CAG/CTG trinuncleotide repeats are fragile sequences that when expanded form DNA secondary structures and cause human disease. We evaluated CAG/CTG repeat stability and repair outcomes in histone H2 mutants in S. cerevisiae. Although the two copies of H2A are nearly identical in amino acid sequence, CAG repeat stability depends on H2A copy 1 (H2A.1) but not copy 2 (H2A.2). H2A.1 promotes high-fidelity homologous recombination, sister chromatid recombination (SCR), and break-induced replication whereas H2A.2 does not share these functions. Both decreased SCR and the increase in CAG expansions were due to the unique Thr126 residue in H2A.1 and hta1Δ or hta1-T126A mutants were epistatic to deletion of the Polδ subunit Pol32, suggesting a role for H2A.1 in D-loop extension. We conclude that H2A.1 plays a greater repair-specific role compared to H2A.2 and may be a first step towards evolution of a repair-specific function for H2AX compared to H2A in mammalian cells.
Integral to the eukaryotic DNA damage response is chromatin structure modifications surrounding the break site (reviewed in House et al., 2014a; Seeber et al., 2013; Price and D'Andrea, 2013). In response to DNA double strand breaks (DSBs) and stalled or collapsed replication forks, the SQEL motif in the H2A C-terminal tail is phosphorylated at Ser129 (H2AX-Ser139 in mammals) by the Phosphoinositide 3-Kinase-Related Kinases (PIKKs), Mec1 and Tel1 (ATR and ATM in mammals) (Downs et al., 2000; Lisby et al., 2004). This modification, termed γH2A (γH2AX in mammals), marks the site of damage and is propagated along the chromatin, detectable up to 50 kb from the break site in yeast (Shroff et al., 2004) and megabases in mammalian cells (Rogakou et al., 1999).
γH2A/γH2AX occurs rapidly upon induction of a DSB, detectable within minutes after damage (Shroff et al., 2004; Rogakou et al., 1999; Paull et al., 2000), and is required to initiate the cascade of histone modifications, chromatin remodeling, and repair factor recruitment and retention necessary for repair. However, other modifiable residues in the H2A tails can also modulate repair factor binding: H2A/H2AX K5ac, K15ub/ac, K36ac, K119ub, and Y142ph all contribute to DNA repair (reviewed in Hunt et al., 2013; Jacquet et al., 2016).
Whereas in humans over a dozen H2A variants have been identified (Albig et al., 1999; Bönisch and Hake, 2012), S. cerevisiae contains just three variants of H2A, encoded by HTA1, HTA2, and HTZ1. HTA1 and HTA2 encode canonical H2A and the two copies are nearly identical in amino acid sequence except for a direct alanine-threonine switch at positions 125/126 in the C-terminal tail (Figure 1B); the underlying DNA sequence is 94% similar. H2A-T126 is phosphorylatable in vivo, even in the absence of DNA damage (Wyatt et al., 2003; Moore et al., 2007). The third H2A variant, H2A.Z, has only 56% amino acid sequence homology to canonical H2A.
H2A modification is a major contributor to DNA repair and may be particularly important in promoting efficient repair at unstable genomic elements. CAG/CTG trinucleotide repeats are in this category, as they can form abnormal secondary structures, such as hairpins and slip-stranded DNA (reviewed in McMurray, 1999; Usdin et al., 2015; Schmidt and Pearson, 2016), and break at a higher frequency than non-repetitive DNA (Freudenreich et al., 1998; Callahan et al., 2003; Nasar et al., 2000). Repair or replication errors within the CAG/CTG repeat can lead to instability, or a change in repeat units. Once expanded (addition of repeat units), the repeat tract is increasingly unstable and prone to further expansion in a length-dependent manner (reviewed in Usdin et al., 2015; Kim and Mirkin, 2013). CAG/CTG repeats are found throughout the human genome but repeat expansion beyond a threshold length of approximately 35 repeats can lead to human disease, including Huntington’s disease, myotonic muscular dystrophy, and several spinocerebellar ataxias (Usdin et al., 2015; Mirkin, 2007). The CAG/CTG repeat is a strong nucleosome-positioning element, shown in vitro by nucleosome assembly assays and visualized by electron microscopy (Godde and Wolffe, 1996; Wang et al., 1994). The intrinsic nucleosome-positioning characteristic of the CAG repeat makes this an interesting and sensitive sequence at which to study the chromatin environment during DNA repair. Further, the unstable nature of the repeat allows us to experimentally test the importance of chromatin and repair factors in promoting high-fidelity repair, since repair errors (errors in synthesis, alignment, processing, etc) can lead to repeat tract length changes.
Secondary structures that occur at CAG/CTG repeats can interfere with DNA transactions, causing stalled or collapsed replication forks, gaps, nicks, and DSBs (Usdin et al., 2015). Repair can proceed via homologous recombination (HR), but this repair itself can be a source of mutagenesis if it does not proceed with high fidelity (reviewed in Polleys et al., 2017). Several steps during HR presumably require nucleosome repositioning or eviction, including resection, strand invasion, copying the template and D-loop extension, and resetting the chromatin structure after repair. Efficient completion of each stage of HR is expected to be important to prevent errors that lead to CAG repeat expansions (Polleys et al., 2017).
We previously described a role for histone H4 acetylation in promoting high-fidelity HR during post-replication repair at CAG repeats (House et al., 2014b). Here, we explore the role of histone H2A in CAG repeat maintenance. In a primary genetic screen for CAG repeat fragility and a secondary screen of CAG repeat instability, deletion of histone H2A.1 increased CAG repeat fragility and expansion frequency. However, deletion of the second copy of this protein, H2A.2, had no effect on repeat fragility or instability. Since histone H2A could be participating in one or more pathways that contribute to repeat stability, several hypotheses have been explored to explain this discrepancy. We found that H2A.1 threonine 126 (T126) is required to prevent CAG expansions and that expansions that arise in the absence of phosphorylatable T126 are dependent on Rad51, Rad52, Rad57, and the Polδ subunit, Pol32. In addition, we show that H2A.1 and H2A.1-T126 are required for efficient SCR at non-repetitive DNA sequences, and are working in the same pathway as the Polδ subunit Pol32. Finally, we show that H2A.1 is specifically important in mediating repair via BIR. Together, these results demonstrate that H2A.1 plays an important role in promoting efficiency and fidelity of recombination during repair. This role is distinct from H2A.2, and our results implicate the T126 residue as important for this distinction.
To identify important factors for maintaining expanded CAG/CTG trinucleotide repeats, a screen was performed for genes that protect against repeat fragility using a yeast artificial chromosome (YAC) end loss assay in the Matα haploid deletion set (screen originally described in Gellon et al., 2011; assay illustrated in Figure 1—figure supplement 1A and reviewed in Polleys and Freudenreich, 2018; Polleys and Freudenreich, 2020). Initial semi-quantitative and quantitative assays showed a 2-fold increase in the rate of 5-FOA-resistance (FOAR) in the hta1Δ mutant compared to the wild-type for a strain containing a YAC with a (CAG)85 repeat tract, whereas the hta2Δ mutant did not deviate from wild-type. Upon further analysis using multiple independent hta1Δ transformants, we observed a wide range of repeat fragility rates and thus were unable to statistically verify the increase over wild-type (Figure 1C; Supplementary file 1). The hta1Δ mutant is mildly sensitive to phleomycin and not sensitive to the other DNA damaging agents tested (MMS, CPT, HU) (Figure 1—figure supplement 2). We conclude that there is likely a mild defect in repair at the CAG repeat in the absence of H2A copy 1.
We next evaluated the contribution of all the histone H2 variants to CAG/CTG repeat stability. Histone genes were deleted and (CAG)85 repeat tract length changes were monitored by PCR analysis (Callahan et al., 2003; House et al., 2014b) (Figure 1D,E,F, Supplementary file 2). Deletion of the genes encoding the two copies of H2A differentially affect CAG repeat stability: expansion frequency is significantly increased in the hta1Δ mutant (7-fold increase over wild-type, p=2.7×10−5) while expansion frequency in the hta2Δ mutant is not significantly changed from wild-type (1.3-fold increase over wild-type, p=0.71) (Figure 1D). Contractions were increased 2.6-fold in the hta1Δ but not the hta2Δ mutant (Figure 1E). Thus, H2A.1 is required to suppress CAG instability while H2A.2 is not.
The H2A.Z variant is encoded by the gene HTZ1 in yeast. Since H2A.Z is deposited at a DSB during repair (Kalocsay et al., 2009), we tested whether it is required to maintain CAG repeat stability. Although the htz1Δ mutant is sensitive to DNA damaging agents (Morillo-Huesca et al., 2010; Papamichos-Chronakis et al., 2011), deleting the HTZ1 gene did not affect repeat expansion frequency, though contractions were significantly increased in the htz1Δ mutant (2.1-fold over WT, p=3.1×10−6) (Figure 1E) (House et al., 2014b). Deleting either HTB gene had no significant effect on repeat expansion frequency (Figure 1D), but contractions were somewhat elevated in the htb1Δ mutant (2-fold over WT, p=0.01)(Figure 1E). Therefore, H2A.1, H2A.Z, and H2B.1 are required to prevent CAG contractions. However, of the histone H2 proteins, only H2A.1 plays a significant role in preventing CAG repeat expansions, suggesting some specialized role for H2A.1 that we have explored further.
Histone H2A is encoded by two nearly identical gene copies in S. cerevisiae, each paired with a copy of H2B and differentially regulated (Norris and Osley, 1987) (Figure 1A). The H2A.1 and H2A.2 protein sequences are identical except for a direct threonine/alanine (T/A) switch in the C-terminal tail (Figure 1B). While transcription from the HTA1-HTB1 gene pair can be upregulated in the absence of HTA2, HTA2-HTB2 is transcribed at a constant rate (Moran et al., 1990). As a result, the H2A pools will be normal in an hta2Δ mutant, whereas an hta1Δ mutant may have a global decrease in H2A. A second pathway of gene dosage compensation exists in which the HTA2-HTB2 gene pair amplifies to form a minichromosome (that also contains the HHT1-HHF1 (H3-H4) gene pair) in the absence of HTA1-HTB1 (35). To distinguish if repeat stability is mediated by the H2A protein sequence or histone levels, the HTA2 sequence was placed under the control of the HTA1 promoter, replacing the HTA1 gene (hta1Δ::HTA2). In this strain, the H2A.2 protein will be expressed at the same level and timing as H2A.1 in a wild-type cell but the H2A.1 protein will not be present. Equal expression of H2A proteins was confirmed by Western blot (Figure 1—figure supplement 1B). We tested strains for gene amplification of HTA1 or HTA2 by qPCR and no instances of gene amplification were detected at the genomic level (Supplementary file 8). If H2A expression level or timing is the major contributor to repair of the CAG repeat, repeat maintenance will be at wild-type levels when HTA2 is expressed from the HTA1 promoter. However, we observed that CAG repeat expansions remained significantly increased from the wild-type in the hta1Δ::HTA2 strain (p=4.2×10−4 to WT, Figure 1D), though a partial suppression of contractions was observed (Figure 1E)(Supplementary file 2). Therefore, while CAG contractions appear to be sensitive to overall histone levels, H2A.2 cannot compensate for H2A.1 in preventing repeat expansions, even when expressed at H2A.1 levels under control of the HTA1 promoter. We conclude that the sequence of H2A.1, not histone levels or subtleties in expression timing, is required to prevent CAG repeat expansions.
Since H2A.2 is not upregulated in the absence of H2A.1, we hypothesized that an hta1∆ mutant may cause local histone depletion or disruption of the chromatin structure at the CAG repeat tract, leading to instability. Although overall bulk chromatin structure was not altered in an hta1∆htb1∆ strain, some areas of the genome were more sensitive to micrococcal nuclease (MNase) digestion (Norris et al., 1988). The CAG repeat is a strong nucleosome positioning element and thus could be more sensitive to H2A.1 depletion than other regions of the genome. To visualize nucleosome positioning at the CAG repeat, we used indirect end-labeling with a probe upstream of the CAG repeat (Figure 1—figure supplement 1A, red line) and measured sensitivity of the chromatin to MNase digestion. A series of discrete, protected fragments (~5 nucleosomes), is observed, indicating several positioned nucleosomes within and flanking the CAG repeat (Figure 2A). The pattern is consistent with canonical ~165 bp spacing between nucleosomes, and suggests there is no visible, major disruption to the chromatin structure in the hta1∆ cells compared to the wild-type (Figure 2A).
To generate a high-resolution nucleosome map of the regions flanking the CAG repeat, we used a custom Illumina BeadArray containing probes spanning 425 bp upstream to 438 bp downstream of the CAG repeat tract on the YAC, including two CAG repeat containing probes and one pure CAG repeat probe, in non-overlapping 30-mers (Supplementary file 3). We hybridized MNase digested mononucleosomes of hta1Δ, hta2Δ, and wild-type (CAG)85 cells to the array to measure nucleosome protection at the CAG repeat. The CAG repeat-containing probes produced a peak in intensity compared to the flanking, non-CAG repeat-containing probes, indicating strong nucleosomal protection at the CAG repeat (Figure 2B). This result confirms previous in vitro data that the CAG repeat is a strong nucleosome binding sequence (Wang et al., 1994; Wang and Griffith, 1995; Volle and Delaney, 2012) and that it positions a nucleosome in vivo on a yeast chromosome. The protection was not reduced in the hta1Δ and hta2Δ mutants (Figure 2B). Thus, both methods used show that there is a positioned nucleosome at the CAG tract that is not altered in the hta1Δ background. We conclude that nucleosome positioning is not the major contributing factor to CAG instability in the hta1Δ mutant; however, subtle differences in chromatin structure in the absence of H2A.1 that are not visible by these assays may still impact repair pathway selection.
H2A.1 and H2A.2 vary in amino acid sequence only by the position of threonine in the C-terminal tail, which occurs at either position 126 in H2A.1 or 125 in H2A.2 (Figure 1B). Previous work measuring steady state levels of 32P-labeled histones showed H2A phosphorylation was decreased ~2 fold in an hta1-T126A mutant, identifying T126 as a phosphorylatable residue (Wyatt et al., 2003). To address the role of specific residues in the H2A C-terminal tail in CAG stability, we introduced point mutations in the endogenous copy of HTA1. When T126 was rendered non-modifiable by mutation to alanine, expansions were significantly increased over wild-type (3.7-fold over WT, p=0.04 to WT) (Figure 3A; Supplementary file 2). Deletion of HTA2 (hta1-T126A hta2Δ) had no further impact on CAG expansion frequency (Supplementary file 2). Expression of hta1-T126A from a plasmid in an hta1Δhta2Δ background resulted in a similar increase in expansions (3.1-fold over WT; p=0.04 to WT)(Figure 3A, right). Therefore, the H2A.1-T126 residue is required for efficient repair of the CAG repeat to prevent expansions. Since phosphorylation of T126 would be lost in the T126A mutant, this suggests the possibility that phosphorylation of this residue is important for its role in repair.
We next asked if constitutive phosphorylation of the T126 residue could promote high-fidelity repair to prevent CAG expansions by introducing the phospho-mimic T126E mutation into the endogenous HTA1 gene. CAG expansion frequency remained elevated in the T126E mutant (4.9%, 3.7-fold over WT, p=0.01) (Figure 3A; Supplementary file 2). This result suggests that either dynamic phosphorylation of H2A.1-T126 or another characteristic of having a threonine at position 126 is important for promoting CAG stability. Given the specific role for H2A.1, this result suggests that the position of the threonine within the tail affects the efficiency with which the H2A copies contribute to DNA replication or repair. CAG contractions were relatively unaffected in both the hta1-T126A and hta1-T126E mutants (Supplementary file 2), therefore this residue primarily protects against expansions.
We also tested the importance of the H2A-S129 residue in repeat maintenance. Despite H2A-S129ph being preferentially detected at a (CAG)155 tract in this same location by ChIP (House et al., 2014b), the ability to phosphorylate H2A-S129 did not significantly affect CAG repeat tract stability; neither expansion nor contraction frequencies were significantly altered from wild-type in the hta1-S129A mutant, expressed either from the genomic copy of HTA1 or from a plasmid (Figure 3A, Supplementary file 2). The double hta1-S129A/T126A mutant expressed from the endogenous HTA1 gene locus did not result in any further increase in expansion frequency from the hta1-T126A single mutant (Figure 3A; Supplementary file 2).
We wanted to evaluate in vivo phosphorylation of the H2A-T126 residue and used a custom antibody raised against a phosphopeptide corresponding to the H2A C-terminal tail phosphorylated at T126. Although this antibody was specific to H2A-T126ph by peptide dot blot (Figure 3—figure supplement 1A) phosphatase treatment of cell extracts only resulted in a 30% reduction in antibody recognition (Figure 3—figure supplement 1B, Supplementary file 9). To test if this antibody could distinguish between H2A isoforms, recognition by the antibody was compared in wild-type (HTA1 and HTA2 both present), hta1Δ (only HTA2 present), hta2Δ (only HTA1 present) and hta1-T126A (genomic copy of hta1 mutated, HTA2 still present). The signal is significantly diminished in the hta1Δ and hta1-T126A strains to 20–30% of WT levels, but not in the hta2Δ mutant (Figure 3B. Supplementary file 9). We conclude that the antibody is specifically recognizing the H2A.1 protein isoform containing threonine at position 126 in the tail, with a low level of background reactivity to the H2A.2 isoform. Importantly, total H2A protein and HTA gene levels remain constant in each mutant (Figure 1—figure supplement 1B; Supplementary file 8) (Libuda and Winston, 2006). Since this antibody mainly detects H2A.1, but not H2A.2, we have designated it ‘H2A.1T126.’.
Since H2A.1 is specifically required to promote repair of the CAG repeats, we wanted to know if the H2A.1 histone variant can be detected at an expanded CAG repeat. We tested recruitment of H2A.1 to the CAG repeat by ChIP during S-phase at time points when we have previously seen measurable increases in repair factors (House et al., 2014b; Sundararajan et al., 2010). Cells containing a (CAG)155 tract were α-factor arrested in G1 and time points were taken after release into fresh media. Whereas H2A.1 levels remain relatively constant at an ACT1 control locus, H2A.1 is specifically enriched at the CAG repeat 40 min into S-phase, when we previously saw peak γH2A at this location (House et al., 2014b), and returns to baseline by 60 min (Figure 3C). We also tested H2B and H3 recruitment to monitor the H2A-H2B dimer and overall nucleosome levels at the CAG repeat. We found no significant differences in H2B or H3 occupancy relative to an endogenous ACT1 locus (Figure 3D), indicating that the recruitment of H2A.1 is specific and not an artifact of increased nucleosome occupancy at this site during S-phase. Western blot analysis of α-factor arrested and released cells showed no increase in H2A.1 across the cell cycle (Figure 3—figure supplement 1C). Interestingly, enrichment of H2A.1 at the CAG tract by ChIP is occurring 20 min after γH2A begins to accumulate and peak Mre11 enrichment at the CAG repeat is detected (House et al., 2014b; Sundararajan and Freudenreich, 2011). Therefore, incorporated H2A.1 peaks after initial damage signaling at the repeat and coincides with maximal levels of γH2A. This transient occupancy supports that H2A.1 is specifically incorporated during DNA repair.
To determine if H2A.1 expression is damage inducible, H2A.1 levels were monitored after exposure to MMS, a DNA base alkylating agent that causes abasic sites that can be converted into single and double strand breaks. In the presence of 0.01% MMS, H2A-S129ph is increased as expected, but H2A.1T126 signal is in fact decreased and is therefore not induced by the DNA damage caused by low levels of MMS (Figure 3E). We also see no change in H2A.1T126 signal when cells are treated with 0.2M HU + 0.03% MMS, a treatment that induces collapsed replication forks (Nagai et al., 2008) (Figure 3—figure supplement 2A). We conclude that H2A.1 expression is not induced by DNA damage, and tentatively conclude that H2A.1T126 phosphorylation is also not likely not induced by DNA damage, since a 30% increase in antibody signal was not observed (as would be expected based on additional antibody recognition of phosphorylated form). A caveat to the conclusion of no damage inducibility is that S129ph could reduce the antibody recognition, precluding detection of an increased signal after MMS treatment. Nonetheless, the conclusion is supported by the lack of sensitivity of the H2A.1-T126A mutant to DNA damaging agents (Figure 1—figure supplement 2) and the fact that we were unable to detect a decrease in antibody signal when kinases known to phosphorylate targets in response to DNA damage and up-regulate the DNA damage response were mutated (Figure 3—figure supplement 2C).
We also tested if H2A-S129ph is altered in the hta1-T126A mutant. Importantly, recognition of H2A-S129ph is not impaired by the hta1-T126A mutation (Figure 3E). Therefore, H2A-S129ph may occur unimpeded when T126 is changed to an alanaine and deficient γH2A formation cannot explain the phenotype in the hta1-T126A mutant. In contrast, there was a reduction in H2A.1T126 antibody recognition in the hta1-S129A mutant (Figure 3E), but given the genetic data that hta1-S129A does not phenocopy the CAG instability profile of the hta1-T126A mutant (Figure 3A) we conclude that the hta1-S129A mutation likely disrupts the antibody epitope rather than affecting H2A.1 T126 modification.
To determine how H2A.1 and T126 contribute to fidelity of CAG repeat repair and prevent expansions, we assessed repeat stability in the absence of DNA repair pathways. If repeat expansions in the hta1Δ mutant are arising through a low-fidelity repair event, expansion frequency will be reduced in the absence of the relevant repair pathway.
We first tested if expansions in the absence of H2A.1 arise through NHEJ by deleting the gene encoding Lif1, a DNA ligase IV subunit, in the hta1Δ and plasmid hta1-T126A mutants. Expansion frequency remained elevated in the hta1Δlif1Δ and hta1-T126A lif1Δ double mutants (Figure 4A and B); therefore, instability in the absence of H2A.1 is not arising through low-fidelity NHEJ. Consistently, Moore et al found no NHEJ defects in a plasmid end-joining assay in an hta1-T126A mutant (Moore et al., 2007). We next surveyed homology-dependent repair pathways. Rad5-dependent post-replication repair can be a source of expansions during low-fidelity repair at CAG repeats (House et al., 2014b). Although expansions in the hta1Δrad5Δ double mutant are somewhat suppressed compared to the hta1Δ single mutant, the difference is not statistically significant (Figure 4A). Likewise, expansions remain elevated in the hta1-T126A rad5Δ double mutant (Figure 4B). Thus, H2A.1 does not appear to be contributing to the fidelity of post-replication repair, however a caveat to this conclusion is that rad5Δ elevates expansions on its own, possibly precluding observation of a suppression. Since expansions in the double mutant are less than additive and are similar to the rad5Δ levels, Rad5 may work upstream of H2A.1. To determine if CAG expansions in the absence of H2A.1 arise through general HR, we measured stability of the CAG repeat in the absence of two key HR proteins: Rad52 and Rad51. Expansions in the hta1Δrad52Δ and hta1Δrad51Δ double mutants are significantly reduced 2.9-fold and 5.2-fold from the hta1Δ single mutant, respectively (p=2.8×10−3 and p=7.7×10−3 to hta1Δ) (Figure 4A). Similarly, expansion frequencies in the hta1Δrad57Δ double mutant are suppressed 4.5-fold (p=1.4×10−3 to hta1Δ) (Figure 4A). Corroborating these results, expansions were also suppressed ~2 fold in the hta1-T126A mutant lacking either Rad52, Rad51 or Rad57, but not to the level of statistical significance due to the somewhat lower starting level of expansions in the hta1-T126A mutant compared to the full HTA1 gene deletion (Figure 4B). Together, these results indicate that expansions in the absence of H2A.1 are arising through Rad51- and Rad52-dependent HR events and suggest that T126 plays a role in regulating HR. Since Rad57 is especially required for SCR (Mozlin et al., 2008), this is consistent with expansions arising during SCR, a pathway previously implicated in causing CAG instability (Kerrest et al., 2009; Nguyen et al., 2017).
To further evaluate the role of H2A.1 in homology-mediated repair events, we assayed the H2A mutants for their ability to undergo SCR using a genetic assay that measures rates of spontaneous unequal SCR as an estimate of overall SCR levels (Mozlin et al., 2008) (Figure 4C). In this assay, misaligned recombination between two ade2 null alleles can result in gene conversion to a functional ADE2 allele and the strain is converted from Trp+Ade- to Trp+Ade+ (Mozlin et al., 2008). SCR is not suppressed from wild-type in the hta2Δ mutant, but is significantly suppressed in the hta1Δ mutant (2.4-fold suppression from wild-type, p=7.9×10−3; Figure 4D). Thus, H2A.1 is required for efficient SCR while H2A.2 is not, mirroring the differential role of the two H2A copies in CAG repeat maintenance. Similarly, SCR levels are decreased 2.7-fold from wild-type in the hta1-T126A mutant integrated at the endogenous HTA1 locus (p=4.9×10−3) (Figure 4D). SCR is also suppressed, though more mildly, in the hta1-S129A mutant integrated at the endogenous HTA1 locus (1.9-fold from wild-type; p=0.03; Figure 4D), in agreement with a previous report that found a mild defect in SCR during repair of a DSB in an hta1-S129A mutant (Conde et al., 2009). Rates of SCR in the hta1-T126A and hta1-S129A mutants are also reduced in the hta2Δ background (Supplementary file 4), indicating that this reduction does not depend on the presence or absence of H2A.2. Since we observe a decrease in SCR in the hta1Δ and hta1-T126A mutants but an increase in CAG expansions that is HR-dependent, we conclude that the defect in efficiency of SCR must occur after the initiation step, such as during the D-loop synthesis. In summary, these results demonstrate that H2A.1 and H2A.1-T126 are required for efficient spontaneous SCR. This supports our conclusion that H2A.1 and H2A.1-T126 are required for proper recombination at the CAG repeat tract to prevent repeat expansions, and extends this finding to non-repetitive DNA sequences.
Histone H2A.1 could affect several steps of homology-mediate repair, including resection, homology search and invasion, D-loop synthesis, gap fill-in, or checkpoint signaling. To test which stages of repair are impacted by H2A.1, we used a single-strand annealing (SSA) system to simultaneously monitor SSA efficiency, repair kinetics, and the checkpoint response after induction of a DSB. In this system, an HO recognition site has been introduced into the LEU2 locus and galactose-induction results in a single HO DSB at this location. At the DSB, resection occurs on both sides of the break and repair via SSA occurs between the two U2 regions of homology (Figure 4E) (Vaze et al., 2002). Repair kinetics are monitored by Southern blotting and DNA damage checkpoint activation is monitored by Western blotting for phosphorylated Rad53. Using this system, we found no difference in viability or repair kinetics between hta1Δ, hta2Δ, and wild-type strains (Figure 4F; Figure 4—figure supplement 1A; Supplementary file 5). This indicates that H2A.1 is not important for resection, alignment, or gap fill-in, which are required steps of repair via SSA. Similarly, we saw no significant defect in the Rad53ph checkpoint response in hta1Δ (Figure 4—figure supplement 1B), suggesting that H2A.1 is not specifically required to mediate or recover from the DNA damage checkpoint response, though some other aspect of cell cycle regulation could be affected. Since repair via SSA does not require invasion into a homologous template or D-loop extension, this result strongly suggests that H2A.1 is specifically important for homology-mediated repair requiring a D-loop.
Since our results indicated that H2A.1 is required during D-loop mediated repair and CAG expansions could occur during DNA synthesis, we tested the role of the Polδ subunit Pol32, which is required for Polδ processivity. Interestingly, SCR is significantly suppressed from the wild-type in the pol32Δ mutant. This establishes that the Pol32 subunit of Polδ is required for D-loop extension during the short tract recombination measured by this assay (less than 1 kb). Notably, SCR suppression in the pol32Δ mutant is epistatic to the absence of H2A.1 and a phosphorylatable H2A.1-T126 residue, as the SCR rate is not further diminished in the hta1Δpol32Δ or hta1-T126A pol32Δ double mutants (Figure 5A). Therefore, these results support that H2A.1 and Thr126 are important in facilitating D-loop extension by Polδ during recombination.
We next tested the role of Pol32 in CAG expansions. In the pol32Δ single mutant, CAG repeat expansions are significantly increased over wild-type (5.8-fold over wild-type, p=2.0×10−3), indicating that processive DNA synthesis by Polδ is required to prevent repeat expansions (Figure 5B). We also found that CAG fragility is significantly increased in the pol32Δ mutant (Figure 5—figure supplement 1A; Supplementary file 1). We tested whether instability in the absence of Pol32 was due to its role in normal replication or recombination-associated DNA synthesis by deleting Rad51 in a pol32Δ mutant (Figure 5B). Expansions in the pol32Δ mutant were suppressed in the absence of Rad51 (to 3.4%, 2.3-fold decrease from pol32Δ, p=0.17), suggesting that expansions are at least in part due to polymerase slippage during D-loop synthesis. The increase in expansion frequency in the pol32Δ mutant is similar to that in the hta1Δ and hta1-T126A mutants. However, CAG expansion frequency drops below the level of each single mutant in the hta1Δpol32Δ and hta1-T126A pol32Δ double mutants (Figure 5B; Supplementary file 2), suggesting that Pol32-dependent synthesis is responsible for some of the expansions occurring in the hta1Δ and hta1-T126A backgrounds. Though the suppression is not statistically significant (for example p=0.12 for hta1Δpol32Δ compared to hta1Δ), it is markedly reduced from the additive levels expected if there was no interaction between the pathways (for example 4.6% in hta1Δpol32Δ vs. 16.6% predicted for additive). Since the expansions in hta1Δ and hta1-T126A are suppressed in the absence of Rad51, Rad52, Rad57, and Pol32 (Figures 4A and 5B), they are likely arising downstream of synapsis and initiation of DNA synthesis during recombination. Taken together, these results suggest that the CAG expansions observed in the absence of H2A.1 occur during a Pol32-dependent recombination process. For example, H2A.1 could promote efficient D-loop extension during replication of the donor strand, thereby preventing opportunity for DNA secondary structure formation.
Pol32 is known to be especially important for BIR, which involves extended D-loop synthesis that can proceed for many kilobases (Lydeard et al., 2007; Anand et al., 2013; Malkova and Ira, 2013) (see also Figure 5D). Considering the importance of Pol32 in preventing CAG expansions and promoting SCR and that it appears to function in the same pathway as H2A.1, we wondered if H2A.1 might also have a role in BIR. To directly test the role of H2A.1 in BIR, we used a system in which a DSB induced by the HO endonuclease can result in a non-reciprocal translocation when repair proceeds via BIR (Anand et al., 2014) (Figure 5C). BIR frequency is suppressed 2.3-fold from wild-type in the hta1Δ mutant (p=1.0×10−4 to WT), but remains at wild-type levels in the hta2Δ mutant (Figure 5D) and the hta1-T126A mutant (Supplementary file 6). Therefore, while H2A.1 is important for BIR, this function is either not regulated by T126 or the T126A phenotype was too subtle to be revealed by this assay. Interestingly, histone loss in response to DSBs has been implicated in promoting recombination and DNA repair rates (Hauer et al., 2017). Therefore, given that the BIR system depends on an induced DSB, altered histone levels in the hta1Δ mutant may increase gene conversions or other conservative repair pathways at the expense of BIR.
After HO-cutting, the cells were also surveyed for repair type by pinning colonies from the YP-Gal plates onto YPD, YPD+Nat and YC-URA. Surviving colonies were used to calculate the percent of colonies that underwent the indicated repair type (Figure 5E). The repair outcome profile in the hta1Δ mutant is altered from wild-type as a greater proportion of cells undergo other types of repair instead of BIR (Figure 5E) (Supplementary file 6). While less severe, the increase in these other types of repair events in the hta1Δ mutant is similar to the phenotype in the pol32Δ mutant, and therefore is likely a consequence of decreased BIR rates (Figure 5E). We conclude that deletion of H2A.1 results in decreased BIR and alterations in repair type frequencies, demonstrating that H2A.1 plays a role in facilitating efficient BIR. Together with the SCR data, this suggests that expansions in the hta1Δ mutant may be due to defective D-loop extension, since this step occurs during both SCR and BIR.
Stemming from an initial observation in a genetic screen that CAG repeat fragility and instability were elevated in an hta1Δ mutant but not in an hta2Δ mutant, we demonstrated that H2A.1 and H2A.2 differentially contribute to homology-mediated repair. The histone subtypes have been documented before to play different functions during the S. cerevisiae life cycle. The absence of gene product from HTA1 and HTB1 led to a constitutive heat shock response after exposure to high temperature, and this phenotype was not rescued with an additional copy of HTA2 and HTB2 (33). Taken with our results demonstrating differential roles for the H2A copies in preventing HR-mediated CAG repeat expansions, facilitating sister chromatid recombination, and promoting BIR, this further supports that H2A.1 is specifically important for repair or recovery from DNA damage. Our data indicate that H2A.1 is more efficient at promoting high-fidelity HR than H2A.2 because of protein sequence; specifically, the position of the phosphorylatable threonine in the H2A.1 C-terminal tail is more advantageous to repair than H2A.2. Both copies of yeast H2A contain the SQEL motif in the C-terminal tail, and therefore both copies are considered homologs of mammalian H2AX. A compelling implication of our result here is that yeast H2A.1 is in fact the closer homolog of mammalian H2AX, as it plays a greater HR-specific role than H2A.2. Like H2A.1, mammalian H2AX also contains a phophorylatable threonine two residues before serine 139. Also similar to H2AX, H2A.1 is initially a smaller proportion of the total H2A pool (Moran et al., 1990; Rogakou et al., 1998). Consequently, yeast H2A.1 and H2A.2 may be more akin to histone variants than histone copies.
H2A-Thr126 has previously been shown to be phosphorylated in vivo but its phosphorylation state after DNA damage and its overall contribution to break repair was unclear (Wyatt et al., 2003; Moore et al., 2007; Harvey et al., 2005; Chambers and Downs, 2007). Using a naturally unstable expanded CAG repeat tract, we have shown that H2A.1-T126 is important for maintaining CAG repeats and plays a role in promoting efficient SCR. Our results indicate that overall levels of HR/SCR are suppressed in the absence of H2A.1-T126, and that the recombination that does take place proceeds with low fidelity, leading to repeat expansions. Although our results implicate a requirement for phosphorylation of H2A.1-T126 in DNA repair, it is formally possible that some other physical property of threonine at position 126 may be important for repair, rather than phosphosphorylation per se. For example, the amino acid sequence in the H2A.1 C-terminal tail with threonine at position 126 may be more advantageous to DNA repair factor recruitment, independent of T126 modification status. Alternatively, H2A.1-T126 could influence neighboring modifications, such as S122 phosphorylation or acetylation of K124 or K127, though our data indicate that T126 does not modulate S129 phosphorylation.
The fact that expansions, an addition of bases, occur in hta1Δ and hta1-T126A mutants is most supportive of a role for H2A.1 in promoting a synthesis step of DNA repair. The expansions are likely arising downstream of synapsis and D-loop assembly since they were suppressed in the absence of Rad51, Rad52, and Rad57 (Figure 4A). Similarly, H2A.1 and T126 play no role in repair via SSA, which does not require formation of a D-loop (Figure 4F; Figure 4—figure supplement 1A; Supplementary file 6). Further, our data placing H2A.1 and T126 in the same pathway as Pol32 in the unequal SCR assay supports a role during D-loop extension. Therefore, we conclude that H2A.1 is most likely promoting efficient D-loop extension during replication of the donor strand. A second possibility is that H2A.1-T126 is required to promote a later step in the process such as D-loop resolution or re-establishment of the chromatin structure of the repaired gap. This could explain why the H2A.1-T126A mutant did not have a discernable effect on BIR, which does not include re-engagement of the extended D-loop with the initiating DNA molecule.
Our data indicated that the role of H2A.1-T126 was not directly dependent on S129ph, however, the two residues could act together to affect an outcome. H2A-S129ph begins to accumulate at the CAG repeat 20 min into S-phase and peaks at 40 min (House et al., 2014b). H2A.1 is either incorporated or phosphorylated at the CAG tract 40 min into S-phase, appearing only after initial accumulation of S129ph but coinciding with peak S129ph. Specific incorporation of H2A.1 at the CAG repeat when DNA damage is occurring is highly suggestive of some role for T126 at that time. Since our antibody also detects phosphorylated H2A.1, it could be that the additional ChIP signal is due to existing H2A.1 becoming phosphorylated locally. Alternatively it may reflect damage-specific incorporation of H2A.1, which may or may not be phosphorylated on T126 (Figure 6A). Levels of H2A.1-T126 decreased upon MMS damage (Figure 3E), and whole-genome proteomics did not identify H2A-T126ph as a DNA damage-inducible modification (Bastos de Oliveira et al., 2015; Smolka et al., 2007; Zhou et al., 2016), though because T126 is surrounded by two lysines (K124, K127), it will end up on a peptide only three amino acids long after trypsin digest, which will not be easily detected by mass spectrometry. Nonetheless, we favor a model in which T126ph is a pre-existing modification independent of S129ph, but that these two modifications may work together to promote efficient repair, for example by binding or excluding certain repair factors in a combinatorial fashion. In this scenario, T126ph would become enriched at the site of damage because of damage-specific incorporation of H2A.1 during repair (Figure 6) and disappear after repair is complete, either by removal of H2A.1 or local dephosphorylation of the T126 residue. Depending on the phosphorylation status of both residues, a repair protein could recognize one or both residues, which could facilitate removal of S129ph-bound proteins or turnover of S129ph/T126ph-containing nucleosomes, influencing the overall progression of the repair process. This idea is supported by the similar increase in expansions in the T126A and T126E mutants, suggesting that dephosphorylation of T126 could be important for repair fidelity. Alternatively, the presence of a threonine three amino acids away from S129ph, independently of its modification state, could influence repair, or the T126 effect may be independent of S129ph, even though they occur coincidentally.
Since we found that H2A.1 is specifically important for D-loop mediated HR repair, ATPase chromatin remodelers are attractive candidates for direct interaction with H2A.1-T126 or H2A.1-T126ph. Interestingly, in G2/M when a sister chromatid is available for recombination, γH2A is dispensible for recruitment of several chromatin modifiers, including Ino80 and subunits of RSC, NuA4, Rpd3, SWI/SNF, and SWR-C (Bennett et al., 2013). H2A.1-T126-mediated recruitment of chromatin remodelers could be important for opening the chromatin structure to allow access by repair factors, remodeling chromatin on the donor strand to facilitate synapsis or D-loop extension, or resetting the chromatin to promote repair resolution. The hta1-T126A mutant displays a defect in telomere positioning effect (TPE) (Wyatt et al., 2003), consistent with a role in reestablishing chromatin structure after repair or replication. We previously found that RSC and NuA4 are required to promote high-fidelity repair of the CAG repeat, and showed that RSC subunits are recruited to the repeat during S-phase (RSC2 peaks at 20 min, RSC1 at 40 min). We concluded that these factors are promoting post-replication repair events (House et al., 2014b), however, it is possible that RSC and NuA4 also have a more general role in any D-loop mediated repair process. Recruitment of any of these proteins to the CAG repeat during repair could facilitate chromatin remodeling to promote efficient D-loop progression or resolution and high-fidelity repair.
Hairpins formed by CAG or CTG repeats interfere with replication and induce joint molecules between sister chromatids, which have been visualized by two-dimensional gel electrophoresis (Kerrest et al., 2009; Nguyen et al., 2017). We propose a model in which gaps caused by replication bypass of the CAG repeat are repaired via SCR, and that incorporation of H2A.1 and/or H2A.1-T126ph ensures efficient remodeling and progression of Pol32-mediated D-loop extension, leading to high-fidelity repair (Figure 6A). In the absence of H2A.1, suboptimal signaling from the H2A.2-T125 residue may inefficiently recruit or retain chromatin remodelers or other DNA repair proteins, impeding D-loop extension due to a chromatin state that is not permissive to extension and copying (Figure 6B). Inefficient progression of the recombination intermediate could lead to Polδ stalling or transient dissociation of the 3’ end of the invading strand, which would allow opportunity for CAG repeat secondary structure formation and slippage during synthesis, leading to repeat length changes in the repaired DNA strand. Alternatively, chromatin remodelers or repair factors influenced by the presence of H2A.1-T126 may be required to reset the chromatin structure after repair. The permissive chromatin state could allow multiple, aberrant invasion events that would increase the opportunity for misalignments during D-loop initiation or elongation, resulting in repeat instability.
Perhaps less commonly, a replication fork stalled by a CAG or CTG hairpin may be converted to a one-ended DSB that could facilitate HR-dependent fork restart, similar to BIR (e.g. broken fork repair or BFR; Malkova and Ira, 2013). If fork restart proceeds with low fidelity, such as if recombination structures are misaligned or hairpins cause strand slippage, mutations can arise. Our data suggest that H2A.1 is required for efficient BIR, and CAG expansions occurring in the hta1Δ background were reduced when Pol32 was deleted, inhibiting BIR. Indeed, recovery from broken forks via BIR/BFR has recently been proposed to cause large-scale expansions of a (CAG)140 repeat tract in yeast (Kim et al., 2017), and our data at the (CAG)85 repeat are consistent with these results. However, in the wild-type background, expansion frequency is significantly increased in a pol32Δ mutant where BIR is suppressed (Figure 5B), indicating that BIR is not the only pathway creating expansions. Our data show that at least some expansions in the pol32Δ mutant occur during recombination (Figure 5B), suggesting that efficient D-loop synthesis through a CAG repeat requires Pol32. It remains possible that some expansions occurring in the pol32Δ mutant are arising by a different mechanism, such as impaired Polδ synthesis during replication. While efficient BIR was dependent on H2A.1, the rates were not affected in the hta1-T126A mutant, thus, the requirement of H2A.1 during BIR is not through the modifiable T126. The BIR assay involves a DSB whereas the spontaneous SCR assay and CAG repeat expansion assay is measuring the response at endogenous DNA lesions that will include a mix of nicks, gaps, and a small proportion of DSBs. We note that CAG fragility, CAG contractions, and BIR, that all involve an initiating DSB, were dependent on H2A.1 but not the T126 residue. Therefore, our data are consistent with H2A.1 levels affecting DSB repair while the H2A.1 sequence-specific role is during gap repair/SCR.
At the occurrence of DNA damage, recombination is thought to be more protective to genome integrity than end joining because the repair is templated. However, recombination can itself be mutagenic if it does not proceed in a regulated manner (Polleys et al., 2017; Guirouilh-Barbat et al., 2014). Turnover of chromatin modifications during repair is an attractive model for facilitating proper repair progression, either by influencing chromatin reorganization (reviewed in Polo, 2015) or by facilitating sequential recruitment and release of repair factors (reviewed in House et al., 2014a; Price and D'Andrea, 2013). Our results demonstrate a genetic interaction between H2A.1 and Pol32 (Polδ) in maintaining CAG repeat stability. The timing and reading of H2A.1-T126 and other chromatin marks may determine how effectively the Polδ complex moves through the donor strand during repair, ensuring that repair is efficient (timely) and that it proceeds with high-fidelity, limiting mutagenic repair outcomes.
Our results suggest that yeast H2A.1 is a closer homolog of mammalian H2AX, whereas H2A.2 is more functionally equivalent to mammalian H2A. Although the density of nucleosomes can vary at different genomic loci, the abundance of histones throughout the genome means they are readily available at the frontlines of DNA damage. In yeast, histone H2A.1 and H2A.2 are both present in a healthy cell and are likely interspersed throughout the genome. Transcript from H2A.2 is present in higher proportion than H2A.1 under normal conditions (cited in Moran et al., 1990), but there must be an adequate amount of H2A.1 present on the chromatin to act in a fashion that promotes DNA repair. This may be similar to H2AX distribution in human cells, which is 2.5–25% of the total H2A pool (Rogakou et al., 1998). Further, in mammalian cells the H2AX histone variant is also required for efficient SCR (Xie et al., 2004). The analogous H2AX threonine (T136) is also phosphorylated in mammalian cells (Li et al., 2010; Bennetzen et al., 2010). A T136V mutation did not decrease survival after IR or DSB-induced HR in mouse cells, but gap repair and repair fidelity were not analyzed (Xie et al., 2010). Thus, it would be interesting to test whether mammalian H2AX T136 plays a role in repair fidelity analogous to the role described here for yeast H2A.1-T126.
In all cases, reference to independent experiments refers to independent biological replicates, which were done from separate starting colonies or cultures that were treated separately (e.g. plated for analysis or prepared separately for DNA or protein preparation).
The yeast strains used in this study are described in Supplementary file 11. Mutant strain construction: Genes were deleted via PCR-based gene replacement with selectable gene markers. Integration of the selectable marker was verified using PCR at the 3’ and 5’ integration junctions and primers internal to the target gene to verify ORF absence. All assays were done with at least two independently created strains.
Assays to measure CAG repeat fragility were performed by fluctuation test as previously described (House et al., 2014b; Polleys and Freudenreich, 2018). Briefly, 10 single colonies with verified (CAG)85 tract lengths were grown 6–8 divisions and URA3 marker loss was tested by plating on FOA-Leu. Mutation frequency was calculated by the Method of Maximum Likelihood (Hall et al., 2009). Rates were evaluated for significant deviation from WT by the student’s t-test, as we were interested in comparison to the wild-type rate. At least three biological replicates were performed for each strain, using at least two independent transformants. See Supplementary file 1 for a list of individual assays. Outliers according to the Grubb’s test were removed.
PCR-based stability assays were performed as described (House et al., 2014b; Polleys and Freudenreich, 2018). Briefly, single colonies with verified (CAG)85 tract lengths were grown 6–8 generations and the tract length was measured in ~100–300 daughter colonies from two independent transformants. PCR products were run in 2% MetaPhor agarose or a custom gel mix on a fragment analyzer (Custom kit DNF945, Advanced Analytical Technologies, Inc). The number of expansions and contractions were evaluated for significant deviation from wild-type using the Fisher’s Exact Test (see Supplementary file 2). This test is appropriate as we are comparing categorical data (for example # expansions in wt vs # expansions in mutant) in small sample sizes (e.g. less than 1000). Sample size was determined by the number needed to determine statistical significance balanced by the practicality of the number of PCR reactions that could be done with available resources.
Chromatin digestion was performed as previously described (Koch et al., 2018, adapted from Wu and Winston, 1997), with the following modifications: spheroblasts were digested with 0, 0.25, 2.5, or 7.5 units of MNase and the DNA pellet RNase A digested for 30 min. The DNA was extracted twice using an equal volume of chloroform and precipitated with NH4OAc and isopropanol. Southern Detection: MNase digested DNA (20–30 μg) was run in 1.5% agarose at 80V for 6 hr and Southern blotted as previously described (Koch et al., 2018). Chromatin structure was probed with a 32P labeled 358 bp PCR fragment amplified from 102 bp upstream of the CAG repeat on YAC CF1. The experiment was repeated multiple times and a representative blot is shown.
Chromatin was isolated and MNase digested as described above, except mononucleosomes were prepared by digesting the chromatin with 10 units of MNase for 15 min. Purified mononucleosomal DNA was amplified using the GenomePlex Whole Genome Amplification kit (Sigma). Amplification and ~150 bp fragment sizes were verified by running in 1.5% agarose. Amplified mononuclesomes were purified with a GenElute PCR clean-up kit (Sigma) and 3’ biotin end-labled (Pierce); DNA was chloroform extracted, and labeling was verified by dot blot according to the manufacturer’s instructions. Purified mononucleosomes were applied to a custom Illumina array that contained YAC CF1 sequence spanning a region 425 bp upstream to 438 bp downstream of the repeat tract in 30 bp non-overlapping probes (Supplementary file 3). For each sample, a 12 μl aliquot of the purified mononucleosomal DNA was mixed with 3 μl of 15.3X SSC buffer containing 2.4% SDS, heated at 95°C for 5 min, and snap cooled. The sample was applied to the sample zone of the microarray and hybridization was carried out overnight at 62°C with humidity. Arrays were washed in stringency wash (0.2% SDS and 2X SSC buffer) with rocking for 5 min, followed by 0.05X SSC buffer with vigorous agitation for 1 min. To reduce non-specific binding of the stain, arrays were incubated with blocker casein in 1X PBS (Thermo) for 5 min. Arrays were then stained for 10 min in blocker casein with 1 µg/ml streptavidin-Cy3 in PBS (Invitrogen). Chips were vigorously agitated in 0.05X SSC buffer for 1 min, dried with compressed air, and scanned with a BeadArray Reader (Illumina) using Direct Hybridization settings with a factor of 5. Signal intensity was exported to Excel via BeadScan. The experiment was repeated twice for WT and 3x for hta1Δ and hta2Δ strains.
Strains were grown in YPD at 30°C with agitation to log phase (OD600 = 1); cells were either untreated or exposed to the following drug treatments at 30°C with agitation: 0.01% MMS for 2 hr, 0.03% MMS for 1 hr, 0.2M HU for 1 hr, 0.2M HU + 0.03% MMS for 1 hr. Lysates were extracted according to Adams et al. (1997) and https://research.fhcrc.org/gottschling/en/protocols/yeast-protocols/protein-prep.html and Western blotted onto PVDF. Blots were probed with anti-H2A (Abcam #13923; 1:5000), anti-H2AT126 (Aves Lab custom antibody; gift from Krebs lab; 1:2500), or anti-H2AS129ph (Abcam; ab15083; 1:2500) in 2.5% milk in 1X PBS (pH 7.4). The signals were detected with HRP-conjugated secondary antibody (1:2500) and ECL (Pierce). Western blot signals were quantified by ImageJ. The fold change in signal from WT was determined by comparing the relative quantification value (rqv) of a mutant to the relative quantification value of the WT (rqv = ratio of the indicated band to the loading control, with background subtracted). Each experiment was repeated at least three times and quantified; individual and mean values, SEM and p-values to control are reported in Supplementary file 9.
The anti-H2A.1T126 custom chicken antibody was generated by Aves Lab (http://aveslab.com). The antibody was raised against the phosphopeptide KKSAKA[pT]KASQEL. An attempt to create a second batch of this antibody by the same procedure was not specific, thus the specificity of antibodies created in this manner is variable. Only experiments with the H2A.1-specific antibody preparation are included.
ChIP was performed in biological duplicate as previously described (Aparicio et al., 2005; House et al., 2014b) using anti-H2A.1T126 (custom preparation, Aves Lab), normal chicken IgY (Millipore AC146), anti-H2B (Abcam ab1790), or anti-H3 (Millipore 05–928) and Dynabeads Protein G (Invitrogen) or anti-IgY agarose beads (Invitrogen) for immunoprecipitation. Time points were taken after release from synchronization in 5 μM α-factor for 1.5 hr at 30°C with shaking. DNA levels were measured by qPCR using SYBR Green PCR mastermix (Roche) or Power SYBR Green PCR Master Mix (Applied Biosystems); qPCR reactions were run in duplicate amplifying a 200 bp fragment 0.6 kb upstream of the CAG repeat or at ACT1. Enrichment at the CAG repeat was determined by the ΔΔCt method; H2B and H3 ChIP was additionally normalized to enrichment at an ACT1 control locus. See Supplementary file 4 for raw IP/Input values, qPCR conditions, and primer sequences.
Assays were performed as previously described (House et al., 2014b; Mozlin et al., 2008). Briefly, Trp+ Ade- cells were grown in YEPD to saturation. Total viable cell count was measured by plating 10−5 dilutions on yeast complete (YC) media and recombinants were selected by plating 10−2 dilutions on YC-Trp-Ade. Recombination rates were calculated by the method of the median and rates were tested for statistical significance using the Student’s t-test (Supplementary file 5). Outliers according to the Grubb’s test were removed.
Assays were modified from Anand et al. (2014). Briefly, colonies on YEPD+Nourseothricin were serially diluted and plated on YEPD and YEP+Galactose in duplicate. Colonies were counted and percent viability on YEP+Galactose was determined by dividing by the number of colonies on YEPD. To determine the frequency of BIR and other types of repair, all YEP+Galactose colonies were pinned onto YEPD, YEPD+Nourseothricin, and YC-URA. BIR frequency was calculated as number of URA+ NAT- cells divided by the number of colonies on YEPD and tested for statistical significance using the Student’s t-test (Supplementary file 7).
FY406 H2A point mutant strains were a gift from Dr. Jessica Downs (Harvey et al., 2005). The H2A point mutants are contained on a CEN plasmid containing either hta1-T126A or hta1-S129A point mutation, a wild-type copy of HTB1, and a HIS3 marker. Both the HTA1/HTB1 and HTA2/HTB2 loci are deleted in these strains so the sole copies of H2A and H2B are expressed from the plasmid. A YAC containing a (CAG)85 repeat tract was introduced by cytoduction as previously described (Callahan et al., 2003). For genomic integration of hta1 point mutants, the HTA1 gene plus 200 bp upstream and 347 bp downstream of the gene was cloned into either the pFA6a-KanMX6 or the pFA6a-HPH vector. The hta1-T126A, T126E, S129A or T126A/S129A point mutations were integrated into the HTA1 plasmid via cloning and PCR-based gene replacement methods integrated the mutations, confirmed by PCR and sequencing. Replacement of HTA2 in HTA1 gene locus: A selectable KANMX6 gene was knocked-in 150 bp downstream of the HTA2 stop codon. The entire HTA2 gene and KANMX6 marker was PCR amplified with primers with 40 bp tails homologous to the HTA1 locus. Integration of HTA2+KANMX6 into the HTA1 gene locus was confirmed by PCR and sequencing. The HTA2 gene at its endogenous locus was deleted.
hta1Δ and hta2Δ strains used in this study (Supplementary file 8) were grown overnight and subjected to DNA extraction via phenol:chloroform and bead beating. Samples were diluted and total genomic HTA1 and HTA2 levels were measured in duplicate by qPCR using Power SYBR Green PCR Master Mix (Applied Biosystems, 4367659). Absolute quantities were determined by comparison to a standard curve and normalized to the control locus ACT1 (Supplementary file 8). Given the high sequence similarity between HTA1 and HTA2, primers were designed such that they were specific to HTA1 or HTA2. Primer sequences are in Supplementary file 8.
Cells were grown to mid-log phase and subsequently treated with 0.035% MMS for 3 hr. For on blot assays, 9.25 × 107 were collected. For in tube assays 1.26 × 109 cells were collected. CIP treatment on-blot: Whole cell lysates were extracted according to Adams et al. (1997) and https://research.fhcrc.org/gottschling/en/protocols/yeast-protocols/protein-prep.html and Western blotted onto PVDF. Post-transfer, membranes were activated, blocked in 5% BSA in 1X TBST for 1 hr and washed with 1X TBS. Membranes were cut in half, placed in 5 mL of 1X Cutsmart buffer and either treated with to 10 U of calf intestinal alkaline phosphatase (CIP) (NEB #M0290S) or received no treatment. Membranes were incubated for 1 hr at 37°C. CIP pre-treatment: Histones were acid extracted according to Jourquin and Géli (2017). For the phosphatase treatment, 0.5 ul of purified histones was resuspended in 1X Cutsmart buffer and incubated with 3 U CIP (NEB #M0290S) for 1 hr at 37°C. Control samples were incubated with no CIP added. Samples were Western blotted onto PVDF and subsequently probed. Visualization and quantification: Blots were probed with anti-H2A (Abcam ab13923; 1:5000), anti-H2AT126 (Aves Lab custom antibody; gift from Krebs lab; 1:2500), or anti-H2AS129ph (Abcam; ab15083; 1:2500) in 5% BSA in 1X PBS (pH 7.4). The signals were detected with HRP-conjugated secondary antibody (1:2500) and ECL (Pierce). Western blot signals were quantified by ImageJ.
Assays were modified from Vaze et al. (2002). For viability assays, colonies were grown for 2–3 divisions in YP-Lactate at 30°C, serially diluted, and plated on YEPD and YEP+Galactose in duplicate. Colonies were counted and percent viability was determined (number of colonies on YEP+Galactose divided by the number of colonies on YEPD). See Supplementary file 6.
Time course experiments were performed as previously described (Vaze et al., 2002). Briefly, 400 mL YP-Lactate was inoculated with 1−3 × 106 cells. Cells were grown overnight at 30°C. Fifty millilitres of cells were removed and then galactose was added to a final concentration of 2% to induce the DSB. Aliquots were removed at the indicated timepoints and DNA was extracted using phenol:chloroform and bead beating. Purified DNA was normalized using a Qubit (dsDNA BR kit; Q32850), digested overnight with KpnI and Southern blotted using a probe specific to LEU2.
Protein extraction was performed as previously described (Foiani et al., 1994). Briefly 5 OD of cells at the indicated timepoints were treated with 20% trichloroacetic acid and bead-beating. Total protein extracts were separated on a 7.5% gel, blotted onto PVDF and probed with anti-Rad53 (Abcam ab166859; 1:1000). Phosphorylated Rad53 is visible as a retarded band on the blot.
mRNA extraction was done on approximately 1.2 × 108 logarithmically growing cells using the Illustra RNAspin Mini kit (GE, 25-0500-70). cDNA synthesis was done using SuperScript II Reverse Transcriptase (Invitrogen, 18064022). mRNA was diluted and POL32 levels were measured in duplicate by qPCR using Power SYBR Green PCR Master Mix (Applied Biosystems, 4367659). Absolute quantities were determined by comparison to a standard curve of genomic DNA and normalized to the control locus ACT1 (Supplementary file 10). Primer sequences are in Supplementary file 10.
Methods in Yeast GeneticsCold Spring Harbor Laboratory Press.
Chromosome rearrangements via template switching between diverged repeated sequencesGenes & Development 28:2394–2406.https://doi.org/10.1101/gad.250258.114
Chromatin immunoprecipitation for determining the association of proteins with specific genomic sequences in vivoCurrent Protocols in Molecular Biology 69:21.3.1–21.321.https://doi.org/10.1002/0471142727.mb2103s69
Site-specific phosphorylation dynamics of the nuclear proteome during the DNA damage responseMolecular & Cellular Proteomics 9:1314–1323.https://doi.org/10.1074/mcp.M900616-MCP200
Histone H2A variants in nucleosomes and chromatin: more or less stable?Nucleic Acids Research 40:10719–10741.https://doi.org/10.1093/nar/gks865
Mutations in yeast replication proteins that increase CAG/CTG expansions also increase repeat fragilityMolecular and Cellular Biology 23:7849–7860.https://doi.org/10.1128/MCB.23.21.7849-7860.2003
The contribution of the budding yeast histone H2A C-terminal tail to DNA-damage responsesBiochemical Society Transactions 35:1519–1524.https://doi.org/10.1042/BST0351519
Is homologous recombination really an error-free process?Frontiers in Genetics 5:175.https://doi.org/10.3389/fgene.2014.00175
Histone degradation in response to DNA damage enhances chromatin dynamics and recombination ratesNature Structural & Molecular Biology 24:99–107.https://doi.org/10.1038/nsmb.3347
Chromatin modifications and DNA repair: beyond double-strand breaksFrontiers in Genetics 5:296.https://doi.org/10.3389/fgene.2014.00296
Histone purification from Saccharomyces cerevisiaeMethods in Molecular Biology 1528:69–73.https://doi.org/10.1007/978-1-4939-6630-1_5
SRS2 and SGS1 prevent chromosomal breaks and stabilize triplet repeats by restraining recombinationNature Structural & Molecular Biology 16:159–167.https://doi.org/10.1038/nsmb.1544
The role of break-induced replicationin large-scale expansions of (CAG)n/(CTG)n repeatsNature Structural & Molecular Biology 24:55–60.https://doi.org/10.1038/nsmb.3334
Break-induced replication: functions and molecular mechanismCurrent Opinion in Genetics & Development 23:271–279.https://doi.org/10.1016/j.gde.2013.05.007
A yeast H2A-H2B promoter can be regulated by changes in histone gene copy numberGenes & Development 4:752–763.https://doi.org/10.1101/gad.4.5.752
Long palindromic sequences induce double-strand breaks during meiosis in yeastMolecular and Cellular Biology 20:3449–3458.https://doi.org/10.1128/MCB.20.10.3449-3458.2000
The two gene pairs encoding H2A and H2B play different roles in the Saccharomyces cerevisiae life cycleMolecular and Cellular Biology 7:3473–3481.https://doi.org/10.1128/MCB.7.10.3473
Methods to study repeat fragility and instability in Saccharomyces cerevisiaeMethods in Molecular Biology 1672:403–419.https://doi.org/10.1007/978-1-4939-7306-4_28
Genetic assays to study repeat fragility in Saccharomyces cerevisiaeMethods in Molecular Biology 2056:83–101.https://doi.org/10.1007/978-1-4939-9784-8_5
Reshaping chromatin after DNA damage: the choreography of histone proteinsJournal of Molecular Biology 427:626–636.https://doi.org/10.1016/j.jmb.2014.05.025
DNA double-stranded breaks induce histone H2AX phosphorylation on serine 139Journal of Biological Chemistry 273:5858–5868.https://doi.org/10.1074/jbc.273.10.5858
Megabase chromatin domains involved in DNA double-strand breaks in vivoThe Journal of Cell Biology 146:905–916.https://doi.org/10.1083/jcb.146.5.905
Nucleosome remodelers in double-strand break repairCurrent Opinion in Genetics & Development 23:174–184.https://doi.org/10.1016/j.gde.2012.12.008
Repeat instability during DNA repair: insights from model systemsCritical Reviews in Biochemistry and Molecular Biology 50:142–167.https://doi.org/10.3109/10409238.2014.999192
Control of sister chromatid recombination by histone H2AXMolecular Cell 16:1017–1025.https://doi.org/10.1016/j.molcel.2004.12.007
Wolf-Dietrich HeyerReviewing Editor; University of California, Davis, United States
Kevin StruhlSenior Editor; Harvard Medical School, United States
Wolf-Dietrich HeyerReviewer; University of California, Davis, United States
In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.
The role of chromatin in homologous recombination is poorly understood, and this work makes the surprising observation that in budding yeast histone H2A.1 but not H2A.2 is required for Rad51-dependent recombination. The authors are able to ascertain that the T126 in HTA.1, which is unique to H2A.1, is critical for this function.
Decision letter after peer review:
The previous reviews from another journal have been considered and evaluated along with the manuscript by an expert Reviewing Editor, who has the following comments:
The previous reviews and responses by the authors address the concerns raised in two rounds of review at another journal. I disagree with the previous editorial decision that the issues surrounding the H2A.1T126 anti-serum jeopardize the main conclusion of the study. The authors appropriately describe the limitations of this anti-serum and carefully interpret the data accordingly. I concur with the reviewers, who after 2 rounds of revisions, support publication of this manuscript.
The only change requested is an edit in the Significance Statement. Please change 'error-free' to 'high-fidelity' or similar. Work in the Strathern and Haber laboratories demonstrated that DSB repair by homologous recombination is associated with a significant increase in mutagenesis.https://doi.org/10.7554/eLife.53362.sa1
[Editors' note: we include below the reviews that the authors received from another journal, along with the authors’ responses.]
My co-authors and I would like to submit the manuscript “Saccharomyces cerevisiae H2A copies differentially contribute to recombination and CAG/CTG repeat maintenance, with a role for H2A.1 threonine 126” for consideration for publication as a Research Article in eLife. This manuscript was previously favorably reviewed at another journal, but ultimately rejected by the editors due to a difference in opinion about using an antibody with less than 100% specificity (but the only one available) in one of the figures. Our results are not dependent on this antibody, though they add to the genetic data, and the reviewers thought our interpretations were appropriate and the experiment using the antibody was a significant addition to the paper. In yeast, there are 2 copies of histone H2A (H2A.1 and H2A.2) that were previously thought to have identical functions. In this manuscript, we characterize a repair-specific role for H2A.1 that is not shared by H2A.2, and can be attributed to a phosphorylatable threonine residue (T126) in the C-terminal tail. We find that H2A.1 is required for efficiency of sister chromatid recombination and for promoting repair fidelity during homologous recombination. Thus, yeast H2A.1 is a closer homolog of mammalian H2AX, which is specifically required for DNA damage repair. This is the first report of this H2A.1-specific function and thus will be of great interest to the readers of eLife. This repair-specific function of H2A.1 will now need to be considered when thinking about the role of histone H2A in the cell.
In the text, we have remedied the typos and nomenclature/genotype concerns, added an explanation of the statistics and more consistently reported the comparisons and p-values within the text instead of relying on the supplemental table. We have also added more supplemental tables that contain all of our raw data. Point-by-point responses can be found in-line below. We thank the editor and the reviewers for your comments and suggestions, as addressing them has further strengthened our story and solidified our conclusions.
Experimentally we have added the following new findings and controls:
– Quantification of all blots, with repetitions and error bars. This allowed us to determine the reactivity of the H2A.1T126 antibody with H2A.1 versus H2A.2 variants. The specificity for H2A.1 (copy 1) is about 70%, which is in line with other histone tail antibodies on the market.
– A phosphatase treatment to show at what level the H2A-T126ph antibody preferentially recognizes the phosphorylated form of H2A.1 in vivo. We found the antibody does not discriminate between the two forms very well in vivo, showing only a 30% preference for the phosphorylated form. Therefore, we have modified our conclusions from the Western blots regarding phosphorylation accordingly.
– ChIP of the histones at the CAG repeat, showing specific enrichment of the H2A.1 variant at the same time γH2AX is maximal, while overall histone/nucleosome levels remain unchanged.
– qPCR of HTA gene levels in the hta1Δ and hta2Δ mutants verify no gene amplification at the genomic level in the strains used in this study.
– Further verification that H2A.1 and T126ph is not induced by MMS, HU, or HU+MMS (quantification of the Western blots using the H2A.1T126 antibody) and spot assays that show no sensitivity to DNA damaging agents in the hta1Δ or hta1-T126A mutants.
– RT-PCR data showing that levels of Pol32 transcript are not changed in the hta1Δ mutant, verifying that the pol32Δ phenotypes are not an indirect consequence of Hta1 affecting POL32 transcription levels.
– CAG fragility and BIR assays using the hta1-T126A mutant (genomic integration; showed no phenotype).
– An SSA assay that demonstrates no role for HTA.1 in a repair reaction that does not involve a D-loop.
– A Rad53ph Western that shows no significant checkpoint recovery defect in the hta1Δ mutant. The SSA and Rad53ph data allowed us to narrow down the step at which H2A.1 is most likely acting (D-loop synthesis).
Our overall conclusions have remained the same since our initial submission of the paper. We conclude that H2A.1 specifically contributes to maintaining fidelity of homology-mediated repair, and our ChIP results suggest repair-specific incorporation of this histone variant. Further, we demonstrated that H2A.1 is genetically in the same pathway as the Polδ subunit Pol32 during sister chromatid recombination (SCR) and HR-mediated CAG repeat expansions. Using several genetic assays, we were able to demonstrate that the role of H2A.1 extends beyond repair of repeats and is broadly required for HR-mediated repair (both SCR and BIR) at non-repetitive sequences as well.
My most serious concern involves the antibody used to detect H2A-T126phos. The supplemental documentation (including peptide sequences, etc) in Moore…Krebs 2007 paper is no longer available, and the paper itself seems self-contradictory with regards to which antiserum was used—the Results section seems to imply that the custom antibody was used, but the Materials and methods refers to an Abcam antibody that has been withdrawn from distribution. In addition, Moore…Krebs do show increased signal with their antibody (whatever its source) in response to menadione (2x), phleomycin and MMS (1.5x). Finally, the supplementary figure included as Figure S3 is not well documented and is, in my opinion, of sub-standard quality.
The various antibodies used in the literature are confusing, and we have added more details about our antibody in the Materials and methods to clarify its origin. The antibody used in the Moore…Krebs 2007 paper was a small batch of test antibody made by Abcam. Dr. Krebs has contacted Genetics and requested that they re-post the missing supplemental figure showing its specificity. Abcam then made a larger batch for general distribution, and this batch was not specific and was pulled from the market.
The antibody used in this paper is different from both of these and was a custom antibody made by Aves raised against the phosphopeptide KKSAKA[pT]KASQEL. This antibody showed specificity to the phosphorylated peptide in vitro(Figure 3—figure supplement 2A, now with better labeling). Later, we ordered a second batch of antibody because we were running low, and two different preparations (from 2 chickens) were both not specific – neither are used for any of the experiments in this paper. We have saved back ~70 μl of our more specific Aves antibody for distribution upon request.
I think that, given the dependence the current paper places on this antibody in assigning the T126A phenotypes to a lack of phosphorylation, it is imperative that validation of the antibody be complete and accessibly documented. In particular, because of differences that can exist between peptides in solution and in the context of the full-length protein, it seems to me critical that a phosphatase-treatment control be included, to quantitatively determine what fraction of the signal on Westerns is due to T126phos, and what fraction might be due to cross-reactivity with the unmodified histone (anti-S129phos would serve as a good control in this). I realize that this is not possible with your current protein preps, but it should be pretty straightforward to do the analysis with acid-acetone extracted histones (see Jourquin F., Géli V. (2017) Histone Purification from Saccharomyces cerevisiae. In: Guillemette B., Gaudreau L. (eds) Histones. Methods in Molecular Biology, vol 1528 for a protocol).
We did the phosphatase experiment per request, with 3 repetitions and quantification. With both in-solution and on-blot phosphatase treatment the T126ph signal was reduced an average of 30% compared to the level before phosphatase treatment. Thus, the antibody has some preference for the phosphorylated form but also recognizes the unphosphorylated form of H2A.1 in vivo.
On the other hand, the antibody shows 70-80% specificity for H2A.1 compared to H2A.2. Thus, it is reasonably specific for H2A copy 1. This level of specificity is in line with other histone tail antibodies that have been used in the literature. For example, The Abcam H3K27ac antibody (ab4729) is extensively used in the literature (it has 723 references listed on the product information page). By peptide array, this antibody is <30% crossreactive with unmodified H3, <20% crossreactive with K9ac, <20% crossreactive with K36ac, and <20% crossreactive with K18ac. It is still acceptably used to demonstrate H3K27ac.
Therefore, although our H2A.1-T126 antibody is not perfect, we feel it is still useful for detecting differences between the two isoforms of H2A, and we have made conclusions accordingly, with appropriate reference to the background isoform cross-reactivity. On the other hand, we have removed all conclusions relating to specificity to the phosphorylated form, as the 30% preference for the phosphorylated form is too weak to draw definitive conclusions. However, based on our experiments showing a phenotype for the hta1-T126A mutation in vivoin multiple assays, we still speculate in the discussion that phosphorylation of HTA.1-T126 may be important for fidelity of sister chromatid recombinational repair, but include the caveat that this could be due to the polarity of the threonine residue and not necessarily T126ph.
In summary, we have modified our language throughout to only conclude that H2A.1 (copy 1) and a threonine at position 126 is important for our various phenotypes but have left in speculation that this is due to T126 phosphorylation.
I think that it is also important, given the relatively high (and, to my impression, variable) reactivity of the antibody with T126A and S129A mutants (compare Figure S2B with S2D), that quantitative measures be made for all western blots using this antibody. If the antibody cannot be validated, or if substantial and consistent quantitative differences in reactivity with wild-type and mutant proteins cannot be documented, then I think that it will be important to remove references to T126 phosphorylation from the paper, although of course speculation in the discussion is always allowed!
We have now added quantification for all Western blots using the H2A.1-T126 antibody, including duplicate experiments done, so that all experimental values obtained are shown with points and error bars on the graphs. See Figures 3B, 3C, Figure 3—figure supplement 1 and 2B. What can be seen in Figure 3B is that the antibody is specific for H2A.1 (copy one, Hta1 protein). However, there is a background level of reactivity to H2A.2 (averaging 20-34% of WT H2A levels, depending on calculation method, Supplementary file 9). The background reactivity to the mutated T126A H2A.1 tail (averaging 26-30% of WT H2A) is similar to the H2A.2 tail background. Mutation of the H2A1.1 tail at S129 (hta1-S129A) also reduces reactivity to a similar 23-24% background level.
Altogether, we conclude the antibody is about 70% specific to the unmutated T126- phosphorylatable version of the H2A.1 tail and has about 30% cross-reactivity to versions of the H2A tail that cannot be phosphorylated on T126.
Statistics. Reviewer 1 (point 4) is being overly gentle. I strongly believe that if an apparent difference is not statistically significant, then one is not justified in calling it a difference. This is a concern that bedevils many of the comparisons, particularly those in Figure 4 and 5. It seems to me that there are two choices here—either a) increase the sample size to the point where there is expected to be sufficient statistical power to distinguish differences, or b) remove mention of these from the text.
This concern would apply to Figure 4B (reductions compared to hta1-T126A) and 5B (reductions compared to hta1Δ). We have tempered our language in response to your concerns about the statistics; however, we have also included an explanation of Fisher’s Exact Test – that this is a conservative test that can often miss biologically relevant differences. The suppression in expansions is consistent between experiments, but at the current level of difference an excess of 500-1000 colonies/PCR reactions would be required for each mutant to lead to a p-value of <0.05 using Fisher’s. This was not feasible to accomplish and would not have changed the overall conclusions. Given the type of data, we believe it is valid to point out consistent trends.
Genotypes in text and figures. I found myself confused as to which experiments were done with mutant alleles integrated at the endogenous locus, mutant alleles integrated at the endogenous locus but with an hta2-deletion, mutant alleles on a plasmid over a deletion of both HTA-HTB loci, etc. It would really help if the nomenclature in the text and figures made this clear. For example, comparing rad51-del with rad51-del hta1-T126A would appear to me to involve not just a difference in HTA1 genotype, but also a difference in the source of both H2A and H2B (plasmid rather than endogenous locus). It would really help if these differences were transparent to the reader.
To address the confusion pertaining to the genotypes, we have more explicitly stated when point mutations are expressed from a plasmid or integrated at the endogenous locus in the text and/or legends. In terms of the comparisons, we have also clarified this throughout. Per the example listed of rad51-del with rad51-del hta1-T126A, you are correct that the T126A mutant is being expressed from the plasmid. However, these two strains were not compared statistically. We clarified that the comparison is being made from hta1-del to hta1-rad51-del, or hta1-T126A to hta1-T126A rad51del etc. All comparative p-values are listed in Supplementary file 2 and we have added the comparisons to hta1-del and hta1-T126A for the contractions as well.
“partial suppression of contractions was observed (Table S2)”
“expansions are somewhat suppressed in the hta1-del rad5-del double mutant” is ambiguous; since expansions are at similar levels in both rad5-del strains. Is the difference between hta1-del and hta1-del rad5-del statistically significant?
No, it is not. We have added a clear statement that the difference between hta1-del and hta1-del rad5-del is not statistically significant. However there is a caveat to concluding that template switch has no role that is now also pointed out.
“and were suppressed approximately 3-fold in the hta1-T126A rad57-del mutant”. I think that the difference is 2-fold.
Figure 1C. “wild-type control” for the hta1-del::HTA2 hta2-del strain would be hta2-del; statistical comparisons should be made between these two strains. Don’t think this will change the outcome.
We added the comparison between hta1::HTA2 hta2 and hta2 single to the sentence (p = 7.5 x 10-3).
Figure 2A. These gels/Southerns are not of sufficient quality to determine if there are differences in nucleosome spacing—all that can be inferred it that there is not a major disruption in chromatin structure (i.e. there are still nucleosomes). I think that these gels either should be redone or they should be dropped from the paper.
The quality of these blots is in-line with the expected quality for this type of experiment. We redid the gels/blots several times and got similar quality data as the experiment presented; we do not think it is possible to improve on it. The fuzziness is due to several factors, 1) MNase cleaves between nucleosomes, so exact cutting sites will vary if nucleosome binding sites are not exactly the same in all cells, leading to slightly different-sized fragments, 2) this was a very long gel and so slight differences in fragment size become more obvious, leading to fuzziness, 3) the CAG repeat positions nucleosomes, but the CAG-85 repeat (255 bp) isn’t necessarily centered on a nucleosome – slight shifting between cells/genomes or binding more than one nucleosome is possible.
We have modified our conclusion to “there is no visible, major disruption to the chromatin structure in the hta1∆ cells compared to the wild-type”, which can be seen from this blot. To complement the less-precise MNase data, we verified that nucleosome protection at the CAG tract is not altered in hta1Δ or hta2Δ mutants using the more precise Illumina array method, however in this method the MNase digestion is done to completion so only one nucleosome binding site is evident. We think the two types of data complement each other, and that this is an important piece of negative data that eliminates one hypothesis for the difference in CAG instability between hta1Δ and hta2Δ strains. We have therefore left Figure 2 as-is.
Figure 4A. What do the carat (^) marks mean?
This has been added to the figure legend – suppression from hta1Δ single.
Tables S1, S4, S5, S6 and corresponding figures. Underlying data from individual experiments need to be included as a supplementary table (I suggest an Excel file with data from individual experiments). In Figure 5E, some measure of statistical significance is needed. I suggest binning the data into BIR and non-BIR outcomes.
Binning BIR and non-BIR outcomes for 5E is identical to the statistical analysis in 5D, so we refer you to 5D for this information. However, to clarify Figure 5E, we have added more details of how repair type was assessed and the outcomes to the Results section. The raw data for all experiments has been added to the supplementary tables in the Excel document. BIR statistics have been added to the BIR supplementary table; a student’s t-test was used to test significant deviation from the wild-type for each repair outcome.
In this study, the authors examine the impact of histone H2A modifications on the stability of a very large CAG/CTG trinucleotide repeat tract, which effectively acts as a fragile site within the DNA. This allows the authors to evaluate the contribution of different repair mechanisms to the stability of the tract (or fragile site) and the impact that H2A-1 modification has on repair pathway selection. The authors demonstrate that the presence or absence of H2A.1 impacts the stability of the long CAG/CTG tract, i.e. there is reduced stability and increased fragility when it is absent. H2A.1 was identified in a screen, H2A.2 does not have the same effect. Given that the difference between the two versions of H2A is a phosphorylatable Thr at position 126 of H2A.1, the authors characterize the importance of this residue in tract stability. Mutation of Thr-126 largely phenocopies the hta1 deletion, with respect to tract expansions. Furthermore, the absence of H2A.1 impacts the repair product profile, apparently suppressing sister chromatid exchange and break-induced replication. Notably, phosphorylation of serine 129, which has previously been implicated in DNA damage response, does not affect tract stability. The role of pol32 deletion suggests a role for efficient D-loop extension in maintaining tract stability.
This is a comprehensive and rigorous study that reveals a role for H2A.1 Thr126 ph in stabilizing CAG/CTG repeats. Beyond that, the authors argue convincingly that H2A.1 more broadly influences DNA repair pathway selection to promote higher fidelity repair, at least with respect to tract length, perhaps via sister chromatid exchange and break-induced replication. This analysis makes the work of interest to those working on repeat tract stability, as well as to the broader community interested in genetics and genomics.
1) The effect of hta1 deletion does not appear to be a result of gross changes in H2A copy number in the cell or altered nucleosome positioning, although subtle differences would be difficult to detect in the assays presented, but could have a significant impact on activity – or pathway selection. Furthermore, there may be effects of cell cycle regulation that are not probed here, nor should they be. But these possibilities should be noted.
We agree that subtle chromatin changes would not be detected, and have modified two sentences to address this possibility. First, in describing the MNase indirect end-labeling result, we have added that there is “no visible, major disruption to the chromatin structure”.
Second, when drawing our conclusion in the paragraph below, we have added “however, subtle differences in chromatin structure in the absence of H2A.1 that are not visible by these assays may still have an impact on repair pathway selection.”
Previous work has shown that gene dosage of HTA2-HTB2 can amplfy to form a minichromosome when HTA1-HTB1 is absent (Libuda and Winston, 2006). We tested whether gene amplification of HTA1 or HTA2 had occurred in any of our strains and found no instances of gene amplification at the genomic level in the strains used in this paper (See new Supplementary file 8).
For cell cycle effects, we have added an experiment to show that hta1 deletion cells have a normal DNA damage checkpoint activation and resolution (by Rad53 phosphorylation). We have noted the possibility that another aspect of cell cycle regulation could be affected.
2) Unlike H2A-S129, the authors demonstrate that H2A-T126ph is not induced by the DNA damaging agents MMS and HU. As a result, the authors suggest that this modified form of H2A is normally present in the chromatin. Is there any evidence for preferential localization of H2A- T126ph at the tracts tested in this study or is it distributed across the genome?
We performed a ChIP experiment to test the association of H2A.1-T126ph (as detected by our antibody) with the CAG repeat, and detected a specific association at 40 min into S phase.
Previous ChIP at this same CAG location showed association of Mre11 at 20 min and γH2AX association also at 20 min and peaking at 40 min., then disappearing. Therefore, our data are consistent with H2A.1-T126ph being associated with replication-associated DNA damage at the repeat.
The antibody we used was purified from chicken eggs, and was quite dilute. Attempts to concentrate it resulted in a significant loss. We found we needed to use about 20 μl per sample for the ChIP to be successful, which amounted to 80-100 μl per cell cycle experiment. As noted in the response to the Editor’s comments, we were running low on antibody, and ordered more, but both new preparations were not specific. Therefore, we are unable to perform any more ChIP experiments, as we want to save back some antibody for requests from colleagues (the dilute preparation works fine for Western blots). We would love to do a genome-wide experiment, but it is not possible at this time.
3) The focus throughout the manuscript is on tract expansions, although contractions were also examined. Notably, while hta1 deletion influences both expansions and contraction, some of the other genetic backgrounds reveal differences in the impact on expansions versus contractions. Can this tell us anything about the mechanisms that result in tract instability – expansion versus contraction? Why does knocking out NHEJ in (lif1 deletion) or PRR (rad5 deletion) in combination with hta1 deletion lead to increased expansions with no detectable change in contraction frequency?
Our lab and others have examined the effect of various repair pathways on CAG repeat expansions and contractions, and there is an extensive literature on this topic. To simplify, gap- filling type repair or HR/PRR involving DNA synthesis tends to lead to expansions (due to slippage or improper flap processing), whereas DSB repair through NHEJ or SSA often leads to contractions. Our data support a model in which H2A.1 is facilitating a step of HR, like D-loop extension, to occur with fidelity. Since this is a synthesis step, expansions can result if it occurs inappropriately (as in the hta1 deletion or hta1-T126A mutant). Contractions could occur during this same process if the template strand forms a structure, or during repair of occasional breaks.
To address your specific question, knocking out PRR (rad5 deletion) on its own mildly increases instability (both expansions p=0.02 and contractions p=0.03). There is no further increase in combination with hta1 delete (hta1rad5), in fact the two mutations are less than additive and are similar to the rad5 levels, suggesting that Rad5 may work upstream of Hta1. We have now mentioned this in the presentation of Figure 4A and 4B.
Deleting lif1 in the hta1 background does suppress the elevated contractions observed in the hta1 background, suggesting that those contractions are occurring through NHEJ of a DSB. This result suggests that the somewhat increased breaks that occur within the CAG repeat in the hta1 background can be healed by NHEJ to cause contractions. We have not focused on this result, as it fits already known and published mechanisms of contractions and is likely not reflecting a specific function of Hta1 in the cell (note that neither fragility nor contractions were increased in the hta1-T126A mutant, in agreement with this interpretation; this data is now included in Figure 1—figure supplement 1B and Supplementary file 2).
We focused on the expansion phenotype in the absence of H2A.1 since it is expansions that are increased in the T126A mutant. We have added the following sentence to the manuscript to more overtly state this:
“Although contraction frequency is increased in the hta1Δ mutant, the frequency is not significantly increased from wild-type in the hta1-T126A or hta1Δ::HTA2 mutant; therefore we conclude that the H2A.1 sequence-specific role is in preventing expansions.”
4) Similarly, the hta126 point mutations (T126A and T126E) both increase expansions but contractions are unaffected, unlike the hta1 deletion, which affects both expansions and contractions. The authors suggest that it is dynamic phosphorylation of this threonine residue that is important. Is this a result of needing dynamic phosphorylation within a single histone core or simply the need to have pools of phosphorylated and unphosphorylated H2A.1 in the cell? Is either the hta1-T126A or hta1-T126E dominant negative? What is the effect on tract stability (expansions and contractions) of co-expressing both mutations?
This question of whether dynamic phosphorylation is occurring on a single histone core is an interesting one, but very difficult to address with the tools at hand. We thought about co- expressing both mutations (T126A, T126E) but since they are the same residue we would either have to introduce 2 copies of H2A.1 in a haploid cell (an unnatural situation with unknown consequences) or do the experiment in a diploid cell, necessitating redoing all the controls and single mutants in this situation. Also, since each single (T126A and T126E) gave a similar phenotype (Figure 3A), it might be hard to distinguish an effect of co-expression. However, we did test the effect of a T126A S129A double mutant on CAG expansions, which did not show any synergism or suppression (now in Figure 3A). The hta1-T126A point mutation does not appear to be dominant negative since the presence or absence of the Hta2 protein did not affect expansions (Supplementary file 2) or DNA damage sensitivity (Figure 1-figure supplement 2).
5) How does the suppression of unequal sister chromatid recombination in the absence of hta1 correlate with a decrease in higher fidelity repair at these fragile sites in the absence of hta1? That is, suppression of unequal SCR would seem to indicate that there is less lower fidelity repair. Would a decrease in misaligned recombination not be expected to stabilize the tracts? I understand that the assay is meant to measure spontaneous recombination between sister chromatids, but it does rely on misalignment.
It is true that one type of infidelity could be the misalignment required for an SCR readout in this assay. A second type of infidelity could occur during the SCR process itself after it initiates, for example during D-loop synthesis or resolution. Since we observe a decrease in SCR in the hta1 mutant but an increase in CAG expansions that are HR dependent, we conclude that the defect in efficiency of SCR must occur after the initiation step, such as during the D-loop synthesis (as a decrease in misaligned recombination would be expected to stabilize the tracts, as you point out). We have now added a sentence to this effect. Our data showing that BIR, a process with extensive D-loop synthesis, is also decreased in the hta1 mutant supports this interpretation.
To further address this point, we have added an SSA assay (Figure 4E), and find no defect in an hta1 mutant. This supports that H2A.1 is not affecting the alignment step of recombination, and this point is now made in the paragraph presenting the SSA data.
1) There seems to be some language missing in the very first sentence of the Abstract.
Added the three missing words – Repetitive and structured.
2) In the fourth paragraph of the Introduction, it would be useful to refer to the tracts as CAG/CTG or CAG:CTG tracts, or some variation of this, because the nucleosome positioning cited was specifically in CTG repeats (which obviously have CAG on the opposite strand).
This has been fixed throughout the Introduction.
3) Page 5 – 4th line from the bottom of the second to last paragraph – it seems that this refers Supplementary file 2.
4) It might be worth discussing the choice of statistical tests and how conservative they might be, particularly given the discussion in a couple of places of results that are interesting but not statistically significant, e.g Figure 4B,D, Figure 5B.
We have added a few sentences describing that Fisher’s is a conservative test and have more explicitly pointed out when trends are interesting but do not reach the level of statistical significance with Fisher’s.
5) There is some discussion of Western blots used to explore phosphorylation within H2A.1 (second paragraph from the bottom). The authors suggest that the reduced signal is a result of reduced accessibility of the antibody, whereas the more straightforward interpretation is that there is reduced phosphorylation, which is the interpretation discussed in the Discussion.
We have now mentioned both possibilities in both the Results and Discussion (this is in regards to reduced antibody recognition of the hta1-T126A and hta1-S129A mutants). We favor that the decreased signal is due to antibody recognition rather than T126ph being dependent on S129ph for three reasons:
1) CAG expansion frequency is elevated in the T126A mutant but not the S129A mutant. If T126ph depended on S129ph, then we would expect the hta1-T126A and hta1-S129A mutants to show equivalent CAG expansion phenotypes. As this is not the case it is unlikely that T126 requires phosphorylation at the S129 residue.
2) S129ph is induced by MMS but T126ph is not. If T126ph was dependent on S129ph, it should also depend on induction of MMS damage.
3) The H2A.1T126ph antibody was raised against a peptide that contained an unmodified serine at position 129. Although the antibody can recognize T126ph in the presence of S129ph (+MMS, Figure 3C), it appears that alanine at position 129 impedes antibody recognition. The signal in the S129A mutant is equivalent to both the T126A and hta1-del mutants, in-line with background reactivity of the antibody.
These interpretations have been expanded upon and clarified in the Results section. We further expand on the interpretation in the Discussion.
In this manuscript, the authors present compelling evidence for a difference between the two different histone H2A proteins in S. cerevisiae with respect to the stability of CAG repeats. A deletion of HTA1 confers instability while a deletion of HTA2 does not.
Furthermore, they provide additional evidence that this is not due to a change in histone levels but rather to the T126 present in HTA1 that is not present in HTA2. Finally, the provide evidence, confirming previous studies, that T126 is phosphorylated. Additional studies examine the instability phenotype when combined with other mutations that affect stability. Overall the results are interesting as they provide the first new information about a qualitative difference between these two H2A proteins since a paper by Norris and Osley in 1987.
1) The main issue with these studies, which is difficult to address, is whether it is the phosphorylation of T126, or some other aspect of T126 that is responsible for the phenotype. There are no results that distinguish between these possibilities, although the latter is not acknowledged and it should be. Both T126A and T126E changes both cause the phenotype, which the authors interpret as a requirement for dynamic phosphorylation, but which could also be interpreted as loss of the T is responsible for the phenotype. In addition, the authors are unable to detect a change in the level of T126 phosphorylation in response to several DNA damaging agents and in several mutant backgrounds. It might be informative to try a different polar amino acid, such as S. If that had a wild-type phenotype, that would argue against the idea that phosphorylation is important.
We concede that some other property of the threonine at that position in the H2A C-terminal tail could contribute to DNA repair and have now explicitly stated this possibility: “Although our results implicate a requirement for dynamic modification of H2A.1-T126 in DNA repair, some other physical property of threonine at position 126 may be important for repair, rather than phosphosphorylation per se.” see Discussion.
Our Western quantifications show that the antibody recognizes H2A.1 over H2A.2 with ~70% specificity, however the new phosphatase experiment showed that the specificity for the phosphorylated form in vivois lower, only 30%. Therefore, we altered our language to be more conservative with regards to conclusions related to phosphorylation of T126. Nonetheless, the presence of H2A.1 at the CAG tract by ChIP at a specific time during S phase coincident with S129ph, does support that it could be the phosphorylated form that is important for the observed phenotypes of the hta1-T126A mutant. Serine would be the closest polar amino acid to substitute for threonine, but we were worried about the interpretation of a T126 to serine mutation, as serine can also be phosphorylated.
2) Much more information should be provided about the screen. What strains were screened – was it something broad like the deletion set, or was it more focused? Please provide details in Results and/or Materials and methods.
This screen was originally published in Gellon et al., 2011 and hta1Δ is listed as a screen positive in the supplement of that paper. We have added that the Stanford deletion set was screened, but reference Gellon 2011 in lieu of adding the methods and results from the screen.
3) Figure S1 – This figure should be included in the main figures in the text as it provides information about the screen and it shows the initial data about the nature of the histone mutations that increase the rate of 5FOA resistance.
The fragility figure has been added to Figure 1.
4) The TRT nomenclature has not been used for many years. I’m don’t see the point of using it here as it might be confusing to those who are not familiar with it. Please just use HTA1-HTB1 and HTA2-HTB2.
This has been changed.
5) The authors point out that there is an amplification of HTA2-HTB2 when there’s a deletion of HTA1-HTB1. Therefore, in the strains used in this paper that contain an hta1 deletion, what has happened to HTA2? Has this duplication occurred? Also, they write that “the H2A.2 gene is not upregulated…” but if this amplification occurs, doesn’t that upregulate expression? In addition, they should use HTA2, not H2A.2 when referring to the gene.
We have clarified that “upregulation” refers to transcriptional upregulation in the text. We have now tested by qPCR for HTA gene levels in the hta1Δ and hta2Δ mutants and found no evidence of gene duplication of HTA2 in any of our hta1Δ mutant strains. These results have been added to supplemental Supplementary file 8.
6) Figure 3 – In the S129A mutant, the level of signal for T126ph is reduced to background – that seen in the T126A mutant. The authors interpret this to mean that the S129A mutant binds the anti-T126ph antisera less well. However, it’s also possible that S129 is required for phosphorylation of T126. The authors should include this possibility.
We have noted this possibility in the Results section. However, we favor that antibody recognition is responsible for the decreased signal for three reasons:
1) CAG expansion frequency is elevated in the T126A mutant but not the S129A mutant. If T126ph depended on S129ph, then we would expect the hta1-T126A and hta1-S129A mutants to show equivalent CAG expansion phenotypes. As this is not the case it is unlikely that T126 requires phosphorylation at the S129 residue.
2) S129ph is induced by MMS but T126ph is not. If T126ph was dependent on S129ph, it should also depend on induction of MMS damage.
3) The H2A.1T126ph antibody was raised against a peptide that contained an unmodified serine at position 129. Although the antibody can recognize T126ph in the presence of S129ph (+MMS, Figure 3C), it appears that alanine at position 129 impedes antibody recognition. The signal in the S129A mutant is equivalent to both the T126A and hta1-del mutants, in-line with background reactivity of the antibody.
These interpretations have been expanded upon and clarified in the Results section. We further expand on the interpretation in the Discussion. We agree that T126ph could be dependent on the presence of a serine 3 amino acids away (but not its phosphorylation), explaining the H2A.1-specific phenotypes, and have suggested this possibility in the Discussion.
7) The authors justification for testing a rad5 mutant is to test if chromatin modifications might make a difference in the hta1 mutant phenotype. This requires change or more justification – how does rad5 relate to chromatin modifications?
Our previous results (House et al., 2014) demonstrated that Rad5-mediated repair at CAG repeats was influenced by chromatin modifications (H4K16ac), which was our original justification/interest in performing those experiments. However, given the reviewer’s concern, we have changed the justification to “Rad5-dependent post-replication repair has previously been shown to be a source of expansions during low-fidelity repair at CAG repeats.”
8) Have the authors tested the level of Pol32 protein in the various hta1 and htb1 mutants to see if it is altered? That could account for some of the phenotypes.
We tested POL32 mRNA levels by RT-qPCR in both WT and hta1-del strains and found no difference (see Supplementary file 10).
The manuscript by House et al. reports a role for H2A.1 threonine 126 phosphorylation in triplet repeat maintenance in yeast. Using mainly genetic approaches, H2A.1 is shown to be required for CAG stability while the other copy of H2A, H2A.2, is dispensable.
Mutation of Thr126 in H2A.1, which is not found in H2A.2, renders yeast cells CAG unstable. The instability of CAG repeats was further shown to require HR and Pol32, which appeared to function through sister chromatid recombination. BIR was also shown to be defective in H2A.1 threonine 126 mutant cells. Overall, this is a well-executed study that reports an interesting differential involvement of H2A genes in yeast that appears to act through the phosphorylation of H2A.1 Thr126. This work is interesting from several angles including a new histone mark involved in CAG repeat stability as well as the suggesting that yeast may contain histone H2A variants within the H2A1 and H2A2 that are normally considered to be core H2A histones. While this work will be of broad interest to researchers in several biological sciences, which make this work a strong candidate for publication, some additional work should be performed to address the following questions.
1) The major issue with this work is that there is no mechanism for how T126p promotes CAG stability. While the publication of this work should not rely on this, some additional experiments/discussion should be added to better place this modification within this pathway. For instance, is this a Mec1/Tel1-dependent event and if so, are these kinases required for CAG stability? Does this modification occur at CAG sites? This is a very important question to support the model as any histone modification could act indirectly. To place this modification at the CAG repeats would go a long way to support the model proposed in Figure 6A. Along the same line of enquiry, are the levels of H2A.1 and H2A.2 the same at CAG repeats or are their differential loading of these histones which would also promote the CAG stability by having phosphorylatable H2A histones in proximity to these repeats.
We performed the requested ChIP experiment and were able to demonstrate that H2A.1 is specifically enriched at the CAG repeat 40 minutes after α-factor release, which is the time at which S129ph at the repeat is maximal (House et al., 2014). This result suggests that H2A.1 is specifically incorporated at replication-associated damage at the expanded repeat tract. Since we do not have an antibody specific to H2A.2, we could not track level of this histone variant at the CAG tract, but we did confirm that H2B and H3 levels are unchanged (Figure 3D, 3E).
We tested the effect of multiple kinases known to phosphorylate in response to DNA damage including Mec1, Tel1, Rad53 (and Rad9), Chk1 and Dun1. None of them showed a significant reduction in H2A.1-T126 antibody signal (Figure 3—figure supplement 1E). However, our new phosphatase experiment data indicate that our antibody shows only a mild preference for the phosphorylated form of the protein, therefore this approach to finding the relevant kinase is not ideal. Since there are over 100 kinases in the yeast genome, identification of the kinase will require a different approach, beyond the scope of this study.
2) The potential crosstalk between S129 and T126 is intriguing. While the Results section proposes that the reduction in T126p by mutation of S129 is perhaps an artifact of the antibody, the discussion suggests that this may be a real result. This is somewhat confusing for the reader. Some additional work should be performed to address this question. Either antibodies can be used to check the specificity of the T126p epitopes or genetic analyses could be performed to see if there are interactions between the loss of T126 and S129. While the genetics seem to support a separation of function between these two modification sites for CAG repeat stability, additional work would help strengthen this idea which is an important one and an issue that is currently unresolved in this manuscript.
The genetics don’t support that T126ph is dependent on S129ph, as we saw no increase in expansions for a S129A mutant. Also, we have now tested CAG expansion frequencies in the S129A T126A double mutant, and see no increase over the T126A level. Therefore, the genetic data do not support a dependence of T126ph on S129ph. We now conclude more clearly that the decrease in antibody recognition due to the S129A mutation is likely due to antibody epitope recognition, rather than a co-dependency of the two modifications. This can be found in the Results section and expanded upon in the Discussion.
3) It would be interesting to perform DNA damage sensitivity assays for H2A.1, H2A.2 and the phospho site mutants to further demonstrate the importance of these different H2A genes and these sites in DNA damage response pathways.
These assays have been added to the supplement – we did not see significant sensitivity of the hta1Δ mutant to the agents tested, although it is perhaps mildly sensitive to camptothecin and phleomycin.
4) Figure 3B should be ran on the same gel. Levels of phosphorylation between different samples are difficult to compare when ran on different gels. Given that these are only 6 samples, there should not be any issues in performing this experiment.
The samples were run on the same gel, but the bar was used to delineate the different treatments. We have removed the bar to address the reviewer’s concern (now Figure 3C). We have also added quantification of this and duplicate experiments.
5) The data in Figure 5E is interesting. For the hta1 mutant, the repair types are only slightly reduced for BIR, especially compared to pol32 mutants. However, this mutant seems to be the only one where GCs are scored. This seems interesting and deserving of a discussion to explain why this may be. Although not required, it would have been nice to have the T126A mutant analyzed here to ensure that all of these phenotypes for hta1 are occurring through this phospho-site.
GCs were scored in each mutant, however the values are small enough to not be clearly visible in the WT. For pol32Δ, this was a mistake which has now been corrected: GCs (maybe due to aborted BIR) are also increased in this mutant (to 9.8%, Supplementary file 7). We think that the increase in alternative repair types in both the hta1Δ and pol32Δ mutants is indicative of the BIR defect, which we have now clarified in the text. It does seem that the proportion of GC events compared to NHEJ is higher for hta1Δ compared to pol32Δ which could indicate that D-loop synthesis is relatively less impaired in hta1Δ, but we were hesitant to make a point about this since the reason is speculative. We also tested BIR in the hta1-T126A mutant and found no significant defect. All raw data for the BIR assays can now be found in Supplementary file 7.
1) The authors should consider editing the first sentence of the Abstract, “DNA are sites of genomic instability”. This sentence distracted this reader from the beginning.
The first three missing words have been replaced – Repetitive and structured.
2) In the Introduction, second paragraph, H2A/H2AX K15 Ub by RNF168 which is bound by 53BP1, as well as TIP60-mediated H2A/H2AX K15 acetylation should be added to the discussion of modifications on H2A histones.
As originally written, this was listed as modifications that have been directly shown to localize to sites of DNA damage, either by ChIP, foci IF, or microirradition. To include K15ub/ac as requested, we broadened the language from “detectable at breaks” to “contribute to DNA repair.”
3) The model in Figure 6B is confusing. It is labelled No H2A.1 but the figure shows this histone without the phosphorylation. I think the authors would like to say “No H2A.1 T126 phosphorylation”. This should be edited for clarity.
We have changed the labeling on the model figure to emphasize the specific role for H2A.1, and changed the color of the histones and labeling in part B to indicate the situation where only H2A.2 (not phosphorylated in our model) is present.
[Editors’ note: the author responses to the re-review process follows.]
The major concern remains that of the antiserum used. It is clear now, from the work now done in characterizing the antiserum, that this is a reagent with complex specificity; in addition to the substantial reactivity remaining after phosphatase treatment, specific reactivity appears to require both T126 and S129, with an indication that phosphorylation of S129 may reduce reactivity. Given the complex nature of the reactivity of this antiserum, it seems clear that experiments using it cannot be interpreted unambiguously.
The antibody shows on average 70% specificity for H2A.1 compared to H2A.2. Thus, it is reasonably specific for H2A copy 1. This level of specificity is in line with other histone tail antibodies that have been used in the literature. For example, The Abcam H3K27ac antibody (ab4729) is extensively used in the literature (it has 723 references listed on the product information page). By peptide array, this antibody is <30% cross-reactive with unmodified H3, <20% cross-reactive with K9ac, <20% cross-reactive with K36ac, and <20% cross-reactive with K18ac. It is still acceptably used to demonstrate H3K27ac.
Therefore, although our H2A.1-T126 antibody is not perfect, we feel it is still useful for detecting differences between the two isoforms of H2A, and we have made conclusions accordingly, with appropriate reference to the background isoform cross-reactivity.
Regarding the “complex specificity”, the antibody was raised against the T126ph form of the H2A.1 tail, so it is natural and expected that it would recognize the amino acid sequence of the tail (e.g. threonine at position 126 and serine at position 129). The fact that the antibody recognizes the sequence it was raised against best (T126ph and S129 not ph) is also not surprising, and we can interpret experiments using it appropriately since we did all the proper control experiments to characterize the reactivity under each condition. Nonetheless, in this version, we have focused on its ability to distinguish between the two H2A isoforms, rather than the state of T126 or S129 phosphorylation.
The antibody is less specific for the phosphorylated form of the H2A.1 tail. Although it shows high specificity in vitro, it has only a 30% preference for the phosphorylated form according to our in vivophosphatase experiments. Therefore, we have removed all conclusions relating to specificity to the phosphorylated form and either moved that data to the supplement or qualified our conclusions based on the recognition of both phosphorylated and non-phosphorylated forms. However, based on our experiments showing a phenotype for the hta1-T126A mutation in vivoin multiple assays, and previous published results that it is phosphorylated in vivo, we still speculate in the discussion that phosphorylation of HTA.1-T126 may be important for fidelity of sister chromatid recombinational repair, but include the caveat that this could be due to the polarity of the threonine residue and not necessarily T126ph. In summary, we have modified our language throughout to only conclude that H2A.1 (copy 1) and a threonine at position 126 is important for our various phenotypes but have left in speculation that this effect could be due to T126 phosphorylation.
Given that these experiments remain a major fraction of the manuscript, I do not think that it would be appropriate to proceed further with the current submission.
The experiments using the antibody are not a major fraction of the paper: they are 3 experiments that constitute part of 1 figure out of 5 figures (3 experimental panels out of 17). Specifically, in this revision, we have focused on the experiments using the antibody to distinguish between the two H2A forms, H2A.1 and H2A.2: Figure 3B shows the specificity for H2A.1. Figure 3 C and D shows that H2A.1 specifically locates to the CAG tract at the same time as H2A-S129ph. Both reviewer 2 and 3 commented that this was an important addition to the paper “The new data showing that H2A.1 is enriched at triplet repeats is an important addition, further demonstrating a direct role for this histone at repeat sequences.” (reviewer 3). Figure 3E shows that the T126A phenotypes are not due to an indirect effect on S129 phosphorylation, which is an important control. These three experiments are focused on the H2A.1 and hta1-T126A phenotypes and therefore support the main points of the paper. We also note that all of our genetic and other data in Figures 1, 2, 3A, 4, and 5 do not depend on this data, so even if it were discounted the story is strong and stands on its own. We disagree that the data obtained using the H2A.1-T126 antibody is a major fraction of the manuscript, but nonetheless, as described above, argue that it adds important information that complements our genetic data.
The reviews are copied below and we hope that they may help you should you decide to revise the manuscript for submission elsewhere.
We are sorry that we cannot be more positive on this occasion, but hope that you appreciate the reasons for this decision.
Comments to the Authors:
This revised manuscript effectively deals with most of the earlier concerns. The apparent role of H2A.1 in D-loop-mediated recombination is very interesting. However, the issue of T126 phosphorylation hasn't been adequately addressed and this limits the mechanistic understanding of the data.
The results for H2A.1 phosphorylation using the T126 antibody are confusing as presented. Basically, the authors have an antibody specific to H2A.1, with some preference for a phosphorylated protein, but not enough to make any real conclusions about the role of T126 (phosphorylation vs. structural). It seems, therefore, that this section should be reduced and re-framed in that context. Perhaps the order of the results in the new section dealing with the antibody specificity could be re-arranged. Perhaps going through the specificity of the antibody for H2A.1 versus H2A.2 then the localization of H2A.1 at the CAG repeat, followed by looking at phosphospecificity (or lack thereof), which limits the interpretation of the role of T126.
We have reframed and restructured this section as suggested by the reviewer, first presenting the specificity to H2A.1 (now Figure 3B) and experiments that show H2A.1-specific phenotypes (regardless of phosphorylation state) (Figures 3C, 3D, 3E). We are upfront about the fact that the antibody was originally raised against the phT126 residue with the aim of testing that modification in our system, but then detail our testing that showed the specificity is greater for the H2A.1 C-tail versus the H2A.2 tail, though there is also a recognition of the T126 phosphorylated form of the H2A.1 tail. (Figure 3—figure supplement 1).
Accordingly, we have re-named the antibody H2A.1-T126 (instead of H2A.1-T126ph) and re-labeled the relevant figures.
Can you not take your histone preps and perform a mass spectrometry analysis to look at H2A.1 phosphorylation in different genetic contexts?
We considered this experiment but believe it will not be successful for the following reasons. There are 4 published papers that investigated the yeast phospho-proteome using mass spectrometry (Smolka et al., 2007; Chen SH et al., 2010, Bastos de Oliveira et al., 2015, and Zhou et al., 2016). The paper from the Elledge lab was particularly comprehensive, using SILAC technology coupled with mass spec identification and testing not only MMS, but HU in S phase cells, and irradiation of mitotically arrested cells. None of the papers detected H2A.1 T126 phosphorylation, though it was detected by 2D protein gels of labeled histones in Wyatt et al., 2003. I communicated with Steve Elledge about potential reasons for this. Because T126 is surrounded by 2 lysines (K124, K127), it will end up on a peptide only 3 amino acids long after trypsin digest, which will not be detected by mass spec (6 residues is the typical minimum peptide detected). We now mention this issue in the text. Steve Elledge said he didn’t think that this short peptide would be detected by SILAC labeling either. Because we don’t have another good way to verify T126ph besides what has already been published and our imperfect antibody, we have focused the revised manuscript on the different H2A subtypes.
I don't see how, as stated by the authors in the discussion, one can determine whether T126ph is dependent on S129ph.
We have clarified our language on this point. Our conclusion is that since the hta1-S129A mutant has no CAG expansion phenotype, and S129ph occurs unimpeded after MMS treatment (DNA damage) in the T126A mutant, the phenotype in the hta1-T126A mutant is not due to an indirect effect on the ability of S129 to be phosphorylated. Figure 3E is included to make this point.
Throughout the discussion there are places where T126ph is assumed to be the relevant form of the protein, including the very last sentence.
We have now modified the discussion to be more even-handed between the following possibilities:
1) T126 may be locally phosphorylated at the CAG repeat during repair and required for repair factor recruitment; 2) The amino acid sequence in the H2A C-terminal tail is required for DNA repair protein recognition (rather than phosphorylation of the residue, the reside itself is required), or 3) modification of T126 affects modification of other tail residues, such as S122ph, K124ac, or K127ac that are then required for efficient repair. We have also modified the last sentence to remove the word ‘phosphorylation’.
The authors argue that the reduced signal with S129A is a result of reduced recognition because of the mutation. Is this not also the case with T126A, then (e.g Figure 3B)?
Yes, that is the case. We conclude that the antibody recognizes the sequence of amino acids in the H2A.1 C-terminal tail, including T126 and S129. Therefore, upon mutation of the T126 residue, the antibody recognition is significantly diminished. As stated, “The signal is significantly diminished in the hta1Δ and hta1-T126A strains to 20-30% of WT levels, but not in the hta2Δ mutant (Figure 3B. Supplementary file 9). We conclude that the antibody is specifically recognizing the H2A.1 protein isoform containing threonine at position 126 in the tail, with a low level of background reactivity to the H2A.2 isoform.” And “we conclude that the hta1-S129A mutation likely disrupts the antibody epitope rather than affecting H2A.1-T126 modification.”
Throughout, some more care is required in describing assays and conditions used for the more general reader. For example, why is the MMS treatment relevant in testing the phospho-specificity of the T126 antibody?
We have added additional explanation to the text describing assays and conditions. MMS is used in the phosphatase experiment in Figure 3—figure supplement 1B to induce S129ph as a positive control, now stated in that legend. It was used in Figure 1-figure supplement 2 spot assays to induce DNA damage, now stated in that legend. It was used in Figure 1-figure supplement and 2 Westerns as a DNA damaging agent to induce S129ph and test whether H2A.1 expression or T126ph was similarly induced, now stated “To determine if H2A.1 expression is damage inducible, H2A.1 levels were monitored after exposure to MMS, a DNA base alkylating agent that causes abasic sites that can be converted into single and double strand breaks.” An interpretation of the results was added a couple sentences later. Better explanations of both the SCR and SSA assays have also been added (see details below).
The new SSA assay cartoon figure is poorly labeled so that the intermediate formed is not clear. How does misalignment lead to gene conversion in the SCR assay?
The SSA figure has been updated to be more clear and more details were added to the figure and legend.
For SCR assay clarification, we have added the sentence (p9), “In this assay, misaligned recombination between two ade2 null alleles can result in gene conversion to a functional ADE2 allele and the strain is converted from Trp+Ade- to Trp+Ade+” and directed the reader to the Mozlin, Fung and Symington, 2008 publication that extensively details this assay for more details. We have also added “X’s” to the diagram to indicate the illustrated cross-over that would yield Ade2+ colonies, and more fully explained the events shown in the legend.
The revised manuscript has been extensively revised and is considerably stronger than the first version. The authors have done an excellent job of responding to all of the comments. I have just two very minor writing comments, listed below.
1) Since genes are deleted, not proteins, the sentence should read, “…deletion of the gene encoding histone H2A.1…” The same correction should be made a few lines below. And later in the manuscript for the lif1 experiment.
These corrections have been made. Changed to: “Deletion of the genes encoding the two copies of H2A differentially affect CAG repeat stability…” The other sentences were also corrected.
2) Change “…hybridization of the antibody…” to “…recognition by the antibody…” or something similar.
Changed to: Although this antibody was specific to H2A-T126ph by peptide dot blot (Figure 3—figure supplement 1B) phosphatase treatment of cell extracts only resulted in a 30% reduction in antibody recognition (Figure 3—figure supplement 1C, Supplementary file 9)
The revised manuscript by House et al. has provided substantial new data to support a role for H2A.1 in triplet stability and various HR pathways. Although a direct role for H2A.1 T126 phosphorylation has been dampened due to issues with the specificity of the antibody, this work still reports some very interesting findings about the differential role of the two H2A genes in yeast, suggesting that yeast in fact may have an H2AX like variant and a canonical H2A. The new data showing that H2A.1 is enriched at triplet repeats is an important addition, further demonstrating a direct role for this histone at repeat sequences. It is unclear why H2A.2 was not similarly ChIPed, perhaps in a H2A.1 mutant, but given the control to show that H3 and H2B levels are constant at these regions, it is reasonable to assume that this is specific for H2A.1. The additional data showing that BIR levels are normal in the H2A.1 T126 mutant compared to H2A.1 delete is also a nice addition, as it suggests additional residues on H2A.1, independent of T126, that are involved in BIR. Although there remains several unanswered questions here, overall, this study is now sufficiently revised to support publication as the authors have addressed all of this reviewer’s previous concerns.
1) Mammalian H2AX T126 mutants have been studied previously. In Xie et al., 2010 (PMID: 20703100), no sensitivity to IR or defects in HR were observed in H2AX KO ES cells reconstituted with H2AX T126V. It would have been perhaps appropriate to discuss this study in the conclusion for the speculation of translating these results with the yeast protein to mammalian H2AX.
A reference to this result and the citation was added and discussed w.r.t. our results in the last two sentences: “A T136V mutation did not decrease survival after IR or DSB-induced HR in mouse cells, but gap repair and repair fidelity were not analyzed (63). Thus, it would be interesting to test whether mammalian H2AX T136 plays a role in repair fidelity analogous to the role described here for yeast H2A.1-T126.”
Given the lack of phenotype in the BIR assay for H2A.1 T126, the model (in Figure 6A) is still confusing since doesn’t this show that T126 is dispensable for BIR, which the model doesn’t suggest.
For the purposes of the model figure, we had to decide whether to show T126 as phosphorylated or not. We chose to show it as we favor a model where phosphorylation has a role, even though it may not be relevant for all steps, but added this explanation in the legend “Though T126ph is shown as present during BIR/BFR for continuity with the previous diagram, our data suggest that the T126 identity has minimal effect on this step in the context of BIR (BFR was not tested).”
When discussing the step during recombination at which H2A.1 is working, we added this explanation of why T126 may be dispensable for BIR: “A second possibility is that H2A.1-T126 is required to promote a later step in the process such as D-loop resolution or re-establishment of the chromatin structure of the repaired gap. This could explain why the H2A.1-T126A mutant did not have a discernable effect on BIR, which does not include re-engagement of the extended D-loop with the initiating DNA molecule.” To better illustrate this point in the model, we removed the T126ph residue from the template strands.
We also added another possible explanation of the effect of hta1 deletion but not hta1-T126A on BIR in the description of the model: “We note that CAG fragility, CAG contractions, and BIR that all involve an initiating DSB were dependent on H2A.1 but not the T126 residue. Therefore, our data are consistent with H2A.1 levels affecting DSB repair while the H2A.1 sequence-specific role is during gap repair/SCR.”
Together, these changes clarify our model and should eliminate the confusion.https://doi.org/10.7554/eLife.53362.sa2
- Catherine H Freudenreich
- Erica J Polleys
The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.
The authors thank L Symington for the SCR assay system, R Anand and J Haber for the BIR assay system, J Haber for the SSA assay system, J Downs for the plasmid H2A point mutant strains, Ryan Hayman and David R Walt for help designing and running the customized Illumina BeadArray, and Danae Schulz for performing the intial screen that identified H2A.1 as a candidate for further study. This work was funded by a National Institutes of Health award to CHF (Project 4 of P01GM105473, James Haber, PI), and an American Cancer Society–Ellison Foundation Postdoctoral Fellowship PF-18-125-10-DMC to EJP.
- Kevin Struhl, Harvard Medical School, United States
- Wolf-Dietrich Heyer, University of California, Davis, United States
- Wolf-Dietrich Heyer, University of California, Davis, United States
© 2019, House et al.
This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.