[Seite 33↓]

3.  Results

3.1. Association of proteins with RF in Drosophila

3.1.1. PCNA is highly conserved in S. cerevisiae, D. melanogaster and mammals

The mechanistic basis for the association of replication factors with RF has been proposed to be mediated via interaction of PBD with PCNA. PCNA functions as a processivity factor for the replicative DNA polymerases (Bravo et al., 1987) (Prelich et al., 1987). To understand whether the PBD-mediated association of proteins with RF is conserved in evolution, we first analyzed the conservation of PCNA protein sequence from divergent eucaryotes, especially the regions shown to interact with PBD of p21Cip1/Waf1 (Gulbis et al., 1996). PCNA is highly conserved in organisms as divergent as yeast, flies and mammals (Fig. 3.1). The residues of PCNA that interact with PBD of p21Cip1/Waf1 are more than 75% conserved in human, Drosophila and yeast. Also the PBD is conserved in homologous proteins from archaebacteria, yeast, worms, flies, amphibians and mammals (Fig. 3.2) (Warbrick et al., 1998) suggesting that the association of proteins to RF mediated by PCNA is a mechanism conserved through evolution. We put this hypothesis to test by analyzing whether the PBD of DNMT1 can associate with RF in Drosophila cells (S2 cells). Further, nothing is known about the mechanism underlying the association of the TS and PBHD of DNMT1 with RF. To understand whether these mechanisms are conserved in divergent eucaryotes, we have analyzed the ability of the TS and PBHD to target to RF in Drosophila cells (S2 cells) as it lacks DNA methylation. (see Section 1.3.5).


[Seite 34↓]

Fig. 3.1. Drosophila and human PCNA are highly similar. Sequence alignment of HsPCNA (human), DmPCNA (Drosophila) and ScPCNA (budding yeast). The protein sequences were aligned using the Jotun-Hein method (in Lasergene software). Residues identical to HsPCNA are boxed. HsPCNA is about 35% identical to ScPCNA and 70% identical to DmPCNA. Regions in HsPCNA that interacts with the PBD of p21 (Gulbis et al., 1996) are in red.


[Seite 35↓]

Fig. 3.2. A conserved PBD in proteins involved in DNA metabolism. Alignment of the PBD from various homologous proteins from different organisms. Accession numbers are indicated in the second column where available. Asterisk denotes the beginning or end of the protein. Identical residues are highlighted in red and conserved substitution is highlighted in green. Association with replication foci requires both the conserved residues and the stretch of peptide rich in basic amino acids (in blue) (Montecucco 1998). (Hs: Homo sapiens; Mm: Mus musculus; Xl: Xenopus laevis; Dm: Drosophila melanogaster; Ce: Caenorhabditis elegans; Sp: Schizosaccharomyces pombe; Sc: Saccharomyces cerevisiae; Mj: Methanococcus jannaschii).

3.1.2. Human PCNA and Drosophila PCNA can associate with RF across the two organisms

To answer our main question whether the three RFTS from mouse DNMT1 can function in Drosophila, we had to first establish whether the replication machinery in the two systems were similar in that a conserved core replication factor from one organism would be targeted to the RF in the heterologous system. To this end, we analyzed the association of human PCNA (HsPCNA) with RF in Drosophila cells (S2 cells) and Drosophila PCNA (DmPCNA) in mouse cells (C2C12 cells). Plasmids encoding HsPCNA, GFP-HsPCNA and GFP-DmPCNA were constructed and introduced into cells by calcium phosphate transfection. Subnuclear localization of HsPCNA in transfected S2 cells was accomplished by indirect immunostaining with anti-PCNA antibody (FL261, Santacruz) that reacted specifically with HsPCNA (Fig. 3.3A). Transfected cells in S-phase showed punctate BrdU staining representing RF and HsPCNA colocalized with these foci (Fig. 3.3B). Also GFP-HsPCNA is targeted to RF in S2 cells (Fig. 3.3B). Similarly, C2C12 cells expressing GFP-DmPCNA showed complete colocalization of GFP-DmPCNA with RF (Fig. 3.3C). Since the available antibody against DmPCNA cross reacts with mouse PCNA, localization of untagged DmPCNA in mouse cells was not possible. Owing to the high sequence similarity of HsPCNA with DmPCNA (Fig. 3.3A) and the conserved [Seite 36↓]structure of PCNA from divergent eucaryotes (Krishna et al., 1994), efficient association of HsPCNA with RF in S2 cells indicates that HsPCNA must be replacing endogenous PCNA in S2 cells. Thus, these results indicate that the basic replication machinery in the two organisms is conserved and core replication factors are inter-changeable.


[Seite 37↓]

Fig. 3.3. Mammalian and Drosophila PCNA associate with RF interchangeably between these divergent organisms. Drosophila S2 and mammalian C2C12 cells were transfected, pulse labeled with BrdU to label sites of active DNA replication, and fixed with formaldehyde followed by immunostaining. (A) S2 cells transfected with plasmid encoding HsPCNA and coimmunostained with anti-PCNA (FL261; Santa Cruz) and anti-BrdU antibodies. Cells positively staining with anti-BrdU were in S phase at the time of BrdU labeling. Anti-PCNA antibody (FL261) specifically stains one transfected cell (arrow) expressing HsPCNA that is in S phase and does not stain endogenous PCNA in the other non-transfected S phase and non-S phase cells. (B) S2 cells transfected with plasmid encoding HsPCNA (top) or GFP-HsPCNA (bottom) and immunostained with anti-BrdU. HsPCNA was detected by coimmunostaining with anti-PCNA antibody (FL261) as in (A). Overlay shows colocalization of HsPCNA and GFP-HsPCNA with BrdU foci. (C) C2C12 cells transfected with plasmid encoding GFP-DmPCNA and immunostained with anti-BrdU. GFP-DmPCNA colocalizes with BrdU foci. Scale bar = 2 µm.


[Seite 38↓]

3.1.3.  Subnuclear localization of DNMT1 in Drosophila cells

In order to determine whether the mechanisms by which PBD, TS and PBHD mediate association with RF in mammalian cells (methylated genome) are conserved in Drosophila (unmethylated genome), we evaluated the ability of these domains to associate with RF in S2 cells. S2 cells were transfected with plasmids encoding chimeric fusions of the N-terminal domain of DNMT1 or the PBD/TS/PBHD with GFP/YFP/ß-gal epitope (summarized in Fig. 3.4). As shown in Fig. 3.5A, NMT2-GFP associates with RF in mouse C2C12 cells. NMT2-GFP has a PBD and therefore it is predicted that this fusion protein would be targeted to RF in Drosophila S2 cells. Surprisingly, NMT2-GFP does not colocalize with RF in S2 cells (Fig. 3.5B). Rather, in S2 cells, NMT2-GFP seems to be diffused throughout the nucleus as well as aggregating in some regions to form large structures. To test whether this is an artifact of the fusion protein in S2 cells, S2 cells were transfected with a plasmid encoding full length untagged DNMT1(s) and the association of DNMT1(s) with RF was analyzed. Like NMT2-GFP, DNMT1(s) is not associated with RF (Fig. 3.5C top panel) in the majority of cells, though in some cells in S-phase DNMT1(s) seemed to partially colocalize with RF (Fig. 3.5C; bottom panel), or with a subset of Drosophila chromatin.

It is surprising that a PBD-containing protein that efficiently associates with RF in mammalian cells does not behave similarly in Drosophila cells even though their basic replication machinery is similar. Reduced efficiency in the targeting of DNMT1 to RF in Drosophila cells could be an artifact of a protein which is very foreign to Drosophila as it lacks a DNMT1 homologue. So we tested the targeting of human DNA Ligase I (HsDNA Lig I), which is a core replication factor involved in ligating Okazaki fragments during synthesis of lagging strand, to RF in S2 cells. DNA Ligase I is conserved in evolutionarily distant organismsand the Drosophila DNA Ligase I (DmDNA Lig I) homologue is 50% identical to HsDNA Lig I (Fig. 3.8). S2 cells were transfected with a plasmid encoding a GFP fusion to HsDNA Lig I (GFP-HsLig) and the subnuclear localization of GFP-HsLig with respect to RF was analyzed. GFP-HsLig showed a diffused nuclear distribution in S2 cells that are in S phase and did not show any discernible colocalization with RF (Fig. 3.6A). In contrast, consistent with previous studies demonstrating association of human DNA Ligase I with RF (Cardoso et al., 1997), GFP-HsLig is efficiently targeted to RF in mouse C2C12 cells (Fig. 3.6B). Unlike NMT2-GFP and DNMT1(s) that aggregates in structures, GFP-HsLig showed a diffused nuclear distribution in all S2 cells observed. This aggregation could be due to the various other functional domains present in the N-terminus of DNMT1. Thus, taking together the subnuclear localization of NMT2-GFP, DNMT1(s) and GFP-HsLig in S2 cells, it appear with plasmid encoding HsPCNA aefficiently targeted to sites of DNA replication in Drosophila cells. Strikingly, the inability of GFP-HsLig to associate with RF in any of the S2 cells in S phase indicates that PBD is unable to function as an RFTS in Drosophila. Partial association of DNMT1(s) with RF in some S2 cells in S phase might mean that the other two RFTS in DNMT1 might be able to target the protein to RF, although at a low efficiency.


[Seite 39↓]

Fig. 3.4. Deletions of DNMT1 used for evaluating the ability of PBD, TS and PBHD to associate with RF in S2 cells. The somatic isoform, DNMT1(l), is illustrated at the top. Various deletions in the N-terminus fused to GFP/YFP/ß-gal epitope are shown. The predicted molecular weights of the fusion proteins are indicated. See Fig 3.12 for Western blot analysis of the fusion proteins. The boundaries of the different domains and deletions are indicated by the amino acid numbers corresponding to the DNMT(l) isoform. See Fig. 1.7 for description of the domains. An SV40 NLS was added in PBHD-YFP for efficient nuclear targeting.

Next, we asked whether the weak targeting of DNMT1(s) to RF in S2 cells can be attributed to the TS and PBHD. In order to answer this, we determined the ability of each of the RFTS from DNMT1 to associate with RF in S2 cells. The plasmid constructs used are summarized in Fig. 3.4. The PBD from DNMT1 and HsDNA Lig I were fused individually to GFP to test the ability of the minimal PBD to associate with RF in Drosophila cells. As shown in Fig. 3.7A, MTPBD-GFP and HsLigPBD-GFP both are diffusely distributed in the nucleus and did not show anyidiscernible localization or exclusion from the RF. The same constructs when expressed in mouse cells show a similar diffused pattern in many cells but are clearly targeted to RF in other cells (see Section 3.2). In S2 cells, none of the transfected cells in S phase showed targeting of both PBDs fused to GFP. Thus, even though the 10 aa PBD plus the 10 aa region rich in basic residues are enough for targeting to RF as observed in mouse cells, it is unable to do so in Drosophila cells. Ability of the PBD to assemble at replication sites requires interaction with PCNA (Chuang et al., 1997) (Montecucco et al., 1998). Inability to target might also be because the PBD from mammalian origin is not able to interact with the endogenous DmPCNA in S2 cells. Although this is unlikely as the PCNA is highly conserved between human and Drosophila (Fig. 3.1), and the PBD from Drosophilaproteins, like p21Cip1/Waf1 and POGO transposase, are similar to mammalian PBD (Fig. 3.2) and can efficiently interact with DmPCNA in vitro (Warbrick et al., 1998). We tested whether HsPCNA would aid targeting of mammalian PBD to RF in Drosophila by co-transfecting plasmids encoding MTPBD-GFP and HsPCNA in S2 cells. As can be seen in Fig. 3.7B, MTPBD-GFP still remained diffused in the nucleus while HsPCNA was associated with RF. Thus, the inability of mammalian PBD to target to RF in Drosophila is not due to the absence of its natural interacting partner (HsPCNA).

The previously described TS-ß-gal epitope tagged fusion used for the mapping of the RFTS in DNMT1 (Leonhardt et al., 1992) (Fig. 3.4) was used to test the ability of the TS to associate with RF in Drosophila cells. Although, as previously shown [Seite 40↓](Leonhardt et al, 1992), this fusion protein associates with RF in mammalian cells, in S2 cells it did not associate with RF (Fig. 3.7C). This fusion protein is not efficiently imported into the nucleus as can be seen from the strong cytoplasmic staining. However, inability to associate with RF cannot be attributed to this because the TS-ß- gal fusion protein can be clearly detected in the nucleus as well. Moreover, in spite of its weak nuclear localization, it efficiently associates with RF in mammalian cells, whereas none of the S2 cells in S-phase showed association of TS-ß-gal with RF. Similarly, expression of PBHD-YFP in S2 cells showed that the fusion protein is unable to associate with RF in S2 cells (Fig. 3.7D). However, PBHD-YFP did not colocalize with RF even in mouse cells (Section 3.2).

Thus, none of the three RFTS from DNMT1 is able to associate with RF in Drosophila cells. Most often these fusion proteins showed a diffused nuclear distribution. Inability of these domains to associate with RF could be due to the following reasons: (a) Possibly these ectopically expressed proteins are produced in high amounts and as a result, the excess protein remains diffused throughout the nucleus making it difficult to discern the population which associates with RF andthat which is unbound and present in the nucleoplasm. However, this seems unlikely because the amount of protein produced is quite low as judged by the weak fluorescence intensity in transfected cells. Also, it is difficult to detect protein expression in transfected S2 cells by western blotting as compared to mammalian cells. Moreover, in S2 cells ectopically expressed HsPCNA and GFP-HsPCNA associates with RF (Fig. 3.3B). (b) Alternatively, in Drosophila, the mechanism behind recruitment of proteins at RF may not be dependent on the PBD, but rather some other unique protein motif might function as an RFTS.


[Seite 41↓]

Fig. 3.5. Subnuclear localization of NMT 2 -GFP and DNMT1(s) with RF in S2 cells. Cells were transfected and pulse labeled with BrdU 24 hrs later followed by indirect immunostaining for BrdU. (A) C2C12 cells transfected with plasmid encoding NMT2-GFP. Colocalization of NMT2-GFP with BrdU is seen in the overlay as yellow. (B) S2 cells transfected with plasmid encoding NMT2-GFP. NMT2-GFP does not colocalize with BrdU. In most cells it aggregates to form structures (arrows in bottom panel) that are excluded from BrdU incorporation sites. (C) S2 cells transfected with a plasmid encoding DNMT1(s). Cells were immunostained with anti-pATH52 to detect DNMT1. In 80-90 % of cells, DNMT1 is excluded from BrdU sites (top panel). In about 10% of S2 cells, DNMT1 partially colocalizes with BrdU incorporated sites (bottom panel). Scale bar = 2 µm


[Seite 42↓]

Fig. 3.6. GFP-HsLig does not associate with RF in Drosophila S2 cells. Cells were transfected with plasmid encoding GFP-HsLig and pulse labeled with BrdU 24 hrs later followed by indirect immunostaining for BrdU. (A) In S2 cells, GFP-HsLig shows diffused nuclear localization in all cells observed. (B) GFP-HsLig colocalizes with BrdU labeled RF in C2C12 cells seen as yellow colour in overlay. Scale bar = 2 µm


[Seite 43↓]

Fig. 3.7. PBD, PBHD and TS do not mediate association with RF in S2 cells. S2 cells were transfected and pulse labeled with BrdU 24 hrs later followed by indirect immunostaining for BrdU. (A) S2 cells transfected with plasmid encoding MTPBD-GFP (top panel) or HsLigPBD-GFP (bottom panel). Both PBD-GFP fusions show diffused nuclear distribution. (B) S2 cells cotransfected with plasmids encoding MTPBD-GFP and HsPCNA. Presence of HsPCNA does not aid association of MTPBD-GFP with RF as MTPBD-GFP still shows diffused nuclear distribution. (C) S2 cells transfected with plasmid encoding TS-ß-gal. ß-gal epitope was detected by indirect immunostaining with anti-ß-gal antibody. (D) S2 cells transfected with plasmid encoding PBHD-YFP. Both TS-ß-gal and PBHD-YFP show diffused nuclear distribution. Scale Bar = 2 µM.


[Seite 44↓]

3.1.4.  Search for an RFTS in Drosophila

In order to determine whether another protein sequence is functioning as an RFTS in Drosophila, we sought to clone a replication protein from Drosophila and identify the RFTS in it. Since DNA Ligase I is a well conserved protein from yeast to humans, the Drosophila genome database and the EST database (BDGP) were searched for putative DNA Ligase I homologues with DNA Ligase I sequence from different organisms (mouse, human, yeast) as query. Three putative homologues were identified in the EST database, viz. LD41868. AT15112, and LD06019 with corresponding predicted proteins CG5602, CG17227 and CG12176 respectively (Fig. 3.8). Comparison of the protein sequences of each of these putative homologues with HsDNA Lig I showed that CG5602 has maximum identity (53.7 %) to HsDNA Ligase I spanning over a large region of the protein as compared to CG17227 (31% identical) and CG12176 (30.3% identical) (Fig. 3.8A). Moreover, comparison of each of the three putative Drosophila homologues with the three mammalian DNA ligases, viz. Ligase I, Ligase III and Ligase IV, indicated that CG17227 and CG12176 are homologues of Ligase III and Ligase IV respectively (Fig. 3.8B). Also, a phylogenetic analysis of the human and putative Drosophila ligases showed that CG5602 is closer to HsDNA Lig I (Fig. 3.8C). Thus, these sequence analyses indicate that the putative DNA ligase I in Drosophila is CG5602 (henceforth called DmDNA Lig I). The EST clone (LD41868) coding for CG5602 was obtained and the sequence was verified. An 81 bp segment in the coding region of the predicted gene was found to be absent in the cDNA clone (Fig 3.9A). The DmDNA Lig I protein sequence was analyzed in order to identify the presence of sequences similar to the PBD or the other RFTS from DNMT1. This revealed the presence of a PBD sequence (designated as DmDNA Lig I-PBD) that had all the features of HsDNA Lig I-PBD required for association with replication foci, viz. the 10 amino acid region interacting with PCNA immediately followed by the 10 amino acid sequence rich in basic residues (Fig. 3.9B). Additionally, two sequences with weak similarity to the conserved PBD were identified and called as PBD-1 and -2 . No sequence with similarity to PBHD and TS were observed. The DmDNA Lig I protein sequence was searched for potential NLS using signature NLS sequences like Caudal NLS, Engrailed NLS, SV40 Large T NLS, Nucleoplasmin NLS, and as well the NLS from HsDNA Ligase I (Montecucco et al., 1995) (Cardoso et al., 1997). Two regions in DmDNA Lig I showed similarity to the nucleoplasmin and T antigen NLS (labeled as NLS-1 and NLS-2, Fig. 3.10A).


[Seite 45↓]

Fig. 3.8. Basis for selection of the EST clone LD41868(CG5602) as the putative DNA ligase I homologue in Drosophila. BLAST search of the D. melanogaster genome and EST database with DNA Ligase I (from different species) identified three similar sequences CG5602, CG17227 and CG12176. (A) Each of the putative DNA ligase protein sequences identified in Drosophila is aligned pairwise with HsDNA Lig I. The percentage identity obtained from the alignment is indicated at the right. CG5602 shows maximum identity to HsDNA Lig I compared to the other two sequences. (B) Table showing the percentage identity obtained from pairwise alignment of the three putative DNA ligase homologues in Drosophila with each of HsDNA Ligase I, III and IV. CG17227 and CG12176 show more identity with HsDNA Lig III and IV respectively. (C) Phylogenetic tree from the comparison of HsDNA ligases and the Drosophila homologues created in Lasergene program (J. Hein method with PAM250 residue weight table. Scale represents the amino acid substitution).


[Seite 46↓]

Fig. 3.9. DmDNA Lig I gene, putative protein and potential PBD. (A) Structure of DmDNA Lig I gene and the predicted protein. The gene structure is illustrated at the top and position of exons are marked by nucleotide numbers. The regions shown to be spliced out are the introns (white boxes). A region in the 5-th exon (marked gray) was found to be absent in the cDNA clone. This region was not detected as an intron using programs to detect introns. Middle shows the LD41868 cDNA clone obtained from ResGen. The translational start is shown by the bent arrow and the protein produced from this cDNA clone is illustrated at the bottom. (B) The DmDNA Lig I protein sequence was searched for presence of a PBD (including the deleted region) using ProfileSearch. The sequence which showed all the features of a PBD essential for association with RF is labeled as DmDNA Lig I-PBD. The other two hits show weak similarity. Asterisk indicates the beginning or end of the protein sequence.


[Seite 47↓]

Fig. 3.10. Deletion constructs used for mapping of the RFTS in DmDNA Lig I. (A) Organization of domains in DmDNA Lig I is shown at the top. Various deletions of DmDNA Ligase I fused to GFP at the N-terminus were made as described in Methods. The predicted molecular weight of the fusion proteins is indicated on the left. (B) Western blot analysis of the chimeric proteins confirmed their protein sizes. COS7 or 293T cells were transfected with the fusion constructs shown in (A) and after 24 hrs of incubation the cells were lysed by boiling in Laemmli's loading buffer and analyzed by immunoblotting with rabbit polyclonal antibody against GFP.

In order to determine whether DmDNA Lig I employs some other protein sequence as an RFTS instead of the conserved PBD, a series of deletion mutants of DmDNA Lig I fused to GFP were made (Fig. 3.10A) and their subnuclear localization with respect to RF was analyzed in both mouse and Drosophila cells. Immunoblotting analysis showed that proteins of the right size were expressed from these plasmids [Seite 48↓](Fig. 3.10B). The full length protein fused to GFP (GdL) did not show association with RF in S2 cells (Fig. 3.11A), whereas it colocalized with RF in C2C12 cells (Fig. 3.11B). None of the deletions, GdL-1, -2, -3 and –4, showed colocalization with RF in both S2 and C2C12 cells (Fig. 3.11A and B). Although GdL-2 and GdL-4 are not efficiently targeted to the nucleus, lack of association with RF cannot be attributed to this as these fusion proteins are not completely excluded from the nucleus. Thus, neither the full length DmDNA Lig I nor any of the deletions fused to GFP associated with RF in Drosophila cells. The fact that only GdL, but none of GdL-1 to 4, associates with RF in mouse cells indicates that this association is mediated by the DmDNA Lig I-PBD. Most importantly, it shows that GdL is folded properly and inability of GdL to associate with RF in Drosophila cells cannot be due to improper folding of the protein.

Fig. 3.11. Subcellular localization of the DmDNA Lig I deletions fused to GFP. Cells were transfected with plasmids encoding the various deletions of DmDNA Lig I fused to GFP and pulse labeled with BrdU after 24 hrs followed by indirect immunostaining for BrdU. (A) None of the fusion proteins associate with RF in S2 cells. (B) Only the full length DmDNA Ligase I fused to GFP (GdL) associates with RF in C2C12 cells. GdL-2 and GdL-4 are not efficiently taken up into the nucleus. Scale Bar = 2 µm.


[Seite 49↓]

The ability of the the minimal DmDNA Lig I-PBD to associate with RF in Drosophila and mouse cells was analyzed. As observed with HsLigPBD-GFP and MTPBD-GFP, DmLigPBD-GFP colocalized with RF in C2C12 cells but showed diffused nuclear distribution in S2 cells (Fig. 3.12A and B). Taken together, these results show that: (i) Dm DNA Lig I associates with RF in mouse cells via the DmDNA Lig I-PBD. (ii) Interestingly, Dm DNA Lig I does not associate with RF in Drosophila cells.

Fig. 3.12. DmLigPBD-GFP does not associate with RF in S2 cells. Cells were transfected with a plasmid encoding DmLigPBD-GFP and pulse labeled with BrdU 24 hrs later followed by indirect immunostaining for BrdU. (A) Colocalization of DmLigPBD-GFP with BrdU in C2C12 cells. (B) DmLigPBD-GFP does not colocalize with BrdU in S2 cells. Scale Bar = 2 µm

Table 3.1 summarizes the ability of the various fusion proteins to associate with RF in Drosophila and mammalian cells. Except for PCNA none of the other replication proteins associate with RF in Drosophila cells. Most notably DmDNA Lig I does not associate with RF in Drosophila cells while it does so in mouse cells. This probably indicates a fundamental difference in the kinetics of association of proteins with RF in Drosophila S2 cells and mammalian cells (see Discussion.


[Seite 50↓]

Table 3.1. Efficiency of association of various fusion proteins with replication foci in Drosophila and mammalian cells

Fusion protein

Percentage cells showing association with RF a

Drosophila S2 cells

Mammalian C2C12 cells

HsPCNA

80% (n = 25)

ND

GFP-HsPCNA

90% (n =20)

100% (n = 20)

GFP-DmPCNA

82% (n = 23)

100% (n = 15)

NMT2-GFP

0% (n = 40)

75% (n = 20)

DNMT1

13% b, (n = 16)

100% (n = 15)

DNMT1 + HsPCNA

15% b, c (n = 27)

ND

GFP-HsDNA Lig I

0% (n = 38)

70% (n = 20)

GFP-HsDNA Lig I + HsPCNA

0% c (n = 19)

ND

MTPBD-GFP d

0% (n = 22)

40% (n = 32)

MTPBD-GFP + HsPCNA

0% c (n = 19)

ND

TS-ß-gal

0% (n =25)

ND

PBHD-YFP

0% (n = 25)

0% (n = 45)

GFP-DmDNA Lig I

0% (n = 40)

62% (n = 24)

a Number in parentheses indicates the number of S phase cells counted.
b Only partial colocalisation with BrdU foci was observed.
c Cotransfection efficiency was about 93%.
d HsLigPBD-GFP and DmLigPBD-GFP behave similar to MTPBD-GFP.


[Seite 51↓]

3.2.  Regulation of subcellular localization of DNMT1 in mammalian cells throughout the cell cycle

3.2.1. Targeting preference of the three targeting domains of DNMT1

Previous studies with synchronized cells have shown that RF form ordered patterns at different stages during S phase (Nakamura et al., 1986) (Nakayasu and Berezney, 1989) (Fox et al., 1991) (O'Keefe et al., 1992). Formation of these patterns has been demonstrated in live cells to involve a gradual and asynchronous assembly and disassembly of the replication machinery (Leonhardt et al., 2000). The spatial pattern of RF indicates the temporal position during S phase (Leonhardt et al., 2000) and this provides a strategy to identify cells at different stages of S phase in an asynchronously growing population. Fig. 3.13 shows the typical patterns of RF in an asynchronous population of C2C12 cells diagnostic of the temporal position in S phase: RF in early-S are evenly distributed throughout the nucleus; in mid-S they are concentrated more around the nucleoli and nuclear periphery; and in late-S they form large toroidal structures at pericentric heterochromatin that is densely stained with Hoechst 33258 (Leonhardt et al., 1992).


Fig. 3.13. Different stages during S phase can be identified in an asynchronously growing population of mammalian cells by the pattern of RF in the nucleus. Mouse C2C12 cells were pulse labeled with 10µM BrdU for 10 min and fixed with ice cold methanol and immunostained with FITC conjugated anti-PCNA antibody (Pharmingen). The cells were fixed again with 3.7% formaldehyde and immunostained with anti-BrdU (Harlan Seralab) antibody followed by detection with Cy5 conjugated secondary antibody. Cells showing different patterns of RF typical of early-, mid- and late-S phases are shown. Non-S phase cells are not labeled with BrdU and show diffused distribution of PCNA. Scale Bar = 2 µm

DNA replication is organized in such a way that the heterochromatin (densely methylated) replicates during late-S phase while euchromatin (sparsely methylated ) replicates earlier (Goldman et al., 1984) (Hatton et al., 1988). Enhanced association of DNMT1 with late-RF might be required for efficient maintenance of the highly methylated DNA in the late replicating heterochromatic regions. In order to determine whether the multiple RFTS in DNMT1 have any selective preference to target to early or late replication foci, various deletions of the N-terminal domain of DNMT1 fused to GFP/YFP were made as summarized in Fig. 3.14A. Western blot analysis of COS7 cells transfected with the various plasmid constructs gave bands at expected sizes indicating that all the fusion proteins are correctly expressed (Fig. 3.14B).


[Seite 52↓]

Fig. 3.14. Deletions of DNMT1 used for studying the preference of PBD, TS and PBHD to associate with RF in mammalian cells. (A) The two isoforms of DNMT1 are shown at the top. DNMT1(l) is the larger somatic isoform and DNMT1(s) is the short isoform expressed in some cells, e.g. oocytes, that differ only in the first 118 amino acids encoding the DMAP binding region. Various deletions of DNMT1(s) were fused to GFP/YFP as listed. The predicted molecular weights of the fusion proteins are indicated. The boundaries of the different domains and deletions are indicated by the amino acid numbers corresponding to the DNMT(l) isoform. The DNMT1(s) isoform was used in making the deletions because the DMAP had no influence on targeting of the protein to RF. See Fig. 1.7 for description of the domains. An SV40 NLS was added in PBHD-YFP for efficient nuclear uptake. F-TS-GFP contains an N-terminal FLAG epitope tag. (B) Western blot analysis of the chimeric proteins showing they are correctly expressed. COS7 or 293T cells were transfected with the fusion constructs shown in (A) and after 24 hrs of incubation the cells were lysed by boiling in Laemmli's loading buffer and analyzed by immunoblotting with the mentioned antibodies.

Firstly, we analyzed whether endogenous DNMT1 is present at RF throughout S phase. C2C12 cells were pulse labeled with BrdU for 10 minutes and immunostained with antibodies against DNMT1 (anti-PATH52 Ab) and BrdU. As shown in Fig. 3.15, endogenous DNMT1 colocalizes with BrdU labeled RF typical of the different stages of S phase. In late-S phase cells two populations of RF are observed - a population of large toroidal shaped foci at the pericentric heterochromatin and a population of smaller foci. The nature of the chromatin in the latter is not known and it could be either euchromatin or heterochromatin. A closer look shows that DNMT1 colocalizes with both populations of RF during late-S phase (marked with arrow and arrowhead). Moreover, DNMT1 colocalizes with RF in all S phase cells. Taken together, this indicates that DNMT1 associates with all sites of DNA replication throughout S phase.


[Seite 54↓]

Fig. 3.15. Endogenous DNMT1 associates with RF throughout S phase. C2C12 cells were pulse labeled with 100 µM BrdU for 10 min and fixed with 3.7% formaldehyde and immunostained for DNMT1 with anti-PATH52 antibody. In the late-S phase cell, arrowhead illustrates an exemplary late replicating centromeric heterochromatin which colocalizes with the Hoechst bright spots, and the arrow shows smaller foci which do not colocalize with Hoechst spots. Scale Bar = 2 µm.

Next, the preference of the various N-terminal deletion proteins to associate with RF during early-, mid- and late-S phase was evaluated. The N-terminal domain fused to GFP (MTN) associated with RF throughout S phase (Fig. 3.16A). Deletion of the PBHD alone (MTN.1, Fig. 3.16B), or both TS and PBHD (MTN.2 and MTN.3, Fig. 3.16C and D) did not affect the ability of the fusion protein to associate with RF. This indicates that the PBD is sufficient for association with RF throughout S phase. Also, a minimal PBD-GFP construct carrying only the 10 amino acid PBD and the 10 basic residues following the PBD reported to be important for association with RF (Montecucco et al., 1998) efficiently associated with the RF throughout S phase (MTPBD-GFP, Fig. 3.16E). Taken together, these observations indicate that the deletion of PBHD and TS does not affect targeting and that the PBD is sufficient for targeting to RF during early, mid (data not shown) and late-S phase. Contrary to earlier reports (Liu et al., 1998), PBHD-YFP fusion protein did not associate with RF at any stage during S phase (Fig. 3.16F). This discrepancy could be because: (a) in their report replication sites were labeled by staining for endogenous DNMT1 and we observe that DNMT1 foci are not always associated with active replication (shown later); (b) their study shows that many DNMT1 foci are excluded of PBHD fusion protein. We observe that in few cells overexpressing PBHD-YFP, the fusion protein accumulates in some structures and these could partially colocalize with RF just as a matter of chance. From our results we conclude that PBHD has no ability to associate with RF. The F-TS-GFP and TS-GFP fusion constructs (Fig. 3.16G and H) are interesting in that they associate with RF only during late-S phase while in early S phase they do not colocalize with RF. Same results were observed with the betagal epitope tagged version which was initially used in mapping the TS in DNMT1 [Seite 55↓](Leonhardt et al., 1992) (Fig. 3.16I). Thus, we consistently observe a specific preference of the TS for targeting to RF during late-S phase. The results of the targeting preference are summarized in Fig. 3.17.


[Seite 56↓]

Fig. 3.16. Subnuclear localization of various deletions of DNMT1 in early- and late-S phase nuclei. Mouse C2C12 cells were transfected with the indicated plasmid constructs and BrdU labeled 24 hrs later followed by indirect immunostaining for BrdU. Cells transfected with TS-ß-gal were immunostained with antibody against ß-gal. The TS associates with RF specifically during late-S phase. Overlay of green and red images are only shown in cases where there is colocalization of the fusion protein and BrdU foci. Scale bar = 2 µm.


[Seite 57↓]

Fig. 3.17. TS associates with replication foci specifically during late-S phase. Summary of the preference of PBD, TS and PBHD to associate with replication foci is illustrated. ‘+’ indicates association; ‘-‘ indicates no association. PBD associates with RF throughout S phase while TS associates with RF only during late-S phase. PBHD does not associate with RF at any stage during S phase.

3.2.2. TS associates with late replicating pericentric heterochromatin

A closer look at Fig. 3.16H (boxed region) brings out two interesting features of the association of F-TS-GFP with RF: (a) F-TS-GFP associates only with late-replicating pericentric heterochromatin which forms large toroidal foci and is excluded from the small RF; (b) even in the large toroidal shaped RF, TS-GFP is actually just around the site of DNA synthesis (BrdU incorporation site) and does not colocalize with the BrdU foci. This is in contrast to the late-S pattern in Fig. 3.16A-E wherein the GFP/YFP fusion proteins are also present at the small RF and completely colocalizes with the toroidal shaped large RF during late-S phase indicating that they associate with all sites of active DNA replication. The large RF are the late-replicating pericentric heterochromatin (Fig. 3.13 and 3.15). These observations suggest that TS per se associates with pericentric regions and not to sites of active DNA synthesis. Moreover, in Fig. 3.16G, it can be observed that during early-S phase, TS-GFP is present in some structures resembling late-S phase structures (such a pattern was also observed in some cells expressing F-TS-GFP). Probably this indicates that TS can associate with pericentric heterochromatin even during early and mid-S phase. This raises two important questions: (1) Is the association with pericentric heterochromatin confined only to the S phase or does it occur during other cell cycle stages? (2) When in S phase does the TS associate with pericentric heterochromatin? The latter question is of significance because during S phase core replication factors (like PCNA, DNA Ligase I) are in a dynamic state of assembly and disassembly which follows a strict spatio-temporal pattern (Leonhardt et al., 2000). The spatio-temporal pattern is a result of a yet undefined higher order chromatin arrangement organizing replicons in clusters (Ma et al., 1998) (Berezney et al., 1995) which fire at specific times during S phase. From the studies on the firing of replicon clusters and the dynamics of PCNA, one can derive that assembly of core replication factors at a site is concomitant with firing of DNA replication while disassembly is accompanied with termination of DNA replication at that site (Sporbert et al., 2002). [Seite 58↓]Hence recruitment to the RF should follow some temporal cue emanating from the RF. Whereas PCNA/DNA ligase I are recruited at all RF throughout S phase, some proteins like pol ε (Fuss and Linn, 2002), HDAC2 (Rountree et al., 2000) and WSTF-ISWI (Bozhenok et al., 2002) are recruited to only late-RF like the TS. Thus, recruitment of proteins to RF is a temporally controlled process and there must be specific temporal cues which discriminate the recruitment of different proteins to RF. By seeking to determine the timing of association of the TS to the late replicating pericentric heterochromatin, we were interested in finding out the temporal cue that determines this association with respect to the one which determines recruitment of PCNA/ DNA ligase I.

To answer these two questions, we monitored the dynamics of F-TS-GFP throughout the cell cycle, and its association with RF during S phase by live cell imaging. A cell cycle marker for live cell analysis was developed by fusing HsDNA Lig I to the DsRed red fluorescent protein (RFP-Ligase) allowing us to follow progression of a cell through different cell cycle stages as well as through the different stages of S-phase. By western blotting we determined that the RFP-Ligase fusion protein is correctly expressed (Fig. 3.18A). Comparison of the subnuclear localization pattern of RFP-Ligase and endogenous DNA Ligase I in C2C12 cells at different cell cycle stages shows that RFP-Ligase behaves like the endogenous DNA Ligase I (Fig. 3.18B). Thus RFP-Ligase can be used to mark actively replicating sites in live cells. To determine the dynamic localization of TS throughout the cell cycle, C2C12 cells were cotransfected with plasmids expressing F-TS-GFP and RFP-Ligase and 24 hrs after transfection the cells which were expressing low levels of the fusion proteins were imaged at about 1 hr intervals for a total of 21 hrs in a chamber maintained at 37ºC. Fig. 3.19 shows the main images from this time series (for complete time lapse, see video 1). At the time when imaging of this cell was started (0 hr), the cell was in early to mid-S phase as can be deciphered from the pattern of RFP-Ligase. At this time, F-TS-GFP was present throughout the nucleus and in some large structures (arrows). Notably, F-TS-GFP did not show any specific association with the early- S phase RF labeled by RFP-Ligase. After about an hour, the cell had entered mid-S phase (image not shown). At 3 hrs, the cell shows a typical mid-S phase pattern and F-TS-GFP still showed the same pattern as at the 0 hr time point. The boxed region shows a newly forming RF which partially colocalizes with the pre-existing F-TS-GFP structure (compare with the 0 hr time point). At about 3.5 hrs from the start, the cell has entered late-S phase where the typical late RF pattern can be observed. Complete colocalization of newly formed large RF with pre-existing F-TS-GFP structures is evident (see boxed region). More importantly, the small RFP-Ligase foci (see boxed region) do not contain F-TS-GFP which is similar to that observed in the fixed cell studies (Fig. 3.16h). At about 8.1 hrs, RFP-Ligase has completely dis-assembled from the RF and is diffused in the nucleus indicating that the cell has entered G2. F-TS-GFP has progressively accumulated at the same sites where it was already visible at 0 hrs. Since RFP-Ligase associated with these sites forming large foci only during late-S phase (3.5 hrs), it can be concluded that these sites correspond to late replicating pericentric heterochromatin. Interestingly, at 8.1 hrs, F-TS-GFP had concentrated more at the large foci and the diffused population had decreased. Probably this indicates that F-TS-GFP gradually accumulates at the pericentric heterochromatin. At about 15.5 hrs, when the cell is in mitosis, F-TS-GFP is retained at the chromatin and in the following G1 (21 hrs), F-TS-GFP is still associated with large structures resembling the pericentric heterochromatin. Thus, F-TS-GFP associates with pericentric heterochromatin throughout the S and G2 phase, and is [Seite 59↓]retained at these sites during mitosis and in the following G1. From these observations it can be concluded that: (1) TS has a strong affinity for pericentric heterochromatin. (2) TS does not function as a replication foci targeting sequence and instead the observation that it has a preference for late RF is due to its affinity for pericentric heterochromatin. (3) There is no temporal cue that determines the association of TS with pericentric heterochromatin and this is independent of the stage of S phase. (4) The association with pericentric heterochromatin is cell cycle independent.


[Seite 60↓]


Fig. 3.18. Development of a cell cycle marker for live cell analysis. Characterization of RFP-Ligase fusion protein expression and subnuclear localization. (A) COS7 cells were transfected with the RFP-Ligase fusion plasmid (shown at top) and after 24 hrs the cells were harvested and extracted by boiling in Laemmli’s sample buffer and analyzed by immunoblotting with an antibody against DNA Ligase I (Cardoso et al., 1997). The fusion protein is correctly expressed. Mock represents transfection of cells in which the plasmid DNA is excluded. (B) RFP-ligase behaves like the endogenous DNA Ligase I throughout the cell cycle in mouse cells. On the left, C2C12 cells were transfected with RFP-Ligase fusion plasmid and pulse labeled with BrdU 24 hrs later followed by indirect immunostaining for BrdU. On the right, C2C12 cells were pulse labeled with BrdU and immunostained with antibodies against DNA Ligase I and BrdU.


[Seite 61↓]
Fig. 3.19. Time lapse analysis of mouse cells expressing F-TS-GFP and RFP-Ligase. C2C12 cells were cotransfected with plasmids encoding F-TS-GFP and RFP-Lig. After 24 hrs the cells were imaged at 1 hr intervals in a Zeiss LSM 510. Three Z sections of 1µm each were taken at each time point. Movement in the Z plane was corrected by comparing the Z sections at different time points and arranging them sequentially so as to include the same structures (RF and F-TS-GFP structures) in each sequential image. Only indicated time points are shown here. Overlay shows the merged image of green and red channels. White arrow indicates two exemplary F-TS-GFP structures that existed at 0 hrs and persist throughout the cell cycle. Box shows zoomed regions. Blue arrows in boxed region indicate colocalization of pre-existing F-TS-GFP structure with newly formed RFP-Lig foci. The full time lapse is in Video-1.

3.2.3. Subnuclear localization of DNMT1 throughout the cell cycle

The strong affinity of TS for pericentric heterochromatin is striking because once associated with these regions, it is visualized there through following cell cycles. This raises the question whether this observation is an artifact or DNMT1 uses the TS to associate with pericentric heterochromatin at any stage during the cell cycle. In order to determine this, we established the localization of endogenous DNMT1 as well as DNMT1(s) fused to GFP (GMT) at G1, S, G2 and M. Fig. 3.20A describes the strategy we used for identifying cells in G1 and G2. To get cells in G1, loosely attached mitotic cells were harvested by mechanical shake off, concentrated by centrifuging and were then laid onto a coverslip. The cells were incubated at 37˚C for about 2 hrs and allowed to become adherent and enter G1. Subsequently the cells were fixed and immunostained. To identify cells in G2, C2C12 cells were pulse labeled with 10 µM BrdU for 10 min and chased for 2-3 hrs in medium containing 100µM thymidine and subsequently fixed and stained for BrdU, DNMT1 and PCNA. It is known that PCNA is localized at RF during S phase and is diffused in the nucleus of non-S phase cells. Cells showing positive BrdU staining and diffused PCNA are cells which were in the late-S phase during the BrdU pulse and have entered G2 during the 2-3 hrs chase. We rule out these cells to be in G1 because the G2 phase in C2C12 cells takes about 4-5 hrs and with our strategy of chasing for 2-3 hrs after BrdU labeling, the BrdU positive cells should be in G2. In G2 phase, DNMT1 is present in large structures reminiscent of the late-RF while PCNA is diffused (Fig. 3.20B). The DNMT1 foci colocalize with BrdU foci indicating that DNMT1 is retained at these late replicating sites even after the cell has exited S phase. This is [Seite 62↓]similar to the behaviour of the TS. In mitotic cells, DNMT1 is associated with the chromatin like the TS (Fig. 3.20C). G1 cells showed a diffused distribution of DNMT1 in the nucleus like PCNA (Fig. 3.20E). Thus, it seems that DNMT1 associates with pericentric heterochromatin during late-S phase and is retained there through G2 and M, and is released from these sites in the following G1. In early-G1 cells, DNMT1 is still associated with the pericentric heterochromatin (Fig. 3.20E). From these studies on fixed cells it is not clear whether DNMT1 is retained at late replicating pericentric heterochromatin in G2 or it is released at the end of S phase and again associates with the pericentric heterochromatin sometime in G2. To sort this out, C2C12 cells were cotransfected with plasmids expressing GFP-DNMT1(s) and RFP-Ligase and the dynamics of GFP-DNMT1(s) was monitored in live cells. The results are shown in Fig. 3.21, Video 2 and Video-3. At the time imaging of this cell was started (0 hrs) the cell was in mid-S phase when DNMT1 completely colocalizes with RFP-Ligase. At 3 hrs, the cell has entered late-S phase when both GFP-DNMT1(s) and RFP-Ligase are present in large structures. The boxed region shows that GFP-DNMT1(s) is present even at the small RF during late-S phase indicating that, like RFP-Ligase, GFP-DNMT1(s) associates with all RF (unlike the TS, see Fig. 3.19, 3.5 hrs). At 5 hrs, the cell is in the very late-S phase when some of the large RFP-Ligase foci present at 3 hrs have disassembled (arrow heads) while GFP-DNMT1(s) is retained at these sites. At 7 hrs, all the RFP-Ligase foci have dis-assembled and the cell has entered G2. At this point, GFP-DNMT1 is retained at the late-replicated regions (arrowheads). During mitosis (8.6 hrs), GFP-DNMT1 is associated with the chromatin and in the following G1 (9.6 hrs) it is diffused. These results match the observations on fixed cell studies and confirm that DNMT1 is retained at late-replicating pericentric heterochromatin through G2 and M and released from these sites only in the next G1.


[Seite 63↓]


Fig. 3.20. Elucidation of the subnuclear localization of endogenous DNMT1 throughout the cell cycle. (A) Strategy for identifying cells in G1, S and G2. (B) DNMT1 is retained at late replicating centromeric heterochromatin during G2. C2C12 cells were pulse-labeled with 10µM BrdU and chased for 2-3 hrs followed by fixation. Cells were immunostained with anti-BrdU, anti-DNMT1 and anti-PCNA antibodies. G2 cells were identified by their positive staining for BrdU and diffused PCNA. (C) DNMT1 is associated with the chromatin during mitosis. Mitotic cells were identified by their morphology and condensed chromosomes. (D) In early G1 cells, when the daughter cells are still connected (cells marked with arrow), DNMT1 is present at the centromeric heterochromatin. (E) DNMT1 shows a diffused subnuclear localization in G1 cells. G1 cells were obtained by seeding mitotic cells collected by mechanical shake off onto cover slips and allowing them to adhere for 2 hrs. The association of DNMT1 with RF during S phase is shown in Fig 3.15. Scale Bar = 5 µm.


[Seite 64↓]


Fig. 3.21. Time lapse analysis of mouse cells expressing GMT and RFP-Ligase. C2C12 cells were cotransfected with plasmids encoding GMT and RFP-Lig. After 24 hrs the cells were imaged at 1 hr intervals as mentioned in Fig. 3.19. Only indicated time points are shown here. Overlay shows the merged image of green and red channels. White arrowheads point to three exemplary GMT foci which assemble concomitantly with RFP-Lig during late-S phase and are retained through the rest of S phase and in G2. Box in the overlay at 3 hrs shows zoomed region where GMT also associates with smaller foci. Green levels are increased to show this. The images of the M phase to G1 phase are from another cell, hence it starts again from 0 hrs. During M, GMT is associated with the chromatin. The daughter cells in G1 show diffused nuclear distribution of GMT. Full time lapse in Video-2.

To test whether retention of DNMT1 at late replicating pericentric heterochromatin in G2 and its association to mitotic chromatin during M is dependent on the TS, a region in the TS was deleted. This deletion (GMT∆TS, Fig. 3.14A) associated with RF throughout S phase (Fig. 3.22) but was not retained at pericentric heterochromatin in G2 and rather showed a diffused distribution (Fig. 3.22). GMT∆TS was not bound to chromatin during mitosis (Fig. 3.22) and in G1 it was again diffused in the nucleus (Fig. 3.22). These results strongly indicate that retention and binding to pericentric heterochromatin during G2 and M is dependent on the TS. Binding of DNMT1 to pericentric regions during their replication might also involve the TS in addition to the PBD, however this cannot be conclusively discerned here. The fact that the TS alone can bind to pericentric regions independent of the cell cycle stage indicates that this interaction could occur also during S phase in the context of the full length DNMT1.


[Seite 65↓]

Fig. 3.22. TS is required for retention of DNMT1 at pericentric regions during G2 and M phases. C2C12 cells were transfected with plasmid encoding GMT∆TS. G1 and G2 cells were identified as described in Fig. 3.20. After 24 hrs of transfection, cells were immunostained with antibody against DNA Ligase I to label RF. GMT∆TS colocalizes with sites of DNA replication throughout S phase. During G2, GMT∆TS shows diffused nuclear distribution, and is excluded from mitotic chromatin. G1 cells show diffused nuclear distribution of GMT∆TS. Scale bar = 5 µm

3.2.4. Comparison of dynamics of TS with DNMT1: what determines release of DNMT1 from pericentric heterochromatin during G1?

Comparison of the localization of TS-GFP (Fig. 3.19) and GFP-DNMT1 (Fig. 3.21) at different cell cycle stages shows that TS associates with pericentric heterochromatin during all cell cycle stages while GFP-DNMT1 is present at these sites only during G2 and M. Two features of GFP-DNMT1 are striking: (1) During S phase, GFP-DNMT1 is targeted to sites of replication and its association with pericentric heterochromatin is delayed until late-S phase when these regions start replicating whereas TS can bind to these sites anytime in S phase. (2) During G1, GFP-DNMT1 becomes diffused while TS stays bound to the pericentric heterochromatin. These observations indicate that: (a) some feature(s) of DNMT1 prevents the association of TS with pericentric heterochromatin until late-S phase and (b) mediates its release in G1 thereby making it diffused.

In order to determine the region responsible for this control, we analyzed the [Seite 66↓]localization of a series of deletion and point mutants of DNMT1 fused to GFP (Fig. 3.14) throughout the cell cycle in fixed or live cells. Previous studies have identified phosphorylation of serine at position 514, which is within the TS, and to which no function has yet been attributed (Glickman et al., 1997). We tested whether phosphorylation of this site controls association of DNMT1 with pericentric heterochromatin by mutating it to a phosphorylation-off (S514A) and phosphorylation-on (S514D) state. All mutants except MTN.1 and GMT-PBD-H167V behaved like DNMT1 in that they associated with RF throughout S phase and assembled at the pericentric heterochromatin only during late S phase, were retained at the pericentric regions during G2 and M phases, and were diffused in the nucleus during G1 (Table 3.2). Deletion of PBHD in MTN.1 did not affect association of the fusion protein to RF, and its retention at pericentric heterochromatin in G2 and M (Fig. 3.23A and B ). But during G1, MTN.1 was not released from the pericentric heterochromatin indicating that PBHD might be important for the release of DNMT1 from pericentric heterochromatin. Full length GMT carrying a deletion of the PBHD, GMT∆PBHD, was not efficiently localized to the nucleus and so its subnuclear distribution could not be evaluated. Previous work has shown that there is an NLS in this region (Cardoso and Leonhardt, 1999).


[Seite 67↓]

Fig. 3.23. PBD delays association of TS in DNMT1 with pericentric heterochromatin until late-S phase. (A) C2C12 cells were transfected with plasmids encoding the various mutants of DNMT1 fused to GFP. 24 hrs after transfection the subnuclear localization of the GFP-fusion proteins at S, G2, M and G1 phases were analyzed as described in Fig. 3.20. MTN.1 (ii), which lacks the PBHD, is retained in pericentric heterochromatin during G1. GMT-PBD-H167V (vii) is associated with pericentric heterochromatin right during early- and mid-S phase like the TS-GFP fusion protein. BrdU staining is shown only in the case of GMT-PBD-H167V. Scale Bar = 5 µm. (B) C2C12 cells transfected with various plasmids were pulse labeled with BrdU 24 hrs later and incorporation of BrdU was detected by immunostaining. The percentage of transfected cells in early-/mid-S phase exhibiting colocalization of the GFP-fusion protein with either BrdU labeled RF or pericentric heterochromatin, or showing diffused nuclear distribution was scored. Black is colocalization with RF, gray is colocalization with pericentric heterochromatin, white is diffused nuclear distribution. Only GMT-PBDmut associates with pericentric heterochromatin during early-/mid-S phase.

Table 3.2: Summary of association (+) or no association (-) of various proteins with RF during different stages of S phase or with pericentric heterochromatin during different stages of the cell cycle. 'Late-S' and 'Throughout' mean association with pericentric heterochromatin occurs only during late-S phase or throughout S phase, respectively.

Fusion protein

Association with RF during:

Association with pericentric heterochromatin during:

Early-S

Mid-S

Late-S

G2

M

G1

S

DNMT1 (Endogenous)

+

+

+

+

+

-

Late-S

MTN

+

+

+

+

+

-

Late-S

MTN.1

+

+

+

+

+

+

Late-S

MTN.2

+

+

+

-

-

-

Late-S

MTN.3

+

+

+

-

-

-

Late-S

TS-GFP

-

-

-

+

+

+

Throughout

F-TS-GFP

-

-

-

+

+

+

Throughout

GMT

+

+

+

+

+

-

Late-S

GMT-S514A

+

+

+

+

+

-

Late-S

GMT-S514D

+

+

+

+

+

-

Late-S

GMT∆TS

+

+

+

-

-

-

Late-S

GMT-PBD-H167V

-/+

-/+

-/+

+

+

-

Throughout


[Seite 69↓]

Mutating the PBD (GMT-PBD-H167V) to a form that cannot bind to PCNA (Chuang et al., 1997) reduced targeting to RF accompanied by association with pericentric heterochromatin right in early-S reminiscent of F-TS-GFP. Like DNMT1, GMT-PBD-H167V was retained at pericentric heterochromatin during G2 and mitosis and showed diffused nuclear distribution in G1. Thus, PBD drives the orchestrated assembly of DNMT1 to RF throughout S phase by delaying TS-mediated association of DNMT1 with the pericentric heterochromatin until late-S phase when these regions replicate. It can be envisaged that the PBD brings DNMT1 to the pericentric heterochromatin during late-S which would then allow the TS to bind to some component of the pericentric heterochromatin and get retained here until the end of mitosis. These results indicate that PBD prevents binding of the TS in DNMT1 with pericentric heterochromatin until late-S phase while PBHD influences its release during G1.

Since in transfected cells there is a high variability in expression of the fusion proteins within the same population of cells, the observed association of GMT-PBD-H167V with pericentric heterochromatin during early-/mid-S phase cells could be a result of excess of the fusion protein available after saturation at the RF. To correct for any mis-representation arising from singular observations, C2C12 cells expressing moderate levels of the transfected plasmids encoding the various deletions, which were in early-/mid-S phase were randomly selected and the percentage cells exhibiting colocalization of the deletion protein with RF or pericentric heterochromatin was evaluated. Only GMT-PBD-H167V was associated with pericentric heterochromatin in a significant number (27.6%) of early-/mid-S phase cells (Fig. 3.23D). Notably, GMT was targeted to RF in all cells and in none of the early-/mid-S phase cells it associated with pericentric heterochromatin. In few cells GMT-PBD-H167V was targeted to RF indicating that the H to V mutation in PBD did not completely abolish binding to PCNA, and therefore association with RF. The N-terminal deletion proteins were associated with the RF in all early-/mid-S phase cells scored. Thus the association of GMT-PBD-H167V with pericentric heterochromatin during early-/mid-S phase cells is not due to overexpression and rather is due to a lack of ability to target to RF.


[Seite 70↓]

3.3.  Biological effects of TS overexpression

The results in the previous section brings out a novel role for the TS in the retention of DNMT1at pericentric heterochromatin in G2 and M. This leads us to the question as to what role DNMT1 has at pericentric regions during G2 and M. We approached this question by asking what is the effect of expression of TS-GFP on cells, arguing that excess TS would prevent association of endogenous DNMT1 with pericentric heterochromatin. Since dense methylation of pericentric regions is important for chromatin stability and cell viability (reviewed in Robertson and Wolffe, 2000), the effect of expression of TS-GFP on cell growth and nuclear morphology was monitored.

3.3.1. Effect on cell viability

From the observations during live cell microscopy, it seems that short-term expression of F-TS-GFP/TS-GFP did not affect cell viability because the cells traversed through the different cell cycle stages and divided normally. To determine the long term effect of expressing TS on cell viability, C2C12 cells were transfected with the different DNMT1 deletion constructs (which have a neomycin resistance marker) and the cells were grown in the presence of neomycin (G418) to select for stably transfected cells. If the fusion protein expressed from the plasmid has a negative influence on cell growth, these cells would not survive and would be lost (Fig. 3.24A). Transfected cells would multiply to form colonies if the protein expressed from the plasmid is not toxic to the cell. Thus, transfected cells would be positively selected for by neo r and negatively selected by how detrimental the expression of the recombinant transgene would be. The cells were cultured in the presence of G418 until all the cells in the mock transfection (no neo r encoding plasmid) were dead. Subsequently, the cells were fixed with methanol, stained with Giemsa stain and colonies were counted. The result shown in Fig. 3.24B is normalized against the GFP control transfection. As can be seen, there is more than 40% reduction in colony formation when cells are expressing the TS fusion protein indicating that the TS has a deleterious effect on cell growth in the long term. A similar reduction was observed in the case of MTPBD-GFP. This would be expected due to the crucial role of PCNA in various cellular processes and binding of MTPBD-GFP would prevent the authentic partners of PCNA from associating with it thereby hampering cell growth. Similar results for the PBD have been reported previously (Mattock et al., 2001). The observation that expression of GMT does not affect cell viability indicates that the TS affects cell growth only when expressed/presented outside the context of the full length protein. This emphasizes the previous observations on the role of other domains in DNMT1, viz the PBD and PBHD, in controlling the activity of the TS. Also, it indicates that fusion of GFP does not have any affect on the normal activity of these proteins. Although we see a negative effect of the TS on cell viability, this assay does not yet pinpoint the process which is affected.


[Seite 71↓]
Fig. 3.24. Expression of TS reduces viability of cells. (A) Methodology for colony forming assay. (B) Images of Giemsa stained colonies. (C) Colony forming efficiency of C2C12 cells transfected with plasmids encoding the mentioned proteins. The number of colonies in each experiment was normalized to the GFP control of the corresponding experiment and expressed as percentage. Results from three independent experiments were averaged; error bars show standard deviation.

3.3.2. Effect on nuclear organization and morphology

To get an insight into the effect of TS in normal chromatin organization, fixed cells expressing the F-TS-GFP/TS-GFP fusion proteins and stained for DNA with Hoechst 33258/TOPRO-3 were screened under the microscope for unusual nuclear structures. Such an analysis brought forth two interesting defects in these cells:

Micronuclei Formation:

Cells expressing F-TS-GFP/TS-GFP and MTPBD-GFP showed almost 2-3 fold increase in micronuclei formation compared to cells expressing GFP or GMT (Fig. 3.25A). Since micronuclei formation reflects chromatin instability (Bonassi and Au, 2002) (Tucker and Preston, 1996), expression of TS-GFP seems to interfere with the integrity of chromatin. The affinity of TS for pericentric heterochromatin indicates that it might influence the maintenance of pericentric heterochromatin. Since many DNA synthesis and DNA repair factors have a PBD, it is quite likely that [Seite 72↓]overexpression of MTPBD-GFP interferes with these basic processes leading to micronuclei formation.


Fig. 3.25. Micronuclei formation and coalescence of centromeric heterochromatin. Transfected cells expressing the mentioned proteins were fixed and stained for DNA with Hoechst 33258/TOPRO-3 and cells exhibiting micronuclei or coalescence of centromeric heterochromatin were scored. About 200-300 transfected cells were counted in each case and the percent values were calculated. The values obtained from three different transfections were averaged and plotted. Bars indicate standard deviation. (A) Percent transfected cells exhibiting micronuclei formation. Inset shows an image of a cell transfected with plasmid expressing TS-GFP exhibiting several micronuclei. (B) Percent transfected cells in which the centromeric heterochromatin is coalesced. Inset shows an image of a cell transfected with plasmid expressing TS-GFP exhibiting coalesced centromeric heterochromatin and a non-transfected cell showing normal centromeric heterochromatin. (C) C2C12 cells were cotransfected with plasmids encoding TS-GFP and CENPB-RFP to label the centromeres. Coalesced heterochromatin consists of relocated and clustered centromeric regions.


[Seite 73↓]

Coalescence of Centromeric Heterochromatin:

Some cells overexpressing the F-TS-GFP and TS-GFP exhibit coalescence of centromeric heterochromatin into large chromocenters. Fig. 3.25B shows the percentage of transfected cells showing coalescent centromeric heterochromatin. In untransfected cells, centromeric heterochromatin is visible as discrete structures that are densely stained with DNA dyes like Hoechst 33258 or TOPRO-3. In cells overexpressing TS-GFP, the centromeric regions coalesce together to form large structures. Cells expressing GMT or MTPBD-GFP never show such structures (Fig. 3.25B). To test whether centromeres had relocated in these coalesced heterochromatin, plasmids encoding TS-GFP and CENP-B-RFP (which labels centromeres) were co-transfected in C2C12 cells. As shown in Fig. 3.25C, these coalesced heterochromatin centers have CENP-B-RFP indicating that indeed these regions are formed by fusion of the centromeric heterochromatin.

These results suggest a role of the TS in the large scale organization of centromeric heterochromatin.


[Seite 74↓]

3.4.  Evolutionary conservation of the regulatory domain of DNMT1

C-5-methyltransferases are classified into five families based on phylogenetic analysis of their catalytic domains (Colot and Rossignol, 1999). One of this is the DNMT1 subfamily constituted by mammalian DNMT1 and counterparts from other animals, plants and fungi. DNMT1 is attributed a maintenance of methylation function primarily due to its strong preference for hemimethylated DNA (Bestor, 1992) (Yoder et al., 1997a) and its redistribution to replication foci during S phase (Leonhardt et al., 1992). However, these studies have been done only with mammalian DNMT1 and it is not known whether the DNMT1 in other organisms has a similar role. Our studies on the ability of the PBD, TS and PBHD of mouse DNMT1 to associate with RF in Drosophila and mammalian cells have shown that only the PBD is capable of mediating association of DNMT1 with RF. Recruitment of several proteins involved in DNA replication/repair to RF seems to be mediated by interaction with PCNA via the PBD suggesting that this is a general mechanism for targeting proteins to RF. In order to check whether the other members of the DNMT1 family can potentially be targeted to replication structures and thereby function as maintenance MTase, we performed sequence analysis on the DNMT1 family and also all known C-5-methyltransferases in eukaryotes to detect the presence of the PBD. Moreover, the role of the TS as a pericentric heterochromatin binding domain is interesting and we sought to identify other proteins harbouring a TS. The results from these sequence analyses are described below.

3.4.1. PBD is present only in metazoan DNMT1 family

Alignment of the PBD from various proteins (Fig. 3.2) shows that conservation of the PBD is confined to just 3-4 residues spanning a region of about 10 amino acid forming a consensus motif QxxI/LxxFF. This is the minimal sequence required for binding to PCNA (Zheleva et al., 2000). A region of about 10 amino acids rich in basic residues immediately following the consensus motif is required for efficient association with RF (Montecucco et al., 1998). To determine whether any of the C-5-methyltransferases from eukaryotes other than the reported mammalian DNMT1 have a PBD that could associate with RF, a profile of the PBD from the known PBDs (including the region rich in basic residues) from different proteins was made and used to search for significant matches in all the known C-5-methyltransferases. Fig. 3.26 shows the alignment of the sequences obtained from such a search. Significant hits for the PBD were obtained in all DNMT1 proteins from metazoan origin. No significant hits were observed in proteins from the metazoan DNMT2 and DNMT3 families. This is consistent with the observation that DNMT3 proteins are not associated with RF (Margot et al., 2001) and DNMT2 is an inactive protein which is not required for de novo or maintenance methylation in embryonic stem cells (Okano et al., 1998b). Some of the plant and fungal MTases show similarity to the consensus PBD motif but lack the adjacent region rich in basic residues, which is important for association with RF. Moreover, analysis of the location of PBD shows that the PBD is present at similar locations in homologous proteins. For example, the PBD in all DNA Ligase I homologues is present in the extreme N-terminus and in the case of metazoan DNMT1, it is located 1/5-th into the protein from the N-terminus. This is not the case for any of the PBD-like sequences [Seite 75↓]identified in the MTases in plants and fungus. The PBD motif being a small region of about 10 aa, many unrelated proteins contain sequences that match the consensus (Dalrymple et al., 2001). A search of the Swissprot database with the PBD profile returns many hits of proteins from bacteria to human (some of which are mitochondrial proteins) whose annotation does not attribute them any role in cell cycle or DNA metabolism indicating that these are false hits. The PBDs identified in the plant and fungal C-5-methyltransferases most probably are false hits considering their unusual locations in the protein and the absence of an adjacent region rich in basic residues. Thus, the PBD-like sequences identified in plants and fungus most likely do not function as authentic PBDs.


Fig. 3.26. Search for a PBD in all C-5-methyltransferases. PBDs characterized in different proteins from metazoa were aligned by PileUp and a profile was generated using ProfileMake. This profile, shown at the top, was used to search all the C-5-methyltransferase families in metazoans, plants and fungi using the program ProfileSearch. The PBDs identified were aligned by Clustal and realigned by visual inspection to give maximal alignment. 100% identity is highlighted in red colour, more than 50% identity is highlighted in yellow, residues in blue form the region rich in basic residues. Sequence names in red are from metazoans, in green are from plants and brown is from fungus. Sequence positions are shown in numbers in parentheses. Square brackets at the right indicate protein accession numbers. The sequence of the most conserved residues in the PBD required for binding to PCNA and association with RF is shown at the top. Most of the PBDs identified in the plant and fungal methyltransferases are either in the C-terminal catalytic domain or in the PBHD or chromodomain and they lack the stretch of basic residues. Mm: Mus musculus; Rr: Rattus rattus; Hs: Homo sapiens; Gg: Gallus gallus; Xl: Xenopus laevis; Xm: Xiphophorus maculatus x helleri; Pl: Paracentrotus lividus; Dm: Drosophila melanogaster; At: Arabidopsis thaliana; Dc: Daucus carota; Ps: Pisum sativum; Nt: Nicotiana tabacum; Le: Lycopersicon esculentum; Zm: Zea mays; Nc: Neurospora crassa.

These sequence analyses show that only metazoan C-5-methyltransferases of the DNMT1 family have a PBD while their plant and fungal counterparts lack a PBD and therefore might not associate with RF. Or, the association of the latter with RF is mediated independent of the PBD.


[Seite 76↓]

3.4.2.  TS is a unique domain present only in DNMT1 family from animals, plant and fungi

The mouse DNMT1 TS mapped earlier (Leonhardt et al., 1992) was used as query to search the non-redundant (nr) database using PSI-BLAST (Altschul et al., 1997), which is a sensitive method to detect weak similarities in protein and nucleotide sequences (Callebaut et al., 1999). Following convergence after four iterations, the sequences with an E-value better than the threshold comprised only C5-methyltransferases from metazoans, plants and fungi belonging to the DNMT1 family (Fig. 3.27). None of the sequences with an E-value worse than threshold, except monocyte leukemia zinc finger protein (mouse), hypothetical zinc finger protein (S. pombe) and CHP rich zinc finger protein, seemed to have any link with chromatin function. The regions of these proteins showed very little similarity to mouse TS and their expected values are very high (> 0.015) to be homologous to the TS. The fact that sequences with a significant E-value all belonged to the C5-methyltransferases of the DNMT1 family indicates that the TS is unique to the DNMT1 family.

The results of the PSI-BLAST showed that in all the plant C-5-methyltransferases there are two regions in the N-terminus that show similarity to mouse TS, designated here as TS-1 and TS-2 (Fig. 3.28). Previous studies have reported two TS domains in MET1 from Arabidopsis (Colot and Rossignol, 1999). Such duplicated TS domains is not observed in fungal and metazoan DNMT1. Though both TS show some similarity to mammalian TS, there is very little similarity within them. TS-2 is detected only through iterative BLAST searches using PSI-BLAST. The expected values obtained from performing a BLAST2 comparison of the mouse TS and Arabidopsis TS-1, mouse TS and Arabidopsis TS-2, and Arabidopsis TS-1 and TS-2 are 2e-12, 0.93, and 3418, respectively. This high E-value between TS-1 and TS-2 is true for all the plant sequences indicating that these duplicated TS might have originated very early in evolution in a common ancestor and subsequently diverged by accumulating many mutations. However, closer inspection of the alignment of the two plant TS and the mouse TS obtained from PSI-BLAST shows that a region is conserved in both (Fig. 3.28, boxes). Analysis of the alignment of all the TS identified shows that this region forms a conserved motif (Fig. 3.29). In fungi, only A. immersus Masc2 has a TS which also has this conserved motif.


[Seite 77↓]

Fig. 3.27. TS is an unique domain present only in the DNMT1 family. The nr database was searched using PSI-BLAST with the mouse TS as query. Subsequent iterations were performed by including the sequences (which were all C-5 methyltransferases) which showed an E value better than threshold. Following four iterations convergence was attained and no new sequences were retrieved. All the sequences showing significant E value belonged to the DNMT1 family.


[Seite 78↓]

Fig. 3.28. TS domain is duplicated in plants and shows a conserved motif. A typical example of the result of PSI-BLAST showing duplicated TS in plants, in this case A. thaliana ( accession no. AAF14882). Box shows the region that has a high degree of conservation in both TS.


[Seite 79↓]


Fig. 3.29. A conserved motif in the TS from metazoans, plants and fungus. TS from mouse (Leonhardt et al., 1992) was used as a query to search the nr database for similar sequences using PSI-BLAST. All significant hits obtained belonged to the DNMT1 family from metazoans, plants and fungus (see also Fig 3.27). These were aligned using Clustal. Here a region of the alignment with maximum conservation is shown. Residues that are identical to the consensus are shaded black, highly similar residues are shaded grey. Asterisk indicates the phosphorylation site (S at position 514) that is conserved in metazoans. In the consensus sequence, letters in red are 100% identical, uppercase is greater than 80% identity and lowercase is similar residues that are conserved at least between two of the three groups of TS: metazoan TS, TS-1 and TS-2. Conserved changes are shown below the consensus. In sequence names Ai is Ascobolus immersus. Other sequence names are as in Fig 3.26. (AtMET is AAF14882).


[Seite 80↓]

In order to determine whether this conserved motif is present in other proteins, a profile of the TS from metazoans, plants and fungus was made and used to search for similar sequences in the Swissprot database. No significant hits were obtained outside the members of the DNMT1 family indicating that this domain is specific to the DNMT1 family. Many of the low scoring hits showed similarity to the TS motif WISTDFADYDLMKPSKEY (residues that are more than 90% conserved in the various TS from DNMT1 family are in bold) but further analysis of their annotation in the database or their homologues did not indicate any function related to chromatin. Two of them, telomerase reverse transcriptase catalytic subunit (mouse accession no. 070372) and RNA-directed RNA polymerase (yeast, accession no. P25328), are chromatin related proteins and the presence of a TS-like motif could be significant but the scores were very low.

Thus, these results indicate that the TS is a unique domain present only in the DNMT1 family. A conserved motif in the TS has been identified and mutational analysis of these residues should provide further insight into their roles.


© 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