|Home | About | Journals | Submit | Contact Us | Français|
The human genome contains a number of methyl CpG binding proteins that translate DNA methylation into a physiological response. To gain insight into the function of MBD2 and MBD3, we first applied protein tagging and mass spectrometry. We show that MBD2 and MBD3 assemble into mutually exclusive distinct Mi-2/NuRD-like complexes, called MBD2/NuRD and MBD3/NuRD. We identified DOC-1, a putative tumor suppressor, as a novel core subunit of MBD2/NuRD as well as MBD3/NuRD. PRMT5 and its cofactor MEP50 were identified as specific MBD2/NuRD interactors. PRMT5 stably and specifically associates with and methylates the RG-rich N terminus of MBD2. Chromatin immunoprecipitation experiments revealed that PRMT5 and MBD2 are recruited to CpG islands in a methylation-dependent manner in vivo and that H4R3, a substrate of PRMT, is methylated at these loci. Our data show that MBD2/NuRD and MBD3/NuRD are distinct protein complexes with different biochemical and functional properties.
Methylation of CpG dinucleotides in regulatory regions of genes is an important mark for epigenetic regulation of transcription (2). Since DNA methylation is passed on to daughter cells during cell division, these methyl CpG marks can be maintained during development and provide epigenetic memory (31). A number of proteins have been identified in the human genome that can specifically bind to methylated CpG residues via a methyl CpG binding domain (MBD) (15, 39). Recruitment of these proteins to promoters containing methylated CpG-rich stretches—CpG islands—is thought to result in modulation of chromatin structure and repression of transcription. The human genome encodes five MBD proteins: MeCP2 and MBD1 to -4 (6, 14, 23). Apart from MBD3, these proteins have been shown to have specific methyl CpG binding activity. Recently a novel protein, Kaiso, was identified as a methyl CpG binding protein even though this protein lacks a classical MBD but appears to bind specifically to methylated DNA via a zinc finger domain (30).
Several MBD proteins have been reported to interact with histone deacetylases (HDACs) as well as histone methyltransferases. MeCP2 has been described to interact with the Sin3/HDAC corepressor complex (18) and Brahma (13), as well as with the histone H3 lysine-9 methyltransferase Suvar 3-9 (12), although these interactions may not be stable since MeCP2 is mostly present inside the cell as a monomer (12, 18, 19, 26). MBD2 and MBD3 have been identified as core subunits of the Mi-2/NuRD complex (9, 27), whereas Kaiso is part of the HDAC-containing N-CoR complex that plays an important role in transcription regulation by nuclear hormone receptors (27, 42, 44). Collectively, these findings suggest a functional link between DNA methylation, histone deacetylation, and histone methylation and indicate that these epigenetic events functionally cooperate to regulate transcription and cellular memory.
MBD2 and MBD3 have both been described as subunits of the Mi-2/NuRD complex. It has been proposed that MBD2, which exhibits methyl CpG binding activity, serves to recruit the MBD3-containing Mi-2/NuRD complex to methylated promoters (44). Knockout studies in mice, however, suggest that MBD2 and MBD3 have distinct nonoverlapping functions: whereas knocking out MBD3 results in embryonic lethality, MBD2-knockout mice are viable and display relatively subtle defects (16). Interestingly, Sansom and coworkers recently showed that the absence of MBD2 protects against intestinal tumorigenesis (32). Thus, although biochemical evidence suggests that MBD2 and MBD3 are part of the same complex, the knockout studies suggest that both proteins have specific or maybe partially overlapping functions.
To gain insights into the protein composition and function of MBD2 and MBD3, we generated stable cell lines expressing tagged versions of these proteins. Purification of the protein complexes revealed that MBD2 and MBD3 are not copurifying but are mutually exclusive. In addition to known Mi-2/NuRD subunits, a 12-kDa protein called DOC-1 was identified as a novel core subunit of both the MBD3 and MBD2 complexes. Furthermore, PRMT5 and its associated cofactor MEP50 were found to copurify with and methylate MBD2 in vitro. Finally, PRMT5 and its H4R3 histone methyltransferase activity were shown to be recruited with MBD2 to CpG islands in a methylation-sensitive manner in vivo, suggesting an unexpected role for an arginine methyltransferase in repression by MBD2. Collectively, these findings provide evidence that MBD2/NuRD and MBD3/NuRD define two distinct protein complexes with different biochemical and functional properties.
Oligonucleotides pRAV-myc1f and pRAV-myc1r encoding a Myc epitope with EcoRI overhangs were cloned into EcoRI-digested vector fragments of pRAV-FLAG (20) to generate pRAV-myc. An EcoRI/BamHI fragment from this vector containing one ProtA domain, two tobacco etch virus (TEV) cleavage sites, and a myc epitope was then ligated with EcoRI/BamHI-digested vector pZ-1-N (Cellzome) to generate retroviral vector pZXN. MBD3, MBD2, and MBD2 lacking the RG stretch were PCR amplified with EcoRI and XhoI overhangs from an MBD3 plasmid (RZPD) and an MBD2 plasmid (image clone collection) which were then ligated into EcoRI/XhoI-digested vector pZXN.
To create a Strep-tagII (2TEV) Myc 2× hemagglutinin (HA) cassette, the 2TEV Myc cassette from pZXN was PCR amplified using a forward primer containing an EcoRI site and a Strep-tagII epitope and a reverse primer with an EcoRI overhang and two HA sites. This fragment was ligated into EcoRI-digested vector psg5-HA TBP. The cassette was PCR amplified again with a forward primer containing a BamHI restriction site and a reverse primer containing one new HA epitope and a NotI restriction site. This PCR product was digested with BamHI and NotI and ligated into BamHI/NotI-digested plasmid pcDNA5/FRT/TO/C-TAP (kind gift from Bernard Luscher) to generate pcDNA5/FRT/TO/stII(2TEV)myc tripleHA. MBD2 was PCR amplified using primers containing NotI and Xho1 restriction sites and ligated into the NotI and XhoI site of pcDNA5/FRT/TO/stII(2TEV)myc tripleHA.
A fragment encoding part of the RG stretch of MBD2 (EGARGGGRGRGR) containing BamHI and EcoRI overhangs was cloned in plasmid pGEX2T (Amersham Pharmacia Biotech). Full-length MBD2, MBD lacking the RG stretch, and MBD3 were PCR amplified with primers containing BamHI and EcoRI overhangs and cloned in BamHI/EcoRI-digested pGEX2T. Primer sequences are available upon request.
MCF7, HEK 293, HeLa, Phoenix, and 293 FLP cells were maintained in Dulbecco's modified Eagle's medium (Invitrogen) supplemented with 10% fetal calf serum, 100 μg of penicillin per ml, and 100 U of streptomycin per ml (Invitrogen) at 37°C in 5% CO2. For 5-aza-2′-deoxycytosine (AzaDc) treatment MCF7 cells were seeded at low density and treated with 1 μM of AzaDc for 72 h. Retroviral stable cell lines were generated according to the following procedure. Phoenix amphotropic packaging cells (2.5× 106 cells) were seeded on a 9-cm dish and transfected 24 h later with 20 μg of retroviral plasmid pZXN-MBD2a or pZXN-MBD2 lacking the RG stretch or pZXN-MBD3 using the calcium phosphate method. After 48 h virus-containing supernatant was filtered through a 0.22-μm-pore-size filter. HeLa or 293 cells (105 each) were seeded in a six-well plate and transduced with 3 ml filtered virus supernatant in the presence of 8 μg/ml of Polybrene for two infectious rounds of 24 h. Cells were then incubated for 24 h in normal medium. The polyclonal population of cells was then selected with 1 μg/ml of puromycin. Clones were then selected, grown in isolation, and screened for recombinant protein expression.
A double stable cell line expressing tagged MBD2 and MBD3 was generated according to the following procedure. 293 FLP cells were transfected using the calcium phosphate method on a 9-cm dish with 2 μg of pcDNA5/FRT/TO/stII(2TEV)myc tripleHA-MBD2 and 18 μg of POG44. After 36 h cells were selected with 100 μg/ml hygromycin. Subsequently, clones were derived and screened for recombinant protein expression and zeocin sensitivity. One good clone was then transduced with virus containing pZXN-MBD3 and was double selected with 100 μg/ml hygromycin and 1 μg/ml of puromycin.
Cell pellets were resuspended in lysis buffer (420 mM KCl, 20% glycerol, 20 mM HEPES, pH 7.9, 0.2 mM EDTA, 5 mM MgCl2, 0.1% Triton X-100, 10 mM β-mercaptoethanol, 1 mM phenylmethylsulfonyl fluoride [PMSF], and complete protease inhibitors [Roche]) and homogenized by 20 strokes with a type B pestle. Extracts were then incubated for 1 hour in a rotation wheel at 4°C to extract nuclear proteins. Lysates were subsequently clarified by ultracentrifugation at 100,000 × g. Whole-cell extracts were aliquoted, snap frozen, and stored at −80°C until further usage.
Whole-cell extracts derived from tandem affinity purification (TAP)-tagged cell lines were diluted with 2 volumes binding buffer (150 mM NaCl, 20 mM Tris-HCl, pH 8.0, 0.1% NP-40, 1 mM dithiothreitol [DTT], 1 mM PMSF, and complete protease inhibitors [Roche]) and then incubated with immunoglobulin G (IgG) Sepharose beads (Pharmacia) for 2 h at 4°C in a rotation wheel. Beads were then washed three times with 10 bead volumes of wash buffer (500 mM NaCl, 20 mM Tris-HCl, pH 8.0, 0.5% NP-40, 1 mM DTT, and 1 mM PMSF) and twice with 10 bead volumes of TEV cleavage buffer (150 mM NaCl, 20 mM Tris-HCl, pH 8.0, 0.1% NP-40, 1 mM DTT, and 0.5 mM EDTA). Beads were then resuspended in 1 bead volume TEV cleavage buffer containing TEV protease and incubated overnight at 4°C in a rotation wheel. TEV eluates were precleared with protein A beads (Pharmacia) and then subjected to immunoprecipitation using Myc antibody (9E11). Immunoprecipitates were washed three times with 10 bead volumes of wash buffer and twice with 10 bead volumes of peptide elution buffer (100 mM KCl, 20% glycerol, 20 mM HEPES KOH, pH 7.9, 0.2 mM EDTA, 0.1% NP-40, 5 mM DTT, and 0.5 mM PMSF). Protein complexes were eluted from the beads by incubation in peptide elution buffer containing 2 mg/ml of Myc peptide at 28°C for 30 min in a thermoshaker. The elution step was carried out twice, and both eluates were pooled.
Endogenous immunoprecipitation assays were performed with HeLa nuclear extract in stringency conditions similar to those for the TAP tag procedure. Antibodies used were MBD3 (IBL, Japan) and MBD2 07-198 (Upstate Biotechnology).
Purified protein complexes were loaded on sodium dodecyl sulfate (SDS)-polyacrylamide gels and run briefly to get rid of detergent and the excess of the peptide used for the elution. The gel lane was then fixed, cut in small pieces, and subsequently reduced and alkylated. Proteins were digested overnight with trypsin (Promega) and eluted from the gel with trifluoroacetic acid. Peptide identification experiments were performed using a nano-high-pressure liquid chromatography Agilent 1100 nanoflow system connected online to a 7-Tesla linear quadrupole ion trap-Fourier transform (FT) mass spectrometer (Thermo Electron, Bremen, Germany) essentially as described previously (28).
MCF7 cells were cross-linked with 1% formaldehyde for 15 min at room temperature, and chromatin was prepared as described previously (3, 41) but excluding CsCl purification. Chromatin was sonicated to an average size of 500 bp. Chromatin derived from 1 million cells was used for each immunoprecipitation in incubation buffer (1% Triton X-100, 150 mM NaCl, 1 mM EDTA, pH 8.0, 0.5 mM EGTA, pH 8.0, 10 mM Tris, pH 8.0, 1 mg/ml bovine serum albumin, and protease inhibitors). Four micrograms of the following antibodies was used for immunoprecipitations: PRMT5 12-303 (Upstate Biotechnology), MBD2 IMG-147 (Imgenex), MBD3 (IBL, Japan), MTA2 PC656 (Oncogene), and anti-dimethyl-histone H4 (Arg3) (07-213) (Upstate Biotechnology). After overnight incubation at 4°C immunoprecipitates were washed twice with 0.1% SDS, 1% Triton X-100, 0.1% deoxycholate, 0.15 M NaCl, 1 mM EDTA, 10 mM Tris (pH 8.0), 0.5 mM EGTA; once with 0.1% SDS, 1% Triton X-100, 0.1% deoxycholate, 0.5 M NaCl, 1 mM EDTA, 10 mM Tris (pH 8.0), 0.5 mM EGTA; once with 0.25 M LiCl, 0.5% deoxycholate, 0.5% NP-40, 1 mM EDTA, 10 mM Tris (pH 8.0), 0.5 mM EGTA; and twice with 1 mM EDTA, 10 mM Tris (pH 8.0), 0.5 mM EGTA. Immunocomplexes were eluted from the beads by adding 1% SDS, 0.1 M NaHCO3 followed by incubation at room temperature for 15 min. Protein-DNA cross-links were reversed in 0.2 M NaCl at 65°C for 4 h, after which DNA was isolated by phenol-chloroform extraction. Real-time quantitative PCR analyses were performed to assess recruitment of the proteins to specific sites. The relative occupancy was derived from the percent recovery of a specific CpG island against the percent recovery of a control BMX region. Means and standard deviations were then calculated from chromatin immunoprecipitation (ChIP) experiments performed from three independent chromatin isolations.
Whole-cell extracts derived from the MBD2 stable cell line or wild-type HEK 293 cells were diluted with 3 volumes of IPP150 (150 mM NaCl, 20 mM Tris HCl, pH 8.0, 0.1% NP-40, 1 mM DTT, and 1 mM PMSF) and incubated with IgG Sepharose beads for 2 h at 4°C in a rotation wheel. Beads were washed three times with 10 bead volumes of IPP150 and then incubated with 1 bead volume of PRMT5 incubation buffer (20 mM HEPES, pH 7.6, 500 mM NaCl, 1 mM MgCl2) in the presence of 0.25 μCi of S-[14C]adenosylmethionine (Amersham). Glutathione S-transferase (GST)-PAH2, GST-MBD2, GST-MBD2 lacking the RG stretch, GST-MBD3, and GST-EGARGGGRGRGR were expressed and purified as described previously (22). These purified proteins were incubated in PRMT5 incubation buffer supplemented with 0.25 μCi of S-[14C]adenosylmethionine (Amersham) in the presence of purified MBD2 complex or a purified Drosophila melanogaster fraction highly enriched for PRMT5/MEP50/USP7 (37). After 2 h of incubation at 30°C products were separated on a 12% SDS-polyacrylamide gel. The gel was then dried and exposed on a phosphoscreen (Bio-Rad) to identify methylated proteins.
Human embryonic kidney (HEK 293) cells stably expressing tagged versions of MBD2 and MBD3 were generated to determine their subunit composition. Gel filtration analysis of whole-cell extracts derived from these cell lines indicated that both proteins are present in high-molecular-weight fractions of approximately 1 to 1.5 MDa (Fig. (Fig.1A1A).
To assess whether the purified complexes were enzymatically active, we performed deacetylation assays on nucleosomal templates acetylated by the SAGA and NuA4 complex (38). Both MBD2 and MBD3 protein complexes displayed robust trichostatin A-sensitive deacetylation activity towards histone H3 and histone H4 (Fig. (Fig.1B).1B). To further investigate the functionality of the purified complexes, electrophoretic mobility shift assays using methylated DNA probes were performed which revealed that only the MBD2 complex was able to bind to methylated DNA. MBD3 shows no affinity for methylated DNA, despite the presence of a highly conserved MBD, as reported previously (40) (Fig. (Fig.1C,1C, compare lanes 2 and 4).
Previous studies have reported that MBD2 and MBD3 are part of the same complex (9). Silver staining of the purified MBD2 and MBD3 complexes revealed a protein of approximately 35 kDa that is lacking in the MBD2 preparation (Fig. (Fig.2A,2A, marked with an arrow). Western blotting identified this protein as MBD3 and revealed the absence of MBD3 in the MBD2 complex (Fig. (Fig.2B,2B, compare lanes 1 and 2). The absence of MBD3 in the MBD2 preparation cannot be explained by a shortage of endogenous MBD3 in 293 cells (Fig. (Fig.2B,2B, lane 3). These data suggest that MBD2 and MBD3 are not part of the same complex but that they may even be mutually exclusive. If MBD2 and MBD3 are, however, present as a heterodimer in the Mi-2/NuRD complex, overexpression of one of the MBDs could cause a shift from an MBD2/MBD3 heterodimer population to an MBD2/MBD2 or MBD3/MBD3 homodimer population. To assess this possibility, we generated a stable cell line expressing MBD2 and MBD3 with different tag combinations. MBD3 was tagged with a ProtA domain and a Myc epitope, whereas MBD2 was tagged with a Strep-tagII and an HA epitope. Western blotting shows that both of these proteins are expressed in the stable cell line (Fig. (Fig.2C,2C, lane 6). Purification of ProtA-Myc-MBD3 on IgG beads resulted in purification of MBD3 (Fig. (Fig.2C,2C, lane 3); tagged MBD2 could not be detected in the immunoprecipitate. Similarly, purification of Strep-tagII-HA-MBD2a on streptactin beads resulted in purification of MBD2, whereas ProtA-Myc-MBD3 did not copurify (Fig. (Fig.2C,2C, lane 1). HDAC1 copurified with tagged MBD2 as well as MBD3, indicating that both tagged proteins assemble in a functional complex. To further substantiate these observations, immunoprecipitation experiments against endogenous MBD2 and MBD3 in HeLa cells were performed (Fig. (Fig.2D).2D). Immunoprecipitation of MBD2 resulted in purification of MBD2 but not of MBD3. Similarly, immunoprecipitation of MBD3 resulted in purification of two polypeptides whereas MBD2 did not copurify. Based on their relative migration we presume these two polypeptides to be MBD3 and the smaller variant MBD3L2 lacking the MBD. Collectively these experiments strongly suggest that MBD2 and MBD3 are mutually exclusive.
To further characterize the purified MBD2 and MBD3 complex, liquid chromatography FT-ICR MS analyses were performed (Table (Table1).1). In agreement with the results described above, the MBD2 complex did not contain MBD3 and vice versa, corroborating and extending our conclusion that MBD2 and MBD3 are mutually exclusive in 293 cells. To unambiguously determine whether the observed mutual exclusiveness of MBD2 and MBD3 is specific for 293 cells or whether this is also true in other cells, a HeLa cell line stably expressing tagged MBD3 was generated. Purification of MBD3 from this cell line resulted in the purification of a Mi-2/NuRD complex lacking MBD2, indicating that in HeLa cells MBD2 and MBD3 are also mutually exclusive (unpublished data).
In 293 cells, Mi-2α and Mi-2β were identified in both the MBD2 and MBD3 complexes. These proteins have previously been characterized by the Schreiber lab as Mi-2/NuRD components CHD3 and CHD4, respectively (35). At present we do not know whether these two isoforms are forming heterodimers or whether Mi-2α and Mi-2β are assembled into distinct complexes. In both complexes, RbAp48 and -46 and HDAC1 and -2, the catalytic module for nucleosomal deacetylation activity, were identified. Furthermore, the highly related p66α and p66β proteins were identified in both complexes as described previously (5). We did not identify peptides matching the histone demethylase LSD1, which has been reported to interact with the Mi-2/NuRD complex (34). A 12-kDa protein called cdk2-associated protein 1 was identified as a novel Mi-2/NuRD component of both the MBD2 and MBD3 complexes. cdk2-associated protein 1 or DOC-1 (deleted in oral cancer 1) is a putative tumor suppressor reported to be inactivated during oral carcinogenesis and colon cancer (36, 43). Furthermore the MBD2 but not the MBD3 eluate contained a large number of peptides matching the arginine methyltransferase PRMT5 as well as its associated protein called MEP50. Finally, several peptides matching different importin α nuclear transport proteins were identified. These proteins were absent in the MBD3 complex, suggesting a specific interaction with MBD2. The association between MBD2 and importins may indicate that MBD2 shuttles between the cytoplasm and the nucleus.
Strikingly all three MTA proteins, MTA1, -2, and -3, as well as two MTA splice variants were identified in the MBD2 and MBD3 complex (Table (Table1).1). It has been suggested that MTA proteins display tissue-specific differential expression giving rise to distinct Mi-2/NuRD complexes (4, 11, 21, 44). Surprisingly, a large number of different posttranslational modifications were identified in all major NuRD subunits (unpublished data). A previous study characterized phosphorylation sites in Mi-2α, Mi-2β, p66α, p66β, HDAC1, and HDAC2 (1). We confirmed the presence of these phosphorylation sites in the NuRD complex and in addition identified a plethora of new sites in the latter as well as in MBD2, MTA1, MTA2, and MTA3. Several posttranslational modifications were detected in conserved domains, and these might therefore have a role in regulating protein-protein or protein-DNA interactions or in fine-tuning of enzymatic activities. Furthermore ~90% of these modifications occurred in highly conserved residues, supporting a role for these modifications throughout evolution.
PRMT5 was detected by FT-MS/MS in the MBD2 complex but was absent from the MBD3 peptide eluate, which was confirmed by Western blotting (Fig. (Fig.3A).3A). Inspection of the amino acid sequence of the subunits of the MBD2 complex revealed that MBD2 has a long stretch of RG repeats N terminal to the MBD (Fig. (Fig.3B),3B), whereas RG repeats are not present in MBD3 or in other subunits. Since the RG motif is a substrate for PRMT5 (10), we tested whether the MBD2 RG stretch is a substrate for PRMT5. Incubating purified MBD2 complex in the presence of S-[14C]adenosylmethionine resulted in a single radioactive band migrating at the position of TAP-tagged MBD2 in the gel (Fig. (Fig.3C).3C). To substantiate these observations, we fused full-length MBD2 or the RG stretch of MBD2 to GST and tested whether the purified MBD2 complex containing PRMT5 could methylate these fusion proteins. As shown in the left panel in Fig. Fig.3D,3D, the MBD2 complex was able to specifically methylate these recombinant substrates but not a GST-PAH2 control. To evaluate whether PRMT5 specifically methylates the RG stretch of MBD2, MBD2 lacking the RG stretch or MBD3 was fused to GST and incubated with purified MBD2 complex. As shown in the right panel in Fig. Fig.3D,3D, the MBD2 complex was able to specifically methylate the RG stretch of MBD2 but not MBD3 or MBD2 lacking the RG stretch. A control for the methylation reaction using a purified PRMT5-containing fraction from Drosophila displayed activity similar to that of the MBD2 complex, thus confirming the specificity of PRMT5 in this assay. In addition, a search for posttranslationally modified peptides in the FT-MS/MS run of the purified MBD2 complex indeed revealed a peptide containing three dimethyl arginine residues (unpublished data). Taken together, these experiments strongly suggest that PRMT5 methylates MBD2 on several arginine residues located in the RG-rich amino acid stretch immediately upstream of the MBD of MBD2 in vitro.
To investigate whether the RG stretch of MBD2 is required for the interaction between PRMT5 and the MBD2 complex, we generated a stable cell line expressing a truncated MBD2 protein starting at the second methionine in the MBD2 sequence, thus lacking the RG stretch. Following purification and FT-MS/MS analysis, Mi-2/NuRD components, the novel core subunit DOC-1, and importin α proteins were present. However, peptides matching either PRMT5 or MEP50 were not identified (Fig. (Fig.3E).3E). Western blotting confirmed that PRMT5 was present in crude extracts but absent in the truncated MBD2 eluate (data not shown). Taken together these results strongly suggest that PRMT5 interacts with the N-terminal RG-rich stretch of MBD2 and methylates this RG stretch.
To assess whether PRMT5 plays a role on chromatin with MBD2, we performed chromatin immunoprecipitation experiments on endogenous proteins in MCF7 breast carcinoma cells. Different CpG island targets which were previously shown to be methylated and bound by MBD proteins in MCF7 cells were analyzed (7, 24, 25) (Fig. (Fig.4A).4A). Chromatin immunoprecipitation using MTA2, MBD2, and MBD3 antibodies followed by real-time quantitative PCR analysis revealed the recruitment of these proteins to two CpG islands, one located close to the first exon of P14ARF and a second CpG island located before the first exon of P16INK4a (Fig. (Fig.4B).4B). Several other tested CpG islands did not recruit MTA2, MBD2, and MBD3. Next, we performed ChIPs using an antibody against PRMT5, and this revealed the recruitment of PRMT5 to the P14ARF and P16INK4a CpG islands. These results provide a functional link between MBD2 and the arginine methyltransferase PRMT5 in vivo.
The biochemical experiments described in this study revealed that MBD2 and MBD3 are mutually exclusive and that PRMT5 interacts with MBD2 but not with MBD3. The ChIP experiments indicate that MBD2 and PRMT5 as well as MBD3 are recruited to the P14ARF and P16INK4a CpG islands. To investigate whether the recruitment of MBD2, PRMT5, and MBD3 to these loci was dependent on CpG methylation, we treated MCF7 cells with AzaDc, a specific inhibitor of DNA methylation, and subsequently investigated the recruitment of MBD2, MBD3, PRMT5, and MTA2 to the P14ARF and P16INK4a CpG islands. As shown in Fig. Fig.4C,4C, AzaDc treatment resulted in a significant loss of MBD2 binding. Strikingly, a reduction of PRMT5 binding to the loci could be observed, indicating that MBD2 and PRMT5 are binding to the P14ARF and P16INK4a CpG islands in a methylation-sensitive manner. In contrast, recruitment of MBD3 to these loci was only moderately affected. MTA2, a protein present in both the MBD2 and MBD3 complexes, was reduced about 50%. Thus, MBD2 and PRMT5 are recruited to the CpG islands in a methylation-dependent manner, whereas MBD3 is only partially affected, supporting the notion that MBD2/PRMT5 and MBD3 are at least to some extent assembling on the CpG islands as distinct complexes. PRMT5 recruitment to promoters is known to correlate with arginine methylation of histones (8). To investigate whether PRMT5 recruitment to the P14ARF and P16INK4a CpG islands correlates with arginine methylation of histone H4, we performed chromatin immunoprecipitation experiments on MCF7 cells treated with or without AzaDc using anantibody against dimethylated histone H4R3. As shown in Fig.Fig.4C4C recruitment of PRMT5 to the P14ARF and P16INK4a CpG islands correlates with an enrichment in the level of arginine dimethylated histone H4R3. Strikingly, treatment of MCF7 cells with AzaDc resulted in a reduction of the level of dimethylated histone H4R3. Thus, PRMT5 recruitment to the P14ARF and P16INK4a CpG islands correlates with histone H4R3 methylation.
In conclusion the experiments described in this study indicate that MBD2 and MBD3 assemble in distinct Mi-2/NuRD-like complexes and are mutually exclusive. Furthermore, PRMT5 binds to and methylates MBD2 and is recruited together with an MBD2-containing Mi-2/NuRD complex to CpG islands in a methylation-dependent manner in vivo.
In this study we set out to gain insights into the function of MBD2 and MBD3. We applied a protein tagging approach to purify MBD2 and MBD3 complexes from mammalian cells. Strikingly, although these proteins have been described to be part of the same complex, we found them to reside in distinct complexes. MBD2 could not be detected in the purified MBD3 complex and vice versa. Independent purification of MBD2 and MBD3 from a double stable cell line expressing MBD2 and MBD3 with different tags to similar levels confirmed their mutual exclusiveness. In addition, endogenous MBD2 did not immunoprecipitate MBD3 and vice versa. Finally, MBD2 could also not be detected in an MBD3 complex purified from HeLa cells stably expressing tagged MBD3 (unpublished data). Feng and Zhang used conventional chromatography to purify the MBD2-containing MeCP1 complex from HeLa nuclear extracts and found MBD3 to copurify, and they argued that these proteins are indeed part of the same complex (9). However, this purified fraction may in fact be a mixture of Mi-2/NuRD complexes, some containing MBD2 and others MBD3. As they likely have a very similar binding affinity for the different chromatographic resins, they would end up in the same fractions throughout the purifications. Jiang and coworkers performed yeast two-hybrid assays and GST pull-downs and found MBD2 and MBD3 to interact directly (17). Although we cannot exclude that a small fraction of MBD2 and MBD3 are interacting, our experiments clearly show that the vast majority of MBD2 and MBD3 proteins independently assemble in distinct Mi-2/NuRD-like complexes, which we propose to call MBD2/NuRD and MBD3/NuRD, respectively. Based on the different subunit compositions, in particular, the presence of PRMT5, MEP50, and the importin complex in MBD2 but not MBD3, we hypothesize that these complexes have different functions inside the cell. This hypothesis is supported by the AzaDc ChIP experiments described in Fig. Fig.44 showing a methylation-dependent MBD2/PRMT5 recruitment to the P14ARF and P16INK4a CpG islands whereas MBD3 recruitment to these loci is largely independent of methylation.
Liquid chromatography-MS/MS and Western blot analyses revealed the association of the arginine methyltransferase PRMT5 and its associated protein MEP50 with the MBD2/NuRD complex, whereas these proteins were lacking in the MBD3/NuRD complex. Whether PRMT5 is a core subunit of the MBD2/NuRD complex or a protein strongly interacting with the MBD2/NuRD complex remains to be determined. PRMT5 is recruited to the MBD2/NuRD complex via the RG-rich N terminus of MBD2. In addition we provided evidence that PRMT5 can methylate this RG stretch of MBD2. Therefore, we hypothesize that the RG stretch of MBD2 might serve a dual purpose as a substrate and as a docking site for PRMT5. PRMT5 has been shown to function in repression of tumor suppressor genes, presumably by adding repressive arginine methyl marks to the histone H3 and H4 tails (29). In agreement with this we found PRMT5 to colocalize with MBD2 on P14ARF and P16INK4a CpG islands, and this correlates with histone H4R3 dimethylation, thus providing a functional link between PRMT5 and MBD2 in vivo.
Since its first description some 7 years ago, the Mi-2/NuRD complex has generally been regarded as one biochemical entity containing a number of core polypeptides. However, our study clearly reveals the presence of MBD2/NuRD and MBD3/NuRD complexes with distinct subunit compositions. Previous observations from a number of labs have revealed the existence of additional Mi-2/NuRD complexes defined for example by different MTA variants, which may allow for a further fine-tuning of different Mi-2/NuRD complexes (4, 11, 21, 33, 44). Finally, in addition to altering protein composition, posttranslational modifications of different Mi-2/NuRD subunits (unpublished data) also may play an important role in regulating its function.
The results described in this study lead us to propose a feed-forward mechanism of repression by different Mi-2/NuRD complexes. The different enzymatic activities gathered within a single protein complex may act synergistically to regulate repression of MBD2 target genes: deacetylation of nucleosomes surrounding the targeting site in combination with the addition of transcriptional repressive arginine methyl marks in the H4 tail by the associated PRMT5 (29). Furthermore, chromatin remodeling catalyzed by the ATPase Mi-2 may occur. The hypoacetylated and arginine methylated nucleosomes surrounding the MBD2/PRMT5 targeting site in turn may provide a binding scaffold for the MBD3/NuRD complex, a complex which has a high affinity for hypoacetylated nucleosomes (data not shown). This results in the co-occurrence of the MBD2/NuRD and MBD3/NuRD complexes on some CpG islands. Further deacetylation of nucleosomes by the MBD3/NuRD complex can then facilitate spreading of deacetylation and maintenance of transcriptional repression. Unraveling the functions unique to each Mi-2/NuRD complex is a challenging task that lies ahead.
We thank Jan van der Knaap for the PRMT5/MEP50 protein fraction, Elly van Tiel for active involvement during a Molecular Biology practical course, and Colin Logie for technical assistance with the FLP recombinase. Furthermore we acknowledge A. Bird, B. Luscher, R. Bernards, and M Knuesel for plasmids; J. Conaway for the 293 FLP cells; and R. Delwel for MBD3 antibody. Finally we thank members of the Stunnenberg lab for discussions and critical reading of the manuscript.
This work was supported by grants from The Netherlands proteomics center.