMT1/ ⌬CAT and DNMT1/CS exhibited a similar repressive effect (52 and 58%) as the wild type DNMT1 (56%, Fig. 5A). No significant repression was, however, noted upon DNMT1/⌬NLS expression, as the NH 2 -terminal truncation retained the protein in the cytoplasm (28). This data reveals the role of the DNMT1 transcriptional repressor domain in impeding methylated rDNA expression.
The NH 2 -terminal domain of DNMT1 interacts with various co-repressors that include histone deacetylase (HDAC) 1/2. To determine whether HDAC plays any role in the DNMT1-mediated suppression of human rDNA gene transcription, HeLa cells were first transfected with methylated pHrD-IRES-Luc with or without DNMT1 expression vector. The cells were then treated with the HDAC inhibitor trichostatin A (TSA). As observed earlier, ectopic expression of DNMT1 inhibited rDNA promoter activity (Fig. 5, A, and B, bottom panel: promoter activity in vector or DNMT1 transfected controls). However, treatment of the cells with 100 nM TSA for 24 h resulted in a ϳ4-fold increase in the promoter activity even in DNMT1 overexpressing cells (Fig. 5B). DNMT1 overexpressing cells showed comparable DNMT1 expression in the presence and absence of TSA (Fig. 5B). This data suggests that DNMT1-me- diated suppression of the rDNA promoter involves HDACs. Next, HDAC1 or HDAC2 were expressed alone or co-expressed with DNMT1 along with methylated pHrD-IRES-Luc in HeLa cells. To determine whether there is any synergy between DNMT1 and HDACs, expression of DNMT1 was titrated to a level where repression of the promoter by DNMT1 alone was Ͻ35%. HDAC2 inhibited the methylated promoter by 25%, whereas in combination with DNMT1 the activity was suppressed by 70% (Fig. 5C). This data suggests that HDAC2 cooperates with DNMT1 to down-regulate the rDNA promoter. On the other hand, HDAC1 did not modulate the effect of DNMT1 on the promoter (Fig. 5C).

DNMT3B-mediated Suppression of rDNA Promoter Does Not Require
Its Catalytic Activity-Disruption of DNMT1, DNMT3B, or both in HCT cells displayed a differential effect on the rDNA transcription and methylation profile of the promoter. The ribosomal RNA transcript level in DNMT3BKO cells was significantly elevated without noticeable demethylation of its promoter (Fig. 1, C and D). To address the underlying mechanism, HeLa cells were transfected with wild type and catalytic site mutants of DNMT3B (CS mutant) along with mock or M.HhaI-methylated pHrD-IRES-Luc. Both the wild type and mutant DNMT3B significantly inhibited the promoter activity irrespective of its methylation status (average 50%, see Fig. 6A). The expression of ectopic wild type and mutant DNMT3B was comparable (Fig. 6A,  lower panel). This data suggests that DNMT3B repress both the endogenous rDNA promoter in the chromatin context (Fig. 1D) and the transfected promoter independent of its catalytic activity.
The NH 2 -terminal domain of the DNMT3B protein does not exhibit any similarity with that of DNMT1 (41). Besides a nuclear localization signal, DNMT3B has a well defined PWWP domain involved in DNA binding (42) and an ATRX homology domain that mediates proteinprotein interactions (25,43) (Fig.  6B). To identify the domain(s) of DNMT3B involved in the transcriptional repression of the rDNA promoter, HeLa cells were co-transfected with pHrD-IRES-Luc and different deletion mutants of mouse DNMT3B (⌬C-terminal, ⌬N-terminal, ⌬PWWP, and ⌬ATRX) (26). Expression of the truncated and wild type proteins was monitored by Western blot analysis (Fig. 6C). As observed earlier, both the wild type and ⌬C-terminal DNMT3B inhibited the unmethylated rDNA promoter to a comparable extent (ϳ50%). However, ⌬PWWP, ⌬ATRX, or ⌬N-DNMT3B could not suppress the promoter activity (Fig.  6D). Similar inhibitory effect of the DNMT3B mutants was also noted with the methylated promoter (data not shown). Thus both PWWP and ATRX domains of DNMT3B are essential to suppress the rDNA promoter.
Because DNMT3B interacts with HDACs through the ATRX domain we investigated whether the HDAC inhibitor can alle- viate its repression on rDNA promoter. Treatment of HeLa cells co-transfected with pHrD-IRES-Luc and the DNMT3B empty vector with 100 nM TSA for 24 h resulted in a ϳ5-fold increase in both mock and methylated-promoter activities (Fig.   6E). The unmethylated and methylated promoter activities increased 3.7-and 4-fold, respectively, when DNMT3B expressing cells were treated with TSA. This data implicates that HDAC(s) is one of the corepressors recruited by DNMT3B to inhibit rDNA promoter activity.
Decitabine and Trichostatin A Synergistically Activate the Methylated rDNA Promoter-Inhibitors of DNMTs and HDACs act synergistically to up-regulate some pol IIdriven promoters that are methylated and silenced (31,44). It was, therefore, of interest to explore whether these two inhibitors exhibit a similar effect on the rDNA promoter. To test this possibility, HCT116 cells were treated with Decitabine (50 nM) alone or a combination of Decitabine and TSA (100 nM). Decitabine at higher dosage displayed a toxic effect on HCT116 cells. Real time RT-PCR analysis demonstrated increased expression of 47 S rRNA in Decitabine-treated cells, which was further up-regulated by TSA exposure (Fig.  7A). A similar result was obtained when HeLa cells were treated with Decitabine (1 M) and TSA (100 nM) alone or in combination (data not shown).
Because the proportion of methylated rDNA promoter varies among 300 -400 copies of the genes, it is difficult to decipher the effect of Decitabine and TSA on methylated versus the unmethylated promoters. The data presented in Fig. 7A, therefore, represents the cumulative effect of the inhibitors on both promoters. To analyze the effect of these two inhibitors on methylated and unmethylated promoters separately, we took advantage of the transient transfection system. Treatment of HeLa cells transfected with the unmethylated promoter resulted in 5.2-fold activation by Decitabine (1 M) and TSA (100 nM), which is essentially an additive effect of Decitabine (1.7fold) and TSA (3.1-fold) treatment alone (Fig. 7B). However, a synergistic activation (9.2-fold) of the promoter was observed following transfection of methylated (versus unmethylated) promoter and treatment with both inhibitors (Fig. 7B). Expo- sure of the methylated promoter separately to Decitabine and TSA resulted in 1.9-and 3.4-fold activation, respectively. This data suggests that the methylated and unmethylated rDNA promoters respond differently to a combination of these two agents.

DISCUSSION
Recent studies (8 -10) have provided important insights into the role of promoter methylation in rDNA expression in Arabidopsis, mouse, and humans. A previous study from our laboratory showed reduced methylation of the rDNA promoter in the primary human hepatocellular carcinoma relative to matching control liver tissue. This is consistent with the elevated levels of rRNA synthesis required for the increased demand of ribosomes in tumors. A comprehensive study on the role of different methyl CpG-binding proteins in rDNA promoter methylation showed that one of the MBDs, MBD2, is associated with the methylated promoter and suppressed its activity (9). The present study explored the role of DNA methyltransferases in the epigenetic regulation of human rRNA synthesis. The presence of DNMTs in the nucleolus and extensive demethylation of the rDNA promoter in DNMT1 and -3B knockout cells is consistent with their role in its methylation. To our knowledge, this is the first demonstration of the localization of DNMTs in the nucleolus in addition to the nucleoplasmic region and their role in rDNA expression. Both DNMT3A and -3B harbor conserved PWWP domains that are involved in localization of these two proteins at pericentric heterochromatin (45). Because the PWWP domain can bind DNA in vitro (42) interaction of DNMT3B with the rDNA promoter might involve the same domain. The present study also revealed that the PWWP domain of DNMT3B along with the ATRX homology domain is involved in DNMT3B-mediated suppression of rDNA promoter activity. The occurrence of these enzymes as well as the methyl C-binding proteins (9) in the nucleolus further attests to functional significance of the DNA methylation machinery in this organelle.
Of the three functional DNMTs in higher organisms (DNMT 1, -3A, and -3B), only DNMT1 and -3B are involved in the methylation mediated suppression of the human rDNA promoter. An interesting observation is the significant synergy between the two proteins in maintaining the CpG methylation of the promoter in colon cancer cell line HCT116. Southern blot analysis showed low level hypomethylation only in MT1KO cells. An increase in the expression of the 47 S rRNA in 3BKO cells without significant alteration in promoter methylation suggests transcriptional repressor function of DNMT3B in rDNA regulation. In this context, a recent study in our laboratory has shown that DNMT3B is involved in the nerve growth factormediated differentiation of PC12 cells and that this process is independent of the catalytic activity of DNMT3B (26). Although ChIP analysis revealed association of DNMT3A with methylated rDNA promoter in HeLa cells; current studies did not reveal its direct involvement in regulating rDNA promoter function. Generation of DNMT3A null cells may address the role of this enzyme in rDNA expression.
The synergistic role of DNMT1 and DNMT3B on rDNA promoter methylation is also consistent with the occurrence of these proteins in a complex following extensive biochemical fractionation (46) as well as the known functional interaction of these two enzymes (33). Although the association of ectopic DNMT1 and -3B with transiently transfected reporter plasmid driven by mouse pol I promoter has been demonstrated earlier (14), its role in rDNA expression has not been studied. The present study has revealed in vivo association of both DNMT1 and -3B with unmethylated as well as methylated rDNA promoters, and suggests that the promoter is methylated first and subsequently suppressed by DNMT1. Increased demethylation of the promoter in MT1KO cells compared with the wild type HCT116 cells further validates this observation. In this context, it should be noted that DNMT1 could also function as de novo cytosine methyltransferase in addition to its role in maintenance methylation of DNA (47). Alternatively, ectopic DNMT1 co-operates with endogenous DNMT3B to initiate methylation of the rDNA promoter.
The effect of HDAC inhibition on rDNA promoter function merits discussion. Previous study has shown that rodent rDNA promoter activity is significantly elevated upon treatment with TSA, the HDAC inhibitor (48). The present study demonstrated a similar effect of TSA on human rDNA promoter acti- vation. Furthermore, combined treatment of cells with the DNA hypomethylating agent Decitabine, a potent inhibitor of DNMT1 (28, 49 -51) and TSA synergistically activated only the transiently transfected methylated rDNA promoter. These two drugs probably exhibit the same synergistic effect on the endogenous methylated promoter. Because mammalian cells harbor multiple copies of rDNA with differential methylation states, it is not possible to distinguish the effect of these drugs on methylated versus unmethylated promoter in vivo. In this context, pol I promoter behaves like pol II promoter where strong synergy between the two agents was observed for several methylated pol II promoters (31,44). We have recently demonstrated that the DNA demethylating agent Decitabine can selectively degrade DNMT1 by a proteasomal pathway (28). The repressor complex that includes DNMT1 and co-repressors such as HDAC may be dislodged following Decitabine-mediated degradation of DNMT1 that could lead to a decrease in HDAC recruitment and the observed synergistic activation of the rDNA promoter.
A unique aspect of epigenetic modulation of rDNA expression is that unlike the mouse system that contains a single CpG in the promoter region, human promoter contains 19 CpG residues. Because the single CpG residue in the mouse promoter is located in the region where the key pol I transcription factor UBF binds, methylation of the cytosine of this dinucleotide prevents mouse rDNA gene transcription. It is likely that the role of three DNMTs, particularly the proteins associated with these enzymes (27,46,(52)(53)(54) in the methylation of the rodent rDNA gene may be distinct from that of the human gene. It is logical to conceive that a distinct set of DNMT-associated proteins participate in the methylation of pol I and II promoters.