The major DNA methyltransferase, Dnmt1, associates with DNA replication sites in S phase maintaining the methylation pattern in the newly synthesized strand. In view of the slow kinetics of Dnmt1 in vitro versus the fast progression of the replication fork, we have tested whether Dnmt1 associates with chromatin beyond S phase. Using time‐lapse microscopy of mammalian cells expressing green‐fluorescent‐protein‐tagged Dnmt1 and DsRed‐tagged DNA Ligase I as a cell cycle progression marker, we have found that Dnmt1 associates with chromatin during G2 and M. This association is mediated by a specific targeting sequence, shows strong preference for constitutive but not facultative heterochromatin and is independent of heterochromatin‐specific histone H3 Lys 9 trimethylation, SUV39H and HP1. Moreover, photobleaching analyses showed that Dnmt1 is continuously loaded onto chromatin throughout G2 and M, indicating a replication‐independent role of Dnmt1 that could represent a novel and separate pathway to maintain DNA methylation.
The higher eukaryotic genome is broadly organized into two structurally distinct chromatin domains, euchromatin and heterochromatin, which functionally translate into active and repressed genetic elements, respectively. At the molecular level, such chromatin domains are characterized by two principal epigenetic marks, DNA methylation and histone modifications. In mammals, DNA methylation predominantly occurs at cytosine residues in CpG dinucleotide sites and is mostly associated with a repressive effect on gene expression (Doerfler, 1983; Bird & Wolffe, 1999; Leonhardt & Cardoso, 2000). Notably, heterochromatic domains are characterized by increased CpG methylation and trimethylation of histone H3 at Lys 9 (H3K9). Stable maintenance of these epigenetic marks through multiple cell divisions is essential for regulated gene expression, development and stability of the genome (Li et al, 1992; Jackson‐Grusby et al, 2001; Peters et al, 2001).
In mammals, methylation of CpG dinucleotides is catalysed by two classes of enzyme—de novo and maintenance DNA methyltransferases. Dnmt3a and Dnmt3b belong to the former class and are required for establishment of new DNA methylation patterns during development (Okano et al, 1999). Dnmt1 has a preference for hemimethylated CpG sites (Bestor, 1992), which are a product of DNA replication. Corroborating this, Dnmt1, but not Dnmt3a and Dnmt3b, relocates to sites of DNA replication (replication foci, RF) as the cell enters S phase (Leonhardt et al, 1992; Margot et al, 2001). These observations led to the proposal that Dnmt1 is the maintenance methyltransferase and that its association with replication sites during S phase ensures the faithful duplication of the DNA methylation patterns post‐replicatively. Three regions of its amino‐terminal domain have been implicated in targeting Dnmt1 to RF, namely, the PCNA binding domain (PBD), targeting sequence (TS) and the polybromo homology domain (PBHD) (Leonhardt et al, 1992; Chuang et al, 1997; Liu et al, 1998). It is presently unclear which domain mediates this association or whether separate domains are required for targeting to early‐replicating euchromatic regions and for late‐replicating heterochromatic regions. It is also unknown how Dnmt1, which shows very slow kinetics in vitro (Pradhan et al, 1999), can cope with the replication fork speed, in particular at densely methylated heterochromatic regions, or whether Dnmt1 function extends beyond S phase. We have, therefore, undertaken a detailed analysis of the molecular basis for the association of Dnmt1 with the replication machinery and with chromatin, throughout the entire cell cycle in living cells. Our in vivo results show that in addition to its binding to replication sites, which is dependent exclusively on PBD, Dnmt1 associates with chromatin (preferentially, constitutive heterochromatin) from late S till early G1 by means of a separate domain called TS. This association is independent of heterochromatin‐specific histone methylation and proteins that bind these methylated histones. Furthermore, it is independent of DNA replication as it occurs de novo during G2 and M phases.
Results and Discussion
We have previously shown that Dnmt1 associates with sites of active DNA replication during S phase (Leonhardt et al, 1992). It has since been accepted that this spatial and temporal coupling of Dnmt1 with the replication machinery ensures the maintenance of DNA methylation patterns at every cell cycle. However, the slow methylation reaction kinetics of the purified Dnmt1 in vitro raises the question as to how the replication‐coupled enzyme efficiently methylates the large number of cytosines, in particular in densely methylated heterochromatin regions. To address this question, we have investigated the cellular distribution of Dnmt1 throughout the entire cell cycle.
The cell cycle stage in vivo was identified using a DNA ligase I fusion with red fluorescent protein (RFP–ligase) as an S‐phase marker (Cardoso et al, 1997). In addition, we generated a green fluorescent protein (GFP) fusion with Dnmt1 (GMT). We followed double‐transfected cells over the cell cycle by taking confocal 3D image stacks every 30–60 min. Time‐lapse analysis (Fig 1) showed that Dnmt1 associated with all RF but, surprisingly, its chromatin association persisted throughout G2 and M phases, in contrast to DNA ligase I. Furthermore, there was a preferential association with late‐replicating heterochromatin (arrowheads), which was no longer detectable in the next G1 phase. We further confirmed that the same was true for the endogenous enzyme, using a variety of antibodies to Dnmt1 (supplementary Fig 1 online and data not shown). This prolonged association of Dnmt1 with chromatin might then allow sufficient time for fully methylating all hemimethylated sites generated during DNA replication.
To clarify how Dnmt1 binds to RF during S phase and which domain is responsible for its continued association with chromatin during G2 and M phases, we generated a series of fluorescently tagged mutants affecting the known domains of Dnmt1. First, we analysed which domain is required for association with RF and, importantly, whether there is any selective preference for early or late RF that differ in their methylation densities. The results are summarized in Fig 2. The N‐terminal domain fused to GFP (MTNG) associated with RF at all stages of S phase and deletion of the PBHD alone (MTNG.1) or both TS and PBHD (MTNY.2 and MTNY.3) did not affect the ability of the fusion protein to associate with RF. Contrary to earlier reports (Liu et al, 1998), PBHD–YFP fusion protein did not associate with RF at any stage during S phase. Interestingly, TS colocalized with late‐replicating heterochromatin. A construct carrying only the 10‐amino‐acid PBD and the following ten residues rich in basic amino acids (PBD–GFP) efficiently associated with RF throughout S phase, and not only during early S as suggested before (Chuang et al, 1997). In summary, association of Dnmt1 with RF throughout S phase depends on PBD, whereas TS shows a preference for late‐replicating chromatin.
Next, we investigated which Dnmt1 domain was responsible for the continued association with chromatin in G2 and M by time‐lapse analysis (Fig 3A) as well as by using the cell cycle staging strategy described in supplementary Fig 1 online (data not shown). The results of the analysis of all Dnmt1 mutants showed that TS alone is responsible for the persistent association with chromatin throughout G2 and M. Similar results were obtained in other cell types and species (supplementary Fig 2 online). Deletion of the TS domain completely abolishes Dnmt1 association with chromatin during G2 and M, but does not prevent the binding of Dnmt1 to RF (Fig 3B). We then tested whether TS interacts preferentially with constitutive heterochromatin (centromeric) or also with facultative heterochromatin (inactive X) by detecting the latter with antibodies against trimethylated K27 residue of histone H3 (Peters et al, 2003; Plath et al, 2003). TS–GFP showed a strong preference for centromeric heterochromatin and only very low levels were detectable at facultative heterochromatin (arrows, Fig 3C). During mitosis, there was still a preferred accumulation at centromeric heterochromatin although TS was also detected along the entire chromosomes (Fig 3A and supplementary Fig 2 online). As histone H3 trimethylation at Lys 9 is a characteristic feature of centromeric heterochromatin (Peters et al, 2003), we tested whether TS association is dependent on this modification or on SUV39H/HP1 binding by expressing TS–GFP in Suv39h1 and Suv39h2 double‐null primary mouse embryonic fibroblasts (PMEFs; Lachner et al, 2001). These cells contain HP1 proteins, but their heterochromatin association is abrogated (Fig 3D). In contrast, we found a clear localization of TS–GFP at chromatin with the same preferential centromeric heterochromatin association in Suv39h1/2 double‐null PMEFs (Fig 3D), which clearly shows that this association is independent of HP1, SUV39H1, SUV39H2 and H3K9 trimethylation.
We then investigated whether Dnmt1 association with chromatin during G2 and M occurs independently of DNA replication. We tested whether the enzyme is de novo loaded during G2 or stays stably bound from S phase by fluorescence photobleaching and recovery analyses (FRAP). In Fig 4A, a cell nucleus is shown before bleaching with dispersed distribution of RFP‐tagged PCNA (used, like DNA ligase I, as a cell cycle progression marker; Leonhardt et al, 2000) and strong association of GFP‐tagged Dnmt1 to heterochromatin, indicating that the cell is in G2 phase. Both fluorophores were bleached in the indicated area (inset), which included a heterochromatic chromocentre (Fig 4A). As expected, RFP–PCNA in non‐S‐phase cells is highly mobile and redistributed very fast after photobleaching, showing in the first post‐bleach image a dispersed distribution in the nucleoplasm (Sporbert et al, 2002). Interestingly, GFP‐tagged Dnmt1 showed essentially complete fluorescence recovery at the bleached heterochromatin focus in about 2 min (half‐time of recovery was 11 s in Fig 3B and 11.7 s for the average curve in Fig 4C), clearly arguing against a unique loading in S phase with static chromatin association throughout G2 and M phases. De novo loading of Dnmt1 was also observed in M phase (data not shown). This clearly shows that there is continuous de novo chromatin binding of Dnmt1 throughout G2/M, which occurs independently of DNA replication.
In summary, Dnmt1 associates throughout S phase with all sites of active DNA synthesis by means of PBD but, in addition, is continuously loaded onto chromatin (preferentially constitutive heterochromatin) from late S throughout G2 and M by means of an independent TS (Fig 5).
We propose that the prolonged association of Dnmt1 with chromatin could allow enough time for the maintenance of the dense methylation of centromeric heterochromatin and might additionally act as a marking and loading platform for subsequent chromatin modifiers. Recent studies in fungi, plants and metazoans have debated the interdependence of DNA methylation and histone modifications in setting up silent heterochromatin domains (Tamaru & Selker, 2001; Soppe et al, 2002; Lehnertz et al, 2003; Tariq et al, 2003). The inherent ability of Dnmt1 to bind to centromeric heterochromatin independently of SUV39H‐mediated H3K9 trimethylation and HP1 suggests separate pathways for establishment and maintenance of stable heterochromatin domains. Furthermore, our data suggest a novel function for Dnmt1 that goes beyond the known replication‐coupled maintenance of DNA methylation patterns. The replication‐independent recruitment of Dnmt1 by means of TS to chromatin could represent a novel and independent pathway to maintain DNA methylation at every cell division.
Cell culture and transfection. Mouse C2C12 myoblasts and HeLa cells were cultured in DMEM supplemented with 20% and 10% FCS, respectively. Suv39H1/2 double‐knockout PMEFs were maintained as described (O'Carroll et al, 2000). Cells were transfected with plasmid DNA using the calcium phosphate–DNA co‐precipitation method followed by a glycerol shock and incubated overnight before performing immunostainings or live‐cell analyses.
Expression constructs. Plasmid constructs encoding translational fusions of mouse Dnmt1, human DNA ligase I and human PCNA with GFP/YFP/RFP were derived from the following vectors: pEGFP‐N2, pEGFP‐C1, pDsRed1‐C1 (Clontech, Palo Alto, CA, USA) and pEVRF0. An SV40‐NLS was included for efficient nuclear targeting of deletion proteins, which lost the native NLS. The PBD insert for PBD–GFP fusion protein was generated by synthesizing sense and antisense oligonucleotides corresponding to both strands of the PBD coding region. The mouse FLAG‐tagged HP1β expression construct was as described (Nielsen et al, 2001). We tested correct expression of the fusion proteins in COS7 or 293T cells by western blotting as described before (Leonhardt et al, 1992).
Immunofluorescence. Cells were fixed with 3.7% formaldehyde in PBS and permeabilized with 0.25% Triton X‐100 for 10 min. The following primary antibodies (diluted in buffer containing 0.2% gelatin) were used: anti‐2x‐trimethyl H3K27 (6523) rabbit antibody (Peters et al, 2003) and anti‐FLAG M2 mouse monoclonal antibody (Kodak, Rochester, NY, USA). Secondary antibodies (Molecular Probes, Leiden, the Netherlands) were conjugated to Alexa Fluor 568. The DNA was counterstained with Hoechst 33258 or TOPRO‐3 and cells were mounted in Mowiol with 2.5% DABCO.
A detailed description of the strategy and methods to identify cells at the different cell cycle stages is included in supplementary Methods online. More than 100 cells from each cell cycle stage were analysed to determine the localization of the wild‐type and mutant Dnmt1 proteins.
Fixed‐cell microscopy. Stained cells were analysed using a Zeiss LSM 510 Meta confocal microscope equipped with a × 63/1.4 NA Plan‐Apochromat objective. Ar laser (488 and 514 nm), HeNe laser 1 (543 nm) and HeNe laser 2 (633 nm) were used to excite the fluorophores. Hoechst 33258‐counterstained nuclei were analysed on an Axioplan 2 microscope and images acquired with a cooled CCD camera (SensiCam) using Zeiss Axiovision software and appropriate filter sets.
Live‐cell microscopy and FRAP analysis. Live‐cell analysis was performed by plating the cells on 40‐mm‐diameter glass coverslips before transfection. At 1 day after transfection, the coverslip was mounted onto the microscope stage and maintained at 37°C using an FCS2 live‐cell chamber and temperature controller (Bioptechs, Butler, PA, USA).
For time‐lapse analysis, images were acquired with a Zeiss LSM 510 using the 488 and 543 nm laser lines at low power (1–5%). Four z‐sections at 1 μm intervals were imaged every 30–60 min and the cells were followed throughout the cell cycle. After image acquisition, the different z‐sections at each time point were aligned manually to correct for movements in the z‐plane.
For FRAP analysis, a region of interest was selected and photobleached by an intense 488 nm Ar laser beam (set to 100% transmission) for 1 s, after which confocal image series were recorded at 2.5 s time intervals. Mean fluorescence intensities of the bleached region were corrected for background and for total nuclear loss of fluorescence over the time course. FRAP data of five nuclei were averaged and the mean curve as well as the standard deviation was calculated and displayed using the GraphPad Prism™ software. The half‐time of recovery was calculated from the single as well as the average curves.
Supplementary information is available at EMBO reports online (http://www.nature.com/embor/journal/v5/n12/extref/7400295‐s1.pdf).
We are indebted to T. Jenuwein for providing Suvar39h1/2 dn fibroblasts and antibodies, S. Tajima and I. Suetake for the gift of 293T cells and the anti‐N‐terminal Dnmt1 antibody, and P. Chambon for the HP1β expression plasmid. We are grateful to M. Grohmann, I. Grunewald, P. Domaing, D. Nowak and K. Zolghadr for their help. This work was supported by grants from the Deutsche Forschungsgemeinschaft and the Volkswagenstiftung to H.L. and M.C.C.
- Copyright © 2004 European Molecular Biology Organization