The Mub1/Ubr2 Ubiquitin Ligase Complex Regulates the Conserved Dsn1 Kinetochore Protein

The kinetochore is the macromolecular complex that assembles onto centromeric DNA and orchestrates the segregation of duplicated chromosomes. More than 60 components make up the budding yeast kinetochore, including inner kinetochore proteins that bind to centromeric chromatin and outer proteins that directly interact with microtubules. However, little is known about how these components assemble into a functional kinetochore and whether there are quality control mechanisms that monitor kinetochore integrity. We previously developed a method to isolate kinetochore particles via purification of the conserved Dsn1 kinetochore protein. We find that the Mub1/Ubr2 ubiquitin ligase complex associates with kinetochore particles through the CENP-CMif2 protein. Although Mub1/Ubr2 are not stable kinetochore components in vivo, they regulate the levels of the conserved outer kinetochore protein Dsn1 via ubiquitylation. Strikingly, a deletion of Mub1/Ubr2 restores the levels and viability of a mutant Dsn1 protein, reminiscent of quality control systems that target aberrant proteins for degradation. Consistent with this, Mub1/Ubr2 help to maintain viability when kinetochores are defective. Together, our data identify a previously unknown regulatory mechanism for the conserved Dsn1 kinetochore protein. We propose that Mub1/Ubr2 are part of a quality control system that monitors kinetochore integrity, thus ensuring genomic stability.


Introduction
Accurate chromosome segregation is essential to avoid the aneuploidy that is associated with cancer and birth defects [1]. Segregation is directed by the kinetochore, the macromolecular protein complex that assembles onto the centromeric region of each chromosome and that interacts with spindle microtubules during mitosis and meiosis [2][3][4][5]. The inner part of the kinetochore binds to centromeric DNA, whereas the outer portion interacts with microtubules. Although greater than 60 kinetochore components have been identified in the budding yeast kinetochore, it remains unclear how the individual proteins assemble onto the centromere to form the macromolecular kinetochore structure [6].
At the base of the kinetochore, the CENP-A centromeric histone H3 variant forms a specific chromatin environment essential for recruiting other kinetochore proteins [7,8]. Components of the CCAN (constitutive centromere-associated network, e.g. CENP-C) closely associate with CENP-A [9,10]. These inner kinetochore components are essential for the assembly of the outer kinetochore [11][12][13]. The outer kinetochore possesses microtubule-binding activity mediated through the KNL1 and Ndc80 complexes in the KMN (KNL1, Mis12, Ndc80 complexes) network [14]. Although the Mis12 complex does not directly bind to microtubules, it is important for the assembly of the KMN and may be a keystone to promote outer kinetochore assembly [14][15][16][17][18][19][20]. Recently, the conserved centromere-binding protein CENP-C Mif2 has been shown to link the inner and outer portions of the kinetochore by binding directly to the Mis12 complex [21,22].
Ubiquitin (Ub)-mediated proteolysis is a widely used cellular system to monitor the quality and quantity of numerous proteins [23,24]. Ubiquitylation of a substrate requires multiple enzymes. Ubiquitin is first activated by an Ub-activating enzyme (E1), transferred to an Ub-conjugating enzyme (E2), and then conjugated to a substrate via an E3 ligase. Although the E3 enzymes largely dictate substrate specificity, efficient ubiquitylation of target proteins requires additional adaptor or cofactor proteins in some cases (e.g. [25,26]). Proteolysis of kinetochore proteins appears to be important to ensure genomic stability. In budding yeast and flies, the CENP-A protein that is the foundation of centromeric chromatin and directs kinetochore formation is degraded to ensure its exclusive localization at the centromere [27,28]. In addition, ubiquitin-mediated degradation of components of the CBF3 and Mis12 complexes regulates kinetochore assembly and function [18,29,30].
Ubr2 was identified as an E3 ligase that regulates the level of the transcription factor Rpn4 through ubiquitylation [31]. Ubiquitylation of Rpn4 by Ubr2 requires an additional factor Mub1 [32], so we will refer to the complex as Mub1/Ubr2 throughout this paper. A recent study showed that a ribonucleotide reductase inhibitor, Sml1, is also targeted by Mub1/Ubr2 [33]. Although mub1D and ubr2D mutants are viable, ubr2D mutant cells are sensitive to increased proteasome activity [31] and defective in the degradation of unfolded cytoplasmic protein [34]. It is not clear whether the Mub1/Ubr2 ligase complex targets additional proteins or cellular processes.
Here, we used a recently developed kinetochore purification method to show that the Mub1/Ubr2 E3 ligase complex interacts with kinetochore particles through the conserved CENP-C Mif2 protein that links inner and outer kinetochore proteins [15]. We found that Mub1/Ubr2 mediate Dsn1 ubiquitylation and regulate its protein levels, especially when the Dsn1 protein is mutated. In addition, Mub1/Ubr2 become important for viability when kinetochore function is compromised. Taken together, these data suggest that Mub1/Ubr2 regulate Dsn1 levels to ensure kinetochore integrity.

Results
Kinetochore particles associate with the E3 ubiquitin ligase Ubr2 through Mub1 We previously developed a technique to isolate functional kinetochore particles via the purification of the Dsn1 kinetochore protein [15]. Although the particles contain components spanning the inner to outer kinetochore, they lack the centromere-binding complex CBF3 and some other inner kinetochore proteins. These data suggest that the particles are not bound to centromeric DNA after purification. Most of the proteins in the preparations that are visible by silver-stained SDS-PAGE are kinetochore components. However, a ,190 kDa band that did not correspond to any candidate kinetochore proteins consistently co-purified with the particles ( Figure 1A). Notably, mass spectrometry (MS) analysis of purified particles from asynchronously growing cells identified the E3 ubiquitin ligase Ubr2 and its cofactor protein Mub1 that had not previously been implicated in kinetochore function [15]. Because the predicted molecular weight of Ubr2 is 216 kDa, we tested whether the unknown candidate is Ubr2 by purifying kinetochore particles from cells expressing GFP-Ubr2. The band shifted upward when kinetochores were analyzed by silver-stained SDS-PAGE, confirming its identity as Ubr2 ( Figure 1A). The Mub1 protein overlaps with non-specifically co-purifying proteins that migrate around 70 kDa (indicated by * in Figure 1A), so we performed an immunoblot to confirm that Mub1 also specifically binds to Dsn1-derived kinetochore particles ( Figure 1B). The association between Ubr2 and Dsn1 was abolished in the absence of Mub1 ( Figure 1C), suggesting that Mub1 acts as an adaptor protein that recruits Ubr2 onto kinetochore particles. Although this result appears to contrast previous studies that showed that Mub1 is not required for Ubr2 to interact with the Rpn4 protein [26], the interaction with Rpn4 was only tested in the presence of high Ubr2 levels.

Author Summary
The flawless execution of cell division is essential to the survival of all organisms. The loss or gain of a single chromosome, the state called aneuploidy, is a hallmark of cancer cells and is the leading cause of spontaneous miscarriages and hereditary birth defects. Segregation is mediated by the kinetochore, the macromolecular complex that assembles on each chromosome and attaches to spindle microtubules to pull chromosomes to opposite poles when cells divide. It is therefore critical to understand how kinetochores are assembled and maintained. Here, we find that the levels of a conserved kinetochore protein are regulated by proteolysis. We propose that cells have quality control systems that ensure kinetochore integrity and thus genome stability.
A silver-stained SDS-PAGE gel of purified CENP-C Mif2 revealed a band that migrates around 190 kDa in addition to the Mif2-Flag band ( Figure 2C). MS analysis of the purified sample showed that Mub1 and Ubr2 are among the most abundant proteins in the sample, suggesting that the 190 kDa band is most likely Ubr2 ( Figure 2D and Table S1). Conversely, when Mub1-Flag was purified, several kinetochore proteins including CENP-C Mif2 were detected ( Figure S2 and Table S2). Our purifications of other kinetochore proteins (e.g. Nuf2, KNL-1 Spc105 , CENP-A Cse4 , Fin1) did not result in significant enrichment of the Mub1/Ubr2 proteins [15,35], suggesting that the association of Mub1/Ubr2 depends on a unique feature of CENP-C Mif2 . The interaction between Mif2 and Ubr2 requires Mub1 ( Figure 2E), whereas the interaction between CENP-C Mif2 and Mub1 is not dependent on Ubr2 ( Figure 2F). These data show that Mub1 requires CENP-C Mif2 to recruit the Ubr2 ligase onto kinetochore particles.  Table S1 for all proteins identified by MS. (E) CENP-C Mif2 association with Ubr2 requires Mub1. CENP-C Mif2 -Flag was immunoprecipitated from cells in the presence (SBY8519) or absence (SBY8911) of MUB1 and analyzed via SDS-PAGE and silver staining. (F) The association between CENP-C Mif2 and Mub1 does not require Ubr2. CENP-C Mif2 -Flag was immunoprecipitated from cells containing Mub1-Myc in the presence (SBY8546) or absence (SBY8572) of UBR2 and analyzed via immunoblot. The Mub1 protein level in the input is higher in ubr2D due to the lack of Ubr2-dependent proteolysis [26]. Note that the CENP-C Mif2 -Flag band overlaps with a background signal in the input. doi:10.1371/journal.pgen.1003216.g002

Mub1/Ubr2 mediate Dsn1 ubiquitylation and degradation
Our identification of a ubiquitin ligase that associates with kinetochore particles suggests there might be a relevant kinetochore target. Although CENP-C protein is ubiquitylated by the viral immediate-early protein ICP0 upon Herpes Simplex Virus infection [36], we found no evidence that the wild-type budding yeast CENP-C Mif2 protein is ubiquitylated in vivo despite its close association with the Mub1/Ubr2 ligase complex (data not shown). We therefore considered Dsn1 as a potential target because we isolated an unstable mutant while studying its phosphoregulation (manuscript in preparation). Dsn1 contains two conserved Aurora B kinase consensus sites (S240 and S250) [19,37,38] that we mutated to analyze the corresponding phenotypes. When both sites were mutated to alanine to block phosphorylation (dsn1-S240A,S250A), the cells were inviable ( Figure 3A). In contrast, the phospho-mimic mutant (dsn1-S240D,S250D) is viable. To analyze the corresponding protein levels, myc-tagged Dsn1 phosphomutants were expressed from the endogenous DSN1 promoter in the presence of wild-type DSN1 to keep the cells alive. Although the protein levels of wild-type Dsn1 and the phospho-mimic mutant were similar, there were lower levels of the Dsn1-S240A,S250A mutant ( Figure 3B). To test whether the inviability of dsn1-S240A,S250A is a result of low protein levels, we expressed dsn1-S240A,S250A from a high copy plasmid, which led to much higher protein levels ( Figure 3C). The overexpressed dsn1-S240A,S250A mutant complements dsn1D ( Figure 3D), supporting the idea that the inviability is at least partially due to reduced Dsn1 protein levels.
We asked whether Dsn1 is a target of Mub1/Ubr2 by testing whether a deletion of MUB1 or UBR2 could stabilize the Dsn1-S240A,S250A mutant protein. Strikingly, Dsn1-S240A,S250A protein levels were restored to near wild-type in both mub1D and ubr2D strains ( Figure 4A). The levels of WT Dsn1 were also increased in the mub1D and ubr2D strains, suggesting that Mub1/ Ubr2 mediate the degradation of WT Dsn1 protein. We therefore analyzed Dsn1 stability by adding cycloheximide to repress translation. The Dsn1 protein levels were higher in the mub1D strain at the start of the experiment and did not decrease as much as in WT cells ( Figure 4B), although there is still some degradation in the absence of Mub1/Ubr2. Because Mub1/Ubr2 target the Dsn1-S240A,S250A protein that lacks Aurora B phosphorylation sites, we also analyzed Dsn1 stability in a budding yeast Aurora B mutant, ipl1-321. We found that Dsn1 is somewhat less stable in an ipl1-321 mutant relative to WT cells and that the degradation is at least partially dependent on Mub1/Ubr2 ( Figure 4C and 4D). These data are consistent with the lack of phosphorylation on residues 240 and 250 leading to the degradation of Dsn1.
We next tested whether Mub1/Ubr2 directly mediate the ubiquitylation of Dsn1. When purified from a cim3-1 proteasome mutant, upper forms of WT Dsn1 as well as the Dsn1-S240A,S250A mutant protein that are characteristic of ubiquitylation were apparent on an immunoblot ( Figure 4E). Indeed, upper bands specific to a Dsn1 immunoprecipitation were recognized by anti-ubiquitin antibodies and were more abundant in cim3-1 mutant cells. Importantly, the ubiquitylation of Dsn1 was dependent on the Ubr2 E3 ligase ( Figure 4E). Taken together, these data strongly suggest that the Mub1/Ubr2 E3 ligase targets the Dsn1 kinetochore protein for degradation.
Our finding that Mub1/Ubr2 regulate Dsn1 protein levels led us to test whether mub1D and ubr2D deletion mutants suppress the lethality of dsn1-S240A,S250A. As predicted, there was significant suppression by deletion of either Mub1 or Ubr2 ( Figure 4F). The mub1D ubr2D double mutant suppressed to a similar extent as the  individual mutants, consistent with Mub1 and Ubr2 functioning in a complex to control Dsn1 protein levels. It was previously found that mub1D and ubr2D mutants exhibit increased proteasome activity because the Rpn4 transcription factor is not degraded [26]. As expected, suppression of the dsn1-S240A,S250A lethality is not due to increased proteasome activity, because suppression occurred even in the mub1D ubr2D rpn4D triple mutant. Taken together, these results show that the dsn1-S240A,S250A mutant is lethal because it has reduced protein levels due to Mub1/Ubr2dependent proteolysis. Our kinetochore particles also interact with another E3 ligase, Psh1 [15], but its deletion does not suppress dsn1-S240A,S250A lethality (data not shown), consistent with its specific regulation of CENP-A [39].

Mub1/Ubr2 are important when kinetochore function is compromised
The regulation of Dsn1 by Mub1/Ubr2 prompted us to test whether they have a role in chromosome segregation. Because Mub1 mediates the Mub1/Ubr2 interaction with kinetochores, we analyzed a mub1D strain. However, chromosome biorientation in metaphase cells (as judged by the characteristic bi-lobed formation of the Mtw1 kinetochore protein [40], Figure 5A) and chromosome segregation (monitored by segregation of a pair of marked sister chromatids [41]) (data not shown) appeared normal in mub1D mutant cells. Interestingly, we failed to identify Mub1/ Ubr2 by MS or immunoblot when we purified centromeric minichromosomes [35], suggesting they may not be core kinetochore components. To determine whether Mub1/Ubr2 localize to endogenous kinetochores, we performed chromatin immunoprecipitation ( Figure 5B) and chromosome spread experiments (data not shown). We did not detect an association of Mub1/Ubr2 with endogenous kinetochores, indicating that Mub1/Ubr2 are not structural components of kinetochores. However, they may transiently associate with kinetochores.
Because we did not detect a major defect in segregation in the absence of Mub1/Ubr2, we considered the possibility that Mub1/ Ubr2 becomes important when kinetochore function is compromised. We analyzed the ndc10-1 temperature sensitive mutant that is completely defective in kinetochore assembly [42] as well as the ndc80-1 mutant that assembles a kinetochore with defects specifically outer kinetochore function [43]. Strikingly, deletion of Mub1/Ubr2 decreased the semi-permissive temperature of these mutants, indicating that their function is important for viability when kinetochores are defective ( Figure 5C and 5D). To determine whether the growth defect in ndc80-1 mub1D strains is due to defective kinetochore function, we monitored kinetochore clustering in the double mutant. Although kinetochores in wildtype yeast cells cluster into two distinct foci at metaphase [40,44,45], cells with defects in microtubule-kinetochore attachments exhibit three or more declustered foci [46]. We visualized kinetochores using GFP epitope-tagged Mtw1 kinetochore protein in ndc80-1 and mub1D ndc80-1 mutant strains. Cells were released from G1 into the cell cycle at 30 degrees and monitored for kinetochore foci after 140 min ( Figure 5E). When the number of kinetochore foci were quantified, 45% of the ndc80 mutant cells exhibited declustered kinetochores (greater than 3 foci) compared to 66% in the ndc80 mub1 cells. These data are consistent with greater than 75% of the cells being arrested in metaphase with high Pds1 levels due to spindle checkpoint activation. Taken together, these data suggest that kinetochore integrity is further compromised by the lack of Mub1 in the ndc80-1 mutant cells.

Discussion
Here, we show that the Mub1/Ubr2 E3 ligase complex regulates the levels of the conserved Dsn1 kinetochore protein via ubiquitin-mediated proteolysis. Mub1/Ubr2 have previously been implicated in proteasome function through the regulation of Rpn4 [26,31], as well as regulation of dNTP levels through Sml1 degradation [33]. Our data suggest that Mub1/Ubr2 also regulate kinetochore function through proteolysis of the Dsn1 protein.
Although Mub1/Ubr2 requires the Rad6 E2 protein to target Rpn4 and Sml1 [26,33], a rad6D does not suppress the lethality of dsn1-S240A,S250A (data not shown), suggesting that Mub1/Ubr2 utilizes a different or additional E2 to target the Dsn1 kinetochore protein. Identification of the E2 that facilitates the ubiquitylation of Dsn1 should shed additional light on the mechanism by which Mub1/Ubr2 target Dsn1 for degradation.
Our data suggest that CENP-C Mif2 is the receptor for Mub1/ Ubr2 on kinetochore particles. Kinetochores purified from mif2 mutants lacked Mub1/Ubr2, and the purification of CENP-C Mif2 , but not other kinetochore proteins (e.g. Nuf2, KNL-1 Spc105 , CENP-A Cse4 , Fin1), resulted in a signification enrichment of the Mub1/Ubr2 proteins [15,35]. Therefore, a unique feature of CENP-C Mif2 or a CENP-C Mif2 -binding protein is important for the interaction of Mub1/Ubr2 with kinetochore particles. However, despite the close association of Mub1/Ubr2 with CENP-C Mif2 , we did not detect ubiquitylation of CENP-C Mif2 or a CENP-C Mif2 mutant allele ( [47] and data not shown). Because we could not detect a stable association of Mub1/Ubr2 with endogenous kinetochores, Mub1/Ubr2 likely bind to kinetochore particles during the purification process. This raises the possibility that Mub1/Ubr2 only transiently associate with kinetochores to regulate Dsn1, and/or that they regulate a pool of Dsn1 that is not associated with kinetochores.
Although Mub1/Ubr2 are not essential during mitotic growth, a deletion of mub1 or ubr2 can restore both the protein levels and inviability of the dsn1-S240A,S250A mutant. These data indicate that Mub1/Ubr2 have a physiological role in regulating kinetochore function in vivo despite their lack of enrichment at endogenous kinetochores. In addition, deletions in Mub1/Ubr2 exhibited negative genetic interactions with the ndc80-1 and ndc10-1 mutants. In the ndc80 mutant cells, kinetochore-microtubule attachments and biorientation defects were exacerbated by the lack of Mub1. This is reminiscent of the spindle checkpoint proteins that are also not essential in budding yeast but become critical to ensure genomic stability when kinetochore-microtubule interactions are compromised [48,49]. Mub1 and Ubr2 mutants display stronger meiotic phenotypes [50], and we note that the monopolin complex that joins sister kinetochores during meiosis I to ensure that sister chromatids segregate to the same pole associates with kinetochores via its interaction with CENP-C and Dsn1 [51]. It will be interesting to determine if the Mub1/Ubr2 complex regulates monopolin binding to kinetochores in meiosis. In addition, although monopolin components are sequestered in the nucleolus during mitosis, Mub1/Ubr2 may provide a back up system to prevent them from linking sister kinetochores during mitosis. An important future goal will be to understand the cellular location of the Mub1/Ubr2 ubiquitylation of Dsn1.
We speculate that Mub1/Ubr2 monitor kinetochore integrity and only become important for mitosis when kinetochores are defective. We were not able to detect a significant Mub1/Ubr2dependent change in Dsn1 stability in ndc10-1 and ndc80-1 kinetochore mutants (data not shown), possibly because there is only a small pool of Dsn1 that is regulated by Mub1/Ubr2. In mammalian cells, Dsn1 is regulated by the SCF-Skp1 ubiquitin ligase [18], so it will be important to test whether Skp1-mediated degradation also regulates Dsn1 in budding yeast. We also detected ubiquitylation and regulation of WT Dsn1 protein levels by Mub1/Ubr2, although the effect was mild. We presume that WT cells would not have a need for robust regulation of Dsn1 by Mub1/Ubr2. A deletion of Mub1/Ubr2 led to a strong increase in the levels of the Dsn1-S240A,S250A mutant lacking Aurora B phosphorylation sites. Degradation of mutant alleles is a hallmark of quality control systems. Some temperature sensitive mutants, although possessing normal or near normal activity, are recognized as aberrant and destroyed by quality control mechanisms [52][53][54][55][56]. Inhibition of the degradation suppresses the temperature sensitivity, similar to the suppression of dsn1-S240A,S250A inviability by deletion of Mub1/Ubr2. One possibility is that kinetochore quality control may be needed to ensure that defective kinetochores do not assemble, or that they are turned over in an attempt to assemble functional kinetochores. For example, ndc80-1 and ndc10-1 mutant cells could accumulate aberrant kinetochores in mub1D and ubr2D mutants, enhancing their temperature sensitivity. It was recently shown that CENP-C Mif2 mediates the interaction between the centromere and outer kinetochore via its interaction with the Mis12 complex that contains Dsn1 [21,22]. One possibility is that Mub1/Ubr2 binding to CENP-C Mif2 regulates kinetochore assembly by preventing stable association of the outer kinetochore via Dsn1-mediated degradation. Consistent with this model, Dsn1 mutants lacking Aurora B phosphorylation sites in other organisms exhibit defects in kinetochore assembly [57,58]. We speculate that this mechanism would help to avoid the formation of ectopic kinetochores outside of the centromere, as well as play a key role in preventing aberrant kinetochores from stably assembling at centromeres.
There are numerous examples in mammalian studies where mutant kinetochore proteins show decreased protein levels, so it will be important to determine whether a similar quality control system operates in other eukaryotes [16,59]. Mub1 has a conserved MYND domain that is implicated in mediating protein-protein interactions [60], and MYND-containing proteins are implicated in protein ubiquitylation in human cells [61,62]. Importantly, many of the MYND-containing proteins are associated with cancer and other diseases (e.g. ZMYND10, ZMYND11/BS69, ETO (eight-twenty-one)/MTG (myeloid translocation gene) family members, DEAF1, and Suppressin (reviewed in [63]). In addition, the stability of Dsn1 orthologs appears to be carefully regulated by chaperones and proteolysis [16,18], so it will be critical to understand how these MYND proteins function and whether they also regulate kinetochores in mammalian cells. Further studies of potential kinetochore quality control mechanisms could shed light on the process of kinetochore assembly as well as the maintenance of genomic stability in all organisms.