Analysis of Protein Phosphatase-1 and Aurora Protein Kinase Suppressors Reveals New Aspects of Regulatory Protein Function in Saccharomyces cerevisiae

Protein phosphatase-1 (PP1) controls many processes in eukaryotic cells. Modulation of mitosis by reversing phosphorylation of proteins phosphorylated by aurora protein kinase is a critical function for PP1. Overexpression of the sole PP1, Glc7, in budding yeast, Saccharomyces cerevisiae, is lethal. This work shows that lethality requires the function of Glc7 regulatory proteins Sds22, Reg2, and phosphorylated Glc8. This finding shows that Glc7 overexpression induced cell death requires a specific subset of the many Glc7-interacting proteins and therefore is likely caused by promiscuous dephosphorylation of a variety of substrates. Additionally, suppression can occur by reducing Glc7 protein levels by high-copy Fpr3 without use of its proline isomerase domain. This divulges a novel function of Fpr3. Most suppressors of GLC7 overexpression also suppress aurora protein kinase, ipl1, temperature-sensitive mutations. However, high-copy mutant SDS22 genes show reciprocal suppression of GLC7 overexpression induced cell death or ipl1 temperature sensitivity. Sds22 binds to many proteins besides Glc7. The N-terminal 25 residues of Sds22 are sufficient to bind, directly or indirectly, to seven proteins studied here including the spindle assembly checkpoint protein, Bub3. These data demonstrate that Sds22 organizes several proteins in addition to Glc7 to perform functions that counteract Ipl1 activity or lead to hyper Glc7 induced cell death. These data also emphasize that Sds22 targets Glc7 to nuclear locations distinct from Ipl1 substrates.

The glycogen-deficient trait used to initially identify glc7 mutations comes from the failure of the Gac1-Glc7 complex activity in cytoplasmic glycogen particles to dephosphorylate glycogen synthase [40,41]. Glc7 further regulates carbon metabolism via association with Reg1 and Reg2 [42,43].
GLC7 is one of several genes that kill yeast cells when they are overexpressed [44]. High-copy GLC7 increases the chromosome gain frequency; a phenotype also shared by ipl1 mutations [23]. Only mutations in SHP1 have previously reported to suppress GLC7 overexpression lethality [45]. A goal of this work was to analyze suppressors of GLC7 overexpression to learn more about the mechanism of lethality, about regulation of Glc7 activity, and function of Glc7 interacting proteins. We discovered that many suppressors of GLC7 overexpression also suppress ipl1. However, we isolated SDS22 mutant genes that could dominantly suppress GLC7, but not ipl1 and vice a versa.

Recessive Suppressors of Glc7 Overexpression
The cause of cell death upon Glc7 overexpression is unknown. Characterization of suppressors of this trait reveals novel aspects of Glc7 function. PP1 enzymes like Glc7 function as holoenzymes containing alternative noncatalytic subunits [3,46,47]. S. cerevisiae possesses several Glc7 noncatalytic subunits and if one or more of them produced a Glc7 holoenzyme responsible for cell death, then deletion of the noncatalytic subunit gene should suppress Glc7 overexpression. Therefore, we tested whether deletions of well characterized noncatalytic genes, GAC1, REG1, GLC8, or REG2 suppressed Glc7 overexpression. Galactose induction of GAL1p-GLC7 overexpressed Glc7 in these assays. Wild-type, gac1, and reg1 cells reduced growth on galactose medium revealing the cell death due to Glc7 overexpression ( Figure 1A). Note that reg1 cells grow poorly on synthetic media; however, sensitivity to Glc7 overexpression was clear from the relative growth on YEP-galactose. In contrast, glc8 and reg2 deletions suppressed GLC7 toxicity ( Figure 1A). Using similar assays, we found that deletions in FIN1, BUB3, and MAD2 failed to suppress Glc7 (data not shown). Therefore, removal of at least two different Glc7 holoenzymes allows cells to tolerate Glc7 overexpression. This suggests that promiscuous dephosphorylation of several Glc7 substrates promoted cell death upon Glc7 overexpression. Furthermore, bypassing SAC function does not promote tolerance to greater Glc7 activity.
We used two schemes to identify dominant suppressors of Glc7 overexpression. The first scheme sought galactose resistant revertants from a GAL1p-GLC7 diploid (Materials and Methods). One revertant, JC907, received the greatest attention because it had a recessive lethal mutation linked to the Glc7 suppressor. Genetic mapping followed by DNA sequencing the SDS22 locus revealed that JC907 contained an SDS22 amber nonsense mutation in the Ser56 codon, SDS22-S56am, in a heterozygous state (Materials and Methods). The recessive lethal trait of SDS22-S56am stems from Sds22 being essential for yeast viability [14]. Our finding the heterozygous SDS22/SDS22-S56am genotype as a suppressor of Glc7 overexpression implicates the Glc7-Sds22 holoenzyme in the Glc7-induiced cell death.
We were curious whether the heterozygous SDS22-S56am allele was special in its suppression of GLC7 or if any SDS22 null allele could suppress. The SDS22-S56am mutation truncates the encoded Sds22 protein such that only residues preceding the leucine-rich repeats, which bind Glc7, would be expressed. We considered the possibility that this truncated Sds22 protein functioned as a dominant-negative. However, results consistent with that contention were not reproducible. Instead we favor the explanation that SDS22-S56am merely functions as a suppressor of Glc7 overexpression because it reduces the concentration of Sds22 in a diploid. The observation that complete heterozygous deletion, sds22D/+, suppressed Glc7 as well as smaller deletions corroborates this conclusion ( Figure 1C). Therefore, the Sds22-Glc7 holoenzyme must also be participate in lethal dephosphorylations that occur upon Glc7 overexpression.

Fpr3 Dominantly Suppresses Glc7 by Reducing Glc7 Protein
We screened libraries of wild-type genes in high-copy, 2 m vectors for genes that suppress Glc7 overexpression to identify additional dominant suppressors of Glc7 (Materials and Methods). Extensive screening identified the FPR3 encoded proline isomerase as the sole dominant, high-copy Glc7 suppressor ( Figure 2A). The S. cerevisiae genome encodes twelve proline isomerases and that encoded by FPR4 is most similar to FPR3 [48]. FPR4 also suppresses Glc7 slightly ( Figure 2A).
We wanted to know how Fpr3 suppressed GLC7 overexpression. Comparison of Fpr3 and Fpr4 amino acid sequences showed two shared domains in addition to the C-terminal proline isomerase homologous domain ( Figure 2B). GLC7 suppression by Fpr3 needs the central Fpr4 homologous region because deletion compromised suppression ( Figure 2C). Surprisingly, deletion of the proline isomerase domain via the ochre nonsense mutation, K302oc, and many other mutations including deletion of residues 294 to 411 did not compromise GLC7 suppression. Therefore, Fpr3 suppresses Glc7 by using a function other than its proline isomerase activity.
Glc7 suppressors could reduce the levels of bulk Glc7 protein levels in the cell. Indeed, Glc7 protein levels declined with increased expression of Fpr3 ( Figure 3). In contrast, glc8 and pcl6 pcl7 mutants that lack phospho-Glc8 or shp1 a Glc7 suppressor that modulates Glc7 nuclear import [12], display no change in bulk Glc7 protein levels. Hence, Fpr3 exploits a unique mechanism of GLC7 suppression; it reduces Glc7 protein levels.

Sds22 and Fpr3 Suppress ipl1
Because Glc7 dephosphorylates kinetochore proteins phosphorylated by Ipl1, several mutations that reduce Glc7 function suppress temperature-sensitive ipl1 mutations [24,38]. Therefore, we tested ipl1 suppression by Fpr3 and Sds22. Wild-type and mutant FPR3 genes suppressed ipl1 identically to their GLC7 suppression; on the contrary, high-copy FPR4 did not suppress ( Figure 4A). This is consistent with the weaker GLC7 suppression by FPR4 compared to FPR3. Notably, ipl1 suppression also did not require the proline isomerase domain of Fpr3. These results are consistent with ipl1 suppression due to reduced Glc7 function.
So far, all suppressors of ipl1 also suppress Glc7 overexpression and vice a versa. High-copy SDS22 suppression of ipl1 was previously reported [24]; however, potential suppression by reducing SDS22 gene dosage was not. To test this possibility, isogenic homozygous ipl1 diploids were constructed. We found homozygous ipl1 diploids to be more temperature-sensitive than ipl1 haploids; however, high-copy SDS22 suppression was evident at 37u and heterozygous sds22D/SDS22 failed to suppress ipl1 ( Figure 4B). Therefore, increases, but not decreases of SDS22 gene dosage suppress ipl1. Suppression of Glc7 overexpression by changing SDS22 gene dosage was completely reciprocal to that of ipl1 suppression. These initial results show that SDS22 uniquely distinguishes suppression of Glc7 overexpression from ipl1 suppression.
We sought SDS22 mutations, which fail to suppress ipl1 in order to delineate important features of Sds22 required for ipl1 suppression. To this end, we isolated two random SDS22 mutant alleles, RM45 and RM107, which compromised ipl1 suppression ( Figure 5A and D), yet encoded proteins expressed at levels comparable to wild-type ( Figure 5E). SDS22-RM45 contains mutations E79G, L159Q, and L295I and SDS22-RM107 mutations F65L, Y141H, and I210T ( Figure 5A). High-copy SDS22-RM45 could not suppress ipl1 even if wild-type Sds22 levels were reduced ( Figure 4B). Thus, SDS22 suppression of ipl1 can be compromised by multiple missense mutations. These findings make it unlikely that high-copy SDS22 suppresses ipl1 strictly by promoter titration because no other mutations were present in these mutant SDS22 genes that failed to suppress.

Sds22 that Fails to Bind Glc7 Suppresses Glc7 Overexpression
Testing GLC7 suppression by high-copy wild-type and mutant SDS22 genes revealed intriguing aspects of Sds22. Wild-type and most mutant SDS22 genes did not suppress Glc7 overexpression ( Figure 6A). The exceptional alleles, SDS22-RM45 and to a lesser degree SDS-D81-171, did suppress Glc7. Suppression by SDS22-RM45 required it to be high-copy; expression from a single-copy vector did not suppress ( Figure 6B). The failure of SDS22-RM45 to complement sds22D ( Figure 5A) and its dominant GLC7 suppression show that it is a dominant negative allele. Such alleles most frequently act via competition with the wild-type protein [50]. Elevation of wild-type SDS22 consistently diminished GLC7 suppression by SDS22-RM45 consistent with a dominant negative activity of Sds22-RM45 ( Figure 6C).
High-throughput studies reported many other proteins bind to Sds22 besides Glc7 [31,51]. We confirmed a subset of these interactions by purifying potential Sds22 binding proteins from yeast as GST fusion proteins and testing if HA3-Sds22 copurified. This methodology verified that Kog1, Nop6, Rvb1, Rvb2, Snf4, and Ygr130C bound to Sds22 ( Figure 7A). The variable yields of HA3-Sds22 copurified were attributable to differences in GST fusion expression and degradation. We intended to use Bub3 as a negative control, but fortuitously discovered that it also bound to Sds22 in this assay. We successfully used histone acetyltransferase subunit, Ahc1, as negative control instead. Binding of Sds22-RM45 to this collection of proteins was indistinguishable from wild-type Sds22 ( Figure 7A).
Two-hybrid assays evaluated Glc7 interaction with mutant Sds22 proteins. As expected from of the lack of leucine-rich repeats, Sds22-S56am did not interact with Glc7 ( Figure 7B). Of the other mutant Sds22 proteins tested, only the RM45 mutant showed no apparent Glc7 affinity. Surprisingly, even Sds22 mutant proteins lacking two or more leucine-rich repeats (D81-124, D81-171, and D243-323) retained detectable Glc7 affinity by this assay. Failure of Sds22-RM45 to bind Glc7 can explain its dominant negative function if this mutant protein diminished the other Sds22-binding proteins from binding to the wild-type Sds22-Glc7 complex.
Dominant ipl1 suppression by high-copy SDS22-S56am ( Figure 4C) suggested that truncated Sds22(1-55) protein bound to proteins other than Glc7 to mediate suppression. High-copy HA3-SDS22-S56am also suppressed ipl1 although it was weaker than SDS22-S56am (data not shown). This result is consistent with the Sds22 N-terminus playing an important role in ipl1 suppression. We sought to analyze Sds22 binding proteins for affinity to Sds22(1-55); however, we were unable to detect HA3-Sds22-  S56am protein from yeast extracts. Instead we used a Sds22(1-25)-LacZ fusion, which showed Sds22 nuclear localization previously [13]. None of the GST fusion proteins tested bound to bgalactosidase ( Figure 8D). In contrast, five of seven GST fusions bound to Sds22(1-25)-LacZ greater than the negative control ( Figure 8C). These results support the binding of Bub3, Kog1, Rvb1, Rvb2, and Snf4 to residues 1-25 of Sds22. Therefore, Sds22 residues 1-25, not only supplies a nuclear localization signal for Sds22, but it promotes binding to at least five other proteins. Note, that these results do not imply simultaneous or direct Sds22 interaction to all or any of these proteins.

Discussion
The molecular details of cell death caused by overexpression of GLC7 remain elusive. However, our discovery that Sds22, Shp1, Reg2, and phospho-Glc8 must all be functional for this cell death strongly implicates the importance of nuclear-localized substrates because Sds22 and Shp1 promote Glc7 nuclear localization [11,12,14]. The Reg2 role in glucose transcription repression and maltose permease proteolysis [42,43] suggests functions of the Reg2-Glc7 holoenzyme in and outside the nucleus. The most parsimonious conclusion at this point is that substrates must be promiscuously dephosphorylated by both Reg2-Glc7 and Sds22-Glc7 holoenzymes to kill yeast cells. Some and perhaps all of these critical substrates reside in the nucleus. Our results illustrate that Reg1 and Reg2 have distinct functions because reg1 fails to suppress Glc7 overexpression, whereas reg2 suppresses ( Figure 1A).
GLC7 suppression could result from displacement of Glc7 from critical substrates, reducing phosphatase activity, or diminishing Glc7 protein levels. We found examples for each of these mechanisms. Phospho-Glc8 increases global Glc7 protein activity via a proposed chaperone function [37,52]; therefore, glc8 mutants have less activity in many if not all Glc7 holoenzymes. Absence of phospho-Glc8 suppresses Glc7 ( Figure 1B). Viable glc7 missense mutations were previously isolated based upon their glycogen-deficient or sporulation-deficient traits [52]. They compromise binding to subsets of Glc7 regulatory subunits and some have enzymatic activity reductions [37,52]. All ten mutant GLC7 genes we tested failed to kill when overexpressed (data not shown). A GLC7 allele with the intron deleted retained the ability to kill when overexpressed. This finding illustrates that death from Glc7 overexpression is not a result of taxing the cellular mRNA splicing machinery. Together, these findings show that it is the fully active Glc7 enzyme that kills when it is overexpressed.
High-copy FPR3 or FPR4 suppression of GLC7 overexpression could be via their proline isomerase activity or by their nucleolar partitioning. Proline isomerization is particularly attractive because there are several conserved prolines found in all PP1 enzymes, proline isomerization is a rate limiting step in protein folding, and PP1 enzymes are notoriously difficult to fold in heterologous systems [47,53,54]. Furthermore, Fpr3 proline isomerase activity inhibits Glc7 to modulate meiotic progression [6]. However, it is clear the proline isomerase domain (PPI) of Fpr3 does not suppress mitotic lethality of Glc7 overexpression because its deletion actually enhanced suppression of GLC7 by Fpr3 ( Figure 2C). In addition to the FPR3-K302oc nonsense mutation, FPR3-V303am, and FPR3(D294-411), which truncate the PPI domain also failed to suppress (data not shown). Fpr3 overproduction did not result in a detectable change in Glc7 subcellular localization (Kelly Tatchell, personal communication). Fpr3 exploits a novel mechanism compared to other GLC7 suppressors; it reduced the total Glc7 protein levels ( Figure 3). We propose the chaperone activity of Fpr3 is exploited like that of some other proline isomerases to catalyze Glc7 degradation [55][56][57].
High-copy FPR3 or FPR4 also suppress ubiquitin ligase tom1 mutations [58]. Tom1 promotes destruction of Dia2, Cdc6 and other proteins [59,60] and high-copy FPR3 might promote destruction of Tom1 targets similar to Glc7. However, we found that glc7 missense mutations or glc8D suppress tom1 (data not shown), which suggests that high-copy FPR3 and FPR4 suppress tom1 via their inhibition of Glc7. These findings further connect Glc7 and Tom1 function to mRNA nuclear export [61].
Through analysis of GLC7 and ipl1suppressors, we learned additional information about Glc7 interacting proteins. Several Glc7 interacting protein genes in high-copy suppress temperaturesensitive ipl1 mutations [24]. Ipl1 phosphorylates several proteins that Glc7 dephosphorylates. Because of this antagonism, reduction of Glc7 activity on these shared substrates suppresses ipl1 temperature-sensitive mutations. High-copy GLC8 or FPR3 suppress ipl1 by reducing activity of all Glc7 holoenzymes. The ipl1 suppression by many of the genes encoding Glc7 interacting proteins was explained by displacement of Glc7 away from Ipl1 [24]. Such displacement allows Ipl1 substrates to increase phosphorylation. These two mechanisms for ipl1 suppression cannot explain how high-copy SDS22 suppresses. Sds22 is nuclear and increasing Sds22 levels would unlikely displace Glc7 from the nucleus [11]. Furthermore, Sds22 promotes Glc7 function [3].
Suppression of GLC7 and ipl1 by SDS22 reveal nuances of Sds22 function. The observation that halving SDS22 gene dosage in SDS22/sds22D diploids suppresses GLC7, but not ipl1 can be rationalized by the former suppression demanding smaller reductions in Glc7 activity than the later. Second, high-copy SDS22 suppression of ipl1, but not GLC7 suggests that Sds22 targets Glc7 to nuclear locations distinct from Ipl1 substrates. Other studies revealed discrete pools of nuclear PP1 [62]. Third, our attempts to reduce Glc7 binding by missense mutations or LRR deletions demonstrate redundant Glc7 binding by the LRR's. Indeed, five LRRs are sufficient to bind PP1 in a histone variant [63].
Explaining suppression of GLC7 and ipl1 by SDS22 alleles is more complicated. Complexity arises because Sds22 is a scaffold, which binds many other proteins. We confirmed binding to Rvb1, Rvb2, Kog1, Nop6, Snf4, and Ygr130C in this work using different techniques than reported previously [31]. We fortuitously showing the extent of deletions, which removed whole repeats and maintained reading frame. Function of each mutant protein was assessed by complementation of sds22D by high-copy mutant genes. JC1378 (sds22D::HIS3/+) was transformed with 2m URA3 SDS22-X-myc3 plasmids, sporulated and at least 20 tetrads dissected. Complementation was indicated by viable His + spore clones. Only SDS22-D273A could complement. B, C, D) Fivefold serial dilutions of JC1126-15B transformed with high-copy SDS22 genes with the indicated genotypes was incubated on -Ura plates at the indicated temperatures. The SDS22-RM12 gene, which suffered a large deletion, serves as a negative control. E) Anti-Myc antibody probed immunoblot of crude extracts of JC1126-15B transformants with high-copy SDS22-X-myc plasmids or pRS316 (empty). Molecular masses in kDa are shown on left. Pgk1 levels in these extracts are shown at the bottom as control. doi:10.1371/journal.pone.0069133.g005 discovered that Sds22 also binds Bub3 (Figure 7). Therefore, overexpression of Sds22 could displace these proteins from their normal location and thus impair their fuction. For example, attenuation of Tor complex 1 function suppresses ipl1 [64]. Kog1 is a Tor complex 1 component [36] and Kog1 binds to Sds22 residues 1-25 ( Figure 8C), which are sufficient to suppress ipl1 ( Figure 4C). Sds22 residues 1-25 are sufficient to bind five of seven Sds22 binding proteins we examined (Figure 8). These Sds22 binding proteins may not bind to Sds22 directly and they most likely bind to more than one segment of Sds22. In fact, highcopy SDS22D2-56 robustly suppresses ipl1 ( Figure 5C). The dominant-negative SDS22-RM45 suppression favors its titration of Sds22 binding proteins other than Glc7 away from wild-type Sds22-Glc7 holoenzyme to achieve GLC7 suppression ( Figure 6C). The ipl1 suppression by high-copy mutant and wild-type SDS22 probably works similarly; however, with differing Sds22-binding protein specificity. With our current data, we cannot specify which Sds22-binding proteins control GLC7 or ipl1 suppression; nevertheless, suppression of GLC7 and ipl1 is reciprocal ( Table 1). Suppression reciprocity concurs with the separate locations of Ipl1 substrates and Sds22 Glc7 targeting.
Mating, transformation, sporulation, and tetrad analysis were performed by procedures previously described [68]. Rich (yeast extract-peptone [YEP]) or synthetic omission media contained glucose, galactose, or raffinose at 2% (w/v) [68]. For growth comparison assays, the cell concentration in exponentially growing cultures was determined by absorbance at 600 nm. By appropriate dilution in water, equal cell numbers were spotted on plates in serial five-fold dilutions.

Plasmid Construction
Most plasmids used in this work are described in Table 3. High-copy plasmids used the 2m origin of replication and low-copy plasmids contained a centromere (CEN). Several plasmids were constructed by recombination in yeast [69]. SDS22 amplified from yeast DNA by PCR and recombined into YCp50 produced plasmid p2431 and contains an SDS22 gene with a NotI site just before the termination codon. Plasmid pAG101 was made by transferring a SDS22 PvuII-NotI fragment to pRS426. A NotI fragment encoding the triplet myc epitope from pMPY-3xMYC [70] was transferred to pAG101 to yield pAG108. Plasmid p2453 was made by recombination of p705-3 with a yeast genomic PCR SDS22 fragment. An AgeI-NotI fragment from p2453 was transferred to pRS314 to make p2518. Plasmids p2603 contains the SDS22(1-25)-lacZ fusion derived from V76B8 in pRS314. Plasmid pRG200 has a URA3 fragment inserted into the XbaI sites internal to the SDS22 coding sequence resulting in removal of codons after 116. The 4851-bp p2533 plasmid was made from p2431 by deleting a SbfI fragment. Plasmids pKC978, pKC1048 and pYT251 contain the GAL1 promoter from pBM272 [71] driving GLC7 transcription in pRS316, YIp5 or pRS314 respectively. Plasmid p2608 has a 3604-bp EcoRI-XbaI SHP1 DNA fragment in pBluescript II KS(+) with URA3 inserted into a coding region BamHI site. Plasmid p2757 was made by swapping restriction fragments with p2518 and p2752. The 2m LEU2 GAL4AD-SDS22 fusion plasmids used in Figure 7 were made by restriction fragment swapping with p2644, which was derived from pACT2-SDS22 [52]. All 2m URA3 CUP1p-GST fusion plasmids [67] used here were DNA sequenced and compared to the S. cerevisiae S288c sequence.

Isolation of Genomic Dominant GLC7 Overexpression Suppressors
Four spontaneous, independent, galactose-resistant revertants of JC482D/pKC1048 were isolated and analyzed. After sporulation and tetrad dissection, two of the revertants produced no galactosesensitive haploid progeny. These revertants likely suffered mutation at the GAL1p-GLC7 locus and were discarded. A third revertant harbored a suppressor mutation unlinked to GAL1p-GLC7:URA3 because approximately one-half the Ura + spores were galactose-resistant. The extragenic suppressor in that revertant was not analyzed further. The fourth revertant, JC907, produced only two viable spores per tetrad. About one-half the viable progeny were Ura + and all Ura + haploids were galactose-sensitive. These observations indicated a dominant GLC7 suppressor in JC907 had a recessive lethal trait and it was genetically unlinked to GAL1p-GLC7. The integrated pKC1048 (URA3) plasmid was evicted from JC907 by 5-fluoro-orotic acid resistance selection [73] to produce strain JC908-2, which continued to display two viable spores per tetrad.

Mapping the Recessive Lethal Trait in JC908-2
Sporulation and tetrad dissection of strain JC908-2 yields two viable haploid spores per tetrad. TnURA3 transposons were integrated into JC908-2 by transformation with plasmids from the Triples collection [13] after NotI digestion. Transposons were chosen that integrated URA3 at 73 distinct locations spaced approximately 150 Kb throughout the genome and thus at least one was guaranteed to show linkage to any locus. Tetrad analysis showed that sds22::TnURA3 from plasmid V76B8 failed to recombine with the JC908-2 recessive lethal trait; either all spores were inviable (sds22::TnURA3/sds22-S56am) or only two viable Uraspores were found (sds22::TnURA3/+). This result indicated the recessive lethal trait was linked to SDS22.
Genetic and DNA sequence analysis showed that SDS22 loci from diploid JC908-2 retrieved by gapped plasmid repair [74] were either wild-type or contained an amber mutation at SDS22 codon 56 (SDS22-S56am). To confirm the SDS22-S56am mutation in JC907, we isolated JC907 derivatives that contained amber nonsense suppressors. The his4-539 mutation in JC907 is an amber mutation. Several spontaneous His + revertants of JC907 were selected and then analyzed by tetrad analysis. Those with extragenic his4-539 suppressors (i.e. those with putative amber suppressors) suppressed SDS22-S56am and allowed viability and became sensitive to galactose. These results confirmed that SDS22-S56am suppressed Glc7 overexpression lethality in a diploid.
High-copy Suppressors of GLC7 Overexpression JC482D/KC1048 was transformed with wild-type yeast libraries constructed in the 2m, YEp13, vector [75]. Suppressors were selected on -Leu -Ura galactose medium. Plasmids retrieved from transformants that grew on this selective medium were analyzed by restriction mapping, DNA sequence analysis, subcloning, and GLC7 suppression analysis. Only plasmids containing FPR3 were isolated by this scheme.

Immunoblotting Experiments
Crude extracts were prepared from exponentially grown cells by glass bead vortexing in extract buffer (50 mM Tris HCl pH 7.5, 150 mM NaCl, 0.1% (v/v) Triton X-100, 1 mM DTT, 10% (v/v) glycerol, 1X Complete Protease Inhibitor Cocktail (Roche Molecular Biochemicals) and 2 mM PMSF [11]. Protein concentrations were determined by Bradford assays using bovine serum albumin standards (Pierce). SDS-PAGE, blotting and detection by chemiluminescence was as described [37] with anti-HA and anti-Pgk1antibodies (Santa Cruz Biotechnology, Inc and Molecular Probes respectively). Densitometry of films used ImageJ software.

GST Pull-down Experiments
Cells that express galactose and copper inducible proteins were grown in minimal medium with 2% (w/v) raffinose to A 600 = 0.7-1.0 and induced with 2% (w/v) galactose and 0.5 mM CuSO 4 for 2-4 hours. GST fusion proteins were purified as described [67] except that binding and washes used buffer containing 250 mM NaCl. In pull-downs with Sds22(1-25)-LacZ, an additional wash with RIPA buffer (50 mM Tris, pH 8, 150 mM NaCl, 1% (v/v) NP40, 0.5% (w/v) sodium deoxycholate, 0.1% (w/v) SDS) was required to reduce the background binding of the negative control GST-Ahc1. In each experiment, the affinity purification was from equivalent crude extract protein masses (1-2 mg). Immunoblots of crude extracts and GST affinity-purified proteins were probed with anti-GST, anti-HA, or anti-b-galactosidase antibodies (Santa Cruz Biotechnology, Inc and Molecular Probes).

Two-hybrid Analysis
DNA binding domain plasmid, pAS1-GLC7, has been described previously [52]. SDS22 mutations were transferred to p2644 by restriction fragment swapping and b-galactosidase activity of PJ69-4A transformants were assayed in triplicate as described [76]. Statistics of activities were compared with a two-tailed t-test assuming unequal variances. Immunoblots probed with anti-Gal4AD (Sigma) antibody showed equivalent Gal4AD-Sds22 expression for each mutant fusion protein.