|Home | About | Journals | Submit | Contact Us | Français|
Tra1 is an essential protein in Saccharomyces cerevisiae. It was first identified in the SAGA and NuA4 complexes, both with functions in multiple aspects of gene regulation and DNA repair, and recently found in the ASTRA complex. Tra1 belongs to the PIKK family of proteins with a C-terminal PI3K domain followed by a FATC domain. Previously we found that mutation of leucine to alanine at position 3733 in the FATC domain of Tra1 (tra1-L3733A) results in transcriptional changes and slow growth under conditions of stress. To further define the regulatory interactions of Tra1 we isolated extragenic suppressors of the tra1-L3733A allele.
We screened for suppressors of the ethanol sensitivity caused by tra1-L3733A. Eleven extragenic recessive mutations, belonging to three complementation groups, were identified that partially suppressed a subset of the phenotypes caused by tra1-L3733A. Using whole genome sequencing we identified one of the mutations as an opal mutation at tryptophan 165 of UPF1/NAM7. Partial suppression of the transcriptional defect resulting from tra1-L3733A was observed at GAL10, but not at PHO5. Suppression was due to loss of nonsense mediated decay (NMD) since deletion of any one of the three NMD surveillance components (upf1/nam7, upf2/nmd2, or upf3) mediated the effect. Deletion of upf1 suppressed a second FATC domain mutation, tra1-F3744A, as well as a mutation to the PIK3 domain. In contrast, deletions of SAGA or NuA4 components were not suppressed.
We have demonstrated a genetic interaction between TRA1 and genes of the NMD pathway. The suppression is specific for mutations in TRA1. Since NMD and Tra1 generally act reciprocally to control gene expression, and the FATC domain mutations do not directly affect NMD, we suggest that suppression occurs as the result of overlap and/or crosstalk in these two broad regulatory networks.
Tra1 is a 3744 amino acid residue protein, essential for viability in Saccharomyces cerevisiae. It is a major constituent of the SAGA and NuA4 transcriptional regulatory complexes [1-3], both with significant roles in gene regulation and DNA repair [4-6]. More recently a putative complex based on mutual associations termed ASTRA, was also found to contain Tra1 . Tra1's mammalian homolog TRRAP was identified because of its interactions with the transcription factors c-myc and E2F . Similarly Tra1 interacts with yeast transcriptional activators to target SAGA and NuA4 to promoters [9-12]. Interestingly, Helmlinger et al.  have recently provided evidence that Tra1 also acts independently of SAGA and NuA4 to regulate gene expression.
Tra1/TRRAP are members of the PIKK (phosphoinositide three-kinase-related kinase) family of proteins, a group that in yeast includes Tor1, Tor2, Tel1 and Mec1 [14,15]. The latter two are structurally and functionally related to ATM and ATR of multicellular eucaryotes. Two additional family members not found in yeast, are the DNA-PKcs (DNA-dependent protein kinase catalytic subunit) and SMG-1, with key roles in DNA repair and nonsense mediated decay of RNA, respectively. The PIKK family members are all large proteins characterized by a common arrangement of C-terminal domains , including a domain that resembles the phosphatidylinositol-3-kinases (PI3K). Unlike the other PIKK molecules, which are protein kinases, Tra1/TRRAP lacks kinase activity [2,8]. Nonetheless, altering residues that parallel key regions of the kinase members of the family affect Tra1 function . On the N-terminal side of the kinase domain is the HEAT and TPR repeat-rich FAT (FRAP-ATM-TRRAP) domain [18-21]. C-terminal to the PI3K domain is the less highly conserved PRD (PIKK regulatory domain), identified in ATM as the site of acetylation by TIP60 .
The ~35-residue FATC domain is at the extreme C-terminus of the PIKK proteins . The critical role of the FATC domain is evident from the finding that addition of a single glycine to the C-terminus of Tra1 abolishes function, and mutations of L3733 or F3744 to alanine result in slow growth in a number of stress conditions . The FATC domain is similarly important for the other PIKK family members; for example, the parallel mutation to L3733A of Tra1 results in a dramatic loss in the kinase activity of SMG-1 . Dames et al.  determined the structure of the isolated FATC domain of S. cerevisiae Tor1. It is predominately helical with a loop at the extreme C-terminus held in place by a disulphide linkage. The helical structure is likely conserved in the family, but not the loop because the two cysteines are only found in the Tor proteins. FATC domains are proposed to be a target for interacting proteins. In a two-hybrid analysis the FATC domain of Mec1 was required for association with the RPA components Rfa1 and Rfa2 . In vivo, the FATC domain of ATM and Tip60 (the mammalian homolog of the NuA4 component Esa1) interact, though this may be indirect [22,27]. Consistent with the mutagenesis analysis of Moritia et al. , Lempiäinen and Halazonetis  suggest the FATC domain interacts with and regulates the activity of the kinase domain.
Nonsense mediated decay (NMD) is a cellular surveillance mechanism present in all eucaryotes that scans for premature stop codons on mRNAs [28,29]. The process is coupled to translation, and results in degradation of potentially deleterious transcripts [30-32]. Premature stop codons can arise through errors in transcription by RNA Polymerase II, failure or inaccurate removal of introns, inaccurate translational starts, ribosomal frameshifting, RNA editing, or errors within the genomic DNA. The mechanism of NMD varies in different organisms, with the molecular details still being determined . In yeast, one model suggests that NMD is triggered if the mRNA binding protein Hrp1 is bound to a downstream sequence element(s) (DSE) when the ribosome encounters a stop codon [34-36]. The DSE is 5' of the native stop codon, with Hrp1 normally being displaced by the passing ribosome. A second model, the faux (false) 3' UTR model, postulates that the length of the 3' untranslated region resulting from a premature stop codon prevents the normal interactions that occur between the ribosome and poly (A) binding protein during the termination of translation. In the absence of these interactions the NMD factors associate with the ribosome, which in turn results in mRNA decay .
The yeast NMD surveillance complex consists of three proteins, Upf1/Nam7, Upf2/Nmd2, and Upf3 (for simplicity referred to as Upf1, Upf2 and Upf3, respectively). Mutations in any of the NMD proteins increase translational read-through and stabilize mRNA containing premature stop codons [30,38-43]. Upf1, an ATP dependent RNA helicase [41,44] is the central component of the pathway, whereas Upf2 and Upf3 regulate Upf1 [42,45]. In metazoans, Upf1 is further regulated by phosphorylation/dephosporylation . The kinase involved is the PIKK family member SMG-1 [45,46].
In addition to removing aberrant transcripts, the NMD pathway plays a significant role in the control of gene expression . Microarray analyses indicate that approximately 10% of yeast genes are affected by loss of NMD; most showing increased levels [48-50]. Approximately half of the changes are the result of direct regulation by NMD . Some of the effects on translational readthrough are due to increased expression of the magnesium transporter Alr1p and the elevated cellular magnesium . Other regulatory functions may arise through programmed ribosomal frameshifting, and altered translational start site selection [52-55].
To identify the genetic network in which Tra1 is involved, we selected for second site suppressors that allow growth of strains with the tra1-L3733A allele on media containing ethanol. An opal mutation at codon 165 of upf1 was identified. Suppression was likely the result of loss of NMD since deletion of upf1, upf2 or upf3 conferred growth of strains containing tra1-L3733A. Suppression was specific for mutations to TRA1; phenotypes arising from deletions of SAGA or NuA4 components were not suppressed. Since deletion of upf1 only reversed the transcriptional defects of the tra1 alleles at a subset of affected promoters, and these same tra1 alleles did not affect NMD, we conclude that loss of NMD suppresses the TRA1 mutations through crosstalk and/or partial overlap of their regulatory networks.
The FATC domain of Tra1 is critical to the protein's function. Mutation of leucine 3733 to alanine results in temperature sensitivity and slow growth in media containing ethanol, Calcofluor white or rapamycin . These phenotypes provide a tool to probe the genetic network in which TRA1 is connected. We selected for extragenic suppressors of the ethanol sensitivity of CY4018, a strain disrupted for the genomic tra1 but viable due to tra1-L3733A on a URA3 centromeric plasmid. Approximately 108 cells were plated onto YPD plates containing 4% ethanol. Forty-three colonies were isolated for further analysis. Each was retested for the suppression after colony isolation, and then examined for whether the mutation was located on the tra1-L3733A containing-plasmid by plasmid shuffling. Eleven strains with extragenic mutations that partially suppressed the slow growth due to tra1-L3733A were identified. The growth of these strains on YPD media and YPD containing 4% ethanol is shown in Figure Figure1a.1a. The strains with the tra1-L3733A suppressor alleles (tentatively termed es2, es12, es35 etc.) were mated with the MATα tra1-L3733A strain CY5522 to determine if the suppressor mutations were dominant or recessive. Figure Figure1b1b shows the analysis for CY5579 (es2), CY5584 (es38), and CY5587 (es41). The diploid strains grew slowly under selective conditions, indicating that the suppressor alleles act recessively. Similarly each of the additional es alleles was recessive. Complementation groups were analyzed through crosses of the es containing strains. As shown in Figure Figure1c,1c, the diploid crosses of es2 with es38, and es41, grew poorly on media containing 4% ethanol, in contrast to the homozygous diploid containing es2. In this way three complementation groups were identified amongst the eleven strains isolated in the screen. One complementation group contained uniquely es2; a second group contained uniquely es35. The third complementation group contained the remaining nine alleles, including es38 and es41. A random spore analysis indicated that the ethanol resistance for the es2, es35 and es38 strains (representing each complementation group) segregated 2:2 suggesting that a single gene was the cause of suppression.
To determine the identity of the suppressor alleles for the es2 and es38/es41 complementation groups, we compared the genome sequence for each of the three strains relative to the parent CY4018. (The third complementation group was not analyzed.) Libraries were prepared and genomic sequencing performed using the ABI SOLiD 4.0 platform at the Centre for Applied Genomics at The Hospital for Sick Children (Toronto, Canada). Approximately 50 million reads were obtained for each sample of which 60% mapped to the reference genome from the Saccharomyces Genome Database. Polymorphisms not found in CY4018 or the other complementation group, were analyzed by visual inspection of the sequencing reads. A causative mutation could not be identified within coding or noncoding sequences for the es38/es41 complementation group. The es2 strain contained an opal mutation at tryptophan codon 165 of UPF1, truncating the 971-residue protein. To confirm that this allele, now designated upf11-164, was responsible for the suppression of tra1-L3733A, CY5579 (upf11-164 tra1-L3733A) was mated with CY4018 (tra1-L3733A), sporulated, and the UPF1 alleles in 8 unrelated spore colonies, four exhibiting slow growth and four fast growth, were isolated by PCR and sequenced. The four spore colonies growing slowly on 4% ethanol contained wild-type UPF1, whereas the fast growing spore colonies contained upf11-164.
In S. cerevisiae, Upf1 is one of three proteins acting in the NMD surveillance complex, the others being Upf2 and Upf3. To determine if loss of this process was responsible for suppression of tra1-L3733A, we analyzed the growth of strains deleted for upf1, upf2 or upf3 in media containing 6% ethanol and at 37°C. As shown in Figure Figure2A,2A, deletion of any of the components of the NMD surveillance complex partially suppressed the slow growth due to tra1-L3733A. Though not documented as a target of NMD, it is possible that loss of NMD suppresses the tra1-L3733A allele by increasing the cellular concentration of the protein. We therefore constructed strains that contain an integrated copy of Flag5-tagged Tra1 or Tra1-L3733A, and examined expression in the presence or absence of upf1 by Western blotting of cell extracts (Figure (Figure2B).2B). The expression of the integrated Flag5-Tra1-L3733A is similar to the wild-type protein (compare lanes 1 and 3). Disruption of upf1 slightly increased expression of Tra1-L3733A (compare lanes 3 and 6). We estimate this increase to be less than 10%, and suggest that it is not sufficient to account for the suppression.
We next addressed the allele specificity of the suppression. Alteration of the terminal phenylalanine of tra1 to alanine results in slow growth in media containing ethanol. A strain deleted for upf1 and containing tra1-F3744A was constructed and its growth compared to the single mutant strains (Figure (Figure3a).3a). Deletion of upf1 suppressed the slow growth due to tra1-F3744A, at least to the same extent as for tra1-L3733A. Deletion of upf1 also suppressed slow growth at 37°C and on media containing 6% ethanol caused by tra1-SRR3413, a triple alanine scanning mutation to residues 3,413-3,415 within the PI3K domain (; Figure Figure3b).3b). Deletions of other components of the SAGA and NuA4 complexes were then examined (Figure (Figure3c3c and and3d).3d). Deletion of ada2 results in slow growth on media containing ethanol; however, unlike the tra1 mutations, the slow growth caused by ada2Δ was not suppressed by upf1Δ (Figure (Figure3c).3c). Disrupting the NuA4 components Eaf3 or Eaf7 results in slow growth at 35°C in media containing 6% ethanol. Deletion of upf1 did not suppress this phenotype for either the eaf3Δ or eaf7Δ strains (Figure (Figure3d).3d). In fact synthetic slow growth was observed on plates containing 6% ethanol for the double mutant strains. This latter result agrees with the synthetic slow growth reported for mutations of esa1 and eaf7 with upf1 and upf3, respectively [56,57].
Alterations to the FATC domain cause slow growth in a number of conditions . We analyzed which of these in addition to the suppression of ethanol and temperature sensitivity are suppressed by upf1Δ. As shown in Figure Figure4,4, deletion of upf1 suppressed the slow growth resulting from tra1-F3744A when galactose is the carbon source, when phosphate is depleted, and when phleomycin (a DNA damaging agent) is present in the media. In contrast deletion of upf1 did not suppress slow growth due to tra1-F3744A in the presence of the cell wall destabilizing agent Calcofluor white, and only had a modest effect with rapamycin. The latter was particularly interestingly, given that deletion of upf1 in the wild-type TRA1 background decreased sensitivity to rapamycin (compare TRA1 UPF1 with TRA1 upf1Δ).
We next addressed whether disruption of upf1 would reverse the transcriptional defects caused by the TRA1 mutations. Expression of two lacZ promoter fusions was examined: PHO5-lacZ and GAL10-lacZ. tral-L3733A and tra1-F3744A decreased activated expression of PHO5-lacZ to approximately one-third of the wild-type level (Figure (Figure5a).5a). Interestingly, the strain containing a disruption of upf1 in the context of wild-type TRA1 also showed decreased expression (approximately 2-fold). Furthermore, disruption of upf1 in the context of either tral-L3733A or tra1-F3744A did not restore transcription. As shown in Figure Figure5b5b expression of GAL10-lacZ was also reduced by tral-L3733A and tra1-F3744A, but in contrast to PHO5-lacZ not by deletion of upf1. In addition, at GAL10 the transcriptional defect due to the tral alleles was partially reversed by upf1Δ. Together these results suggest that in some, but not all, cases disruption of upf1 may suppress tra1-induced phenotypes by regulating transcription (most likely indirectly) of common genes.
Our previous studies with TRA1 have linked phenotypes resulting from its mutation with transcriptional change [17,23]. The observation that disrupting upf1 did not restore transcription in all cases, suggested two models for how loss of NMD might suppress the FATC domain mutations. The first predicts a direct link between NMD and Tra1. If Tra1 is a negative regulator of NMD, Tra1-L3733A might increase NMD, decreasing the level of certain mRNA transcripts, which in turn could cause growth-related phenotypes. Loss of the NMD pathway would reverse the effect. The interaction between SAGA component Sgf29 and Upf1 , and the phosphorylation of Upf1 by the PIKK member SMG-1 in metazoans [45,46,59], are consistent with the possibility of direct regulation. The second model predicts a less specific interaction between Tra1 and NMD. The tra1 mutations through its action in SAGA, NuA4 or independently  alter expression of a set of genes. Some of these genes, or genes with epistatic relationships, may be regulated through mRNA turnover and/or translational readthrough by processes involving NMD. This possibility is enhanced by the scope of both networks and their generally reciprocal nature. By eliminating NMD the expression of genes that directly or indirectly intersect in the pathways may return to near normal, effectively compensating for diminished Tra1 activity. To differentiate between these models, we analyzed whether tral-L3733A alters NMD. We constructed a PGK1-lacZ fusion (PGK1fs-lacZ; Figure Figure6a)6a) frameshifted at codon 166 and including those sequences required for NMD, -544 to 891 [34,60]. Previous mRNAseq experiments have indicated that PGK1 expression in YPD media is relatively unaffected by mutations within the FATC domain of Tra1 . Deletion of upf1 results in a 5-fold increase in expression of β-galactosidase from the PGK1fs-lacZ fusion in comparison to a wild-type UPF1 strain (Figure (Figure6b),6b), verifying the use of this construct to monitor NMD. Indicative of Tra1 not having a direct role in NMD, expression of PGK1fs-lacZ was unaffected by tral-L3733A. In light of this result and the broad role for NMD in gene expression, we thus favor a model whereby deletion of NMD suppresses the tra1 alleles through its reciprocal action on an overlapping set of genes.
We have identified a genetic interaction between components of the nonsense mediated decay pathway and tra1, through random selection for mutations that suppress the ethanol sensitivity of a tra1-L3733A strain. Our initial selection identified a nonsense mutation at codon 165 of UPF1. Similar suppression was found with deletions of upf1, upf2 and upf3 indicating that loss of nonsense mediated decay was the likely cause. Our study demonstrates a relationship between the Tra1 and NMD regulatory networks, and further emphasizes the general importance of NMD in gene expression.
The suppression mediated by deletion of upf1 was specific for certain phenotypes arising from tra1 mutations. Slow growth due to high temperature, ethanol, low phosphate, galactose as the primary carbon source, and phleomycin was suppressed. Slow growth due to rapamycin or Calcofluor white was not. The suppression mediated by upf1Δ was also specific for mutations within TRA1, not suppressing deletions of the SAGA component Ada2 or the NuA4 components Eaf3 or Eaf7. In the case of the latter two disruptions, synthetic slow growth was observed.
We propose the suppression of tra1-L3733A and tra1-F3744A caused by loss of NMD is the result of crosstalk and/or direct overlap in the networks regulated by these genes. In this model, altered (most often decreased) expression resulting from the tra1 mutations would be partially reversed by reducing mRNA turnover (or enhancing translational readthrough) in the NMD deficient backgrounds. Also along this line, a gene regulated by NMD may encode a protein that directly or indirectly regulates some of the functions of Tra1. Our reasons for preferring this indirect mechanism for suppression are the following. First, consistent with this model nonsense mediated decay has a broad role in gene regulation. Some of NMD's roles relate to the removal of aberrant transcripts that have acquired nonsense codons. Other roles relate to control mechanisms that utilize the pathway to remove mRNAs that would otherwise be functional. As such NMD influences approximately 10% of yeast genes [48,49]--likely an underestimate of the extent of its control as this is in rich media. Clearly the breadth of the NMD effect indicates that NMD is of global importance in yeast regulatory pathways, not only affecting aberrant transcripts. The majority of genes are upregulated in response to loss of NMD; this contrasts to the generally more prevalent decreased expression observed upon mutation of tra1 [17,23]. Thus the prevalence of genes affected and reciprocal nature of loss of NMD and Tra1 could result in their neutralization in a double mutant background. Indeed, it is the scope of NMD that likely explains why its loss can also suppress mutations of other globally important yet functionally diverse factors required for gene expression (for example: TAF6, TAF9 and RAP1 , PAF1 , and BRE1 ).
This indirect model for suppression accommodates the finding that only a subset of tra1-L3733A phenotypes is suppressed by loss of NMD. Suppression would require that Tra1 and NMD regulate key genes responsible for the phenotype (also see below). Similarly, specificity for TRA1 mutations may be accounted for if other NuA4 and/or SAGA components influence genes to an extent that is not sufficiently reversed by NMD. Finally, the finding that loss of NMD affects promoter-dependent events of some tra1-effected genes (for example GAL10) but not others (for example, PHO5) is more consistent with an indirect mechanism for suppression that could act through distinct genes and/or steps in gene expression, rather than by restoring Tra1 function.
Alternative models for suppression by upf1Δ are possible if Tra1 were involved in NMD. The finding that tra1-L3733A and tra1-F3744A did not alter expression of an internally frame-shifted PGK1 reporter plasmid, suggests that Tra1 is not directly involved in NMD. The lack of direct relationship between Tra1 and NMD is consistent with Tra1 acting in the nucleus, whereas in S. cerevisiae NMD is primarily a cytoplasmic process regulating post-transcriptional events .
We have not pinpointed the genes whose regulation by NMD allows suppression of the phenotypes caused by tra1-L3733A. Because of epistatic relationships, the key genes affected by tra1-L3733A and NMD may not be identical. It is also possible that small changes in multiple target genes could cause suppression; this would make identification of relevant targets difficult. Expression screening of mRNAs to detect changes in profiles may be complicated because the genes are likely stress induced, and will differ from condition to condition. Moreover, some aspects of NMD relate specifically to translation, and will not be seen by RNA profiling. Nonetheless, we have compared gene expression profiles in YPD media for tra1-L3733A  and upf1Δ  strains. Of the 79 genes with reduced expression due to tra1-L3733A (twofold or greater), six display increased expression of twofold or greater with upf1Δ (one has decreased expression). This ratio, given that it approximates the 10% of the genome regulated by NMD, does not support a specific overlap in the pathways, however it does emphasize the generally reciprocal nature of Tra1 and NMD. The six genes with reciprocal changes in expression are YBL107C, YER187W, MIP6 (YHR015W), JMN1 (YMR294W), YNR071C and YOL014W. Of these only MIP6 and JMN1 have characterized functions. MIP6 is of potential relevance since it encodes a protein with putative RNA binding motifs, and was identified in a two-hybrid analysis as interacting with the Mex67, an mRNA export factor .
When expressed on a centromeric plasmid from the DED1 promoter, Tra1-L3733A is less abundant than the wild type protein . We do not observe this decrease when FATC mutations are integrated into the genome. For this reason the functional experiments performed in this analysis were with genomically encoded tra1-L3733A and tra1-F3744A. Based on the recent results of Stirling et al. , who show a link between chromosome instability and components of the ASTRA complex, we believe that the plasmid versions may be less well expressed due to decreased stability of the plasmid for the mutant versions of tra1.
In a recent independent selection for suppressors of tra1-F3744A, we identified two alleles of the ASTRA component Tti2 (Genereaux et al. Genetics, in press). The tti2 alleles also suppress tra1-L3733A, but not deletions of components of SAGA or NuA4 components. The tti2 alleles, unlike upf11-164, acted dominantly. This as well as complementation experiments with TTI2 (not shown) suggests that es38/41 and es35 represent additional independent suppressor alleles.
We have demonstrated a genetic interaction between TRA1 and genes of the nonsense mediated decay pathway. In a recessive manner, deletion of upf1 partially suppressed the growth related defects of tra1-L3733A, tra1-F3744A and tra1-SRR3413, mutations within the FATC and PI3K domains. The suppression was specific for TRA1 mutations; no effect was seen for deletions of other SAGA or NuA4 components. A subset of phenotypes attributable to the FATC mutations was suppressed; furthermore, not all transcriptional defects were reversed by deletion of upf1. We suggest that the suppression relates to the breadth and overlapping, yet generally reciprocal nature of the gene regulatory pathways in which Tra1 and the NMD components are involved.
Strains for selection of suppressor mutations are derivatives of KY320 (; see Table Table1)1) and the isogenic MATα strain CY4413. CY1021 contains a genomic disruption of tra1 and is maintained by a plasmid copy of myc-tagged TRA1 expressed from the DED1 promoter . CY3003  and CY4018 were obtained from CY1021 by plasmid shuffling and contain myc9-tagged tra1-L3733A expressed from the DED1 promoter on YCplac22  or a URA3 centromeric plasmid (YCplac22u) derived from YCplac22 by switching TRP1 to URA3, respectively . CY5522 is the MATα equivalent of CY4018 and was generated by mating CY4413 with CY4018. After sporulation, a MATα spore colony was isolated that required the plasmid copy of tra1-L3733A for growth. Strains carrying tra1-L3733A and extragenic suppressors (hereafter defined as es alleles) es2 (CY5579), es12 (CY5580), es35 (CY5581), es36 (CY5582), es37 (CY5583), es38 (CY5584), es39 (CY5585), es40 (CY5586), es41 (CY5587), es42 (CY5588), and es43 (CY5750) were derived from CY4018 using the selection scheme described below. CY5666 (es2), CY5758 (es38), and CY5603 (es41) are MATα equivalents of CY5579, CY5584, and CY5587, respectively, and were made after mating with CY5522. MATα spore colonies carrying the suppressor were selected based on their ability to grow at high temperature and on plates containing 4% ethanol.
Yeast strains deleted for upf1 (CY5932), upf3 (CY5936) and upf2 (CY5934) were derived from the diploid strains, BY46214, BY44702 and BY41905 , respectively, by selecting Kanr spore colonies. These MATa strains were mated with CY4057 (tra1-L3733A; ), to yield after sporulation and selection of Kanr His - spore colonies CY5972 (upf1Δ tra1-L3733A), CY5983 (upf3Δ tra1-L3733A), and CY5996 (upf2Δ tra1-L3733A). Similarly, CY6030 (upf1Δ tra1-F3744A) was generated after mating of CY4351 (tra1-F3744A) and CY5932. CY5939 (upf1Δ::nat1) was obtained by gene replacement of Kanr with nat1. Double mutant strains of upf1Δ::nat1 with ada2Δ (CY5979), eaf3Δ (CY5980), and eaf7Δ (CY5976) were obtained after mating and sporulation of the diploids of crosses of CY5939 with BY4282, BY7143, and BY2940, respectively. CY6111 (tra1-SRR3413 upf1Δ) was derived after mating of CY2220  and CY5932, and selecting Ura + Kanr spore colonies. CY6004, CY6005 contain a TRA1 allele with a N-terminal Flag5-tag integrated with a URA3 marker.
Growth comparisons were performed on YP media containing 2% glucose (YPD) selective plates after 3-5 days at 30°C unless stated otherwise. Standard concentrations used for the selections are as follows: 7.5 μg/ml Calcofluor White (Sigma-Aldrich, Inc.), 4% or 6% ethanol, 1.0 μg/ml phleomycin (Sigma-Aldrich, Inc.), and 2 nM rapamycin (LC Laboratories, Woburn Ma). We note that the KY320 background is more sensitive to ethanol than the BY4741 background, accounting for the use of either 4% or 6% ethanol, respectively.
Myc-tagged tra1 alleles on the centromeric TRP1 (YCplac22) or URA3 (YCplac33) plasmids have been described [17,23]. PHO5 (-452 to +47) and GAL10 (-595 to -245) promoter-lacZ fusions in the LEU2 centromeric plasmid YCp87 have been described [17,72]. The PGK1-lacZ fusion with a frameshift mutation (PGK1fs-lacZ) was constructed by synthesizing the gene from -544 to +891 by PCR using oligonucleotides 5'-GCGGATCCACGTGGCCTCTTATCGAG-3' and 5'-CTCAAGCTTCCTTGGTGTTGGCATCAGCAGAG-3', digesting the gene with Asp718, creating blunt ends with the Klenow fragment of DNA polymerase and religation.
Six cultures of CY4018 were grown overnight in YPD. Approximately 100 million cells were plated onto 200 YPD plates containing 4% ethanol and incubated at 30°C for 5 days. Fast growing colonies were colony purified and retested. YCplac22-tra1-L3733A was transformed into each potential suppressor strain and YCplac22u-myc9-tra1-L3733A shuffled out on 5-fluoroorotic acid. Strains that retained their ability to grow on 4% ethanol were defined as containing extragenic suppressors.
Genomic DNA was prepared from10 mL of lyticase treated cells . Approximately 5 μg of DNA from each sample was sent to the Centre for Applied Genomics (Toronto, Ontario). DNA library construction and next-generation sequencing using paired-end reads was performed at the Centre. Samples were sequenced using the Applied Biosystems SOLiD 4.0 next-generation sequencing platform. The sequencing was performed in a single lane with multiplexing that included 11 additional unrelated samples. The Saccharomyces cerevisiae genome sequence was downloaded from the Saccharomyces Genome Database (SGD; ) on March 24, 2011. Custom Shell and Perl scripts were written for the sequencing analysis. The program Bowtie , allowing up to three mismatches per read, was used to map the colorspace reads to each chromosome of the yeast genome and obtain mapped reads in SAM format (Sequence Alignment/Map; ). The VCF (variant call format) from SAMtools  was used to obtain a raw list of polymorphisms from the mapped reads. Those reads with a Phred quality score below 20 were eliminated to obtain a filtered list of polymorphisms. A custom Perl script was written to eliminate the background polymorphisms found in wild-type samples.
Yeast strains containing lacZ-promoter fusions were grown to stationary phase in media lacking leucine. Assays with PHO5-lacZ in media depleted of phosphate and GAL10-lacZ in media containing galactose as the sole carbon source were performed as described in Mutiu et al.  and Brandl et al. , respectively, with ο-nitrophenol-β-D-galactosidase as substrate and normalizing values to cell density. Assay of PGK1fs-lacZ was performed similarly in YPD media. Results presented are from a minimum of four replicates with the standard errors indicated.
Western blotting was performed using PVDF membranes and anti-Flag (M2; Sigma-Aldrich) antibody as described previously .
SK and CJB performed the experiments and co-wrote the manuscript. GBG and SK performed the bioinformatics analysis. All authors have read and approved the manuscript.
We would like to thank Megan Davey and Stephen Hoke for critically reading the manuscript, Julie Genereaux for technical assistance, Ivan Sadowski for sharing sequencing lanes, and Joe Mymryk for plasmids. This work was supported by CIHR grant MOP10845 to CJB. SK was supported by a National Science and Engineering Research Council of Canada Studentship.