Rapid epigenetic adaptation to uncontrolled heterochromatin spreading

Reviewer #1:

In the subsection headed “Mst2 and Epe1 are required to counteract the high activity of Clr4”, the authors state: “these cells are able to efficiently silence an otr::ura4⁺ reporter gene.” No data is shown to support this claim and should probably be added to the supplemental section.

We have removed this claim.

Reviewer #2:

My main concern is that the suppression of the mst2 ∆epe1∆ phenotype via silencing of clr4⁺ is a phenomenon that, while interesting and cool, is entirely specific to a mutant situation. Presumably, this never occurs in wild-type cells. In that sense, this paper is more about investigating a mutant pathology than revealing how wild-type cells function.

Although the generation of an epigenetically silenced clr4⁺ allele only happens in mst2∆ epe1∆ background, we can extrapolate that any other mutations or environmental insults resulting in a massive increase in heterochromatin assembly can trigger a similar response. Moreover, our results demonstrate the functional redundancy of Mst2 and Epe1 in controlling heterochromatin assembly in wild type cells.

Minor points:

1) The authors state that the constitutive domains at the centromeres in mst2Δ are in good agreement with the wild type. This statement is only true for the chromosome 1. According to the Figure 1D, we can clearly observe a decrease of the H3K9me signal at the centromere 2 and 3 in mst2Δ compare to the wild type. How can the authors explain these discrepancies?

We have included detailed microarray data around centromere 2 and 3 in Figure 1–figure supplement 1. The apparent higher amount of H3K9me2 at these regions in wild type cells was due to higher values of only a few probes. For the majority of probes, the values are very similar between wild type and mst2∆ cells. We also used thinner lines in Figure 1D to better visualize this effect.

2) The authors' epe1Δ is at odds with those of Zofall (Zofall et al., 2012) who reported subtelomeric spread in epe1Δ cell.

We have included detailed microarray data of the four telomeric regions of chromosome I and II (Figure 1–figure supplement 1). In all cases, the effect of epe1∆ on telomeric heterochromatin is very minor. We currently do not know the reason for the discrepancy with Zofall et al., 2012. It is known that epe1∆ shows variable phenotypes in centromere silencing (Trewick et al., 2007), suggesting that there might be two different sub-populations of cells with different behaviours. We have added a description of this discrepancy in the text (in the second paragraph of the subsection headed “Mst2 regulates histone turnover at heterochromatin”).

3) How do the authors explain that only few heterochromatics islands can show a spread of H3K9me in the triple mutants mst2Δ epe1Δ swi6Δ?

We observed spreading of H3K9me at the majority of heterochromatic islands in mst2∆ epe1∆ swi6∆ cells. Due to space limitations, we only showed one in the original version. We have now included detailed data of all heterochromatin islands in Figure 3–figure supplement 1.

4) Is RNAi required for ectopic silencing establishment?

We found that RNAi is not required for ectopic silencing establishment at the clr4⁺ locus. We crossed mst2∆ dcr1∆ with epe1∆ dcr1∆ and found that all mst2∆ epe1∆ dcr1∆ cells can still recover, consistent with our ChIP analysis that there is H3K9me at the clr4⁺ locus in these cells (Figure 5–figure supplement 1).

5) In the Introduction, the authors should indicate that the establishment of constitutive heterochromatin at repetitive DNA elements requires RNA interference at the centromeres but not at other element such as telomeres (Ragunathan et al., 2014).

We disagree with the reviewer's statement. It has been shown that the DNA repeats at telomeres contribute to heterochromatin establishment (Kanoh et al., 2005). The ability to establish heterochromatin at telomeres in RNAi mutant, as described in Ragunathan et al., 2014, is because of redundant heterochromatin establishment pathways at telomeres (Kanoh et al., 2005).

6) The authors do not mention the recruitment of RNAi to these specific meiotic genes by the Mmi RNA surveillance machinery (Hiriart et al., 2012).

We added a sentence in the fourth paragraph of the Introduction to describe the recruitment of RNAi to meiotic genes.

7) The authors state that they identified a few heterochromatic islands with a low level of H3K9me. This analysis should be explained more clearly in the Methods.

We stated in the Methods section the specific cut off used to determine heterochromatin islands. A list of the heterochromatin islands was included in Supplementary File 1.

8) For a better understanding, the authors should add on the x axis the title “Chromosome position” on the ChIP-chip figures.

We added axis label “Chromosome position” for all ChIP-chip figures.

9) In the subsection headed “Misregulation of heterochromatin affects the fitness of mst2∆ epe1∆ cells”, the two mutants mst2Δ and epe1Δ appear to have a very different phenotype on telomeres (Figure 1E).

Even though mst2∆ and epe1∆ have similar effects on histone turnover, ectopic heterochromatin assembly (Figure 1), and on suppression of RNAi mutants (Reddy et al., 2011 and Trewick et al., 2007), they behaved differently in the spreading of heterochromatin domains at telomeres and centromeres (Figure 1), as well as heterochromatin maintenance in the absence of initiation signals (Ragunathan et al., 2014), suggesting that these two mutants are not identical.

10) Figure 2A, how many days of colony growth?

The picture shown in Figure 2A is 6 days of growth after dissection of tetrads. We added the description in the figure legend.

11) Figure 2B, the time of growth is not described.

The dilution analyses were performed after 1 day of growth in rich medium. We added the description in the figure legend.

12) Figure 2G, how did the RNAi factors score in this screen?

RNAi factors were not identified in our screen. We added a description of this fact in the text (in the subsection headed “Misregulation of heterochromatin affects the fitness of mst2∆ epe1∆ cells”). We also showed a tetrad dissection of an mst2∆ dcr1∆ and epe1∆ dcr1∆ cross in Figure 2–figure supplement 2.

13) At the end of the subsection headed “Increased heterochromatin spreading is responsible for the initial growth defects of mst2∆ epe1∆ cells”, add clr4⁺ locus.

We changed the text to “mei4⁺ and clr4⁺ loci”.

14) Figure 4D, mst2Δ Flag-clr4⁺ is missing.

We added a panel in Figure 4D to show different combinations. There are no growth defects associated with mst2∆ Flag-clr4.

Reviewer #3:

1) The histone turnover experiment (Figure 1) would benefit from the inclusion of total H3 control ChIP in WT and mutant backgrounds. Also, since histone turnover is known to be much greater on highly transcribed genes such as act1, it would be wise to also compare turnover levels at dh to repressed genes (e.g. nmt1 or other) where histone turnover should be low and thus provide a more accurate comparison. This would also provide an internal control to confirm that differences in histone turnover levels at different loci are being detected (i.e. act1 v nmt1).

We normalized our data to a non-transcribed mating-type region, which showed low histone turnover (Aygun et al., 2013), in new Figure 1C. The results are similar to those without such normalization.

2) What is the frequency of fast growing colonies with clr4⁺ epialleles (and rik1⁺ epialleles) as opposed to the frequency of genuine mutations that inactivate clr4 or genes encoding other heterochromatin components?

Our data suggest that the clr4⁺ epiallele is much more prevalent than genuine genetic mutations. For example, experiments in Figure 4 showed that the suppressor is epigenetic rather than genetic, as cells inherited the silenced clr4⁺ epiallele can revert back to normal. The results shown are representative of multiple independent suppressor strains. The accurate determination of the frequency of epiallele vs. genuine mutations requires genome sequencing of independent suppressor strains.

3) Quantitative PCR ChIP, RT-PCR, ChIP-chip: It is not immediately clear from figures how ChIP signals were quantified and what is presented. My assumption was that it is %IP shown. Only when I read the methods/legends did I discover that all values are normalized relative to the act1 gene. For ChIP the Y axis in all figures should be clearly labelled H3-FLAG/H3K9me2/ at dg/clr4/rik1 relative to act1 (or H3K9me2 ratio dg/act1 etc). Likewise, clr4/rik1 transcript/mRNA relative to act1 in all relevant figures.

For ChIP-chip analyses, details should be provided for how H3K9me2 profiles were obtained and the methods utilized should conform with normally accepted practices for microarray data processing and presentation. From Figure 1 to 3, it looks as if all values may be plotted relative to a control locus? The processing employed and units used need to be clearly described and indicated.

We added labels in all relevant figures when a control locus was employed for normalization. For ChIP-chip analysis, we normalized data in order to average results from different experiments. We explained in the Methods section about the details of data processing.

4) The data in Figure 5 is discussed in a section entitled: “The mechanism of heterochromatin assembly at the clr4⁺ locus”. Here the authors just show that manipulation of sequences flanking the 3' end of the clr4 gene prevents heterochromatin assembly over clr4 in mst1 epe1 double null cells. They find that dcr1/RNAi, Mmi1/exosome and Pab2/3' end processing pathways are not involved. Red1 should be also tested as it is more directly involved in the recruitment of Clr4 to some heterochromatin islands. It would also make sense to test Rrp6 and Mtl1.

We changed the section title to: “Sequence 3’ to clr4⁺ is required for heterochromatin assembly at the clr4⁺ locus in mst2∆ epe1∆ cells”. We attempted to generate triple mutations of mst2∆ epe1∆ red1∆ and mst2∆ epe1∆ red1∆ rrp6∆. However, we were unable to generate such strains. Both red1∆ and rrp6∆ cells are extremely sick, making it difficult to determine whether the lethality is due to the cumulative sickness of these strains or because such strains cannot generate clr4⁺ epialleles. Mtl1 is an essential gene, and we currently do not have mtl1 mutant.

No explanation is given for how they conclude from RNA-seq data (Figure 5A) that the clr4 transcript runs through the convergent meu6 gene. Is the same sized transcript detected by northern with meu6 and clr4 probes? Is 5' RACE consistent with a clr4 (sense)-meu6 (antisense) RNA being synthesized? Regardless, the experiments presented do not actually reveal the mechanism of heterochromatin assembly at the clr4 locus in mst2 epe1 null cells. How does this 3' region nucleate heterochromatin? If it is something to do with the 3' UTR of the clr4 RNA then a more precise insertion of a known transcription terminator downstream of the Clr4 ORF should prevent heterochromatin assembly.

Does the clr4 locus have anything in common with the rik1 locus and other H3K9me2 islands detected in mst1 epe1 swi6 null cells? Apart from meu6-clr4, are there other H3K9me2 islands specific to mst1 epe1 swi6 null cells that might be informative with respect to mechanism?

Our RNA-seq data is strand specific, which clearly indicated that the clr4⁺ transcript covers the meu6⁺ coding sequence and that the meu6⁺ mRNA levels are very low. Such a conclusion is consistent with the annotation of genome in Pombase based on results of Marguerate et al., 2012 and Rhind et al., 2011. Therefore, we believe a 5'-RACE experiment is unnecessary.

We currently do not know the mechanism that induces heterochromatin assembly at the clr4⁺ locus, except that sequences 3' to clr4⁺ is required for heterochromatin assembly. But we did rule out most of the known heterochromatin establishment mechanisms, such as RNAi, Mmi1, and convergent transcription, indicating that a novel mechanism is involved. We are actively dissecting the mechanisms that establish this ectopic heterochromatin, including precisely manipulating transcription termination at the clr4⁺ locus, and we believe such an experiment is out of the scope of the current study.

We found that a common theme of the clr4⁺ and rik1⁺ loci is that both are adjacent to genes up-regulated during meiosis, although whether there is any causal relationship is still unknown. Apart from meu6-clr4, there are other H3K9me2 islands in mst2∆ epe1∆ swi6∆ cells, such as mug8, SPAC8c9.04, and sol1. However, meu6-clr4 does not show accumulation of H3K9me in mst2∆ epe1∆ swi6∆ cells. Therefore, they are not likely to be formed through similar mechanisms.

Related to this, did the screen of the deletion library (Figure 2D) identify any additional genes that suppress the mst2 epe1 slow growth phenotype but do not correspond to known heterochromatin components? If so, presumably they might suggest how meu6-clr4 is silenced and these should at least be discussed.

There are a number of additional mutants identified, which are listed in Supplementary file 2. Due to the fast generation of the clr4⁺ epiallele, such an experiment is intrinsically noisy and many of the identified factors are simply noise. The screen is not designed to identify factors required for heterochromatin assembly at the clr4⁺ locus. Theoretically, such a mutation will result in cells that cannot recover, which can be tested by repeated pinning of the cells to measure their growth. However, there are many strains in the library causing growth defects, and the linkage of genes to three different loci will result in many SGA spots showing growth defects, thus the noise will be too high for such experiments to be informative.

5) More extensive analysis is required to allow the authors to conclude that clr4-R406H “mildly affected silencing” and that “these cells are able to efficiently silence the otr::ura4⁺ reporter”. A simple 5-FOA plate, with no -ura4 plate for comparison, is not sufficient. Are H3K9me2 levels of dg/dh repeats reduced? Are dg/dh transcript levels increased?

We have removed this claim. We have included RT-PCR analysis of otr::ura4⁺ transcript levels, shown in Figure 6–figure supplement 1.