Characterization of Dnmt3b:thymine-DNA glycosylase interaction and stimulation of thymine glycosylase-mediated repair by DNA methyltransferase(s) and RNA

Abstract

Methylation of cytosine residues in CpG dinucleotides plays an important role in epigenetic regulation of gene expression and chromatin structure/stability in higher eukaryotes. DNA methylation patterns are established and maintained at CpG dinucleotides by DNA methyltransferases (Dnmt1, Dnmt3a, and Dnmt3b). In mammals and many other eukaryotes, the CpG dinucleotide is underrepresented in the genome. This loss is postulated to be the result of unrepaired deamination of cytosine and 5-methylcytosine to uracil and thymine, respectively. Two thymine glycosylases are believed to reduce the impact of 5-methylcytosine deamination. G/T mismatch-specific thymine-DNA glycosylase (Tdg) and methyl-CpG binding domain protein 4 can both excise uracil or thymine at U.G and T.G mismatches to initiate base excision repair. Here, we report the characterization of interactions between Dnmt3b and both Tdg and methyl-CpG binding domain protein 4. Our results demonstrate (1) that both Tdg and Dnmt3b are colocalized to heterochromatin and (2) reduction of T.G mismatch repair efficiency upon loss of DNA methyltransferase expression, as well as a requirement for an RNA component for correct T.G mismatch repair.

Figure 1

Interaction and genomic localization of Dnmt3b and Tdg in EC cells. (a) Endogenous Dnmt3b was immunoprecipitated from F9 EC cell nuclear extracts cells with increasing concentrations (0.5, 1.0, 2.0 μg) of α-Dnmt3b antibody. An immunoblot of the pull down product with α-Tdg antibody confirmed that Tdg coprecipitated with Dnmt3b. (b) Genomic localization of Dnmt3b and Tdg by chromatin immunoprecipitation (ChIP). Dnmt3b and Tdg were immunoprecipitated from crosslinked chromatin in P19 EC cells. PCR was performed using primers designed to amplify the centromeric minor satellite repeats. (b′) Dnmt3b and Tdg were immunoprecipitated from crosslinked chromatin. The complexes were purified by several wash steps and subjected to a second ChIP (i.e. ReChIP) with the antibody indicated above. PCR was performed using primers designed to amplify the pericentromeric major satellite repeats or the intracisternal A particle long terminal repeats (IAP LTR). Control PCR reactions utilized total input DNA, DNA isolated using species-matched normal IgG antibody and DNA isolated in the absence of immunoprecipitating antibody.

Figure 2

Tdg and Mbd4 make multiple interactions with Dnmt3b. (a) Schematic diagram of Dnmt3b showing the location of characterized domains and conserved MTase motifs of Dnmt3b. Refer to text for details. (b) Interaction between Tdg and Mbd4 and deletion mutants of Dnmt3b tested by coimmunoprecipitation and immunoblotting. When the PWWP domain and the region of the C-terminus that interacts with Tdg (identified in Figure 3c) were deleted from Dnmt3b, neither Tdg, nor Mbd4 precipitated with Dnmt3b. (c) Interaction between Tdg and Mbd4 and C-terminal truncation mutants of Dnmt3b tested by coimmunoprecipitation and immunoblotting. Note that while deletion of the complete C-terminal region of Dnmt3b led to loss of interaction, deletion of motifs IV-X did not reduce binding between Dnmt3b and either Tdg or Mbd4. This identifies a region containing the conserved MTase motif I as critical for interaction with both glycosylases. IB, immunoblot. (d) Summary of the results displayed in (b) and (c). (e) Tdg interacts with the PWWP and catalytic domains of Dnmt3b in vitro. ³⁵S-Met-radiolabeled, myc-tagged Dnmt3b as well as myc-tagged Dnmt3b PWWP and MTase IVI domains and HA-tagged Tdg were generated by in vitro transcription/translation in rabbit reticulocyte lysates. Immunoprecipitations were performed with α-cMyc antibody. Lane 1, (positive control) cMyc-p53 immunoprecipitates the SV40 large T antigen. Lane 2, coimmunoprecipitation of HA-Tdg and cMyc-Dnmt3b. Lane 3, the isolated Dnmt3b PWWP domain (aa 229-301) effectively immunoprecipitates Tdg. Lane 4, a portion of the Dnmt3b C-terminus encompassing conserved DNA MTase motifs I-VI (aa 580-754) also interacts with Tdg.

Figure 3

DNA methyltransferases stimulate base excision repair of T·G mismatches. A: Design and testing of oligodeoxyribonucleotides (ODNs) used in a mismatch repair assay. (a) Sequence of the double-stranded ODN. The sense strand contains either a normal CCGG site or a mutated CTGG site. The antisense strand contains either C or 5mC within an MspI/HpaII recognition site. Positions of variable nucleotides y and x are shown in bold. The MspI/HpaII recognition sequence is indicated with a bar. The asterisk indicates the position of the ³²P label. (b) Predicted sensitivity to MspI or HpaII digestion of the control C·GC and C·G^mC ODNs compared to the mismatched T·GC and T·G^mC ODNs. (c) MspI digestion of the control ODNs results in the expected 26 nt long radiolabeled fragment from both controls, while a 26 nt long fragment is only observed after HpaII digestion of the unmethylated control. In contrast, T·G mismatch ODNs are refractory to digestion by either enzyme. M, MspI. H, HpaII. B: De novo DNA methyltrasnferases stimulate repair of T·G mismatches. (a) Evaluation of Dnmt and Tdg protein levels in J1 and Dnmt null ES cells. Nuclear extracts prepared from each cell line were immunoblotted using antibodies specific for Dnmt1, Dnmt3a, Dnmt3b and Tdg to verify the expression of Tdg and the absence of specific Dnmt expression in each cell line relative to wild-type J1. Immunoblot for Pcna served as a loading control. (b) Predicted T·GC ODN sensitivity to digestion by MspI or HpaII depending on extent of repair/DNA methylation. (c) Typical results from one of three independent experiments in which the T·GC ODN was incubated with wild-type J1 or Dnmt null ES cell nuclear extracts followed by digestion with either MspI or HpaII. The expected 26 nt repair product produced as a result of digestion is indicated by the arrowhead. In the absence of either Dnmt3b and/or Dnmt3a there is a reduction in repair efficiency when compared to wild-type nuclear extracts (refer to text for discussion of Dnmt1 null extracts). As a control (last lane), T·GC ODN was used in a repair assay but was not digested. Lack of a 26 nt fragment indicates the ODN is being repaired and not merely nicked 5′ to the mismatched thymine. (d) Quantitation of differences in the extent of mismatch repair by extracts from normal ES cells (J1) and ES cells nullizygous for the indicated Dnmts. Error bars represent the mean (± SD) of three independent experiments. [*] P < 0.005

Figure 4

Treatment of nuclear extracts with RNase A leads to random cleavage of the T·G^mC ODN. (a) Diagram depicting an experimental protocol devised to examine methylation of repaired T·G^mC ODN in normal and Dnmt null ES cell lines. Following repair of the T·G^mC ODN using RNase A-treated ES cell nuclear extracts, the ODN was purified and heat denatured. It was then annealed to a 5-fold excess of cold antisense CG strand. This will generate an unmethylated or hemimethylated ODN depending on the methylation status of the repaired cytosine. MspI will digest either ODN but HpaII digestion will be inhibited if the repaired cytosine is methylated. (b) Results of the experiment diagrammed in (a). The T·G^mC ODN appears to be degraded by each RNase-treated nuclear extract in a consistent pattern irrespective of methyltransferase expression. As a control, the C·GC ODN was incubated with RNase-treated J1 nuclear extract. The expected 26 nt digestion product is indicated by an arrow.

Figure 5

Evidence for the existence of an endogenous RNA component essential to T·G mismatch repair. (a) T·GC ODN was treated with RNase A (Lanes 1, 2), J1 nuclear extract (Lanes 3, 4) or aliquots of a J1 nuclear extract were treated with DNase-free RNase A for 30 min (Lanes 5, 6), with RNase A blocked with an excess of RNase A inhibitor (Lanes 7, 8), or treated with RNase A for 30 min followed by addition of RNase A inhibitor for 30 min and then supplemented with 10 μg of J1 total RNA (Lanes 9, 10). The untreated or treated aliquots were used in T·GC repair assays. The expected 26 nt repair product is indicated by the arrowhead. Both the untreated extract and the extract treated with blocked RNase A were competent for repair. The extract treated with RNase A alone was not able to repair the ODN. However, when extracts treated with RNase A followed by RNase A inhibitor were supplemented with J1 total RNA (lanes 9, 10), they regained the ability to correctly repair the T·GC ODN. Naked ODN incubated with RNase A for 16 hs at 37°C was not detectably degraded (lanes 1, 2). M, MspI. H, HpaII. (b) J1 nuclear extracts treated sequentially with RNase A and RNase inhibitor were supplemented with J1 total RNA, E. coli tRNA or RNase-treated salmon sperm genomic DNA as indicated. Reactions were split in half and either digested with HpaII or not digested. Addition of tRNA or genomic DNA resulted in non-specific cleavage of the ODN phosphodiester backbone (lanes 6, 8 respectively). M, sizing markers. H, HpaII. Ø, no digestion.