[Seite 81↓]

4.  Discussion

In mammalian cells replication of the genome takes place at distinct subnuclear sites called replication foci (RF). Various replication factors are localized to these sites via specific protein sequences called RFTS. Also DNMT1, an enzyme that maintains the DNA methylation pattern after DNA replication, localizes at RF. Association of DNMT1 with RF is reported to be mediated independently by three regulatory sequences. To understand the evolutionary conservation in the organization of these subnuclear structures and the mechanisms for the assembly of the replication machinery, we have analyzed the ability of mammalian replication factors and DNMT1 to associate with RF in Drosophila and mammalian cells. To dissect the function of the three regulatory sequences of DNMT1 and their role in controlling the subnuclear distribution of DNMT1 throughout the cell cycle, detailed analyses of various DNMT1 mutants to associate with RF in mammalian cells were carried out.

4.1. The mammalian DNMT1 does not associate with RF in Drosophila cells

The observation that immunostaining for PCNA and BrdU incorporation sites gives a punctate pattern indicates that in Drosophila S2 cells, like mammalian cells, DNA replication is organized in discrete structures or RF (Fig. 3.3). This is consistent with previous observation that replication in Drosophila Kc cells occurs at subnuclear foci (Ahmad and Henikoff, 2001). PCNA is highly conserved between human and Drosophila (Fig. 3.1), and HsPCNA and DmPCNA can efficiently associate with RF across the two organisms (Fig. 3.3). This indicates that PCNA, a conserved central core at the replication fork, is structurally and functionally exchangeable at the RF between mammals and Drosophila. However, in Drosophila cells neither the N-terminal domain of DNMT1 nor the three individual RFTS of DNMT1 showed any association with RF (Fig. 3.5B and Fig. 3.7). The full length DNMT1(s) showed partial association with RF in about 10% of Drosophila cells in S phase (Fig. 3.5D and Table 3.1). One possible explanation for this partial association is that DNMT1 associates with particular features of the genome, which in an unsynchronized cell population would yield a partial association with RF (see also Section 4.3 and Outlook for further discussion). It is surprising that DNMT1 having a functional PBD, which is reported in homologous proteins from a wide range of eucaryotes including Drosophila (Warbrick et al., 1998), is unable to associate with RF in Drosophila cells. Due to the general inability of PBD-GFP fusions from different origins (DNMT1, HsDNA Ligase I, DmDNA Ligase I) to associate with RF in Drosophila cells, we argued that in Drosophila another domain might function as an RFTS. To identify a domain that functions as an RFTS in Drosophila, a series of deletions of DmDNA Ligase I fused to GFP were analyzed for their ability to associate with RF in Drosophila cells. Surprisingly neither the full length DmDNA Ligase I nor any of the deletions associated with RF in Drosophila cells (Fig. 3.11A). However, DmDNA Ligase I associated with RF in mammalian cells, which is mediated by the PBD as deletion of the PBD abrogated this association (Fig. 3.11B). This also indicates that DmDNA Ligase I fused to GFP is expressed and folded properly and that the inability [Seite 82↓]to associate with RF in Drosophila cells cannot be attributed to improper expression or misfolding. The fact that no localization at Drosophila RF was observed does not rule out transient and short lived associations but clearly no distinguishable RF structures assemble as in mammalian cells. These results suggest a fundamental difference between Drosophila and mammalian cells. While replication in Drosophila cells seems to rely on transient interaction with highly mobile factors, mammalian cells seem to have evolved stable higher order structures possibly to cope with the challenges of replicating a genome 22 times as big as the fly genome.

4.2. Targeting of DNMT1 to RF in mammalian cells is driven solely by the PBD

Analysis of the association of various deletions of DNMT1 fused to GFP/YFP with RF showed that the PBD is sufficient for targeting of DNMT1 to RF (Fig. 3.16 A-E). Contrary to earlier reports (Liu et al, 1998), PBHD was unable to associate with RF in any of the cells in S phase (Fig. 3.16F). However, in their work they observe only „partial“ targeting of PBHD to RF in „some“ cells. We observe that the PBHD-YFP fusion forms some aggregates in the nucleus in overexpressing cells (judged by YFP fluorescence intensity) and as a matter of chance these aggregates could colocalize with RF. Moreover, PBHD (also called BAH domain) is present in many proteins involved in chromatin remodeling and gene silencing (Callebaut et al., 1999). The only other example of a protein involved in DNA replication that contains the PBHD/BAH is S. cerivisiae OrcI where the BAH domain is dispensable for DNA replication while it is required for transcriptional silencing (Bell et al., 1995). Thus similar to these proteins, PBHD/BAH domain in DNMT1 might have a role pertaining to chromatin alterations rather than having a role in targeting the protein to replication structures. The crystal structure solution of the BAH domain from S. cerevisiae Orc1 suggests it is primarily a protein interaction domain whose specificity is determined by a highly divergent region which juts out of a main scaffold formed by the conserved region (Zhang et al., 2002). It is not yet known what binds to the PBHD/BAH domain of DNMT1. However, it has been shown that HDAC1 interacts with DNMT1 and a part of this interaction domain lies within the conserved region of PBHD/BAH of DNMT1 (Fuks et al., 2000). Thus, based on these reports the PBHD domain in DNMT1 might be involved in some chromatin remodeling activities.

Here we observe that the TS associates only with late-RF (Fig. 3.16G, H), which is consistent with the previous observation that the TS associates mostly with late-RF (Leonhardt et al., 1992). Further analysis of the temporal course of TS association with late-RF showed that this association is attributable to the affinity of the TS for pericentric heterochromatin (Fig. 3.19). Thus, the TS does not have any inherent ability to associate with RF per se, and the observed association with late-RF is coincidental with its association with pericentric heterochromatin.

Taking together these observations on the ability of the three reported RFTS from DNMT1 to target to RF, it is clear that PBD is the only domain which mediates targeting of DNMT1 to RF. A point mutation in the PBD (H at position 167 was mutated to V) to a form that abrogates PCNA binding ability (Chuang et al., 1997) resulted in reduced association of the mutant GMT-PBD-H167V with replication sites (Fig. 3.23A and Table 3.2). About 30% of cells in S phase showed association of [Seite 83↓]GMT-PBD-H167V with RF. We attribute this to some weak binding of the mutated PBD with PCNA. It was suggested that the PBD from DNMT1 might play a role in targeting to RF during early-S while the other domains would target DNMT1 to RF during other S phase stages (Chuang et al., 1997). In opposition to this view, our results show that PBD solely mediates association of DNMT1 with RF throughout S phase.

4.3. The TS mediates cell cycle dependent binding of DNMT1 to pericentric heterochromatin

Previous studies have shown that DNMT1 forms a punctate pattern colocalizing with RF during S phase (Leonhardt et al., 1992), and is diffused in the nucleus in G2 (Rountree et al., 2000). In this study we observe that the TS domain alone showed a strong affinity for pericentric heterochromatin and remains associated with these sites in later cell cycle stages (Fig. 3.16 and Fig. 3.19). The subnuclear localization of DNMT1 and deletion mutants throughout the cell cycle was analyzed by two independent approaches: by immunostaining of endogenous DNMT1 in fixed cells at different cell cycle stages and by following the dynamics of a GFP-DNMT1(s) fusion protein (GMT) in live cells throughout the cell cycle. Both approaches unequivocally showed that from late-S through G2 and M until early-G1 DNMT1 is present at pericentric heterochromatin. It is diffused in the nucleus during G1 phase and associates again with the RF in S phase (Fig. 3.20 and Fig. 3.21). During M phase, endogenous DNMT1 (Fig. 3.20E) and GMT (Fig. 3.21) apparently are bound to the whole chromatin and is not confined only to the pericentric heterochromatin unlike F-TS-GFP, which is concentrated at pericentric heterochromatin (Fig. 3.19). This could be due to the other functional domains in DNMT1. Deletion of the TS from DNMT1 resulted in loss of association with pericentric heterochromatin during G2 and M phases (Fig. 3.22) indicating that the TS is solely responsible for retention of DNMT1 at these sites following S phase in G2 and M phases. Importantly, this deletion mutant efficiently associated with the RF at pericentric heterochromatin during late S phase, which should be mediated by the PBD. Thus, the TS is responsible for the specific association of DNMT1 with late-replicating pericentric heterochromatin during G2 and M.

The TS domain alone and the full length DNMT1 display differing subnuclear localization throughout the cell cycle. Whereas TS-GFP has a strong affinity for pericentric heterochromatin throughout the cell cycle (Fig. 3.19), DNMT1 is present at these sites only during late S phase and is retained here through G2 and M, to be released in G1 (Fig. 3.20 and Fig. 3.21). Thus, the activity of the TS in DNMT1 is controlled in such a way that it can associate with its target sites only during late-S, G2 and M. By analyzing the subnuclear localization of a series of point and deletion mutations in the DNMT1 fused to GFP throughout the cell cycle, we observe that a point mutation in PBD, abrogating PCNA binding ability (Chuang et al., 1997), resulted in reduced association of the mutant with replication sites and led to association with pericentric heterochromatin right during early- and mid-S phase (Table 3.2, Fig. 3.23). This pattern resembles the localization of the TS-GFP fusion protein. Thus, during S phase, PBD controls the localization of DNMT1 and overrides the strong affinity of the TS for pericentric heterochromatin. Three possibilities can be [Seite 84↓]envisaged to mediate this overriding effect of the PBD over the TS: (i) The PBD-PCNA interaction at the replication fork has a higher affinity and this might be dominant over the TS binding to pericentric heterochromatin. However, it is observed that overexpressed GFP-DNMT1 fusion does not associate with pericentric heterochromatin during early-/mid-S phase indicating that the overriding effect of the PBD over the TS cannot be explained as a simple competition between PCNA for the PBD and the pericentric heterochromatin for the TS. (ii) The TS is masked by other domains in DNMT1 which prevents its binding to pericentric heterochromatin. Binding of PCNA would result in a conformational change resulting in unmasking of the TS that makes it available for interaction with pericentric heterochromatin. (iii) Interaction of DNMT1 with PCNA allosterically activates the TS allowing binding to pericentric heterochromatin. In the latter two possibilities, the TS would get unmasked or allosterically activated only at the RF upon binding to PCNA. Thus the TS would get an opportunity to bind pericentric heterochromatin only during replication of these regions. Further experiments are required to sort out these three possibilities.

A fusion protein lacking the PBHD behaved similar to GFP-DNMT1 throughout S, G2 and M but was retained at pericentric regions during G1 (Table 3.2, Fig. 3.23 (ii)). Thus, the PBHD domain via its yet unidentified interacting partner seems to play a role in the release of DNMT1 from pericentric heterochromatin during G1.

4.4. Alternative mechanism of inheritance of DNA methylation in plants and fungi

It is interesting to note that maintenance of the two forms of epigenetic information, nucleosome state in S. cerevisiae (Zhang et al., 2000) and DNA methylation (this study), both are linked to PCNA. This makes sense considering the central place PCNA occupies in the replication machinery. Since DNA replication involves opening of the chromatin structure and formation of nascent unmethylated DNA, the proteins involved in epigenetic maintenance might have evolved in such a way that their activity is coupled to the DNA replication process itself. Sequence analysis aimed at identifying a PBD in the DNMT1 homologues in plants and fungi showed that they lack such a motif (Fig. 3.26) while the N-terminal domain of these proteins have the TS (duplicated in the case of plants; illustrated in Fig. 4.1) and the PBHD/BAH domains (Fig. 3.28) (Callebaut et al., 1999). Considering the conservation of PBD from archaebacteria to higher eucaryotes (Warbrick et al., 1998) (Dalrymple et al., 2001), the absence of a PBD in the DNMT1 homologues in plants and fungi is intriguing and suggests that maintenance of DNA methylation is uncoupled from the very process of DNA replication in these organisms. We propose that this could reflect the role DNA methylation has in these organisms and the distribution of methylated residues in the genome. In mammals, DNA methylation occurs in repetitive elements (transposons and other repeats) and also in coding regions except for CpG islands of active promoters (Yoder et al., 1997b). Methylation of coding regions is proposed to play a role in transcriptional regulation and there exists evidence correlating tissue specific methylation of promoters with transcriptional silencing in mammals (Futscher et al., 2002). The distribution of DNA [Seite 85↓]methylation in repeats as well as coding regions would have demanded from DNMT1 a mechanism that scans the whole genome indiscriminately for methylation patterns, which have to be inherited in the newly synthesized DNA. This could be best accomplished by targeting DNMT1 to sites of DNA replication, thereby creating a selective advantage confered by the PBD in DNMT1. In plants and fungi, genomic methylation is restricted mostly to transposons and other repeat elements (Rabinowicz et al., 1999) (Goyon et al., 1996), which are typically excluded from gene rich regions. Here, it is believed that DNA methylation mainly protects host genome from insults by invading foreign DNA/transposons (Martienssen and Colot, 2001). We hypothesize that, in these organisms, the modus operandi of the DNA methylation machinery might be by specifically targeting to repeats. This could be mediated by the TS sequence in the DNMT1 homologues in plants and fungi, which is supported by our observation that the TS has an innate ability to target to pericentric repeats. Thus, maintenance of DNA methylation in plants and fungi might be operating by an alternative, and probably evolutionarily older mechanism that targets DNMT1 to repeat elements. Consistent with this hypothesis is the observation that an N-terminal region in Dim2, required for methylation of duplicated genes inactivated by RIP (repeat induced point mutation) (Kouzminova and Selker, 2001), has a TS-like motif identified in this study. Dim2 is a DNMT1 homologue in N. crassa that is required for all known DNA methylation in this fungus. Sequence analyses using a profile of the conserved motif in TS detected the TS motif in the N-terminal domain of Dim2 at an analogous position as that of TS in Masc2 (illustrated in Fig. 4.1), which is the DNMT1 homologue in the fungus A. immersus. In this context, it is worth pointing out that DNMT1 might also function as a de novo methyltransferase by acting as the primary step in methylating foreign DNA/repetitive elements by virtue of being targeted to DNA repeats. This is supported by deletion of dim2 in N. crassa which causes complete loss of all known DNA methylation indicating that it is a de novo methyltransferase (Foss et al., 1993). Thus, TS mediated targeting to repetitive elements might be a conserved mechanism for silencing foreign DNA by both de novo and maintenance methylation. It would be very interesting to know how this targeting occurs. In the case of human DNMT1, the TS harbours a Zn binding domain and a bipartite DNA binding domain (Zn-2 and DB in Fig. 3.14) identified in vitro (Chuang et al., 1996), which might be involved in targeting to repetitive elements. However, the corresponding region from the closely related murine DNMT1 is reported to have no DNA binding activity. Moreover, extensive deletion analysis of the TS (Leonhardt et al., 1992) has shown that presence of the DNA binding domain and the Zn binding region is not sufficient for association with pericentric heterochromatin. Also, deletion of one of the motifs of the bipartite DNA binding domain does not abrogate association with pericentric heterochromatin (TS-ß-gal, Fig. 3.16i). Alternatively, targeting of TS to repetitive DNA elements might occur via protein-protein interactions with other not yet identified proteins bound to the TS. Since the TS associates with pericentric heterochromatin at any cell cycle stage, these factors should be constitutive components of these heterochromatin regions.


[Seite 86↓]
Fig. 4.1. Domain organization in DNMT1 from metazoan, plants and fungi. The TS is duplicated in plants. A TS-like motif identified in N. crassa Dim2 is also illustrated. BAH corresponds to PBHD. Names of organisms are abbreviated as in Fig. 3.26.

Our hypothesis that plant and fungal DNMT1 are not targeted to RF is based only on sequence analyses and it could still be possible that the DNMT1 homologues in plants and fungi are targeted to RF by a mechanism independent of the PBD. It will be interesting to test this experimentally by studying the subnuclear distribution of DNMT1 during S phase and other cell cycle stages in plants and fungi. It would be also interesting to analyze the organization of functional domains of the DNA methyltransferases in the invertebrate chordate Ciona intestinalis in which the DNA repeats in the genome are not methylated while genes are methylated (Simmen et al., 1999).

4.5. Mechanism of maintenance of epigenetic information by DNMT1 in mammals

Genetic information is modulated by two epigenetic marks, DNA methylation and histone modifications, which have to be maintained at every round of DNA replication for controlled gene expression and genomic integrity. It has been shown that DNMT1 interacts with HDAC1 and 2 which has led to the proposal that DNMT1 plays a role in maintenance of deacetylated histones in late-replicating heterochromatin (Robertson et al., 2000) (Rountree et al., 2000). Based on our results, we propose the following model describing the mechanism of epigenetic maintenance by DNMT1 (Fig. 4.2). As the cell enters S phase, PCNA assembles throughout the nucleus at sites where DNA replication initiates (Leonhardt et al., 2000) (Somanathan et al., 2001) and DNMT1 associates with these sites via its PBD (Step-1, Fig. 4.2). DNMT1, with its preference for hemi-methylated DNA (Gruenbaum et al., 1982) (Bestor and Ingram, 1983) (Yoder et al., 1997a), then methylates the newly replicated DNA. This dynamic distribution of DNMT1 follows that of PCNA throughout S phase. At late RF, where pericentric alpha- and gamma-satellite DNA also replicate (Ten Hagen et al., 1990) (O'Keefe et al., 1992) (Leonhardt et al., 1992), DNMT1 is primarily recruited via interaction with PCNA followed by binding to a yet unknown but constitutive component of pericentric heterochromatin via its TS (Step-2, Fig. 4.2). At the cessation of DNA replication, PCNA disassembles from these sites while DNMT1 is retained there (Step-3, Fig. 4.2). We propose that the TS mediated retention of DNMT1 at late replicating pericentric heterochromatin throughout G2 has the following roles in the maturation of centromeric regions: (i) By being retained at pericentric heterochromatin throughout G2, DNMT1 would complete methylation of [Seite 87↓]newly synthesized DNA (Step-4, Fig. 4.2). This is important because DNMT1 being catalytically slow (Pradhan et al., 1999) would require more time to finish methylation of the densely methylated pericentric heterochromatin. This is supported by observations that a significant amount of micrococcal nuclease resistant DNA is methylated several hours following DNA synthesis (Geraci et al., 1974) (Woodcock et al., 1982) (Davis et al., 1985). (ii) Retention of DNMT1 at pericentric heterochromatin during G2 might play a role in the formation of condensed chromatin by recruiting HDACs which would deacetylate the newly deposited histones. It has been proposed that DNMT1 mediates chromatin maturation by recruiting HDAC2 to late RF (Rountree et al., 2000). Our observation that TS mediates retention of DNMT1 at pericentric regions during G2 supports this proposal and extends this proposed role of DNMT1 in chromatin maturation to G2.

Fig. 4.2. Hypothetical model explaining the mechanism of inheritance of DNA methylation pattern by DNMT1. The events at a replication fork during early-S and late-S are shown. (1) DNMT1 is recruited at early replication foci interacting with PCNA via the PBD. The catalytic domain (cat) in DNMT1 methylates the newly synthesized DNA strand. (2) At late replication foci, DNMT1 is again recruited by PCNA and methylates the newly synthesized strand. (3) At the cessation of DNA replication, PCNA disassembles from the DNA strand but DNMT1 is retained at these sites via the TS. DNMT1 continues methylating the pericentric heterochromatin. (4) During G2, DNMT1 is retained at the pericentric regions where it completes DNA methylation and acts as a docking site for histone modifying enzymes like HDAC1/2 and thereby aids in maturation of newly synthesized heterochromatin.

The in vivo role of TS in maintaining and/or establishing DNA methylation patterns by DNMT1 is supported by rescue experiments with ES cells deficient in DNMT1 (Dnmt1-/-). Wild type ES cells are capable of differentiation while Dnmt1-/- ES cells are not. This inability to differentiate can be rescued by expression of Dnmt1 but not mutant forms that lack the TS (Gaudet, F. and Leonhardt, H., unpublished results). Deletion of the TS in DNMT1, however, does not abolish catalytic activity [Seite 88↓](Margot et al., 2000). Thus, the role of the TS in mediating association of DNMT1 with pericentric heterochromatin is essential for maintenance of DNA methylation in vivo. Furthermore, overexpression of TS in mammalian cells, likely preventing the access of endogenous DNMT1 to pericentric heterochromatin, causes large scale chromatin reorganization and increases micronuclei formation suggestive of genomic instability (Fig. 3.24 and Fig. 3.25). Thus the association of DNMT1 with heterochromatin mediated by the TS is vital for the maintenance of epigenetic information and the organization and stability of the genome.


© Die inhaltliche Zusammenstellung und Aufmachung dieser Publikation sowie die elektronische Verarbeitung sind urheberrechtlich geschützt. Jede Verwertung, die nicht ausdrücklich vom Urheberrechtsgesetz zugelassen ist, bedarf der vorherigen Zustimmung. Das gilt insbesondere für die Vervielfältigung, die Bearbeitung und Einspeicherung und Verarbeitung in elektronische Systeme.
DiML DTD Version 3.0Zertifizierter Dokumentenserver
der Humboldt-Universität zu Berlin
HTML-Version erstellt am:
11.05.2004