Tra1 is an essential component of the Saccharomyces cerevisiae SAGA and NuA4 complexes. Using targeted mutagenesis, we identified residues within its C-terminal phosphatidylinositol-3-kinase (PI3K) domain that are required for function. The phenotypes of tra1-P3408A, S3463A, and SRR3413-3415AAA included temperature sensitivity and reduced growth in media containing 6% ethanol or calcofluor white or depleted of phosphate. These alleles resulted in a twofold or greater change in expression of ∼7% of yeast genes in rich media and reduced activation of PHO5 and ADH2 promoters. Tra1-SRR3413 associated with components of both the NuA4 and SAGA complexes and with the Gal4 transcriptional activation domain similar to wild-type protein. Tra1-SRR3413 was recruited to the PHO5 promoter in vivo but gave rise to decreased relative amounts of acetylated histone H3 and histone H4 at SAGA and NuA4 regulated promoters. Distinct from other components of these complexes, tra1-SRR3413 resulted in generation-dependent telomere shortening and synthetic slow growth in combination with deletions of a number of genes with roles in membrane-related processes. While the tra1 alleles have some phenotypic similarities with deletions of SAGA and NuA4 components, their distinct nature may arise from the simultaneous alteration of SAGA and NuA4 functions.
IN eukaryotic cells the post-translational modification of nucleosomes by multisubunit complexes is a key aspect of transcriptional regulation (reviewed in Berger 2002). Histone modifications including acetylation, methylation, ubiquitylation, and phosphorylation can directly alter chromatin structure or act as a recruitment signal for additional factors (Strahl and Allis 2000). As well as regulating transcriptional initiation, nucleosome modifications affect transcriptional elongation and other nuclear processes such as DNA replication, DNA repair, and RNA export (Iizuka and Smith 2003).
The Saccharomyces cerevisiae Spt-Ada-Gcn5-Acetyltransferase (SAGA) complex modifies chromatin and provides an interface between DNA-binding transcriptional regulators and the basal transcriptional machinery (reviewed in Green 2005). The structural core of SAGA is composed of a subset of the TBP-associated factors (TAFs) (Grant et al. 1998a; Wu et al. 2004), with Spt7, Ada1, and Spt20 also being required for the integrity of the complex (Horiuchi et al. 1997; Roberts and Winston 1997; Sterner et al. 1999). The histone acetyltransferase Gcn5/Ada4 activates and represses transcription by modifying histones H3 and H2B (Brownell et al. 1996; Grant et al. 1997; Kuo et al. 1998; Wang et al. 1998; Ricci et al. 2002). In turn, the Ada proteins, Ada2 and Ngg1/Ada3 regulate the activity and substrate preference of Gcn5 (Balasubramanian et al. 2001). Further regulation is provided by the interaction of Spt3 and Spt8 with the TATA-binding protein (Eisenmann et al. 1992, 1994; Dudley et al. 1999). Recruitment of SAGA to promoters is mediated by Tra1, an essential 437-kDa protein (Grant et al. 1998b; Saleh et al. 1998) that interacts directly with transcriptional activators (Brown et al. 2001; Bhaumik et al. 2004; Fishburn et al. 2005; Reeves and Hahn 2005). The mammalian ortholog of Tra1, TRRAP is required for transcriptional regulation by myc, p53, E2F, and E1A (Mcmahon et al. 1998; Bouchard et al. 2001; Deleu et al. 2001; Ard et al. 2002; Kulesza et al. 2002). Its deletion results in defects in cell cycle progression and early embryonic lethality (Herceg et al. 2001).
Tra1 is also a component of the multisubunit NuA4 complex (Allard et al. 1999) that preferentially acetylates histones H4 and H2A, the catalytic subunit being the essential protein Esa1 (Smith et al. 1998; Clarke et al. 1999). NuA4 associates with acidic activation domains, probably through Tra1-mediated interactions, and activates transcription in an acetylation-dependent manner (Vignali et al. 2000; Nourani et al. 2004). Acetylation by NuA4 is also critical for nonhomologous end joining of DNA double-stand breaks and for replication-coupled repair (Bird et al. 2002; Choy and Kron 2002; Downs et al. 2004). The role of NuA4 in repair was highlighted by the finding that it acetylates the histone variant Htz1, which is intimately involved with these processes (Keogh et al. 2006).
Aside from its length, a distinguishing feature of Tra1 is its C-terminal domain of ∼300 amino acids that is related to the phosphatidylinositol-3-kinase (PI3K) domain found in several key cellular regulators including ATM, DNA-PK, and FRAP (Keith and Schreiber 1999). The group also shares less well-defined sequences flanking the PI3K domain called the FRAP-ATM-TRRAP (FAT) and C-FAT domains (Bosotti et al. 2000). Unlike other members of the family, Tra1 and TRRAP lack the signature motifs of kinases and kinase activity (Bosotti et al. 2000). The exact role of the PI3K domain is thus unclear; although in human cells, the PI3K domain of TRRAP is required for cellular transformation by myc and E1A (Park et al. 2001).
We have used a mutagenesis approach to determine the structure/function relationships of Tra1, focusing in particular on the PI3K domain. We have identified eight mutations in the PI3K domain that result in cellular inviability and three that result in temperature-sensitive growth and reduced growth on media containing 6% ethanol. Characterization of these temperature-sensitive alleles at the permissive temperature confirms a role for the PI3K domain in transcriptional regulation. These mutations confer altered expression of ∼7% of the yeast genome and were distinct from those affecting individual components of either SAGA or NuA4 complexes.
MATERIALS AND METHODS
Yeast strains and growth:
Yeast strains are listed in Table 1. TRA1 alleles contained on TRP1 centromeric plasmids were transformed into CY1021 and the wild-type copy was displaced by plasmid shuffling (Sikorski and Boeke 1991). CY1896 is a derivative of Y5565 that has been gene replaced with tra1-SRR3413 and selected for through placement of Tn10LUK at the downstream BstBI site. CY2437 and CY2438 are derivatives of KY320 and similarly contain integrations of wild-type TRA1 and tra1-SRR3413, respectively, in a background lacking BamHI and SalI restriction sites within TRA1.
Growth comparisons were performed on selective plates after 3–5 days at 30° unless stated otherwise. Assays were performed in duplicate on independently constructed strains. Scoring for PI3K domain mutations was done in relation to CY1524, which contains TRA1SB, the background allele used to construct mutations within the PI3K domain. TRA1SB is Flag-tagged and contains a BamHI site that converts N3580A and a SalI site of nucleotides A9714G and T9717G. Scoring for the viability of non-PI3K domain mutations was relative to CY1020 (Saleh et al. 1998).
Construction of DNA molecules:
lacZ reporter constructs were cloned as his3-lacZ fusions into the LEU2 centromeric plasmid YCp87 (Brandl et al. 1993). ADH2-lacZ contains promoter sequences (−440 to −150; Saleh et al. 1998). PHO5 (−452 to +47), INO1 (−517 to +40), and HIS4 (−545 to +29) were engineered by PCR as BamHI–HindIII fragments and cloned into YCp87 (see supplemental Table 1 at http://www.genetics.org/supplemental/ for a listing of oligonucleotides).
Construction of TRA1 alleles:
Mutations of DSP171AAA, EFS184AAA, KEL345AAA, VLL837AAA, VRL867AAA, KKK1686AAA, GLW2677AAA, GYH2939AAA, TLP2972AAA, HFQ3168AAA, SRR3413AAA, and PFR3621AAA were engineered into myc-TRA1-Ycplac111 (Saleh et al. 1998). Each allele was synthesized in two PCR reactions. Reactions contained a listed oligonucleotide and the appropriate outside flanking primer with a unique cloning site (oligonucleotides 842, 2542-3, 2337-1, 2323-2, 2323-1, 2542-4, 2337-2, and 2346). Mutagenized codons were replaced with a NotI site such that 5′ and 3′ halves could be ligated sequentially into pUC vectors. Fragments were moved into myc-TRA1-Ycplac111 using the outside restriction sites. Mutations within the coding region of the PI3K domain (including additional alleles of EER3416AAA and PFR3621AAA) were constructed in TRA1SB. Flag epitopes were introduced at an N-terminal NotI site. Mutagenized alleles were constructed by PCR and ligated as SalI–BamHI (for alterations N-terminal to N3580) or BamHI–SacI (for alterations C-terminal to N3580). Flanking primers used in the PCR reactions were 4293-1 and 4225-3 or 4249-1 and 2346/4479-1 depending on the placement relative to N3580. YNG2 and GCN5 were similarly cloned into this molecule.
Tandem affinity purification (TAP)-tagged molecules were cloned into YCpDed-TAP, a derivative of the URA3-containing centromeric plasmid YCplac33 (Gietz and Sugino 1988) that contains a DED1 promoter driving expression of a TAP epitope. The DED1 promoter was synthesized by PCR using oligonucleotides 4149-1 and 4149-2 and cloned as a PstI–BamHI fragment into YCplac33. The TAP epitope was subsequently cloned as a BamHI–SacI fragment into this molecule after PCR using oligonucleotides 4149-4 and 4168-1 and pFA6a as the template (provided by Kathy Gould). The coding sequence for the Flag-epitope was cloned into this molecule at the engineered NotI site to give YCpDed-TAP-Flag. TRA1 was introduced into this vector as a NotI–SacI fragment.
TAP was carried out as described by Rigaut et al. (1999). One liter of cells grown in YPD to an A600 ∼1.5 were ground in liquid nitrogen (Saleh et al. 1997). Further steps were at 4°. Cell extract suspended in 5 ml of IPP150 with protease inhibitors (1 mm phenylmethylsulfonylfluoride, 0.1 mm benzamidine hydrochloride, 2 μg/ml pepstatin A, 2 μg/ml leupeptin, and 0.1 mg/ml trypsin inhibitor) was cleared by centrifugation at 147,000 × g and incubated with 600 μl of IgG-Agarose (Sigma-Aldrich Canada, Oakville, Ontario) for 2 hr. After washing with 30 ml of IPP150 and 10 ml of TEV cleavage buffer, beads were suspended in 1 ml of cleavage buffer and incubated with 100 units TEV protease for 2 hr. Three microliters of 1 m CaCl2 were added to the eluent and then diluted with 3 ml of calmodulin binding buffer. The protein was incubated with 600 μl of calmodulin–sepharose (Stratagene, La Jolla, CA) for 1 hr, washed with 30 ml of calmodulin binding buffer, and protein eluted.
HIS4-lacZ cells grown in minimal media to an A600 of ∼1.5 were pelleted, washed in LacZ buffer, and concentrated approximately fivefold. β-Galactosidase was determined using o-nitrophenol-β-d-galactosidase as substrate, standardizing to cell density (Ausubel et al. 1998). Analysis of PHO5-lacZ, INO1-lacZ, and ADH2-lacZ was performed similarly except that overnight cultures were washed three times in water and then grown to an A600 ∼1.5 in YPD depleted of phosphate (Han et al. 1988)or depleted of inositol or in minimal media containing 3% ethanol (v/v) and 0.1% glucose.
Gal4 affinity chromatography:
Amino acid residues 764–882 of Gal4 were expressed as a GST fusion in the vector pGEX1. One liter of Escherichia coli cells was induced with 0.6 mm isopropyl β-d-thiogalactopyranoside for 6 hr at 37°. Harvested cells were resuspended in lysis buffer [40 mm Tris (pH 8.5), 20 mm NaCl, 10% glycerol, 0.08% NP-40, 0.2 mm EDTA (pH 8.0), 1 mm DTT] with protease inhibitors, treated with 1 mg/ml lysozyme on ice for 1 hr, and broken by sonication. Cleared extract was incubated on ice with 1 ml glutathione sepharose 4B (GE Healthcare, Waukesha, WI) and washed three times with 10 ml of 140 mm NaCl, 2.7 mm KCl, 10 mm Na2HPO4, 1.8 mm KH2PO4 and once with 10 ml of lysis buffer. TAP-Tra1, purified on IgG-Agarose from 1.0-liter cultures of CY1962 and CY1963 containing HA-tagged Ada2 (Saleh et al. 1997), was applied to Gal4-containing glutathione sepharose in 3 ml of lysis buffer, rotated at 4° for 60 min, and washed with lysis buffer, and bound protein was eluted with 10 mm reduced glutathione. Thirty microliters of the eluent were separated on a 10% SDS–PAGE gel, transferred to PVDF membrane, and blotted with anti-HA antibody.
Genetic analyses—identification of an intragenic suppressor of tra1-SRR3413:
Approximately 2 × 106 cells of yeast strain CY1531 (tra1-SRR3413) were plated onto YPD containing 6% ethanol. Two colonies were isolated after 5 days of growth and Tra1 expressing plasmids were recovered (Hoffman and Winston 1987). The SalI–SacI fragment was recloned into TRA1SB and found sufficient for suppression. Systematic genetic-array analysis on yeast strain CY1896 (tra1-SRR3413) was performed as described by Tong and Boone (2006).
RNA purification and microarray analysis:
Yeast cells were grown in YP media containing 2% glucose to an A600nm = 2.0 at 30°. RNA was purified from 108 cells after glass bead disruption as described previously (Mutiu and Brandl 2005; Mutiu et al. 2007). Microarray analyses were performed using the Agilent yeast oligo array kit (Mogene, LC, St. Louis). Comparisons were made with wild-type strain CY1524 and performed in duplicate with dye reversal. Array data for the confirmed set of protein-coding genes were log transformed and averages obtained. Genes displaying an expression change >1.5-fold, where the direction of the response differed between replicates, were discarded. Profiles were normalized by subtracting the mean response of all genes from the response of each gene and then dividing by the standard deviation of all responses within a profile. An initial cluster analysis using a subset of the compendium data (Hughes et al. 2000) that included deletion strains with gene ontology terms indicating gene expression identified microarray profiles of strains with deletions of cka2, ckb2, gcn4, hat2, isw1, mbp1, rpd3, rtg1, sap30, sin3, ste12, and yap3 as being most similar to tra1-SRR3413. Agglomerative hierarchical clustering based on the average linkage of uncentered correlations was performed on the profiles from these strains and strains with deletions of SAGA (Ingvarsdottir et al. 2005) and NuA4 (Krogan et al. 2004) components using CLUSTER 3.0 software (Eisen et al. 1998). Genes not appearing in at least two of the microarray profiles were excluded. The data were visualized using MapleTree (http://rana.lbl.gov/EisenSoftware.htm). Gene ontology terms associated with gene clusters were determined using FunSpec (Robinson et al. 2002).
Chromatin immunoprecipitation assays:
Assays were performed essentially as described by Kuo and Allis (1999). For assays comparing acetylated histone H3, 50 ml of culture was grown to an A600nm ∼ 2.0. To equalize amounts of input DNA, cell extract was reverse crosslinked and PCR reactions performed on serial dilutions. For immunoprecipitations, DNA equal to that contained in 50 μg (protein) of cell extract from the wild-type sample was incubated with protein A sepharose 4 Fast Flow (Pharmacia, Piscataway, NJ), transferred to clean tubes, and rotated overnight with 1 μl of antibody (anti-H3, ab1791, Abcam, Cambridge, MA; anti-AcH3, Lys9, 07-352, Upstate Biotechnology). Precipitate was removed by centrifugation and the supernatant incubated with protein A beads. After washing and reversal of crosslinks, products were analyzed by PCR (Ricci et al. 2002) using primers for the PHO5 (−550 to −300), ADE1 (−505 to 51), PHO84 (−199 to 74), PGK1 (−249 to 28) promoters or PHO5 gene (1035–1322). PCR conditions were 5 min at 95°, followed by 25 cycles of 30 sec at 95°, 52°, and 72°, and then 5 min at 72°. Assays of acetylated histone H4 were performed with 250-ml cultures grown to A600nm ∼ 2.0 in YPD, 0.5 mg of chromatin (in a total volume 0.5 ml), and protein A/G plus 2 μl of anti-AcH4, Lys8 (ab1760, Abcam). For immunoprecipitations using TAP-Flag-Tra1, chromatin was prepared in RIPA buffer with washing and elution as described by Alekseyenko et al. (2006).
Yeast extract prepared by glass bead disruption (Brandl et al. 1993) or grinding in liquid nitrogen (Saleh et al. 1997) was separated by SDS–PAGE and transferred to PVDF membranes (Roche Applied Science, Indianapolis) using a wet transfer system in 48 mm Tris, 40 mm glycine, 0.0375% SDS, and 20% methanol for 1 hr at 100 V. Anti-Flag antibody (M2, Sigma-Aldrich Canada ) and anti-HA antibody (ascites fluid from the 12CA5 cell line) were used at a ratio of 1:4000. Secondary antibody (anti-mouse IgG HRP; Promega, Madison, WI) used at a ratio of 1:10,000 was detected using a 20% solution of Immobilon Western (Millipore). Densitometric scanning of films was performed using AlphaImager 3400 software (Alpha Innotech, San Leandro, CA).
Telomere length assays:
Plasmid linearization assay:
Strains were transformed with pVL106 (Lundblad and Szostak 1989). Leu+ Ura+ colonies were grown to saturation in media lacking uracil and then YPD. Ten microliters were spotted onto leucine-depleted plates containing 5-fluoroorotic acid (5-FOA). Values reported are the percentage of 5-FOA resistant, Leu+ colonies in the mutant strain as compared to wild type, normalized to the number of cells plated.
Terminal restriction fragment determination:
TRA1 alleles were transformed into CY1021 and plasmid shuffling was performed (generation zero). Ten-milliliter saturated cultures were grown from these colonies at 30° or 34° and DNA was prepared (Ausubel et al. 1998). Each culture was also diluted 1:1000 into YPD and grown again to saturation, and then the process was repeated. Genomic DNA was digested with XhoI and Southern blotting performed (Teng and Zakian 1999).
Fractionation of Tra1-containing complexes on HiTrap Q sepharose Fast Flow:
Eighteen milligrams of whole-cell extract containing HA-Ada2p and Tra1SB or Tra1-SRR3413 was prepared in 40 mm Tris-HCl (pH 7.7) buffer containing 20 mm NaCl, 10% glycerol, 0.08% Nonidet P-40, 0.2 mm EDTA, and 0.1 mm dithiothreitol. Extracts were treated with protamine sulfate (0.2%), rotated at 4° for 30 min, cleared by centrifugation at 147,000 × g, and applied to a HiTrap Q sepharose Fast Flow column (1.0 ml; flow rate of 1.0 ml/min; Amersham Pharmacia Biotech). After washing with 4 ml of extraction buffer, protein was eluted with a 16-ml gradient of 0–1.0 m NaCl. Aliquots (20 μl) of 200-μl fractions were separated by SDS–PAGE and analyzed by Western blotting with anti-HA antibody. Twenty-five milligrams of whole-cell extract containing TAP-Flag-Yng2 and Tra1SB or Tra1-SRR3413 were fractionated and blotted using anti-Flag antibody.
The size of Tra1 presents a significant challenge for analyses of structure and function using molecular genetic approaches. Because of the difficulty in identifying and characterizing random mutations, we undertook a targeted mutagenesis approach, initially constructing 12 triple-alanine scanning mutations scattered in regions that might be relevant for function (Table 2). These alleles were expressed on a TRP1-containing centromeric plasmid and transformed into yeast strain CY1021 that contains a disruption of the genomic copy of TRA1 complemented by wild-type TRA1 expressed on a URA3-containing centromeric plasmid (Saleh et al. 1998). Each allele was analyzed for growth properties after shuffling out wild-type TRA1 on media containing 5-FOA. Phenotypes examined included mating efficiency; temperature sensitivity (37°); growth on media depleted for inositol; amino acids and phosphate; fermentation of galactose and raffinose; and sensitivity to hydroxyurea, methylmethanesulfonate (MMS), ultraviolet irradiation, γ-irradiation, high osmolarity, cycloheximide, and tunicamycin. Two of the alleles tested, DSP171AAA and EFS184AAA altered potential phosphorylation sites. (The number indicates the first residue of the three. We do not include the AAA for each allele.) VLL837 altered a putative LXXLL motif. Many of the other altered residues are conserved in yeast and/or human Tra1 homologs. Only 2 of the 12 alleles, both with mutations within the PI3K domain, had discernible phenotypes. tra1-PFR3621 would not support growth on media containing 5-FOA and therefore is a nonfunctional allele. tra1-EER3416 showed reduced growth on all media tested, as well as slight temperature sensitivity.
Targeted mutation of the PI3K domain:
On the basis of the above results, we directed additional mutations to the PI3K domain. As we were concerned that mutations might simply alter the fold of the PI3K domain, we focused on mutagenizing codons for amino acid residues conserved in the Tra1/TRRAP family of molecules but not in the broader PI3K domain family. In the absence of a structure for the Tra1 PI3K domain, the structure of PI3K-γ (Walker et al. 1999; see supplemental Figure 1 at http://www.genetics.org/supplemental/) allows an initial prediction of the positions of these residues. The PI3K domain consists of N-terminal and C-terminal lobes connected by a loop between β7 and β8. For PI3K-γ, ATP interactions are found with the β3–β4 loop (P-loop), the end of β5, and the β7–β8 loop. The region between α6 and β9 corresponds to the catalytic loop. The C-terminal lobe also contains an activation loop that is required for substrate binding of the PI3K molecules (Bondeva et al. 1998) and likely binds phosphatidylinositols in PI3K-γ.
The phenotypes of strains containing the mutagenized tra1 alleles are shown in Table 3. Of the 19 alleles with targeted mutations in the PI3K domain, 8 did not support viability. Two of the 8, R3456A and R3567A, were single-amino-acid substitutions; the others, WRR3317, DIE3351, RFL3374, FRK3544, PFR3621, and RDE3668 were triple-alanine scanning mutations. Five of the eight mutations were within predicted loop regions (WRR3317, DIE3351, RFL3374, R3567, and PFR3621). F3443, which aligns with the α3–β6 loop, can be considered in this group since strains containing tra1-F3443A were marginally viable in rich media but inviable under conditions of stress. If their positioning within loops is correct, it is not likely that these residues are important for structure but instead may be involved in protein–protein interactions or have regulatory roles. In agreement with this, R3567 of Tra1 aligns with R947, an essential residue within the catalytic loop of PI3K-γ (Dhand et al. 1994; Stack and Emr 1994). Likewise it is required for the activity of Tra1, though for Tra1/TRRAP the size of the residue may be the critical factor since human TRRAP contains a leucine at this position. Similarly, PFR3621 aligns with the activation loop and despite the lack of kinase activity, these residues are required for the activity of Tra1. (We note that the phenotype of PFR3621 may in part be due to reduced expression levels; not shown.) Three mutations giving rise to inviability were not in predicted loops. R3456 is in a comparable position to G877 of PI3K-γ, which is found within β7 and conserved in many of the lipid kinases (Walker et al. 1999). FRK3544 and RDE3668 align with α6 and α9, respectively, and are uniquely conserved in the Tra1/TRRAP molecules.
Eleven alleles supported viability in rich media. Their growth characteristics under a variety of conditions that reflect ada, spt, or NuA4 phenotypes are shown in Table 3. Other conditions and assays tested but without substantial effects were growth on MMS (also see below), tert-butylhydroperoxide, galactose, raffinose, 6-azauracil, 1.0 m NaCl, and tunicamycin. In addition, none of these alleles significantly changed the rate of loss of a URA3-containing centromeric plasmid after growth on nonselective conditions or resulted in a significant defect in mating (not shown). K3571A and N3580A had little effect on growth under any of the conditions tested. Four of the alleles, P3446A, DKL3491, P3662A, and N3617A, showed a very weak classic ada phenotype (Berger et al. 1992), being partially resistant to overexpression of VP16. (Note that a deletion of gcn5 would score as ++++ on this scale.) Some of these alleles were also slightly sensitive to 6% ethanol, characteristic of an ada2 disruption (Takahashi et al. 2001).
Three alleles, tra1-SRR3413, tra1-P3408A, and tra1-S3463A, resulted in a common and distinct phenotype. (EER3416 had a similar phenotype but to a lesser extent.) Strains containing these alleles were temperature sensitive for growth at 37°, had dramatically reduced growth in media containing elevated concentrations of ethanol, and grew poorly in media lacking phosphate. Plates showing the growth of strains containing these alleles under some of the conditions are shown in Figure 1A. The two alleles with the greatest temperature sensitivity (tra1-SRR3413 and tra1-P3408A) resulted in somewhat reduced growth on media depleted of inositol or containing benomyl, a microtubule-destabilizing agent. These latter two phenotypes are shared with deletions of some Spt components of SAGA (Roberts and Winston 1997) and with components of the NuA4 complex (Lemasson et al. 2003; Bittner et al. 2004; Krogan et al. 2004), respectively. The tra1 alleles displayed similar sensitivity to ethanol as does deletion of ada2 (not shown), and like ada2 they had reduced growth on calcofluor white (Figure 1A), suggesting that the effect results from changes in cell-wall properties or the ability of the strains to respond to changes in cell-wall integrity (Takahashi et al. 2001). We examined the growth properties of strains containing these alleles in more detail (Figure 1B). At 34°, the tra1-SRR3413, tra1-P3408A, and tra1-S3463A alleles resulted in a reduced growth rate in rich media relative to the TRA1SB allele. Upon heat shock the mutant strains showed an immediate growth arrest but did recover; however, 4 hr after the shift to 37°, growth stopped. To ensure that the effects of these mutations were not due to expression on a plasmid or the background N3580A mutation from the BamHI site within TRA1SB, tra1-SRR3413 lacking the N3580A mutation was integrated into the genome. As shown in Figure 1C, the integrated allele showed similar slow growth at elevated temperature and in the presence of calcofluor white as did the plasmid version.
Since the NuA4 complex has a role in DNA double-strand-break repair (Bird et al. 2002), we examined strains containing Tra1-SRR3413, P3408A, and S3463A more closely for sensitivity to MMS compared to a deletion of the NuA4 component Vid21 (vacuole import and degradation)/Eaf1. tra1-SRR3413 showed a slight sensitivity to MMS but dramatically less so than deletion of vid21/eaf1 (not shown). In support of only minor changes in double-strand-break repair, none of the three alleles resulted in enhanced sensitivity to γ-irradiation at 30° (not shown).
To evaluate whether the growth phenotypes might result from reduced expression of the altered forms of Tra1, the levels of Flag3x-tagged Tra1-P3408A, Tra1-S3463A, and Tra1-SRR3413 were examined by Western blotting of crude yeast extracts (Figure 2A). Each of these forms was expressed at a level approximately equivalent to Tra1SB, although we observed slightly elevated levels of Tra1-S3463A. We focused additional studies on tra1-SRR3413 since it displayed the most dramatic phenotypes. To determine if Tra1-SRR3413 associated with components of NuA4 and SAGA complexes, a TAP-tagged version of the molecule was purified by tandem affinity purification and the presence of Spt7 and Flag-tagged Yng2 determined by Western blotting (Figure 2B). SRR3413 interacted with Flag-Yng2 and with two forms of Spt7, suggesting its presence in NuA4, SAGA, and SAGA-like (SLIK) complexes. To further examine if Tra1-SRR3413 is found within SAGA and NuA4 complexes, crude extracts containing Tra1-SRR3413 were fractionated on a HiTrap Q sepharose column and the elution profile of HA-Ada2 and Flag-Yng2 was compared to that found in a strain containing Tra1SB (Figure 2C and 2D). The elution profiles for Tra1-SRR3413 varied little with Tra1SB, suggesting that it associates in the appropriate high molecular mass complexes.
Gene expression in the conditional tra1 strains:
To determine if the tra1-SRR3413, tra1-P3408A, and tra1-S3463A alleles resulted in changes in transcriptional regulation, we examined expression of PHO5, HIS4, INO1, and ADH2 using lacZ-reporter fusions (Figure 3A). Assays were performed at 30° under conditions that would induce each specific promoter. Strains deleted for gcn5, spt7, and yng2 were examined for comparison. The tra1 mutations resulted in expression of PHO5 at a level from 3 to 17% of that seen in the wild-type (TRA1SB), the extent paralleling the severity of the growth defects. Expression of HIS4 and INO1 was relatively unchanged in the tra1 backgrounds. ADH2 was decreased on average to 60% wild-type levels for the three tra1 mutants with the extent again paralleling the growth defect. The pattern of expression seen in the tra1 strains most closely resembled that seen upon disruption of gcn5; expression of HIS4 is reduced upon disruption of spt7 and yng2, unlike the tra1 strains and INO1 is reduced upon disruption of spt7. Elevated temperature did result in enhanced effects of tra1-SRR3413 as expression of INO1-lacZ was reduced to ∼50% of that seen in TRA1SB when cells were grown at 35° (Figure 3B).
A more detailed analysis of gene expression in the tra1-SRR3413, tra1-P3408A, and tra1-S3463A strains was obtained by performing microarray analysis. The Agilent yeast oligo array representing all independent yeast ORFs was probed with cRNA produced from cells grown logarithmically at 30° in YP media containing 2% glucose (note that this differs from the data shown in Figure 4, which were performed in inducing conditions). Two-dye experiments were performed comparing expression in the mutant strains to that found in a strain containing TRA1SB. For each mutant, the experiment was performed in duplicate with dye reversal. The tra1-SRR3413 allele resulted in a change in expression of twofold or greater of ∼7% of the protein encoding genes with ∼70% having decreased expression. The 25 genes showing the greatest increase and decrease in expression for each mutant are shown in Table 4 (the full data set has been submitted to the GEO database at NCBI under accession no. GSE6847). The genes affected in the three strains were highly similar, consistent with the mutant alleles having a common defect. Gene ontology of those most affected indicates a group of upregulated genes are involved with ion transport and cell cycle regulation, while a number of the downregulated genes have roles in nucleotide and amino acid metabolism.
Hierarchical cluster analysis was performed with the array data for the tra1-SRR3413 strain, strains with deletions of SAGA and NuA4 components, and a subset of the compendium data set (Hughes et al. 2000) including those genes linked to transcription. tra1-SRR3413 clustered in a leaf with a deletion of ada2, although the depth of the branch is indicative of only weak similarity (Figure 4). Profiles of strains with deletions of the NuA4 components (eaf5, yaf9, eaf1/vid21, and yng2) showed less similarity with tra1-SRR3413, than ada2, spt3, or spt8. Overall, the relatively low level of similarity indicates that the transcription pattern of tra1-SRR3413 results from the cumulative effects of this allele on both SAGA and NuA4 complexes and/or from there being one or more independent roles for Tra1.
Brown et al. (2000) previously identified a tra1 allele that resulted in decreased binding to the Gal4 transcriptional activation domain. To address whether the transcriptional changes observed for tra1-SRR3413 were due to a defect in its interaction with transcriptional activation domains, we compared the ability of Tra1-SRR3413 and Tra1SB to associate with the Gal4 activation domain (Gal4AD). Tandem- affinity-purified Tra1SB and Tra1-SRR3413, prepared from strains containing HA-tagged Ada2, were chromatographed on GST-Gal4AD columns. Association of SAGA with Gal4 was detected by Western blotting for HA-Ada2 (Figure 5A). SAGA containing Tra1-SRR3413 bound Gal4AD to the same extent as SAGA-containing Tra1SB (compare lanes 4 and 6), suggesting that the mutation does not decrease association with transcriptional activators. This is consistent with the region required for activator interaction being contained within amino acid residues 2233–2836 (Brown et al. 2001). Recruitment of Tra1-SRR3413 to the PHO5 promoter in vivo was examined by chromatin immunoprecipitation using the TAP-tagged derivatives of Tra1. As shown in Figure 5B, Tra1-SRR3413 was recruited as efficiently to the PHO5 promoter as Tra1SB (compare lanes 2 and 3, top; input DNAs are shown in the bottom).
To determine if Tra1-SRR3413 alters acetylation of either histone H3 or histone H4, chromatin immunoprecipitations were performed. We first compared the ratio of acetylated histone H4 (at lysine 8) to total histone H3 at the ADE1 and PHO5 promoters for the strain containing tra1-SRR3413 with strains containing deletions of the NuA4 components Vid21/Eaf1 and Yng2. Figure 6A shows the relative level of acetylated histone H4 as a percentage of that found in the wild-type strain. For both ADE1 and PHO5 promoters, tra1-SRR3413 resulted in a decrease in relative acetylation to ∼60% of that found in the TRA1SB strain. This decrease was comparable to that found upon deletion of yng2 and somewhat less than deletion of vid21/eaf1. For each mutant strain, a smaller effect was observed for sequences at the 3′ end of the PHO5 gene. Acetylation of histone H3 (at lysine 9) was then analyzed at PHO5, PHO84, and PGK1 promoters (Figure 6B), the latter being SAGA-independent. tra1-SRR3413 resulted in a decrease in the ratio of acetylated histone H3 to total histone H3 at PHO5 and PHO84 promoters that was ∼35% of that found in the wild-type tra1SB strain, approaching that seen in strains deleted for spt7 and ada2. At the PGK1 promoter, the decrease in acetylation of histone H3 was less apparent (∼70% of wild type for tra1-SRR3413). Although we cannot exclude an indirect role of Tra1 on acetylation, the simplest interpretation of these results is that the PI3K domain of Tra1 is required for optimal histone acetylation by both the NuA4 and SAGA complexes.
Tra1-SRR3413 results in shortened telomere length:
The similarity of Tra1 to many of the PI3K-domain-containing proteins involved in telomere biology (for example, Tel1 and Mec1) led us to test if the tra1-SRR3413, tra1-P3408A, and tra1-S3463A alleles affected telomere length. As an initial test we assayed the effects of these alleles on the ability of strains to generate and stably maintain linearized plasmids after transformation with circular plasmids containing inverted repeats of telomeric sequences (Lundblad and Szostak 1989). Strains containing tra1-SRR3413, tra1-P3408A, and tra1-S3463A were transformed with the LEU2-containing centromeric plasmid pVL106 (kindly supplied by V. Lundblad), which contains URA3 flanked by inverted telomeric sequences. Cells were grown in nonselective media and plated onto media containing 5-FOA and lacking leucine. 5-FOA-resistant/Leu+ colonies were counted and normalized to the number of total cells plated. As shown in Figure 7A, tra1-SRR3413 resulted in a reduced ability to linearize and maintain pVL106. To directly examine telomere length, Southern blotting was used to compare the terminal restriction fragments (TRFs) of chromosomes in the tra1-SRR3413 and TRA1SB strains. CY1021 containing wild-type TRA1 on a URA3-containing plasmid was transformed with TRA1SB or tra1-SRR3413 on TRP1-centromeric plasmids. The wild-type TRA1 allele was lost after selection on 5-FOA, establishing generation zero. Strains were grown through multiple generations, genomic DNA was isolated, and TRF length was determined after digestion of the genomic DNA with XhoI. Cells were grown at both 30° and 34°. The latter significantly reduces growth rate and from the linearization assay appeared to suppress the effects of SRR3413 on telomere shortening (not shown). As shown in Figure 7B, at 10 generations of growth, the TRFs were of approximately equal lengths in TRA1SB and tra1-SRR3413 strains. By 50 generations, TRFs were appreciably shorter in the tra1-SRR3413 strain, with a further shortening apparent at 100 generations when cells were grown at 30°. The median difference was ∼120 and 140 bp at 50 and 100 generations, respectively. The effect of SRR3413 was not apparent when cells were grown at 34°, perhaps the result of increasing the period available for telomere extension.
To examine the molecular basis for the tra1-SRR3413 phenotype, we initiated a suppressor analysis. tra1-SRR3413 was introduced into yeast strain CY1021 by plasmid shuffling. Approximately 2 × 106 cells were plated onto YPD containing 6% ethanol. Growth of two colonies was observed after 5 days. To determine if these were intragenic, the tra1-expressing plasmid was isolated from both strains and the phenotype reexamined after plasmid shuffling in yeast strain CY1021. In both cases the plasmid enabled growth in YPD containing 6% ethanol. The 3′ SalI–SacI fragment of each allele was sequenced and found to contain the SRR3413 mutation and a common mutation converting H3530Y. This mutation was shown to be sufficient for suppression by recloning the SalI–SacI fragment into TRA1SB and plasmid shuffling. H3530Y suppressed the growth defects associated with SRR3413 on both 6% ethanol-containing plates (Figure 8A) and at 37° (not shown). Allele specificity was examined by introducing H3530Y into wild-type and tra1-P3408A-containing backgrounds. H3530Y on its own had no effect on growth but did suppress the ethanol sensitivity due to P3408A and therefore is a general suppressor of the ethanol-sensitive class of tra1 alleles. As shown in Figure 8B, the effects of H3530Y are most likely due to suppression of transcriptional defects since this mutation suppressed the slow induction of PHO5 seen in a tra1-SRR3413 strain.
To detect genes whose disruption in parallel with tra1-SRR3413 generates a synthetic slow-growth phenotype, we performed a systematic genetic array (SGA) analysis in which a strain containing tra1-SRR3413 was crossed into the array representing the collection of ∼4700 viable haploid deletion strains. Candidate genetic interactions, scored as slow-growing colonies through two independent screenings, were examined individually after generation of double-mutant strains by tetrad dissection. Growth for these strains was examined on synthetic complete media at 34°. The 23 genes identified and their functions are listed in supplemental Table 2 at http://www.genetics.org/supplemental/. Consistent with the finding that tra1-SRR3413 resulted in sensitivity to ethanol and calcofluor white, 10 of these appear involved in cell-membrane and cell-wall processes. This number is greater if indirect roles, for example a contribution to phosphatidylserine synthesis by SER1 and SER2, are considered. The other prevalent groups represented were genes involved with transport and mitochondria function. A hierarchal cluster analysis was performed with the tra1-SRR3413 interacting genes and the data sets of Tong et al. (2004) and Pan et al. (2006). tra1-SRR3413 did not cluster closely with any of the strains used in these analyses. In fact, no common synthetic growth defects were found among those listed in the Saccharomyces Genome Database for NuA4 and SAGA components and those for tra1-SRR3413.
Tra1 is one of the largest yeast proteins and is essential for viability. As a component of both NuA4 and SAGA complexes, Tra1 interacts directly with DNA-binding transcriptional regulators recruiting these HAT complexes to promoters. On the basis of the similarity of this domain with that found in other cellular regulators it is likely that the PI3K domain of the molecule has a significant function. Indicative of a key role for the PI3K domain, eight targeted mutations within this region resulted in loss of viability. Although we lack detailed structural information for the Tra1/TRRAP subfamily, the alignment of these residues in comparison to the structure of porcine PI3K-γ is interesting. Residues R3567 and PFR3621 are not unique to Tra1/TRRAP and are found in the catalytic loop and activation loop, respectively. As is the case for the kinase members of the PI3K family (Dhand et al. 1994; Stack and Emr 1994), R3567 is required for function of Tra1. Similarly, PFR3621 is generally conserved within the activation loop of the PI3K family. The activation loop plays a critical role in substrate binding for these molecules (Bondeva et al. 1998), suggesting a similar essential role for this region in Tra1. Substrates for Tra1 are unknown. Potentially these could include phospholipids; however, we have detected only very weak binding of recombinant Tra1-PI3K domain to PIP strips (Echelon Biosciences, Salt Lake City; not shown).
Eight alleles with mutations in the PI3K domain supported viability and displayed phenotypes, which fell into two groups. The first group, tra1-P3446, -DKL3491, -P3612, and -N3617, was slightly resistant to overexpression of VP16, although to a much lesser extent than the ada genes, and was partially sensitive to hydroxyurea, characteristic of defects in DNA replication. This latter phenotype has not been previously ascribed to SAGA or NuA4 components and may reflect a novel function of one or both of these complexes or perhaps defects in expression of genes involved in this process. The second group of alleles containing mutations of P3408, SRR3413, S3463, and to a lesser extent EER3416, was more marked. These alleles resulted in temperature-sensitive growth and reduced growth in low-phosphate media and in media containing 6% ethanol. The latter result has also been noted for deletions of ada2 and on the basis of their reduced growth in the presence of calcofluor white, this phenotype likely reflects changes in cell-wall properties (Takahashi et al. 2001) or the response to cell-wall stress. Unlike deletions of Ada components, the tra1-P3408A, -SRR3413, and -S3463A alleles did not result in resistance to overexpression of VP16, nor is their temperature-sensitive growth indicative of deletions of ada genes in the KY320 strain background. tra1-P3408 and tra1-SRR3413, the alleles with the most pronounced phenotypes, were partially sensitive to the microtubule-destabilizing agent benomyl. This is characteristic of deletions of NuA4 components, but again for the tra1 alleles the effect was not as severe. The tra1-SRR3413 allele was the only allele showing sensitivity to MMS but dramatically less so than for deletions of components of the NuA4 complex. Growth of tra1-P3408, SRR3413, and S3463 strains was modestly reduced on media depleted of inositol, a characteristic of deletions of spt genes; however, unlike deletion of spt7, the inositol auxotrophy of the tra1 alleles was not the result of reduced expression of INO1. Cumulatively, these phenotypes suggest that Tra1 possesses functions overlapping yet distinct from SAGA or NuA4 components.
The transcriptional effects of the tra1-P3408A, SRR3413, and S3463A alleles were compared to deletions of spt7, gcn5, and yng2 using lacZ fusion reporters. The transcriptional change for the tra1 alleles was less than that seen upon deletion of the NuA4 or SAGA components but most closely resembled that seen upon deletion of gcn5. Microarray analyses confirmed the transcription defect arising from these alleles and that they were acting by altering a common function. Cluster analysis vs. the expression profiles seen for deletions of SAGA and NuA4 components showed the greatest similarity, though not robust, to ada2. Comparison of the growth phenotypes of strains containing deletion of strains carrying single and double mutations of ada2Δ0 and tra1-P3408A confirmed that the tra1 mutations were not functioning exclusively through effects on the ADA genes (data not shown). This lack of similarity suggests that the transcriptional effects of the tra1 alleles result from the integrative effect of disturbing histone acetylation by both SAGA and NuA4 complexes and/or the possibility that the PI3K domain of Tra1 has one or more roles distinct from other members of these complexes. The mechanism by which the tra1 mutations affect histone acetylation is unclear; interestingly, in the three-dimensional structure of the SAGA complex revealed by electron microscopy, Tra1 and Gcn5 are positioned on opposite sides of a nucleosome-sized cleft (Wu et al. 2004), suggesting that Tra1 may play a role in orienting nucleosomes for acetylation.
The finding of a generation-dependent telomere shortening in the tra1-SRR3413 strain points to a role of Tra1 in telomere elongation or maintenance. The shortened telomeres may be the indirect result of Tra1 regulating expression of genes required for these processes. In this regard 11 genes whose deletion results in shortened telomeres (Askree et al. 2004) have decreased expression of ≥1.5-fold in the tra1-SRR3413 strain. Most notable of these is EST3 with decreased expression of 2.4-fold. Alternatively, Tra1 may play a direct role in telomere elongation or maintenance. We do not believe that the Tra1 effect is mediated solely through one of SAGA or NuA4-dependent histone acetylation, as component genes were not found in the genomewide screen of Askree et al. (2004) or in our independent analyses of Spt7, Gcn5, and Eaf7 (data not shown). Additional studies will be required to determine if Tra1 is recruited to telomeres and the role of the PI3K domain in telomere elongation.
The ethanol and calcofluor white sensitivity of tra1-SRR3413 as well as its genetic interactions with genes involved in cell-wall and membrane processes is consistent with the involvement of SAGA in the response to stress (Huisinga and Pugh 2004). Tra1 may be required to trigger transcriptional changes necessary when membrane processes are disturbed during cellular stress. Associations between SAGA components and membrane processes have previously been suggested by the finding that Gcn5 and Spt20 are required for the unfolded protein response (Welihinda et al. 1997, 2000). Interestingly, deletion of a number of genes involved in vesicular trafficking, including vps9, vps15, vps28, vps34, vps36, vps39, bro1, and sur4, results in both telomere shortening and ethanol sensitivity similar to that in tra1-SRR3413. The other two gene deletions with both of these phenotypes are bem2 and sit4, each of which is involved in cytoskeletal and cell-wall organization. (Expression of these genes is not reduced in the tra1-SRR3413 background.) Other studies connect NuA4 components with membrane processes; most notably, vid21/eaf1 (vacuole import and degradation) was identified in a screen for defects in sorting of carboxypeptidase Y to the vacuole (Bonangelino et al. 2002), is sensitive to ethanol (Fujita et al. 2006), and interacts with Vac8, which is required for several aspects of vacuolar function (Tang et al. 2006).
Given the importance of PI3K domains in key signaling proteins and the conservation of this domain in the Tra1/TRRAP family of molecules, it is reasonable to consider possible activities of the PI3K domain. In this regard P3408, SRR3413, and S3463 map proximally to what would be the cleft between N- and C-terminal lobes of the PI3K-γ structure. P3408, in fact, aligns with K833, which in PI3K-γ interacts with the α-phosphate of ATP and is covalently modified by the kinase inhibitor wortmanin (Wymann et al. 1996). While clearly an approximation of the Tra1 structure, S3463 may be located within the cleft at a position that putatively could be involved with substrate interaction. Though speculative, it is attractive to propose that the putative cleft of Tra1 is a substrate-binding pocket, where the domain catalyzes the transfer of a modifying unit other than phosphate moieties, transfers water as a hydrolase, or perhaps has a role in the nuclear signaling by inositol polyphosphates (for example, Steger et al. 2003).
We thank Brian Shilton, Catherine Bateman, David Litchfield, and Greg Gloor for their comments on the manuscript; Victoria Lundblad, Fred Winston, Kathy Gould, Joan Curcio, V. Zakian, and Raymund Wellinger for reagents; Shayna Oldegard, Dana Abrassart, Monica Piasecki, Courtney Coschi, and Kim Grant for assistance in preparation of DNA constructs; and Joe Martens and Melissa Bradford for assistance with the chromatin immunoprecipitations. This work was supported by Canadian Institutes of Health Research grant to C.J.B. S.M.T.H. is a holder of a Natural Sciences and Engineering Research Council studentship. C.H. was supported by an Ontario Graduate Scholarship in Science and Technology scholarship. A.I.M. is a holder of a Western Graduate Research Scholarship.
Microarray data from this article have been submitted to the GEO database at NCBI under accession no. GSE6847.
↵1 These authors contributed equally to this work.
↵2 Present address: Department of Molecular and Cellular Biology, University of Guelph, Guelph, ON N1G 2W1, Canada.
Communicating editor: M. Hampsey
- Received April 12, 2007.
- Accepted July 10, 2007.
- Copyright © 2007 by the Genetics Society of America