Histone H3G34R mutation causes replication stress, homologous recombination defects and genomic instability in S. pombe

Recurrent somatic mutations of H3F3A in aggressive pediatric high-grade gliomas generate K27M or G34R/V mutant histone H3.3. H3.3-G34R/V mutants are common in tumors with mutations in p53 and ATRX, an H3.3-specific chromatin remodeler. To gain insight into the role of H3-G34R, we generated fission yeast that express only the mutant histone H3. H3-G34R specifically reduces H3K36 tri-methylation and H3K36 acetylation, and mutants show partial transcriptional overlap with set2 deletions. H3-G34R mutants exhibit genomic instability and increased replication stress, including slowed replication fork restart, although DNA replication checkpoints are functional. H3-G34R mutants are defective for DNA damage repair by homologous recombination (HR), and have altered HR protein dynamics in both damaged and untreated cells. These data suggest H3-G34R slows resolution of HR-mediated repair and that unresolved replication intermediates impair chromosome segregation. This analysis of H3-G34R mutant fission yeast provides mechanistic insight into how G34R mutation may promote genomic instability in glioma. DOI: http://dx.doi.org/10.7554/eLife.27406.001


Introduction
The genomic DNA of eukaryotes is packaged into chromatin, which regulates all DNA transactions including transcription, replication and repair. The basic subunit of chromatin is the nucleosome, which is made up of 147 bp of DNA wrapped around a core octamer of histone proteins. Proteins that regulate chromatin dynamics are frequently mutated in cancer, and recently mutations in the histone genes themselves have been discovered (Schwartzentruber et al., 2012;Wu et al., 2012;Behjati et al., 2013). One such mutant is the Glycine 34 to Arginine (G34R) mutant of histone H3.3, identified as a frequent somatic mutation in pediatric high-grade cortical glioma (pHGG). H3.3 is a non-canonical H3 variant, whose deposition is not linked to replication and which can accumulate in post-mitotic cells and in areas of high transcriptional activity (Skene and Henikoff, 2013). Notably, G34R mutations are only found in H3.3, primarily in one of two H3.3 genes (H3F3A) and not in the replication coupled H3 proteins H3.1 and H3.2, which are encoded by a total of 13 genes. G34R mutations are commonly found in tumors additionally mutant for the tumor suppressor p53 and ATRX, a chromatin remodeler involved specifically in nucleosomal deposition of the H3.3 variant histone at telomeres and pericentric heterochromatin (Goldberg et al., 2010;Lewis et al., 2010;Drané et al., 2010). While p53 and ATRX mutations may contribute to the genomic instability exemplified by these gliomas, the specific role of the G34R H3.3 mutant in disease remains a mystery.
One published study showed G34R nucleosomes had reduced K36me2/3 on the same H3.3 tail (Lewis et al., 2013), suggesting a potential role in regulating H3K36 methylation in gliomas. Furthermore, mutations in SETD2, the enzyme that performs H3K36 di-to tri-methylation in mammals (Edmunds et al., 2008), have also frequently been identified in pHGG (Fontebasso et al., 2013). However, SETD2 mutations are not found in conjunction with histone mutations, suggesting the importance of disruption of the H3K36me3/ SETD2 axis in cortical pHGG. To gain biological insight into the role of the G34R mutant, we engineered fission yeast to express either wild-type or G34R mutant histone H3. In fission yeast, a single enzyme, Set2, performs all methylation on H3K36 (Morris et al., 2005). We hypothesized that H3-G34R mutation may reduce Set2 function and influence important chromatin-templated processes such as DNA transcription, replication and repair, which are highly conserved between fission yeast and human.
Here, we show that histone H3-G34R mutation specifically reduced tri-methylation, but not dimethylation of H3K36, and caused a decrease in H3K36 acetylation. However, G34R mutants displayed different phenotypes to set2D, suggesting that defective H3K36 modification may not be the sole defect in G34R. H3-G34R mutants showed chromosome instability and sensitivities to DNA damaging agents which are distinct from cells lacking Set2. H3-G34R mutants had defects in HRdirected repair, which may be attributed to a delay in DNA repair dynamics at compromised replication forks. Together, our work provides valuable insight to the potential role of H3-G34R mutations in pediatric high-grade gliomas. eLife digest Children suffering from a brain cancer called high-grade glioma rarely recover because there are no therapies that can effectively target this disease. Recently, mutations in a gene that encodes a protein called histone H3 were found in children's glioma cells. Histone proteins bind to DNA to help package it into cells. They help to protect the DNA from damage, control when genes are switched on or off, and influence how the DNA is copied when cells are preparing to divide to produce two daughter cells. However, little was known about why one of the histone H3 mutations is associated with high-grade gliomas.
Humans and other animals have many different versions of the histone H3 protein, which makes it difficult to study a mutation that only affects one particular version. Therefore Yadav et al. used fission yeast -which they engineered to only contain one form of histone H3 -to study what effect the mutation has on cells.
The experiments show that fission yeast cells with the mutant form of histone H3 took longer to replicate their DNA and were more likely to die when exposed to chemicals that interfere with DNA duplication. Furthermore, the mutant cells were slower at repairing certain types of damage to DNA and were more likely to go on to divide before DNA duplication and repair were completed. As a result, the daughter cells produced were more likely to have gained or lost whole chunks of their DNA. This phenomenon is known as chromosomal instability and is often seen in cases of highgrade glioma in children.
The findings of Yadav et al. give new insight into how a specific mutation affecting the histone H3 protein may act in cells. Future experiments will need to confirm whether this mutation also has a similar effect on children's glioma cells, which may help to provide new options for treating this cancer.

Results
In contrast to the complexity of mammals, where fifteen genes encode three different histone H3 proteins (H3.1, H3.2, H3.3), fission yeast have three genes that code for a single histone H3 protein ( Figure 1a) (Matsumoto and Yanagida, 1985). Fission yeast H3 has features of both H3.1 and H3.3 proteins of higher eukaryotes (Figure 1-figure supplement 1a). To simplify the analysis of histone H3 and H4 mutants, strains have been derived that express only one H3 and one H4 gene (hht2 + and hhf2 + ) (Mellone et al., 2003). These strains maintain histone protein levels similar to those in wild-type strains (Figure 1-figure supplement 1b), likely through upregulation of hht2 + and hhf2 + expression. To assess the effect of G34R mutation, we introduced mutations into hht2 + (Figure 1a), and derived strains that express only the mutant histone H3 gene (Figure 1-figure supplement  1c). These strains are viable at a range of temperatures and appear to have normal chromosomal integrity (Figure 1-figure supplement 1d,e). 'Sole copy' histone H3 and H4 strains were used throughout this study and are named H3-WT and H3-G34R in the text, and 'WT' and 'G34R' in the figures. Additional mutants were all in the sole copy H3-WT background unless indicated otherwise by inclusion of (3xH3) denoting 3 copies of H3/H4.

Post translational modification of H3K36 is altered in H3-G34R mutants
In mammalian cells, ectopic expression of H3.3G34R reduced K36me2/3 methylation on the mutant histone tail (Lewis et al., 2013). Five enzymes can methylate H3K36 in mammals, with only SETD2 responsible for H3K36me3 (Edmunds et al., 2008). In fission yeast, a single enzyme (Set2) performs all methylation on H3K36 (Morris et al., 2005). We began our investigation by determining if H3-G34R mutation reduces Set2 H3K36 methyltransferase activity. Because of the proximity of G34R to K36, we tested whether antibodies can specifically recognize K36 methylation states in the G34R mutant. Using di-or tri-methylated K36 H3 peptides as targets, we identified two anti-K36me3 antibodies that were minimally affected by G34R mutation, and one antibody against K36me2 whose binding was weakened~10 fold by the G34R mutation ( Figure 1b, Figure 1-figure supplement 1f). Figure 1-figure supplement 1f-g serve to demonstrate reproducibility of the H3K36me3 methylation pattern with an alternate H3K36me3-specific antibody. Using these antibodies, we found that H3K36me3 was markedly reduced in H3-G34R compared to H3-WT cell extracts, but that H3K36me2 levels were unchanged ( Figure 1c, Figure 1-figure supplement 1g). As we observed that binding of the H3K36me2 antibody is reduced~10 fold on G34R peptides, these results suggest that K36me2 is elevated in H3-G34R cells. Thus Set2 function is altered in H3-G34R cells, with specific reduction of H3K36me3 and retention/ accumulation of H3K36me2.
Since endogenously tagged Set2-FLAG protein levels were similar in H3-WT and H3-G34R cells (Figure 1-figure supplement 1h), the reduction in H3K36me3 is not caused by a loss of set2 + expression in H3-G34R cells. Additionally, chromatin immunoprecipitation (ChIP) analysis revealed that Set2-FLAG recruitment to chromatin was not reduced in H3-G34R cells (Figure 1d), suggesting that the H3-G34R mutation hinders Set2 activity but not access to K36. We therefore asked if we could force the generation of H3K36me3 in H3-G34R cells by overexpression of the Set2 methyltransferase ( Figure 1-figure supplement 1i). Western analysis of extracts prepared from cells that overexpress Set2 showed that pSet2-FLAG restored K36me2 and me3 in H3-WT set2D strains but was able to generate only low levels of H3K36me3 in both H3-G34R and H3-G34R set2D cells (Figure 1e). Thus the H3-G34R mutation reduces Set2 activity on the H3 tail, leading to a specific reduction in tri-methylation of H3K36, and this effect cannot be bypassed by overexpression of Set2.
In addition to methylation of H3K36, this residue is also regulated by acetylation (Morris et al., 2007). To address whether H3-G34R impedes acetylation at H3K36, we used quantitative mass spectrometry to perform targeted analysis of acetylation of tails of histone H3 and H4 in histones acid extracted from H3-WT, H3-G34R and set2D strains (Figure 1f) (Kuo and Andrews, 2013). Calibration was performed to monitor the elution time of the acetylated and propionylated tryptic peptides and transitions were created to study acetylation of H3-WT and H3-G34R as well as the H4 tails (see Supplementary files 1 and 2 and Materials and methods). These studies showed acetylation of K36 on histone H3 was greatly reduced in H3-G34R strains and somewhat reduced in set2D. Additionally, there was a slight upregulation of H3K18 acetylation and possibly enhanced K27 acetylation in H3-G34R, although K27 acetylation in H3-WT was highly variable. set2D cells showed more widespread changes, with a slight reduction of H4 K16 acetylation, some induction of H3K9 and K18 acetylation  and a marked increase in H3K14 acetylation. In summary, H3-G34R cells showed a focal pattern of chromatin changes centered on inefficient tri-methylation and reduced acetylation of histone H3K36, whereas set2D effects were more widespread. H3K36me3 and H3K36 acetylation have been shown to be cell cycle regulated (Li et al., 2013;Pai et al., 2014), with K36me3 accumulating in G1 and early-mid S phase. We assessed the consequence of H3-G34R mutation on K36me3 accumulation at several stages of the cell cycle, using cells synchronized in G1, S, G2 and M phases (Figure 1-figure supplement 2). Although levels of H3K36me3 were reduced in H3-G34R cells compared to H3-WT at all stages of the cell cycle (Figure 1-figure supplement 2a,b,c), H3K36me3 did accumulate in both H3-WT and H3-G34R cells synchronized in G1, indicative that the G34R mutation does not override cell cycle control of Set2 function. Additionally, this experiment also demonstrates that H3K36me2 levels are higher throughout the cell cycle in H3-G34R compared to H3-WT cells.
H3-G34R mutants are transcriptionally distinct from set2 cells Post-translational modifications of H3K36 are important for transcriptional control, raising the possibility of transcriptional disruption in H3-G34R cells. Cells that lack the H3K36 methyltransferase Set2 or that express a mutant of Set2 that allows only di-methylation of K36 show a similar widespread transcriptional upregulation in fission yeast (Suzuki et al., 2016;Matsuda et al., 2015), suggesting that H3K36me2 is not sufficient for silencing. Since H3-G34R cells exhibit reduced H3K36me3 and possibly enhanced K36me2, as well as a reduction in H3K36Ac, we asked how H3-G34R mutation influences transcriptional control.
In genome-wide RNA-Seq analyses, set2D had many transcripts with altered regulation (753 genes were up and 149 genes downregulated) (Figure 2a,b) in an organism with~5100 protein coding genes and~1500 non coding RNAs (Supplementary file 3). Unexpectedly, the H3-G34R mutant showed more mild transcriptional effects (repression of 69 genes and upregulation of 172 genes). This difference in gene regulation may stem from additional chromatin changes such as increased H3K14Ac in set2D.
143 genes were deregulated in both H3-G34R and set2D cells: 97 were coordinately regulated, with 9 genes downregulated, and 88 genes upregulated in both set2D and H3-G34R ( Figure 2b). In most other cases (44/46), transcripts accumulated in set2D but were repressed in H3-G34R. As can be seen on chromosome-wide plots (Figure 2c,d), transcripts that are differentially regulated between H3-G34R and set2D map to regions adjacent to subtelomeric regions of chromosome 1 and 2 (ST or sub-subtelomeric domains [Matsuda et al., 2015;Buchanan et al., 2009]), which are not heterochromatic and are normally expressed at low levels (Mata et al., 2002). We confirmed gene regulation trends for 2 genes from ST domains by real time qPCR (Figure 2e), and additionally show that transcript levels for these genes in the double mutant, H3-G34R set2D, are similar to set2D cells. Thus H3K36me2 is perhaps critical for the repression in ST domains seen in H3-G34R cells. Intriguingly, these domains have been recently shown to associate with Shugoshin (Sgo2) to form highly condensed subtelomeric 'knob' regions which are transcriptionally silenced and delayed for replication (Matsuda et al., 2015;Tashiro et al., 2016).
Set2 has also been shown in both S. cerevisiae and S. pombe to repress antisense transcription (Nicolas et al., 2007;Shim et al., 2012;Venkatesh et al., 2016). Focusing on just these antisense transcripts, of which there are 635 in fission yeast, we found that, compared to H3-WT cells, 318 were differentially regulated in set2D, and 62 in H3-G34R mutants, with the vast majority being upregulated (313 and 57, respectively) ( Figure 2f). 47 antisense transcripts were targeted by both Set2 and H3-G34R, and 43 of these accumulated in both mutants. Therefore H3-G34R cells, similar to    In conclusion, interpretation of the transcriptional profiling results is complex. There is some overlap with Set2-mediated control as shown by similar regulation of some sense and some antisense genes, but also differences as evidenced by opposite regulation of transcription with sub-subtelomeric domains. We conclude that suppression of antisense gene transcription appears linked to trimethylation of K36, as it is reduced in both H3-G34R and set2D, and that repression of subsubtelomeric domains may be linked to dimethylation of K36 in H3-G34R, as it is abolished in set2D or compound set2D H3-G34R mutants that lack dimethylation of K36.

Heterochromatin is maintained in H3-G34R cells, but cells exhibit genomic instability
RNA-Seq provides information on unique transcripts. To determine whether constitutive heterochromatin that forms on repetitive sequences was affected by H3-G34R mutation, we monitored transcripts from pericentromeric loci and the subtelomeric tlh genes in H3-WT and H3-G34R cells (Figure 2-figure supplement 1a), with clr4D serving as a control for loss of heterochromatin (Ekwall et al., 1996;Allshire et al., 1995). Transcript levels were similar in H3-G34R and H3-WT, suggesting that heterochromatin is intact in H3-G34R mutants, whereas set2D showed a slight upregulation in subtelomeric tlh transcripts.
Additionally we monitored H3K9me2, a conserved heterochromatin mark, and the Swi6 HP1 protein which binds to that mark (Nakayama et al., 2001;Bannister et al., 2001). As shown in Figure 2-figure supplement 1b and c, H3-G34R mutation had no impact on H3K9 methylation or Swi6 HP1 recruitment to heterochromatic regions . ChIP against Cnp1 (Castillo et al., 2007), the fission yeast CENP-A protein that underlies and supports kinetochore assembly, showed no change in Cnp1 recruitment to centromeric central core sequences in H3-G34R mutant cells (Figure 2-figure supplement 1d). Together, the lack of apparent effect on kinetochore or heterochromatic structures indicates that the architecture of centromeric and telomeric heterochromatin is intact in H3-G34R cells.
We next tested genomic stability in H3-G34R mutants by monitoring the frequency of chromosome loss using a non-essential minichromosome Ch16 (Niwa et al., 1989). As expected, H3-WT cells showed low levels of loss of Ch16 (0.9% cells) (Mellone et al., 2003) and cells lacking the heterochromatin protein Swi6 HP1 displayed high frequencies of Ch16 loss (6.4%, Figure 3a). Interestingly, we found that H3-G34R cells lost minichromosomes at an elevated frequency (3.5%), while minichromosome loss in set2D was not significantly upregulated (1.6%). We further tested for specific chromosome segregation defects in H3-G34R strains by monitoring frequencies of lagging chromosomes on late anaphase spindles, using anti-tubulin antibodies to stain the spindle, and DAPI to stain DNA ( Figure 3b) . Counting only late anaphase cells (spindle >10 microns), we found that chromosomes mis-segregated in 7.4% of H3-G34R cells compared to 0.9% in H3-WT, 1.3% in set2D and 25.6% in clr4D cells that lack heterochromatin. Furthermore, in cells synchronized using an nda3-KM311 early mitotic block and release (Hiraoka et al., 1984), and stained for tubulin and DNA,~38% of H3-G34R cells showed stretching of DNA or aberrantly positioned DNA on mid to late anaphase spindles compared with~14% of H3-WT cells ( Figure 3c). Together, these data show that the H3-G34R cells exhibit elevated chromosomal instability, which is independent of heterochromatin or kinetochore defects. The nature of the chromosome segregation defects in H3-G34R cells is suggestive of a defect in DNA condensation or resolution of sister chromatids after DNA replication (Saka et al., 1994).

H3-G34R mutants show DNA damage sensitivity upon replication stress
Since H3-G34R cells show increased chromosomal instability, and are reduced in histone H3 K36 acetylation and tri-methylation, which are modifications that are linked to DNA damage responses (Pai et al., 2014;Pfister et al., 2014;Jha and Strahl, 2014;Li et al., 2013), we tested H3-G34R strains for sensitivity to DNA damaging agents. Cells were plated on media containing hydroxyurea (HU; a ribonucleotide reductase inhibitor that depletes dNTPs), camptothecin (CPT; an agent that blocks topoisomerase 1 ligase activity), or methyl methanesulfonate (MMS; an alkylating agent) (Figure 4a). At the concentrations used, each of these agents predominantly affects DNA replication (HU and MMS), or requires DNA replication to inflict DNA damage (CPT). We found that chronic exposure to each genotoxin resulted in significant sensitivity of H3-G34R cells compared with H3-WT cells. Conversely, when H3-G34R and H3-WT cells were exposed to gamma irradiation (gIR) or bleomycin, two replication-independent genotoxins, H3-G34R mutants showed no sensitivity, whereas set2D cells were sensitive to gIR and bleomycin (Figure 4b and c) as seen previously (Pai et al., 2014). Thus H3-G34R, but not set2D cells appear selectively sensitive to agents that generate DNA replication stress.
To further address the relationship of the H3-G34R mutant with Set2, we tested genetic interactions between set2D and H3-G34R. In the absence of damage, there were no synthetic growth defects in the set2D H3-G34R double mutant. set2D were more sensitive than H3-G34R to MMS and a. b.
c.    CPT, but showed similar sensitivity to H3-G34R on HU. Additive growth defects were seen for set2D H3-G34R double mutants on CPT and MMS, but not with HU ( Figure 4d). Repair of damage from CPT or MMS-induced lesions likely involves the use of Holliday junctions, whereas recovery from HUinduced stress generally does not . The enhanced sensitivity of double mutants to CPT and MMS may indicate that set2D and H3-G34R mutants are defective in distinct aspects of HR-mediated repair requiring assembly of Holliday junctions. We further asked whether overexpression of Set2 could compensate for the DNA damage sensitivities of H3-G34R cells. Whereas pSet2-FLAG overexpression fully compensated for set2D growth defects on damaging agents, no growth advantage was conferred on H3-G34R cells ( Figure 4e). However, since Set2 overexpression cannot compensate for H3K36me3 in H3-G34R cells (Figure 1e), it is unclear whether the DNA damage sensitivity of H3-G34R cells is linked to the lack of efficient tri-methylation on K36.

H3-G34R cells have intact DNA replication checkpoints
We have shown that H3-G34R mutants are sensitive to chronic exposure to replication stress. DNA damage checkpoint signaling is conserved in fission yeast ( Figure 5a) (Melo and Toczyski, 2002). Following DNA damage, the upstream sensor kinases Rad3 (ATR) and Tel1 (ATM) phosphorylate histone H2A on a C-terminal serine to generate gH2A and activate the downstream DNA damage checkpoint kinases Chk1 (Capasso et al., 2002) and the replication checkpoint kinase Cds1 (Lindsay et al., 1998).
To address whether DNA damage checkpoint signaling is functional in H3-G34R cells, we first monitored phosphorylation of H2A (gH2A). Western analyses of cells treated with HU or MMS showed no differences in gH2A phosphorylation levels in H3-G34R or set2D strains compared to H3-WT, indicative that the function of upstream kinases is not perturbed in H3-G34R cells ( Figure 5b). Next, we monitored phosphorylation-related mobility shifts of Chk1-HA and found that Chk1 was efficiently activated by MMS treatment in H3-G34R cells, suggesting that DNA damage checkpoint signaling was normal (Figure 5c) (Walworth and Bernards, 1996). To assess Cds1 replication checkpoint function, we performed HU treatment of H3-G34R cells since, in cds1D cells, prolonged DNA synthesis occurs both during HU treatment and after release. This causes an accumulation of single stranded DNA (ssDNA) that can be visualized as large foci of replication protein A (RPA) (Sabatinos et al., 2012). Cells were arrested in HU, and RPA foci monitored 6 hr post-release. In contrast to the 'flares' of RPA seen in cds1D after release, H3-G34R showed no accumulation of RPA foci (Figure 5d), supporting that the Cds1 pathway is functional in H3-G34R mutants.
Finally, we tested H3-G34R in combination with checkpoint mutants for their ability to arrest after a prolonged exposure to genotoxin (HU), which can be monitored by elongation of the cells as an indication of proper checkpoint arrest. Control cells lacking both checkpoints (cds1D chk1D) failed to arrest after HU treatment, as indicated by their short length and entry into lethal mitoses with incompletely replicated chromosomes (Figure 5e) (Lindsay et al., 1998). Cells lacking either cds1 or chk1 elongated properly, likely arrested by compensation of the other kinase. H3-G34R cds1D and H3-G34R chk1D double mutants also arrested normally as elongated cells, suggesting that both checkpoints are intact in H3-G34R (Figure 5e). In summary, DNA damage checkpoint signaling appears to be functional in H3-G34R cells.

H3-G34R cells show defects in homologous recombination pathways
Thus far, we have shown that H3-G34R mutants exhibit genomic instability that does not involve disruption of heterochromatin (Figure 3), that they are sensitive to DNA replication stress (Figure 4), and have functional DNA damage checkpoint signaling ( Figure 5). We next assessed DNA damage repair pathways in H3-G34R cells (Figure 6a). First we asked how cells recover from replication stress. After HU treatment, completion of DNA replication occurs without the need for Origin Replication Complex (ORC) activity. However, some mutants that are sensitive to replication stress, such   as a mutant of the PTIP homolog (Cho et al., 2003), brc1D, require ORC to recover from HU arrest (Bass et al., 2012). We tested whether H3-G34R cells depend on ORC to recover from HU, by use of a temperature sensitive allele in ORC, orp1-4. Cells were synchronized with HU at 25˚C, and released at either 25˚or 36˚C for 4 hr prior to plating ( Figure 6b). As expected, any orp1-4 cells that are released to 36˚C die (green and purple bars), whereas orp1-4 H3-WT cells that are HU arrested and released at 25˚C survive. Intriguingly, unlike orp1-4 brc1D which die due to severe mitotic defects (Bass et al., 2012) (red bars, Figure 6b), we found that orp1-4 H3-G34R survive. Thus, H3-G34R cells do not rely on ORC to recover from HU-imposed replication stress. Next, we systematically mutated genes that are key to well defined DNA repair pathways in H3-G34R and H3-WT cells. We mutated genes for post-replication repair (PRR), non-homologous end joining (NHEJ), and homologous recombination (HR) in H3-WT and H3-G34R cells (Figure 6c-e). To test PRR, we generated combination mutants of H3-G34R with either rad18 (important for translesion synthesis) or ubc13 (important for template switching) . H3-G34R mutants with either of these mutations show synergistic sensitivity to MMS, suggesting that the PRR pathways are functional in H3-G34R cells (Figure 6c). Additionally we monitored epistasis with a mutant in the NHEJ pathway. Combination mutants of H3-G34R were generated with mutants in Ku70, a subunit of the Ku complex responsible for recognizing double strand breaks (DSBs) (Manolis et al., 2001). H3-G34R showed synergistic defects with ku70D when assessed on MMS compared with either single mutant, indicating that NHEJ is functional in H3-G34R cells (Figure 6d).
Furthermore, we assessed the genotoxin sensitivity of H3-G34R in combination with a mutant necessary for HR-directed repair. Combination mutants were generated with rad51D (Rad51 encodes a strand exchange protein that forms a filament on DNA and facilitates the search for homology). In contrast to mutations with PRR or NHEJ proteins, H3-G34R rad51D double mutants showed epistasis with single mutants when plated on MMS (Figure 6e). This suggests that H3-G34R cells have a defective HR pathway.
To extend this analysis, we developed an assay to directly monitor the frequency of HR-directed repair ( Figure 6-figure supplement 1a). We transformed leu1-32 mutant cells (that bear a single nucleotide mutation in leu1 that renders cells auxotrophic for leucine) with a 576 bp fragment of wild type leu1 + and scored leu1 + transformants that can only arise by HR, normalizing for transformation efficiencies of strains using plasmid DNA ( Figure 6f). As expected, rad51D cells were very defective with HR rates of only~3% of H3-WT cells. H3-G34R and set2D strains were also defective (~35% efficiency) and the double H3-G34R set2D mutant showed an even stronger defect, with a 90% drop in HR efficiency. The reduction in HR in set2D was not due to the genetic background, as we obtained similar results when set2D was assessed in a 3xH3/H4 genetic background ( Figure 6figure supplement 1b). Thus both H3-G34R and set2D reduce HR efficiency but via distinct mechanisms, as additive defects are seen in the set2D H3-G34R double mutant. We note that our finding of an HR defect in set2D differs from the previously published result of enhanced HR and defective NHEJ in 3xH3/H4 set2D cells as described in Pai et al., 2014. Their study utilized an engineered non-essential minichromosome with an HO break site. The authors monitored four different markers for loss, and interpret combinations of marker loss as NHEJ-repair, gene conversion, loss of heterozygosity and loss of the chromosome. However, due to the nature of the assay, there are many possible causes of apparent marker loss which may cloud interpretation. Our simple and direct approach to measure HR relies on a site specific gain of marker function at a defined locus. We note that we have just assessed HR at one site, and it is quite feasible that distinct outcomes may be obtained by assessing HR at different genomic regions. We sought to determine if the difference in interpretation of Set2 function was indeed caused by the different assays or by strain differences between the set2D backgrounds. We performed the transformation-based HR assay in the set2D strains used in Pai et al., 2014, and in NHEJ defective ku70D cells (Figure 6-figure supplement  1c), and found that all set2D strains tested showed a defect in HR, whereas HR efficiency was  Figure 6. H3-G34R cells show defects in homologous recombination pathways. (a) Scheme depicting DNA damage repair pathways. (b) Cells of indicated genotypes were cultured at 25˚C in the presence or absence of 11 mM HU for 6.5 hr and then washed and either kept at 25˚C (permissive) or shifted to 36˚C (restrictive) for 4 hr. Cells were then plated at 25˚C to determine cell viability. Numbers were normalized to untreated 25˚C control values for each genotype, and results were averaged from 2 independent biological replicates. (c-e) Serial 5-fold dilution assays of H3-WT or H3-G34R cells and double mutants with (c) PRR pathway mutants rad18D or ubc13D, (d) NHEJ mutant ku70D or (e) HR mutant rad51D plated on media ± MMS treatment. (f) Homologous recombination assay based on correction of leu1-32 mutation by HR. Cells of indicated genotypes were transformed with a leu1 gene fragment to measure HR, or plasmid to measure transformation efficiency. Relative HR efficiency is shown as 100% for H3-WT, and results are averaged from 6 independent experiments with error bars representing SEM. Black asterisks reflect significant differences with H3-WT cells (p<0.05), and red asterisks, significant differences from both H3-G34R and set2D (p<0.01). See also Figure 6-figure supplement 1. (g) Percentage of cells that form foci of either RPA (Rad11-GFP), Rad52-GFP, or Rad51 before or after treatment with 0.05% MMS for 4 hr. Immunostaining with GFP antibodies was used for monitoring RPA and Rad52 while anti-Rad51 antibodies were used to detect Rad51. Cells were counted from 2 independent experiments with errors representing SEM. See also images in Figure 6-figure supplements 1 and 2. Figure 6 continued on next page strongly enhanced in ku70D, as HR is now the sole method of DNA break repair available. We conclude that the difference in interpretation of whether set2D cells are deficient in HR lies with the use of distinct assays.
In fission yeast, NHEJ is restricted to G1, while HR operates in S and G2 phases of the cell cycle (Ferreira and Cooper, 2004). One possible cause of the decreased HR in H3-G34R cells and the dependence of H3-G34R cells on NHEJ for viability (Figure 6d) is an alteration in the cell cycle, with an extension of G1. We measured the generation time during exponential growth and found that H3-G34R had an extended doubling time of 201 min (±4 min) compared to 180 min (±10 min) for H3-WT, and 194 min (±8 min) for set2D in rich media at 32˚C. To determine if the growth delay of H3-G34R mutants was linked to a particular cell cycle stage, we monitored cell cycle kinetics using FACS sorting of cells synchronized in G1 or in metaphase and then released back into the cycle at 25˚C. As shown in Figure 6-figure supplement 2, cdc10-M17 (G1) arrested and released cells begin DNA synthesis at 40 min after release. H3-WT cells have completed DNA synthesis by 80-100 min, whereas S phase is prolonged in H3-G34R cells and not completed until 140-160 min after release. It is difficult to draw conclusions from the nda3-km311 arrest and release experiment since fission yeast do not undergo cytokinesis until G1, which complicates analysis of G1 cells. From these data we can surmise that H3-G34R cells spend longer in S phase when released from a G1/S block.
Next we asked whether there were defects in localization of HR proteins. We evaluated foci formation of RPA, Rad51, and Rad52 (Figure 6g and Figure 6-figure supplement 3a). Consistent with enhanced replication stress in H3-G34R cells, H3-G34R cells showed increased numbers of RPA and Rad52 foci (Bass et al., 2012). As expected, all cells showed increased foci for all three proteins following MMS treatment. Surprisingly however, fewer MMS-treated H3-G34R cells showed Rad51 and Rad52 foci compared with H3-WT and set2D. The reduction in foci was not linked to a change in protein expression as Rad51 and Rad52 protein levels were similar in MMS treated H3-WT and H3-G34R cells (Figure 6-figure supplement 3b). Additionally, Rad52 foci co-localized with RPA, suggesting that Rad52 foci occur at sites of DNA damage ( Figure 6-figure supplement 3c). In summary, H3-G34R cells exhibit elevated marks of replicative stress, and on MMS-damage, have diminished foci formation for Rad51 and Rad52, which may be linked to the HR defect seen in H3-G34R cells.
We questioned whether the chromosome segregation defects in H3-G34R may be related to delayed resolution of replicative intermediates. To address this, we monitored RPA localization in anaphase cells following cdc10 block and release, and found that H3-G34R cells exhibit anaphase bridges, with RPA marking ultrafine bridges between DAPI staining masses ( Figure 6-figure supplement 3d). We are unable to determine the frequency of these bridges in anaphase cells since attempts to co-stain the spindle hampered visualization of bridges, but they were easy to identify in synchronized populations suggesting that they are frequent events. This supports that chromosome segregation defects in H3-G34R cells are linked to the delayed resolution of replication intermediates.
H3-G34R mutants exhibit delayed recovery from replicative stress HR-directed repair is often utilized to restart replication forks following stress. Because we observed a defect in HR, and a defect in Rad51 and Rad52 foci formation in MMS-treated H3-G34R cells, we speculated whether this was due to a delay in Rad51/Rad52 filament formation, and thus a defect in replication restart. To address whether the kinetics of recovery were affected, cells were arrested with HU, washed, and then counted at intervals following release (Figure 7a). Checkpoint-defective cds1D cells do not recover from the arrest. While H3-G34R do delay proliferation for longer than H3-WT (3 hr rather than 2 hr), the proliferation rates at later time points appear similar. To directly monitor the rate of DNA replication, we arrested cells with HU and on release, monitored de novo DNA synthesis by EdU incorporation, gating on cells that had fully incorporated EdU (Figure 7b). Interestingly, de novo DNA synthesis is initially delayed in H3-G34R cells but at later time points after HU release, the rates of DNA synthesis appear similar in H3-WT and H3-G34R cells.
To further evaluate the kinetics of DNA damage response, we monitored gH2A phosphorylation ( Figure 7c) and counted cells that bear Rad52 foci (Figure 7d) during MMS treatment and recovery. MMS treatment led to a rapid induction of gH2A signal in H3-WT cells, such that at 45 min treatment, maximal signal was achieved which was retained during treatment. Surprisingly, gH2A signal accumulated more slowly in H3-G34R cells, and peaked at 45 min into the recovery period, at higher levels than seen in H3-WT cells. On recovery from MMS, gH2A remained intensely phosphorylated at later time points in H3-G34R cells compared with H3-WT cells (see Figure 7c).
Similar to gH2A, Rad52 foci accumulated faster in MMS treated H3-WT cells than H3-G34R, even though more Rad52 foci were present in untreated H3-G34R cells than in H3-WT (Figure 7d). During recovery, H3-WT cells with Rad52 foci dropped to~50% by 180 mins while Rad52 foci continued to accumulate in H3-G34R, peaking at 90 min following release, before dropping. These data suggest that the kinetics of DNA damage response are different between H3-WT and H3-G34R cells, with H3-G34R mutants showing a slower induction of damage response, as well as accumulation of recombination proteins at later timepoints following damage, which may contribute to the HR-mediated repair defect in H3-G34R mutants.

Discussion
In this study, we generated single copy histone mutants in fission yeast with the aim of studying the behavior of the H3-G34R mutant. This system allows a detailed examination of the possible consequences of this histone mutant, divorced from the complexity of mammalian histone regulation.
Using this system, we show that H3-G34R mutants have focal histone modification changes, with reduced tri-methylation and acetylation, but retention of di-methylation on the nearby K36 residue on histone H3. These results are consistent with some perturbation of Set2 function and we demonstrate that this defect cannot be overridden by overexpression of Set2 as, even under these conditions, H3-G34R cells fail to accumulate normal levels of H3K36me3.
Furthermore, the transcriptional consequence of H3-G34R mutation is limited, with fewer genes affected than in set2D. Indeed for genes whose transcription is altered in both set2D and H3-G34R, only some are coordinately regulated. Notably, these include antisense genes, which suggests that trimethylation of H3K36 is important for suppression of antisense gene transcription. Interestingly, genes showing discordant regulation map to specific chromosomal loci-the ST regions. Transcriptional upregulation in ST domains in set2D is likely a direct consequence of loss of Set2 function on H3K36, since the histone H3 K36A mutant shows similar upregulation of expression of ST domains (Matsuda et al., 2015). Of interest, ST regions have recently been shown to be the most highly condensed regions of the genome, forming 'knobs', with condensation linked to Set2 function (Matsuda et al., 2015), but the biological implications of knob formation are not yet known.
Perhaps the most consequential findings from this work are the effects of H3-G34R mutation on genome stability. We see evidence for enhanced chromosome loss and for defects in chromosome segregation, with accumulation of cells that have trailing or 'stretched' chromosomes. We find that RPA coats ultrafine anaphase bridges in mitotic H3-G34R cells, marking DNA that links chromosome masses, which is indicative of unresolved replication problems. These data are consistent with our finding that H3-G34R cells are sensitive to replication stress, and that HR pathways are defective in these cells. These phenotypes are unlikely to be due to transcriptional deregulation in H3-G34R cells since transcription of DNA damage genes, under non-perturbed conditions, is unaltered. Instead, as summarized in Figure 8, these findings may derive from the altered kinetics of signal response to damage in H3-G34R cells, with slowed accumulation of gH2A and delayed formation of Rad52 foci, features which may also contribute to the delayed replication of H3-G34R cells during recovery from Reduced K36me3 Reduced K36ac Figure 8. Model illustrating how HR defects in H3-G34R mutants may compromise genomic integrity. Model illustrates the key defects in H3-G34R cells and how these may contribute to genomic instability. Replication forks provide an endogenous source of DNA damage, as they are fragile structures that are prone to stalling and collapsing. (left) To cope with replication stress, cells activate the checkpoint signaling pathway by sensing ssDNA via Rad3. Rad3 interacts with ssDNA through RPA and phosphorylates H2A, which helps to transduce the checkpoint response to downstream kinases. For lesions that require HR-directed repair, nucleosome removal and subsequent 5' to 3' DNA resection exposes ssDNA which is initially coated with RPA, and then replaced by Rad51, facilitated by the Rad52 protein. The search for a homologous template can begin once sufficient Rad51-ssDNA filament has been formed and strand invasion of the Rad51-ssDNA filament generates Holiday junctions which need to be resolved prior to cell division. (right) In H3-G34R cells, HR is defective. Checkpoint signaling is functional, but appears delayed with slower accumulation of gH2A possibly due to impaired nucleosome removal and delayed resection, which would limit the amount of Rad3 on RPA/DNA. Retention of gH2A indicates a defect in nucleosome Figure 8 continued on next page replication stress. H3-G34R mutation may thus impair HR-mediated repair of DNA damage induced DSBs.
How might H3-G34R mutation elicit these effects? One possibility may be via modulation of H3K36 methylation signal. However this seems unlikely given that cells that lack Set2 show no defects in chromosome segregation, and the HR defects that we uncovered in set2D cells are not epistatic to HR defects in H3-G34R cells. Another alternative is that the altered kinetics for gH2A phosphorylation and Rad52 foci formation during MMS treatment and recovery are indicative of chromatin defects in the H3-G34R cells. H2A phosphorylation occurs within the C-terminal tail of H2A, which projects out from the DNA and occupies a territory close to the N-terminal tail of H3 within the nucleosome structure (Du and Briggs, 2010). Incorporation of an additional basic charge in the H3 tail close to where the H3 tail exits the H3 core domain may have conformational consequences and the H3-G34R tail may modulate kinase activity on the C-terminal H2A tail ( Figure 8). However, checkpoint signaling is clearly functional in H3-G34R cells, even if it is delayed ( Figure 5). Once damage has occurred, H3-G34R cells retain gH2A for longer than H3-WT ( Figure 7c). It is thought that, in order for gH2A to be dephosphorylated, nucleosomes must first be removed from chromatin where they can be acted upon by phosphatases in free nuclear pools (Keogh et al., 2006;Jablonowski et al., 2015). Retention of gH2A may stem from a defect in efficient chromatin remodeling, thereby preventing timely repair. In support of this, an H2A mutant that mimics gH2A shows decreased interaction with the chromatin remodeler Fun30 (Eapen et al., 2012).
H3-G34R cells show defects in HR-directed repair. Nucleosome removal from damaged chromatin facilitates resection and subsequent coating by RPA. After sufficient resection of DNA, Rad52 helps to facilitate the exchange from RPA-coated DNA to a Rad51-coated filament, to allow for homology searching and HR repair. The presence of a gH2A mimetic can inhibit the rate of resection (Eapen et al., 2012). Here, we show that H3-G34R mutants are specifically deficient for HR ( Figure 6e) with prolonged retention of gH2A and delayed Rad52 foci formation (Figure 7c,d). It is possible that H3-G34R cells are defective in timely nucleosome removal and resection (see Figure 8). Following this model, delayed HR-mediated repair in H3-G34R cells, or delayed resolution of replicative intermediates in undamaged H3-G34R cells, may allow accumulation of unresolved Holliday junctions prior to mitosis, and contribute to the chromosome segregation defects shown in this study ( Figure 3 and Figure 6-figure supplement 3d). We note that consistent with our findings, clearance of gH2A.X foci in human cells that lack H3K36me3 due to deletion of SETD2 is delayed (Hacker et al., 2016), and such cells are defective for HR and exhibit genomic instability (Pfister et al., 2014).
Clearly an important consideration is that our data was obtained in an organism where the whole population of histone H3 is mutant. In pHGG, only one allele of four that encodes H3.3 protein is affected in cells with 13 other genes that encode wild type H3.1 and H3.2 proteins. Unlike the H3K27M mutation that dominantly blocks PRC2 activity, G34R mutation does not dominantly affect K36 methylation (Lewis et al., 2013;Zhang et al., 2017). Instead, we suggest that the H3.3 G34R mutant may be greatly enriched at specific chromosomal loci. H3.3 is deposited into regions of constitutive heterochromatin including repetitive telomeric and pericentromeric domains, and at regions of high transcriptional activity through the activity of specialized chaperones (reviewed in Kallappagoudar et al., 2015). These sites of H3.3 deposition are already subject to replicative stress. As cells traverse S phase, the replicative delay incurred by the presence of G34R mutant H3.3 at these sites and impaired resolution of replication intermediates could greatly enhance genomic instability and offer a potent source of genomic perturbations to facilitate tumor development (Macheret and Halazonetis, 2015). removal and resection which in turn, could account for the delayed Rad52 foci formation in H3-G34R cells and likely delayed Holiday junction resolution following repair. H3-G34R cells show elevated chromosome segregation defects without defects in heterochromatin or kinetochore chromatin, suggestive that a delayed resolution of HR products contributes to genomic instability. DOI: 10.7554/eLife.27406.016 Materials and methods

Strain generation
Histone mutant strains were generated as described (Mellone et al., 2003). Gene mutagenesis used standard PCR-based procedures, and strains are listed in Supplementary file 4. All crosses used random spore analysis with nutritional/ drug selection and PCR verification (including verification of loss of additional H3/H4 genes) and sequencing of H3.2 allele. Two independent clones for each genotype were used in nearly all experiments, and all experiments were performed at least twice.

Yeast growth media
Fission yeast were maintained on rich (YES), or pombe minimal with glutamate (PMG) media with appropriate supplements (Moreno et al., 1991). PMG is Edinburgh minimal medium with glutamate.
JPO-4159: TTTGTTAAATCATATGTCGACATGCAGACGGCATCATCTCT, JPO-4160: ACCCGGGGATCCTCTAGAACCTACAGGAAAGAGTTACTCAAGAATAAGAAT Histone purification for mass spectrometric studies ( Figure 1f) Histones were purified following a previously described purification protocol with some modification (Sinha et al., 2010). A 150 mL culture was inoculated to a density of 1.4 Â 10 6 cells/mL in 4X YES media from a starter culture grown overnight in 4X YES at 25˚C. The cells were grown to a density of 3.6 Â 10 7 cells/mL and harvested by spinning at 3000 RPM for five minutes in a benchtop centrifuge. Cells were washed with H 2 O containing 10 mM sodium butyrate. The cells were then washed with NIB buffer (250 mM sucrose, 20 mM HEPES pH 7.5, 60 mM KCl, 15 mM NaCl, 5 mM MgCl2, 1 mM CaCl2, 0.8% Triton X 100, 0.5 mM spermine, 2.5 mM spermidine, 10 mM sodium butyrate, 1 mM PMSF, and Sigma yeast protease inhibitor). The pellet was frozen on dry ice and stored at À80˚C. For lysis, the pellet was resuspended in 2 mL of NIB and transferred to a bead beater tube along with chilled acid washed glass beads. The sample and beads were frozen on dry ice and bead beaten for a total of 10 min at max power. The sample was collected by 'piggy backing' into a 50 mL Oak Ridge tube at 3000 RPM at 4˚C in a benchtop centrifuge. The samples were then pelleted by centrifuging at 20,000 RPM for 10 min at 4˚C in a Beckman Avanti centrifuge J-30I centrifuge using a JA25.50 rotor. The supernatant was discarded and the pellet was washed in 15 mL of NIB. The pellet was then resuspended in 10 mL of 0.4N H 2 SO 4 and sonicated for one minute at max power before incubation for two hours on a rotating wheel at 4˚C. The sample was pelleted by centrifuging at 20,000 RPM at 4˚C for 10 min and the supernatant was transferred to a new tube along with 5 mL of 5% buffer G (5% guanidine HCl and 100 mM potassium phosphate buffer pH 6.8) where the pH was adjusted to 6.8 using 5N KOH. 0.5 mL of Bio-Rex pre-equilibrated in 5% buffer G was added to the sample and incubated at room temperature with rotation overnight. The resin was then washed 2X with 20 mL of 5% buffer G and incubated with 3 mL of 40% buffer G (40% guanidine HCl and 100 mM potassium phosphate buffer pH 6.8) for one hour at room temperature to elute the bound protein. Buffer exchange and concentration was performed against 5% acetonitrile with 0.1% TFA to a final volume of 150 ml and the sample was stored at À80˚C. Protein concentration was measured and 10 mg of sample was run on a gel for quality control.

UPLCÀMS/MS analysis
Histone samples were TCA precipitated, acetone washed, and prepared for mass spectrometry analysis as previously described (Kuo and Andrews, 2013). A Waters (Milford, MA) Acquity H-class UPLC system coupled to a Thermo (Waltham, MA) TSQ Quantum Access triple-quadrupole (QqQ) mass spectrometer was used to quantify modified histones. Selected reaction monitoring was used to monitor the elution of the acetylated and propionylated tryptic peptides. Transitions were created to study acetylation to pombe H3 wild-type and mutants as well as the H4 tails. The detailed transitions for peptides of H3 that vary in sequence from xenopus are reported in Supplementary file 2 and the peptides used are listed in Supplementary file 1; the transitions for the xenopus peptides have been previously reported (Kuo et al., 2014).

QqQ MS data analysis
Each acetylated and/or propionylated peak was identified by retention time and specific transitions. The resulting peak integration was conducted using Xcalibur software (version 2.1, Thermo). The fraction of a specific peptide (Fp) is calculated as F p =I s /( P I p ), where I s is the intensity of a specific peptide state and I p is the intensity of any state of that peptide.
2. Minichromosome loss. (Figure 3a). The minichromosome (Ch16) (Niwa et al., 1989) bears an ade6-216 allele which can complement the ade6-210 allele present within the strain background. Loss of Ch16 causes loss of complementation of function of ade6 + and accumulation of red pigment when cells are grown on limiting adenine media. Strains containing Ch16 were grown in PMG -Leu (to maintain Ch16) at 32˚C to a density of 5 Â 10 6 cells/mL. Cells were diluted in PMG (no additives) to a final concentration of 5À10 Â 10 3 cells/mL. Cells were plated onto PMG agar supplemented with amino acids and nucleobases with limiting (10% normal concentration) adenine, incubated at 25˚C for 5 days and then transferred to 4˚C to let red color develop. Ch16 loss frequencies were calculated by counting half sectored colonies (and those with >50% red but <100% red) and dividing by the total number of white, white sectored and less than half red colonies, excluding red colonies from the analysis as they have lost the minichromosome before plating. Results represent data from multiple (2-4) independent cultures of cells, and 5À10 Â 10 3 colonies were scored for each strain.
3. Lagging chromosome analysis. (Figure 3b). Chromosome missegregation frequencies were obtained as previously described (Mellone et al., 2003). 220 late anaphase cells were scored for presence of lagging chromosomes for each genotype, using two independent cultures of strains.
4. Mitotic arrest and release (Figure 3c). nda3-km311 mutant (Hiraoka et al., 1984) in H3-WT and H3-G34R backgrounds was used to synchronize cells in mitotic phase. Cells were grown in YES to 5 Â 10 6 cells/ml at 32˚C, then were filtered to rapidly collect cells and transferred to media at 18˚C for 8 hr. Cells were released to warm media at 32˚C and aliquots fixed at 15 min intervals in 3.7% paraformaldehyde for tubulin and DAPI staining as described in lagging chromosome analysis method. 2. Native WCE and chromatin enrichment (NIB buffer). Cell lysates were fractionated using a method modified from (Ekwall et al., 1997;Sinha et al., 2010). Briefly, cell pellets from 50 ml cultures in YES were washed with NIB [0.25 M sucrose, 60 mM KCl, 15 mM NaCl, 5 mM MgCl 2 , 1 mM CaCl 2 , 2.5 mM spermidine, 20 mM HEPES, 10 mM N-butyric acid, 0.8% Triton X-100 (w/v) with 1:1000 protease inhibitors (Sigma (P8215)) and 1 mM phenylmethylsulfonyl fluoride (PMSF)] and resuspended in 1 ml NIB. Bead beating was performed for 3 min in presence of 500 ul glass beads (Sigma, G8772). After removal of beads, 500 ul of lysate was taken as 'total' extract. The rest was centrifuged at 13,000 rpm for 15 min at 4˚C. Supernatant was saved as 'soluble' fraction. Pellet was washed three times in 1 ml NIB buffer, and resuspended in 500 ul NIB buffer as 'chromatin enriched' fraction. Samples were heated with sample buffer prior to PAGE and western. This extract type was used for Figure 1-figure supplement 1c. 3. Denatured extracts in Urea (Figure 5c) Previously published protocols were used for Chk1-HA western blot (O'Connell et al., 1997;Bass et al., 2012). 25 ml cultures of strains were treated ±0.05% MMS for 1 hr. MMS was inactivated with freshly made 5% sodium thiosulphate. Cells were disrupted with glass beads (Sigma, G8772) using a bead beater and extracted into urea lysis buffer (8 M urea, 100 mM NaH 2 PO 4 , 10 mM Tris pH 8, 60 mM b-glycerophosphate, 1% Triton X-100 with protease inhibitor cocktail (Sigma P8215) and 1 mM PMSF). The extract was cleared by centrifugation at 13,000 rpm for 10 min, and the supernatant was boiled in SDS sample buffer.
Western blot quantification used Image Studio Lite from LiCor. Equal sized squares were drawn using the software, which automatically assigns signal intensity for each band after subtracting background signal. The signal intensities were used to quantify the images, relative to signal intensity for total H3 bands.
5. Peptide sequences used for assessment of K36 methyl Abs (Figure 1, and Figure 1-figure supplement 1) are listed in Supplementary file 1. Dot blots used a 10-fold serial dilution series with 2 ul spots of 1 mM, 0.1 mM, 0.01 mM and 0.001 mM peptides spotted on activated PVDF 0.2 micron membrane. Spots were air dried for 1.5 hr, blocked in 5% BSA in PBST at RT for 1 hr, incubated with primary Ab for 1 hr at RT, washed with PBST, incubated with HRP-conjugated anti-rabbit secondary Ab for 30 min, washed with PBST and then developed with enhanced chemiluminescence and images captured by LiCor imaging. Anti-H3 K36 methylation Abs are listed above and peptide loading was verified by ponceau staining.

Chromatin immunoprecipitation
ChIP assays (Figure 1d, Figure 2-figure supplement 1b,c) were performed as described previously (Alper et al., 2013). Antibodies for H3K9me2: (Abcam ab1220) and Swi6: (Thermo Scientific PA 1-497). Real time PCR was used to quantify enrichment at heterochromatic loci relative to adh1 + . Set2-3xFLAG ChIP used Anti-FLAG: (Sigma F1804) and monitored Set2 association with act1 + or clr4 + loci as a ratio of signal from input DNA. For Cnp1 ChIPs (Figure 2-figure supplement 1d), 10 ul anti-Cnp1 antiserum (Kniola et al., 2001) was used per ChIP. Crosslinks were reversed by heating at 100˚C for 12 min in the presence of 100 ul Chelex-100 resin (10% slurry in dH 2 O; BioRad), followed by treatment with Proteinase K (2.5 mg/ml) 30 min at 55˚C with shaking and heat-inactivation of Proteinase K (10 min, 100˚C). DNA was recovered and used in qPCR analysis. Primer sequences used for q-PCR are listed in Supplementary file 5.

Transcript analysis
RNA seq studies ( Figure 2): Hot phenol extraction was used to prepare the RNA (Leeds et al., 1991). Cultures were grown overnight in YES at 25˚C to a density of 2.5 Â 10 6 cells/ml in a 25 mL culture. The cells were pelleted by centrifugation and washed in DEPC H 2 O. The pellet was resuspended in 750 ul TES Buffer [50 mM Tris-HCl pH 7.5, 10 mM EDTA, 100 mM NaCl, 0.5% SDS made in DEPC H 2 O] along with an equal volume of 5:1 phenol:chloroform pH 4.7 and incubated at 65˚C for one hour with vortexing for 10 s every 10 min. The samples were then cooled on ice and centrifuged for five minutes at 13,000 RPM. The aqueous phase was transferred to a 2 mL phase lock tube and an additional phenol:chloroform extraction was performed. After centrifugation, the aqueous phase was transferred to a new tube and an equal volume of chloroform was added. To precipitate the RNA, the aqueous phase was transferred to a 2 mL microcentrifuge tube and three volumes of ice cold ethanol and 3M NaOAc pH 5.2 were added and the samples were kept at À20˚C to precipitate the RNA. The next day, samples were centrifuged at 14,000 RPM and 4˚C for 15 min. The pellet was washed with ice cold 70% ethanol and air dried for 30 min. A Turbo DNAse (Ambion) reaction was set up with 100 ug of RNA in a 150 ul reaction containing 5 ul of Turbo DNAse. The reaction was incubated at 37˚C for 30 min and another 5 ul of Turbo DNAse was added and incubated for an additional 30 min prior to the removal of the DNAse using 50 ul of inactivation beads. An RNeasy Mini kit (Sigma) was used to further clean up and concentrate the RNA which was eluted in a final volume of 30 ul of DEPC H 2 O.