TORC1 is a conserved multisubunit kinase complex that regulates many aspects of eukaryotic growth including the biosynthesis of ribosomes. The TOR protein kinase resident in TORC1 is responsive to environmental cues and is potently inhibited by the natural product rapamycin. Recent characterization of the rapamycin‐sensitive phosphoproteome in yeast has yielded insights into how TORC1 regulates growth. Here, we show that Sch9, an AGC family kinase and direct substrate of TORC1, promotes ribosome biogenesis (Ribi) and ribosomal protein (RP) gene expression via direct inhibitory phosphorylation of the transcriptional repressors Stb3, Dot6 and Tod6. Deletion of STB3, DOT6 and TOD6 partially bypasses the growth and cell size defects of an sch9 strain and reveals interdependent regulation of both Ribi and RP gene expression, and other aspects of Ribi. Dephosphorylation of Stb3, Dot6 and Tod6 enables recruitment of the RPD3L histone deacetylase complex to repress Ribi/RP gene promoters. Taken together with previous studies, these results suggest that Sch9 is a master regulator of ribosome biogenesis through the control of Ribi, RP, ribosomal RNA and tRNA gene transcription.
The budding yeast Saccharomyces cerevisiae synthesizes ∼2000 ribosomes per minute per cell under optimal growth conditions. To ensure the proper transcription, processing and assembly of the 78 ribosomal proteins (RPs) and 4 ribosomal RNAs (rRNAs) that constitute a ribosome, hundreds of trans‐acting assembly factors and other cofactors for translation regulation, tRNA biosynthesis or purine/pyrimidine synthesis must also be produced from a set of coregulated genes known as the ribosome biogenesis (Ribi) regulon (Jorgensen et al, 2004; Wade et al, 2006). Ribosome biogenesis thus requires the coordinated activities of all three RNA polymerases (RNAPI, II and III) and can commandeer up to 70–80% of total nuclear transcriptional capacity (Warner, 1999). To limit unnecessary energy expenditure under stress and/or starvation conditions, the synthesis of ribosomal components and their Ribi cofactors must be tightly controlled. This regulation is imposed in part by nutrient‐ and stress‐sensitive signalling networks. Most notably, the TOR and PKA kinases regulate Ribi at the transcriptional level (De Virgilio and Loewith, 2006).
The TOR kinases in yeast (TOR1 and TOR2) are the catalytic subunits of two functionally distinct multiprotein complexes, TORC1 and TORC2 (De Virgilio and Loewith, 2006). In optimal growth conditions, TORC1 is active and promotes growth both by stimulating anabolic processes, such as protein synthesis, and by inhibiting catabolic processes, such as macroautophagy (De Virgilio and Loewith, 2006). Binding of the natural product macrolide rapamycin or exposure to cellular stress rapidly inactivates TORC1 (Heitman et al, 1991; Urban et al, 2007), inhibits anabolic processes and induces catabolic processes.
Ribosome biogenesis is the principal anabolic process that is stimulated by TORC1. The assembly of new ribosomes begins with the TORC1‐dependent activation of RNAPI, II and III to promote the transcription of rRNA, RP and Ribi, and tRNA genes, respectively (Zaragoza et al, 1998; Jorgensen et al, 2004; Martin et al, 2004; Schawalder et al, 2004; Wade et al, 2004; Rudra et al, 2005). TORC1‐dependent signals are mediated by a number of effectors kinases (Breitkreutz et al, 2010), the best characterized of which is Sch9, an AGC family Ser/Thr kinase and direct substrate of TORC1 (Urban et al, 2007). Sch9 regulates RNAPIII by phosphorylating and inactivating Maf1, a conserved repressor of RNAPIII activity (Upadhya et al, 2002; Oficjalska‐Pham et al, 2006; Roberts et al, 2006; Huber et al, 2009; Lee et al, 2009; Wei and Zheng, 2009). TORC1 also stimulates RNAPI in both an Sch9‐dependent and ‐independent manner, in part through regulation of the transcription initiation factor Rrn3 (Peyroche et al, 2000; Laferte et al, 2006; Huber et al, 2009). Finally, TORC1 regulates the activity of RNAPII at the large cohort of Ribi and RP genes, at least in part via Sch9 (Jorgensen et al, 2004; Urban et al, 2007).
Ribi and RP genes exhibit overlapping but non‐identical kinetics of regulation due to their differing promoter structures, which bind both common and distinct transcription factors (Ju and Warner, 1994; Hughes et al, 2000; Jorgensen et al, 2004; Wade et al, 2006; Lempiainen and Shore, 2009). Among these factors, Fhl1/Ifh1/Crf1, Sfp1 and Hmo1 have all been shown to mediate TORC1 signals to RP and/or Ribi genes (Jorgensen et al, 2004; Martin et al, 2004; Schawalder et al, 2004; Wade et al, 2004; Rudra et al, 2005; Lempiainen et al, 2009; Singh and Tyers, 2009). However, genetic analysis suggests that Sch9 functions in parallel to Fhl1/Ifh1 and Sfp1 (Jorgensen et al, 2004; Lempiainen et al, 2009), by an as yet undeciphered mechanism.
Ribi gene promoters are enriched in two distinct motifs, the PAC element (Polymerase A and C) and the RRPE element (rRNA processing element) (Hughes et al, 2000; Jorgensen et al, 2004; Wade et al, 2006). Several independent studies recently identified two Myb‐like HTH transcription factors, Dot6 and Tod6, as cognate factors for the PAC element (Badis et al, 2008; Freckleton et al, 2009; Zhu et al, 2009). Dot6 and Tod6 act as transcriptional repressors whose function is antagonized by signals from TORC1, Sch9 and PKA (Lippman and Broach, 2009). Despite its apparent lack of a DNA‐binding domain, Stb3 recognizes the RRPE element, where it can apparently act as either a transcriptional activator (Liko et al, 2007) or a transcriptional repressor (Liko et al, 2010), depending on growth context. Like Dot6 and Tod6, Stb3 appears to lie downstream of the TORC1–Sch9 axis (Liko et al, 2010).
Histone acetylation within Ribi and RP gene promoters is also regulated by TORC1. Inhibition of TORC1 by rapamycin induces both the release of the Esa1 histone acetyltransferase and the recruitment of the Rpd3 histone deacetylase at RP gene promoters, and thereby represses transcription (Rohde and Cardenas, 2003; Humphrey et al, 2004). Rpd3 is also recruited to Ribi gene promoters upon TORC1 inhibition (Humphrey et al, 2004). Importantly, in the absence of Rpd3, but not other histone deacetylases, rapamycin effectively fails to repress both RP and Ribi regulons (Humphrey et al, 2004). Rpd3 resides in two functionally distinct complexes, RPD3L and RPD3S, which are characterized by specific essential subunits such as Sds3 and Sap30 (RPD3L) and Rco1 (RPD3S) (Carrozza et al, 2005b). RPD3S is thought to suppress spurious intragenic transcription by deacetylating histones in coding regions, while RPD3L functions to repress transcription initiation when recruited to promoter regions by various DNA‐binding factors (Carrozza et al, 2005a, 2005b).
The mechanism whereby TORC1 controls the Ribi and RP regulons remains only poorly characterized. Our recent analysis of the rapamycin‐sensitive phosphoproteome revealed that Stb3, Dot6 and Tod6 become hypophosphorylated upon TORC1 inhibition in an Sch9‐dependent manner, suggesting that these factors may be the missing effectors of TORC1 in the control of RNAPII‐mediated transcription of Ribi and RP genes (Huber et al, 2009). Consistently, rapamycin treatment elicited dephosphorylation of Stb3 and Dot6 in a similar survey for novel TORC1 effectors (Soulard et al, 2010). Here, we demonstrate that Sch9 directly phosphorylates the Stb3, Dot6 and Tod6 transcription factors to abrogate the repression of Ribi and RP genes. Upon Sch9 inhibition, Stb3 and Dot6/Tod6 cooperate to recruit the RPD3L histone deacetylase complex to Ribi gene promoters, whereas Stb3 is responsible for recruitment of RPD3L to RP gene promoters. These observations consolidate the TORC1‐ and Sch9‐dependent mechanisms of transcriptional regulation of the major promoter classes during ribosome biogenesis.
Sch9 directly phosphorylates Stb3, Dot6 and Tod6 in vivo and in vitro
We first sought to determine whether the TORC1‐dependent phosphorylation of Stb3, Dot6 and Tod6 is specifically mediated by Sch9. We also included PKA in these analyses as PKA has been shown previously to phosphorylate Stb3 and Dot6 in vitro (Budovskaya et al, 2005; Deminoff et al, 2006) and this kinase often has overlapping functions with Sch9 (De Virgilio and Loewith, 2006; Zaman et al, 2008; Ramachandran and Herman, 2011). We used analogue‐sensitive alleles of SCH9 (sch9as) and PKA (pkaas: tpk1astpk2astpk3as as PKA activity is encoded by three partially redundant TPK genes in yeast) (Jorgensen et al, 2004; Yorimitsu et al, 2007), which encode a point mutation in the kinase domain that specifically renders the enzyme sensitive to inhibition by the bulky ATP analogue 1NM‐PP1 (Shokat and Velleca, 2002). We observed that HA‐tagged Stb3 analysed from untreated cells migrated as a smear in SDS–PAGE, suggesting phosphorylation on multiple sites (Supplementary Figure S1). This migration pattern was not dramatically altered by rapamycin treatment or by inhibition of Sch9as and/or PKAas with 1NM‐PP1 (Figure 1A; Supplementary Figure S1). We and others have previously observed that Sch9 and PKA share preferences for Ser/Thr residues preceded by an Arg at the −3 position and Arg/Lys at −2, that is, a consensus motif of R[R/K]x[S/T]* (Huber et al, 2009; Lee et al, 2009). We probed for potential changes in Stb3 phosphorylation using an antibody against this phosphorylated consensus motif (see Supplementary Figure S1A for antibody specificity controls). We observed rapid dephosphorylation of Stb3 after Sch9 inhibition but not after PKA inhibition, suggesting that Stb3 phosphorylation at the consensus R[R/K]x[S/T] sites is specifically regulated by Sch9 in vivo (Figure 1A). Inhibition of either Sch9 or PKA also resulted in an apparent rapid dephosphorylation of both Dot6 (Figure 1B) and Tod6 (Figure 1C). Combined inhibition of the two kinases caused a more pronounced dephosphorylation of both substrates, suggesting that Sch9 and PKA act in parallel on Dot6 and Tod6.
To determine if the three transcription factors might be direct substrates of Sch9, recombinant Stb3, Dot6 and Tod6 were tested for their ability to be phosphorylated by Sch9 in an in vitro kinase assay. A constitutively active form, Sch93E (Urban et al, 2007), but not catalytically impaired Sch9, could directly phosphorylate all three proteins in vitro (Figure 1D–G). Together, these in vivo and in vitro results suggest that Sch9 directly phosphorylates Stb3, Dot6 and Tod6.
Sch9 regulates Ribi and RP genes via Stb3, Dot6, Tod6 and the histone deacetylase RPD3L
Next, we wished to examine the biological function for Stb3, Dot6 and Tod6 phosphorylation downstream of TORC1/Sch9. Dot6 and Tod6 function as repressors of Ribi/RP transcription (Lippman and Broach, 2009) presumably via their intrinsic affinity for PAC promoter elements, which are enriched upstream of Ribi genes (Freckleton et al, 2009; Zhu et al, 2009). Stb3 also appears to function as a downstream transcriptional effector of Sch9, but curiously seems to act as both a transcriptional activator and repressor (Liko et al, 2007, 2010). To clarify the role of Stb3 and to comprehensively characterize the functional relationships of Stb3, Dot6 and Tod6 as downstream effectors of Sch9, we used mRNA‐Seq transcriptome profiles to delineate the sets of genes that respond to each factor. As expected (Jorgensen et al, 2004), inhibition of Sch9 resulted in downregulation of many genes (12% of 5025 genes detected were downregulated >1.5‐fold), among which Ribi (79% of 457 genes; P<10−31) and RP (99% of 137 genes; P<10−37) genes were highly enriched (Figure 2A). As observed in the mRNA‐Seq data and confirmed by RT–qPCR (Supplementary Figure S2A and B), this repression was partially alleviated when STB3 or DOT6 and TOD6 were deleted. This effect was not due to a prior lack of induction of the Ribi/RP genes in these deletion strains (Supplementary File F1 and data not shown). Strikingly, disruption of STB3 abrogated the repression of RP genes following Sch9 inhibition (P<10−37; Figure 2A; Supplementary Figure S2A–C; Supplementary Table S1); conversely Sch9 appeared to regulate Ribi genes preferentially via Dot6 and Tod6 (P<10−31). The combined deletion of all three factors had an additive effect, as observed by the near total failure of Sch9 inhibition to repress Ribi/RP gene expression in the stb3Δ dot6Δ tod6Δ background. In agreement with the known DNA motif preferences, Ribi genes that harboured only RRPE elements in their promoters were preferentially regulated via Stb3 (P<10−4; Figure 2A; Supplementary Figure S2C; Supplementary Table S1), whereas genes bearing only PAC elements were preferentially regulated by Dot6/Tod6 (P<0.001).
In addition to these effects, we observed hyper‐induction of Gcn4 target genes upon Sch9 inhibition in a dot6Δ tod6Δ strain and to an even greater extent in a stb3Δ strain (Supplementary Figure S2B). Although we have not pursued this effect further, we suspect that imbalanced repression of Ribi genes under these conditions may result in increased GCN4 mRNA translation, and consequently in the observed increase in expression of Gcn4 targets.
Stb3 was originally identified as a binding partner of Sin3 (Kasten and Stillman, 1997), a component of the RPD3L and RPD3S histone deacetylase complexes. Dot6 and Tod6 also bind to the RPD3L complex (Shevchenko et al, 2008). In addition, Rpd3, which is the shared catalytic subunit of these deacetylase complexes, is implicated in Ribi and RP gene regulation downstream of TORC1 (Rohde and Cardenas, 2003; Humphrey et al, 2004). The interaction of Stb3, Dot6 and Tod6 with RPD3 could thus explain their reported activities as transcriptional repressors. We, therefore, determined whether either of the RPD3 complexes function downstream of Sch9 in the regulation of Ribi/RP genes. We examined the transcriptional profile caused by Sch9 inhibition in strains that lacked Rpd3, Rco1 (a specific and essential component of RPD3S) or Sds3 (a specific and essential component of RPD3L) (Carrozza et al, 2005b). While the absence of Rco1 had virtually no effect, deletion of the RPD3L components dramatically alleviated the repression of Ribi/RP genes that would otherwise be observed upon Sch9 inhibition (Figure 2A; Supplementary Figure S2A and B). These observations strongly suggest that Stb3, Dot6 and Tod6 repress transcription of the Ribi and RP regulons via RPD3L.
Effects of Stb3, Dot6 and Tod6 on cell growth and cell size
Cells in which Ribi gene expression is compromised exhibit a reduced growth rate and a small cell size (Jorgensen et al, 2004). The absence of Sch9 activity in particular causes a severe small size (Whi) phenotype and a marked growth defect (Jorgensen et al, 2002). We determined if these phenotypes are mediated via Stb3, Dot6, Tod6 and the RPD3L complex. Our wild‐type strain had a doubling time (td) of ∼90 min and a mode cell volume of 41 fL (Figure 2B–D; Supplementary Figure S3C and D; Supplementary Table S2); inhibition of an sch9as strain with 1NM‐PP1 increased the doubling time to about 170 min and decreased mode cell volume to about 20 fL. This decreased cell size is attributable to a reduced critical size threshold at the point of commitment to division (Start), as estimated by the half height value of the daughter cell (left hand) side of the size distribution (half height daughter size for wild type=24 fL and for sch9as with 1NM‐PP1=12 fL). Deletion of STB3 in the sch9as strain strongly suppressed the slow growth phenotype (td=125 min) and partially suppressed the cell size defect (daughter size=14 fL). Importantly, the stb3Δ strain itself was wild type in size (Supplementary Figure S3C), thereby demonstrating that the size rescue effect was epistatic rather than additive. Deletion of DOT6 and TOD6 more potently rescued the size defect (daughter size=17 fL) but suppressed the slow growth phenotype to a lesser extent (td=135 min). As the dot6Δ tod6Δ strain had a somewhat larger size than wild type (Supplementary Figure S3C), the increase in size was more pronounced in the sch9as background (42% increase in sch9as versus 23% increase in wild type), again consistent with an epistatic interaction. Combined deletion of the three genes further suppressed both phenotypes (td=115 min; daughter size=20 fL), consistent with the observed synthetic transcriptional effects. As predicted, deletion of either RPD3 or SDS3 fully recapitulated the size suppression by the deletion of STB3, DOT6 and TOD6 (daughter sizes=20 and 22 fL, respectively). However, the rpd3Δ and sds3Δ mutations only slightly rescued the growth rate phenotype of sch9as cells (td=145 and 140 min, respectively), in part due to an Sch9‐independent slow growth phenotype (td=115 min in both strains). Deletion of the RPD3S subunit, RCO1, did not suppress either sch9 phenotype (td=175 min doubling time; daughter size=12 fL). Independent flow cytometric determination of total cell protein content confirmed that the suppression of cell size defects was due to differences in biomass and not due to increased vacuolarization (Supplementary Figure S3E). We note that even the strongest size suppression effects did not fully restore size to wild type, suggesting that additional Sch9 effectors likely influence the size threshold. Collectively, these results indicate that Stb3 and Dot6/Tod6 function in concert with the RPD3L complex to regulate growth rate and cell size.
Sch9 regulates RNA polymerase I and III via Stb3, Dot6 and Tod6
TORC1/Sch9 regulates rRNA transcription by RNAPI through an Rrn3‐independent pathway (Huber et al, 2009; Wei and Zheng, 2009; Philippi et al, 2010); in addition, Sch9 regulates rRNA processing (Huber et al, 2009). Since the Ribi regulon contains many genes involved in rRNA transcription and processing, and since RPs appear to have indirect roles in these processes (Reiter et al, 2011), we wondered if Sch9 indirectly promotes rRNA transcription and/or processing via the Ribi and RP regulons. We assayed RNAPI transcription by Miller chromatin spreads (Figure 3A and B) and found that Sch9 inhibition reduced the number of polymerases per 35S rRNA gene by >60%. This defect was partially suppressed by the disruption of STB3, DOT6/TOD6 or RPD3. We then performed metabolic pulse‐labelling experiments with 3H‐uracil to examine effects on rRNA synthesis and processing (Huber et al, 2009). Sch9 inhibition reduced the incorporation of radioactivity into rRNA and caused a defect in 27S–25S and 20S–18S rRNA maturation (Figure 3C). Disruption of STB3 partially rescued both effects, while disruption of DOT6 and TOD6 resulted in a relative accumulation of 27S and 20S species, suggesting that rRNA transcription was partially restored but that the processing phenotype was not. Again, deletion of all three transcription factors, or deletion of RPD3 alone, further rescued the rRNA transcription and processing defects. Interestingly, the downregulation of 5S rRNA and tRNA production upon Sch9 inhibition was also partially ameliorated by elimination of STB3, DOT6/TOD6 or RPD3L function (Supplementary Figure S4). These findings reveal that Sch9 regulates both rRNA transcription/processing and tRNA synthesis via the same cohort of Rpd3‐associated effectors that govern Ribi and RP gene expression.
Phosphorylation of Stb3, Dot6 and Tod6 regulates their repressive activities
Given that Stb3, Dot6 and Tod6 phosphorylation negatively correlated with their repressive activity (Figures 1A–C and 2A), we mutated the known Sch9 sites in each protein and assessed the phenotypic consequences. Five sites in Stb3 match the R[R/K]x[S/T] Sch9/PKA consensus motif, three of which (S254, S285 and S286) have been reported in phosphoproteomic profiles (Chi et al, 2007; Bodenmiller et al, 2008; Huber et al, 2009; Soulard et al, 2010; Stark et al, 2010). Replacing all three sites with non‐phosphorylatable Ala residues yielded a variant (Stb33A) that was not recognized by a phosphospecific antibody in vivo (Supplementary Figure S1A and B). Overexpression of STB33A but not wild‐type STB3 impaired cell growth and decreased RP gene expression (Figure 4A; Supplementary Figure S5A). While this RP defect was partially suppressed in an sds3Δ strain (Supplementary Figure S5A and B), consistent with a previous report (Liko et al, 2010), the toxicity of STB33A overexpression was not suppressed by disruption of RPD3L function (Figure 4A). Unlike previous studies (Liko et al, 2010), we detected only marginal suppression of STB33A toxicity upon disruption of the HOS2 histone deacetylase gene. Furthermore, deletion of HOS2 suppressed neither the defects in growth rate nor Ribi or RP gene expression of an sch9as strain in our genetic background (Supplementary Figure S5C and D). These negative results are consistent with the previous reports that Hos2 has virtually no influence on the broad transcriptional profile elicited by rapamycin (Humphrey et al, 2004).
We had previously mapped five phosphorylation sites in Dot6 and six sites in Tod6, all of which corresponded to the R[R/K]x[S/T] consensus except for S247 in Dot6, which still fits a minimal RxxS consensus (Huber et al, 2009). Dot6 and Tod6 variants in which these sites were substituted with Ala residues (Dot65A and Tod66A) were not as robustly phosphorylated by Sch9 in an in vitro kinase assay (Figure 1E and F). Overexpression of DOT65A and TOD66A, but not wild‐type alleles caused a severe slow growth phenotype that was largely suppressed by deletion of SDS3 (Figure 4B). Because of the severity of the DOT65A‐ and TOD66A‐associated growth phenotypes, we verified this result by conditional expression of each allele from the GAL1 promoter (Supplementary Figure S6A). On galactose medium, overexpression of the non‐phosphorylatable Dot6 and Tod6 variants again caused a strong growth phenotype, which was suppressed by deletion of RPD3 or SDS3 but not RCO1. Similar results were also obtained with a doxycycline‐inducible system (Supplementary Figure S6B and C). In agreement with the growth defects, Ribi and RP gene transcription was strongly repressed upon overexpression of the non‐phosphorylatable variants of Dot6 and Tod6 compared with their wild‐type counterparts, and was restored upon disruption of RPD3L function (Supplementary Figure S6D). Collectively, these data demonstrate that Sch9‐dependent phosphorylation antagonizes the repressive functions of Stb3, Dot6 and Tod6.
Negative feedback regulation of Tod6 and homeostatic control of Ribi/RP genes
In several assays for Tod6 phosphorylation, we noticed that Tod6 levels were decreased when the protein was hypophosphorylated (Figure 1C). Tod6–5HA levels were diminished in a sch9as strain (the sch9as allele is slightly hypomorphic; Jorgensen et al, 2004) and further decreased upon treatment with 1NM‐PP1 (Figure 5A). Similarly, the Tod66A–5HA phosphosite mutant was less abundant as compared with the wild‐type protein (Figure 5B and C). In part, this effect was due to diminished expression of TOD66A–5HA as compared with wild‐type TOD6–5HA mRNA (Figure 5C). The reduced abundance of the TOD66A–5HA transcript levels was not due to cis destabilization effects because expression of this allele also decreased the endogenous TOD6 transcript (Figure 5D). These observations are consistent with the fact that TOD6 is part of the Ribi regulon (Wade et al, 2006), with the downregulation of TOD6 upon Sch9 inhibition (Supplementary File F1), with the regulated recruitment of RPD3L to the TOD6 promoter (see below) and with the specific effects of TOD6 on Ribi gene expression (Supplementary Figure S7). In addition, the Tod66A–5HA protein/mRNA ratio was significantly decreased compared with the wild‐type ratio (Figure 5C), suggesting that Tod6 dephosphorylation may destabilize the protein. We were unable to directly test this hypothesis because the effects of both translational inhibitors and nutrient shifts on Sch9 activity (Urban et al, 2007) would confound interpretation of wild‐type Tod6 half‐life data. We conclude that Tod6 negatively represses its own transcription in a manner that is antagonized by Sch9‐dependent phosphorylation, and that Tod6 phosphorylation may promote its stability. Altogether these data suggest that Tod6 is part of a homeostatic control mechanism of Ribi gene transcription involving negative feedback loops on its expression and possibly stability.
RPD3L is recruited to Ribi/RP gene promoters by Stb3, Dot6 and Tod6
As Stb3, Dot6 and Tod6 each interact with RPD3L subunits (Kasten and Stillman, 1997; Shevchenko et al, 2008), we hypothesized that these factors may recruit the RPD3L deacetylase complex to RP and Ribi promoter DNA. Coimmunoprecipitation experiments confirmed that each factor interacted with RPD3L and, somewhat surprisingly, revealed that the interactions were not sensitive to Sch9 activity (Supplementary Figure S8A and B). We next asked if RPD3L was recruited to the relevant promoter regions upon Sch9 inhibition and whether this recruitment was dependent on Stb3 and/or Dot6/Tod6. We used a TAP‐tagged version of Sds3 for chromatin immunoprecipitation followed by high‐throughput sequencing (ChIP‐Seq). RPD3L‐associated sequences were detected using the MACS algorithm and mapped to downstream ORFs on both DNA strands (Supplementary File F2). In comparison to a mock ChIP signal obtained with untagged Sds3, 875 peaks were mapped to the promoters of 943 genes in the Sds3–TAP immunoprecipitations. Of these, 223 peaks in 271 promoters were upregulated >1.5‐fold in an sch9as strain treated with 1NM‐PP1 compared with the untagged control strain (Figure 6A; Supplementary Figure S9A and B; Supplementary Table S3). Analyses of these RPD3L ChIP‐Seq profiles revealed that RP promoters were highly enriched (P<10−9) and that Ribi promoters were enriched to a lesser extent (P<0.01). Assessment of the genetic dependencies of these profiles revealed that RPD3L peaks that depended on Stb3 were highly enriched in RP gene promoters (P<10−12), whereas the majority of RPD3L peaks in Ribi gene promoters were preferentially but not exclusively dependent on Dot6 and Tod6 (20 Ribi genes; P<0.05). The Stb3‐ versus Dot6/Tod6‐dependent RPD3L recruitment to RP and Ribi gene promoters was significantly correlated with the Stb3‐ versus Dot6/Tod6‐dependent regulation of RP and Ribi genes by Sch9 (Supplementary Figure S10; P<10−4). In the absence of all three repressive transcription factors, the RPD3L peaks at the promoters of both RP and Ribi promoters were equivalent to those in an SCH9 control strain (Figure 6A; Supplementary Figure S9C). These ChIP‐Seq data were confirmed by conventional ChIP–qPCR experiments for selected Ribi and RP genes (Supplementary Figure S11A). Overall, the influence of STB3 and DOT6/TOD6 deletions on RPD3L recruitment closely paralleled their effects on transcriptome profiles (Figure 2A; Supplementary Table S1). The combination of ChIP‐Seq, transcriptome profiling and protein–protein interaction data strongly support a model whereby Stb3, Dot6 and Tod6 serve to physically recruit the RPD3L histone deacetylase complex upstream of Ribi and RP genes.
Our finding that RPD3L was recruited to RP gene promoters almost exclusively by Stb3 was somewhat puzzling given that Stb3 was identified as a sequence‐specific partner for the RRPE element that is strongly enriched upstream of Ribi and not RP genes (Liko et al, 2007). We, therefore, located Stb3 interaction regions in the genome by repeating the ChIP‐Seq analyses with an Stb3–TAP fusion protein. Compared with RPD3L, Stb3 showed strong occupancy at fewer loci, almost all of which were upregulated upon Sch9 inhibition (635 out of 636 peaks) and coincident with the Sch9‐sensitive RPD3L peaks in Ribi and RP promoters (Figure 6B; Supplementary Figure S9A and B). The Stb3 ChIP‐Seq data were also confirmed by ChIP–qPCR for selected Ribi and RP genes (Supplementary Figure S11B). ChIP–qPCR also revealed that Dot6 interactions with the LTV1 and NOP14 promoters were strongly upregulated upon Sch9 inhibition (Supplementary Figure S11C). Tod6 also occupies these loci, but was not overtly responsive to Sch9 inhibition. This apparent lack of regulation should be interpreted with caution, however, as Tod6 levels are markedly lower in a sch9as strain and decrease further upon 1NM‐PP1 treatment (Figure 5A; Supplementary Figure S8B). In contrast to Stb3, we observed that both Dot6 and Tod6 preferentially occupied Ribi gene promoters as the ChIP–qPCR upstream of RP genes was only slightly above background (Supplementary Figure S11C).
As Stb3, Dot6 and Tod6 each showed detectable binding upstream of Ribi and RP genes, we asked whether their recruitment was interdependent. We thus examined the interactions of Stb3–TAP at the LTV1 (Ribi) and RPL9A (RP) promoters in a 1NM‐PP1‐treated sch9as strain upon DOT6 and/or TOD6 deletion. The single deletions had little or no impact on the Stb3 signal at either loci. In the dot6Δ tod6Δ background, however, Stb3 recruitment upstream of LTV1 was reduced to background levels, whereas its interaction with the RPL9A promoter was only slightly affected (Figure 6C). We then asked whether Dot6/Tod6 recruitment to these loci was dependent upon STB3. The interaction of both Dot6 and Tod6 was partially impaired at the LTV1 promoter and completely absent at the RPL9A promoter in the stb3Δ background (Figure 6D). These ChIP data suggest that upon Sch9 inhibition, Stb3 and Dot6/Tod6 bind upstream of Ribi genes in a cooperative manner to recruit the RPD3L histone deacetylase complex, while RPD3L tethering to RP gene promoters depends mainly on Stb3, with only a minor influence of Dot6/Tod6. These observations consistently correlate with the relative effects of Stb3, Dot6 and Tod6 on the transcriptional regulation of Ribi and RP genes.
We show here that the three transcriptional repressors, Stb3, Dot6 and Tod6, are directly phosphorylated by the Sch9 kinase to allow physiological regulation of the Ribi and RP regulons in response to TORC1‐mediated nutrient signals. These observations build on our previous analysis of the rapamycin‐sensitive phosphoproteome (Huber et al, 2009), and on previous genetic data that implicate Stb3, Dot6 and Tod6 in the regulation of Ribi and RP genes downstream of TORC1/Sch9 and PKA (Lippman and Broach, 2009; Liko et al, 2010). Our analysis shows that Sch9‐dependent regulation of RP gene transcription is mediated primarily by Stb3 whereas Sch9‐dependent regulation of Ribi gene transcription is mediated by Dot6, Tod6 and, to a lesser extent, Stb3. We provide a mechanistic basis for the repressive activity of the three transcription factors, namely that the histone deacetylase complex RPD3L is recruited to Ribi and RP gene promoters upon Sch9 inhibition in an Stb3‐ and Dot6/Tod6‐dependent manner. A model that encapsulates these results is presented in Figure 7.
Several observations suggest that RPD3L recruitment is the primary mechanism by which these three transcription factors repress Ribi and RP transcription: (i) Stb3, Dot6 and Tod6 interact physically with RPD3L (Supplementary Figure S8A and B) (Kasten and Stillman, 1997; Shevchenko et al, 2008); (ii) disruption of RPD3L phenocopies the derepression of Ribi and RP genes caused by deletion of STB3, DOT6 and TOD6 when Sch9 is inhibited (Figures 2A–C and 3A and B; Supplementary Figure S2A); (iii) the relative dependency on Stb3 versus Dot6/Tod6 for RPD3L recruitment to Ribi and RP gene promoters correlates with the transcriptional regulation of these regulons (Figures 2A and 6A; Supplementary Figure S10); and (iv) overexpression of non‐phosphorylatable Dot6 and Tod6 variants impairs growth in an RPD3L‐dependent manner (Figure 4B; Supplementary Figure S6A–D).
Recently, it has been suggested that Stb3 acts by recruiting the Hos2 histone deacetylase to repress transcription because deletion of HOS2 but not RPD3 suppressed the growth defect caused by Stb3 overexpression (Liko et al, 2010). However, in our strain background, deletion of HOS2 does not suppress the severe growth phenotype observed upon Stb33A overexpression (Figure 4A), nor does it alleviate the repression of Stb3‐dependent Ribi and RP genes observed upon inhibition of Sch9 (Supplementary Figure S6B). In agreement with our data, two independent studies showed that Rpd3, but not Hos2, couples TORC1 signals to Ribi and RP genes (Rohde and Cardenas, 2003; Humphrey et al, 2004). Nevertheless, it is puzzling that RPD3L disruption did not suppress the Stb33A‐induced slow growth phenotype in our strain background (Figure 4B). Although several explanations can be envisioned, at present we favour the hypothesis that Stb33A represses Ribi/RP transcription not only via recruitment of RPD3L but also by displacing and/or otherwise interfering with Fhl1/Rap1 function. This hypothesis is based on our ChIP‐Seq profiles, which indicate that RPD3L and Stb3 chromatin binding sites are very close to, if not coincident with, Fhl1 and Rap1 binding sites (Figure 6B). This observation suggests that Stb33A could effectively interfere with Fhl1/Rap1 functions, for example by antagonizing recruitment of the critical transcriptional activator Ifh1. The predictions of this model remain to be tested.
How does Sch9 regulate Stb3, Dot6 and Tod6? Sch9 was previously shown to antagonize the nuclear accumulation of Stb3 (Liko et al, 2010). Notably, unlike many other regulated transcription factor complexes, we did not detect any change in Stb3 affinity for RPD3L upon Sch9 inhibition (Supplementary Figure S8A). Nucleocytoplasmic shuttling and/or the affinity of its association with chromatin are other potential points of Stb3 regulation. In contrast to Stb3, Dot6 and Tod6 localize to the nucleus in rapidly growing cells and this localization is not apparently regulated by Sch9 (data not shown). Similar to Stb3, the interactions between Dot6 or Tod6 and RPD3L also did not seem to be altered upon Sch9 inhibition (Supplementary Figure S9B). Thus, we hypothesize that Sch9‐dependent phosphorylation regulates the ability of Dot6 and Tod6 to interact with chromatin. As the recruitment of Stb3 and Dot6/Tod6 appears to be interdependent (Figure 6C and D), these factors may physically interact at chromatin and/or alter the local chromatin environment to facilitate mutual interactions.
Delineating the precise roles of individual phosphorylation events in Stb3, Dot6 and Tod6 will be a challenge. Large‐scale mass spectrometric studies (Chi et al, 2007; Bodenmiller et al, 2008; Huber et al, 2009; Soulard et al, 2010; Stark et al, 2010) suggest that these factors are phosphorylated on >50 sites in total. Furthermore, although we could not confirm previous reports that Stb3 phosphorylation is regulated by PKA (Budovskaya et al, 2005), we did confirm that Dot6 and Tod6 phosphorylation is regulated downstream of PKA (Figure 1B and C) (Deminoff et al, 2006). Thus, like Maf1 (Huber et al, 2009; Lee et al, 2009; Wei and Zheng, 2009; Ramachandran and Herman, 2011), Dot6 and Tod6 represent a convergence point for growth regulation mediated by TORC1 and PKA.
Previously, we showed that Sch9 regulates RNAPI and III transcription (Huber et al, 2009). Here, we demonstrate that at least part of this regulation is mediated on a relatively short timescale by Stb3 and Dot6/Tod6 (30 min Sch9 inhibition; Figure 3A–C). As we have been unable to detect binding of any of these transcription factors to the rDNA locus (Supplementary Figure S11A–C), we speculate that this regulation occurs indirectly via the expression of Ribi genes, many of which encode factors involved in RNAPI/III transcription (Jorgensen et al, 2004; Wade et al, 2006). However, as deletion of STB3, DOT6 and TOD6 only partially blocks the reduction in rRNA transcription observed upon Sch9 inhibition, it is likely that another target of Sch9 involved in RNAPI regulation remains to be identified.
We also found that Stb3 influences rRNA processing. This observation was not in itself surprising as many Ribi gene products are implicated in this process (Jorgensen et al, 2004; Wade et al, 2006). However, the result that Stb3, which regulates primarily RP rather than Ribi transcription, seems to have an exclusive role in regulating 27S–25S and 20S–18S rRNA processing (Figure 3B) was unexpected. We note that RP synthesis and incorporation into the 40S and 60S ribosomal subunits is coupled with the maturation of the corresponding rRNAs (Ferreira‐Cerca et al, 2005; Poll et al, 2009; Reiter et al, 2011).
We have shown previously that Ribi influences not only growth rate but also the critical cell size threshold (Jorgensen et al, 2002, 2004). This effect of Ribi on cell size is independent from effects on protein synthesis per se (Jorgensen et al, 2004). We found that Stb3 affects the rate of cell growth more potently than Dot6/Tod6 (Figure 2B), consistent with the predominant effects of Stb3 on RP gene expression. In contrast, Dot6/Tod6 have a greater influence on cell size than Stb3 (Figure 2C), also consistent with their primary effect on the Ribi regulon. As deletion of STB3, DOT6 and TOD6 did not completely alleviate the small size caused by Sch9 inhibition, we again infer the existence of other Sch9 effectors for cell size. These observations will serve as a starting point to further dissect how the Ribi machinery communicates growth potential to the cell division machinery at Start.
Finally, our studies have uncovered an additional feedback loop in ribosome biogenesis, namely the autoregulation of Tod6 activity in the control of Ribi gene expression (Figure 5C and D; Supplementary Figures S6 and S7). This negative transcriptional feedback mechanism in principle imposes homeostatic control on Ribi (Supplementary Figure S6), and may act in concert with previously reported signalling feedback loops on Sch9 and Sfp1 (Jorgensen et al, 2004; Mnaimneh et al, 2004; Lempiainen et al, 2009). Understanding the complex interactions of these feedback loops as the primary means of establishing homeostatic control of ribosome biogenesis will be an intriguing and challenging area of future research.
Materials and methods
Yeast strains and growth assays
S. cerevisiae strains and plasmids are described in Supplementary Tables S4 and S5, respectively. Strains were constructed according to the standard protocols. Unless specified otherwise, rapamycin was used at 200 nM (from a 1‐mM stock solution in 90% ethanol, 10% Tween‐20), and 1NM‐PP1 at 300 nM (from 1 mM or 10 mM stocks in DMSO).
For growth rate assays, cells growing exponentially were diluted in the indicated media to an OD600 of ⩽0.025 and 200 μl aliquots were dispensed in transparent 96‐well plates. Wells were loaded with medium without cells to serve as reference. The plates were incubated at 30°C and OD600 was measured every 15 min for each well in a Sunrise microplate reader (Tecan, Switzerland). Reference values measured from medium alone were subtracted from all measurements. Doubling times were calculated from the slopes of linear regressions on the OD600 after Log2 transformation as a function of time once the cultures reached an OD600 of 0.2.
Cell size assays were performed on exponential phase cultures using a Z2 Coulter counter, as described previously (Jorgensen et al, 2002).
Denaturing protein extraction and immunoprecipitation
Denaturing protein extracts were performed using the TCA‐Urea method as described previously (Urban et al, 2007). For immunoprecipitation, denatured extracts were diluted 10‐fold in native lysis buffer (PBS 10% glycerol 0.5% Tween‐20) supplemented with 10 mM NaF, 10 mM p‐nitrophenylphosphate, 10 mM Na2P2O4, 10 mM β‐glycerophosphate, 1 × Roche protease inhibitor cocktail and 1 mM PMSF and cleared by centrifugation at full speed in a microcentrifuge. A total of 10 μl anti‐HA sepharose beads was added to the supernatants, incubated for 2 h at 4°C and washed 3 × with native lysis buffer. Immunoprecipitated proteins were analysed by western blotting with anti‐HA and anti R[R/K]x[S/T]* (Cell Signaling) antibodies.
Sch9 kinase assays
TAP–Sch9 variants were expressed and purified as described previously (Huber et al, 2009) except that magnetic beads coated with rabbit IgG (Invitrogen) were used instead of glutathione‐coated sepharose beads and the purified kinase was not eluted from the beads. Recombinant GST, GST–Stb3, GST–Dot6 and GST–Tod6 fusion protein variants were expressed using the pGEX6P1 system as described previously (Huber et al, 2009) except that the proteins were eluted using PBS 20% glycerol 0.5% Tween‐20 supplemented with 20 mM reduced glutathione for 15 min at room temperature. Sch9 kinase assays were performed as described previously (Huber et al, 2009).
RNA extracts were purified of genomic DNA contaminations using RNeasy kits (Qiagen). Libraries were prepared from pools of equivalent amounts of RNA from four independent experiments and sequenced using version 4 kits and single read flow cells on Genome Analyzer IIx machine according to the manufacturer's instructions (Illumina). Sequencing reads were mapped to the transcriptome of S. cerevisiae using the QSeq software with standard parameters (DNAStar). Minimal thresholds of 20 reads in each condition and 100 reads in at least one condition were applied to all transcripts. The relative expression of 5083 transcripts meeting these criteria was calculated by normalizing their reads counts to the total number of reads in each condition. A summary of the libraries’ sequencing and reads mapping to yeast ORFs is shown in Supplementary Table S6. We defined as Ribi genes the first 500 non‐RP genes in the ranking list of Wade et al (2006). Raw sequence files are available in the GEO database (http://www.ncbi.nlm.nih.gov/geo/; accession number: GSE29122).
‘Miller’ chromatin spreads were prepared according to French et al (2003). All areas of the grids containing dispersed nucleoli were photographed. The number of RNA polymerase I molecules on each visible 35S rRNA gene was counted by hand on enlarged micrographs. To be included, the chromatin strand on which the gene occurred needed to be sufficiently long to encompass the entire length of the gene.
3H‐uracil pulse‐labelling assays
3H‐uracil pulse‐labelling assays were performed as described previously (Huber et al, 2009) with the following modifications. Briefly, cells were made prototroph and grown in SC‐URA at 25°C. In all, 10 ml aliquots were removed and pulsed with 25 μCi 3H‐uracil for 20 min. Cold uracil was added at a 100‐fold molar excess and cells were grown 20 more minutes before harvest and total RNA extraction. RNA was resolved by gel electrophoresis and transferred to membranes. Its loading was controlled by ethidium bromide staining and the incorporation of 3H‐uracil was determined by exposure to phosphorimager screens.
Quantitative western blotting
The Li‐Cor infrared fluorescent system was used for quantitative western blotting. All antibodies were incubated in PBS 0.01% Tween‐20 (PBST) supplemented with 5% BSA. Washing steps were performed using PBST. After 1 h blocking in PBST 5% BSA, membranes were probed overnight with mouse anti‐HA and rabbit anti‐Hog1 antibodies. Membranes were washed three times with PBST and the primary antibodies were detected with anti‐mouse and anti‐rabbit secondary antibodies coupled to the infrared dyes IRDye800® (Rockland, PA, USA) and IRDye680® (Li‐Cor, NE, USA), respectively. After three washes, fluorescence was detected using the Odyssey® IR imaging system (Li‐Cor). The ImageJ software was used for quantification (Abramoff et al, 2004).
ChIP assays were performed and quantified by qPCR using the SYBR Green system as described previously (Bianchi et al, 2004) with slight modifications. Briefly, cells were fixed with 1% formaldehyde for 10 min at room temperature. Fixation was stopped with 125 mM glycine and cells were harvested by centrifugation. The extracted chromatin was sheared in a bioruptor sonicator (Diagenode) for 20 min (30 s on; 30 s off) at full power. IPs were performed using pan‐mouse IgG beads (Dynal, Invitrogen) and were quantified using primers (Supplementary Table S7) for the indicated loci and normalized by qPCR DNA purified from the IP input. IP efficiency was normalized with a similar quantification for the ADH4 locus as a control.
For ChIP‐Seq experiments, ChIPs were scaled up by a factor of six (240 ml of culture at an OD600 of 0.5) and chromatin was sheared in aliquots of 300 μl for 30 min instead of 20 with otherwise identical settings. ChIPs were repeated three times (untagged controls were repeated nine times). The immunopurified DNA was pooled, crosslinks were reversed overnight at 65°C and the DNA was purified using Qiagen PCR purification kits. The DNA was eluted with 40 μl of the supplied elution buffer and stored at −80°C until further analysis. Libraries were prepared using ChIP‐Seq sample preparation kits (Illumina) according to the manufacturer's instructions. DNA fragments ends were repaired using a mix of Klenow DNA polymerase, T4 DNA polymerase and T4 polynucleotide kinase. DNA was then purified and 3′ A overhangs were added using a Klenow fragment (3′–5′ exo minus). DNA was purified again and ligated to adapters. In all, 190±10 bp fragments were selected using the E‐Gel SizeSelect system (Invitrogen) and purified. Fragments with adapters were finally enriched with 18 cycles of PCR and purified. High‐throughput sequencing of libraries was performed on a Genome Analyzer IIx machine (Illumina) each in a separate channel. Sequencing reads were mapped on SGD1.01 genome assembly using Bowtie 0.12.1 (Langmead et al, 2009), with parameters ‐n 3 ‐‐best ‐‐strata ‐‐solexa1.3‐quals ‐a ‐m 20 and viewed with the USCS Genome Browser (Kent et al, 2002). Libraries sequencing and reads alignment to the yeast genome are summarized in Supplementary Table S8. The MACS algorithm v 22.214.171.124 (Zhang et al, 2008) was used to detect Sds3–TAP binding peaks by comparing the mapped read counts obtained from the sch9asSDS3–TAP versus the untagged sch9as strains with parameters ‐‐mfold=2 ‐‐P value=1e−5. Raw sequence files are available in the GEO database (GSE29124).
Supplementary data are available at The EMBO Journal Online (http://www.embojournal.org).
The authors declare that they have no conflict of interest.
Supplementary File F1
Supplementary File F2
We thank David Shore and RL laboratory members for comments on draft manuscripts. We are grateful for all the high‐throughput sequencing performed at the Genomics platform of the National Center of Competence in Research ‘Frontiers in Genetics’ and their computational analysis performed at the Vital‐IT Center for high‐performance computing of the Swiss Institute of Bioinformatics (http://www.vital-it.ch). AH was recipient of a fellowship from the Novartis Foundation for Biology and Medicine. RL is the recipient of a professorship from the Swiss National Science Foundation (PP00P3‐110770/3100A0‐108114) and receives generous support from the canton of Genève, the National Centers of Competence in Research ‘Frontiers in Genetics’ and ‘Chemical Biology’, the Leenaards Foundation, and the European Research Council (ERC‐2007‐StG 206173‐TOR signalling). ALB acknowledges funding by the NIH (GM63952). MT acknowledges support from the Wellcome Trust, a Royal Society Wolfson Research Merit Award and the Scottish Universities Life Sciences Alliance. JR is supported by the SystemsX.ch initiative and the EPFL.
Author contributions: Miller chromatin spreads were performed by SF. Cell sizes were measured by HT. SY performed western blot analyses (Supplementary Figures S5A and S6C). MS carried out western blot analysis (Figure 1B) and protein determinations by flow cytometry. MPP performed RT–qPCR measurements (Supplementary Figure S5B). High‐throughput sequencing data mapping and analysis was carried out by JR. All other experiments and analyses were carried out by AH. MT, ALB and RL contributed to experimental design and supervision. AH wrote the manuscript with the help of HT, SF, MT, ALB and RL.
- Copyright © 2011 European Molecular Biology Organization