New ubiquitin-dependent mechanisms regulating the Aurora B–protein phosphatase 1 balance in Saccharomyces cerevisiae

Protein ubiquitylation is a common and complex protein modification in which the highly conserved ubiquitin protein is covalently linked to the ε amino group of lysine residues. Most of these modification events regulate protein stability and quality control via the proteasome or through autophagy, but ubiquitylation can result in a variety of other outcomes, including changes to subcellular localization, protein–protein interactions or enzymatic activity (reviewed in Rape, 2018). The major substrate for ubiquitylation is ubiquitin itself, in which seven conserved lysine residues and the N-terminal methionine residue can serve as substrates for additional rounds of ubiquitylation, leading to polyubiquitylated proteins. Different linkages in polyubiquitin can adopt unique structures, leading to different fates for a protein (Komander and Rape, 2012). Specificity is dictated by a diverse array of proteins containing ubiquitin-binding domains (UBDs), which bind to specific sets of ubiquitylated proteins. Ubiquitin-binding proteins associated with the proteasome escort mostly K48- and K11-linked polyubiquitin to the proteasome (Chau et al., 1989; Jin et al., 2008) or to the Cdc48–Ufd1–Npl4 segregase (Ye et al., 2003), whereas proteins with affinity for monoubiquitin or polyubiquitin formed with other linkages regulate ubiquitylated proteins in other ways (Komander and Rape, 2012). Finally, the attachment of ubiquitin can be reversed by de-ubiquitylating enzymes (DUBs), which make this modification highly dynamic (Mevissen and Komander, 2017).

Protein ubiquitylation is carried out by the ordered activities of the E1 ubiquitin-activating enzyme, which transfers ubiquitin to an E2 ubiquitin-conjugating enzyme (Lorenz et al., 2013). Ubiquitin is then transferred to the substrate lysine residue, usually with the assistance of an E3 ubiquitin ligase. There is a clear hierarchy in the abundance and conservation of these enzyme classes. One or a few highly conserved E1 enzymes activate ∼10 to ∼70 E2s, depending on the species (Jue et al., 2015). E3s number in the hundreds (Finley et al., 2012). Given this hierarchy, interest in specific ubiquitin-dependent processes has logically focused on E2 and E3 enzymes. However, the structures of several E1–E2 complexes indicate that different E2s associate with unique structural elements of the E1 (Olsen and Lima, 2013; Lv et al., 2017), suggesting a potential regulatory role for the E1 enzyme. In addition, Uba1, the sole E1 enzyme in Saccharomyces cerevisiae, is post-translationally modified by phosphorylation, ubiquitylation and succinylation at multiple sites (Albuquerque et al., 2008; Holt et al., 2009; Swaney et al., 2013; Fang et al., 2014), providing potential mechanisms for regulating specific interactions with E2 enzymes.

Whether or not E1 contributes to selectivity, the effects of impaired E1 activity in model systems have demonstrated the wide array of processes regulated by ubiquitylation. Study of loss-of-function mutations in Drosophila melanogaster UBA1 have shown that mutations in this E1 reduce fly lifespan and impair motor function (Liu and Pfleger, 2013). Characterization of a Caenorhabditis elegans uba1-1 mutant revealed roles for ubiquitylation in regulating sperm fertility, control of body size and sex-specific development (Kulkarni and Smith, 2008). In S. cerevisiae, the temperature-sensitive uba1-204 allele has been used to show that ubiquitin conjugates are necessary for targeting the ubiquitin-binding adaptor protein Rad23 to the proteasome (Ghaboosi and Deshaies, 2007). Similarly, mammalian cell lines that harbor mutations in the E1 gene or in other genes that affect E1 function have demonstrated important roles for ubiquitylation in the heat-shock response and in regulation of cell cycle progression (Kulka et al., 1988; Salvat et al., 2000; Lao et al., 2012; Sugaya et al., 2014).

Some mutations in E1 confer thermosensitivity and lead to cell cycle arrest, which provided early evidence that ubiquitylation is required for cell cycle progression through both S and G2 phases (Finley et al., 1984; Ayusawa et al., 1992; Nishitani et al., 1992). The anaphase-promoting complex (APC/C) and Skp1–Cdc53–F-box (SCF) E3 ligases are responsible for targeting cell cycle-specific substrates to the proteasome (reviewed in Wickliffe et al., 2009). However, several lines of evidence also show that ubiquitylation regulates proteasome-independent cell cycle events. For example, in human cells, ubiquitylation of components of the chromosome passenger complex (CPC) trigger CPC binding to centromeres during prometaphase (Vong et al., 2005). The CPC is a four subunit protein complex, composed of Aurora B kinase (Ipl1 in S. cerevisiae) and three auxiliary subunits (Carmena et al., 2012). Prior to anaphase, Aurora B/Ipl1 phosphorylates kinetochore proteins, which destabilizes their attachment to spindle microtubules, and allows correction of improper kinetochore–microtubule interactions (reviewed in Tanaka et al., 2005). Thus, Aurora B activity creates unattached kinetochores and activates the spindle assembly checkpoint (SAC), which monitors bipolar chromosome attachment to the mitotic spindle and delays anaphase until proper bi-orientation is achieved (Pinsky et al., 2006b). In fission yeast and human cells, Aurora B is localized to the inner centromere prior to anaphase via its CPC partner proteins, and this localization is controlled by histone H3 and H2A phosphorylation (Kelly et al., 2010; Yamagishi et al., 2010). In human cells, a cullin-based ubiquitin ligase (Cul3) interacts with three different substrate-specific adaptor proteins to ubiquitylate Aurora B and target it differently to the spindle midzone and midzone microtubules (Sumara et al., 2007; Maerki et al., 2009). Ubiquitylated Aurora B is subsequently removed from chromosomes by the AAA+ ATPase Cdc48 (known as p97 and VCP in mammals) and its adaptor proteins Ufd1-Npl4, which facilitates nuclear reformation at the end of mitosis (Ramadan et al., 2007). Following anaphase, Aurora B migrates from chromosomes to the central spindle where it plays a role in cytokinesis (Carmena et al., 2012).

The protein phosphatase PP1 (Glc7 in S. cerevisiae) opposes the CPC activity, dephosphorylating kinetochore proteins to stabilize kinetochore–microtubule interactions and silence the SAC (Pinsky et al., 2009; Vanoosthuyse and Hardwick, 2009; Liu et al., 2010; Wurzenberger et al., 2012). Therefore, Aurora B and PP1 activities must be tightly balanced to ensure fidelity of mitotic chromosome segregation. In addition to opposing Aurora B at kinetochores, PP1 has many additional regulatory roles, ranging from carbohydrate metabolism to learning and memory (reviewed in Virshup and Shenolikar, 2009; Bollen et al., 2010). The activity of the highly conserved PP1 catalytic subunit is tightly controlled by a large collection of binding proteins that associate with PP1 via short conserved motifs, the most well studied of which is the RVxF motif found in many PP1-binding proteins (Bollen et al., 2010). Outside these motifs, most PP1-binding proteins are not highly conserved but allow PP1 to serve as the catalytic subunit for a large number of tissue- and cell type-specific phosphatase holoenzymes. Remarkably, Sds22 and Ypi1 (also known as inhibitor 3 or PPP1R11 in mammals), two of the most highly conserved PP1-binding partners, have been implicated in regulating the PP1 phosphatase activity that opposes Aurora B (Hisamoto et al., 1995; Peggie et al., 2002; Bharucha et al., 2008; Posch et al., 2010; Wurzenberger et al., 2012; Eiteneuer et al., 2014). Genetic studies have also identified the AAA+ ATPase Cdc48 and its adaptor Shp1 (known as p47 or NSFL1C in humans) (Zhang et al., 1995; Cheng and Chen, 2010; Böhm, 2011; Robinson et al., 2012; Cheng and Chen, 2015) as Glc7 regulators. Mutations in any of these regulatory proteins lead to mitotic delay or arrest, reduce Glc7 nuclear localization and cause formation of Glc7 aggregates (Peggie et al., 2002; Bharucha et al., 2008; Cheng and Chen, 2010, 2015). The observation that Glc7 aggregate formation is blocked by cycloheximide in cdc48 and shp1 mutants led Cheng and Chen (Cheng and Chen, 2015) to propose that these regulators participate in the folding or maturation of nascent Glc7.

Genetic studies using conditional IPL1 mutant alleles have been instrumental in identifying factors that regulate Ipl1, Glc7 and the kinetochore (Tung et al., 1995; Cheeseman et al., 2002; Peggie et al., 2002; Zhang et al., 2005; Pinsky et al., 2006a; Bharucha et al., 2008; Ng et al., 2009; Cheng and Chen, 2010; Tatchell et al., 2011; Robinson et al., 2012; Makrantoni et al., 2017). Surprisingly, considering the importance of ubiquitylation in cell cycle progression, relatively few IPL1 suppressors or enhancers alter direct components of ubiquitylation pathways. Mutations altering the ubiquitin-conjugating enzyme Rad6 and its E3 partner Bre1 weakly suppress the temperature sensitivity of ipl1-2 mutants indirectly, through Set1-dependent methylation of kinetochore protein Dam1 (see Latham et al., 2011 and below). A deletion mutation of the E3 SLX5 has the opposite effect, causing ipl1-321 mutants to grow more slowly (Ng et al., 2009). We report here that a missense mutation in the E1 ubiquitin activating enzyme UBA1 suppresses ipl1-2. Characterization of this UBA1 allele reveals novel roles for ubiquitylation in Ipl1 and Glc7 regulation.

A previously uncharacterized ipl1-2 suppressor mutation, denoted revertant 100 (rev100), segregates as a single Mendelian allele for which slow growth co-segregates with suppression of ipl1-2. rev100 suppresses the temperature sensitivity of ipl1-2 up to 33°C (Fig. 1A). In an IPL1 WT background, rev100 confers slow growth at 24°C, sensitivity to low (14°C) and high (37°C) temperatures and sensitivity to caffeine (Fig. 1B). rev100 causes slight sensitivity to 0.1 M CsCl in growth medium but higher resistance to 0.1 M LiCl in growth medium (Fig. 1B). To identify the mutation responsible for ipl1-2 suppression, whole-genome sequence analysis was performed on pooled DNA from six backcrossed rev100 ascoporal clones. A total of 62 differences between the rev100 mutant DNA and our parent strain (KT1113) were found. Of these, 18 were observed in all sequence reads (Table S1). Given the range of physiological pathways impacted by ubiquitin, a likely candidate for the rev100 mutation is the non-synonymous T to C transversion in UBA1, which results in a tryptophan to arginine missense mutation at residue 928 in Uba1, the essential ubiquitin-activating E1 enzyme. We confirmed that the T to C transversion is present in a rev100 mutant strain by Sanger sequence analysis of PCR-amplified UBA1 DNA. A cross between a rev100 strain (KT3300) and a trp3::KanMX deletion strain (EG2365-1C) gave a map distance of 2.4 map units (41P:2T:0NPD). This tight linkage would be predicted if the rev100 mutation lies in UBA1, which is adjacent to the TRP3 locus. Finally, we transformed an ipl1-2 rev100 strain with a replicating yeast shuttle vector containing UBA1. The UBA1-containing plasmid restored the temperature sensitivity and complemented the growth defects caused by rev100 (Fig. 1C). Taken together, these results indicate that rev100 is a new allele of UBA1. Henceforth, we refer to the suppressor mutation in rev100 as uba1-W928R.

A key role of Ipl1 is to ensure correct bipolar attachment of chromosomes to spindle microtubules by phosphorylating kinetochore proteins, which destabilizes attached microtubules. Loss of Ipl1 activity results in increased rates of chromosome loss and cell death (Chan and Botstein, 1993; Biggins et al., 1999). To determine whether uba1-W928R suppresses chromosome loss in the ipl1-2 mutant, we monitored rates of chromosome III loss in diploid strains homozygous for both ipl1-2 and uba1-W928R. Diploid yeast strains monosomic for chromosome III are able to mate. Therefore, acquisition of mating ability by diploid strains can be used as a read-out for chromosome III loss. Diploid yeast strains were grown at 24°C to mid-log phase and shifted to 30°C. Cell viability and mating ability were assayed after the temperature shift. As expected, the homozygous ipl1-2 diploid with WT UBA1 showed loss of cell viability and an increase in the mating ability at 30°C (open circles in Fig. 1D). In contrast, the viability and mating ability of the homozygous uba1-W928R ipl1-2 strain did not change following the temperature shift (filled triangles in Fig. 1D). We also monitored the loss of lacO-tagged chromosome IV in strains expressing a GFP-tagged lac repressor (Straight et al., 1997). In ipl1-2 mutant cells, 39% of cells grown at 30°C for 4 h either had an abnormal segregation pattern of the lacO array or no fluorescence. Loss of fluorescence can occur upon the loss of chromosome XV, which bears the GFP–lacI gene. The missegregation frequency was reduced to 3.7% in the ipl1-2 uba1-W928R strain (Fig. 1E). Taken together, these results indicate that uba1-W928R rescues viability and suppresses chromosome loss due to ipl1-2.

Uba1 transfers an activated ubiquitin to an E2-conjugating enzyme, which binds Uba1 via the ubiquitin fold domain (UFD). The UFD undergoes distinct movement during the E1–E2 handoff, made possible by a flexible β-hairpin linker between the UFD and the main body of Uba1 (Lee and Schindelin, 2008; Olsen and Lima, 2013). W928, which is conserved in all E1 proteins (Fig. 2A), is located adjacent to the β-hairpin linker (residues 915 to 927), tethering the main body of the E1 enzyme to the UFD (Fig. 2B). Interestingly, studies in S. pombe that compared Uba1 structures with and without the E2 substrate show that UFD rotation is achieved by bending that begins immediately after W928, with a complementary twisting of the β-hairpin to maintain interactions between the Uba1 core (adenylation domain) and the UFD linker (Olsen and Lima, 2013). These specific contacts permit attainment of the correct UFD conformation to facilitate E2 enzyme recruitment. We reasoned that a tryptophan to arginine mutation at this critical residue could alter conformational changes needed for binding of some E2 enzyme substrates. To investigate possible structural consequences of the tryptophan to arginine mutation in silico, we modeled the change with PyMOL using the published S. cerevisiae Uba1 structure (PDB: 3CMM) (Fig. 2C). Replacing the W928 residue (red) with a positively charged arginine residue breaks hydrogen bonding with T583. Also, R928 is aligned away from the hydrophobic pocket (purple), further increasing repulsive forces from the adjacent R581 (green) (Fig. 2C, right panel), which may distort the rotation of the UFD core. Thus, the uba1-W928R mutation is likely to disrupt the conformation between the UFD and the main body of Uba1, and could alter the overall level of protein ubiquitylation, as has been reported previously for the uba1-204 allele (Ghaboosi and Deshaies, 2007). To test this possibility, we grew wild-type (WT), uba1-W928R and uba1-204 strains at 24°C to mid-log phase, raised the temperature to 37°C, and monitored protein ubiquitylation by immunoblot analysis. As previously reported (Ghaboosi and Deshaies, 2007), protein ubiquitylation drops rapidly in the uba1-204 strain after the temperature shift (Fig. 2D). In contrast, protein-conjugated ubiquitin levels are reduced in the uba1-W928R strain but remain constant at all temperatures. Furthermore, the increased level of protein ubiquitylation that was observed for the WT strain after an 1 h incubation at 37°C was not recapitulated in the uba1-W928R strain. The uba1-204 allele does not suppress ipl1-2 (Fig. S1), likely because uba1-204 strains retain Uba1 function at 30°C, the non-permissive temperature for ipl1-2 cells.

To address the extent to which the ubiquitylation defect of uba1-W928R impacts protein turnover, we tested for a possible defect in N-end rule proteasomal degradation (Varshavsky, 1996) by using a copper-inducible URA3 reporter gene coupled to an N-end degron (Varshavsky, 2005; Hanna et al., 2006). The rate of growth of strains expressing this reporter in the absence of uracil and presence of copper inversely reflects the rate of degradation of the Ura3 reporter protein. A rad6Δ strain was used as a positive control, since the requirement for Rad6 as the E2 enzyme for the N-end rule pathway is well established (Dohmen et al., 1991). As expected, we observed that a rad6Δ mutant displays a strong growth advantage over WT when cultured on medium lacking uracil. The uba1-W928R strain grows better than WT but not as well as rad6Δ cells (Fig. 2E). To confirm that the turnover of the Ura3 reporter protein is reduced in the uba1-W928R strain, we performed cycloheximide-chase analyses. We observed rapid degradation of the Ura3 reporter protein in the WT strain, with a half-life of ∼10 min. In contrast, the uba1-W928R mutant shows stabilization of the Ura3 reporter protein accompanied by an increased steady-state level (Fig. 2F). The rad6Δ mutant exhibits very high levels of Ura3. Thus, uba1-W928R causes a partial defect in proteasomal degradation via the N-end rule pathway. Since Rad6 is the downstream E2 enzyme largely responsible for N-end-rule-mediated protein degradation, this result also suggests that the uba1-W928R mutation impairs Rad6-dependent ubiquitylation.

Rad6 also has a well-characterized role in post-replicative DNA repair (Cejka et al., 2001; Giannattasio et al., 2005). To test the possibility that uba1-W928R causes a defect in DNA repair, we assayed the sensitivity of WT and uba1-W928R strains to UV irradiation. A rad9Δ strain, defective in the DNA damage-induced cell cycle checkpoint (Weinert and Hartwell, 1988, 1990), was used as the positive control. We observed that uba1-W928R mutant strains are more sensitive to UV radiation than WT cells (Fig. 2G). The uba1-W928R rad9Δ double mutant is more sensitive than either of the single mutants (Fig. 2G; Fig. S2), indicating that uba1-W928R cells are defective in DNA repair. These results provide further evidence that uba1-W928R reduces ubiquitylation mediated by Rad6.

uba1-W928R cells are larger (Fig. 3D) and slower growing (Fig. 1A) than WT cells. To determine whether uba1-W928R mutant cells are delayed in mitotic progression, we synchronized cells in G1 using α-factor, released them into fresh medium and assayed levels of Pds1 (securin), which accumulates in mitosis and is rapidly degraded at the onset of anaphase. The strains also contain GFP–Spc42, a component of the spindle pole body, allowing direct monitoring of spindle length. As shown in the top panel of Fig. 3A, Pds1 rapidly accumulates after release from the α-factor block in WT cells and is largely degraded before 90 min, indicating that most of the cells have entered anaphase by 90 min. In contrast, uba1-W928R cells are delayed in mitosis for an additional 30–60 min. We also measured the distance between the two spindle pole bodies in these cells and confirmed that formation of long spindles correlates with Pds1 degradation. At 120 mins after release from α-factor, 60% of the uba1-W928R mutant cells have long spindles (defined by distance between the spindle pole bodies greater than 2.5 µm) as compared to 20% for the WT (Fig. 3B). Thus, uba1-W928R cells exhibit a delay in anaphase entry. Such a delay could result from a failure to terminate the spindle assembly checkpoint (SAC). To test whether SAC activation is responsible for the mitotic delay observed for uba1-W928R cells, we compared Pds1 kinetics in mad1Δ uba1-W928R cells and uba1-W928R single-mutant cells. The kinetics of Pds1 appearance and disappearance are identical in uba1-W928R and uba1-W928R mad1Δ cells (Fig. 3C), indicating that the mitotic delay does not result from a failure to terminate the SAC.

The mitotic spindle lengths inferred from distances between the two spindle pole bodies are longer for uba1-W928R cells than for WT cells. To confirm that the spindle pole body distances accurately reflect spindle length, we performed live-cell microscopy on uba1-W928R strains expressing GFP-tagged tubulin. Consistent with the increased distances between spindle pole bodies, we observed that uba1-W928R cells show abnormally long spindles (Fig. 3D). The average spindle length at late anaphase, just before spindle disassembly, is 6.5 µm for the WT (n=5), compared to 11.2 µm for the uba1-W938R mutant (n=5). We followed spindle dynamics in individual cells by time-lapse microscopy and found that spindle disassembly is delayed in uba1-W928R (Fig. 3D, Movies 1 and 2). The average time between anaphase onset and spindle breakdown is 18 min for the WT, compared to 51 min for uba1-W928R cells. The spindles of only two of five uba1-W928R cells disassembled by 60 min.

This increase in spindle length may be due, in part, to an increase in cell size. However, we note that more than 10% of uba1-W928R cells have curved spindles (Fig. 3D), indicating that spindle disassembly may be influenced by uba1-W928R. Microtubule motor proteins are required to scale the anaphase spindle length to the cell length (Straight et al., 1998; Rizk et al., 2014). Similar to what is seen in uba1-W928R, cells containing mutations/deletions in the kinesin-related motor protein Kip3 exhibit delayed cell cycle progression through anaphase and display abnormally long anaphase spindles (Straight et al., 1998; Rizk et al., 2014). kip3 mutants are also synthetic lethal with cin8, which encodes another kinesin motor protein (Pan et al., 2004). We tested for genetic interactions between uba1-W928R and mutations in microtubule motor genes, DYN1, CIN8, KIP2, KIP3, KIP1 and KAR3. The most striking genetic interaction that we observed was synthetic lethality between cin8 and uba1-W928R (Table S2). cin8 mutations have been reported to interact genetically with many mutants, but the phenotypic similarity of kip3 and uba1-W928R, and the cin8 uba1-W928R synthetic lethality, are consistent with the possibility that ubiquitylation directly or indirectly influences the activity of Kip3. Taken together, these results suggest that both defective motor protein function and slower degradation of Pds1 may contribute to the mitotic cell cycle delay exhibited by uba1-W928R.

Given that uba1-W928R mutants are defective in protein ubiquitylation and degradation via the N-end rule pathway, a likely explanation for the ipl1-2 suppression phenotype is that the Ipl1 protein is stabilized in the uba1-W928R mutant. To test this hypothesis, we first assayed the phosphorylation state of histone H3 S10 (H3-S10), a known substrate of Ipl1 and Glc7 (Hsu et al., 2000). As a control, we also tested the phosphorylation state of H3-S10 in the glc7-127 mutant, which restores H3-S10 phosphorylation in an ipl1-2 mutant (Hsu et al., 2000). The uba1-W928R mutation in the ipl1-2 strain raises H3-S10 phosphorylation to levels that are comparable to those in of WT (Fig. 4A). The restoration seen with uba1-W928R is less effective than that seen with glc7-127, consistent with the fact that glc7-127 is a much stronger ipl1-2 suppressor. The partial restoration of H3-S10 phosphorylation by uba1-W928R is consistent with the possibility that this mutation either increases Ipl1 kinase activity and/or reduces the opposing Glc7 phosphatase activity.

To determine whether uba1-W928R increases Ipl1 protein levels, we tagged the IPL1 and ipl1-2 loci with a 13× Myc C-terminal epitope (13Xmyc) and confirmed that the ipl1-2-13Xmyc strain has a similar profile of temperature sensitivity to the untagged ipl1-2 mutant (Fig. S3A). Ipl1 and Ipl1-2 protein levels are increased 2.5- to 3-fold in the uba1-W928R mutant (Fig. 4B). The half-life of Ipl1-2–13Xmyc is 1.5 h in the UBA1⁺ strain but more than 4 h in the uba1-W928R mutant background (Fig. 4D). To test whether an increase in Ipl1 protein level is sufficient to suppress ipl1-2, we expressed Ipl1 or Ipl1-2 from the galactose-inducible GAL1 promoter in an ipl1-2 strain and assessed growth at high temperatures. We observed that increased protein levels of either Ipl1 or Ipl1-2 allowed growth of ipl1-2 cells even at 37°C (Fig. 4C, upper panel). The ipl1-2 strain expressing Ipl1-2 from the GAL1 promoter grows well on galactose medium but not on glucose medium, when expression is greatly reduced. Interestingly, expression of Ipl1 from the GAL1 promoter fully suppresses the temperature sensitivity of ipl1-2 even on glucose medium, suggesting that only a low level of Ipl1 is sufficient to suppress ipl1-2 (Fig. 4C, lower panel). Taken together, these results suggest that uba1-W928R suppresses ipl1-2 at least in part by stabilizing the Ipl1-2 protein and, thus, increasing its protein level and activity. Consistent with this, we observe that Ipl1-2 accumulates upon proteasome inhibition (Fig. S3B).

The elevation of Ipl1-2 levels due to uba1-W928R could account entirely for the ipl1-2 suppression phenotype. However, we observe that uba1-W928R exhibits strong genetic interactions with glc7 mutants (see below), suggesting the possibility that ubiquitylation affects Glc7 function. To test this, we first determined Glc7 protein levels in uba1-W928R strains. Unlike Ipl1, Glc7 protein levels are not elevated in uba1-W928R, even after treatment with MG132, which blocks proteasomal degradation (Fig. 5A). Consistent with these observations, Glc7 protein stability remains unchanged in the uba1-W928R mutant (Fig. 5B).

Glc7 mutations causing reduced activity are known to suppress ipl1-2, presumably by restoring the balance of phosphatase-to-kinase activity (Hsu et al., 2000). Even though Glc7 protein levels are unchanged in the uba1-W928R mutant, we sought to determine whether Glc7 is functionally compromised. We first analyzed meiotic progeny of crosses between uba1-W928R and two glc7 mutant strains (glc7-127 and glc7-S99L), which were shown previously to suppress ipl1-2 (Hsu et al., 2000; Robinson et al., 2012). uba1-W928R glc7 double mutants are very slow growing, failing to form viable macro colonies in 3 days (Fig. 5C). This could result from the increase in Ipl1 levels, which in the absence of fully functional Glc7 would lead to loss of cell growth and/or viability. If this were the case, then decreasing Ipl1 activity in the uba1-W928R glc7 mutant should alleviate the growth defect. To test this, we compared growth rates of uba1-W928R glc7-127 spore clones, with or without the ipl1-2 mutant allele. Ipl1-2 is known to have reduced kinase activity at 24°C, the temperature at which the spore clones were incubated (Hsu et al., 2000). No difference in growth rates were noted for uba1-W928R glc7-127 mutants, irrespective of their IPL1 genotype (Fig. 5D), arguing that the slow growth of uba1-W928R glc7 mutant strains is independent of Ipl1 activity. These observations are more consistent with the hypothesis that some aspect of Glc7 function is impaired in uba1-W928R mutants.

Synthetic lethality between uba1-W928R and glc7 mutations could be explained if proper Glc7 function requires a ubiquitylation-dependent process. In this scenario, reduced ubiquitylation in the uba1-W928R mutant would negatively impact on Glc7 function, and further impairment due to a glc7 mutation would lead to the inviability of the uba1-W928R glc7 double mutant. If this were the case, mutations in positive regulators of Glc7 also should show synthetic growth defects with uba1-W928R. To test this, we analyzed meiotic progeny of a cross between an sds22-D2N D199N mutant strain that was isolated in an ipl1-2 suppression screen (Robinson et al., 2012) and uba1-W928R. Sds22 is a leucine-rich repeat protein and an essential positive regulator of Glc7 (MacKelvie et al., 1995; Peggie et al., 2002; Pedelini et al., 2007). As observed with glc7 mutants, the sds22-D2N D199N mutant strain also exhibits strong negative genetic interactions with uba1-W928R (Fig. 5E), further strengthening the hypothesis that some aspect of Glc7 function is regulated by ubiquitylation.

Previously characterized ipl1-2 suppressors that act by compromising Glc7 function, such as sds22-6, ypi1-GFP, cdc48-3 and shp1^ts, either reduce the nuclear-to-cytoplasmic ratio of Glc7 or cause Glc7 aggregation (Pedelini et al., 2007; Bharucha et al., 2008; Cheng and Chen, 2010, 2015; Robinson et al., 2012). We examined WT and uba1-W928R cells expressing Glc7–mCitrine by fluorescence microscopy. We observed that the Glc7 nuclear-to-cytoplasmic ratio is significantly reduced in uba1-W928R cells (Fig. 5F). More strikingly, 62% of the uba1-W928R cells show bright Glc7 foci in the nucleus. The average intensity of these spots is 2-fold higher than the nuclear Glc7 fluorescence and 4.5-fold higher than Glc7 fluorescence in the cytoplasm. 97% of Glc7 foci are located adjacent to, but not overlapping the area occupied by the nucleolar protein, Nop56 (Fig. 5F; Fig. S4A). We also observed that the percentage of cells possessing Glc7 foci increases significantly when uba1-W928R cells are shifted from 24°C to restrictive temperatures. 73% of these cells show foci when shifted to 33°C for 2 h and 85% of uba1-W928R cells show bright Glc7 foci when cultured at 14°C overnight (Figs S4B,C). Since Glc7 function and localization are perturbed in the uba1-W928R mutant, some aspect of Glc7 function is likely to be dependent on ubiquitin.

The sensitivity of uba1-W928R strains to radiation and their defect in the N-end rule degradation pathway (Fig. 2F,G) suggest that the E2 Rad6 is not properly thioesterified to ubiquitin in uba1-W928R cells. To identify additional E2 ubiquitin-conjugating enzyme(s) that may be impacted by uba1-W928R, we tested mutations in nine of the eleven E2 ubiquitin-conjugating enzymes for traits similar to those found in uba1-W928R strains. Gene deletions were used for the non-essential E2 enzymes Rad6, Ubc4, Ubc5, Ubc7, Ubc8, Pex4, Ubc11 and Ubc13. The polyubiquitylation-defective cdc34^tm mutant (Cocklin et al., 2011; Lass et al., 2011) was used for the essential E2 enzyme Cdc34 (which associates with SCF E3 ligases). Each of these mutations was combined with either ipl1-2 or glc7-127 and assayed for suppression of temperature sensitivity or a synthetic growth phenotype, respectively.

The majority of the E2 deletion mutations do not suppress ipl1-2 (Fig. S5A), nor do they confer a slow growth phenotype with glc7-127 (Fig. S5B). Only two E2 enzyme mutations, cdc34^tm and ubc4Δ, suppress ipl1-2 (Fig. 6A). We observed very weak ipl1-2 suppression by rad6Δ (previously reported in Latham et al., 2011). Interestingly, uba1-W928R is a better ipl1-2 suppressor than either cdc34^tm or ubc4Δ (Fig. 6A), suggesting that suppression of ipl1-2 by uba1-W928R is conferred by defects in more than one E2 enzyme. Similarly, only rad6Δ and cdc34^tm confer a slow growth phenotype with glc7-127 (Fig. 6B). While the cdc34^tm mutation suppresses ipl1-2 and also displays genetic interaction with glc7-127, the ubc4Δ mutant suppresses ipl1-2 but does not show any additive growth defects with glc7-127 (Fig. S5B). Conversely, the rad6Δ mutation exhibits strong growth defects in combination with glc7-127, but is a very poor ipl1-2 suppressor. Thus, our genetic evidence shows that perturbations in the activities of particular E2 enzymes lead to distinct effects on Ipl1 and Glc7 function, suggesting the existence of multiple independent ubiquitin-dependent pathways that influence Ipl1 and Glc7 activities.

uba1-W928R confers several distinct growth traits, including sensitivity to low and high temperatures, caffeine, CsCl, UV radiation, and resistance to LiCl (Fig. 1). Of the E2 enzyme mutants tested, only rad6Δ and cdc34^tm confer similar phenotypes to uba1-W928R (Fig. S6A). Moreover, as shown in Fig. 6A,B, genetic interaction of rad6Δ and cdc34^tm with glc7-127 and the ipl1-2 suppression by cdc34^tm resemble those caused by uba1-W928R. Since rad6Δ and cdc34^tm mutations result in multiple growth defects similar to those observed for uba1-W928R, it is plausible that such traits of uba1-W928R may be due to alteration of interactions with Rad6 and/or Cdc34. As shown in Fig. 6B, the single rad6Δ and cdc34^tm mutations confer slow growth with glc7-127, but this genetic interaction is not as strong as the synthetic lethality of uba1-W928R and glc7-127. We reasoned that if uba1-W928R fails to interact normally with both rad6Δ and cdc34^tm, then the rad6Δ cdc34^tm double mutant should show synthetic lethality with glc7-127. To test this, we assayed growth of meiotic progeny from diploid strains heterozygous for rad6Δ, cdc34^tm and glc7-127. The rad6Δ cdc34^tm glc7-127 triple mutants grow very slowly (Fig. 7A). rad6Δ cdc34^tm mutants also give rise to small colonies, but these are significantly larger than those formed by the rad6Δ cdc34^tm glc7-127 triple mutants (Fig. 7B). These results suggest that Rad6 and Cdc34 could be the key E2 enzymes acting downstream of Uba1 to affect Glc7 function.

Another trait conferred by uba1-W928R is the accumulation of bright Glc7 nuclear foci (Fig. 5F). To determine whether rad6Δ or cdc34^tm mutants also influence Glc7 localization, we examined Glc7–mCitrine fluorescence in cdc34^tm and rad6Δ strains. Bright foci of Glc7–mCitrine were observed in 68% of the cdc34^tm cells (Fig. 7C). The foci are located close to or overlapping the area of nucleolar Nop65–mRFP fluorescence, as observed in uba1-W928R mutant cells. We note that the uba1-W928R cdc34^tm double mutant does not show enhancement of these defects (Fig. S6B). Glc7 foci are also present in cells lacking Rad6 or containing an alteration in its E3-binding partner, Bre1 (Fig. 7C). Taken together, these results indicate that Cdc34, Rad6 and Bre1 have previously unknown roles in regulating Glc7 localization and function. Suppression of ipl1-2 by rad6Δ and bre1Δ (Latham et al., 2011) was thought to occur by indirectly affecting the kinetochore protein Dam1. Our results suggest that Rad6 and Bre1 also regulate Glc7, which could contribute to ipl1-2 suppression.

We report here the identification and characterization of a novel UBA1 allele, uba1-W928R, that suppresses ipl1-2 defects. Ubiquitylation affects many pathways whose perturbation could result in suppression of ipl1-2. However, our observation that the uba1-W928R mutation partially restores histone H3-S10 phosphorylation in the ipl1-2 mutant (Fig. 4A) indicates that the Ipl1–Glc7 kinase–phosphatase balance is altered. Since either an increase in Ipl1 activity or a decrease in Glc7 activity would result in both the restoration of HH3-S10 phosphorylation and suppression of ipl1-2 growth defects, we were surprised to find that both Ipl1 and Glc7 are affected by the uba1-W928R mutation.

The uba1-W928R mutation increases the stability of WT Ipl1 and Ipl1-2 protein (Fig. 4B,D). Since increased expression of the Ipl1-2 protein restores temperature-independent growth to the ipl1-2 strain, we conclude that the increase in Ipl1-2 protein levels due to uba1-W928R accounts for at least part of the suppression mechanism. This implies that one or more components of the CPC is ubiquitylated, and the simplest explanation is that Ipl1 is ubiquitylated and targeted for degradation. In metazoans, ubiquitylation regulates Aurora B by both proteasome- and non-proteasome-dependent mechanisms. A Cul3-based E3 ligase ubiquitylates Aurora B and targets it to the spindle midzone, which is necessary for coordinating faithful chromosome segregation and completion of cytokinesis (Sumara et al., 2007; Maerki et al., 2009). However, there is also evidence that ubiquitylation of Aurora B by the E3 ligase BARD1 targets it to the proteasome for degradation (Ryser et al., 2009). Aurora B is also known to bind to the APC/C subunits Cdh1 and Cdc20 in vivo, suggesting that Aurora B is degraded by the proteasome via the APC/C (Nguyen et al., 2005). Surprisingly, to date there has been no evidence for ubiquitin-dependent Ipl1 regulation in S. cerevisiae. Our work suggests that the E2 enzymes Cdc34 and Ubc4 are candidates to act downstream of uba1-W928R to regulate Ipl1, since cdc34^tm and ubc4Δ each suppresses ipl1-2. Cdc34 is the main, if not only, E2 enzyme partner of SCF ubiquitin–protein ligase complexes (Petroski and Deshaies, 2005). It colocalizes with β-tubulin on mitotic spindles during anaphase in mammalian cells and has well-defined roles in regulating cell cycle progression (Feldman et al., 1997; Kus, et al., 2004). Ubc4 is the key E2 that partners with Ubc1 to polyubiquitylate key cell cycle regulators targeted by the APC/C (Rodrigo-Brenni and Morgan, 2007). Moreover, Ubc4 has been reported to play an essential role in degradation of the mitotic cyclin Cdc13 in Xenopus and clam oocytes (Seino et al., 2003). Since uba1-W928R is a stronger ipl1-2 suppressor than either of the single E2 mutants, multiple pathways may independently regulate Ipl1, similar to ubiquitylation of Aurora B in metazoans.

The uba1-W928R mutation also alters Glc7 localization, and cells containing this mutation in combination with one of several glc7 mutations are very slow-growing or inviable. Since we ruled out the possibility that the genetic interactions are due to increased Ipl1 activity in the context of reduced phosphatase activity, we favor the possibility that ubiquitylation plays a positive role in Glc7 function. One potential role for ubiquitylation is in the context of Cdc48-dependent maturation of Glc7. Cheng and Chen (Cheng and Chen, 2015) showed that newly synthesized Glc7 aggregates in temperature-sensitive cdc48-3 mutants at the restrictive temperature and in cells depleted of the Cdc48 adaptor protein Shp1. They also showed that Cdc48–Shp1 interacts transiently with Glc7, Sds22 and Ypi1, which together may be required to prevent misfolding of Glc7. Since substrates for the Cdc48 segregase are usually ubiquitylated, it is possible that ubiquitylation of Glc7, Sds22 and/or Ypi1 is necessary for the binding of Cdc48, which then acts as a foldase to prevent Glc7 misfolding. Interestingly, the PP1-related phosphatases Ppz1 and Ppz2 also interact transiently with Cdc48–Shp1 and require Cdc48–Shp1 chaperone function to prevent misfolding (Cheng and Chen, 2015). We note that ppz1Δ ppz2Δ double mutants are synthetic lethal with uba1-W928R (Fig. S7A), suggesting that uba1-W928R causes impaired Cdc48–Shp1 binding to multiple phosphatase substrates.

Although neither we nor Cheng and Chen (Cheng and Chen, 2015) were able to detect ubiquitylated forms of Glc7 by immunoblotting, both yeast Glc7 and human PP1 have been reported to be ubiquitylated at multiple lysine residues in proteomic screens for ubiquitylated proteins (Fig. S7C) (Wagner et al., 2011; Starita et al., 2012; Swaney et al., 2013). Also, the direct Cdc48–Shp1 substrate need not be Glc7. In global screens, the Glc7 regulators Ypi1, Shp1 and Cdc48 have been reported to be ubiquitylated in vivo (Swaney et al., 2013), but the function of this modification has not been investigated for any of these proteins. Sds22 protein levels are twice as high in uba-W928R mutants as in WT (Fig. S7B), suggesting that stability of Sds22 is determined by ubiquitylation. Therefore, given our results and that multiple components of the Glc7 maturation pathway are known to be ubiquitylated, we propose that ubiquitylation of Glc7 and/or its regulators influences Glc7 localization and activity.

Glc7 aggregation and reduced nuclear localization are hallmarks of mutations affecting genes in the Glc7 maturation pathway. sds22-6, ypi1-GFP, cdc48-3 and shp1^ts mutants exhibit both nuclear and cytoplasmic Glc7 aggregates. However, the Glc7 foci observed in uba1-W928R, rad6Δ, cdc34^tm and bre1Δ cells are nuclear, adjacent to the nucleolar protein Nop56. One of our future goals is to determine whether the aggregates observed in uba1-W928R differ biochemically from those that form due to alteration of Glc7 regulatory proteins. This would inform us about the specific step(s) in Glc7 maturation that require ubiquitylation. It is possible that, while perturbation of the Glc7 regulatory pathway causes misfolding of newly synthesized Glc7 (Cheng and Chen, 2015), ubiquitylation may be required at a later step. Consistent with this, uba1-W928R cells exhibit nuclear Glc7 foci even after cycloheximide treatment (Fig. S7D).

As mentioned above, the E2 conjugating enzymes Rad6 and Cdc34 also influence Glc7. Mutations in rad6 and cdc34 show synthetic growth defects with glc7-127 and cause Glc7 foci formation (Figs 6B and 7B). Although the principal function of Cdc34 is to regulate protein levels of SCF target substrates, it is also known to play a role in nuclear protein quality control, along with the San1 E3 ligase (Gardner et al., 2005). The bright Glc7 nuclear foci observed in the cdc34^tm mutant suggest that Cdc34 may be required for clearing misfolded nuclear Glc7, so that in its absence, Glc7 accumulates and forms foci. There has been no evidence to date for a role of Rad6 in protein quality control. Rad6 is known to partner with three RING E3 ligases, Rad18, Ubr1 and Bre1. The Rad6–Ubr1 pair polyubiquitylates N-end rule substrates (Dohmen et al., 1991) and, since we do not observe changes in Glc7 protein levels in the uba1-W928R mutant, it is unlikely that this pathway regulates Glc7 function. Along with Rad18, Rad6 specifically monoubiquitylates K164 on proliferating cell nuclear antigen (PCNA) to mediate repair of post-replicative DNA damage (Hoege et al., 2002). The Rad6–Bre1 pair monoubiquitylates histone H2B on K123 (equivalent to mammalian H2B K120), which in turn regulates histone H3K4 methylation and gene silencing (Wood et al., 2003). Interestingly, this H2B modification also affects mitosis, by indirectly reducing Ipl1-mediated phosphorylation of the kinetochore substrate Dam1 (Latham et al., 2011). Our data suggest a new role for Rad6–Bre1 in maintaining Glc7 localization. Both rad6Δ and bre1Δ accumulate Glc7 foci and show genetic interaction with the glc7-127 mutant allele (Fig. 7C; Fig. S7E). Since the only known substrate of Bre1 (human RNF20 and RNF40) to date is histone H2B K123, this modification may indirectly affect Glc7 localization. Alternatively, Bre1 may have unidentified substrates beyond histone H2B (previously proposed in Turco et al., 2015).

In summary, our new UBA1 allele illuminates two new mechanisms for ubiquitin-dependent regulation of the Ipl1–Glc7 balance in budding yeast, by affecting Ipl1 protein stability, and Glc7 localization and activity. Since yeast Glc7 is over 80% identical to mammalian PP1 and the essential function of Glc7 can be complemented by human PP1 isoforms (Gibbons et al., 2007), knowledge of ubiquitin-mediated regulation of Glc7 could provide novel mechanisms for modulation of PP1 in metazoans.

The yeast strains used in this study are listed in Table S3 and are congenic to strain JC482 (Cannon and Tatchell, 1987). Yeast strains were grown on medium containing 1% yeast extract, 2% peptone, and 2% glucose (YPD) or synthetic complete (SC) medium (Sherman et al., 1986) or modifications of these media as stated. Filter sterilized 0.25M caffeine stock in H₂O (Sigma 68F-0817) was added to autoclaved rich medium and used for serial dilution tests. Strains were sporulated at 24°C on medium containing 0.1% yeast extract, 0.05% glucose and 1% potassium acetate. Yeast transformation, manipulation of Escherichia coli, and the preparation of bacterial growth medium were performed as described previously (Sherman et al., 1986). Tetrad dissection and analysis were performed as described previously (Rose et al., 1990). The Myc-tagged variants of ipl1-2 and SDS22 were generated by the method of Longtine (Longtine et al., 1998) with primers listed in Table S4. PCR products with the tagged alleles were introduced into the WT diploid strain KT1112/KT1113 by transformation. G418-resistant transformants were sporulated, and protein extracts from G418-resistant meiotic progeny were assayed for the presence of the tagged protein by immunoblotting. Strains containing deletions of E2 enzyme genes were made by using PCR to amplify deletion cassettes from genomic DNA of appropriate strains from the yeast deletion collection (Winzeler et al., 1999) and transforming the WT KT1112/KT1113 strain with the PCR products. Primers used are listed in Table S4. Chromosome III loss in diploids and segregation of GFP-tagged chromosome IV were assayed as described previously (Tatchell et al., 2011). The rev100 mutant was isolated as elaborated previously (Tatchell et al., 2011; Robinson et al., 2012). To assay the sensitivity of yeast cells to UV light, cultures were grown to logarithmic phase at 24°C in YPD medium, plated onto YPD plates, exposed to UV irradiation in a CL-1000 Ultraviolet Crosslinker (UVP) and then incubated for 4 days at 24°C before counting colonies.

The E. coli strain DH5α was used to amplify all plasmids. DH5α was grown in LB medium supplemented with ampicillin to select for plasmids. Transformation of chemically competent bacterial cells was as described previously (Sambrook et al., 1989). Restriction enzymes (Promega), DNA ligase (NEB), and high-fidelity DNA polymerase for PCR (Bio-Rad and Phenix) were used according to the manufacturer's instructions. Yeast genomic DNA was isolated using the YeastStar kit (Zymo Research), and yeast plasmid DNA was isolated using the Zymoprep kit (Zymo Research). Plasmids were purified from E. coli using the GenElute miniprep kit (Sigma-Aldrich). DNA restriction fragments were purified from agarose gel slices using the ZymoClean kit (Zymo Research).

To identify the mutation responsible for suppression and slow growth, we carried out serial backcrosses of a rev100 mutant strain to our unmutagenized parent strain and prepared genomic DNA from six MATα rev100 ascosporal clones from the fourth backcross. Whole-genome sequence analysis on the pooled DNA was performed using the Illumina MiSeq. DNA libraries were prepared from 1 ng of genomic DNA using the Illumina Nextera XT kit (Illumina, Inc., San Diego, CA) according to the manufacturer's protocol. In brief, DNA was simultaneously fragmented and tagged with unique adapter sequences in a tagmentation reaction. A limited-cycle PCR reaction was used to amplify the insert DNA and add index sequences to both ends of the DNA. Libraries were analyzed on an Agilent Bioanalyzer 2100 HS DNA assay (Agilent Technologies, Santa Clara, CA) to determine average size. Libraries were normalized, pooled and denatured. A 1.0% library of 12.5 pM PhiX was spiked in as an internal control. The library pool was sequenced on an Illumina MiSeq, with a read length of 2×250 base pairs. Base calling and quality scoring were performed with Illumina Real Time Analysis software (RTA). Illumina MiSeq Reporter software was used to demultiplex reads, generate FASTQ files and align indexed reads to S. cerevisiae (saCer2). Alignment was performed using the Burrows–Wheeler aligner (BWA; https://arxiv.org/abs/1303.3997v2). Variants were then called in MiSeq Reporter using the Genome Analysis Toolkit (GATK).

Live yeast cells from log phase cultures were placed on a 2% agarose pad and imaged using a CoolSNAP HQ charge-coupled device camera through an Olympus UPlanFl 100×1.3 NA objective. Filters from Chroma Technology were used to image cells expressing proteins tagged with mCitrine, mCherry and mRFP. SlideBook 6 software (Intelligent Imaging Innovations), was used to control camera acquisition, fluorescence filter wheels, and the z-axis stepping motor (Ludl Electronic Products). Fluorescence images (binned 2×2) were acquired in a series of z-axis planes (0.5 μm apart). For measuring fluorescence intensity, images in nine planes were converted into a z-axis projection and the maximum fluorescence in nine adjacent pixels was measured using ImageJ software. Distances between spindle pole bodies, lengths of spindles and diameter of spore clones were measured using the ImageJ software. For cell cycle analysis, overnight cultures were diluted 1:20 and grown to log phase at 24°C. Cultures were synchronized by addition of α-factor (0.1 mg/ml final concentration) and incubated for a further 2 h. Cells were collected by centrifugation at 200 g, washed twice by centrifugation, and resuspended in YPD containing nocodazole (15 µg/ml final concentration) or DMSO. Aliquots were taken every 30 min for western blotting and every 20 min for microscopy. To study spindle dynamics, fluorescence microscopy of WT and uba1-W928R cells expressing GFP-tagged Spc42 was performed. Cells were synchronized using α-factor, released in fresh medium and aliquots were taken every 20 min for microscopy. Time-lapse microscopy to quantify spindle elongation rate was performed by imaging synchronized cells expressing Tub1–GFP every 3 min, and spindle lengths were measured using ImageJ.

Total protein extracts were prepared from yeast cultures in log phase grown at 24°C and shifted to restrictive temperature when indicated. For drug treatments, log phase cultures were split in half, followed by addition of DMSO/autoclaved water or the indicated concentration of drug [MG132 (75 µM), cycloheximide (200 µg/ml)] to the culture. Cells were collected and lysed in tricholoracetic acid using glass beads (Davis et al., 1993), and total protein extracts were electrophoresed on Tris-glycine gradient gels (Criterion; Bio-Rad Any kDa or 4–20%) or 12% SDS-PAGE gels. Gels were blotted onto nitrocellulose (Amersham Protran) or PVDF (Bio-Rad) for H3-S10 detection. Immunoblot analysis was performed as described previously (Stuart et al., 1994). The following antibodies were used: anti-ubiquitin (1:1000, FK2, SMC-550 StressMarq Biosciences), anti-Myc (1:1000, 9E10; Evan et al. 1985), anti-GFP (1:2000, JL-8; Kozubowski et al. 2003), anti-HA (1:1000, 12CA5; Wilson et al. 1984), anti-Histone H3 (1:1000, ab1791, Abcam), anti-phospho-H3-S10 (1:1000, 06-570, EMD Millipore) with horseradish peroxidase (HRP)-conjugated secondary antibody (Bio-Rad 170-5046 and 170-5047) and subsequent detection using ECL reagents (Clarity Western Bio-Rad). The anti-Glc7 goat serum was a gift from Andreas Mayer, University of Lausanne, Switzerland. Protein levels were quantified from immunoblots using the ChemiDoc Touch imaging system with Image Lab 6.0 software (Bio-Rad). Anti-phosphoglycerate kinase (Pgk1) (Life Technologies) was used as a loading control. H3-Ser10 phosphorylation was detected as described in Tatchell et al. (2011).

We thank Alan Tartakoff (Case Western Reserve University) for the NOP56-mRFP strain, John Hanna (Harvard Medical School) for the URA3 reporter-degron strain, Mark Goebl (Indiana University School of Medicine) for the cdc34^tm strain, Raymond Deshaies (Caltech) for the uba1-204 strain and the UBA1-pRS306 plasmid and Sue Biggins (Fred Hutchinson Cancer Research Center) for the IPL1-pGAL plasmid.