The Budding Yeast Cdc48Shp1 Complex Promotes Cell Cycle Progression by Positive Regulation of Protein Phosphatase 1 (Glc7)

The conserved, ubiquitin-selective AAA ATPase Cdc48 regulates numerous cellular processes including protein quality control, DNA repair and the cell cycle. Cdc48 function is tightly controlled by a multitude of cofactors mediating substrate specificity and processing. The UBX domain protein Shp1 is a bona fide substrate-recruiting cofactor of Cdc48 in the budding yeast S. cerevisiae. Even though Shp1 has been proposed to be a positive regulator of Glc7, the catalytic subunit of protein phosphatase 1 in S. cerevisiae, its cellular functions in complex with Cdc48 remain largely unknown. Here we show that deletion of the SHP1 gene results in severe growth defects and a cell cycle delay at the metaphase to anaphase transition caused by reduced Glc7 activity. Using an engineered Cdc48 binding-deficient variant of Shp1, we establish the Cdc48Shp1 complex as a critical regulator of mitotic Glc7 activity. We demonstrate that shp1 mutants possess a perturbed balance of Glc7 phosphatase and Ipl1 (Aurora B) kinase activities and show that hyper-phosphorylation of the kinetochore protein Dam1, a key mitotic substrate of Glc7 and Ipl1, is a critical defect in shp1. We also show for the first time a physical interaction between Glc7 and Shp1 in vivo. Whereas loss of Shp1 does not significantly affect Glc7 protein levels or localization, it causes reduced binding of the activator protein Glc8 to Glc7. Our data suggest that the Cdc48Shp1 complex controls Glc7 activity by regulating its interaction with Glc8 and possibly further regulatory subunits.


Introduction
The evolutionarily conserved and highly abundant eukaryotic AAA (ATPase associated with various cellular activities) ATPase Cdc48 (also known as TER94 in Drosophila and as p97 and VCP in mammals) has emerged as an important motor and regulator for the turnover of ubiquitylated proteins [1][2][3][4]. It converts chemical energy released by ATP hydrolysis into mechanical force in order to drive the segregation of ubiquitylated substrate proteins from stable protein complexes, membranes, and chromatin [1,[3][4][5]. Cdc48 plays central roles in the proteasomal degradation of protein quality control targets, cell cycle regulators, and transcription factors [3,4,6]. Recently, Cdc48 has also been implicated in the lysosomal degradation of proteins delivered via autophagic and endosomal pathways [3,4,7]. Thus, Cdc48 is involved in the three major routes of regulated intracellular proteolysis in eukaryotes. In addition, Cdc48 has been shown to function in non-proteolytic processes in the fusion of homotypic membrane vesicles of the Endoplasmic Reticulum, Golgi apparatus, and the nuclear envelope [8,9].
The involvement of Cdc48 in such diverse cellular processes requires tight control of its activity. Indeed, a large number of cofactor proteins regulate central aspects of Cdc48 function, including its subcellular localization and substrate specificity [2,3,10]. The mutual exclusive binding of two major cofactors, the heterodimer Ufd1-Npl4 and Shp1 (also known as p47 in vertebrates), defines two distinct Cdc48 complexes, Cdc48 Ufd1-Npl4 and Cdc48 Shp1 , which are specialized in proteasomal and nonproteasomal pathways, respectively [10][11][12]. Cofactor binding to Cdc48 appears to be hierarchical, as additional cofactors bind to the Cdc48 Ufd1-Npl4 and Cdc48 Shp1 complexes in order to further fine-tune their cellular function [10,13]. Cofactors interact with Cdc48 by virtue of one or more Cdc48 binding modules, among them the ubiquitin-like UBX domain [10,[14][15][16] and the linear binding site 1 (BS1) motif (also known as SHP box) [17][18][19]. UBX domain containing proteins constitute the largest family of Cdc48 cofactors [10]. In the budding yeast Saccharomyces cerevisiae, seven UBX proteins were identified and shown to bind Cdc48 [20,21]: Shp1 itself (also known as Ubx1) and Ubx2 through Ubx7. In addition to their carboxyl-terminal UBX domain, Shp1, Ubx2 and Ubx5 possess an amino-terminal UBA domain mediating the binding of ubiquitin and ubiquitylated substrates [20,[22][23][24], and thus exhibit the prototypical architecture of substrate-recruiting adaptors for Cdc48 [20,25,26]. So far, cellular functions were identified for only few of the yeast UBX proteins and include roles in ER-associated protein degradation [22,27,28], lipid droplet homeostasis [29], and UV-induced turnover of RNA polymerase II [24]. By contrast, the role of Shp1 is still poorly understood. Shp1 has been implicated in the proteasomal degradation of a Cdc48 model substrate [20], but the physiological relevance of this finding remains unclear. More recently, Shp1 has been shown to bind the autophagy factor Atg8 and to be involved in autophagosome biogenesis [30]. However, the severe phenotypes of shp1 mutants suggest that Shp1 has additional, more critical cellular functions [20,31].
The SHP1 gene was first identified in a genetic screen for suppressors of the otherwise lethal over-expression of GLC7, the sole catalytic subunit of protein phosphatase 1 (PP1) in yeast [32]. Two shp1 (suppressor of high-copy PP1) alleles tolerated the overexpression of GLC7 and, in turn, exhibited phenotypes reminiscent of glc7 loss-of-function mutants. shp1 null mutants are inviable in the W303 strain background [31] and have reduced PP1 activity in other backgrounds [32,33], consistent with the model that Shp1 is a positive regulator required for normal Glc7 activity [32][33][34]. However, the mechanism by which Shp1 influences Glc7 activity is unknown. It has been proposed that Shp1 positively affects Glc7 activity by a yet undefined indirect mechanism [32][33][34] or by controlling the nuclear localization of Glc7 [31].
Glc7 regulates numerous cellular processes including glycogen metabolism, glucose repression, RNA processing, meiosis and sporulation, DNA damage recovery, actin organization, cell wall morphogenesis, and mitosis (reviewed in [34,35]). A mitotic function of PP1 was first discovered in the fission yeast S. pombe [36,37] and subsequently also shown to exist in higher eukaryotes such as Drosophila and mammals [38,39]. In S. cerevisiae, PP1 is crucial for proper chromosome segregation and, consequently, several different glc7 mutants have been shown to arrest at or before anaphase onset [40][41][42].
Accurate distribution of the replicated genome during cell division is essential for viability and depends on proper chromosome segregation. During mitosis, two physically connected sister chromatids must be faithfully segregated to mother and daughter cell, an event controlled by the spindle assembly checkpoint (SAC) [43,44]. In order for the yeast metaphase to anaphase transition to occur, each kinetochore must attach to a single microtubule of the mitotic spindle [43][44][45]. The SAC prevents anaphase onset by keeping the APC/C Cdc20 ubiquitin ligase complex inactive. Once proper bi-polar attachment is achieved, active APC/C Cdc20 ubiquitylates the mitotic substrate Pds1 (securin), which in turn is rapidly degraded by the 26S proteasome resulting in cleavage of cohesin and sister-chromatid separation [43,44,[46][47][48].
Shp1 has previously been implicated in the regulation of several cytosolic functions of Glc7 [32,33,[60][61][62]. In this study, we identify the Cdc48 Shp1 complex as a critical positive regulator of Glc7 activity towards mitotic Ipl1 substrates including Dam1. We show that shp1 mutants exhibit a SAC-mediated cell cycle delay resulting from reduced Glc7 activity, which in turn is caused by the lack of a specific Cdc48 Shp1 function. Moreover, we provide evidence that Cdc48 Shp1 regulates Glc7 activity by controlling its interaction with regulatory subunits rather than affecting Glc7 protein levels or localization.

Results
shp1 mutants are impaired in growth and mitotic progression In order to study cellular functions of Shp1, we generated shp1 null mutants completely lacking Shp1 in the DF5 strain background by mutating the start codon (shp1-7) or by deleting the entire coding region (Dshp1). In contrast to the W303 background, where SHP1 is essential (data not shown; [31]), DF5 shp1 null mutants are viable but grow slowly and are hypersensitive against multiple stress conditions, including high and low temperature and various cell-damaging agents (Fig. 1a) [20]. In addition to these growth phenotypes, shp1 null cells also exhibit an altered cell cycle distribution (Fig. 1b). Analysis of the DNA content of asynchronous cultures by flow cytometry showed a significant accumulation of cells with 2n DNA content in Dshp1 compared to the wild-type. At 25uC, the proportion of wild-type cells in the G1/S and G2/M cell cycle phases was nearly equal (4363% and 5663%, respectively), whereas Dshp1 cultures contained only 2766% cells in G1/S, but 7366% in G2/M. At the non-permissive temperature of 14uC, Dshp1 cultures contained a majority of cells in G2/M, hardly any cells in G1/S, and a significant amount of sub-G1 material potentially indicating the presence of inviable cell remnants (Fig. 1b).
The cell cycle delay was further analyzed by releasing wild-type and Dshp1 cultures at 25uC from a G1 cell cycle arrest induced by the mating pheromone a-factor (Fig. 1cd). The FACS profiles of samples taken at various time points after release show that wildtype and Dshp1 cells both entered G2/M approximately 60-80 min after release (Fig. 1c). However, whereas wild-type cells initiated G1 of the following cell cycle after about 120 min, the number of Dshp1 cells in G1 started to increase only after 160 to 180 min. This G2/M delay of Dshp1 was confirmed by the analysis of cyclin levels by Western blot (Fig. 1d). As judged by the degradation of the G1/S cyclin Cln2 and the onset of expression of the mitotic cyclin Clb2, wild-type and Dshp1 strains both entered G2/M 60-80 min after release. Wild-type cells initiated the next cell cycle at about 120 min, as indicated by the decrease in Clb2 and increase in Cln2 levels. By contrast, the majority of Dshp1 cells remained in G2/M with high Clb2 levels and undetectable Cln2 levels until 160 min after release. Note that the increased Clb2 levels observed in a-factor arrested Dshp1 cells are not caused by defective mitotic exit resulting in G1 entry with high Clb2 levels (data not shown), but are due to a less efficient G1 arrest observed in shp1 mutants (see Fig. 1c). In summary, our data show that Shp1 is required for normal mitotic progression.

Shp1 functions in growth and mitotic progression require Cdc48 binding
All known cellular functions of Shp1 and its mammalian homologue p47 are believed to be based on its role as an adaptor of Cdc48/p97 [8,10,20,30,31], suggesting that the mitotic phenotype of shp1 null mutants described above may involve Cdc48 as well. However, Cdc48 is essential, and conditional cdc48 mutants exhibit pleiotropic phenotypes including defects at several stages of the cell cycle [63][64][65][66][67], thus complicating a meaningful interpretation with respect to Shp1-dependent mitotic defects.
To overcome the limitations of conditional cdc48 alleles, we engineered shp1 alleles encoding Shp1 variants specifically impaired in Cdc48 binding. To this end, sets of amino acid residues in the UBX domain and the binding site 1 (BS1) motif of Shp1 critical for Cdc48 binding were mutated separately and in combinations (Fig. 2a). In addition, key residues in a potential second BS1 motif preceding the SEP domain (Kay Hofmann and A.B., unpublished) were also mutated. Finally, deletion variants lacking the entire UBX and UBA domain, respectively, were constructed. All shp1 alleles were introduced into DF5 by chromosomal integration in single copy under control of the SHP1 promoter. As expected from previous reports [18,68], deletion of the UBA domain did not interfere with Cdc48 binding at all, and deletion of the entire UBX domain or separate mutation of UBX domain or BS1 residues resulted only in partial loss of Cdc48 binding in immunoprecipitation experiments (Fig. 2b). In contrast, the simultaneous mutation of key residues in the UBX domain and in one or both BS1 motifs in the shp1-a1 and shp1-b1 alleles led to a complete loss of Cdc48 binding. Phenotypic analysis showed that both alleles confer temperature sensitivity, indicating that this shp1 phenotype involves Cdc48 binding (Fig. 2c). Next, we analyzed the shp1-a1 and shp1-b1 mutants for potential mitotic defects. Intriguingly, like the shp1 null mutants, the FACS profiles of the Cdc48 binding-deficient mutants revealed an accumulation of cells in G2/M (Fig. 2d), and a delayed mitotic progression was observed with elevated Clb2 levels until 180-200 min after release from G1 arrest (Fig. 2e). These results demonstrate for the first time that the mitotic defects of shp1 mutants are due to the lack of a specific, Shp1-mediated function(s) of Cdc48 during cell cycle progression.

The mitotic delay of shp1 mutants involves SAC activation
The metaphase to anaphase transition is controlled by the spindle assembly checkpoint (SAC) through inhibition of the APC/C Cdc20 ubiquitin ligase complex until chromosome biorientation is achieved [43,44,69]. In order to test if the early mitotic delay of shp1 mutants is the result of SAC activation, we analyzed the stability of Pds1 (budding yeast securin) in wild-type and shp1-7 cultures released from G1 arrest (Fig. 3a). Pds1 was expressed approximately 40 min after the release both in wild-type and shp1-7. However, whereas Pds1 was completely degraded 100 min after release in wild-type, it was significantly stabilized and detectable until the end of the time course in shp1-7. These results indicate a prolonged SAC activation in shp1 and pinpoint the mitotic delay of shp1 to the metaphase to anaphase transition.
Mutants with spindle or kinetochore defects are hypersensitive to microtubule depolymerizing agents [69][70][71][72] and often depend on the presence of an intact SAC for viability [73][74][75]. Consistent with the observed SAC activation, shp1-7 was indeed found to be hypersensitive towards benomyl (Fig. 3b). Furthermore, we detected a strong negative genetic interaction approaching synthetic lethality between shp1-7 and a deletion mutant of a central SAC component, Dmad2 (Fig. 3c). Of note, surviving shp1-7 Dmad2 cells displayed a more even G1/S versus G2/M distribution than the shp1-7 single mutant (Fig. 3d), further supporting the notion that the mitotic delay of shp1-7 is caused by SAC activation.
The mitotic phenotype of shp1 mutants is caused by reduced Glc7 activity shp1 mutants were originally identified based on their ability to tolerate elevated Glc7 levels [32], and Shp1 has been proposed to be a positive regulator of Glc7 [32][33][34][35]. To test if the mitotic phenotype of shp1 mutants is related to Glc7 function(s), we  analyzed genetic interactions between shp1 and the conditional glc7 allele, glc7-129, which at the non-permissive temperature confers a cell cycle arrest at the metaphase-anaphase transition [76]. Furthermore, the mitotic arrest of glc7-129 was reported to depend on the SAC [49]. Intriguingly, the shp1-7 glc7-129 double mutant was inviable at all temperatures tested (Fig. S1a and data not shown), indicating overlapping cellular functions of Shp1 and Glc7. As expected, the synthetic lethality of shp1-7 glc7-129 could be suppressed by a centromeric plasmid encoding wild-type SHP1 (Fig. 4a). Importantly, when we tested the Cdc48 binding-deficient alleles shp1 DUBX , shp1-a1, and shp1-b1, their ability to suppress the lethality of shp1-7 glc7-129 correlated with the ability of the respective gene products to bind Cdc48, demonstrating that an intact Cdc48 Shp1 complex is required for the viability of glc7-129.
To confirm that Shp1 is involved in mitotic functions of Glc7, we next tested genetic interactions between SHP1 and the major nuclear Glc7 regulatory subunit, SDS22 (Fig. 4b). Indeed, we observed synthetic lethality of the shp1-7 sds22-6 double mutant as well, strongly suggesting that Shp1 is critical for a mitotic function(s) of Glc7. Finally, we analyzed genetic interactions between SHP1 and IPL1, the gene encoding the yeast Aurora B kinase homologue. Ipl1 has been described to antagonize mitotic functions of Glc7 at the kinetochore, and the correct balance of Ipl1 kinase and Glc7 phosphatase activities is crucial for unperturbed mitotic progression [50,52,55,59,77]. Importantly, we observed a clear mutual suppression of the shp1-7 and ipl1-321 temperature sensitivities. The shp1-7 ipl1-321 double mutant grew better at 33uC and 37uC than either single mutant (Fig. S1b), suggesting that reduced Ipl1 activity partially alleviates the defects caused by reduced Glc7 activity in shp1-7. However, this positive genetic interaction between shp1 and ipl1-321 was not observed for Shp1 variants proficient in Cdc48 binding (Fig. 4c), again confirming that Shp1 regulates mitotic Glc7 activity in its capacity as a Cdc48 cofactor. Together, our genetic analysis strongly suggests that shp1 mutants possess limiting mitotic Glc7 activity leading to unbalanced Ipl1 activity.
To directly address if nuclear substrates of Glc7 are hyperphosphorylated in shp1, the phosphorylation state of histone H3 in wild-type and shp1, glc7, and ipl1 mutants was analyzed at the nonpermissive temperature of 35uC (Fig. 4d). Residue S10 of histone H3 is phosphorylated by Ipl1 during mitosis, and the phosphorylation level is governed by the balance of Ipl1 kinase and Glc7 phosphatase activities [53]. As expected, the glc7-129 and ipl1-321 mutants exhibited increased and decreased phosphorylation of histone H3, respectively. In shp1-7 cells, the phosphorylated form was strikingly increased, directly demonstrating that Glc7 phosphatase activity is impaired in shp1-7. Importantly, the accumulation of phosphorylated histone H3 in shp1-7 was efficiently suppressed in the ipl1-321 shp1-7 double mutant, resulting in a normal ratio of phosphorylated and total histone H3. Thus, Shp1 indeed controls the balance of Ipl1 and Glc7 activities towards their nuclear target, histone H3.
We next tested if the supposed limiting Glc7 activity in shp1 mutants can be overcome by raising cellular Glc7 levels. To this end, wild-type GLC7 was expressed under the control of the strong, inducible MET25 promoter (Fig. 5a). As expected [32], induction of GLC7 over-expression on methionine-free medium was toxic for wild-type, but not shp1-7. Importantly, the Cdc48 bindingdeficient allele shp1-a1 also tolerated GLC7 over-expression (Fig. 5a), again indicating that regulation of Glc7 activity by Shp1 is Cdc48-dependent. FACS analysis of wild-type cells confirmed that GLC7 over-expression was highly toxic (Fig. 5b, top row). Remarkably, however, GLC7 over-expression was not only tolerated by the shp1-7 and shp1-a1 mutants, but in fact and shp1-a1 cultures synchronized by a-factor arrest/release were analyzed by Western blot against the mitotic cyclin Clb2 (e) as described in the legend to Fig. 1d. doi:10.1371/journal.pone.0056486.g002 Unbalanced Ipl1 and Glc7 activities give rise to chromosome segregation defects [50,53,59], suggesting that shp1 mutants may be impaired in chromosome segregation as well. Indeed, yeast cells depleted of Shp1 were recently shown to exhibit defective chromosome bi-orientation [31]. Using strains containing a lacO array integrated at the LEU2 locus of chromosome III and expressing GFP LacI and the spindle pole body marker Spc42 Mars , we analyzed chromosome segregation in wild-type and shp1 mutants by live-cell fluorescence microscopy (Fig. 5cd). Compared to wild-type, cultures of shp1-7 and shp1-a1 contained significantly more large budded cells with a short spindle and unseparated chromosomes III, and significantly less cells with a long spindle and two separated chromosomes III (Fig. 5c). This finding is fully consistent with the metaphase to anaphase delay described above. Of note, shp1-7 and shp1-a1 also showed a significant increase in cells with chromosome segregation defects (15-20% of largebudded cells in comparison to 3% in the wild-type), as well as some aberrant spindles, confirming that Shp1 is required for faithful chromosome segregation. Importantly, and in line with the FACS data shown in Fig. 5b, over-expression of GLC7 in the shp1 mutants suppressed both the metaphase to anaphase delay and the chromosome segregation defects.
Taken together, these results demonstrate for the first time that nuclear Glc7 activity is reduced in shp1 and that the mitotic phenotype of shp1 results from limiting Glc7 activity.

Dam1 hyper-phosphorylation causes growth defects of shp1 mutants
The phosphorylation state of the kinetochore protein Dam1 is critical for proper microtubule attachments during mitosis [55,[78][79][80]. Since Dam1 has been identified as a common substrate of Ipl1 kinase and Glc7 phosphatase activities [54][55][56]81,82], we next analyzed the phosphorylation state of Dam1 in shp1 mutants. To this end, shp1, glc7 and ipl1 mutants were shifted to 35uC, and phosphorylation of Dam1 was analyzed by Western blot (Fig. 6a). Compared to wild-type cells, Dam1 was indeed hyper-phosphorylated in shp1-7, as judged by the reduction of the faster migrating non-phosphorylated form and the relative increase of the slower migrating phosphorylated form of Dam1. Of note, the increase of Dam1 phosphorylation in shp1 was comparable to that observed in glc7-129 cells. As expected, ipl1-321 cells exhibited strongly reduced Dam1 phosphorylation under these conditions. It has previously been shown that the hypo-phosphorylation of Dam1 in ipl1 mutants can be partially suppressed by a reduction of Glc7 phosphatase activity in glc7 mutants [56]. In line with the reduced mitotic Glc7 activity in shp1, the shp1-7 ipl1-321 double mutant indeed exhibited a roughly equal distribution of phosphorylated and non-phosphorylated Dam1 (Fig. 6a).
To elucidate if the hyper-phosphorylation of Dam1 in shp1 mutants was responsible for the observed growth defects, we made use of previously described phosphorylation site mutants of Dam1 [55]. To this end, we transformed shp1-7 with high copy number plasmids carrying wild-type DAM1 or dam1 mutated in residues S20 and S292, major target sites for Ipl1 [55,82]. Whereas neither the empty vector control nor wild-type DAM1 had an influence on the growth of shp1-7, over-expression of the dam1SA phospho-  Fig. 2b is indicated at the right. (b) Synthetic lethality of shp1-7 and sds22-6. Growth of haploid progeny of one tetrad from the cross of shp1-7 with sds22-6 carrying YC-SHP1 was analyzed on control (YPD) and 59FOA plates as described above. (c) Positive genetic interaction between ipl1-321 and shp1 mutants defective in Cdc48 binding. Growth of haploid progeny of one tetrad from the cross of shp1-7 with ipl1-321 carrying a centromeric plasmid for the expression of the indicated wild-type and mutant SHP1 alleles was analyzed at the indicated temperatures. The ability of the shp1 mutant gene products to bind Cdc48 is indicated at the right. (d) Hyper-phosphorylation of histone H3 in shp1-7. The phosphorylation state of histone H3 in the indicated WT and mutant strains at 35uC was analyzed by Western blot using an antibody recognizing phosphorylated residue Ser10 (pH3) and total H3, respectively. The ratio of the signal intensities (pH3/total H3) is given at the bottom. doi:10.1371/journal.pone.0056486.g004 Figure 5. The mitotic phenotype of shp1 mutants is caused by reduced Glc7 activity. (a) shp1 mutants tolerate over-expression of GLC7. Wild-type (WT) and shp1-7 and shp1-a1 mutant cells expressing GLC7 from an integrative plasmid under the control of the inducible P MET25 promoter (P MET -GLC7) were analyzed for growth at 25uC in the presence (+Met (off)) and absence (2Met (on)) of methionine in the growth medium. The respective strains carrying an empty integrative plasmid (P MET ) served as control. (b) Over-expression of GLC7 suppresses the mitotic delay of shp1 mutants. The strains described in panel (a) were analyzed for cell cycle distribution by FACS in the absence and presence of methionine as indicated. (c) Over-expression of GLC7 suppresses the chromosome segregation defect of shp1 mutants. Sister chromatid separation of wild-type, shp1-7 and mutant incapable of being phosphorylated on residues 20 and 292 enabled shp1-7 cells to grow robustly at 30 and 35uC (Fig. 6b) and, albeit very weakly, at 37uC (data not shown). Conversely, overexpression of the dam1SD mutant mimicking constitutive phosphorylation of residues 20 and 292 was detrimental for the growth of shp1-7 at all temperatures (Fig. 6b). These data show for the first time that Dam1 is a critical target of unbalanced Ipl1 and Glc7 activities in shp1 mutants.
We next tested if reduced Dam1 phosphorylation could also ameliorate the more severe growth defect of shp1 in the absence of an intact SAC, i.e. in the shp1-7 Dmad2 double mutant (Fig. 6c). Indeed, over-expression of the dam1SA phospho-mutant was able to partially suppress the growth defect of shp1-7 Dmad2 both at ambient temperature and at the non-permissive temperature of 37uC. These results suggest that a reduction of phosphorylated Ipl1 target sites on Dam1 is sufficient to partially restore productive kinetochore-microtubule attachments and to limit chromosome mis-segregation in shp1-7 to an extent that significantly improves viability.

The nuclear localization of Glc7 is intact in shp1 mutants
In order to investigate potential reasons for the reduced Glc7 activity in shp1 mutants, we analyzed Glc7 protein levels and subcellular localization using epitope-tagged Glc7 variants. Because both over-expression and epitope-tagging of Glc7 can affect viability [32,83,84], we generated strains expressing carboxyl-terminally tagged Glc7 from its authentic chromosomal locus as the sole source of Glc7 activity. Based on their normal growth at 30uC and 37uC, we concluded that cells expressing Glc7 GFP and Glc7 3myc are not defective in critical aspects of Glc7 function in the DF5 strain background (Fig. 7a). In contrast, expression of Glc7 3HA causes a partial-loss-of-function phenotype reflected in temperature-sensitive growth. The functionality of the Glc7 GFP fusion protein was further confirmed by flow cytometry revealing a wild-type cell cycle distribution (Fig. 7b).
Using the functional, epitope-tagged Glc7 3myc protein we were able to show a physical interaction between Glc7 and Shp1 at endogenous expression levels by immunoprecipitation for the first time (Fig. 7c). The interaction was confirmed in a reciprocal experiment, where Glc7 3myc was co-immunoprecipitated with Shp1 3HA (Fig. 7d). This physical interaction between Shp1 and Glc7 could suggest that Shp1 directly controls the half-life or cellular localization of Glc7. Because we could not detect differences between wild-type and shp1 cells in the protein levels of endogenous, untagged or epitope-tagged Glc7 (data not shown; see e.g. input lanes in Fig. 7g), we performed a thorough analysis of Glc7 subcellular localization by confocal spinning disk live-cell microscopy of cells expressing Glc7 GFP (Fig. 7e). Consistent with previous reports [84][85][86][87][88][89], the majority of Glc7 GFP was detected in the nucleus of wild-type cells, with additional diffuse cytosolic staining and a distinct localization at the septum of medium and large budded cells. shp1-7 and shp1-b1 cells showed a very similar distribution of Glc7 GFP with respect to nuclear, cytosolic, and septum localization, and no aberrant localization or aggregation of Glc7 GFP was observed ( Fig. 7e; Figs. S2, S3). Quantification of the intensity of the nuclear versus cytosolic GFP signal revealed a slight decrease of nuclear Glc7 in the shp1 mutants to approximately 80% of the wild-type signal (Fig. 7f). Interestingly, in the course of these experiments, it became evident that the nuclear localization of Glc7 GFP is influenced by the presence of untagged Glc7 in shp1 mutants, but not wild-type cells. The additional expression of GLC7 from a plasmid resulted in a notable reduction of the nuclear Glc7 GFP signal in shp1-7 GLC7 GFP (Fig. S3). This effect appears to correlate with the expression level of untagged Glc7, as a highcopy plasmid encoding GLC7 under control of the strong ADH1 promoter reduced the nuclear Glc7 GFP signal even further (Fig.  S3). It should be stressed, however, that lack of Shp1 clearly did not abolish nuclear localization of Glc7 GFP if this is the sole source of Glc7 (Fig. 7ef).
The nuclear localization of Glc7 was further assessed by analyzing the interaction of endogenous, untagged Glc7 with its nuclear targeting factor Sds22 [84,86,90]. Co-immunoprecipitation of Glc7 with Sds22 from lysates of wild-type and shp1-7 strains demonstrated that the interaction between Glc7 and Sds22 was not reduced in shp1-7 (Fig. 7g). Because the majority of Sds22 interacts with Glc7 in the nucleus [90], these data strongly suggest a normal nuclear localization of the Sds22-Glc7 complex in shp1-7. Taken together, our results make the possibility that the nuclear localization of Glc7 is grossly affected in shp1 null mutants highly unlikely.

The interaction between Glc7 and Glc8 is impaired in shp1 mutants
As neither protein levels nor cellular localization of Glc7 were severely affected in shp1 mutants, we considered the possibility that Shp1 may influence the interaction of Glc7 with one or more of its numerous regulatory subunits. Because the physical interaction of Glc7 with its mitotic regulator Sds22 was unaltered in shp1 (Fig. 7g), and no genetic interaction between SHP1 and the putative Glc7 kinetochore recruitment factor FIN1 [91] could be detected (data not shown), we next analyzed genetic interactions between shp1 and glc8 mutants (Fig. 8a). Glc8 is a positive regulator of Glc7, whose Glc7 activating function depends on phosphorylation of residue T118 by the cyclin-dependent kinase Pho85 [34,92,93]. Deletion of GLC8 has been shown to cause a reduction of Glc7 phosphatase activity and cellular glycogen levels [94], but does not result in significant growth or cell cycle defects (Fig. 8a and data not shown). Some glc7 alleles are synthetic lethal with Dglc8 and glc8 mutants that cannot be phosphorylated on residue T118 [93], indicating that activation by Glc8 becomes essential under conditions of reduced Glc7 activity. Interestingly, the shp1-7 Dglc8 double mutant was found to be inviable as well (Fig. 8a), demonstrating that either Glc8 or Shp1 is required for viability of the DF5 strain background, presumably to ensure sufficient Glc7 activity. Consistent with this hypothesis, expression of GLC8 restored growth of shp1-7 Dglc8, whereas the glc8-T118A allele was shp1-a1 mutant cells expressing GLC7 under the control of the inducible P MET25 promoter was analyzed at 25uC in the presence (P MET -GLC7 off) and absence (P MET -GLC7 on) of methionine in the growth medium. Large-budded cells (n.300 for each condition) were sorted into four classes based on the relative orientation of the GFP LacI-marked chromosomes III and the spindle pole body (SPB) marker Spc42 Mars : I, normal metaphase spindle; II, normal anaphase spindle; III, meta-/anaphase spindle with segregation defect; IV, aberrant number of SPBs. Error bars indicate binomial standard errors. The distribution of the five cell types over the four classes is non-random with high statistic significance according to a Pearson's chi-squared test of independence (X 2 (12) = 123.931; p.0.001). All pairwise differences within classes I-III between (i) wild-type and shp1 mutants without overexpression of GLC7, and (ii) shp1 mutants with and without GLC7 over-expression are statistically significant with p,0.01 according to Fisher's exact test.

(d) Representative examples of large-budded cells falling into the four classes analyzed in panel (c)
. Upper row, fluorescence microscopy of GFP LacI-marked chromosomes III (chr. III) and Spc42 Mars -marked SPBs; lower row, differential interference contrast (DIC) microscopy. The asterisks mark an additional unbudded cell in class I that was not included in the analysis. doi:10.1371/journal.pone.0056486.g005 unable to rescue the synthetic lethality (Fig. 8b), proving that growth of shp1-7 indeed requires the activation of Glc7 by Glc8. Similar to the genetic interaction between shp1 and glc7 (Fig. 4a), the ability of Shp1 to bind Cdc48 is required for viability of strains lacking GLC8 (Fig. 8b), again demonstrating that the Cdc48 Shp1 complex is necessary for normal regulation of Glc7 activity.
Glc8 has been demonstrated to physically interact with Glc7 in vivo and in vitro [92]. We therefore analyzed the effect of a shp1 null mutation on the Glc7-Glc8 interaction by co-immunoprecipitation, using lysates from asynchronous and a-factor-arrested cultures of wild-type and shp1-7 GLC8 3HA strains. The shp1-7 GLC8 3HA strain was viable and did not exhibit any additional growth defect compared to shp1-7 (data not shown), indicating that Glc8 3HA is functional. Intriguingly, the interaction of endogenous Glc7 with Glc8 3HA was significantly reduced in asynchronously growing as well as in G1-arrested shp1-7 cultures (Fig. 8cd). Quantification of the Glc7 co-immunoprecipitation with Glc8 in asynchronous cultures of wild-type and shp1-7 revealed that the Glc7-Glc8 interaction in shp1-7 was reduced by approximately 50% (Fig. 8e). These data show that Shp1 is required for a normal physical interaction between Glc7 and its activator Glc8.
We next tested if the reduced interaction between Glc7 and Glc8 in shp1-7 is responsible for shp1 phenotypes. To this end, we over-expressed GLC8 3HA under the control of the MET25 promoter (Fig. 8f). Importantly, the over-expression of GLC8 3HA partially suppressed the growth defects of shp1-7, as improved growth at 35uC and weak, but detectable growth at the nonpermissive temperature of 37uC was observed (Fig. 8f). In contrast, over-expression of glc8-T118A 3HA was toxic in shp1-7 cells (Fig. 8f), presumably because excess non-activating Glc8-T118 protein competed with endogenous Glc8 for Glc7 binding. In summary, our data show that the Cdc48 Shp1 complex is important for the activation of Glc7 by Glc8, and that lack of Glc8-mediated activation contributes critically to the phenotype of shp1 mutants.

Discussion
This study addresses the relationship of Shp1, a major Cdc48 cofactor, and Glc7, the catalytic subunit of budding yeast PP1. We found that shp1 mutants exhibit a variety of severe phenotypes, including a significant mitotic delay during progression from metaphase to anaphase. We were able to show that the mitotic phenotype of shp1 mutants is caused by limiting nuclear Glc7 activity towards mitotic substrates, resulting in their hyperphosphorylation due to unbalanced Ipl1 kinase activity. By engineering shp1 alleles specifically defective in Cdc48 binding, we established that Shp1 regulates Glc7 in its capacity as a Cdc48 cofactor. Importantly, we could demonstrate that Shp1 and Glc7 interact physically, and that the Cdc48 Shp1 complex is required for normal interaction of Glc7 with Glc8.
shp1 mutants were originally found to exhibit reduced Glc7 activity towards glycogen phosphorylase, decreased glycogen accumulation, and defective sporulation [32]. Other shp1 phenotypes attributed to reduced Glc7 activity include defective vacuolar degradation of fructose-1,6-bisphosphatase through the vacuole import and degradation (Vid) pathway [60], impaired V-ATPase activity [61], and impaired glucose repression [62]. Here, we provide several lines of evidence that shp1 mutants also possesses a significant defect in mitotic Glc7 activity. First, the genetic interactions between shp1 and glc7, sds22, mad2, and ipl1 all point towards impaired nuclear function(s) of Glc7 in shp1. Second, overexpression of GLC7 in shp1 restored a normal cell cycle distribution and suppressed chromosome segregation defects. Third, the nuclear Glc7 substrates histone H3 and Dam1 are hyperphosphorylated in shp1 in an Ipl1-dependent manner. Together with the previously described cytosolic and vacuolar processes, the elucidation of its involvement in mitotic Glc7 functions under- Dephosphorylation of Ipl1/Aurora B substrates at kinetochores is a well-established and evolutionarily conserved mitotic function of Glc7/PP1 [95][96][97]. Recently, however, Glc7 has also been implicated in silencing of the SAC [98][99][100][101], raising the possibility that impaired SAC silencing rather than chromosome attachment defects causes the mitotic delay of shp1. The observed stabilization of Pds1 in shp1 (Fig. 3a) as well as the suppression of the mitotic delay by over-expression of GLC7 (Fig. 5b) would in fact be consistent with both scenarios. However, the suppression of the chromosome segregation defects indicating defective chromosome bi-orientation by over-expression of GLC7 (Fig. 5c) provides evidence that disturbed kinetochore-microtubule attachments are the primary cause for the mitotic delay in shp1. Furthermore, while glc7 mutants defective in SAC silencing are rescued by deletion of MAD2 [99], the pronounced negative genetic interaction observed for the shp1-7 Dmad2 double mutant also strongly argues against a causative role of impaired SAC silencing for the mitotic phenotype of shp1 (Fig. 3c). It rather shows that SAC inactivation/deletion in the continued presence of mitotic defects is highly detrimental to shp1. According to our data, the key mitotic defect of shp1 mutants is the unbalanced Ipl1 activity at the kinetochore. This conclusion is not only supported by the positive genetic interaction between shp1-7 and ipl1-321, but also underlined by the observed hyperphosphorylation of the Ipl1 targets H3 and Dam1, which is suppressed in the shp1-7 ipl1-321 double mutant.
At the kinetochore, the delicate balance between Ipl1 and Glc7 activities is believed to control cycles of association and dissociation of spindle microtubules that ultimately lead to proper bi-polar attachment and thus satisfaction of the SAC and mitotic progression [57,58,91]. The essential microtubule-binding protein Dam1 has been shown to be a critical target of Ipl1 [54,55,82] and Glc7 [56] during this process. Dam1 is the central component of the heterooligomeric Dam1/DASH complex located at the plus ends of spindle microtubules. There, the Dam1/DASH complex recruits the Ndc80 complex and thereby ensures dynamic coupling of microtubule plus ends with kinetochores [79,80]. Of note, Ndc80 recruitment is abolished by Ipl1-mediated phosphorylation of Dam1 or by phospho-mimicking mutations in Ipl1 target sites of Dam1 [79,80]. Importantly, our results show for the first time that Dam1 is hyper-phosphorylated in shp1, and that this altered modification significantly contributes to the severe phenotype of shp1 mutants. Using the dam1SA and damSD alleles, we set out to mimic the effects of ipl1 and glc7 loss-of-function mutations, respectively, on this specific target. Intriguingly, altering the relative abundance of Dam1 phospho-sites in shp1 almost perfectly phenocopied the genetic interactions of shp1 with ipl1 and glc7. Over-expression of dam1SA allowed robust growth up to 35uC similar to the shp1-7 ipl1-321 double mutant, whereas overexpression of dam1SD was toxic, albeit this effect was less severe than that observed for the shp1-7 glc7-129 double mutant. These results clearly show that Dam1 hyper-phosphorylation is a major cause of shp1 phenotypes related to mitotic functions of Glc7 and Ipl1.
Our analysis of the mitotic phenotype of viable, logarithmically growing shp1 mutant cells in the DF5 strain background is largely consistent with the results of a recent study using the temperaturesensitive cdc48-3 allele and a P GAL -3HA-SHP1 allele for the conditional depletion of Shp1 in the W303 strain background [31]. The authors of that study concluded that Cdc48 and Shp1 are important for the kinase to phosphatase balance at the kinetochore and proposed that Cdc48 Shp1 regulates the nuclear localization of Glc7. Our study goes beyond their analysis and The asterisk marks a cross-reactive band of the Glc7 antibody. (d) Lysates of strains expressing Glc7 3myc and Shp1 3HA as indicated were subjected to immunoprecipitation with anti-HA antibody and analyzed for co-precipitation of Glc7 3myc . The asterisk marks the immunoglobulin heavy chain of the HA antibody. Quantification of the Glc7 signal in the IP lanes relative to the heavy chain signal revealed a more than eight-fold difference between the Shp1 3HA lane and the negative control. (e, f) Nuclear localization of Glc7 GFP in shp1 mutants. WT, shp1-7 and shp1-b1 cells expressing Glc7 GFP as sole source of Glc7 were analyzed by confocal spinning disk microscopy. (e) Representative z-stack projections generated with ImageJ. Scale bars: 5 mm. (f) Quantification of the GFP signal in equal areas of nucleus versus cytosol in single z-slices of confocal images. (g) Normal binding of Glc7 to Sds22 in shp1-7. Lysates of WT and shp1-7 cells expressing Sds22 3myc as indicated were subjected to immunoprecipitation with anti-myc antibody and analyzed for co-precipitation of untagged Glc7. The asterisk marks the immunoglobulin light chain of the myc antibody. doi:10.1371/journal.pone.0056486.g007 differs in certain central aspects. We were able to demonstrate strong positive and negative genetic interactions, respectively, of shp1 null and Cdc48 binding-deficient alleles with ipl1-321 and glc7-129. Importantly, we proved that the cell cycle and chromosome segregation defects of shp1 null and Cdc48 bindingdeficient mutants are efficiently suppressed by increased Glc7 levels. Finally, we established an increased Dam1 phosphorylation in shp1 mutants, which can be suppressed by a reduction of Ipl1 activity.
One likely explanation for the differences between the two studies relates to the strains used by Cheng and Chen. In particular, the use of the cdc48-3 strain poses problems due to its pleiotropic phenotypes. Besides defects in the kinetochore-microtubule attachment reported by Cheng and Chen, cdc48-3 has been shown to be impaired in G1 progression [64,66], spindle disassembly at the end of mitosis [65], transcription factor remodeling [102], UV-induced turnover of RNAPolII [24], ERAD [103,104], and autophagy [30]. As long as specific targets of Cdc48 at the kinetochore remain unknown, it is therefore almost impossible to differentiate between direct and secondary effects of the cdc48-3 allele on cell cycle progression. Furthermore, Cheng and Chen state that the observed mitotic phenotypes of cdc48-3 were generally more severe than those of Shp1-depleted cells. This finding is likely to reflect the involvement of alternative Cdc48 cofactors, in particular Ufd1-Npl4, in Shp1-independent functions of Cdc48 during the cell cycle. Taken together, the uncertainties in the interpretation of cdc48-3 phenotypes underscore the importance of designing specific Cdc48 binding-deficient shp1 alleles. The shp1 alleles presented in this study enabled us to study genetic interactions and the effect of GLC7 over-expression in the absence of unrelated pleiotropic defects and thus allowed us to formally conclude for the first time that the regulation of Glc7 activity indeed requires the Cdc48 Shp1 complex.
The major discrepancy between this study and the study by Cheng and Chen relates to the cellular localization of Glc7 in the absence of Shp1. While these authors found that depletion of Shp1 leads to the loss of Glc7 accumulation in the nucleus, our microscopy data of strains expressing a fully functional Glc7 GFP fusion protein as the sole source of Glc7 indicated only a moderate reduction of nuclear Glc7 in shp1 (Fig. 7ef). These data are Figure 8. Impaired interaction between Glc7 and Glc8 in shp1. (a) Synthetic lethality of shp1-7 Dglc8. Growth of haploid progeny of one tetrad from the cross of shp1-7 with Dglc8 carrying YC33-SHP1 was analyzed on control (YPD) and 59FOA plates as described in the legend to Fig. 4b. (b) Synthetic lethality of Dglc8 with shp1 mutants defective in Cdc48 binding. Dglc8 shp1-7 double mutant cells carrying YC33-SHP1 and a LEU2-based centromeric plasmid for the expression of the indicated wild-type and mutant SHP1 and GLC8 alleles were analyzed as described in the legend to Fig. 4a. (c, d, e) Reduced physical interaction between Glc7 and Glc8 in shp1-7. Lysates of asynchronous (c) or a-factor-arrested (d) wild-type and shp1-7 cells expressing Glc8 3HA were subjected to immunoprecipitation with anti-HA antibody and analyzed for co-precipitation of endogenous, untagged Glc7. The asterisks mark a cross-reactive band of the Glc7 antibody (Glc7 blots) and the immunoglobulin light chain of the HA antibody (Glc8 blots), respectively. In panel (c), irrelevant lanes were removed from the figure. However, all lanes shown were on the same Western blot and exposed and processed identically. (e) Quantification of three independent experiments as in panel (c), showing the ratio of the Glc7 and Glc8 signal intensities. (f) Overexpression of GLC8 partially suppresses the temperature sensitivity of shp1-7. Wild-type (WT) and shp1-7 cells expressing the indicated GLC8 3HA alleles from an integrative plasmid under the control of the inducible P MET25 promoter were analyzed for growth at the indicated temperatures in the presence (+Met (off)) and absence (2Met (on)) of methionine in the growth medium. doi:10.1371/journal.pone.0056486.g008 supported by a normal co-immunoprecipitation of Glc7 with its nuclear targeting subunit Sds22 in shp1 (Fig. 7g), and they are in agreement with data from biochemical fractionation experiments [32]. There are two potential explanations for the discrepancy of our data with those by Cheng and Chen. First, we found that the nuclear localization of Glc7 GFP in shp1 is reduced in the presence of additional, untagged Glc7 (Fig. S3) for unknown reasons. Cheng and Chen used a strain expressing GFP Glc7 in addition to endogenous Glc7, raising the possibility that these conditions prevented a nuclear localization of the tagged Glc7 variant. Second, Cheng and Chen performed microscopy 12 hours after promoter shut-off under conditions of ongoing cell death, whereas our analysis was performed with logarithmically growing shp1 cells. Altogether, considering the available experimental evidence, a gross reduction of nuclear Glc7 levels in shp1 null mutants appears unlikely. In line with this conclusion, cytoplasmic Glc7 functions in glycogen metabolism and in the Vid pathway are affected in shp1 mutants as well [32,60], also arguing against impaired nuclear localization of Glc7 as the critical defect in shp1.
Besides the genetic interactions between glc7 and shp1 mutants, the present study showed for the first time that Shp1 and Glc7 also interact physically (Fig. 7cd). We currently do not know if this interaction is direct or indirect, for instance bridged by regulatory subunits of Glc7. While Shp1 lacks a classical RVxF motif (data not shown), which mediates the binding of many PP1 regulatory subunits [34,105,106], a number of Glc7 subunits interact through other motifs (reviewed in [34,106]). Alternatively, Cdc48 Shp1 could interact with ubiquitylated Glc7 or an ubiquitylated Glc7 interactor. Consistent with this possibility, we found that Glc7 is ubiquitylated in vivo (data not shown), in agreement with proteomics studies [107][108][109]. Clearly, the molecular basis for Shp1 binding to Glc7 remains to be elucidated in future studies.
The identification of a physical interaction between Shp1 and Glc7 raises the intriguing possibility that Cdc48 Shp1 controls Glc7 cellular functions by modulating binding of regulatory subunits. While we failed to detect Shp1-dependent differences in the interactions of Glc7 with Sds22 (Fig. 7g) and Reg1 (data not shown; see [60]), we found a strikingly reduced binding between Glc7 and Glc8 in shp1 (Fig. 8cde). Because Glc8 is considered a substrate-independent, major activator of Glc7, the reduced interaction could at least partially explain the broad spectrum of Glc7 functions affected in shp1 mutants. This interpretation is strengthened by the finding that GLC8 over-expression partially suppressed the temperature-sensitivity of shp1 (Fig. 8f). However, the reduced binding of Glc8 to Glc7 cannot be the sole cause of the pleiotropic Glc7-related phenotypes of shp1. The much less severe phenotypes of Dglc8 clearly show that GLC8 is not strictly required for viability in an otherwise unperturbed cell, suggesting that more complex mechanisms for the positive regulation of Glc7 activity must exist. Furthermore, the synthetic lethality of shp1 and Dglc8 (Fig. 8a) cannot be explained on basis of the reduced interaction between Glc7 and Glc8 observed in the shp1-7 single mutant. We therefore favor the hypothesis that the Cdc48 Shp1 complex controls the balance of Glc7 interactions with additional regulatory subunits, perhaps by mediating the dissociation of certain regulatory subunits from Glc7 by virtue of the segregase mechanism underlying other cellular functions of Cdc48 [3][4][5]. Interestingly, regulatory subunits appear to exist in excess over Glc7 in vivo [110], suggesting that they compete for binding to Glc7 [34]. In support of a competitive model for Glc7 binding, over-expression of several regulatory subunits has been shown to re-direct cellular Glc7 activity [56,111]. We speculate that in such a scenario of competitive subunit interactions, loss of Cdc48 Shp1 segregase activity would stabilize the interaction of certain regulatory subunits with Glc7, resulting in reduced binding of Glc8 and additional, yet unknown, subunits required for mitotic progression. The future identification of additional Cdc48 Shp1 targets involved in Glc7 regulation will be critical for the experimental evaluation of this hypothesis.

Plasmids
Plasmids used in this study are listed in Table 1. Unless specified otherwise below, genomic fragments of wild-type and mutant alleles were PCR-amplified from genomic yeast DNA or plasmids and cloned into yeast shuttle vectors [112] using standard techniques; details are available upon request from the authors. For the construction of YEplac195-P ADH -GLC7, the GLC7 coding region was PCR-amplified from cDNA and cloned via BamHI/ PstI into YEplac195 [112] modified to carry the ADH1 promoter and terminator in its EcoRI/BamHI and PstI/SphI sites, respectively (pAB1376). For the construction of YIplac128-P MET25 -GLC7, the P MET25 promoter was subcloned from pUG36 [113] into the SacI and XbaI sites of YIplac128 [112] modified to carry the coding sequence for a carboxyl-terminal 3HA epitope tag and the ADH1 terminator (PstI/NlaIII fragment of pYM1 [114] subcloned into YIplac128 via PstI/SphI) (pAB1165). The GLC7 coding sequence including the stop codon was cloned into pAB1165 via SpeI/PstI (pAB1280). The coding sequences (excluding the stop codon) of GLC8 or glc8-T118A were cloned into pAB1165 via XbaI/PstI. YIplac211-GFP-LacI was constructed by subcloning the P HIS3 -GFP-LacI-NLS fragment from pAFS135 [115] into YIplac211 [112] via KpnI/XbaI (pAB2040). shp1 mutant alleles including shp1-7 (first codon mutated from ATG to ACC) and alleles carrying combinations of mutations in Cdc48 binding sites as depicted in Fig. 2a (F396/P397/I398RG; R360RA; F306/Q309/Q311/R312/L313RAAAAA; F201/ R204/F206/R207/L208RAAAAA; details available from the authors upon request) were generated from pAB1847 by sitespecific mutagenesis using the QuikChange II XL kit (Stratagene) and verified by sequencing.

Yeast strains and media
All strains used in this study are derivatives of DF5 [116] and listed in Table 2. Chromosomal deletions and fusions with epitope tags or fluorescent proteins were generated using standard methods [114,117,118]. For the characterization of Cdc48 binding-deficient shp1 alleles (Fig. 2a-c), the respective YIplac211-shp1 plasmids were linearized in URA3 with StuI and transformed into Dshp1::kanMX6. Transformants were selected on SC-Ura media for correct integration into the ura3-52 locus, and expression of Shp1 was confirmed by Western blot. For the construction of the shp1-7, shp1-a1 and shp1-b1 mutant strains by a pop-in pop-out strategy, the respective YIplac211-shp1 plasmids were linearized within the SHP1 open reading frame by restriction digest with BamHI and transformed into DF5a. Transformants were selected for Ura prototrophy, and correct integration was verified by colony PCR. Positive clones were streaked out twice on 59FOA, and single colonies were picked. These clones were then analyzed by colony PCR for the presence of the full-length shp1 allele to exclude aberrant pop-out events, by Western blot for Shp1 levels, and for temperature sensitivity. Similar approaches were used to transfer the conditional alleles glc7-129, ipl1-321, and sds22-6 into the DF5 strain background. All mutant alleles were finally sequenced, either by cloning the open reading frame together with 1 kb of upstream and downstream flanking sequences, or by directly sequencing a PCR product of the mutated region. Double mutants were constructed by crossing the respective conditional allele with the shp1-7 mutant carrying YCplac33-SHP1.
Yeast was cultured in standard YPD and SC media [119]. For the induction of the P MET25 promoter, cells were first grown in SC media supplemented with 2 mM methionine, washed twice with H 2 O, and then transferred to SC medium lacking methionine.

a-factor arrest/release
Overnight cultures of wild-type and mutant strains were diluted to an OD 600 nm of 0.1 (0.15 for shp1 mutants) in 50 ml YPD. The cultures were then grown at 25uC for approximately four hours until reaching an OD 600 nm of 0.3-0.35. 10 mM a-factor (central core facility, Max Planck Institute of Biochemistry, Martinsried, Germany) in DMSO were added, and the cells were allowed to arrest for three hours at 25uC. Directly before addition of a-factor, a control sample from the asynchronous culture was collected, and the pellet was frozen in liquid nitrogen. The efficiency of the arrest was determined by FACS analysis and/or Western blot for Clb2 levels after three hours of arrest. The cultures were then washed two times with equal volumes of YPD and resuspended to a final OD 600 nm of approximately 0.5 in YPD. The released cultures were incubated at 25uC, and at various time-points an amount of cells corresponding to 1 ml of OD 600 nm = 0.6 was removed from the culture, quickly pelleted and frozen in liquid nitrogen. Subsequently, the cell pellets were lysed by TCA precipitation and resuspended in 50 ml HU/DTT buffer (8 M urea, 5% SDS, 0.2 M Tris pH 6.8, bromophenol blue, 0.1 M DTT). Fluctuations in the levels of Clb2 and other cell cycle marker proteins were analyzed by Western blot.

FACS analysis
Analysis of DNA content by FACS was performed exactly as described [120] using a BD FACS Calibur and CellQuest Pro software or a BD FACS Canto and FACS Diva software.

Microscopy
Yeast strains were grown in appropriate, sterile-filtered SC media to avoid high background fluorescence. Cells were immobilized by incubation on cover slips coated with 1 mg/ml concanavalin A (Type 5, Sigma Aldrich) for at least 30 min. Cells from logarithmically growing cultures were directly spotted on the cover slip, shortly incubated and sealed with Vaseline. Spinning disk confocal microscopy of Glc7 GFP localization (Fig. 7e) was performed using a microscope setup described previously [122]. In brief, cells expressing Glc7 GFP were analyzed using a laser-based spinning disk confocal microscope (Andor Technology). Filtered images (Semrock emission filters in a Sutter filter wheel) were captured with a D-977 iXon EMCCD+ camera (Andor Technology) after twofold magnification (Andor Technology) by using a 1006TIRFM/1.45 objective (Olympus). Z-Stacks were recorded with a spacing of 0.2 mm over the entire cell (10-25 planes). Images were processed with ImageJ software (http://rsbweb.nih. gov/ij/) and the MBF ImageJ for Microscopy collection of plugins (http://www.macbiophotonics.ca/imagej/). For quantification of the Glc7 GFP signal, single Z-slices of confocal images that had been recorded under identical conditions were used. The average GFP fluorescence intensity was measured in an area of equal size in the nucleus and cytoplasm using ImageJ software, and the ratio was calculated. Fluorescence microscopy of Glc7 GFP localization upon additional expression of untagged GLC7 (Fig. S3) was carried out using a Zeiss Axiovert 200 M microscope equipped with an Axio Apochrom (Zeiss) 1006/1.4 oil objective and the filter set #10 (FITC). Images were captured using an AxioCam MRm TV2/30 0.636 (Zeiss) camera and AxioVision LE software.
For the analysis of sister chromatid separation (Fig. 5cd), cultures were grown to log-phase in SC medium+/22 mM methionine, harvested, and resuspended in sterile filtered medium. 1.4% low-melting agarose was added in equal volume to mount the samples on cover slips. Microscopy was carried out on a Nikon TiE inverted live cell system with a motorized Prior Z-stage and Perfect Focus System using a 1006 1.45 NA objective (Nikon). Eleven Z-Stacks (spacing 0.3 mm) were recorded with a Photometrics HQ2 camera and analyzed using Nikon NIS Elements software. For differential interference contrast (DIC) microscopy, a single snap-shot was taken. All images were recorded using identical exposure times. Medium-to large-budded cells in each Table 1. Plasmids used in this study.

Plasmid
Description Source class were counted using NIS Elements software by assessing the position of the GFP LacI-marked chromosome III relative to the SPBs (Spc42 Mars ).

Immunoprecipitation
Yeast cultures were grown in YPD to an OD 600 of 0.7, harvested and washed once with cold ddH 2 O/1 mM PMSF. Cells were then lysed in IP buffer (50 mM Tris/HCl pH 7.5, 100 mM KCl, 5 mM MgCl2, 0.1% NP-40, 10% glycerol, 10 mM NaF, 2 mM PMSF, complete protease inhibitor cocktail (Roche)) by addition of zirconia beads (Biospec) and vortexing. After lysis, the NP-40 concentration was raised to 1%, and the extracts were centrifuged at 2,600 g for 5 min, followed by centrifugation at 20,000 g for 25 min. An input sample (10 ml) was taken prior to antibody addition and denatured by addition of an equal amount of HU/DTT buffer and incubation at 65uC for 10 min. The supernatants were incubated with 20 ml pre-coupled HA antibody, 4.5 ml myc antibody, or 2 ml Shp1 antibody and rotated at 4uC overnight. Immunocomplexes were then either bound to 20 ml protein A sepharose beads (GE Healthcare) for three hours (4uC), or directly washed (pre-coupled HA antibody) four times (600 ml IP buffer/1% NP-40 for 10 min, 800 ml IP buffer/1% NP-40 8 min, 800 ml IP-buffer 5 min, 1 ml IP buffer). Bound proteins were eluted by incubation with 25 ml HU/DTT buffer for 10 min at 65uC and analyzed by Western blot. Figure S1 Genetic interactions of shp1 with glc7 and ipl1. (a) Synthetic lethality of shp1-7 glc7-129. Growth of haploid progeny of one tetrad from the crossing of shp1-7 with glc7-129 carrying YC33-SHP1 was analyzed on control (YPD) and 59FOA plates as described in the legend to Fig. 4b. (b) Positive genetic interaction between shp1-7 and ipl1-321. Growth of haploid progeny of one tetrad from the crossing of shp1-7 with ipl1-321 was analyzed at the indicated temperatures. (TIF) Figure S2 Nuclear localization of Glc7 in shp1 mutants. Asynchronous logarithmic cultures of wild-type or the indicated shp1 mutants expressing Glc7 GFP as the only source of Glc7 and the nuclear envelope marker Pom34 Mars were grown at RT and analyzed by live-cell fluorescence microscopy. Scale bar 5 mm. Fluorescent images are z-stack projections, DIC a single image of the focus plane. (TIF) Figure S3 GLC7 expression levels influence the nuclear localization of the Glc7 GFP fusion protein in shp1-7. (a) Wild-type (WT) or shp1-7 strains expressing Glc7 GFP as the only source of Glc7 were transformed with either empty YC plasmids or plasmids encoding the GLC7 gene under control of its own promoter (YC-GLC7) or the ADH promoter (YEpADH-GLC7). Asynchronous logarithmic cultures of the indicated strains were analyzed by livecell fluorescence microscopy. GFP (Glc7) fluorescence, DIC images, and the overlay are depicted. (b) Lysates of the cultures used in (a) were analyzed by Western blot against Glc7 and Cdc48 (loading control). For comparison, WT and shp1-7 expressing endogenous untagged Glc7 are also shown. The asterisk marks a cross-reactive band of the Glc7 antibody. (TIF)