SUMOylation targets shugoshin to stabilize sister kinetochore biorientation

The accurate segregation of chromosomes during mitosis relies on the attachment of sister chromatids to microtubules from opposite poles, called biorientation. Sister chromatid cohesion resists microtubule forces, generating tension which provides the signal that biorientation has occurred. How tension silences the surveillance pathways that prevent cell cycle progression and correct erroneous kinetochore-microtubule remains unclear. Here we identify SUMOylation as a mechanism that promotes anaphase onset upon biorientation. SUMO ligases modify the tension-sensing pericentromere-localized chromatin protein, shugoshin, to stabilize bioriented sister kinetochore-microtubule attachments. In the absence of SUMOylation, Aurora B kinase removal from kinetochores is delayed. Shugoshin SUMOylation prevents its binding to protein phosphatase 2A (PP2A) and release of this interaction is important for stabilizing sister kinetochore biorientation. We propose that SUMOylation modulates the kinase-phosphatase balance within pericentromeres to inactivate the error correction machinery, thereby allowing anaphase entry in response to biorientation.


Introduction 17
Mitosis divides the nucleus to produce two genetically identical daughter cells. Prior 18 to mitosis, DNA replication produces sister chromatids, linked together by the cohesin 19 complex. Sister chromatids are aligned at metaphase, thus allowing microtubule 20 spindles to be captured by kinetochores assembled on centromeres. The correct form 21 of attachment is termed 'biorientation', meaning that the kinetochores on the two 22 sister chromatids are attached to microtubules emanating from opposite spindle poles. 23 Biorientation creates tension, because cohesin holding sister chromatids together 24 resists the pulling force of microtubules [1]. The fulfilment of biorientation allows 25 securin degradation and, consequently, the activation of the protease separase, which 26 cleaves cohesin, triggering sister chromatid separation (reviewed in [2]). 27 The conserved shugoshin protein plays key roles in promoting biorientation in 28 mitosis and preventing cell cycle progression where biorientation fails [3,4]. Budding 29 yeast possesses a single shugoshin gene, SGO1. Sgo1 localizes to both the core 30 ~125bp centromere, where the kinetochore resides, and the surrounding ~20kb 31 cohesin-rich chromosomal region called the pericentromere [5]. The kinetochore-32 localized Bub1 kinase promotes Sgo1 enrichment at the pericentromere through 33 phosphorylation of S121 on histone H2A [6][7][8]. Sgo1, in turn, recruits condensin and 34 protein phosphatase 2A, PP2A-Rts1, to the pericentromere and maintains the 35 chromosome passenger complex (CPC) containing Aurora B kinase at centromeres 36 during mitosis [9,10]. Condensin at pericentromeres is thought to bias the 37 conformation of the sister chromatids to favour biorientation. The CPC recognizes 38 erroneous microtubule-kinetochore attachments and destabilizes them, thereby 39 maintaining the activity of the spindle assembly checkpoint (SAC) to prevent 40 anaphase entry (reviewed in [11]). In vertebrate cells, PP2A-B56 protects cohesin in 41 pericentromeres from removal by the so-called prophase pathway, which removes 42 cohesin through a non-proteolytic mechanism that is independent of separase [12,13]. 43 In budding yeast, PP2A-Rts1 is recruited by shugoshin despite the absence of the 44 prophase pathway [9,10,14]. Instead, PP2A-Rts1 has been implicated in ensuring the 45 equal segregation of sister chromatids during mitosis, since mutants failing to recruit 46 PP2A-Rts1 to the centromere are unable to respond to a lack of inter-sister 47 kinetochore tension and mis-segregate chromosomes upon recovery [9,14]. 48 Sgo1 both directs and responds to cell cycle cues as chromosomes establish 49 and achieve biorientation, upon which anaphase entry is triggered. A key signal that 50

Sumoylation does not promote Sgo1 removal from chromatin under tension 211
Sgo1 is released from pericentromeres under tension [8] but whether this is critical for 212 the metaphase-anaphase transition remained unclear. To address this, we asked if 213 artificial tethering of Sgo1 to the kinetochore can prevent tension-dependent removal: 214 GFP-binding protein (GBP) -tagged Sgo1 was produced from a galactose-inducible 215 promoter (replacing endogenous Sgo1) in cells where the kinetochore protein Mtw1 216 was tagged with GFP (Mtw1-GFP) ( Figure 4D). On its own, pGAL-SGO1-GBP 217 expression caused a modest metaphase delay compared to the Mtw1-GFP control, 218 presumably due to mild overexpression ( Figure 4D). However, Sgo1-GBP expression 219 in a strain producing Mtw1-GFP resulted in a severe delay in metaphase ( Figure 4D). 220 This delay required Sgo1 association with PP2A-Rts1 and/or CPC, because 221 kinetochore tethering of the Sgo1-3A mutant protein, which has lost these 222 interactions, resulted in a more modest metaphase delay ( Figure 4D). Therefore, 223 kinetochore-associated Sgo1 prevents anaphase onset in a manner depending on its 224 ability to bind PP2A-Rts1 and/or CPC, showing that tension-dependent removal of 225 Sgo1 is critical for anaphase entry. 226 Based on these findings, we considered that Siz1/Siz2 may promote anaphase 227 entry by triggering the release of Sgo1 from chromosomes upon sister kinetochore 228 biorientation. In wild type metaphase-arrested cells, chromatin immunoprecipitation 229 followed by qPCR (ChIP-qPCR) showed that Sgo1 associates with a centromeric site 230 in the absence, but not presence, of spindle tension and this pattern was largely 231 unchanged in siz1∆ siz2∆, sgo1-2R, sgo1-4R or sgo1-5R cells with reduced 232 SUMOylation ( Figure 4E and F; Figure S4F and G). We confirmed that 233 SUMOylation is dispensable for the tension-dependent release of Sgo1 during 234 anaphase by live cell imaging. In wild type cells, Sgo1-GFP first appeared as a bright 235 focus which dissociated upon splitting of Mtw1-tdTomato foci at metaphase and this 236 occurred with similar timing in siz1∆ siz2∆ cells ( Figure 4G, Figure S4H). Instead, 237 the observed delay in anaphase entry occurred after bulk Sgo1 removal ( Figure 4G). 238 Hence, although tension-dependent release of Sgo1 is critical for anaphase entry, this 239 occurs independently of Sgo1 SUMOylation. 240 241 Sgo1 SUMOylation is required for stabilizing biorientation 242 Our findings indicate that SUMOylation neither targets Sgo1 for STUbL-mediated 243 destruction nor does it facilitate Sgo1 removal under tension. To probe the mechanism 244 connecting Sgo1 SUMOylation to timely anaphase onset, we visualized the efficiency 245 of sister kinetochore biorientation in the Sgo1 SUMO mutants. We analysed the initial 246 establishment of sister kinetochore biorientation by monitoring the separation of sister 247 CEN4-GFP foci as spindles reformed after nocodazole wash-out, while maintaining a 248 metaphase arrest (by depletion of Cdc20, Figure 5A). In contrast to sgo1-3A cells 249 which have impaired biorientation, likely due to impaired CPC binding[10], all 250 mutants with reduced Sgo1 SUMOylation except sgo1-5R (siz1∆ siz2∆, sgo1-2R and 251 sgo1-4R) showed proficient sister kinetochore biorientation ( Figure 5B). Therefore, 252 error correction and sister kinetochore biorientation pathways are functional in the 253 absence of Sgo1 SUMOylation. 254 Next, we assessed the stability of biorientation in the SUMO mutants. Cells 255 were released from nocodazole washout, and the separation of CEN4-GFP was 256 monitored as cells progressed into anaphase. A single CEN4-GFP focus was observed 257 initially, and two CEN4-GFP foci appeared upon attachment of sister kinetochores to 258 microtubules from opposite poles. Stable attachment led to further separation of the 259 two CEN4-GFP foci, which eventually segregated to opposite poles in anaphase 260 ( Figure 5C). In contrast, if attachments are unstable, the two CEN4-GFP foci 261 reassociate prior to their splitting and segregation. In sgo1∆ and sgo1-3A mutants, 262 which are defective in sensing and correcting attachment errors, the visualisation of 263 two CEN4-GFP foci was delayed and the number of missegregation events was 264 increased, but the levels of reassociation of two CEN4-GFP foci was similar to wild 265 type ( Figure 5D and E, Figure S5A and B). siz1D siz2D and sgo1-4R mutants, in stark 266 contrast, were proficient in the initial establishment of biorientation and did not show 267 increased missegregation ( Figure 5D and E, Figure S5A and B). Instead, both mutants 268 showed ~15% increase in the number of cells in which the two CEN4-GFP foci 269 reassociated, indicative of unstable biorientation. We conclude that Sgo1 270 SUMOylation is important to maintain the bioriented state. 271

272
The metaphase delay in siz1∆ siz2∆ is rescued by inactivating mutations in CPC/SAC 273 Unstable biorientation in SUMO-deficient mutants is expected to generate unattached 274 kinetochores and engage the SAC, potentially explaining the metaphase delay of these 275 cells. Consistent with this idea, deletion of MAD2 partially rescued the metaphase 276 delay of siz1∆ siz2∆ cells ( Figure S6A). The CPC-dependent error correction pathway 277 is likely responsible for the instability of kinetochore-microtubule interactons in the 278 SUMO mutants because inhibition of the CPC component, Ipl1 (ipl1-as1) also 279 reduced the metaphase delay of siz1∆ siz2∆ cells ( Figure 6A). Consistently, Sgo1 280 interaction with PP2A-Rts1 and/or CPC, is important for the metaphase delay in the 281 absence of SUMOylation because both time course analysis ( Figure S6B) and live 282 cell imaging ( Figure 6B) revealed that sgo1-3A siz1∆ siz2∆ cells spent less time in 283 metaphase than siz1∆ siz2∆ cells. Interestingly, however, unlike Ipl1 ( Figure 6A), the 284 PP2A regulatory subunit, Rts1, was largely dispensible for the metaphase delay of 285 siz1∆ siz2∆ cells ( Figure S6C). Therefore, the CPC-dependent error correction 286 pathway is responsible for the metaphase delay observed in the absence of Sgo1 287

Sgo1 SUMOylation promotes Ipl1 relocalization 290
The kinase activity of Ipl1 (Aurora B kinase) is required for error correction and Ipl1 291 is known to re-localize from centromeres to the spindle mid-zone upon the 292 establishment of biorientation [25,26]. Unstable biorientation in Sgo1 SUMOylation 293 mutants suggested that this removal may be incomplete. We monitored Ipl1-GFP and 294 its co-localization with Mtw1-tdTomato by imaging cells released from G1. In both 295 siz1∆ siz2∆ and sgo1-4R, Ipl1 recruitment to the kinetochore-proximal regions 296 occurred normally ( Figure 6C). However, as kinetochores separated, Ipl1-GFP 297 persisted close to kinetochores in the Sgo1 SUMO mutants ( Figure 6C). Increased 298 centromeric Ipl1 was also measured by ChIP in SUMO-deficient cells arrested in 299 metaphase with kinetochores under tension ( Figure S7A and B). Hence, Sgo1 300 SUMOylation facilitates the re-localization of Ipl1from centromeres to prevent 301 persistent error correction and SAC activation. 302 303 Sgo1 SUMOylation is incompatible with PP2A-Rts1 binding 304 Interestingly, the coiled-coil domain of Sgo1 is both required for its SUMOylation 305 ( Figure 3) and for PP2A-Rts1 binding [10,27]. This raised the question of whether 306 Sgo1 SUMOylation also impacts PP2A-Rts1 binding. Sgo1 SUMOylation was 307 increased in the Sgo1-3A mutant ( Figure 7A), suggesting that PP2A-Rts1 binding 308 normally dampens Sgo1 SUMOylation. Structural modelling of budding yeast Sgo1-309 PP2A-Rts1 interaction revealed that PP2A-Rts1 binding to Sgo1 would be 310 incompatible with SUMOylation on these residues ( Figure S7C). We used an in vitro 311 binding assay to test the effects of Sgo1 SUMOylation on Rts1 binding. Purified Sgo1 312 was SUMOylated on beads in vitro, beads were stringently washed to remove 313 components of the SUMO reaction and subsequently incubated with cell-free extract 314 from sgo1Δ Rts1-9Myc cells. This revealed that, as expected, Rts1-9Myc bound 315 robustly to unSUMOylated Sgo1, however Rts1 binding was greatly reduced by Sgo1 316 SUMOylation, consistent with the prediction that SUMOylation and PP2A-Rts1 317 binding are mutually exclusive ( Figure 7B). Similarly, immunoprecipitated Rts1-318 9Myc bound in vitro SUMOylated Sgo1 less well than unSUMOylated Sgo1 ( Figure  319 7C). Analysis of the SUMO-deficient Sgo1-4R protein showed that, unexpectedly, 320 binding of Rts1 was reduced to a similar extent to Sgo1-3A, even in the absence of 321 SUMOylation ( Figure 7B). Despite the fact that both mutants fail to bind PP2A-Rts1, 322 they result in very different outcomes in vivo: while sgo1-3A shows defective initial 323 biorientation of sister kinetochores and chromosome mis-segregation after nocodazole 324 washout, sgo1-4R does not ( Figure 5B and E). Conversely, unstable sister kinetochore 325 biorientation and a metaphase delay is observed in sgo1-4R but not sgo1-3A cells 326 ( Figure 3D, Figure 5D, Figure 6B, Figure S6B). This indicates that a failure to bind 327 PP2A-Rts1 cannot be the primary cause of these defects. Instead, sister kinetochore 328 biorientation and chromosome segregation defects in sgo1-3A cells are attributed to 329 defective CPC maintenance at kinetochores [10] while, conversely, our data indicate 330 that CPC persists at kinetochores in sgo1-4R cells causing unstable biorientation and a 331 metaphase delay ( Figure 6C). Collectively, these data indicate that the ability of Sgo1 332 to bind and release CPC underlies the establishment and stabilization of biorientation, 333 respectively ( Figure S7D). 334 335

Dissociation of shugoshin and PP2A-Rts1 stabilizes sister kinetochore biorientation 336
To understand the importance of the Sgo1-PP2A-Rts1 interaction in stabilizing sister 337 kinetochore biorientation, we asked whether tethering of Rts1 to wild type Sgo1 or 338 Sgo1-4R could rescue the instability of attachments. Although forcing Rts1 339 interaction with Sgo1-4R throughout the cell cycle did not affect initial biorientation 340 ( Figure 7D), this state was unstable, as judged by the increased number of cells 341 switching between two and one CEN4-GFP dots ( Figure 7E), though chromosome 342 segregation was ultimately successful in the majority of cells ( Figure 7F). This further 343 confirms that loss of Rts1 binding is not the cause of unstable sister kinetochore 344 biorientation in sgo1-4R cells. However, interestingly, forced interaction between 345 Rts1 and wild type Sgo1 also resulted in frequent switches ( Figure 7E). Therefore, 346 release of the Sgo1-Rts1 interaction is important to stabilize bioriented sister 347 kinetochore-microtubule attachments. Together with our finding that PP2A-Rts1 348 binding is incompatible with Sgo1 SUMOylation, this suggests that PP2A-Rts1 349 dissociation as a result of Sgo1 SUMOylation, is also important to stabilize 350 biorientation. 351

Identification of shugoshin regulators 354
Starting with an unbiased genetic screen we have identified SUMO ligases as 355 negative regulators of the pericentromeric hub that responds to a lack of tension 356 between kinetochores. Sgo1, the central pericentromeric adaptor protein is one key 357 target of the Siz1/Siz2 SUMO ligases. Kinetochore-microtubule interactions are 358 unstable in Sgo1-SUMO deficient cells (siz1∆ siz2∆ and sgo1-4R). Persistent cycles 359 of kinetochore detachment and re-attachment engage the SAC, explaining why a 360 failure to SUMOylate Sgo1 results in a metaphase delay. Consistently, we find that 361 inactivation of components of the error correction pathway (Ipl1, Mad2) advanced 362 anaphase timing in siz1∆ siz2∆ cells. 363 Sgo1 inactivation and removal from kinetochores is essential for timely 364 anaphase entry ([18], Figure 4D) and here we have identified one mechanism that 365 contributes to this inactivation. However, Sgo1 also prevents anaphase onset by 366 inhibiting separase independently of securin [18]. PP2A-Cdc55 dependent 367 dephosphorylation of separase and, potentially also cohesin itself is likely to underlie 368 this inhibition [28,29]. Notably, ZDS2, a negative regulator of PP2A-Cdc55 [30], was 369 also isolated in our screen along with HOS3, the cohesin deacetylase [31-33] 370 indicating that further mechanisms await discovery. 371 372

SUMOylation of Sgo1 ensures efficient entry into anaphase 373
How does Sgo1 SUMOylation regulate anaphase entry? We found that Sgo1 374 SUMOylation is dispensable for its tension-dependent release from pericentromeres 375 and that, although SUMOylation promotes Sgo1 degradation indirectly, this is not 376 required for efficient anaphase entry. Instead, our work suggested that Sgo1 377 SUMOylation likely promotes anaphase entry by silencing the error correction 378 process, as biorientation was highly unstable in Sgo1 SUMO-deficient mutants 379 ( Figure 5D). Remarkably, we found that Ipl1 removal from kinetochores was 380 incomplete in the Sgo1 SUMO mutants ( Figure 6C), suggesting a key role of this 381 modification in modulating the subcellular localization of Ipl1. 382 Meanwhile, we showed that Rts1 binds preferentially to unSUMOylated Sgo1,383 and that tethering Sgo1 to Rts1 destabilized biorientation in a similar way as the Sgo1 384 SUMO mutants. These findings suggest that Sgo1 SUMOylation-mediated Rts1 385 dissociation has an important role in stabilizing microtubule-kinetochore attachment. 386 Interestingly, PP2A-B56 dampens the effects of Aurora B to allow initial attachments 387 in human cells [34][35][36], suggesting a potential explanation for our observations. 388 Overall, our findings indicate that SUMOylation modulates the kinase-phosphatase 389 balance at the kinetochore to dampen CPC activity and allow initial kinetochore-390 microtubule attachments ( Figure 7G). Different mutants affect this balance in distinct 391 ways, leading to the observed outcomes on the establishment and stabilization of 392 biorientation ( Figure S7D)    Cells carrying Spc42-tdTomato and Cdc14-GFP were synchronized in G1 in media containing 2% raffinose. 25 µM copper sulfate was added to induce pCUP1-SIZ2 expression. After releasing from G1, 0.2% galactose was added to induce pGAL-SGO1 expression. The duration of metaphase was estimated by the time taken between the separation of the spindle pole bodies (two Spc42-tdTomato foci) and the dispersal of Cdc14-GFP from the nucleolus. (C) Metaphase duration is shown for wild type (AMy24115), pGAL-SGO1 (AMy27596), pGAL-SGO1 pCUP1-SIZ2 (AMy27738) and pCUP1-SIZ2 (AMy27952) strains. (D and E) Siz1 and Siz2 are required for timely anaphase onset. Metaphase duration was determined as the time between formation of a short bipolar spindle (YFP-Tub1) and release of Cdc14-GFP from the nucleolus from live cell imaging. (D) Schematics and representative images are shown. (E) Metaphase duration is shown for wild type (AMy24174) and siz1∆ siz2∆ (AMy24313) strains. (F) and (G) The metaphase delay of siz1∆ siz2∆ cells is partially rescued by SGO1 deletion. Wild type (AMy1290), sgo1∆ (AMy8466), siz1∆ siz2∆ (AMy8465) and siz1∆ siz2∆ sgo1∆ (AMy12110) strains carrying PDS1-6HA were released from a G1 arrest. Spindle morphology was scored after anti-tubulin immunofluorescence and the percentages of short (metaphase) spindles are shown (top) and Pds1 levels were analysed by anti-HA Western blot (bottom). Pgk1 is shown as a loading control. Arrows and asterisks indicate SUMO-Sgo1-6HA and unmodified Sgo1-6HA, which binds non-specifically to the resin, respectively. (C) Sgo1 is SUMOylated by Siz1 and Siz2 in vitro. Purified Sgo1 was incubated with 1 µM E1, E2, E3, SUMO and ATP or missing one component as indicated. Reaction was incubated at 30°C for 3 h. (D) Sgo1 SUMOylation occurs in metaphase. Cells carrying SGO1-6HA and 7xHIS-SMT3 (AMy7655) were released from G1, harvested at the indicated intervals, and SUMOylation was analysed as described in (A). Cell cycle stage was monitored by scoring spindle morphology after anti-tubulin immunofluorescence. (E) Chromatin association promotes Sgo1 SUMOylation. Sgo1 SUMOylation was determined in wild type (AMy7654), bub1∆ (AMy10098), bub1-KD (catalytically inactive Bub1 kinase, AMy10102), sgo1-100 (AMy26334) and sgo1-700 (AMy26336) strains. (F) Sgo1 SUMOylation is lost upon the establishment of tension between sister kinetochores. Cells carrying pMET-CDC20 and either 7xHIS-SMT3 (AM9641) or empty vector (AMy26342) were arrested in metaphase by depletion of Cdc20 either in the presence of benomyl and nocodazole (no tension) or DMSO (tension).

Yeast strains and plasmids
Yeast strains are derivatives of W303 and are listed in Supplementary Table S3. Plasmids and primers are listed in Supplementary Tables S4 and S5, respectively. StuI-digested AMp1239 was transformed into a CDC14-GFP strain to make the YFP-TUB1 CDC14-GFP parent strain. Genes were deleted or tagged using PCR-based transformation. K-R mutant plasmids were generated using Quikchange II XL sitedirected mutagenesis kit (Agilent), with primers listed in Table S5. K-R mutants were PCR amplified from the resulting plasmid using primers AM16 and AM3177 and were integrated into an sgo1∆ strain (AMy827). 7HIS-SMT3 and HIS-UBI plasmids were kind gifts from Dr. H. Ulrich.

Yeast growth and synchronization
Unless otherwise stated, yeast strains were grown in YEP supplemented with 2% glucose and 0.3 mM adenine (YPDA). For the benomyl sensitivity assays, plates were made by adding 10 μg/mL benomyl (Sigma) or the equivalent volume of DMSO (solvent control) to boiling media.
To synchronize cells in G1, overnight cultures were inoculated to OD600 = 0.2 and grown for 1 h at room temperature, before diluting back to OD600 = 0.2. α-factor was added to 5 μg/mL for 90 min and then re-added to 2.5 μg/mL for every 90 min, until > 95% cells exhibited shmooing morphology. To release cells from G1, α-factor was washed out using 10 ´ volume relevant media. For pMET-CDC20-containing strains, cells were arrested in G1 in methionine dropout medium. After α-factor wash-out, cells were released into YPDA (+ DMSO or + 30 μg/mL benomyl and 15 μg/mL nocodazole) + 8 mM methionine for 1 h. 4 mM methionine and DMSO/15 μg/mL nocodazole were re-added for 1 h.

Metaphase duration measurements by live cell imaging and mitotic time course
Synthetic complete/dropout media were used for growing and washing cells for live cell imaging. Cells released from G1-arrest were loaded onto μ-slide dishes (Ibidi) coated with concanavalin-A (Sigma). Images were taken at indicated time intervals using a Zeiss inverted microscope, in a temperature-controlled chamber (25°C for glucose-based media and 30°C for raffinose-based media). Movies were analyzed using the ImageJ software.

Analysis of in vivo SUMOylation
Cultures were inoculated to OD600 = 0.2 in 200 mL synthetic dropout media and grown for 4 h at room temperature. Equal OD of cells were collected for samples in the same experiment. Cell pellets were resuspended in 20 mL cold H2O and incubated with 3.2 mL solution containing 1.85 M sodium hydroxide and 7.5% βmercaptoethanol. After 20 min incubation on ice, 1.65 mL 100% trichloroacetic acid was added and cells were incubated on ice for a further 20 min. Cell pellets were drop-frozen in liquid nitrogen, and subsequently lysed by bead-beating in 300 µL buffer A (6 M guanidine hydrochloride, 100 mM sodium phosphate buffer pH 8.0, 10 mM Tris-HCl pH 8.0). Lysate was diluted three-fold in buffer A and 10 µL was saved as input controls. Lysate was applied to a column packed with 600 μL 50% slurry Ni-NTA agarose beads (Qiagen), washed twice with buffer A + 0.05% Tween-20, twice with buffer B (8 M urea, 100 mM sodium phosphate buffer pH 6.8, 10 mM Tris-HCl pH 6.8) + 0.05% Tween-20 and once with buffer B + 0.05% Tween-20 + 20 mM imidazole. SUMOylated proteins were eluted by buffer B + 0.05% Tween-20 + 200 mM imidazole. Input and elute samples were concentrated by centrifugation using Vivaspin columns (Sartorius).

Mass spectrometry
A pGAL-SGO1 strain with 7HIS-SMT3 was inoculated to 0.2 OD in 2% raffinose media. After 3 h growth at room temperature, cells were diluted back to 0.2 OD and 2% galactose was added to induce SGO1 overexpression. Cells were harvested after 3 h and SUMOylated proteins were purified as described above. Proteins eluted from the Ni-NTA column were separated on a 4-12% NuPage Bis-Tris gel (Invitrogen) and the gel slice encompassing SUMOylated Sgo1 (between ~100 kDa and 135 kDa, based on immunoblotting of a parallel gel) was excised for mass spectrometry analysis.

Cloning, expression and purification of recombinant Sgo1
Full-length wild type SGO1 was amplified from plasmid AMp899, to replace SMT3 in plasmid AMp773 by Gibson assembly using primers AM8849 to AM8852. V5 tag was inserted by Gibson assembly using plasmid AMp970 and primers AM8866 to AM8869 to generate N-terminal tobacco etch virus (TEV) protease cleavable Hisx7-V5 tag-tagged SGO1 (AMp1738) under the control of a pCUP1 promoter. sgo1-4R was PCR amplified from plasmid AMp1340 and ligated into AMp1738 by Gibson assembly using primers AM8850, AM8851, AM9124 and AM 9125 to generate Nterminal TEV protease cleavable Hisx7-V5 tag-tagged sgo1-4R (AMp1783).
A protease-deficient yeast strain (AMy8184) was transformed with the resulting plasmids (AMp1738 or AMp1783) and the transformants were inoculated into 8 L uracil dropout media. When OD600 reached 0.5 -0.7, 0.5 mM CuSO4 was added to induce Hisx7-V5 -SGO1 (or sgo1-4R) expression. Cells were harvested 6 h after induction. Cell pellets were snap frozen in liquid nitrogen and ground to powder in a ball breaker machine (Retsch). All purification steps were performed at 4°C or on ice.   For analyzing biorientation in cells going into anaphase ( Figure 5E and 7F), nocodazole-arrested cells were loaded onto the ONIX Microfluidic Perfusion System (CellAsic) and visualized with a Zeiss inverted microscope coupled to an EMCCD camera at 25°C. Imaging started as soon as cells were released into methionine dropout media without drugs. Table S1. Complete list of high copy suppressors of GAL-SGO1 sickness identified in the screen shown in Figure S1A.  Deletion of CDC55 partially alleviates the metaphase delay phenotype of the siz1∆ siz2∆ mutant. Mitotic time course analysis was performed as described in Figure 1F, for the following strains: wild type (AMy8467), cdc55∆ (AMy8779), siz1∆ siz2∆ (AMy8452) and siz1∆ siz2∆ cdc55∆ (AMy8637).  Schematics describing the truncation mutants generated for Sgo1. The conserved coiled-coil and basic domains are highlighted in red and blue, respectively. Results from (B) are summarized on the right. (B) Sgo1 is likely to be SUMOylated in the first 208 amino acids. In vivo SUMOylation was assessed for the following Sgo1-6HA tagged strains as described in Figure 2A, together with the indicated negative controls: wild type (AMy7654), sgo1∆2-108 (AMy14764), sgo1∆ 2-208 (AMy14765) and sgo1∆2-308 (AMy14766). Unmodified Sgo1 bands are marked with asterisks. In addition to Lys124 SUMOylation, identified by mass spectrometry, the region between amino acids 41 and 108 is likely to be SUMOylated. In vivo SUMOylation was assessed for the following Sgo1-6HA tagged strains: wild type (AMy7654), sgo1∆2-40 (AMy18194), sgo1∆2-40 K124R (AMy18476) and sgo1∆2-108 K124R (AMy16540). (C-F) Characterization of unSUMOylatable Sgo1 mutants. (C) The Sgo1-K124R mutants does not show a metaphase delay. Mitotic time course analysis as described in Figure 1F was performed for wild type (AMy8467) and sgo1-K124R (AMy24448) strains carrying The degradation of Sgo1 depends on SUMO-conjugating protein Ubc9. The cells were synchronized in nocodazole and released into medium with α-factor to ensure arrest in G1. (B and C) Sgo1 half-life is increased in slx5∆ and siz1∆ siz2∆ mutants. (B) Scheme describing the cycloheximide chase experiment. Cells were arrested in G1 throughout the experiment and pGAL1-SGO1-9MYC expression was initially prevented by growth of cells in raffinose. Subsequently, a pulse of Sgo1 was provided by the addition of galactose, after which de novo Sgo1-9Myc synthesis was quenched The distance between Spc42-tdTomato dots was measured by ImageJ and the average distance was calculated for each time point. (B) The initial establishment of biorientation is unaffected in siz1∆ siz2∆ and sgo1-4R cells. The time point at which a cell first displayed two CEN4-GFP dots was defined as the timing of the initial establishment of biorientation. Wild type (AMy8467), siz1∆ siz2∆ (AMy8452), sgo1-3A (AMy8964) and siz1∆ siz2∆ sgo1-3A (AMy8755) strains carrying PDS1-HA and SGO1-9MYC were analysed as described in Figure 1F. (C) Deletion of RTS1 did not rescue the metaphase delay of the siz1∆ siz2∆ mutant. Mitotic time course was performed as in (A) for the following strains: wild type (AMy8467), rts1∆ (AMy20909), siz1∆ siz2∆ (AMy8452) and siz1∆ siz2∆ rts1∆ (AMy17284). were measured by ChIP-qPCR using wild type (AMy26686), siz1∆ siz2∆ (AMy23194), sgo1-2R (AMy26684), sgo1-4R (AMy26692) and sgo1-5R (AMy26691) carrying IPL1-6HA, together with a no tag control (AMy2508). Cells were arrested in metaphase by depletion of Cdc20 in the presence or absence of spindle tension. Error bars represent standard errors calculated from 5 biological repeats. * = P < 0.05. (B) Ipl1 protein levels are unchanged in Sgo1 SUMO-deficient mutants. Protein extracts from (A) were analyzed by anti-HA and anti-Kar2 (loading control) western blotting. (C) Structural modelling predicts that SUMOylation on the coiled-coil domain of Sgo1 is incompatible with Sgo1-PP2A interaction. S.c. Sgo1-PP2A interaction was modelled based on structural information obtained from cocrystallized human Sgo1(51-96) and PP2A using Phyre2 web portal (www.sbg.bio.ic.ac.uk/phyre2) [54]. Potential consequence of symoylation was modelled using the molecular graphic program PyMOL (The PyMOL Molecular Graphics System, Version 2.0 Schrödinger, LLC). According to this model, Lys64 and Lys70 are critically positioned at the binding surface with no room to accommodate a bulkier modification such as sumoylation. Lys56 is exposed to the solvent, but the attachment of SUMO (highlighted in gold) is expected to result in steric clashes with PP2A and weaken Sgo1-PP2A binding. Structural information is unavailable beyond Leu82 and so Lys85 could not be included in this model. (D) Model for role of Sgo1 SUMOylation in stabilizing the bioriented state. In wild type cells, Sgo1 brings both PP2A-Rts1 and CPC to centromeres and PP2A-Rts1 enhances CPC localization. A minor pool of Sgo1 is dynamically SUMOylated and this both prevents PP2A-Rts1 binding and directly or indirectly promotes CPC removal, dampening its activity at kinetochores. Upon tethering of PP2A-Rts1 to Sgo1, release of PP2A-Rts1 cannot occur and CPC activity persists. In sgo1-3A cells, the interaction with both PP2A-Rts1 and CPC is absent and error correction is defective due to a failure to maintain CPC. In sgo1-4R cells, PP2A-Rts1, but not CPC binding is lost. The failure to SUMOylate, along with the absence of PP2A-Rts1, means that the phosphatase cannot undergo its capture and release by Sgo1. Either as a consequence of this and/or other effects of Sgo1 SUMOylation, CPC is maintained on kinetochores resulting in ectopic error correction and destabilization of kinetochoremicrotubule attachments. siz1D siz2D mutants show similar behaviour to sgo1-4R except that Sgo1 is expected to retain the ability to bind PP2A-Rts1, so that absence of SUMOylation both prevents CPC removal and PP2A-Rts1 release.