Rett-causing mutations reveal two domains critical for MeCP2 function and for toxicity in MECP2 duplication syndrome mice

1) The detailed study of R306C adds strength to the paper and shows evidence for the C-terminus basic cluster to have a possible DNA binding role. However, some additional follow-up experiments are desirable but not required to confirm results shown in Figure 7C. For example, (i) comparison to other MeCP2 mutations close to the basic cluster would be useful (ii) proof that DNA binding of the full length MeCP2-R306C is also impaired.

We have performed additional EMSA experiments to address (i) if other mutations in the basic cluster behave similarly and (ii) if the DNA binding of full-length MeCP2-R306C is impaired. We chose to pursue two particularly interesting RTT-causing mutations within the basic cluster: arginine to histidine at residue 306 (R306H) and lysine to glutamic acid at residue 304 (K304E). Both of these mutations change a positively charged amino acid to a relatively uncharged one (histidine) or negatively charged one (glutamic acid), whereas the other mutations in the cluster retain the positive charge. Additionally, within the cluster, R306H is the second most common RTT-causing missense mutation after R306C, indicating that this residue could be of particular importance. Using the same 274-340 amino acid fragment of MeCP2, we found that neither R306H nor K304E bind DNA, thus supporting our argument that the basic cluster flanking R306 is important for DNA binding. We included these findings in the revised Figure 7.

In the context of the full-length protein, we examined the binding of WT MeCP2 and MeCP2-R306C and found that both bind DNA with equal affinity in this in vitro assay. However, this is expected, as the primary mechanism of MeCP2 binding to DNA is via the methyl-CpG binding domain (MBD), thus making subtle differences in binding affinities extremely difficult to detect in vitro. Stronger proof that there is indeed a difference in the DNA binding of full-length MeCP2-R306C comes from the in vivo ChIP experiments. In Figure 7B, it is clear that MeCP2-R306C-EGFP binding is decreased in vivo when compared to the wild type MeCP2-EGFP. Since the MBD is intact in both proteins and the only difference is the R306C mutation, it is safe to conclude that the mutation affects DNA binding in the context of the full-length protein. We made this point more clearly in the revised text.

2) Two of the most pronounced phenotypes reported in MeCP2 overexpression in mice are context dependent fear conditioning and motor coordination impairments on the rotating rod (Collins et al., 2004; Na et al., 2013, which should also be cited). It would strengthen the findings that mutations in either functional domain do not produce phenotypes when expressed with endogenous MeCP2, if the authors examined the R111G-Tg and the R306-Tg mice in these behavioral paradigms. 

We thank the reviewers for the suggestion. We added the missing citation and performed the contextual conditioned fear and rotarod assays in the R111G Tg (new Figure 2E,H) and R306C Tg (new Figure 5D,I,K) mice. We found that there was no difference in these two lines when compared to WT littermates, thus strengthening our conclusion that both functional domains are necessary to see MECP2 duplication phenotypes.

3) The authors demonstrated no behavioral effects of R111G-Tg and R306C-Tg mice in comparison to wild type animals. The authors should address the extent of interaction of R306C-Tg and R111G-Tg (those expressing mutant MeCP2 with endogenous MeCP2) with components of the Sin3a and NCoR repressor complexes. In the R306C-Tg and R111G-Tg mice, does MeCP2 interact with the corepressors equally to what is observed in the wild type animals? 

While we agree with the reviewers that this is an interesting question; the challenge of proving that endogenous MeCP2 interacts with the corepressor complexes in the transgenic animals lies in lack of good commercial MeCP2 antisera. The MeCP2 antibodies are not efficient, hence the reason we added an EGFP tag to our mutant protein. To pull down both WT and mutant MeCP2, it would be necessary to combine the MeCP2-EGFP allele with either the MeCP2-R111G or the MeCP2-R306C allele, on the Mecp2 null background in one mouse. This will take 8 months or so and yet will not impact conclusions as explained in the next paragraph.

We can safely conclude that the WT endogenous MeCP2 is functional as it rescues all the R111G or R306C loss of function phenotypes: the transgenic mice behave exactly like WT animals indicating the endogenous protein is fully functional. Lastly our conclusion that both functional domains are necessary to cause phenotypes in the duplication mice would not be impacted by the outcome of endogenous MeCP2 IP. For example, MeCP2-R111G interacts with both corepressor complexes similarly to WT MeCP2 (Figure 3B), so we would expect the two versions of the protein to interact with corepressors to the same extent in the duplication context. Yet we already know that duplication mice present with phenotypes whereas the R111G Tg mice do not. This indicates that interaction with the corepressors is not sufficient to cause duplication phenotype and requires the methyl-CpG binding domain (MBD).

4) The immunofluorescence staining in Figure 6A demonstrates nuclear punctuated staining of R306C, indicative of DNA binding to heterochromatin. However, the EMSA data of R306C in Figure 7C shows its inability to bind DNA. These two pieces of evidences are somewhat conflicting so the authors should comment on this point.

MeCP2 predominantly binds DNA via its methyl-CpG binding domain (MBD); what we show in our study is that the basic cluster can also bind DNA. To prove the latter point one has to use a fragment (amino acids 274-340 of MeCP2, thus lacking the MBD) in the in vitro EMSA assay. In contrast, the immunofluorescence in Figure 6A depicts full-length MeCP2-R306C localizing to heterochromatin in vivo. We chose to use the C-terminal fragment for the EMSA assay to uncover more subtle binding differences which would otherwise be obscured by the strong MBD binding. The in vivo ChIP assay, which is more quantitative than IF, shows that, on the whole, R306C binds a little less than WT protein in spite of its intact MBD, consistent with our conclusion that the basic cluster contributes some binding. We have clarified the discussion of these data in the text.