Screening of chromatin regulatory factors interact with Sld3, Sld7, and Cdc45

To identify chromatin regulatory factors that interact with Sld3, Sld7, and Cdc45, we cloned 26 genes that were listed by the keyword “histone acetyltransferase” in the Saccharomyces Genome Database web site (https://www.yeastgenome.org/). Each of them were cloned into the vectors for the yeast two-hybrid (Y2H) assay, DNA-binding domain (DB) vector pST1667 and the activation domain (AD) vector pST548 (Tanaka 2019), respectively, and their capability of interaction with any of Sld3, Sld7, and Cdc45 was tested by the Y2H color assay (Bartel and Fields 1995).

Isolation of mutants with diminished interactions between Sld3 and Esa1

To construct the mutant library, SLD3 and ESA1 ORF were PCR-amplified with Taq polymerase (MightyAmp ver. 2; Takara Bio Co., Otsu, Japan) and cloned into the BamHI sites of the Y2H DB vector pST1667 and the Y2H AD vector pST548 (Tanaka 2019), respectively, using the In-Fusion enzyme (Takara Bio Co,). Mutant libraries of DB-Sld3 and AD-Esa1 were introduced into the Y2H host TAT-7 strain harboring AD-Esa1 and DB-Sld3, respectively. After 1 day of incubation, the original plates were replicated, and the incubation was continued. After colony formation, cells were lifted from the replicated plates and subjected to the Y2H color assay (Bartel and Fields 1995). “White” candidates were picked from the original plate and re-streaked for confirmation of the white phenotype. In parallel, whole-cell extracts were prepared from the re-streak and analyzed by western blotting. Candidates generating full-length or near-full-length protein were stored as secondary candidates. Plasmids were recovered from secondary candidates, and the white phenotype in the Y2H color assay was confirmed (tertiary candidates). From tertiary candidates, clones with endogenous promoters were constructed in YCp-type plasmids and clones that could support cell growth were selected by plasmid shuffling; the resultant mutants were denoted sld3^Esa1(−) and esa1^Sld3(−).

Construction of yeast strains

When a plasmid such as GALp-3/4/75 was integrated into yeast genome, integrated copy number of the plasmid was confirmed by monitoring the change in the length of targeted region by Southern blotting. To construct the strain harboring the sld3^Esa1(−)-3 or esa1^Sld3(−)-2 allele as the sole copy of the respective genes, the gene replacement method using URA3/5-FOA selection called the “pop-in/pop-out” method (Boeke et al. 1987) was employed. The alleles in the SLD3 or ESA1 locus of the resultant clone were confirmed by DNA sequencing.

Other methods

Flow cytometry was performed as described (Tanaka and Diffley 2002b). Replication intermediates were detected using alkaline gels as described (Tanaka et al. 2011). Preparation of yeast whole-cell extracts and western blotting were performed as described (Tanaka et al. 2011). Y2H assay was performed as described (Tanaka et al. 2013). The plasmid loss rate was measured as described (Tanaka and Diffley 2002a).

Anti-silencing effects depend on the replication function of 3–7–45

The Sld3–Sld7–Cdc45 complex functions in origin firing; specifically, it associates with the Mcm2–7 complex at the licensed origin in a DDK-dependent manner during the firing reaction (Heller et al. 2011; Tanaka et al. 2011). In a previous study, we observed that untimely activation of origins caused by 3–7–45 expression required the expression of all three genes; expression of any one or two of them did not induce untimely origin firing except in the case of simultaneous expression of Sld3 and Sld7, which triggered very weak activation of origins (Tanaka et al. 2011). The anti-silencing effect caused by 3–7–45 expression recapitulated this, i.e., the anti-silencing effect was observed only when all three genes were expressed; a very weak anti-silencing effect was observed when Sld3 and Sld7 were simultaneously expressed (Supplementary Figure S1C).

If the anti-silencing effect caused by 3–7–45 expression depends on its replication-related function, which involves the association of 3–7–45 with the DDK-phosphorylated Mcm2–7 complex at licensed origins, then the anti-silencing effect must depend on origin licensing. Replication origins are licensed from the end of M-phase until they are activated in S-phase. Therefore, if the duration of time when cells harbor licensed origins is changed, it would affect the anti-silencing effect of 3–7–45. To test this possibility, we used the Δswe1 and Δmih1 mutants. Because Swe1 and Mih1 regulate the timing of onset of M-phase via activation of M-phase CDK in opposite manners, and M-phase entry is accelerated in Δswe1 and delayed in Δmih1, the length of G1 phase is longer in Δswe1 than in the wild type, and shorter in Δmih1 (Russell et al. 1989; Booher et al. 1993). The anti-silencing effect of 3–7–45 was strengthened in Δswe1 cells and weakened in Δmih1 cells (Supplementary Figure S1D). On the basis of these observations, we concluded that the anti-silencing effect is a result of the origin association of 3–7–45.

Sld3 interacts with Esa1

The chromatin regulatory functions of 3–7–45 remain unknown. To begin to investigate this issue, we postulated that any of Sld3, Sld7, and Cdc45 interacts with chromatin regulatory factors. To identify such interactors, we performed systematic Y2H assays and found that Sld3 interacts with Esa1, the budding yeast ortholog of lysine acetyl transferase 5 (Kat5) and the catalytic subunit of NuA4 histone acetyl transferase (HAT) (Figure 2A and Supplementary Figure S2A) (Clarke et al. 1999). Esa1 also exhibited a weak interaction with Sld7 but did not interact with Cdc45 in this assay.

Open in a separate window

Figure 2

Sld3 interacts with Esa1. (A) Y2H analysis was performed with plasmids encoding the indicated proteins. (B) Δeaf3, Δeaf5, Δeaf6, Δeaf7, and Δyaf9 cells in the wild type (Wt) or sld3-6 background were spotted and grown under the indicated conditions. Strains used were W303-1a Δbar1, YST2327, YST2329, YST2330, YST2331, and YST2551 (Wt background, from top) and YYK16, YST3406, YST2313, YST2314, YST3407, and YST3408 (sld3-6 background, from top). (C) HMLα::URA3-harboring Wt, Δeaf3, Δeaf5, and Δeaf6 cells harboring the GALp vector or GALp-3–7–45 were spotted and grown under the indicated conditions. Strains used were YST1788, YST2345, YST2346, and YST2347 (GALp vector, from top) and YST1790, YST2351, YST2352, and YST2353 (GALp-3–7–45, from top).

NuA4 subunits interact genetically with Sld3

NuA4 HAT consists of 13 subunits including Esa1 (Allard et al. 1999). Although Esa1 is essential for cell growth, some subunits of NuA4 are not. To elucidate the functional relationship between Sld3 and NuA4, we deleted several of the non-essential subunits of NuA4, Δeaf3, Δeaf5, Δeaf6, Δeaf7, and Δyaf9 in cells harboring temperature-sensitive alleles of SLD3, sld3-5, and sld3-6 (Kamimura et al. 2001). Some of the mutant alleles exhibited a genetic interaction (Figure 2B and Supplementary Figure S2, B and C). Δeaf3, Δeaf5, and Δeaf6 exacerbated the sensitivity of the sld3-6 mutant to high temperature and HU (Figure 2B). By contrast, Δeaf7 and Δyaf9 exhibited an opposite effect in the same background (Figure 2B and Supplementary Figure S2B). No such effect was observed in another sld3-ts allele, sld3-5 (Supplementary Figure S2C). All of the deletion mutants weakly exacerbated the phenotype of sld3-5. This difference between sld3-5 and sld3-6 backgrounds might be due to differences in the properties of alleles. Collectively, the results of this experiment revealed a genetic interaction between NuA4 and SLD3.

Based on the existence of a genetic interaction between Sld3 and NuA4 subunits, we next asked whether Δeaf3, Δeaf5, and Δeaf6 affect the anti-silencing effect of 3–7–45 in the HMLα::URA3 cells, as these three deletions exhibited similar synthetic effects in both sld3-5 and sld3-6 backgrounds. Although Δeaf6 had no obvious effects, Δeaf3 and Δeaf5 weakened the anti-silencing effect (Figure 2C). This indicates that NuA4 somehow contributes to the anti-silencing caused by 3–7–45 expression.

Deletion of the NuA4 subunit Eaf5 affects the activity of origins in the silenced locus HMLα

To further elucidate the genetic interactions between Sld3 and NuA4 subunits, we asked whether NuA4 has an effect on the activities of dormant origins located in the silenced locus HMLα. To answer this question, we performed a plasmid loss assay with centromeric plasmids harboring the early replication origin ARS1, a short fragment derived from HMLα that only contains ARS301, or the whole region of HMLα plus ARS302–320 (hereafter denoted as ARS1, ARS301, and ARS301–320; for details, see Supplementary Figure S3, A and B). The ARS301-containing short fragment has characteristics of a late replication origin even in a plasmid context (Santocanale and Diffley 1996; Bousset and Diffley 1998), probably due to its chromatin structure. Therefore, we first investigated whether the HMLα-derived fragments were silenced on the plasmid. The URA3 reporter gene, which was inserted proximal to or between the HMLα-derived origins, was silenced in the ARS301 and ARS301–320 plasmids, but not in the ARS1 plasmid (Supplementary Figure S3C). We concluded that HMLα-derived fragments are silenced in the plasmid context, as in the chromosome context.

ARS301 and ARS301–320 plasmids were lost at higher rates than the ARS1 plasmid (Figure 3A), suggesting poorer DNA replication of these origins. Importantly, this high plasmid loss rate was suppressed by the deletion of SIR2, which encodes silencing factor (Rine and Herskowitz 1987), Δsir2 (Figure 3A), implying that the higher loss rate of ARS301 and ARS301–320 plasmids was related to silencing. Deletion of the NuA4 subunit Eaf5 increased the loss rate of ARS301 and ARS301–320 plasmids, but not that of the ARS1 plasmid (Figure 3A), indicating that the deletion interfered with the replication of these plasmids. Conversely, 3–7–45 expression suppressed the high loss rate of ARS301 and ARS301–320 plasmids (Figure 3B). Finally, the combination of Δeaf5 and 3–7–45 expression resulted in an intermediate loss rate for the ARS301 and ARS301–320 plasmids (Figure 3C). Therefore, we concluded that the NuA4 complex, at least Eaf5, positively affects not only the anti-silencing effect of 3–7–45 but also the activation of dormant origins in HMLα.

Open in a separate window

Figure 3

Plasmids harboring HMLα origins replicate more efficiently when 3–7–45 is expressed, and less efficiently in the Δeaf5 background. (A) Plasmid loss rates were measured using plasmids harboring the indicated ARSs (ARS1: pST1969; ARS301: pST1970; and ARS301–320: pST1971; see Supplementary Figure S3, A and D, for details). Strains used were W303-1a Δbar1 (Wt), YST1930 (Δsir2), and YST2329 (Δeaf5). (B) Plasmid loss assay was performed with host strains harboring either GALp vector (YST1038) or GALp-3–7–45 (YST1054). (C) Plasmid loss assay was performed with Δeaf5 strains harboring either GALp vector (YST2482) or GALp-3–7–45 (YST2489). For each experiment, the average of four independent results is shown. Error bars: SD. Statistical significance in the differences was tested by the Student's t-test. **P < 0.01, *0.01 < P < 0.05, and n.s. (not significant): P > 0.05. Actual P-values are also shown.

The Sld3–Esa1 interaction is important for the anti-silencing effects of 3–7–45 and the activation of replication origins in the silenced locus HMLα

Experiments with Δeaf5 suggested a positive role for NuA4 in the anti-silencing effect of 3–7–45 and the activation of dormant origins in the silenced locus HMLα. However, deletions of non-essential subunits revealed a complex genetic interaction with sld3-6 (Figure 2B and Supplementary Figure S2B), and no interaction between Eaf5 and Sld3 could be detected in the Y2H assay (Supplementary Figure S4A). Because an Esa1–Sld3 interaction was identified in the same assay (Figure 2 and Supplementary Figure S4A), we investigated the role of that interaction, rather than the NuA4 complex itself, in the role of 3–7–45 in anti-silencing and DNA replication of dormant origins in the silenced locus. For this purpose, we isolated Sld3 mutants with diminished interaction with Esa1. Because Sld3 is essential for growth, we sought to isolate viable mutants for further analysis. Ultimately, we obtained three such mutants, sld3^Esa1(−)-1, -2, and -3 (Materials and Methods, Figure 4A). These three Sld3^Esa1(−) mutants exhibited a significant reduction in the Esa1 interaction in the Y2H assay; in particular, Sld3^Esa1(−)-3 seemed to lose the interaction completely. Importantly, all mutants maintained interactions with other replication factors such as Sld7, Cdc45, Dpb11, Orc2, and Mcm6 at levels similar to those in the wild type (Figure 4A), indicating that they only diminished or lost the ability to interact with Esa1.

Open in a separate window

Figure 4

Mutants that diminish the Sld3–Esa1 interaction exhibit weaker anti-silencing effects and higher plasmid loss of HMLα origins. (A) Isolation of Sld3 mutants with diminished interactions with Esa1. Y2H analysis was performed with the plasmids encoding the indicated proteins. (B) HMLα::URA3 cells expressing Sld3 wild-type or Sld3^Esa1(−) mutant proteins with Sld7 and Cdc45 were serially diluted, spotted, and grown on the indicated plates. Strains used were YST1862 (Δsir2), YST1861 (Δsir1), UCC3515 (−), YST2827 (+GALp vector), YST2828 (Sld3 (wt)), YST2832 (Sld3^Esa1(−)-2), YST2834 (Sld3^Esa1(−)-1), and YST2831 (Sld3^Esa1(−)-3). (C) Plasmid loss assay was performed in the sld3^Esa1(−)-3 (YST2870) and esa1^Sld3(−)-3 (YST2741) strains (Wt: YST2843). The average of three independent results is shown. Error bars: SD. Statistical significance in the differences was tested by the Student's t-test. **P < 0.01, *0.01 < P < 0.05, and n.s. (not significant): P > 0.05. Actual P-values are also shown.

Mapping of mutated amino acids in Sld3^Esa1(−) mutants revealed that all three mutants have amino acid substitutions in the Sld3/Treslin homology domain, which is located in the middle of the Sld3 protein and conserved in Sld3 orthologs (Sanchez-Pulido et al. 2010) (Supplementary Figure S4B). Moreover, both Sld3^Esa1(−)-1 and -3 harbor an extra amino acid substitution in the C-terminal region, but these mutations seemed not to affect the Sld3^Esa1(−) phenotype, as mutations in the Sld3/Treslin homology domain (N178F and N189Y in the Sld3^Esa1(−)-1 and P157T, T193S, and V264L in the Sld3^Esa1(−)-3) were sufficient for a diminished interaction with Esa1, and a single amino acid substitution in the C-terminal region alone (S459L in the Sld3^Esa1(−)-1 and P623Q in the Sld3^Esa1(−)-3) did not affect the Sld3–Esa1 interaction (data not shown). These findings imply that the Sld3/Treslin homology domain is important for the Sld3–Esa1 interaction.

To evaluate the importance of the Sld3–Esa1 interaction, we expressed Sld3^Esa1(−) simultaneously with Sld7 and Cdc45; we observed that the anti-silencing effect was weaker in HMLα::URA3 cells expressing Sld3^Esa1(−) than in those expressing wild-type Sld3 (Figure 4B). This effect was roughly proportional to the decrease in the Y2H signal, as this effect was stronger with the Sld3^Esa1(−)-1 and -3 than that with Sld3^Esa1(−)-2, which still shows very weak interaction with Esa1 in the Y2H assay (Figure 4, A and B). These observations indicate that the Sld3–Esa1 interaction is important for the anti-silencing effect caused by 3–7–45 expression. To observe the effect on the activity of dormant origins in HMLα, we performed a plasmid loss assay. The loss rate of both the ARS301 and ARS301–320 plasmids was higher in the sld3^Esa1(−)-3 background than in the wild type, although the loss rate of the ARS1 plasmid was not changed.

We also isolated two viable Esa1 mutants with diminished interactions with Sld3, esa1^Sld3(−)-1 and -2 (Supplementary Figure S4C). Both mutants had multiple amino acid substitutions in the middle of the Esa1 protein, and at least one substitution was located in the MYST-type HAT domain. Of the 445 amino acids in Esa1, the N-terminal portion (aa 1–146) was dispensable for the Sld3 interaction, and most of the remaining part of the protein (aa 147–445) was a MYST-type HAT domain (aa 162–433) (Supplementary Figure S4A). Therefore, the MYST-type HAT domain is important for the interaction with Sld3. As in the case of sld3^Esa1(−)-3, the loss rate of ARS301 and ARS301–320 plasmids was elevated in esa1^Sld3(−)-2 (Figure 4C). Collectively, these data indicate that the Sld3–Esa1 interaction has a positive effect on the activation of origins in the silenced locus.

The anti-silencing effect caused by 3–7–45 expression is transient

Chromatin silencing is maintained over generations. In this study, by monitoring the activity of a URA3 gene inserted into the HMLα locus, we found that 3–7–45 expression counteracted the silencing of the locus. This effect was strengthened in the Δswe1 mutant, which has a longer G1 phase than the wild type, and the opposite effect was observed in the Δmih1 mutant, which has a shorter G1 phase than the wild type. 3–7–45 transiently associates with the Mcm2–7 complex on licensed origins when they fire, and origins are licensed in the G1 phase, implying that the anti-silencing effect of 3–7–45 is related to their association to licensed replication origins.

On 5-FOA plates, colonies expressing 3–7–45 were smaller than those expressing vector control, indicating that the level of Ura3 was higher in the 3–7–45 cells; by contrast, the number of colonies formed was largely unaffected (Figure 1D). Therefore, the anti-silencing effect of 3–7–45 seems to be transient, rather than representing a switch in status. When Sld3^Esa1(−) mutants were expressed, colony size increased, but once again the number of colonies was not affected (Figure 4B, +5-FOA panel). Therefore, we propose the following model: when 3–7–45 is associated with licensed origins, local chromatin structure is transiently changed via the function of Esa1. Although our data cannot strictly exclude the possibility that Esa1 locates silenced origins prior to the association of Sld3, it would be difficult to explain the transient effect of 3–7–45 expression in such a situation.

Interestingly, in the in vitro replication assay with chromatinized template, Esa1 accumulated on template chromatin DNA when S-phase extract was added. This requires Mcm2–7 loading and subsequent DDK treatment (Kurat et al. 2017). Because Sld3 associates with the Mcm2–7 complex at licensed origins in a DDK-dependent manner (Heller et al. 2011; Tanaka et al. 2011; Deegan et al. 2016), these in vitro data are consistent with our model.

In chromatin immunoprecipitation (ChIP) assays, we tried to detect association with Esa1 or a reduction in the level of the silencer protein Sir2 at the HMLα locus following 3–7–45 expression, but we detected no changes (data not shown). Because Sld3 transiently associates with licensed origins during the firing reaction, the association of Esa1 with silenced origins might occur in a very short window and ChIP analysis might not be sensitive enough to detect such transient changes.

Genome silencing is stably maintained throughout the cell cycle and over generations. Hence, our finding that temporary anti-silencing, or fluctuation of silencing, can occur when an origin fires reveals a novel mode of silencing regulation.

Target and the role of Esa1 in the firing reaction

Esa1, a lysine acetyltransferase and the catalytic subunit of the NuA4 complex, acetylates four conserved lysines in the histone H4 N-terminal tail at positions 5, 8, 12, and 16 (Dhar et al. 2017). Therefore, transient recruitment of Esa1 by Sld3 might transiently change the modification levels of local proteins. Using ChIP with the specific antibodies against H4K5Ac, H4K8Ac, H4K12Ac, and H4K16Ac raised by the Kimura group (Hayashi-Takanaka et al. 2015), we tried to detect changes in these modifications in 3–7–45 expressing cells. However, as in the case of the Esa1-ChIP, we could not detect significant changes, probably because of technical limitations. Further analysis that is aimed at identifying Esa1 targets, including non-histones, will be necessary to understand the precise role of Esa1 in the firing reaction. The aforementioned in vitro replication assay (Kurat et al. 2017) would be very useful for elucidating the function of Esa1 in the firing reaction.

Is the full NuA4 complex required for anti-silencing and untimely activation of HMLα origins caused by 3–7–45 expression?

Budding yeast NuA4 consists of 13 unique subunits, including Esa1 (Allard et al. 1999). We tested the genetic interactions between NuA4 and Sld3 by combining deletions of non-essential subunits of NuA4 (Δeaf3, Δeaf5, Δeaf6, Δeaf7, and Δyaf9) with sld3-ts alleles. Although we detected genetic interactions (Figure 2 and Supplementary Figure S2), the strengths and directions of these interactions were complex. Although this might be partly due to differences between sld3-5 and sld3-6 (see below), the structure of the NuA4 complex might also affect the phenotype. For example, the 13 subunits of NuA4 can be separated into subcomplexes, one of which is the TINTIN triad consisting of Eaf3, Eaf5, and Eaf7 (Cheng and Cote 2014; Rossetto et al. 2014). Δeaf3 and Δeaf5 exacerbated the growth defects of sld3-6 to different extents, whereas Δeaf7 suppressed the growth defects (Figure 2B and Supplementary Figure S2B). By contrast, in sld3-5, Δeaf5 and Δeaf7 weakly exacerbated the growth defects, whereas Δeaf3 did not (Supplementary Figure S2C). Moreover, although both Δeaf5 and Δeaf6 exacerbated the phenotypes of sld3-5 and sld3-6, neither Eaf5 nor Eaf6 interacted with Sld3 in the Y2H assay. Together, these findings suggest the existence of different role(s) for NuA4 subunits or other subcomplexes.

We are deeply grateful to M. Miyazawa-Onami and A. Hosoe for their technical assistance, to H. Kimura for antibodies against acetylated histones, and to B. Stillman for the SB49 antibody. We also thank H. Araki, M. Yagura, and Y. Tanaka for help and discussion.

S.T. is responsible for all aspects of the study.