D.2 DNA methylation in chromatin
D.2.3 DNA methylation of mono-nucleosomes
Radioactive DNA methyltransferase assays with Dnmt3a and Dnmt3b2 were performed with three different substrates: free DNA (MF79/80; 342bp), mono-nucleosomes with symmetric linker DNA (MF79/80; 342bp) and mono-nucleosomes without linker DNA (MF124/125;
147bp) (C.3.5).
In a reaction setup with substrate saturating conditions, both Dnmt3a and Dnmt3b2 methylated free DNA and mono-nucleosomes with linker DNA equally well although the overall methylation efficiency differed from Dnmt3a to Dnmt3b2. This is in agreement with the notion that DNA and mono-nucleosomes with long linker DNA were bound similarly well (C.3.3).
Affirmative, comparable maximum velocities for the methylation of DNA and mono-nucleosomes with linker DNA (220bp, 5S rDNA) were found for Dnmt3a and Dnmt3b (Robertson et al. 2004; Takeshima et al. 2006). However, the amount of CpG sites applied were 5-15 fold higher for mono-nucleosomes indicative of different KM values for DNA and mono-nucleosomes although similar association constants for both substrates were determined in EMSA experiments (Robertson et al. 2004; Takeshima et al. 2006).
Interestingly, nucleosomal DNA was hardly methylated at all (C.3.5). Quantitative analysis revealed a 35-fold and a 27-fold reduction of absolute DNA methylation towards mono-nucleosomes (147bp) of Dnmt3a and Dnmt3b2 respectively. This clearly indicates that DNA wrapped around the histone octamer represents a major obstacle to DNA methylation by the de novo DNA methyltransferases. From the extremely low activity of mono-nucleosomes (roughly 3% compared to ‘naked’ DNA), it is hard to judge whether it is caused by residual free DNA from the nucleosome assembly or from specific methylation within the nucleosome.
Therefore, a detailed single-molecule analysis of DNA methylation with the method of bisulfite conversion (E.2.13) was performed (C.3.6). In contrast to the incorporation of 3H- or
14C labeled methyl-groups, following scintillation counting or PAGE-based quantitation respectively, this technique allows both analysis of each potential CpG target site and discrimination between sense and anti-sense strands.
The bacterial DNA methyltransferase M.SssI which served as a reference, efficiently methylated both the (+) and the (-) strand of the ‘naked’ and the nucleosomal template. DNA methylation did not occur within the nucleosomal core which in accordance with previously described data (Kladde et al. 1999) (Okuwaki and Verreault 2004).
Similarly, Dnmt3b2 methylated both the upper and the lower strand, but less efficient than M.SssI indicating possible sequence preferences (Handa and Jeltsch 2005). Residual DNA methylation was found at CpG sites located at the entry/exist sites of the nucleosome and for
two positions within the nucleosome. The methylation pattern suggests that residual methylation in the 3H-assay originates from a low-level of nucleosomal methylation.
Other data on the methylation of nucleosomes by Dnmt3b pointed out that Dnmt3b was able to methylate within the nucleosome (Takeshima et al. 2006). However, the DNA methylation read-out was based on an error prone radioactive assay analyzing MNase treated and gel-purified nucleosomal fragments. The method of bisulfite treatment is superior since single molecules can be analyzed and possible contamination of free DNA directly be seen.
The reaction efficiency of Dnmt3a towards DNA and nucleosomal matrices was extremely low which originated from a non-saturated reaction setup regarding the amount of template used. Nevertheless, the overall picture of DNA methylation in chromatin by Dnmt3a resembles the results given by Dnmt3b2. In fact, latest DNA methylation experiments using optimal reaction conditions (data not shown) reveal a similar picture of nucleosomal DNA methylation by Dnmt3a.
In contrast, published data claimed equal methylation efficiencies of Dnmt3a towards free DNA and mono-nucleosomes (147bp) (Gowher et al. 2005b) as determined from gel separation based quantitation. However, their bisulfite analysis of nucleosomal DNA showed a methylation efficiency comparable to that of Dnmt3b2, indicating that indeed DNA methylation is restricted within the nucleosome.
In agreement with my results, recent data showed that Dnmt3a scarcely methylated the DNA within the core region, but preferentially methylated the linker DNA between two nucleosomes in vitro (Takeshima et al. 2006; Takeshima et al. 2008).
Our data suggest that DNA methylation within nucleosomes does not occur in vivo at all considering nucleosomal arrays and higher order chromatin structures. In addition, histone H1 was found to inhibit DNA methylation of linker DNA by Dnmt3a (Takeshima et al. 2008).
Nucleosomes represent a major barrier for de novo DNA methylation in vitro indicating that enzymatic activities like chromatin remodeling enzymes that move or assemble/disassemble nucleosomes are required.
As described in the introduction, Dnmt3a and Dnmt3b were found in complexes with remodeling enzymes (A.3.8). Recently, it was demonstrated that different chromatin remodeling machines with intrinsic binding preferences of nucleosomes establish specific nucleosome positioning patterns (Rippe et al. 2007b). Accordingly, in vivo distinct chromatin structures could be created defining DNA accessibility for DNA methyltransferases and other chromatin associated factors. In fact, nucleosome remodeling dependent silencing by NoRC was shown to be a prerequisite for histone H4 deacetylation, H3K9 methylation and DNA methylation at a defined CpG site in the upstream control element of the rDNA gene promoter (Santoro and Grummt 2005).
From a mechanistic point of view, it will be interesting to decipher the underlying molecular mechanisms of chromatin remodeling and DNA methylation.
Since DNA methylation does not occur within the nucleosome, the first question to follow up is whether moving of histones by chromatin remodeling enzymes is sufficient to create accessibility for CpG site methylation. DNA methyltransferase assays with simultaneous
nucleosome ‘sliding’ could be performed. In vitro chromatin remodeling reactions are substoichiometric in regard to remodeling enzymes, hence once the nucleosome is moved, the remodeler should dissociate and not interfere with binding of Dnmts. This notion would imply that the reaction would follow an endpoint and therefore, the released nucleosome position would represent a low-affinity target for the remodeler. With a different chromatin remodeling enzyme, the remodeling reaction could be much more dynamic, i.e. the adjusted nucleosome position also serves as a substrate for the remodeling enzyme and is then moved back to its original position. Whether a ‘window of opportunity’ for DNA methylation between the forward and reverse reaction exists remains elusive. In this regard, it would be intriguing to see whether DNA methylation can affect the DNA sequence/nucleosome affinity of the remodeling machine. If the affinity was lowered by DNA methylation, the nucleosome would be moved back to its original position but would not be relocated again, leading to an equilibrium on the side of the ‘starting’ position but with methylated DNA. If the affinity was increased, the equilibrium would be shifted to the side of the second nucleosome position.
Therefore, modulation of substrate affinities of chromatin remodeling machines through DNA methylation could be a trigger to adjust certain nucleosome positions conferred with a certain transcriptional state. So far unknown active DNA demethylation enzymes or other remodelers with a different substrate affinity could reverse this process.
Alternatively to nucleosome ‘sliding’, disassembly of the nucleosome could be required for full methylation, following nucleosome assembly analogous to chromatin formation after replication.