Methylation of histone H3K23 blocks DNA damage in pericentric heterochromatin during meiosis

1) The EZL enzymes need to be expressed, purified, and their methylase activities toward H3K23 and H3K27 on either H3 or nucleosomes with K23R or A as internal controls need to be demonstrated.

Respectfully, we believe that that these experiments, while desirable, are not essential to the conclusions of the paper. The activity of a methyltransferase in vitro does not unambiguously define enzyme lysine preference in vivo. In vitro site specificity for methyltransferases (as is the case with most other chromatin-associated enzymes) can vary depending on the substrate, reaction conditions, the stability of the protein, and the associated complex members present in the assay. The EZH2 class of methyltransferases (and by extension, the homologous Tetrahymena EZL methyltransferases) can be particularly sensitive to in vitro assay conditions since other PRC2 complex members are essential for activity. Despite these limitations of the in vitro biochemistry approaches, we have spent over two years attempting to purify the Tetrahymena Ezl3 methyltransferase and complex using Tetrahymena cells, as well as heterologous cell systems.

Recombinant Ezl3 does not purify in a soluble form, and while we have previously tagged and purified endogenous Ezl1 complex (Liu et al., 2007, and data not shown), our efforts to similarly tag endogenous Ezl3 have failed to produce viable strains. In the revised manuscript we have tried to soften language throughout regarding direct modification of H3K23 by Ezl3. Since our exhaustive efforts to characterize Ezl3 in vitro have been hindered by insurmountable technical barriers that will likely take many additional months or years to overcome, and given the scope of this manuscript, we ask that the reviewers provide some leeway in their suggested biochemical experiments. We also point out that EZH methyltransferases are well-documented histone methyltransferases, and ascribing Ezl3 as the key enzyme responsible for H3K23 methylation is the Occam's razor model for our data. Moreover, as knockdown of Ezl3 does not change any other histone methylation site, there is no evidence that Ezl3 causes an indirect, cascade effect that leads to H3K23 methylation. Accordingly, we hope it will not be viewed as a major detraction to exclude this data.

2) In the revised study, you and co-authors should present a statistical analysis of the results. Are they significantly different as they stand? More importantly, you and co-authors should prepare some additional wild-type and mutant strains and add those to the analysis. In Tetrahymena, it should be very straightforward to generate a set of mutants to compare with the wildtype.

For our revised manuscript, we generated and mated a larger group of Tetrahymena strains, and performed statistical analysis on our progeny viability results. This data (in revised Figure 6E) supports our original conclusion that progeny viability is indeed significantly reduced in cells missing H3K23me (p-value < 0.005). The revised figure includes 10 new wild type crosses, and 6 newly generated and crossed mutant strains. We thank the Reviewers for these suggestions as they further support an important point of this manuscript.

3) It is also very important to demonstrate in the revised manuscript more directly that there are new DSBs in the pericentric heterochromatin.

We were able to develop an assay combining FISH and IF (immunoFISH) on Tetrahymena (as suggested by the Reviewers) to provide additional evidence that meiotic DSBs become enriched in pericentric regions in cells missing H3K23me3, and we include this new data in the revised Figure 6D. A FISH probe for the germline-limited highly-repetitive Tlr1 sequence, which is present in hundreds of copies at pericentric regions in the micronucleus, has a pattern of strong overlap with γH2AX signal in cells that are missing H3K23me3, in contrast to the non-overlapping pattern of these markers in wild type cells. We have also moved the model figure to Figure 6–figure supplement 3, as unambiguous validation of this molecular model is not experimentally possible at this time.

4) The mouse data, and to some degree the C. elegans data, are definitely out of place. Please either further develop this data per the reviewers' suggestions (as included below) or just cut this portion out of the paper and just go with the Tetrahymena portion of the study.

We agree with the Reviewers that the mouse data is not as developed as our Tetrahymena and C. elegans data. In the original draft, we had analyzed mouse germ tissue to strengthen the link between H3K23me3 and pericentric or otherwise heterochromatic DNA regions that we observed in Tetrahymena, since these sequences are not well characterized in the ciliate model. We feel the new immunoFISH analysis of meiotic Tetrahymena chromatin that we include in our revised submission better resolves the localization of pericentric regions, and we will remove the mouse data from this submission as per the Reviewers’ request (see revised Figure 7). However, we feel that the C. elegans data remains valuable to maintain in this manuscript since the multicellularity and morphology of C. elegans, and our analysis of the available meiotic mutants, all provide critical support of our central claim that the majority of H3K23me3 is confined to meiotic chromatin. The C. elegans data also underscores that biological roles of H3K23me3 are likely to be distinct from other heterochromatic histone PTMs like H3K27me3 and H3K9me3. These major points would be less convincing if our data was limited to the single-celled Tetrahymena model. Also, we were not attempting to suggest that our data proves that H3K23me impacts C. elegans DSB localization, and the confusing language has been removed in our revised draft.