The molecular chaperone Hsp90 regulates heterochromatin assembly through stabilizing multiple complexes in fission yeast

Heterochromatin is a highly condensed form of chromatin in eukaryotes. In the fission yeast Schizosaccharomyces pombe, large blocks of constitutive heterochromatin domains coat centromeres, telomeres, the silent mating-type locus and rDNA regions, and are essential for functional organization of these chromosomal domains (Cam et al., 2005; Grewal and Jia, 2007; Moazed, 2009). It has been increasingly clear through extensive studies in the past decades that the RNA interference (RNAi) pathway is one of the main pathways mediating heterochromatin formation and post-transcriptional silencing in fission yeast and a variety of multicellular organisms, such as plants like Arabidopsis thaliana (Martienssen and Moazed, 2015). The RNAi-induced transcriptional silencing (RITS) complex first identified in fission yeast directly regulates heterochromatic gene silencing, and initiates and maintains heterochromatin formation at centromeres (Verdel et al., 2004). It also contributes partly to forming heterochromatin at telomeres, rDNA regions and the mating-type locus (Martienssen and Moazed, 2015; Motamedi et al., 2004; Noma et al., 2004; Verdel et al., 2004). The RITS complex is composed of Argonaute (Ago1), Chp1 and Tas3, in which Tas3 bridges Ago1 and Chp1 to form a linear architecture and uses the single-stranded guide siRNAs to target homologous chromatin containing nascent noncoding RNAs for silencing (Jain et al., 2016; Schalch et al., 2011; Verdel et al., 2004). The RITS complex receives short interfering RNAs (siRNAs) passed from the argonaute chaperone (ARC) complex, which consists of Ago1, Arb1 and Arb2 (Buker et al., 2007). The association of RITS with nascent transcripts via Ago1 leads to the recruitment of the Clr4–Rik1–Cul4 (CLRC) methyltransferase/ubiquitin ligase complex to chromatin via the bridging protein Stc1 (Bayne et al., 2010; Hong et al., 2005), which is followed by additional cycles of histone 3 lysine 9 (H3K9) methylation (H3K9me) (Zhang et al., 2008). The H3K9me mark provides a binding site for the chromodomain-containing proteins Swi6, Chp1 and Chp2 (Huisinga et al., 2006). RITS also recruits the RNA-dependent RNA polymerase complex (RDRC), which contains the conserved proteins Rdp1 (fission yeast RdRP), Hrr1 and Cid12, and is required for synthesis of double-stranded RNAs substrate for Dcr1 (fission yeast Dicer) to generate duplex siRNAs (Colmenares et al., 2007; Motamedi et al., 2004). This step promotes both further dsRNA and siRNA production and a targeted H3K9 methylation and then completes a positive self-reinforcing feedback loop. Recent studies have revealed that Ago1 (fission yeast Argonaute) and the 3′-5′ exonuclease Triman generate Dicer-independent primal small RNAs (priRNAs) and mature siRNAs to initiate de novo assembly of heterochromatin at centromeric repeats (Marasovic et al., 2013). priRNAs are degradation products of abundant transcripts that bind to Ago1 and target antisense transcripts that result from bidirectional transcription of DNA repeats (Halic and Moazed, 2010).

In addition to the RNAi pathway-mediated cis-acting post-transcriptional gene silencing (cis-PTGS) mechanism, several parallel mechanisms also contribute to heterochromatic silencing at both post-transcriptional and transcriptional levels in S. pombe. A multienzyme effector complex called the Snf2/HDAC-containing repressor complex (SHREC; composed of the histone deacetylase Clr3, the SNF2 chromatin-remodeling protein Mit1, and additional proteins Clr1, Clr2 and Ccq1) distributes throughout all major heterochromatin domains to regulate histone acetylation and nucleosome positioning, and thus transcriptional gene silencing (TGS) (Motamedi et al., 2008; Sugiyama et al., 2007). On the other hand, another SNF2 family protein, Fft3, facilitates heterochromatin inheritance, rather than its de novo assembly, by suppressing histone turnover in dividing cells (Taneja et al., 2017). Also, two stress-activated ATF/CREB proteins, Atf1 and Pcr1, as well as Clr5, and the shelterin complex (consisting of six subunits Taz1, Rap1, Poz1, Tpz1, Pot1 and Ccq1) act at the silent mating-type locus and at telomeres to initiate heterochromatin formation, respectively (Hansen et al., 2011; Jia et al., 2004; Kanoh et al., 2005; Kim et al., 2004). Among these factors, Atf1, Pcr1 and Clr5 likely recruit the histone deacetylase Clr6 to the mating-type region, while the shelterin subunit Ccq1 recruits CLRC and SHREC to chromosome ends to establish heterochromatin by directly associating with these two complexes (Hansen et al., 2011; Sugiyama et al., 2007; van Emden et al., 2019; Wang et al., 2016).

Heat-shock proteins (Hsps) are molecular chaperones that control protein folding and function. Hsp90 is one of the most conserved and abundant members of molecular chaperone protein family, present from bacteria to mammals, and is an essential component of the protective heat-shock response. The role of Hsp90, however, extends well beyond stress tolerance, and it participates in many important biological processes (Taipale et al., 2010; Young et al., 2001). Hsp90 may regulate the function or turnover of its associated proteins (known as clients), which is dependent on its ATPase activity (Taipale et al., 2010). Previous biochemical purifications revealed that Argonaute proteins frequently co-purify with heat-shock proteins, including Hsp90, in human and mammals (Höck et al., 2007; Landthaler et al., 2008; Liu et al., 2004; Maniataki and Mourelatos, 2005; Tahbaz et al., 2001). Further studies in human, Drosophila and plants demonstrated that Hsp90 chaperone machinery is directly required to load small RNA duplexes into Argonaute proteins or to stabilize unloaded Argonaute complexes and facilitate the efficient formation of the pre-RNA-induced silencing complex (pre-RISC) in vitro (Iki et al., 2010; Iwasaki et al., 2010; Johnston et al., 2010; Miyoshi et al., 2010; Pare et al., 2009).

The fission yeast S. pombe has been an invaluable model to study the assembly of heterochromatin domains, among which the assembly of centromeric heterochromatin depends on the processing of long non-coding repeat RNAs into double-stranded siRNAs (Moazed, 2009; Volpe et al., 2002). Fission yeast shares the same RNAi mechanisms with higher eukaryotes, and each of the major RNAi factors is encoded by single genes (dcr1⁺, ago1⁺ and rdp1⁺) (Volpe et al., 2002). Hsp90 in fission yeast is encoded by the swo1⁺ gene and has been shown to be required for stability of the tyrosine kinases Wee1 and Mik1 and the type II myosin Myo2 (Aligue et al., 1994; Goes and Martin, 2001; Mishra et al., 2005). Given the fact that Hsp90 is critical for RNAi-mediated heterochromatin formation and silencing in higher eukaryotes, we wondered whether a similar mechanism exists in fission yeast. Here, through genetic analyses and chromatin immunoprecipitation followed by sequencing (ChIP-Seq), we show that fission yeast Hsp90 is indeed required for heterochromatic gene silencing or H3K9me enrichment at all major constitutive heterochromatin regions and an artificial heterochromatin locus. Consistent with a recent independent study (Okazaki et al., 2018), we also demonstrate that Hsp90 promotes RNAi pathway-mediated heterochromatin assembly through maintaining the integrity of the RITS and ARC complexes. More strikingly, we further find that Hsp90 plays much broader roles in heterochromatin nucleation, maintenance and transmission, especially at the mating-type locus and subtelomeres, by promoting the stabilization or assembly of RNAi-independent factors and complexes, such as the SNF2 family nucleosome remodeling factor Fft3, the shelterin complex subunit Poz1 and the SHREC. In addition, Hsp90 is required for proper recruitment of the CLRC by the shelterin component Ccq1 to subtelomeres and targeting of the SHREC and Fft3 to mating type locus and/or rDNA regions. Thus, our work demonstrate that fission yeast Hsp90 regulates both RNAi-dependent and -independent heterochromatin assembly and gene silencing through stabilizing multiple effectors and effector complexes and modulating their heterochromatin target-directed localization.

To examine the potential role of Hsp90 in heterochromatin organization in vivo, we first chose a temperature-sensitive mutant swo1-26, which carries a mutation in the ATPase domain from glycine to cysteine at residue 84 (G84C) (Alaamery and Hoffman, 2008; Aligue et al., 1994) (Fig. 1A). Hereafter, this mutant is denoted hsp90-G84C. By genetic crosses, we generated hsp90-G84C yeast strains in which a ura4⁺ or ade6⁺ marker gene was inserted into the outermost pericentromeric heterochromatin region of chromosome 1 (otr1R::ura4⁺ and otr1R::ade6⁺, respectively), at the silent mating-type locus (mat3M::ura4⁺), within the telomere-associated sequences (TAS) of chromosome 2 (tel2L::ura4⁺) or within rDNA repeats (rDNA::ura4⁺) (Allshire et al., 1995; Ekwall et al., 1999; Nimmo et al., 1998; Thon and Klar, 1992; Thon and Verhein-Hansen, 2000) (Fig. S1A). At both the permissive temperature of 25°C and the semi-permissive temperature of 28°C, hsp90-G84C mutants carrying the ura4⁺ reporter gene showed a slightly but appreciably better growth on medium without uracil and a mild loss of growth on medium containing the counter-selective drug 5-fluoroorotic acid (5-FOA) compared to wild-type cells (Fig. 1B–E; Fig. S1B), indicating a weak loss of heterochromatic gene silencing. Similarly, de-repression of the reporter otr1R::cade6⁺ was also observed in the hsp90-G84C mutant (Fig. S1C). Consistent with these results, the levels of transcripts from the reporter genes, as measured through real-time reverse transcriptase quantitative PCR (RT-qPCR), increased moderately in hsp90-G84C cells, validating the impaired heterochromatin silencing at all tested regions (Fig. 1B–E; Fig. S1B,C). Interestingly, the hsp90-G84C mutation significantly enhanced the heterochromatic silencing defects at centromere and rDNA regions, but not at the mating type locus and telomeres, in cells lacking either Ago1 or Dcr1, two essential effectors of the RNAi machinery (Fig. 1B–E; Fig. S1B,C). Furthermore, the transcript levels of native repeats within four major heterochromatin regions were also largely increased in the hsp90-G84C mutant (Fig. 1F). These data suggest that, as for its homologs in higher eukaryotes, fission yeast Hsp90 also participates in heterochromatic gene silencing.

Our above results suggest that the fission yeast Hsp90 might localize at heterochromatin regions in the genome to facilitate RNAi-mediated gene silencing. However, we failed to faithfully detect Hsp90 localization at the major heterochromatin regions by chromatin immunoprecipitation (ChIP), in which the unspecific background was unacceptably high (data not shown). This was most likely due to the fact that Hsp90 is one of the most abundant molecular chaperones within eukaryotic cells and has multiple functions; thus, probably only a small subset of its protein pool would colocalize with heterochromatin. To circumvent this difficulty, we sought to directly remove Hsp90 only at heterochromatin regions. For this purpose, we engineered a Hsp90 protein that could be cleaved by tobacco etch virus protease (Dougherty et al., 1989) by inserting two TEV recognition sites with a 9-amino-acid sequence (ENLYFQGAS) right behind residue 220 and 330 within the N- and middle-domain of Hsp90, respectively [denoted Hsp90(tev)] (Fig. 2A). To inactivate only the heterochromatin pool of Hsp90, we fused TEV protease sequence with CFP and two tandem sequences of chromo domain (CD) of Swi6 (2xCD-CFP-TEV) (Fig. 2A), which binds to K9-methylated histone H3 (H3K9me), which is mostly present at major heterochromatin regions (Ekwall et al., 1995; Kawashima et al., 2007; Kitajima et al., 2006; Sadaie et al., 2004). The engineered Hsp90 protein expressed from bacteria was efficiently cleaved by TEV protease in vitro (Fig. 2B). Also, in addition to being localized at other heterochromatin sites, the 2xCD-CFP-TEV fusion protein was indeed localized at centromeric regions, as indicated by its juxtaposition to kinetochore protein Cnp3 (Tanaka et al., 2009) (Fig. 2C).

Intriguingly, yeast cells harboring both Hsp90(tev) and 2xCD-CFP-TEV showed strong expression of the otr1R::ade6⁺ reporter gene, demonstrated by both ameliorated growth on medium containing limited adenine and significantly enhanced mRNA level of the ade6⁺ gene detected by RT-qPCR (Fig. 2D). We noticed that the silencing of the ade6⁺ gene was weakly lost in the hsp90(tev) cells alone (Fig. 2D), suggesting insertion of cleavage sites for TEV protease caused some defects in Hsp90 function, such as insufficient targeting of Tas3 to centromeres (see our data below). As a control, we also tested the effect of a TEV protease fused with a C-terminal sequence of kinetochore protein Cnp3 (Cnp3C-TEV), which exclusively localizes to the centromeric central core region (Yokobayashi and Watanabe, 2005) (Fig. 2C), on expression of otr1R::ade6⁺ (Fig. 2D). We found that Cnp3C-TEV only negligibly influenced the silencing of the otr1R::ade6⁺ (Fig. 2D), suggesting that Hsp90(tev) was cleaved in a pericentromeric heterochromatin region-specific manner and only heterochromatin regions, and not central core region-localized Hsp90, is involved in gene silencing control.

RNAi-dependent heterochromatin formation is tightly linked to centromere function and proper segregation of chromosomes in S. pombe. Mutant strains that are defective in centromeric silencing usually exhibit lagging chromosomes, higher minichromosome loss rates, and increased sensitivity to microtubule-destabilizing drugs such as thiabendazole (TBZ) (Hall et al., 2003). Because our above results showed that cells carrying TEV protease-cleavable Hsp90(tev) and heterochromatin-localized 2xCD-CFP-TEV were defective in pericentromeric heterochromatin silencing, we assumed that these cells probably also experienced difficulty in faithful segregation of chromosomes. As expected, these cells were indeed hypersensitive to TBZ, indicating that chromosome segregation is not robust in these mutant cells (Fig. 2E). In contrast, we did not observe any TBZ sensitivity in the hsp90-G84C mutant (data not shown), supporting the idea that specifically abolishing pericentromere-localized Hsp90 leads to stronger heterochromatin gene silencing defects than compromising Hsp90 activity globally. It is noteworthy that the stronger influence of 2xCD-CFP-TEV than Cnp3C-TEV on pericentromeric gene silencing and TBZ sensitivity in cells harboring Hsp90(tev) is not attributable to either significantly elevated total protein level of 2xCD-CFP-TEV or enhanced accumulation of 2xCD-CFP-TEV at centromeres compared to Cnp3C-TEV (Fig. 2F,G).

It has been previously shown that tethering the Tas3 subunit of RITS complex to a modified ura4⁺ transcript (ura4-5BoxB), using the site-specific RNA-binding protein λN, can induce heterochromatin formation and ura4⁺ silencing ectopically (Fig. S2A) (Bühler et al., 2006). This artificial heterochromatin assembly at ura4-5BoxB site requires RNAi components, the histone methyltransferase Clr4 and the heterochromatin protein Swi6 (Bühler et al., 2006). To test whether Hsp90 protein is also required to maintain the Tas3-λN-induced silencing of the ura4⁺ gene, we examined the effect of mutation hsp90-G84C on ura4-5BoxB silencing. As shown in Fig. S2B, growth of tas3-λN ura4-5BoxB cells on counter-selective 5-FOA medium was abolished in the hsp90-G84C mutant, similar to what is seen for the ago1Δ mutant, suggesting the relieved gene silencing at ura4-5BoxB locus. These results support that fission yeast Hsp90 participates in RNAi-mediated gene silencing, and that this possibly occurs through Ago1, as in higher eukaryotes, such as humans, Drosophila and plants (Iki et al., 2010; Iwasaki et al., 2010; Johnston et al., 2010; Miyoshi et al., 2010; Pare et al., 2009). Detection of ura4 transcripts by RT-qPCR confirmed that the hsp90-G84C mutation led to even more significant de-repression of ura4-5BoxB than ago1Δ (Fig. S2C), possibly due to the fact that hsp90-G84C mutation destabilizes not only Ago1-containing complexes but also some RNAi-independent factors involved in heterochromatic silencing, such as Fft3 and SHREC components (see below).

H3K9me is a conserved hallmark for heterochromatic domains in eukaryotes (Grewal and Rice, 2004; Hall et al., 2002; Lachner et al., 2003; Nakayama et al., 2001; Rea et al., 2000). In fission yeast, histone methyltransferase Clr4 catalyzes H3K9 methylation, thereby creating binding sites for the HP1 homolog Swi6, which is essential for the assembly and spreading of heterochromatic structures (Nakayama et al., 2001). One comprehensive genomic mapping analysis revealed that prominent peaks of H3K9me and Swi6 associate with major heterochromatic loci including centromeres, subtelomeres, rDNA and the mat locus (Cam et al., 2005).

To further examine whether our observed heterochromatic gene silencing defects in the hsp90-G84C mutant is coupled to compromised assembly of heterochromatin, we performed ChIP followed by sequencing (ChIP-Seq) to monitor the level of dimethylated H3K9 (H3K9me2) in the whole genome. Our results showed that H3K9me2 enrichment was altered at subtelomeric regions and rDNA regions in the hsp90-G84C mutant compared to wild-type cells even at 25°C, but at mating-type locus the change was observed only at 37°C (Fig. 3). Specifically, the subtelomeric regions in hsp90-G84C mutant cells were shrunk dramatically at chromosomes 1 and 2 (Fig. 3B). As controls, ago1Δ and tas3Δ mutants only showed decreased H3K9me2 at centromeres, but not at other regions, where multiple mechanisms are required for heterochromatin formation. Quite unexpectedly, we noticed that H3K9me2 spreads further in the tas3Δ mutant than wild-type cells at subtelomeric regions for unknown reasons, which deserves further investigation in the future. By examining some representative sites using ChIP-qPCR, we confirmed the reduction of H3K9me2 at heterochromatic regions, especially at pericentromeres (dg in otr), subtelomeres (SPAC212.12, fmt1⁺) and rDNA (28S), in hsp90-G84C cells cultured at 25°C (Fig. 3E). Note that the reduction of H3K9me2 at the pericentromeric dg site could not be observed in ChIP-Seq profiles (Fig. 3A). In addition, H3K9me2 modification was also specifically disrupted at pericentromeres, telomeres and rDNA in hsp90(tev) 2XCD-TEV cells (Fig. S3). Overall, our combined data from hsp90-G84C and hsp90(tev) mutants demonstrated that H3K9me2 modification was altered to various degrees at different sites, with the rDNA region the most strongly reduced, and the mating-type locus and pericentromeres the least influenced. It is possible that compromised heterochromatin assembly or maintenance at certain loci, such as the rDNA region and internal subtelomeres, leads to release and redistribution of heterochromatin-related factors such as Swi6 to ‘buff off’ the heterochromatin formation defects. This mechanism has been suggested for the restoration of pericentromeric heterochromatin assembly in RNAi mutants upon eliminating telomeric shelterin components (Tadeo et al., 2013). Nevertheless, our above data suggest that Hsp90 is indeed involved in promoting heterochromatin assembly or spreading at major heterochromatin sites.

One previous study revealed that the mammalian Hsp90 can associate with and stabilize GERp95 protein, a member of PAZ protein family (Tahbaz et al., 2001). Subsequently, GERp95 was identified as Ago2 and shown to participate in RNAi pathway in human cells (Liu et al., 2004). Recent in vitro studies in human, mammalian, Drosophila and plant cells have demonstrated that Hsp90 plays a vital role in loading small RNA duplexes onto Argonaute proteins, which are the core factor in RNAi machinery (Iki et al., 2010; Iwasaki et al., 2010; Johnston et al., 2010; Miyoshi et al., 2010; Pare et al., 2009). In fission yeast, Ago1, Chp1 and Tas3 form a RNAi effector complex, the RITS complex, which links siRNAs to heterochromatin assembly (Verdel et al., 2004). As a molecular chaperone protein, Hsp90 may contribute to effector complex formation in the RNAi pathway. Therefore, we wondered whether the fission yeast Hsp90 helps the RNAi-dependent assembly of heterochromatin through maintaining the integrity of the RITS complex. We first examined whether the protein level of mutant protein Hsp90-G84C itself is altered. Interestingly, the amount of Hsp90 itself was not altered in hsp90-G84C cells (Fig. 4A), suggesting that the function, rather than the quantity, of the Hsp90 protein is affected by the G84C mutation. Next, we tested the protein levels of two major components of the RITS complex, Ago1 and Tas3, and found that the amount of Tas3 but not Ago1 was significantly reduced once the hsp90-G84C cells were shifted to restrictive temperature of 37°C for 4 h (Fig. 4B). These data suggest that Hsp90 is required to maintain the proper abundance of Tas3 protein in cells.

We then examined how the altered protein level of Tas3 affects the RITS complex formation in vivo. Notably, the low level of Tas3 in hsp90-G84C cells could only recruit minimal amount of Ago1 detected by the co-immunoprecipitation assay at elevated temperatures (Fig. 4C). This is in contrast to a previous study, which showed that the interaction between Ago1 and Tas3 is not affected by hsp90-A4 (Okazaki et al., 2018). Accordingly, our ChIP assay revealed that targeting of Tas3, and therefore possibly also the RITS complex, to centromeres, telomeres and rDNA regions was indeed compromised in hsp90-G84C cells (Fig. 4D). Since Ago1 is also involved in assembly of the ARC complex, containing Ago1, Arb1 and Arb2, which is responsible for passing siRNAs to RITS complex, we next investigated the interaction between Ago1 and Arb1 in the mutant cells. Interestingly, even at 25°C, the amount of Arb1-associated Ago1 in the immunoprecipitates from hsp90-G84C cells was much reduced compared to wild-type samples (Fig. 4E), which was similarly observed previously with the hsp90-A4 mutant (Okazaki et al., 2018). It is noteworthy that the amount of Arb1 was also drastically decreased in hsp90-G84C cells at high temperatures (Fig. 4E). However, neither the protein abundance of Ago1, Tas3 and Arb1 nor the interaction between Ago1 and Tas3 or Arb1 was affected in hsp90(tev) 2XCD-TEV cells (Figs S4, S5A,B). This was most likely due to the possibility that the Hsp90 cleaved by heterochromatin-localizing TEV protease only takes up a very small portion of total cellular Hsp90 pool, which may escape the detection capability of immunoblotting and co-immunoprecipitation, but the specifically affected portion could potentially cause heterochromatin formation defects. Supporting this assumption, we found that the targeting of Tas3 to not only centromeres, but also telomeres, mating-type locus and rDNA, was abolished in hsp90(tev) 2XCD-TEV cells (Fig. S6A). This is because, although Swi6 prominently associates with centromeres through its CD domain, it also associates with other heterochromatic loci including subtelomeres, the mat locus and rDNA (Cam et al., 2005). Taken together, these results strongly suggest that Hsp90 contributes to the formation and stabilization of the RITS complex and the ARC complex in vivo, and to the active recruitment of at least RITS to major heterochromatin sites, especially at high temperature.

As a molecular chaperone protein, Hsp90 may execute its function through directly associating with its clients (Taipale et al., 2010). Previous biochemical purifications revealed that Argonaute proteins frequently co-purify with heat-shock proteins, including Hsp90, in human and mammals (Höck et al., 2007; Landthaler et al., 2008; Liu et al., 2004; Maniataki and Mourelatos, 2005; Tahbaz et al., 2001). Although purifications of the RITS complex and the ARC complex in fission yeast did not identify Hsp90 as a permanently bound component (Buker et al., 2007; Motamedi et al., 2004; Verdel et al., 2004), it is possible that, among the peptides identified by purification of those protein complexes, Hsp90 was usually considered as an unspecific background protein, or alternatively, that the physical interactions between Hsp90 and subunits of the RITS or ARC complexes is transient. To find out whether Hsp90 indeed interacts with the RITS or ARC complexes, we performed co-immunoprecipitation assays, in which tagged Ago1, Tas3 or Arb1 was immunoprecipitated. Intriguingly, Hsp90 could be unambiguously detected in the immunoprecipitates when Flag–Ago1, Tas3–TAP or Arb1–TAP was used as the bait (Fig. S7A), supporting the notion that fission yeast Hsp90 is associated with both RITS and ARC complexes in vivo. It is worth noting that the removal of unspecific binding of Hsp90 to IgG beads, which was most likely due to its unusually high abundance, was only possible when high-salt concentrations were used during immunoprecipitation procedures.

Members of the Argonaute/PIWI family contain two conserved domains, PAZ and PIWI, in which PAZ is flanked by an N-terminal domain (N) and a middle domain (M). Whereas the PAZ domain has been shown to be important for siRNA binding (Lingel et al., 2003; Song et al., 2003; Yan et al., 2003), the PIWI domain of Argonaute proteins is responsible for the endonucleolytic cleavage of a target RNA (Liu et al., 2004; Song et al., 2004). MBP pulldown assays were performed with a series of MBP fusions of Ago1 fragments expressed in bacteria and yeast extract containing Hsp90–GFP (Fig. S7B). Our results indicated that the N-terminal domain of Ago1 is the major portion responsible for its association with Hsp90, while the PAZ and PIWI domains have only minimal or no interactions with Hsp90 (Fig. S7C).

In S. pombe, both RNAi pathway-mediated and -independent mechanisms operate in the establishment, maintenance and transmission of heterochromatin. Our genetic analyses showed that the hsp90-G84C mutation with either ago1Δ or dcr1Δ had an additive effect on defective heterochromatic silencing (Fig. 1B–E; Fig. S1B,C), suggesting that Hsp90 may also affect RNAi-independent factors involved in heterochromatin formation. To investigate this possibility, we first used a well-characterized strain in which the cenH sequence at the silent mating-type region is replaced with ade6⁺ (KΔ::ade6⁺) (Ayoub et al., 1999) (Fig. 5A). KΔ::ade6⁺ cells display various levels of ade6⁺ expression (Fig. 5B), because deletion of cenH causes specific defects in the RNAi-dependent establishment of the silenced state and heterochromatin (Hall et al., 2002). In wild-type cells, once the heterochromatin-enriched ade6-off state is established, it is stably inherited and rarely switches back to the ade6-on state. However, hsp90-G84C mutation compromised this ade6-off state in a clonally variegated manner (Fig. 5B), indicating that maintenance or transmission of heterochromatin was disrupted. As two sequence-specific DNA binding proteins Atf1 and Pcr1 act in a mechanism parallel to the RNAi pathway at the mating-type locus to initiate heterochromatin (Jia et al., 2004; Kim et al., 2004), we examined whether their protein stability was affected by hsp90-G84C. Surprisingly, the abundance of both Atf1 and Pcr1 was not reduced in the hsp90-G84C mutant (Fig. S8A). As it has been shown that Atf1 and Pcr1 are responsible for targeting the multienzyme effector complex SHREC to the mating-type locus (Sugiyama et al., 2007; Yamada et al., 2005), we next tested whether SHREC was defective in hsp90-G84C mutant. Interestingly, although the abundance of the scaffold protein Clr1 was only slightly reduced, its interaction with the histone deacetylase (HDAC) subunit Clr3 was significantly compromised in the hsp90-G84C mutant (Fig. 5C), suggesting that the assembly of intact SHREC is largely disrupted. Consistent with this, much less Clr3 was enriched at mating-type locus and also at rDNA region in hsp90-G84C cells compared to in wild-type cells (Fig. 5D). Both the protein abundance of Clr3 and its binding at mating-type locus was unaltered in hsp90(tev) 2XCD-TEV strains (Figs S4 and S6B).

Recently, Fft3 has been identified as a fission yeast homolog of the SNF2 family chromatin remodeler and as a factor uniquely required for heterochromatin inheritance, rather than for de novo assembly, particularly at subtelomeric regions and at the mat locus when the cenH heterochromatin nucleation center is replaced (KΔ::ura4⁺), but not at pericentromeric regions (Taneja et al., 2017). Since our ChIP-Seq data showed that H3K9me2 distribution was most decreased at the subtelomeric regions and mat locus, but not at the pericentromeric regions in the hsp90-G84C mutant (Fig. 3), we wondered whether Fft3 could be a substrate of fission yeast Hsp90. Interestingly, Fft3 was unstable in hsp90-G84C cells (Fig. 5E), suggesting that Hsp90 promotes heterochromatin inheritance after nucleation by stabilizing Fft3. Our ChIP data confirmed that the binding of Fft3 at mat locus was specifically lost in the hsp90-G84C mutant (Fig. 5F). However, neither the protein level of Fft3 nor its binding to the mat locus was affected in hsp90(tev) 2XCD-TEV strains (Figs S4 and S6C), further corroborating previous findings that Fft3 uniquely functions at the mat locus to promote heterochromatin inheritance.

Together, these data support the idea that Hsp90 is required to stabilize not only the RNAi-pathway effector complexes RITS and ARC, but also the RNAi-independent factors SHREC and Fft3, and contributes to both the establishment and transmission of heterochromatin through these factors prominently at the mat locus in S. pombe.

In S. pombe, heterochromatin assembly at native telomeres also requires redundant pathways, including at least the shelterin complex and RNAi mechanism. We observed that H3K9me2 enrichment at internal subtelomeric regions was severely compromised in hsp90-G84C mutant (Fig. 3B). To test whether altered stability or assembly of shelterin was responsible for those defects, we employed a ura4⁺ reporter gene inserted near telomeric repeats of a mini-chromosome Ch16 (TEL::ura4⁺ or m23::ura4⁺) (Fig. 6A), which lacks tlh1⁺ and TAS sequences and whose silencing is dependent entirely on shelterin (Kanoh et al., 2005; Nimmo et al., 1994; Tadeo et al., 2013; Wang et al., 2016). Interestingly, the hsp90-G84C mutation abolished the silencing of the reporter gene to a similar degree to that seen in ccq1Δ or clr4Δ mutants (Fig. 6B), indicating that Hsp90 is required for maintaining the proper function of shelterin on heterochromatin formation at telomeric repeats. Consistent with this, hsp90-G84C also caused a significantly elevated transcript levels of the endogenous subtelomeric non-coding RNAs (TERRA) corresponding to both telomeric repeats and TAS (Bah et al., 2012; Greenwood and Cooper, 2012), indicating derepression of TERRA (Fig. 6C). Intriguingly, the silencing defects of TEL::ura4⁺ in the hsp90-G84C mutant could be completely rescued by mst2Δ or epe1Δ (Fig. 6B), two mutations that have been previously demonstrated to promote heterochromatin spreading (Ayoub et al., 2003; Reddy et al., 2011; Wang et al., 2015, 2013, 2012a).

Internal to the TAS at left telomere 1, a repetitive DNA element within the tlh1⁺ gene mediates the H3K9me2 enrichment and recruitment of Swi6 through the RNAi pathway (Kanoh et al., 2005). We found that both hsp90-G84C and ccq1Δ mutants displayed strong silencing defects at tlh1⁺, which was even stronger than that for the RNAi component mutant ago1Δ (Fig. 6C). Whereas the transcription of some subtelomeric genes, such as SPAC212.12 and fmt1⁺, which sit further towards the centromere compared to tlh1⁺ on the left of telomere 1, was not severely affected by hsp90-G84C, ccq1Δ or ago1Δ (Fig. 6C). Quite unexpectedly, transcription of tlh1⁺, SPAC212.12 and fmt1⁺ in hsp90-G84C mutant was not tightly correlated with levels of heterochromatin hallmark H3K9me2 at these sites (compare Fig. 6C and Fig. 3B,E). These observations suggest that the silencing of these subtelomeric genes must be mediated via a mechanism independent of H3K9me.

We next examined the protein levels of all six components of shelterin complex by western blotting and found that only the abundance of Poz1, but not its other components, was reduced significantly in the hsp90-G84C mutant (Fig. 7A; Fig. S8B). Indeed, the proper connection between shelterin components Ccq1 and Taz1 appeared to be intact (Fig. 7B), whereas the interaction between Ccq1 and the CLRC subunit Clr4, but not the SHREC subunit Clr3, was disrupted (Fig. 7C,D). These data suggest that very likely efficient recruitment of CLRC to telomeric repeats and subtelomeres by Ccq1 is compromised in the hsp90-G84C mutant. Consistent with this idea, Clr4 enrichment at telomeres was indeed much less efficient in the hsp90-G84C mutant (Fig. 7E), although it remained unchanged in the hsp90(tev) 2XCD-TEV strains (Fig. S6D). Collectively, these results demonstrate that Hsp90 contributes to telomeric and subtelomeric heterochromatin assembly by stabilizing the interaction between the shelterin complex and CLRC.

So far, the compositions of all major effector complexes involved in heterochromatin formation operated directly or indirectly by the RNAi machinery in fission yeast, such as RITS, ARC, RDRC, CLRC and SHREC complexes, have been identified, mainly based on biochemical purifications (Bayne et al., 2010; Buker et al., 2007; Colmenares et al., 2007; Hong et al., 2005; Jia et al., 2005; Motamedi et al., 2004; Noma et al., 2004; Sugiyama et al., 2007; Verdel et al., 2004). Occasionally, some critical factors required for RNAi-dependent heterochromatin assembly, have been revealed by genetic screens or hint-directed genetic analyses. For example, the splicing factors Cwf10, Prp10 and Cwf14 (Bayne et al., 2008; Kallgren et al., 2014), the spliceosomal U4 snRNA (Chinen et al., 2010), the Spt-Ada-Gcn5-acetyltransferase (SAGA) complex subunit Sgf73 (Deng et al., 2015) and the inner nuclear membrane protein complex Lem2–Nur1 (Banday et al., 2016; Barrales et al., 2016) have been identified by these approaches. In this study, we provided another example where genetics helped to identify a novel factor, Hsp90, that broadly regulates the stability or assembly of factors or complexes involved in heterochromatin formation in S. pombe.

As one member of the most conserved and abundant molecular chaperone protein family, Hsp90 protein plays diverse roles in many important biological processes, well beyond stress tolerance (Taipale et al., 2010; Young et al., 2001). In fission yeast, Hsp90 is encoded by the swo1⁺ gene and so far its functions have been attributed to only a few processes, such as cell cycle control, cytokinesis and glucose/cAMP signaling (Alaamery and Hoffman, 2008; Aligue et al., 1994; Goes and Martin, 2001; Mishra et al., 2005). It has been shown that the fission yeast Hsp90 contributes to maintaining stability of the tyrosine kinases Wee1 and Mik1 and type II myosin Myo2 (Aligue et al., 1994; Goes and Martin, 2001; Mishra et al., 2005). A few previous studies have revealed that Hsp90 is involved in promoting the formation of the RNAi effector complexes in plants and animals in vitro (Iki et al., 2010; Iwasaki et al., 2010, 2015; Miyoshi et al., 2010), and evidence for the fission yeast counterpart of Hsp90 is required for siRNA generation and RNAi-dependent heterochromatin assembly has been provided in a recent study (Okazaki et al., 2018).

In this study, initially through genetic analyses, we found that the fission yeast Hsp90 is indeed required for heterochromatic gene silencing at all major constitutive heterochromatin regions and an artificial heterochromatin site. Our ChIP-seq and ChIP-qPCR data further provided evidence that heterochromatin assembly marked by H3K9me2 is actually compromised at major constitutive heterochromatin sites in the hsp90 mutant. All these defects seem to be caused by unstable Tas3, Arb1, Fft3 and Poz1 proteins, disrupted Argonaute-containing RNAi effector complexes RITS and ARC, compromised assembly of SHREC and insufficient recruitment of CLRC by shelterin to chromosome ends.

Based on our detailed analyses on effects of hsp90 mutants [either hsp90-G84C or hsp90(tev) 2XCD-TEV] on protein stability, complex assembly and the heterochromatin site-binding capability of some RNAi-dependent and -independent factors, we propose that Hsp90 regulates heterochromatin assembly, maintenance and transmission at different sites with different priorities (Fig. 8). For example, RITS and SHREC are general targets of Hsp90 for post-transcriptional and transcriptional gene silencing at all four major heterochromatin domains in the genome, whereas Fft3 and SHREC are prominently protected by Hsp90 for heterochromatin maintenance and inheritance at the mating-type locus, and shelterin complex only at the telomeres, for recruiting CLRC and local heterochromatin spreading (Fig. 8). Although the mechanisms for this division of labor remain elusive, one possibility is that Hsp90 employs different co-chaperones for directed targeting to different clients. Thus, identifying the potential specific Hsp90 co-chaperone involved in heterochromatin formation at distinct sites would be an interesting task for future studies.

Intriguingly, one recent study identified Hsp90 and a nucleocytoplasmic type-I Hsp40 protein Mas5 as novel silencing factors in fission yeast, and demonstrated that both Hsp90 and Mas5 contribute to RNAi-dependent heterochromatin assembly only at pericentromeric regions, but not at the rest of the constitutive heterochromatin sites (Okazaki et al., 2018). Although in their study, Okazaki et al. (2018) used a newly isolated hsp90-A4 mutant which carries a different mutation (R33C) from the hsp90-G84C mutant (carrying the mutation G84C) used in our study, both studies found largely similar defects in heterochromatic gene silencing and heterochromatin assembly at pericentromeric regions and lend independent corroboration to each other. This is most likely due to the fact that both mutations lie within the ATPase domain of Hsp90, and possibly cause a decrease in the ATPase activity. However, the effects of Hsp90 on RNAi-independent factors, such as SHREC, shelterin and Fft3, were likely overlooked by Okazaki et al. (2018) in their hsp90-A4 mutant. This was probably because they did not examine the silencing defects in hsp90-A4 ago1Δ or hsp90-A4 dcr1Δ and therefore missed the exacerbated phenotypes in double mutants, which could suggest the defects in RNAi-independent factors. We obtained this clue from our observations that hsp90-G84C mutation showed exacerbated defects in heterochromatin silencing at pericentromeres, and rDNA regions when combined with ago1Δ or dcr1Δ, mutants of two essential effectors of the RNAi machinery (Fig. 1; Fig. S1). Since Hsp90 is an essential component of the protective heat-shock response and has multiple clients, it is fairly possible that more factors involved in heterochromatin assembly are affected by Hsp90 in S. pombe but are not examined in this study and thus remain to be uncovered.

It has been well-established that, in addition to the RNAi pathway, redundant pathways also mediate the formation of heterochromatin at mating-type locus, rDNA regions or subtelomeres. For example, two ATF/CREB family proteins Atf1 and Pcr1, the Sir2-dependent pathway, the shelterin complex, the fission yeast homolog of Erh1 (enhancer of rudimentary) and the carbon catabolite-repression 4-NOT (CCR4-NOT) multi-subunit protein complex (Cam et al., 2005; Freeman-Cook et al., 2005; Hall et al., 2002; Jia et al., 2004; Kanoh et al., 2005; Kim et al., 2004; Shankaranarayana et al., 2003; Sugiyama et al., 2016). We assume that probably one or more of these extra factors also requires functional Hsp90 to act properly during heterochromatic silencing (Fig. 8).

One major mechanism for Hsp90 to fulfill its functions is through regulating the function or turnover of its associated proteins (i.e. its clients), which is dependent on ATP activity (Taipale et al., 2010). However, the effects of mutations R33C and G84C on ATPase activity may differ to a certain degree, because based on the ATP binding and structural analyses on the S. cerevisiae Hsp90 homolog, the residue Arg32 (Arg33 in the S. pombe protein) does not directly contact ATP, whereas Gly83 together with adjacent Asp79 and Thr171 (corresponding to Gly84, Asp80 and Thr172, respectively in the S. pombe protein) have direct contact with ATP (Mishra et al., 2016; Panaretou et al., 1998; Prodromou et al., 1997). Interestingly, the budding yeast Gly83 is also among the binding sites for Hsp90 inhibitor and antitumor agent geldanamycin (Prodromou et al., 1997). Indeed, a few observations in our study are slightly different from those of Okazaki et al. (2018). For example, the protein level of Ago1 detected by western blotting is decreased in hsp90-A4 mutant but not in hsp90-G84C mutant, but conversely, the protein level of Tas3 is decreased in hsp90-G84C mutant but not in the hsp90-A4 mutant. Accordingly, the integrity of RITS complex in vivo revealed by co-immunoprecipitation efficacy of Tas3 and Ago1 is disrupted in the hsp90-G84C mutant but not in hsp90-A4 mutant. Although possibilities exist that the binding efficiency and protective capability of mutant Hsp90s (Hsp90-G84C and Hsp90-A4) to Ago1 and Tas3 are quite different, or each mutation has distinct effects on the catalytic cycle of the ATPase, the exact reasons for these differences await further detailed investigations. Nevertheless, in addition to maintaining the intact integrity of ARC and RITS complexes, fission yeast Hsp90 may also promote the generation of siRNAs and loading siRNA duplexes onto Argonaute proteins as in Drosophila and plants, which is an ATP-dependent process (Iki et al., 2010; Iwasaki et al., 2010; Miyoshi et al., 2010). This should be an interesting issue for future studies. Collectively, both our findings and those of a previous study demonstrate that the mechanism of Hsp90 mediating the assembly and integrity of Argonaute-containing RITS complex in vivo is very likely to be conserved throughout eukaryotic evolution (Kato et al., 2018; Okazaki et al., 2018), and that Hsp90 is a global regulator of heterochromatin assembly in S. pombe.

In animals, Piwi-interacting RNAs (piRNAs) are 26–30 bases long and act as germ line-specific small non-coding RNAs that have evolutionarily conserved functions in silencing of mobile genetic elements (transposons) and maintenance of genome integrity (Ghildiyal and Zamore, 2009; Juliano et al., 2011; Siomi et al., 2011). In Drosophila, Hsp90 together with its co-chaperone Hop (Hsp70/90-organizing protein homolog) form a complex with piRNA-binding protein Piwi and mediate piRNA biogenesis, transposon silencing and epigenetic suppression of hidden genetic variations (Gangaraju et al., 2011; Karam et al., 2017; Specchia et al., 2010). Also, it has been shown in Drosophila, cavefish and plants, that Hsp90 functions as a driver for morphological evolution and phenotypic variations (Queitsch et al., 2002; Rohner et al., 2013; Rutherford and Lindquist, 1998). Thus, it would be interesting to investigate in the future whether fission yeast Hsp90 is also involved in transposon silencing and epigenetic buffering of genetic and phenotypic variations of some traits.

Schizosaccharomyces pombe strains used and created in this study are listed in Table S1. Liquid cultures or solid agar plates consisting of rich medium (YE5S) or minimal medium (Pombe glutamate or PMG5S) containing 4 g/l sodium glutamate as a nitrogen source with appropriate supplements were used as described previously (Forsburg and Rhind, 2006; Moreno et al., 1991). G418 disulfate (Sigma-Aldrich), hygromycin B (Sangon Biotech) or nourseothiricin (clonNAT; Werner BioAgents) was used at a final concentration of 100 μg/ml where appropriate. ura4⁺ silencing was assessed by growth on PMG5S without uracil or with 1 mg/ml 5-FOA, which is toxic to cells expressing ura4⁺. ade6⁺ silencing was assessed by growth on YE5S without or with 0.5 mg/l adenine (YE5S with low adenine). For serial-dilution assays, three serial 10-fold dilutions were made, and 5 μl of each was spotted on plates with the starting cell number of 10⁴. Genetic crosses and general yeast techniques were performed as described previously (Forsburg and Rhind, 2006; Moreno et al., 1991).

To construct strains expressing TEV protease fused with CFP and two tandem sequences of the chromo domain (CD) of Swi6 (i.e. 2xCD-CFP-TEV), two NdeI restriction sites within CFP and TEV gene sequence respectively in vector pNATZA21-CFP-9myc-TEV (a kind gift from Yoshinori Watanabe, Laboratory of Chromosome Dynamics, Institute of Molecular and Cellular Biosciences, University of Tokyo, Japan) were first destroyed by site-directed mutagenesis. A BglII-SacI fragment carrying the kanMX6 cassette derived from pFA6a-kanMX6 (Bähler et al., 1998) was ligated to the similarly digested modified pNATZA21-CFP-9myc-TEV in which two NdeI sites were removed; this resulted in pKANZA21-CFP-9myc-TEV. A DNA fragment harboring two tandem sequences of CD of Swi6 was amplified by PCR using pHBKA81-CFP-2xCD (a kind gift from Yoshinori Watanabe) as template and then inserted into pKANZA21-CFP-9myc-TEV using NdeI sites. This finally generated pKANZA21-CD-CD-CFP-9myc-TEV, and this vector was then linearized by ApaI digestion and integrated at a so called ‘Z locus’ adjacent to the zfs1⁺ gene on chromosome 2 (Tada et al., 2011) using the kan^R marker after transformation.

To construct strains expressing the TEV protease-cleavable Hsp90, a 400 bp fragment downstream of coding sequence of hsp90⁺ gene was first amplified by PCR using yeast genomic DNA as template and cloned into pFA6a-13myc-hphMX6 (Sato et al., 2005) using SacI and EcoRV sites. Then the full-length coding sequence of hsp90⁺ gene without a TAA stop codon was amplified by PCR using yeast genomic DNA as template and ligated to the above vector using XmaI sites and generated the pFA6a-hsp90⁺-13myc-hphMX6-400 bp hsp90⁺ 3′ sequence. The 9-amino-acid target site sequence (ENLYFQGAS) for TEV protease (abbreviated as tev) (Dougherty et al., 1989) was inserted between E220 and E221 and between K330 and R331, respectively, in pFA6a-hsp90⁺-13myc-hphMX6-400 bp hsp90⁺ 3′ sequence by site-directed mutagenesis. This resulted in pFA6a-hsp90⁺(tev)-13myc-hphMX6-400 bp hsp90⁺ 3′ sequence which carries two TEV recognition sites within hsp90⁺ and a 400 bp 3′-untranslated region. This vector was then linearized with BamHI and transformed into hsp90-G84C mutant cells to replace the endogenous hsp90 gene. Transformants resistant to high temperature (37°C) and hygromycin B were selected and confirmed by PCR and sequencing.

To generate the vector for recombinant fusion protein production of MBP-Hsp90(tev), the gene fragment carrying two TEV recognition sites within hsp90⁺ was amplified by PCR using the vector pFA6a-hsp90⁺(tev)-13myc-hphMX6-400 bp hsp90⁺ 3′ sequence as template and then inserted into pMAL-2c (New England BioLabs, Ipswich, MA) using BamHІ and PstІ sites.

To generate the vectors for recombinant fusion protein production of MBP–Hsp90 [1–704 amino acids (aa), full-length], MBP–Hsp90(1–254 aa), MBP–Hsp90(221–540 aa) and MBP–Hsp90(541–704 aa), a gene fragment was amplified by PCR using plasmid pFA6a-hsp90⁺-13myc-hphMX6-400 bp hsp90⁺ 3′ sequence as template and then inserted into pMAL-2c using the BamHІ and PstІ sites.

To generate the vectors for recombinant fusion protein production of MBP–Ago1 (1–834 aa, full-length), MBP–Ago1(1–120 aa), MBP–Ago1(1–215 aa), MBP–Ago1(210–330 aa) and MBP–Ago1(334–834 aa), a gene fragment was amplified by PCR using plasmid pGEX-4T-1-GST-spAgo1 (a kind gift from Leemor Joshua-Tor, Cold Spring Harbor Laboratory, USA) as template and then inserted into pMAL-2c using SalІ and PstІ sites.

C-terminal TAP epitope tagging of endogenous Arb1 and GFP tagging of Fft3 was undertaken by PCR-based gene targeting (Bähler et al., 1998).

All yeast transformation was performed using the lithium acetate method (Bähler et al., 1998; Keeney and Boeke, 1994). Transformants were selected on YE5S plates containing G418 or hygromycin B. Correct tagging and integration was verified by using a colony PCR method or sequencing as described previously (Chen et al., 2017).

Cells were cultured in YE5S at 25 or 28°C to an optical denisty at 600 nm (OD₆₀₀)=0.4–0.5 and total cellular RNA was isolated by using the TRIZOL method. cDNA was prepared using oligo(dT) or random oligonucleotide primers and PrimeScript reverse transcriptase (purchased from Takara). Real-time PCR was performed in the presence of SYBR Green using an Applied Biosystems 7900HT light cycler. Relative RNA levels were calculated using the ΔCT method and normalized to act1⁺ levels. Primer sequences are shown in Table S2.

For co-immunoprecipitation and western blot experiments, cells were lysed by bead disruption using FastPrep24 homogenizer (MP Biomedical) in NP40 lysis buffer (6 mM Na₂HPO₄, 4 mM NaH₂PO4, 1% NP-40, 150 mM NaCl, 2 mM EDTA, 50 mM NaF, 0.1 mM Na₃VO₄) plus protease inhibitors as previously described (Wang et al., 2012b). For co-immunoprecipitations involving Hsp90, NP40 lysis buffer with a higher salt concentration of 900 mM NaCl was used to avoid unspecific binding. Proteins were immunoprecipitated by IgG Sepharose beads (GE Healthcare) (for Tas3-TAP and Arb1-TAP) or mouse monoclonal anti-Flag antibodies (Sigma-Aldrich). Immunoblot analysis of cell lysates and immunoprecipitates was performed using peroxidase-anti-peroxidase soluble complex (Sigma-Aldrich), rabbit polyclonal anti-Myc (GeneScript), rat polyclonal anti-HA (Roche), mouse monoclonal anti-GFP or rabbit polyclonal anti-Hsp90 (generated at Xiamen University antibody facility using recombinant S. pombe Hsp90 as antigens) primary antibodies (at 1:1000–1:2000 dilutions). Cdc2 was detected using rabbit polyclonal anti-PSTAIRE (sc-53, Santa Cruz Biotechnology) as a loading control (1:1000 dilution).

The standard procedures of chromatin immunoprecipitation were used as previously described (Lu and He, 2018; Volpe and Demaio, 2011). The final enrichment of each target sequence relative to a control act1 sequence was calculated as previously described (Wang et al., 2013). For ChIP-Seq, log-phase wild-type cells were grown at 29°C, and log-phase hsp90-G84C mutant cells were grown either at 25°C or grown at 25°C until OD₆₀₀=0.2, then shifted to 37°C for 4 h in YE5S medium. Then 10⁹ cells were harvested and digested by Zymolyase 20T with a final concentration of 0.25 mg/ml at 37°C for almost 1 h. The chromatin was digested into mononucleosome by micrococcal nuclease (EN0181, Thermo Fisher Scientific) to a final concentration of 240 U/ml at 37°C for 20 min. Anti-H3K9me2 (ab1220, Abcam), anti-FLAG (F1804, Sigma-Aldrich), Anti-Myc (ab9132, Abcam), anti-GFP (ab290, Abcam) and IgG Sepharose beads (GE Healthcare) were used in immunoprecipitation. Libraries of DNA were prepared by using a commercial high throughput library preparation kit (KK8301, KAPA Biosystems). The sequencer (Ion PGM™ System, Life Technologies) was used for next-generation sequencing according to the manufacturer's protocols. ChIP-seq raw data was aligned to the assembly genome S. pombe ASM294v.2.22 with BWA and analyzed as previously described (Lu and He, 2018). All ChIP-seq sequencing data sets generated in this study were submitted to NCBI GEO database under the accession number GSE140617.

All recombinant bacterially produced MBP fusion proteins of full-length or truncated Hsp90 or Ago1 were expressed in Escherichia coli BL21(DE3) cells and purified on amylose beads (MBP; New England BioLabs) according to the manufacturer's instructions and as previously described (Wang et al., 2012b), except that lysis buffer containing a higher salt concentration of 900 mM NaCl was used. Purified proteins were incubated with clarified whole-cell yeast lysates made from cells expressing Flag-ago1 or hsp90-GFP for 1–2 h at 4°C, followed by SDS-PAGE, Coomassie Blue staining or western blot analysis to examine the association between full-length or truncated Hsp90 and Ago1.

MBP-Hsp90(tev) fusion protein was expressed in Escherichia coli BL21(DE3) cells and purified on amylose resin. About 20 μg purified MBP-Hsp90(tev) was suspended in 30 µl TEV buffer (50 mM Tris-HCl, pH 8.0, 0.5 mM EDTA, 1 mM DTT) with or without adding 10 U AcTEV protease (Invitrogen) and incubated at 30°C for 2 h. The samples were then subjected to SDS-PAGE, followed by Coomassie Blue staining.

CFP or tdTomato fusion proteins were observed in live or cold methanol-fixed cells. Photomicrographs were obtained using a Perkin Elmer spinning-disk confocal microscope (UltraVIEW^® VoX) with a 100× NA 1.49 TIRF oil immersion objective (Nikon) coupled to a cooled CCD camera (9100-50 EMCCD; Hamamatsu Photonics) and spinning disk head (CSU-X1, Yokogawa). Image processing and analysis was performed using Metamorph 7.1 software (Molecular Devices, Sunnyvale, CA) and Adobe Photoshop.

Experiments for quantification of qPCRs, protein levels, co-immunoprecipitation efficacy and fluorescent signal intensities of Cnp3C-CFP-9myc-TEV, 2xCD-CFP-9myc-TEV and Cnp3-tdTomato were repeated three times, and the mean value and standard deviation (s.d.) for each sample was calculated. In order to determine statistical significance of our data, two-tailed paired-sample t-tests were performed and P-values were calculated using GraphPad Prism 7. P<0.05 was considered statistically significant.

We thank Drs Li-lin Du (National Institute of Biological Sciences, China), Robin Allshire (The University of Edinburgh, UK), Shiv Grewal (National Institutes of Health, USA), Danesh Moazed (Harvard Medical School, USA), Marc Buhler (Friedrich Miescher Institute for Biomedical Research, Switzerland), Janet Partridge (St. Jude Children's Research Hospital, USA), Leemor Joshua-Tor (Cold Spring Harbor Laboratory, USA), Mohan Balasubramanian (University of Warwick, UK) and Yoshinori Watanabe (The University of Tokyo, Japan) for kindly providing the yeast strains or plasmids.

The peer review history is available online at https://jcs.biologists.org/lookup/doi/10.1242/jcs.244863.reviewer-comments.pdf