The essential Sth1p is the protein most closely related to the conserved Snf2p/Swi2p in Saccharomyces cerevisiae. Sth1p purified from yeast has a DNA-stimulated ATPase activity required for its function in vivo. The finding that Sth1p is a component of a multiprotein complex capable of ATP-dependent remodeling of the structure of chromatin (RSC) in vitro, suggests that it provides RSC with ATP hydrolysis activity. Three sth1 temperature-sensitive mutations map to the highly conserved ATPase/helicase domain and have cell cycle and non-cell cycle phenotypes, suggesting multiple essential roles for Sth1p. The Sth1p bromodomain is required for wild-type function; deletion mutants lacking portions of this region are thermosensitive and arrest with highly elongated buds and 2C DNA content, indicating perturbation of a unique function. The pleiotropic growth defects of sth1-ts mutants imply a requirement for Sth1p in a general cellular process that affects several metabolic pathways. Significantly, an sth1-ts allele is synthetically sick or lethal with previously identified mutations in histones and chromatin assembly genes that suppress snf/swi, suggesting that RSC interacts differently with chromatin than Snf/Swi. These results provide a framework for understanding the ATP-dependent RSC function in modeling chromatin and its connection to the cell cycle.
EUKARYOTIC DNA is organized into nucleosomes and higher-order chromatin structure that inhibit initiation of transcription by RNA polymerase II in vitro and in vivo (Grunstein 1990; Kornberg and Lorch 1992). Conversion of genes from transcriptionally repressed to transcriptionally active states often occurs with changes in nucleosome positioning in vivo by replication-independent mechanisms (Crostonet al. 1991; Felsenfeld 1996; Wolffe 1996). This structural remodeling of chromatin can be accomplished by several cellular mechanisms, including modification of histones, the binding of gene-specific transcriptional factors, and the activity of protein complexes capable of reconfiguring nucleosomes. Several eukaryotic multiprotein complexes have been implicated in such chromatin-remodeling processes, including the Saccharomyces cerevisiae Snf/Swi (Winston and Carlson 1992; Carlson and Laurent 1994; Peterson and Tamkun 1995; Kingstonet al. 1996), mammalian BRG1 and hbrm (Imbalzanoet al. 1994; Kwonet al. 1994; Wang et al. 1996a,b), and Drosophila NURF, CHRAC, ACF, and BRM complexes (Dingwallet al. 1995; Tsukiyamaet al. 1995; Itoet al. 1997; Mizuguchiet al. 1997; Varga-Weiszet al. 1997).
The yeast SNF and SWI genes were identified originally as mutants defective in the expression of specific genes (Neigeborn and Carlson 1984; Sternet al. 1984) and have since been implicated in the transcription of several differently regulated promoters (Abramset al. 1986; O’Haraet al. 1988; Estruch and Carlson 1990; Laurentet al. 1990; Happelet al. 1991; Peterson and Herskowitz 1992), although they are not required for transcription of all genes. The Snf/Swi proteins were also shown to assist activation by several sequence-specific transcriptional activators expressed in S. cerevisiae (Laurent and Carlson 1992; Peterson and Herskowitz 1992; Yoshinagaet al. 1992). The yeast Snf/Swi complex is composed of 11 proteins and has an apparent mass of 2 MD (Cairnset al. 1994; Petersonet al. 1994). Genetic evidence provided the first clue that the Snf/Swi proteins function in transcriptional activation by relieving chromatin-mediated repression. Growth and transcriptional defects of snf/swi mutants are suppressed by mutations in genes encoding histones H2A and H2B (Hirschhornet al. 1992), H3, H4 (Prelich and Winston 1993; Krugeret al. 1995), nonhistone chromatin assembly factors, including Spt4p, Spt5p, and Spt6p (Neigebornet al. 1987; Sternberget al. 1987; Kruger and Herskowitz 1991; Swanson and Winston 1992; Hirschhornet al. 1995; Bortvin and Winston 1996), and an HMG1-like protein (Kruger and Herskowitz 1991). The chromatin structure of the SUC2 promoter is also altered in snf2, snf5, and swi1 mutants (Hirschhornet al. 1992; Matallanaet al. 1992; Gavin and Simpson 1997; Wu and Winston 1997). Biochemical evidence for a direct role in nucleosome remodeling was provided by experiments in which purified yeast Snf/Swi complex was shown to disrupt histone-DNA contacts enabling activators to bind to nucleosomal sites in vitro in an ATP-dependent manner (Côtéet al. 1994).
The yeast Snf2p protein contains motifs similar to those of DNA-dependent ATPases and DNA helicases (Gorbalenyaet al. 1989; Laurentet al. 1992; Henikoff 1993) and is the prototypal member of a large protein superfamily. Representatives of the superfamily include prokaryotic and eukaryotic proteins that participate in a variety of nuclear processes, including transcription, chromosome segregation, and DNA repair and recombination (see Carlson and Laurent 1994). Phylogenetic analysis of proteins within this superfamily generates several subfamilies believed to have similar functions, one of which includes in addition to Snf2p, the S. cerevisiae Sth1p, Drosophila melanogaster brm, and human hbrm and BRG1 (brm/SWI2 related gene) proteins (Eisenet al. 1995). Importantly, recombinant Snf2p has a DNA-dependent ATPase activity (Laurentet al. 1993), and purified yeast Snf/Swi is capable of a similar ATP hydrolysis activity (Cairnset al. 1994; Petersonet al. 1994). Moreover, the nucleosome restructuring activities of the yeast Snf/Swi and human hSWI/SNF complexes require the Snf2p and BRG1 ATPases, respectively (Côtéet al. 1994; Kwonet al. 1994), and the Drosophila brm protein in the BRM complex and the Drosophila ISW1 protein present in the NURF, ACF, and CHRAC ATP-dependent nucleosome remodeling factors are likewise believed to contribute the necessary ATP hydrolysis activity (Tsukiyamaet al. 1995; Tsukiyama and Wu 1995; Itoet al. 1997; Varga-Weiszet al. 1997). Snf2p and its relatives therefore provide several independent chromatin-modeling complexes with the ATP-dependent enzymatic activity required to rearrange nucleosomal DNA.
In addition to transcription by RNA polymerase II, dynamic chromatin-remodeling activities are important in several other basic cellular processes that have been conserved during eukaryotic evolution, any of which might require multiprotein complexes with activities similar to those of the Snf/Swi, BRG1, hbrm, Brm, NURF, ACF, or CHRAC complexes. These processes include transcription by RNA polymerases I and III, the replication, repair, and recombination of DNA, epigenetic regulation, retroelement integration, and the assembly and maintenance of stable mitotic chromosomes. Interestingly, the human counterparts of Snf/Swi have been shown to associate with nuclear matrix or scaffold (Reyeset al. 1997). Moreover, in addition to their roles in remodeling chromatin, ACF and CHRAC functions are also necessary for the assembly and proper spacing of nucleosomes, implying roles in DNA replication (Itoet al. 1997; Varga-Weiszet al. 1997).
We have initiated studies with yeast Sth1p/Nps1p (Laurentet al. 1992; Tsuchiyaet al. 1992), the most closely related protein to Snf2p in the yeast genome. Sth1p shares several regions of homology with Snf2p including a 518-amino-acid ATPase/helicase-related domain, a 75-amino-acid bromodomain of unknown function, which is present in the Snf2p subfamily proteins and in several other proteins (Hayneset al. 1992; Laurentet al. 1992; Tamkunet al. 1992; Jeanmouginet al. 1997) and four shorter regions flanking the ATPase domain (see Figure 4A). The ATPase/helicase-related domains of Snf2p and Sth1p are interchangeable (Laurentet al. 1993). There are, however, functional differences between SNF2 and STH1. STH1 is essential for mitotic growth of yeast cells, while SNF2 is not. Increased gene dosage of either SNF2 or STH1 fails to compensate for loss of the other and a DNA-bound LexA-Snf2p hybrid protein activates transcription of a target gene dramatically, whereas a LexA-Sth1p fusion protein activates transcription only very weakly (Laurentet al. 1992). In addition, STH1 depletion experiments suggest a role for Sth1p in mitotic cell cycle progression (Tsuchiyaet al. 1992).
Significantly, Sth1p was shown to be a component of an abundant 16-protein complex with the ability to remodel the structure of chromatin (RSC) in vitro (Cairnset al. 1996). RSC shares several features with the Snf/Swi complex. First, RSC and Snf/Swi alter nucleosome structure in vitro, as measured by changes in DNAse I digestion patterns. Second, both remodeling activities require ATPase activities that are stimulated equally well by single-stranded, double-stranded, or nucleosomal DNA. Finally, at least four subunits of RSC are essential homologs of the Snf2p, Snf5p, Swi3p, and Swp73p components of Snf/Swi (Cairnset al. 1996; Caoet al. 1997). However, in contrast to Snf/Swi, RSC is essential and abundant, suggesting that the two complexes carry out distinct functions.
Here we take biochemical and genetic approaches to elucidate the essential function(s) of the yeast Sth1p. We show that an Sth1p fusion protein expressed in yeast has DNA-stimulated ATPase activity, which is abolished by mutation of the nucleoside triphosphate (NTP)-binding site, the same mutation that previously eliminated STH1 function in vivo. To study the phenotype caused by loss of STH1 gene function, three temperature-sensitive mutations were isolated. All mutations map to the highly conserved ATPase/helicase-related domain: one allele causes cells to arrest in G2/M, and the other two cause cells to arrest asynchronously, suggesting additional essential roles of Sth1p in the cell. We also demonstrate that the C-terminal bromodomain motif of Sth1p is required for wild-type STH1 function. The aberrant morphology of cells lacking portions of the bromodomain suggests that these sequences are required for a discrete function. Finally, genetic interactions between an sth1 mutation and previously identified mutations affecting nucleosome structure suggest that RSC contacts chromatin in a manner that is different from that of Snf/Swi.
MATERIALS AND METHODS
Strains, genetic methods, and media: Saccharomyces cerevisiae strains are listed in Table 1. The Escherichia coli strains were XL1-Blue (Stratagene, La Jolla, CA) and GM48 (F- thr leu thi lacY gelK gelT ara fhuA tsx dam dcm supE44). Rich (YPD), synthetic complete (SC), and sporulation media were prepared as described previously (Roseet al. 1990). YPgal, YPsuc, YPraf, YPeth/gly, and SCgal media were 2% in galactose, sucrose, raffinose, ethanol and glycerol, and galactose, respectively. For the analysis of temperature-sensitive strains, logarithmic-phase cultures growing at 30° (the permissive temperature) were shifted to 37° (the fully restrictive temperature). Yeast strains were transformed by the lithium acetate procedure (Itoet al. 1983). Standard genetic procedures of crossing, sporulation, and tetrad analysis were followed (Roseet al. 1990).
Preparation of GST-Sth1p fusion proteins for ATPase assays: The pDJ84, pDJ84KR, pDJ85, and pDJ86 plasmids (Table 2) expressed GST-Sth1288-1359, GST-Sth1K501R288-1359, GST-Sth1288-1014, and GST-Sth1K501R288-1014 fusion proteins, respectively, from the GAL1,10 promoter in yeast strain BLY36. Cells were grown in SC media lacking uracil and expression of the fusion proteins was induced by addition of 4% galactose. Exponentially growing cells at 30° were harvested by centrifugation, washed with sorbitol buffer [0.3 m sorbitol, 0.1 m NaCl, 5 mm MgCl2, and 10 mm Tris-HCl (pH 7.4)] and resuspended in lysis buffer (sorbitol buffer containing 2 mm phenylmethylsulfonyl fluoride, 0.1 mm leupeptin, 0.1 mm aprotinin, and 0.001 mm pepstatin A). All steps were performed at 4° according to the protocol described by Mitchell et al. (1993). Cells were broken by vortexing with glass beads, and protein concentration determined using a colorimetric assay (BioRad, Richmond, CA) with bovine serum albumin as a standard. A 50% (w/v) slurry of glutathione agarose beads (Sigma, St. Louis) was added to protein extracts at a ratio of 10 μl slurry per milligram of protein. The solution was incubated for 30 min at 4° on a nutator. Proteins adsorbed to the beads were washed three times with sorbitol buffer including 1% NP-40 and once with sorbitol buffer alone. Adsorbed GST fusion proteins were released from the beads by elution buffer [10 mm glutathione, 50 mm Tris-HCl (pH 7.5), 5% glycerol, 5 mm DTT] twice at room temperature for 2 min. Beads were centrifuged at 8160 × g for 1 min in a microcentrifuge and the supernatants were collected and pooled. Proteins were analyzed by 6% SDS-polyacrylmide gel electrophoresis. As judged by Tris-Glycine polyacrylamide gel electrophoresis, the eluted GST-Sth1p fusion proteins constituted >85% of the total protein.
ATPase assay: The hydrolysis of ATP was measured as the release of inorganic phosphate, 32Pi, from [γ-32P]ATP. The standard assay mixture (20 μl) contained 4 μl of reaction buffer A [50 mm MgCl2, 250 mm NaCl, 25% glycerol, 0.1 mg/ ml bovine serum albumin, 5 mm dithiothreitol, 5 mm ATP, 100 mm HEPES (pH 7.0)], 1 μCi of [γ-32P]ATP (6000 Ci/ mmol; Amersham Corp., Arlington Heights, IL), 9 μl of GST-Sth1p fusion protein (or an equal volume of the vector-only control) immobilized on agarose beads in sorbitol buffer or 9 μl of GST-Sth1p fusion protein in elution buffer (or an equal volume of the eluate from the vector-only control), and 6 μl of H2O or nucleic acid.
After 30 min at 37°, the reaction was terminated by the addition of 1 μl of 0.5 m EDTA. A sample (1 μl) of the reaction mixture was then spotted onto a poly(ethyleneimine)-cellulose thin-layer chromatography plate (J. T. Baker, Phillipsburg, NJ or Selecto Scientific), and ascending chromatography was carried out in 0.8 m LiCl and 0.8 m acetic acid. The amount of 32Pi formed was quantitated using a phosphorImager (Molecular Dynamics, Sunnyvale, CA) and normalized to the total amount of radioactivity applied to each lane. Double-stranded (ds) DNA was prepared from pBluescript KS- (Stratagene). Partially purified nucleosomal TRP1-ARS1-derived plasmid DNA (Rothet al. 1990) was also tested. Correctly positioned nucleosomes on the chromatin episome were confirmed by micrococcal nuclease digestion and indirect end-label analysis (B. C. Laurent, unpublished results). Alternatively, the hydrolysis of ATP was measured by a colorimetric assay (Lanzettaet al. 1979).
Disruption of the chromosomal STH1 locus: Diploid strains BLY29 and BLY18 were transformed to histidine independence with the 2.3-kb EcoRI-XbaI fragment of pDJ63 and the 4.1-kb EcoRI-BamHI fragment of pDJ63-2 to replace the wild-type STH1 genes with sth1-Δ2::HIS3 and sth1-Δ3::HIS3 alleles, creating BLY30 and BLY81, respectively. The correct structures of gene replacements were confirmed by Southern blot analysis.
Isolation of temperature-sensitive alleles of STH1: The centromere-containing plasmid pDJ64 (TRP1) carrying the wild-type STH1 gene was mutagenized in vitro by incubating 12 μg of plasmid DNA in a solution of 1 m hydroxylamine (Sigma), 50 mm sodium pyrophosphate (pH 7.0), 100 mm NaCl, and 2 mm EDTA for 1 hr at 75° as described (Sikorski and Boeke 1991). The solution was extracted twice with phenol:chloroform:isoamylalcohol (25:24:1) and precipitated twice with ethanol, and the excess hydroxylamine removed by gel filtration through a small Sephadex G-50 column. Mutagenized plasmid DNA in 10 mm Tris-HCl (pH 8.0), 1 mm EDTA was then used directly to transform BLRY30-1, a yeast strain carrying a deletion of the chromosomal STH1 gene and a copy of the wild-type STH1 gene on a second centromere-containing plasmid, pDJ67 (URA3; Sikorski and Boeke 1991). Approximately 6000 transformants grown selectively for both plasmids were replicated to supplemented synthetic complete (SC) medium containing 5-fluoroorotic acid to counterselect against the plasmid carrying the wild-type STH1 gene. Transformants (2000) carrying only the mutagenized plasmid were screened for conditional growth (lethality at 37°). A total of 300 transformants showing a temperature-sensitive (Ts-) phenotype were retested for conditional growth on nonselective (YPD) plates to eliminate cells bearing mutations in the plasmidborne selectable marker gene. Forty-six mutants were conditional for growth also on YPD plates. Plasmid DNA was recovered from 10 of these colonies. Restriction fragment swapping experiments confirmed that three of the mutations are located within STH1. Plasmids p376-1-12, p14-1-7, and p137-1-10 from these temperature-sensitive strains were digested with NheI and XbaI to recover a 5.3-kb fragment containing the sth1-1ts, sth1-2ts, and sth1-3ts mutant alleles, respectively. These fragments were cloned into the integrating vector pRS306 (URA3), creating psth1-1ts, psth1-2ts, and psth1-3ts, which were then used to replace the wild-type STH1 locus by transformation and subsequent growth on SC 5-fluoroorotic acid plates. Temperature-sensitive and Ura- colonies were analyzed by genetic complementation to confirm integration of the sth1-ts alleles; pDJ67 (STH1 CEN6) and pBL50 (STH1 2 μm) complemented the growth defect at 37°. Genetic crosses to an isogenic STH1 strain confirmed that all the mutations are recessive. p14-1-7 was digested with EcoRI and SmaI and the recovered 6.4-kb fragment was cloned into pRS316 to create pDJ2ts.
The locations of each mutation were determined by interchanging fragments of the wild-type STH1 gene with homologous fragments of the mutant allele. The 1.1-kb BglII-AgeI fragments of p376-1-12 and p14-1-7 and the 1-kb AgeI-StuI fragment of p137-1-10 were subcloned into pUC19 and sequenced by the dideoxy-chain termination method (Sangeret al. 1977) with a Sequenase 2.0 kit (United States Biochemical, Cleveland). As confirmation, “hot start” PCR amplification (Chouet al. 1992) using ampliwax PCR gems (Perkin Elmer, Norwalk, CT) was first performed directly on genomic DNA isolated from the sth1-1ts and sth1-3ts strains or from psth1-1ts, psth1-2ts, and psth1-3ts with primers MK011 (5′-TCAATTACGAGGTTTAGAATGGATG-3′) and MK012 (5′-TTTGTCCAATTCTGTGAGCTCTATC-3′). The PCR products were purified by Wizard PCR preps (Promega, Madison, WI) and then analyzed directly by thermal cycle sequencing using SequiTherm DNA polymerase (Epicentre Technologies, Madison, WI) and primers MK011, MK012, and MK016 (5′-ATGCTCAATCGAAGTTATCATTCAC-3′) to detect the sth1-1ts, sth1-3ts, and sth1-2ts lesions, respectively. Allele-specific PCR amplification (Aultet al. 1993) of genomic DNA from BLY48 (sth1-2ts) using primers MK013 (5′-CGTTAATTCTAGTTTCTCTTGGGTACC-3′) and MK030 (5′-GCACTGTTGAACTTTGTT TTGCT-3′; the C to T substitution mutation is set boldface) confirmed the presence of the sth1-2ts mutation. The nucleotide changes and predicted amino acid changes are as follows: sth1-1ts is C-1514 to T, S505F; sth1-2ts, C-1937 to T, P646L; and sth1-3ts, C-2417 to T and C-2642 to T, S806L and T881M, respectively.
Budding morphology: To determine the mitotic index, cells grown to midlog phase at 30° were diluted to a density of 4 × 106 cells/ml into YPD medium prewarmed to 37°. We found that the phenotype of cells diluted first into prewarmed YPD and then shifted to 37° was more dramatic than that of cells shifted directly to 37°. At several time points, aliquots were removed, fixed with 3.7% formaldehyde in PBS [100 mm NaCl (80 mm Na2HPO4, 20 mm NaH2PO4, pH 7.3)], and the number of single, small-budded, and large-budded cells was determined microscopically.
DNA flow cytometry: Early- to midlog phase cultures of temperature-sensitive and isogenic wild-type strains grown in YPD medium with 2% glucose at 30° were split and diluted into prewarmed 30° or 37° YPD media to concentrations of approximately 4 × 106 cells/ml. At various time intervals, approximately 2 × 107 cells were removed, sonicated briefly, and fixed in 70% ethanol at room temperature for 12 hr.
After two washes in 50 mm Tris-HCl (pH 7.8), fixed cells were resuspended in 1 ml of the same buffer containing 2 mg/ml RNAse A and incubated at 37° for 12 hr on a nutator. Cells were pelleted, resuspended in 0.5 ml of 55 mm HCl containing 5 mg/ml pepsin (Sigma), and incubated at 37° for 30 min. Cells were washed once in 1 ml of TMN buffer [200 mm Tris-HCl (pH 7.5), 211 mm NaCl, and 78 mm MgCl2] and then resuspended in TMN buffer containing 50 μg/ml propidium iodide (Sigma). The fluorescence intensities of stained cells were measured after 10-30 min on a Becton Dickinson (San Jose, CA) 900 FACScan machine.
RNA analysis: Cells were grown in YPD liquid media to a concentration of 0.5-1.0 × 108 cells/ml, and total RNA was isolated using RNeasy spin columns according to the manufacturer’s protocol (QIAGEN Inc., Chatsworth, CA) and quantitated by measuring absorbence at 260 nm. RNAs (20 μg) were size fractionated in a 1.5% agarose gel containing formaldehyde and transferred to a GeneScreen membrane (DuPont/New England Nuclear, Boston) and UV cross-linked. The filter was successively hybridized to 32P-labeled probes prepared by PCR amplification of CLB2pic19, genomic (S288C) DNA, or FC204 primed by the primer pairs CLB2-1 (5′-GGCGTT GGATCAAGCACTGAGGTAG-3′) and CLB2-2 (5′-GCCCCT CTTCTCATTCATGCAAGGTC-3′), CSE4-1 (5′-GAAGATCA CTCAGTAACGTCAACAGGC-3′) and CSE4-2 (5′-GCGGAGA GCACTTGTCGAACC3′), and CLN3-1 (5′-ATGGCCATATTGAAGGATACC-3′) and CLN3-2 (5′-CAGCGAGTTTTCTTG AGGTTGCTAC-3′) to generate PCR fragments of CLB2, CSE4, and CLN3, respectively. All probes were labeled by random priming. Filters were hybridized in 5× SSC, 50 mm sodium phosphate (pH 7.0), 1× Denhardt’s solution, 0.1% SDS, 100 μg/ml denatured salmon sperm DNA, and 5% formamide at 42° for 12-16 hr, followed by three 15-min washes with 0.1× SSC and 0.1% SDS at 50°.
Immunofluorescence microscopy: Cells grown to early- or midlog phase were fixed with formaldehyde for 2 hr and processed for indirect immunofluorescence microscopy as described (Pringleet al. 1991). Microtubule structures were observed by staining with YOL1/34 rat monoclonal anti-yeast α-tubulin primary antibody (Sera Lab) diluted 1:100 and fluorescein isothiocyanate-conjugated goat anti-rat immunoglobulin G secondary antibody (Sigma) diluted 1:100. DNA was visualized by staining with 4′,6-diamidino-2-phenylindole (DAPI). Epifluorescence and Nomarski differential interference contrast photomicrography were performed with a Nikon Labophoto-II photomicroscope.
Sth1p has DNA-stimulated ATPase activity: To determine whether Sth1p has ATPase activity, the conserved C-terminal 1072 amino acids of the protein including the central 518-amino-acid ATPase/helicase-related domain (Laurent et al. 1992, 1993) were expressed in yeast. As a negative control, an analogous mutant fusion protein containing a single amino acid substitution (K501R) within the NTP-binding motif (Walkeret al. 1982), which abolishes STH1 function in vivo (Laurentet al. 1992), was also expressed.
The wild-type protein preparation exhibited ATPase activity with a specific activity of 355 pmol of inorganic phosphate (Pi) released per microgram per minute in the absence of added nucleic acid, an activity that was sixfold higher than that of the mutant protein or the GST moiety alone (Figure 1). The hydrolysis activities associated with GST and GST-Sth1K501R preparations are probably due to the presence of copurifying contaminants. Partially purified nucleosomal DNA and dsDNA stimulated the ATPase activity modestly (Figure 1). The ATPase activity of GST-Sth1p in the absence of added DNA suggests that this preparation contains nucleic acid. Prior treatment of a preparation of GST-Sth1p with DNase I did not significantly reduce the ATPase activity, nor did addition of dsDNA stimulate GST-Sth1p activity following DNAse I treatment (data not shown), although we cannot rule out the possibility that DNA bound by GST-Sth1p is inaccessible to DNase I. The specific activity of the Sth1p ATPase, 124 pmol Pi/pmol protein/min in the presence of DNA, is comparable to that of the purified Snf/Swi (Cairnset al. 1994; Côtéet al. 1994) and RSC (Cairnset al. 1996) complexes, with an optimal NaCl or KCl concentration of 50-150 mm, an optimal MgCl2 concentration of 1-2 mm, and an optimal pH range of pH 6.0-8.0. The finding that none of the other putative components of RSC has homology to DNA-dependent ATPases (B. R. Cairns, personal communication) together with the similarities of the biochemical properties of the Sth1p ATPase to those of purified RSC, strongly suggest that Sth1p provides RSC with the ATP hydrolysis activity.
In addition, a GST-Sth1p fusion protein expressing the ATPase-related domain alone possessed ATPase activity comparable to that of GST-Sth1 (data not shown), suggesting that the C-terminal 346 amino acids, including the bromodomain, are dispensable for ATP hydrolysis activity in vitro. In summary, the in vitro biochemical data agree with the genetic evidence that the K501R mutation abolishes STH1 function in vivo (Laurentet al. 1992), indicating that the ATPase activity is required for RSC function in vivo.
Temperature-sensitive mutations in STH1 map to the conserved ATPase/helicase-related region: To characterize the phenotype of cells defective in STH1, temperature-sensitive mutations were generated by random mutagenesis of the cloned gene. Three independent temperature-sensitive sth1 mutations were identified in a screen of ∼2000 transformants using the plasmid shuffle technique (Sikorski and Boeke 1991). The integrated alleles segregated 2+: 2- for growth at 37° in crosses to a wild-type strain. All of the mutations are recessive and were complemented by plasmids containing the wild-type STH1 gene.
The three mutants, sth1-1ts, sth1-2ts, and sth1-3ts, displayed a range of different growth phenotypes at permissive (30°) and nonpermissive (37°) temperatures relative to the isogenic wild-type strains (Figure 2A). None of these sth1-ts mutants appears to show a first cycle arrest and the sth1-2ts and sth1-3ts alleles are more severe than the sth1-1ts allele (Figure 2A). Significantly, all of the mutations alter amino acids in the highly conserved ATPase/helicase-related domain and these residues lie within or between the signature helicase motifs (Figure 2, B and C; Gorbalenyaet al. 1989; Bork and Koonin 1993; Henikoff 1993). Each of the amino acids mutated is highly conserved among all members of the Snf2 protein subfamily, further indicating the functional importance of conserved sequences within the ATPase-related domain.
sth1-1ts and sth1-3ts mutants incubated at 37° expressed Sth1p levels that were similar to those at 30°, and sth1-2ts mutants expressed Sth1p levels that were only three- to fivefold reduced compared to Sth1p levels at 30° (data not shown). Sth1p levels in the three sth1-ts mutants grown at 30° were comparable to those in wild-type cells grown at 30° or 37°. These results suggest that the arrest phenotypes of sth1-ts mutants reflect temperature-dependent conformational changes induced at the nonpermissive temperature, rather than proteolysis of the mutant proteins.
The Sth1-1, Sth1-2, and Sth1-3 proteins have wild-type ATPase activity in vitro: Because each of the sth1-ts alleles alters amino acids within the conserved ATPase/helicase-related domain, which alone is sufficient for ATP hydrolysis in vitro, we tested whether the mutations affect ATP hydrolysis activity. The specific activities of partially purified preparations of GST-ATPase/helicase-related domain fusion proteins expressing the S505F, P646L, S806L, and T881M substitutions were 120, 115, and 166 pmol Pi/pmol protein/min, respectively, comparable to that of the wild-type GST-Sth1p (124 pmol Pi/pmol protein/min). This result indicates that these mutations do not affect the ability of Sth1p to hydrolyze ATP in vitro and suggests further that this domain carries out other functions in addition to ATP hydrolysis.
sth1-ts mutants exhibit cell cycle and non-cell cycle phenotypes: To determine whether sth1-ts mutants show morphological abnormalities, cultures of logarithmically growing sth1-1, sth1-2, and sth1-3 temperature-sensitive mutants were diluted into prewarmed 30° or 37° YPD and examined microscopically at several time intervals. One of the mutants, sth1-3ts, exhibited a dramatic increase in the number of large-budded cells when grown at the nonpermissive temperature. At least 75% of sth1-3ts cells arrested with a budded phenotype and of these, 60% were large budded (Figure 3A), suggesting a block in mitosis specifically at the G2/M transition. These large-budded cells exhibited a heterogenous morphology: approximately half contained a divided nucleus, and half contained nuclei that were bilobed and extended across the bud neck (Figure 3B), characteristic of cells in G2 or M phase. This suggests that in a large fraction of sth1-3ts mutant cells the DNA was replicated (the nucleus had divided) but that cell division had not occurred. Indirect immunofluorescence microscopy of anti-α-tubulin-stained sth1-3ts mutant cells showed shortened intranuclear microtubule spindles compared to sth1-3ts cells grown at 30° (Figure 3B) or to wild-type cells at 30° or 37° (data not shown), indicating that poleward elongation of microtubule spindles does not occur in sth1-3ts cells. Examination of the DNA content of sth1-3ts cells by flow cytometry confirmed a perturbation of the cell cycle distribution (Figure 3C). Together, these results suggest that STH1 is required for progression of cells through the cell cycle, specifically through the G2/M phase transition.
To compare the phenotype exerted by the sth1-3ts allele to the phenotype caused by an sth1 null allele in an isogenic background (S288C), we transformed diploid strain BLY81 (sth1-Δ3/STH1 ura3/ura3) with a URA3 centromere-derived plasmid carrying the wild-type STH1 gene expressed from the inducible GAL10 promoter. Following 12 hr of growth in 2% glucose, 73% of an sth1-Δ3 haploid strain carrying the GAL10-STH1 plasmid for viability arrested with a budded phenotype (data not shown), a bud morphology bias very similar to what we observed following a shift of the conditional sth1-3ts mutant to the nonpermissive temperature. A previously described nps1/sth1 deletion allele also resulted in G2 arrest, in STH1 depletion experiments of similar design (Tsuchiyaet al. 1992). This result indicates that the phenotype caused by the sth1-3ts mutation is indistinguishable from that caused by an sth1 null mutation in cells of the S288C genetic background.
In contrast to the sth1-3ts allele, the sth1-1ts and sth1-2ts mutations arrested cells at random positions throughout the cell cycle and flow cytometric analysis of the arrested mutants confirmed the random cell type distribution (data not shown). The similar growth defects of sth1-3ts and sth1-2ts mutants argue against the lesser penetrance of the temperature-sensitive defect in sth1-2ts mutants. In addition, the primary defects of the sth1-1ts and sth1-2ts mutations do not appear to affect synthesis of the gene product. We favor the model that the sth1-1ts and sth1-2ts mutations are defective in additional essential function(s) of Sth1p required throughout the cell cycle, although it is possible that Sth1p has instead a single important cellular function at G2/M and that the sth1-1ts and sth1-2ts are leaky hypomorphic alleles.
The sth1S806L and sth1T881M mutations function synergistically: To further analyze the nature of the sth1-3ts temperature-sensitive growth defects and to determine the relative contributions of each point substitution to the phenotype, we constructed centromere (URA3) plasmids bearing each of the two sth1-3ts point mutations, designated psth1S806L and psth1T881M, and a plasmid containing both of the mutations, psth1S806L,T881M. Diploid strain BLY81 (sth1-Δ3/STH1 ura3/ura3) was transformed with each of these plasmids, and Ura+ transformants were sporulated and subjected to tetrad dissection to recover sth1-Δ3 haploid strains expressing these plasmids. sth1T881M cells were moderately temperature-sensitive (Ts-) for growth, but much less so than sth1S806L,T881M cells (Figure 3D), which had a terminal arrest morphology similar to that of arrested sth1-3ts cells. Significantly, sth1S806L cells grew as well as cells carrying the wild-type STH1 gene at both temperatures (Figure 3D). These results suggest that the two mutations function synergistically to confer the growth defects of sth1-3ts cells.
The bromodomain of Sth1p is required for wild-type function: The presence of several highly conserved sequences suggests that Sth1p contains discrete functional domains. In addition to the ATPase/helicase-related domain, the evolutionarily conserved bromodomain at the C terminus is a candidate for one such domain. The functional importance of the bromodomain, which extends from amino acids 1274 to 1359, including a 22-amino-acid extension predicted to form a third α-helix, here designated α3 and recently called helix C (Jeanmouginet al. 1997; Figure 4A), was investigated by transforming diploid strain BLY81 (sth1-Δ3/STH1 ura3/ura3) with centromere (URA3) plasmids pDJ66, pIN11, pIN14, and pIN15 carrying the sth1Δ1338, sth1Δ1349, sth1Δ1274-1335, and sth1Δ1274 deletion alleles, respectively (Figure 4A). Ura+ transformants were sporulated and subjected to tetrad analysis to recover sth1-Δ3 spore clones carrying sth1 or STH1 plasmids for viability. All sth1 segregants were thermosensitive for growth on YPD, although the mutants displayed a range of temperature sensitivity (Figure 4, A and B). Segregants carrying sth1Δ1338 and sth1Δ1349 were extremely thermosensitive and failed to grow at 37°, and those carrying sth1Δ1274-1335 exhibited moderate Ts- growth defects. sth1Δ1274 segregants displayed Ts- defects intermediate between those of sth1Δ1338/sth1Δ1349 and sth1Δ-1274-1335 (Figure 4, A and B).
To understand the nature of the Ts- growth defects, cells were analyzed microscopically for morphological differences after a shift to 37°. sth1-Δ3 cells carrying sth1Δ1338 or sth1Δ1349 exhibited an extremely tight growth arrest and exhibited an unusual morphological defect. Eighty-one percent of the arrested sth1Δ1338 cells have a budded phenotype: 11% are small budded, 39% are large budded, and 31% have unusual, highly elongated buds (Figure 4C). Approximately one half of the population of elongated cells is multiply budded at 37° (see Figure 4C). DAPI staining revealed that the majority of cells with elongated buds contained stained nuclei at the mother bud neck (Figure 4C), indicating a delay in nuclear division. sth1Δ1349 cells displayed a similar, although less pronounced terminal arrest morphology (data not shown). sth1 cells carrying sth1Δ1338 or sth1Δ1349 grown permissively were larger than sth1 cells expressing wild-type Sth1p but did not contain abnormal elongated buds (Figure 4C; data not shown). Importantly, despite the thermosensitivity of cells lacking Sth1p amino acids 1274 to 1335 or 1274 to 1359, these cells do not arrest their growth at the nonpermissive temperature and exhibit nearly wild-type budding morphologies, with no elongated or multiple buds (data not shown).
To test the possibility that the arrest of sth1Δ1338 and sth1Δ1349 cells reflects a block in the cell cycle at G2 or M phases, these cells were examined by flow cytometry at several time points after a shift from 30° to 37°. The majority of sth1Δ cells carrying either sth1Δ1338 or sth1Δ1349 contained a 2C DNA content (Figure 4D). Even at the permissive temperature, a significant proportion of the mutants were large budded and had G2/M DNA contents (Figure 4, C and D), indicating a delay in mitosis. Thus, these sth1 cells undergo DNA replication and nuclear migration to the mother bud neck, consistent with arrest in the cell division cycle at the G2/M boundary.
Levels of the C-terminally truncated Sth1 proteins detected in sth1 cells carrying sth1Δ1338, sth1Δ1349, sth1Δ1274-1335, or sth1Δ1274 grown at 30° or 37° were comparable to those of the full-length Sth1p, as judged by immunoblotting using anti-Sth1p antibody (data not shown). Together, these results indicate that the bromo-domain is required for wild-type function. In other experiments, we showed that the sth1Δ1338 allele on a high copy plasmid introduced into a wild-type strain does not confer a dominant negative phenotype (data not shown). The morphological differences of cells expressing sth1Δ1338 or sth1Δ1349, which lack all or part of the predicted α3 helix, respectively, compared to those expressing sth1Δ1274-1335 or sth1Δ1274, which lack either two or all three of the predicted bromodomain α-helices, respectively, further suggest a critical role for the α3 helix in cells that also express the remaining sequences of the bromodomain.
Mitotic DNA damage checkpoint is not activated by the sth1-3ts or sth1Δ1338 alleles: One hypothesis to explain the G2/M arrest in strains with either the sth1-3ts or sth1Δ1338 mutations is that lesions in Sth1p activate one or more cell cycle checkpoints. To test whether the arrest in either the sth1-3ts or sth1Δ1338 cells requires the RAD9 checkpoint, we compared the growth of sth1-3ts or sth1Δ1338 single mutants with sth1-3ts or sth1Δ1338 cells from which RAD9 had been deleted, following a shift to 37°. The RAD9 checkpoint is activated by damaged and incompletely replicated DNA (Weinert and Hartwell 1988). For example, RAD9 is required for the G2 arrest of cells with mutations in the genes coding for DNA ligase, DNA polymerase α, and DNA polymerase γ, and rad9 cells carrying mutations of any of these genes exhibit greatly reduced viability (Weinertet al. 1994). The rad9Δ mutation did not alleviate the cell cycle arrest conferred by either of the sth1-ts alleles (data not shown). In addition, the viability of the sth1-3ts rad9Δ or sth1Δ1338 rad9Δ double mutants was comparable to that of the sth1-3ts or sth1Δ1338 mutant (data not shown), indicating that the arrest of sth1-3ts or sth1Δ1338 and the RAD9 checkpoint were independent. We also tested whether mutation of a second G2/M DNA damage and replication repair checkpoint gene, RAD53/MEC2, which encodes a protein kinase signal transducer believed to function downstream of Rad9p (Weinertet al. 1994; Elledge 1996), is required for the sth1-3ts arrest. The mec2-1 allele also failed to alleviate the G2/M arrest caused by sth1-3ts (data not shown) and did not enhance the lethality of sth1-3ts cells. These results indicate that the RAD9 and MEC2 genes necessary for monitoring DNA metabolism are not required for the sth1-3ts or sth1Δ1338 arrest.
Transcription of genes required for cell cycle progression through G2/M upon inactivation of STH1: Despite the inability of a LexA-Sth1p fusion protein to activate gene expression (Laurentet al. 1992), other roles for Sth1p in transcription are still possible. One explanation for the G2/M arrest of sth1-3ts mutants shifted to the nonpermissive temperature is that STH1 is required specifically for transcription of a gene or set of genes necessary for cell cycle progression. For example, temperature-sensitive mutations in yTAFII145 that cause a G1 arrest were shown to abolish transcription of the G1/S cyclin genes CLB5 and CLB6 (Walkeret al. 1997). To test whether Sth1p is required for expression of G2/M-specific genes, we analyzed RNA levels of two important genes required for progression through the G2/M transition, CLB2 and CSE4, upon inactivation of STH1. CLB2 encodes a B-type cyclin important for progression from G2 into M and is transcribed only during G2/M (Suranaet al. 1991), and CSE4 encodes a chromatin-associated protein necessary for the segregation of chromosomes, whose inactivation arrests cells in mitosis (Stoleret al. 1995). Transcription levels of CLB2 and CSE4, and as a control, CLN3, which encodes a G1 cyclin that is expressed constitutively (Nashet al. 1988), were unaffected in the sth1-3ts mutant at 3, 6, or 10 hr following a shift to the nonpermissive temperature (Figure 5). Thus, the G2/M arrest in sth1-3ts cells cannot be explained by a specific defect in the transcription of either CLB2 or CSE4.
STH1 is required for nonglucose carbon source utilization: To gain insight into the genetic targets of Sth1p, the growth phenotypes of sth1-ts mutants on fermentable and nonfermentable carbon sources other than glucose were tested at semipermissive temperatures (Figure 6). The mutants grew nearly as well as isogenic wild-type strains on glucose, but showed a variety of growth defects on other carbon sources (Figure 6). sth1-1ts mutants were defective for growth on ethanol + glycerol and significantly defective for growth on raffinose, sucrose, and galactose (Figure 6). sth1-2ts mutants were defective for growth on raffinose, but capable of some growth on sucrose, galactose, and ethanol + glycerol (Figure 6). sth1-3ts mutants were defective for growth on all nonglucose carbon sources tested, further distinguishing the three sth1-ts mutants. These pleiotropic defects suggest that Sth1p functions in a general cellular process affecting several metabolic pathways.
One explanation for the broad mutant phenotype is that Sth1p functions in some general aspect of transcription, like the Snf/Swi proteins, despite the finding that STH1 function was not required for expression of the G2/M-specific transcripts, CLB2 or CSE4. Therefore, expression of the SUC2 and GAL10 promoters, under glucose-derepressing or galactose-inducing conditions, respectively, was tested by quantitating β-galactosidase activities in sth1-ts strains carrying either a snf/swi-dependent SUC2-LEU2-lacZ reporter gene (Sarokin and Carlson 1985), or a GAL10-lacZ 2-μm reporter plasmid (Westet al. 1984) grown at the permissive and nonpermissive temperatures. The SUC2 reporter gene was capable of derepressed levels of lacZ expression in the sth1-1ts and sth1-3ts strains grown at 37° that were within twofold of those in the mutants grown at 30°, and levels of induced lacZ expression from the GAL10-lacZ reporter plasmid in the sth1-1ts and sth1-3ts mutants grown at 37° were reduced two- to threefold compared to 30° (data not shown). Taken together, these results indicate that sth1-ts mutants shifted to the nonpermissive temperature were only slightly defective in transcription of these promoters. The modest effects at SUC2 and GAL10 alone cannot therefore account for the conditional lethality caused by the sth1-ts alleles.
The sth1-3ts allele causes synthetic lethality and synthetic sickness with mutations in histone and nonhistone chromatin-associated proteins: Mutations in genes encoding histones and nonhistone chromatin-assembly proteins were shown previously to suppress the transcriptional and/or growth defects caused by snf/swi mutations. Because aspects of the chromatin-remodeling activities of Snf/Swi and RSC are similar (Cairnset al. 1996), we examined whether RSC interacted genetically with any of several previously identified mutations that suppress snf/swi mutations. Specifically, we tested whether the growth defects caused by the sth1-3 temperature-sensitive mutation were suppressed by spt6/ssn20-1, Δhta1-htb1, or point mutations in histone H3, hht2-2 (E105K) and hht2-3 (T118I), or histone H4, hhf2-8 (V43I), although suppression of Ts- growth is distinct from suppression of transcriptional defects. An spt6 mutation was not capable of suppressing the lethality associated with a complete deletion of STH1 (Laurentet al. 1992).
Surprisingly, we found that the sth1-3ts spt6/ssn20-1 double mutants exhibited more severe growth defects than the sth1-3ts or spt6/ssn20-1 mutant (Figure 7A, compare row 4 with rows 2 and 3), suggesting that the spt6/ssn20-1 mutation functions synergistically with the sth1-3ts mutation. In contrast, deletion of one of the copies of histones H2A and H2B neither suppressed nor enhanced the sth1-3ts phenotype (Figure 7A, compare row 6 with rows 2 and 5).
The partially dominant H3 mutation, hht2-2 (E105K), also caused a synthetic sickness with the sth1-3ts mutation (Figure 7B, compare row 4 with rows 2 and 3), although a second partially dominant H3 mutation, hht2-3 (T118I), did not (Figure 7B, compare row 5 with rows 3 and 4). The hht2-2 and hht2-3 mutations in an STH1 background grew as well as wild-type strain BLY76 at all temperatures assayed (Figure 7A, row 1; data not shown). We next tested the interactions between sth1-3ts and the H3 point mutations in strains in which the mutant variants were the sole source of HHT2 by transforming an hht2Δ sth1-3ts double mutant with low-copy plasmids expressing the hht2-2 or hht2-3 alleles. Importantly, the sth1-3ts hht2Δ double mutant recovered from the diploid BLY49-2D × BLY157 displayed growth defects that were more severe than those of either sth1-3ts or hht2Δ (Figure 7B, compare row 7 with rows 2 and 6), suggesting that like the partially dominant hht2-2 point mutation and ssn20-1, deletion of HHT2 caused synthetic sickness with sth1-3ts. Significantly, we were unable to recover viable sth1-3ts hht2Δ cells transformed with the hht2-2 plasmid, suggesting that the sth1-3ts hht 2-2 double mutant was synthetically lethal. Colonies of 100-200 cells were detected microscopically after 4-day growth on transformation plates at 30°, but these cells were inviable (data not shown). sth1-3ts hht2Δ cells transformed with the hht2-3 plasmid were viable but displayed a severe synthetic sickness even at the fully permissive temperature (Figure 7B, compare row 9 with rows 7 and 8). The synthetic phenotypes with histone H3 point and deletion mutations argue that these interactions are between three loss-of-function mutations, and are not allele specific.
Similarly, we examined sth1-3ts hhf2-8 interactions in a strain in which both copies of histone H4 were deleted and the only copy of H4 was provided by the point mutation. As for the hht2-3 mutation, we found that hhf 2-8 caused a synthetic sickness with the sth1-3ts mutation (Figure 7C, compare row 4 with rows 1-3).
In summary, the synthetic sickness and synthetic lethality caused by the sth1-3ts mutation with the ssn20-1, hht2-2, hht2-3, hht2Δ, and hhf2-8 mutations (as partially dominant and/or recessive mutations) suggest that Sth1p functions in a parallel genetic pathway with Spt6p and histones H3 and H4.
Yeast Snf/Swi and RSC are distinct nucleosome remodeling complexes: Despite similarities, at least five lines of evidence indicate that Snf/Swi and RSC have distinct functions in the cell. First, RSC is approximately 10-fold more abundant than Snf/Swi in yeast cells, suggesting that it plays a broader role. Second, the RSC and homologous SNF/SWI genes do not function in redundant pathways. Third, unlike the case in snf/swi mutants, SUC2 derepression and GAL10 induction are only slightly defective in rsc mutants, suggesting that the sth1-ts growth defects on sucrose, raffinose, and galactose are not due to primary defects in transcription at these promoters. Fourth, RSC is essential for mitotic growth and Snf/Swi is not. Finally, mutations in histone and nonhistone chromatin assembly genes that partially suppress the transcriptional or growth defects of snf/swi enhance the temperature-sensitive defects of the sth1-3ts mutation, suggesting differences in the way in which Snf/Swi and RSC interact with chromatin.
Point substitutions within conserved helicase motifs: Genetic and biochemical analyses of the lethal and conditional lethal mutations localized to the conserved ATPase/helicase-related domain of Sth1p reveal that this region is essential for several Sth1p functions, including DNA-stimulated ATP hydrolysis, progression through the mitotic cell division cycle, and one or more general cellular processes required for growth on several nonglucose carbon sources. On the basis of a comparison to the recently solved X-ray crystal structure of a DNA helicase (Subramanyaet al. 1996), the amino acids altered by the sth1 mutations are predicted to have roles in the enzymatic activity of Sth1p.
None of the point substitution mutations recovered from the mutagenesis of STH1 affects the ability of the encoded Sth1 protein to hydrolyze ATP in vitro. However, the unconditional lethality of the K501R mutation could explain the failure to recover conditional ATPase-sensitive alleles; alternatively, it is possible that the sth1-ts mutations affect the ATPase activity of RSC in vivo. Importantly, the sth1-3ts allele confers G2/M arrest, suggesting that cell cycle progression and in vitro ATP hydrolysis functions are separable. Assembly of RSC may also occur independently of ATP hydrolysis, as Sth1p expressing the ATPase-sensitive K501R allele binds as well as the wild-type to Sfh1p, a yeast Snf5p homolog, in vivo in the two-hybrid assay (data not shown). This result supports a model in which ATPase activity is downstream of RSC assembly, and is consistent with the finding that mutation of the corresponding amino acid of Snf2p permitted assembly of Snf/Swi (Petersonet al. 1994; Richmond and Peterson 1996).
The sth1-ts mutants exhibit cell cycle and non-cell cycle phenotypes. The similarity of arrest phenotypes caused by the sth1-3ts allele and the sth1Δ deletion allele in STH1 depletion experiments strongly suggests that sth1-3ts affects STH1 function. The differences in phenotypes of the three sth1-ts mutants suggests a model that Sth1p carries out important functions not only at G2/M but also throughout the cell cycle, although we cannot rule out the possibility that sth1-1ts and sth1-2ts are leaky alleles. The importance of the ATPase/ helicase-related domain for STH1 function is further underscored by the finding that six out of seven of the disease-causing mutations in XH2, a human SNF2-related gene, are missense alleles located throughout the corresponding ATPase/helicase-related domain (Gibbonset al. 1995). Ultimately, it will be important to characterize the link between the ATP hydrolysis, chromatin modeling, and cell division cycle functions of RSC in vivo.
Functional importance of the Sth1p bromodomain: A function for the highly conserved bromodomain, a protein sequence motif found in one to five copies in several different proteins, is currently unknown. Of the proteins in which its functional importance has been tested, only Gcn5p requires the bromodomain for full complementation, although it is dispensable for Gcn5p histone acetyltransferase (HAT) activity (Candauet al. 1997). Interestingly, the bromodomain of the human GCN5 (hGCN5) HAT was shown recently to interact with the Ku70 subunit of the DNA-PK holoenzyme that phosphorylates hGCN5 thereby repressing HAT activity, suggesting a role for the bromodomain in the negative modulation of protein activity (Barlevet al. 1998). The bromodomain in Snf2p, the most closely related protein to Sth1p in the S. cerevisiae genome, is dispensable (Laurentet al. 1993). In contrast, deletion mutant analysis demonstrates the functional importance of the Sth1p bromodomain and further suggests that this region carries out a discrete function. Although all STH1 C-terminal truncation mutants are temperature sensitive for growth, sth1Δ1338 and sth1Δ1349, which lack all or a portion of the bromodomain α3 helix/helix C (Jeanmouginet al. 1997), show the strongest phenotype and arrest in G2/M of the cell cycle. These cells arrest with abnormally elongated buds, a phenotype not detected in two additional sth1 mutants that arrest at the same point in the cell cycle: sth1-3ts cells and sth1Δ null cells in which the expression of GAL1,10 promoter-driven STH1 is repressed. Several of these cells were multiply budded and bud growth was often hyperpolarized, terminal phenotypes that are reminiscent of the G2/M cak1 (Kaldiset al. 1996) or the G1/S sextuple clb1-6 and the cdc34 (Schwobet al. 1994) mutants, suggesting defects in mitotic cyclins. This result suggests that the Sth1p bromodomain is engaged in a novel cellular function. The morphological differences of cells expressing sth1Δ1338 and sth1Δ1349 compared to those expressing sth1Δ1274-1335 and sth1Δ1274 further suggest a functional communication between α3 and the two amino-terminal α-helices of the bromodomain. We speculate that amino acids on the surface of the α3 helix regulate intramolecular or intermolecular protein-protein interactions to modulate Sth1p activity.
Roles of the Sth1 protein in the mitotic cell cycle: Several mutations in STH1 lead to a mitotic cell division cycle arrest at the G2/M transition. At least four models could explain this result. The first is that Sth1p is a Cdc (cell division cycle) protein with a direct role in regulating cell cycle progression. Our data suggest a role for Sth1p at, or just preceding, the G2/M transition, although the mitotic indices and microcolony assays of sth1-3ts and sth1Δ1338 cells indicate that these are not first-cycle arrest mutants.
The second model is that RSC is required indirectly for the mitotic cell division cycle. According to this model, mutation of one or more RSC components leads to loss of RSC function or decreased levels of RSC, either of which causes a cell cycle arrest at the G2/M stage. The finding that a conditional mutation in another RSC component, SFH1, also arrests cells at the same position in the cell cycle (Caoet al. 1997) argues in favor of this model. In addition, two of the other proteins in the Snf2p subfamily regulate cell cycle progression (Dunaiefet al. 1994; Stroberet al. 1996). At present, we can rule out a requirement for transcription of two genes whose function is necessary for cell cycle progression through mitosis, CLB2, a B-type cyclin and CSE4, which is required for chromosome condensation, as transcription of neither was dependent on STH1.
A third model is that the sth1 mutants activate a G2/M checkpoint control pathway. Significantly, mutations in the Drosophila lodestar gene, a SNF2 homolog, result in chromatid tangling and fragmentation at anaphase (Girdham and Glover 1991), defects that trigger mitotic checkpoints in yeast. At least three independent checkpoint pathways exist in S. cerevisiae to ensure the accurate transmission of genetic material during mitosis (Weinert and Hartwell 1988; Hoytet al. 1991; Li and Murray 1991; Lew and Reed 1995). So far, we have shown that neither the RAD9 nor the MEC2 genes needed for DNA damage and replication checkpoints (Murray 1995) are required for the sth1-3ts arrest. However, it is still possible that loss-of-function mutations in any of the other genes involved in the G2/M checkpoint (Paulovichet al. 1997) could alleviate the cell cycle arrest of the sth1-3ts allele. Alternatively, the sth1-3ts allele could activate a novel checkpoint pathway that surveys the chromatin structure of DNA.
The fourth model is that RSC is required for proper assembly or organization of chromatin. Histones are synthesized in late G1-early S phase of the cell cycle and are required for nucleosome assembly during S phase (Osley and Lycan 1987; Norriset al. 1988; Linget al. 1996). Moreover, other experiments indicate the need for histones in progression through G2 (Norris and Osley 1987; Kayneet al. 1988; Morganet al. 1991; Megeeet al. 1995). A role for RSC in mitotic chromosome condensation or sister chromatid segregation, like the highly conserved yeast Smc or Cse4 (related by sequence to histone H3) proteins, respectively (Stoleret al. 1995; Strunnikovet al. 1995), is also consistent with the terminal morphology produced by sth1-3ts cells. Importantly, the Drosophila ACF and CHRAC multiprotein complexes that each contain the Snf2p-related protein ISWI, have recently been shown to function in the ATP-dependent assembly of chromatin (Itoet al. 1997; Varga-Weiszet al. 1997). In addition, human homologs of Snf/Swi have been shown to associate with nuclear matrix, suggesting a role in organizing chromatin (Reyeset al. 1997). By analogy, RSC could also direct assembly of chromatin in mitotic yeast cells, in addition to its established role in nucleosome reconfiguration in vitro. The synthetic sickness and lethality observed between the sth1-3ts mutation and spt6/ssn20-1 and histone H3 and H4 mutations support this hypothesis.
We reasoned that if RSC plays a general role in organizing chromatin, the structure of bulk chromatin would be altered with inactivation of STH1 function. To test this, nuclei prepared from sth1-3ts cells and isogenic wild-type cells grown at various times following a shift to 37° were digested with micrococcal nuclease. No major differences in micrococcal nuclease sensitivity were detected in the STH1+ and sth1- nuclei (data not shown), suggesting that Sth1p does not function in establishing or maintaining bulk chromatin. Further genetic and biochemical experiments will help us to distinguish among the models above, although the role of RSC in the cell cycle could be explained by a combination of these models.
Sth1p and chromatin modeling: Purified RSC can remodel nucleosomes in an ATP-dependent manner in vitro (Cairnset al. 1996) although it remains to be shown that RSC has such activity in vivo. Significantly, several of the mutations in histone and chromatin assembly genes that suppress snf/swi failed to suppress the growth defects caused by temperature-sensitive mutations in STH1. Instead, we found that the sth1-3ts allele is synthetically sick or synthetically lethal with mutations in the histone H3, H4, and SPT6 genes, implying that these genes function synergistically in parallel genetic pathways.
The specific histone H3 and H4 mutations examined here affect histone-DNA interactions to varying degrees (Kurumizaka and Wolffe 1997; Lugeret al. 1997; Wechseret al. 1997) and all alter amino acids within the histone fold regions of H3 and H4, either within α-helices or in loops linking α-helices (Lugeret al. 1997). Inexplicably, the E105K mutation, which is synthetically lethal with sth1-3ts, has relatively minor effects on nucleosome structure (Kurumizaka and Wolffe 1997) and is predicted to have little direct effect at the mononucleosome level (Lugeret al. 1997).
Several models could explain the synthetic enhancement between mutations in sth1-3ts and mutations in histones H3, H4, or SPT6. One possibility is that Sth1p is required for expression of the histone or SPT6 genes. At present, we favor a model that the synthetic enhancement phenotypes between mutations in sth1-ts and mutations in SPT6 and histones reflect functional interactions. Our genetic analysis showed that the strongest synthetic phenotypes were between sth1-3ts and hht2, hhf2, and spt6. Interestingly, Bortvin and Winston (1996) have shown that Spt6p interacts directly with histones H3 and H4 (and to a lesser extent with H2A-H2B) and probably plays a role in nucleosome assembly. Our results suggest that RSC and Snf/Swi interact oppositely with histones and chromatin-associated proteins and that RSC plays an important role in the organization or assembly of chromatin.
Tatayana Makarova is acknowledged for assistance with the ATPase experiments. We thank Shulan Wu for RNA analyses, Matvey Brokhin for help in characterizing the sth1-2ts mutant, and Tang Yi for contributions to the bulk chromatin analyses during a research rotation. We also thank Mary Ann Osley, Steve Kurtz, Michelle Treitel, Rong Jiang, Brad Cairns, Maureen McLeod, Fred Volkert, and Yixue Cao for many fruitful discussions and for comments on the manuscript, and all colleagues for kindly providing plasmids and strains. This work was supported by grants from the American Cancer Society (NP-871) and the March of Dimes Basil O’Connor Starter Scholar Research Award (5-FY93-0870; 5-FY94-0759) to B.C.L. B.K.B. was supported by U.S. Public Health Service grant GM47238 awarded to Fred Cross, and M.P.K. was supported by U.S. Public Health Service grant GM52908 awarded to Robert T. Simpson.
Communicating editor: F. Winston
Note added in proof: An sth1/nps1 temperature-sensitive allele has been shown recently to activate the MAD1-dependent checkpoint and cause alterations in centromeric chromatin structure (Tsuchiya, E., T. Hosotani and T. Miyakawa, 1998. Nucleic Acids Res. 26: 3286-3292).
- Received May 19, 1998.
- Accepted July 23, 1998.
- Copyright © 1998 by the Genetics Society of America