A highly conserved binding pocket on PP2A-B56 is required for shugoshin binding and cohesion protection

The shugoshin proteins are universal protectors of centromeric cohesin during mitosis and meiosis. The binding of human Sgo1 to the PP2A-B56 phosphatase through a coiled coil (CC) region is believed to mediate cohesion protection during mitosis. Here we undertook a structure function analysis of the PP2A-B56-Sgo1 complex, revealing unanticipated aspects of complex formation and function. We establish that a highly conserved pocket of the B56 regulatory subunit is required for Sgo1 binding and cohesion protection. Consistent with this, we show that Sgo1 blocks the binding of PP2A-B56 substrates containing a canonical B56 binding motif. Surprisingly, we identify B56 and Sgo1 mutants that prevent complex formation yet support cohesion protection and normal mitotic progression. This suggests that Sgo1 and PP2A-B56 have cohesion protection activity independently of complex formation. Collectively our work provides important insight into cohesion protection during mitosis.

The B56 subunit of the holoenzyme confers substrate specificity by binding to interactors that target the phosphatase to its substrates.Most B56 interactors bind via a conserved LxxIxE peptide motif that engages a highly conserved pocket on B56 present in all five B56 isoforms (Hertz et al, 2016;Wang et al, 2016;Wang et al, 2020;Wu et al, 2017).A number of important mitotic regulators such as BubR1, Kif4A and RacGAP1 bind to PP2A-B56 through a LxxIxE motif to regulate specific dephosphorylation events.There are five isoforms of B56 (B56a, b,g, d and e) that display distinct localization patterns during mitosis (Bastos et al, 2014;Foley et al, 2011;Vallardi et al, 2019).
However, there are limited data demonstrating that Sororin and SA2 are directly dephosphorylated at centromeres by the PP2A-B56-Sgo1 complex.
In addition to recruiting PP2A-B56, the shugoshin proteins also recruit the chromosomal passenger complex (CPC) to centromeres through their CC region (Kawashima et al, 2007;Tsukahara et al, 2010;Vanoosthuyse et al, 2007).The shugoshin-dependent localization of CPC to the centromere could also contribute to cohesion protection (Hengeveld et al, 2017).
Although the interplay between shugoshin recruitment of PP2A-B56 and the CPC to centromeres is not fully established, recent work suggests that the ability of Sgo1/2 to recruit the CPC and PP2A-B56 are distinct activities (Bonner et al, 2020).These observations underscore the complexity of the CC region of the shugoshin proteins.
Crystallographic studies have determined the human PP2A-B56g in complex with a fragment of human Sgo1 comprising residues 51-96, which represents most, but not the entire N-terminal CC domain (Xu et al., 2009).This Sgo1 fragment displays less affinity to PP2A-B56g than longer N-terminal fragments of Sgo1, but sufficient affinity to efficiently bind to PP2A-B56g under crystallization conditions using high protein concentrations.The structure revealed that the Sgo1 fragment forms a dimer which engages several residues of the last C-terminal HEAT repeat of B56g and makes contacts to the PP2A catalytic subunit (Fig. 1A).Although the crystal asymmetric unit shows a 1:1 interaction between Sgo1 peptide strands and PP2A holoenzymes, the Sgo1 peptide strands are arranged into a parallel CC homodimer, where one fragment is related to the other by a two-fold crystallographic symmetry axis (depicted as chain A and Asym in Fig. 1A).This arrangement allows them to interact symmetrically with PP2A enzymes on both sides.Thus, one PP2A-B56g holoenzyme displays interactions with residues from both of the two alpha helices forming one Sgo1 CC region in the crystal, which is again consistent with biochemical experiments showing that dimerization of Sgo1 is required for binding to PP2A-B56g (Tang et al, 1998;Xu et al., 2009).In the PP2A-B56g-Sgo1 structure, the LxxIxE binding pocket of B56g is fully exposed and indeed the N-terminal region of Sgo1 does not appear to contain any recognizable LxxIxE motif.
These observations raise the possibility that the PP2A-B56-Sgo1 complex can make higher order complexes with LxxIxE containing proteins, which could be important for mitotic cohesion protection.We explored this possibility, which revealed unanticipated aspects of the PP2A-B56-Sgo1 complex important for understanding cohesion protection.

Sgo1 and LxxIxE motifs compete for binding to PP2A-B56
We first determined if Sgo1 can bind to PP2A-B56 in complex with LxxIxE containing proteins.We generated stable inducible HeLa cell lines that express YFP-tagged B56a (stable inducible HeLa cell lines used throughout unless indicated) and arrested cells in prometaphase using nocodazole.Mitotic cells were collected by mitotic shake-off and YFP-B56a was purified using a YFP affinity resin.This enriches the entire PP2A-B56a holoenzyme on the beads (Fig. 1B) and co-purifies LxxIxE containing proteins such as BubR1 and Kif4A (Hertz et al., 2016).We then incubated the purified YFP-B56a with either recombinantly expressed and purified full-length Sgo1 or an N-terminal fragment of Sgo1 spanning residues 1-154 and washed the complexes (Fig. 1C).As a control, we treated YFP-B56a purifications with buffer instead of Sgo1.Strikingly, both BubR1 and Kif4a bound to PP2A-B56a in the control samples but were efficiently displaced in the presence of Sgo1(Fig.1D).We performed a similar experiment in the presence of a high affinity LxxIxE peptide or the control peptide LxxAxA.The LxxIxE peptide efficiently displaced BubR1 and Kif4A as expected but also reduced Sgo1 binding (Fig. 1E).These results suggest that Sgo1 might engage the conserved LxxIxE binding pocket of B56a for binding.To further confirm this, we used a panel of B56a mutants that have mutations in the LxxIxE binding pocket and analyzed their ability to bind Sgo1.YFP-B56a variants were purified from prometaphase arrested cells and Sgo1 and BubR1 binding was analyzed.Interestingly, all B56a mutants unable to bind BubR1 failed to co-purify Sgo1 (Fig. 1 F).The reason why the LxxIxE peptide does not fully displace Sgo1, in contrast to the B56 mutants, could reflect that the PP2A-B56-Sgo1 complex is very stable once formed.
Collectively, these results indicate that LxxIxE motif-containing proteins and Sgo1 compete for a common binding surface on PP2A-B56a.

The LxxIxE binding pocket of PP2A-B56 is required for cohesion protection
The involvement of the B56a LxxIxE binding pocket in Sgo1 binding was surprising, given that Sgo1 binds the less conserved C-terminal HEAT repeat of B56g in the reported structure of the PP2A-B56g-Sgo1 complex (Fig. 2A).To further analyze this, we investigated the B56a R222E LxxIxE pocket mutant in depth for Sgo1 binding and cohesion protection.We compared this to a B56a mutant (B56a 5A), in which all residues at the reported structural interface with Sgo1 were mutated (B56a 5A:Y365A, H377A, Y381A, L384A, M388A) (Fig. 2A).First, we compared the binding of PP2A-B56a to Sgo1 and LxxIxE containing mitotic regulators.Consistent with the reported structure of the PP2A-B56g-Sgo1 complex, we found that YFP-B56a 5A bound less Sgo1 while maintaining its interactions with BubR1 and Kif4A (Fig. 2B).In contrast, B56a R222E (mutation in the LxxIxE binding pocket) lost both binding to Sgo1 and LxxIxE containing proteins.In a reciprocal experiment, cells stably expressing FLAG-tagged B56a variants were transfected with YFP-Sgo1 and then YFP-Sgo1 was affinity-purified from mitotic cells.Again, we observed impaired binding to both B56a R222E and 5A, with the latter mutant retaining more binding to Sgo1 (Fig. 2C).A similar result was obtained using YFP-Sgo2 (Fig. EV1A).These experiments strengthen the conclusion that the LxxIxE binding pocket of B56 is an important binding determinant for the shugoshin proteins.
We next analyzed the ability of the B56a mutants to support cohesion protection.All B56 isoforms were depleted by RNAi and cells were induced to express RNAi resistant YFP-B56a variants at endogenous levels (Fig. EV1B).This in our hands did not affect Sgo1 or Sgo2 localization to centromeres (Fig. EV1C-D).Cells were synchronized in prometaphase using nocodazole and chromosome spreads were stained with CREST and DAPI to analyze cohesin integrity.The distance between the two peak intensities of CREST was measured, as premature cohesin removal results in longer distances.Indeed, depleting all B56 subunits increased the distance between centromeres, which was rescued by expressing B56a wild type (WT) (Fig. 2D-E).As anticipated from the interaction studies, B56a R222E did not support cohesion protection at all while B56a 5A surprisingly did (Fig. 2D-E).To further substantiate this result, we performed live cell imaging of the same conditions.Removing Sgo1 and consequently centromeric cohesin results in prolonged mitotic arrest because of activation of the spindle assembly checkpoint.Similarly, depleting all B56 isoforms resulted in a prolonged arrest which was rescued by YFP-B56a WT and 5A but not the R222E mutant, thus paralleling the chromosome spread results (Fig. EV1E-G).Consistent with our binding experiments (Fig. 2B-C), only YFP-B56a WT displayed localization to chromosomes as observed by live cell imaging (Fig. EV1F).We analyzed the YFP-B56a 5A phenotype over a range of expression levels and even low levels of expression rescued the B56 RNAi.
These results show that mutating the LxxIxE binding pocket of B56a abolishes cohesion protection while the reported interface for binding the Sgo1 CC appears less critical for this.
To establish that B56a R222E can assemble an active PP2A holoenzyme capable of cohesion protection, we artificially recruited the B56a mutants to the centromere by fusing them to the centromere-targeting domain of Cenp B (CB).We then asked if in the absence of Sgo1, these B56a mutants supported cohesion protection (Fig. EV2A-B for Sgo1 depletion).We performed chromosome spreads and measured the distance between CREST peak intensities.All variants of CB-B56a rescued the cohesion defect when Sgo1 was depleted, arguing that they form functional PP2A complexes (Fig. 2F-G).The results do not exclude the possibility that binding of a LxxIxE interactor to PP2A-B56 is required for cohesion protection under physiological conditions.
Collectively, our analysis of B56a R222E shows that this mutant is defective in Sgo1 binding and cohesion protection.At present, we do not know if the defect in cohesion protection is due to loss of binding to Sgo1 and/or a LxxIxE interactor.

Sgo1 mutations affecting PP2A-B56 binding
We were puzzled by the fact that the B56a 5A mutant fully supported cohesion protection despite showing a clear reduction in Sgo1 binding.This suggested that binding of PP2A-B56 to Sgo1 might not be strictly required for cohesion protection.To explore this further, we investigated the consequence of mutating the residues in Sgo1 involved in binding the C-terminal region of B56.We generated a Sgo1 mutant (Sgo1 4A) where the four residues (L83, K87, Y90, C94) contacting B56g in the reported structure were mutated to alanine residues (Fig. 3A and Fig. EV2C).As a comparison, we used a previously reported Sgo1 3A mutant (Y57, N60, K62 to alanine) which contains three mutated residues at the interface with the PP2A catalytic subunit (Xu et al., 2009).The reported interface with the PP2A catalytic subunit involves residues from both alpha helices of the Sgo1 CC region (Fig. 3A).
Stably expressed YFP-Sgo1 variants were purified from mitotic cells at two salt concentrations (50 mM and 150 mM NaCl) and binding to PP2A-B56a was determined (Fig. 3B-C).Both Sgo1 4A and 3A showed a strong reduction in binding to PP2A-B56 components and only Sgo1 4A maintained some residual binding at the low salt concentration (Fig. 3B).
To analyze binding in cells, we took two separate approaches.Firstly, we employed an assay where LacI-GFP fusions of Sgo1 variants are localized to a LacO array on chromosome 1 in U2OS cells, which allows visualization of PP2A-B56 and CPC recruitment (Fig. 3D).Mitotic cells expressing LacI-GFP fusions of Sgo1 full length protein and 1-130 were stained for PP2A-C or CPC components (Aurora B and Borealin), and signals were quantified and normalized to GFP.Compared to Sgo1 WT, both Sgo1 4A and Sgo1 3A mutants recruited PP2A-C less efficiently (Fig. 3D, F and Fig. EV2 A-E).Consistent with the low salt purifications (Fig. 3B), Sgo1 4A recruited slightly more PP2A-C than Sgo1 3A.In contrast, we observed more subtle variations in recruitment of CPC components in the full length Sgo1 constructs while there was no difference using the Sgo1 1-130 constructs (Fig. 3G, H and Fig. EV2D, E).
In a second approach, we fused the TurboID tag to the N-terminus of Sgo1 and following addition of biotin to mitotic cells for 1 hour, we enriched biotinylated proteins under stringent purification conditions (Fig. 3I and Fig. EV 3F).Subsequent analysis of samples by label free quantitative mass spectrometry revealed labelling of PP2A-B56 components as well as CPC components and other centromeric proteins (Fig. EV3G, Table EV1).Using this approach, we compared PP2A-B56 binding between the different Sgo1 variants.This revealed a significant (p-value<0.05,log2 fold change>1) reduction in biotinylation of B56 subunits in both Sgo1 3A and 4A compared to Sgo1 WT (Fig. 3J, Table EV1).Consistent with the LacO array results, we observed a reduction in labeling of CPC components Aurora B and Borealin in Sgo1 3A and 4A compared to Sgo1 WT while INCENP labelling was less affected.In particular, we noted a significantly stronger reduction in Borealin labeling in Sgo1 3A (log2 fold change=3.24,p<0.05) compared to Sgo1 WT and Sgo1 4A.This subunit has been reported to bind Sgo1 directly (Tsukahara et al., 2010), and thus might be more sensitive to mutations in Sgo1 that affect CPC recruitment.Consistent with our results it has been reported that Sgo1 3A is less efficient in localizing Ipl1 (Aurora B) in budding yeast (Verzijlbergen et al, 2014).Furthermore, we noted lower levels of Plk1 labeling in Sgo1 3A compared to the other Sgo1 proteins (Table EV1).Plk1 has been shown to regulate MEI-S322 and human Sgo1 localization to centromeres but we have not noticed any obvious differences in centromere localization between the different Sgo1 variants (Clarke et al, 2005;Tang et al., 2006).
Collectively, we define two Sgo1 mutants showing a strong reduction in PP2A-B56 binding and which also have reduced levels of CPC recruitment capacity.

Impairing PP2A-B56 binding to Sgo1 does not prevent cohesion protection
To determine the ability of Sgo1 mutants to support cohesion protection, we depleted endogenous Sgo1 and expressed RNAi-resistant Sgo1 mutants.Cells were arrested in mitosis; chromosome spreads were prepared and distances between kinetochore pairs were measured (Fig. 4A-B).Sgo1 depletion resulted in complete loss of cohesion, which was rescued by Sgo1 WT and Sgo1 4A but not by Sgo1 3A.In this experimental setup, we observe protection of cohesin along chromosome arms likely due to overexpression of exogenous Sgo1.The lack of cohesion protection in Sgo1 3A is consistent with data from budding yeast meiosis (Xu et al., 2009).To further substantiate these findings, we performed a live cell analysis of cells complemented with the different Sgo1 variants and monitored mitotic progression (Fig. 4C-E).Sgo1 depletion induced a strong mitotic arrest which was rescued by Sgo1 WT and Sgo1 4A but not by Sgo1 3A, consistent with the chromosome spread results.From the live cell analysis, it was obvious that Sgo1 3A expressing cells did not align chromosomes (Fig. 4D).We analyzed all Sgo1 variants at a similar fluorescent intensity as well as over a range of fluorescent intensities, and this revealed that even low levels of Sgo1 4A expression was sufficient to support Sgo1 function.
Given the fact that Sgo1 4A bound slightly more PP2A-B56 compared to Sgo1 3A, this could be sufficient to make Sgo1 4A functional if complex formation is required for cohesion protection.If this was the case, we reasoned that a partial depletion of B56 regulatory subunits would impact the function of Sgo1 4A more than Sgo1 WT.To test this, we incorporated a B56 RNAi depletion step in our Sgo1 complementation protocol (Fig. EV4A).This resulted in partial depletion of all B56 regulatory subunits (Fig. EV4B) and cells died faster in mitosis compared to Sgo1 depleted cells (Fig. EV4C).It also increased the mitotic timing of Sgo1 WT complemented cells as expected due to the role of PP2A-B56 in establishing kinetochore-microtubule interactions (Foley et al., 2011).However, Sgo1 4A was not more sensitive than Sgo1 WT to depletion of B56 regulatory subunits and progressed slightly faster through mitosis.The median time of Sgo1 4A in this experiment was lower than Sgo1 3A complemented cells with endogenous levels of B56 (112 min Sgo1 4A (Fig. EV4C, Sgo1+B56 RNAi) vs 224 min Sgo1 3A (Fig. 4E, Sgo1 RNAi).
The results of the Sgo1 4A mutant show that the binding to PP2A-B56 can be lowered substantially without affecting Sgo1 function, consistent with our analysis of B56a 5A.
An important discovery from our work is that the highly conserved LxxIxE binding pocket of B56 subunits is required for Sgo1 and Sgo2 binding and cohesion protection.This was surprising based on the reported structure of the human PP2A-B56g-Sgo1 complex and the fact that the B56 binding region of Sgo1 and Sgo2 lacks a recognizable LxxIxE motif.One possibility is that the solved structure, which used only a short fragment of Sgo1, does not fully recapitulate the PP2A-B56-Sgo1 complex and crucial aspects of the structure are yet to be uncovered.We anticipate that full length Sgo1 binds in a manner that engages the LxxIxE binding pocket of B56 as well as the C-terminal HEAT repeat as reported in the structure.Consistent with this, in vivo cross-linking mass spectrometry identified peptides of Sgo1 cross-linked to residues in close proximity to the LxxIxE binding pocket on B56 (Herzog et al, 2012).Given the strong conservation of the B56 LxxIxE binding pocket, our results explain why Sgo1 co-purifies with all isoforms of B56 (Kitajima et al., 2006).In addition to this, specific sequence elements present in B56a might further favor Sgo1/2 binding (Vallardi et al., 2019).An implication from our results is that Sgo1 and LxxIxE motifs compete for binding to PP2A-B56 which could regulate dephosphorylation during mitosis.
Our work also raises the question of whether PP2A-B56 and Sgo1 are two independent pathways for cohesion protection or whether they act in the same pathway as anticipated from their binding to each other.Experiments with expression of Sororin or Rec8 mutants that cannot be phosphorylated or artificial recruitment of PP2A to cohesin/centromeres does not discriminate between the two models because it could simply be that enhancing the activity of the PP2A-B56 pathway bypasses the need for the Sgo1 pathway.The only way to establish this is to generate separation of function mutants which we have done here.
The results of B56a 5A and Sgo1 4A mutants reveal that a substantial reduction in PP2A-B56-Sgo1 complex formation can be tolerated without any impact on cohesion protection.
Two interpretations are possible based on these results: i) low levels of PP2A-B56 binding to Sgo1 is sufficient or ii) binding of PP2A-B56 to Sgo1 is not required.We tested this by lowering the levels of B56 in Sgo1 4A complemented cells, which favored the later interpretation.Collectively, our results with B56 5A and Sgo1 4A support a two-pathway model for cohesion protection during mitosis, though we cannot exclude that residual complex formation supports function.
In contrast, our analysis of B56a R222E and Sgo1 3A favor that Sgo1 and PP2A-B56 act in the same pathway to protect cohesion.To make this conclusion requires that the B56a R222E and Sgo1 3A mutants specifically disrupt the PP2A-B56-Sgo1 complex.For B56a R222E, we know that this is not the case as binding to LxxIxE motifs is also disrupted.Our analysis of Sgo1 3A also reveals other changes in the proximity assays, arguing that more work is required to fully establish that the only defect in Sgo1 3A is a loss of PP2A-B56 binding, or cohesion loss in this mutant could be partially attributed to loss of other regulators.
Another Sgo1 mutant that has been analyzed previously is Sgo1 N61 that was originally identified in MEI-S322 (Tang et al., 1998).This mutation prevents PP2A-B56 binding and results in a defect in cohesion protection (Tang et al., 2006;Xu et al., 2009).However, the N61 mutation causes a destabilization of MEI-S322 and this residue is not making any contact to PP2A-B56 in the reported structure (Tang et al., 1998;Xu et al., 2009).It could be that Sgo1 N61 mutation affects the integrity of the CC region which causes the cohesion defect independently of affecting PP2A-B56.Indeed, the MEI-S322 8 mutation (V35E in the CC) does not affect B56 subunit interactions yet is compromised in cohesion protection arguing for additional functions of the CC region beyond PP2A-B56 binding (Pinto & Orr-Weaver, 2017;Tang et al., 1998).
Collectively, our work provides important insight into the protection of cohesion during mitosis and pinpoints important questions that needs to be addressed in future studies.

Cloning
Standard cloning methods were used throughout the study.pcDNA5/FRT/TO vector was used unless otherwise stated.B56a variants were generated in our previous study (Hertz et al., 2016).B56a 5A, Sgo1 3A, and Sgo1 4A mutant constructs were synthesized by GeneArt (Thermo Fischer Scientific).BamHI and NotI were used to subclone B56a and Sgo1 constructs with various tags (YFP, FLAG, or TurboID).Full-length Sgo2 was amplified by PCR and inserted in pcDNA5/FRT/TO-YFP vector.For YFP-CenpB-B56a constructs, CenpB domain was amplified by PCR and inserted into pcDNA5/FRT/TO-YFP vector by standard restriction cloning, followed by subcloning of B56a variants into the vector using BamHI/NotI.

Cell Culture
HeLa FRT/TRex cells (gift from S. Taylor) were used throughout the study, unless otherwise stated.Stable cell lines were generated using the T-Rex doxycycline Flip-In system (Invitrogen).For synchronization, 2.5 mM thymidine and 200 ng/uL nocodazole were used.

Expression and purification of recombinant hSgo1
BL21 (DE3) Gold E.coli cells expressing hSgo1 FL and truncations (hSgo1 1-154 ) were grown at 37°C/200rpm to an optical density of 1.5 (OD600) and induced overnight at 18°C with 0.35mM IPTG.Cells were resuspended in lysis buffer containing 20mM Tris.HCl pH 8, 500 mM NaCl, 1mM EDTA and supplemented with complete EDTA-free cocktail tablets (1 tablet/50ml cells; Roche) and 0.01mg/ml DNase (Sigma) and 1mM PMSF.The lysate was sonicated at 60% amplitude for 8 minutes (2s on, 2s off) and centrifuged at approx.58000 x g for 50 minutes at 4°C and the protein was batch purified using chitin beads (NEB).Post lysis and high salt chaperone wash, the chitin beads were washed with 3 CV of 20mM Tris.HCl, 500mM NaCl, 50mM DTT and incubated at RT overnight.The next day, the protein was eluted with the lysis buffer without DTT.The elutions were analysed for protein quality on an SDS-PAGE, and the elutions containing hSgo1 were pooled and dialysed in 20mM Tris HCl pH 8, 125mM NaCl, 4mM DTT overnight at 4°C.The next day, the dialysed sample was loaded onto a HiTrap Q HP (GE Healthcare) anionic exchange column.The excess DNA contamination was separated from hSgo1 by providing a 50% salt gradient over 20CV in an ÄKTA start system (GE Healthcare).The samples containing hSgo1 were pooled, concentrated and the pure protein was finally obtained by a final size exclusion chromatography step with the column equilibrated with 20mM Tris.HCl, 200mM salt and 5mM DTT (Superdex 200 Increase 10/300, GE Healthcare).

Immunoprecipitation and Competition Assays
Inducible, stable cell lines expressing indicated YFP-tagged bait were lysed in a low salt lysis buffer (50 mM Tris-HCl pH 7.5, 50 mM NaCl, 1 mM EDTA, 1 mM DDT, 0.1% NP-40, protease-and phosphatase inhibitors), unless otherwise stated.In some experiments, the same lysis buffer with 150mM NaCl was used.Lysates were immunoprecipitated with GFPtrap beads (ChromoTek) according to the manufacturer's recommendation.The beads were washed three times with wash buffer (50 mM Tris-HCl pH 7.5, 1 mg/mL BSA, 20% glycerol, and 1 mM DTT) and eluted in 2x sample buffer.For the peptide competition assays, a peptide containing LxxIxE motif (LPRSSTLPTIHEEEELSLC) or a control mutant peptide that was unable to bind B56 (LPRSSTLPTAHAEEELSLC) was used.For the competition assay with Sgo1 proteins, purified full-length hSgo1 or hSgo1 1-154 described above were used.The peptides/proteins were incubated with cell lysates 30 minutes prior to the addition of GFPtrap beads.
IP samples were resolved with 4-12% Bis-Tris gels (Thermo Fischer Scientific), transferred to PVDF membranes.LI-COR Odyssey imaging system was used for visualization, and signals were quantified using Image Studio software (LI-COR).

Chromosome Spreads
Indicated cells were seeded in a 6-well plate, RNAi knockdown was performed, and cells were synchronized using thymidine followed by nocodazole treatment.48 h after RNAi transfection, mitotic cells were collected by shake-off.After hypotonic treatment with KCl, cells were spun onto microscopy slides with a Shandon Cytospin centrifuge (Thermo Fischer), fixed with 4% paraformaldehyde, and then immunocytochemistry was performed.
Representative images were taken with a 100x objective on a DeltaVision fluorescent microscope under the same condition.The distance between the two peak intensities of CREST or YFP-CB was measured for 5 kinetochore pairs using imageJ and averaged for a single cell.At least 45 cells from minimum of 3 independent experiments were analyzed.

Live Cell Imaging
Live cell imaging was performed using a DeltaVision fluorescent microscope.Cells were seeded in a 8-well ibidi dish (ibidi) a day before filming, the media was changed to Leibovitz's L-15 (Life Technologies) immediately before the filming.Indicated channels were recorded at 7-8 minute intervals and data were analyzed using SoftWoRx (GE Healthcare).The time from nuclear envelope breakdown (NEBD) to anaphase was measured in single cells.

Immunofluorescence
Cells were seeded in a 6-well plate and treated as indicated.A day before fixing, cells were transferred in an 8-well ibidi dish.Cells were fixed in 4% PFA for 10 min at room temperature, and standard immunocytochemical methods were used.Fluorescent microscopy was performed using a DeltaVision fluorescent microscope.To ensure quantitative image quality, the imaging parameters were kept constant for a given experiment.

LacO-LacI assay
Sgo1-LacI-GFP constructs were cloned in a pAceBac1-CMV background (Hadders et al, 2020).Bacmids were generated using the Bac-to-Bac system in conjunction with EMBacY cells (Berger et al, 2004;Bieniossek et al, 2012).Baculovirus was then produced by transfection of bacmids into Sf9 cells using standard procedures.P2 viruses were harvested after 5 days, filtered (0.2 um) and stored at 4°C till use.The lacO-LacI assays were performed as previously described in Hadders et al 2020.Briefly, U-2 OS LacO Haspin CM (CRISPR Mutant) cells were seeded on glass coverslips followed directly by addition of recombinant baculovirus encoding the Sgo1-LacI-GFP variants or LacI-GFP as a control.
After 3-4 hours S-trityl-L-cysteine (STLC; 20 µM) was added overnight to block cells in mitosis.The next morning cells were fixed in 4% PFA (v/v) in PHEM buffer (60 mM HEPES KOH, 20 mM PIPES KOH, pH 6.8, 5 mM EGTA, and 1 mM MgCl2) for 10-15 minutes followed by permeabilization in ice cold methanol for a minimum of 1 hour.
For immunofluorescence, cells were washed with PBS with 0.01% Tween 20 (PBST), followed by blocking with 3% BSA in PBST for ± 30 min.Cells were then incubated with primary antibodies diluted in 3% BSA in PBST for 2 h followed by washing three times, again with PBST.Cells were then incubated with secondary antibodies, GFP-Booster and DAPI (500 ng/ul) in PBST + 3% BSA for 1 h.Coverslips were washed again, twice with PBST, followed by a final wash with PBS, before mounting onto glass slides using Prolong Diamond Antifade Mountant (Thermo Fisher Scientific).
Fluorescence images were acquired on a DeltaVision imaging system (GE Healthcare), upgraded with a seven-color InsightSSI Module & TruLight Illumination System Module using a UPlanSApo 60×/1.40objective and a CoolSnap HQ2 camera (Photometrics).3D zstacks were collected and deconvolved using Softworx v6.Presented images are deconvolved maximum intensity projections.Quantifications were performed using an inhouse-developed macro in ImageJ that sets a threshold (Otsu) based on the GFP channel followed by measurement of all channels within this region of interest.

Figure 1 .
Figure 1.Sgo1 and LxxIxE motifs compete for binding to PP2A-B56 (A) Structure of the PP2A-B56g-Sgo1 complex (adapted from Xu et al, PDB: 3FGA).The Sgo1 coiled-coil homodimer interacts with both PP2A catalytic and B56 regulatory subunits.The model shows a LxxIxE peptide bound to B56 at its conserved binding pocket.(B) YFP pull down from cells stably expressing YFP (control) or YFP-B56a enriches the entire PP2A-B56a holoenzyme on the beads.PP2A-A, scaffold subunit; PP2A-C, PP2A catalytic subunit.(C) Coomassie-stained SDS-PAGE of the purified Sgo1 full length (FL) and Sgo1 1-154 .(D) Competition assay with the purified Sgo1 proteins shown in (C).Binding of YFP-B56a to indicated proteins was determined.Representative of 3 independent experiments.(E) Peptide competition assay with a WT LxxIxE peptide or a mutated variant that does not bind B56 (LxxAxA).Binding of YFP-B56a to indicated proteins was determined and quantified by LiCor.(F) YFP-B56a pull down from cells stably expressing the indicated LxxIxE binding pocket variants of B56a and subsequent immunoblotting of indicated proteins.Representative of 4 independent experiments.

Figure 2 .
Figure 2. Sgo1 binding to the LxxIxE binding pocket of PP2A-B56 is required for cohesion protection (A) Structure of the reported PP2Ag-B56-Sgo1 binding interface and residues mutated in the B56a 5A mutant are shown.(B) IP of YFP-B56a from cells stably expressing the B56a WT, R222E, and 5A followed by immunoblotting of indicated proteins.(C) Reciprocal IP of (B).YFP-Sgo1 expression construct was transfected into cells stably expressing FLAG-B56a WT, R222E and 5A, followed by YFP IP and immunoblotting of indicated proteins.(D) Representative images of chromosome spreads from the indicated conditions.E) Quantification of (D).The distance between the two peak intensities of CREST was measured for 5 kinetochore pairs and averaged for a single cell and plotted.The data are from 4 independent experiments and the mean and SD are indicated.(F) Sgo1 RNAi and rescue with the indicated B56a variants fused to YFP and the Cenp B centromere-targeting domain (CB).Representative images of chromosome spreads are shown.CB targets all the rescue constructs (green) to the centromere.(G) Quantification of (F).The distance between

Figure 3 .
Figure 3. Sgo1 mutations in the coiled-coil domain affect PP2A-B56 binding (A) Structure of the reported PP2A-B56g-Sgo1 binding interfaces and residues mutated in the Sgo1 3A and 4A mutants are shown.3A refers to Y57A, N60A, and K62A mutations at the PP2A-C binding interface.4A refers to L83A, K87A, Y90A, and C94A mutations at the B56 binding interface of Sgo1.(B-C) IP of YFP-Sgo1 from cells stably expressing the Sgo1 WT, 3A, and 4A using different salt conditions (50mM NaCl (B) or 150 mM (C)) followed by immunoblotting of indicated proteins and quantification by LiCor.(D-H) Mitotic U-2 OS LacO Haspin CM cells expressing Sgo1-LacI-GFP variants or LacI-GFP (control) were stained for PP2A-C (D) or CPC components, Aurora B (E) and Borealin (not shown).PP2A-C (F), Aurora B (G) and Borealin (H) signal intensity was quantified, normalized to GFP, and then plotted.Each circle represents an individual cell, and the mean fluorescent intensity is indicated.Representative of at least 3 independent experiments.(I) Schematic of the TurboID-Sgo1 approach.(J) Table summarizing the Log2 differences between Sgo1 WT, 3A and 4A.

Figure 4 .
Figure 4. Impaired PP2A-B56 binding to Sgo1 does not prevent cohesion protection (A) Representative images of chromosome spreads from Sgo1 RNAi treated cells stably expressing the indicated YFP-Sgo1 variants.All the YFP-Sgo1 rescue constructs (green) localize to the centromeres (CREST, red).(B) The distance between the two peak intensities of CREST was measured for 5 kinetochore pairs from the chromosome spreads in (A) and averaged for a single cell.The data are from 3 independent experiments and the mean and SD are indicated.(C) Experimental protocol of the live cell imaging shown in (D).(D) Sgo1 RNAi and rescue with the indicated Sgo1 RNAi-resistant constructs was performed.Representative still images captured during the live cell imaging showing DIC and YFP-Sgo1 WT, 3A, and 4A localization during mitosis.Time (min) from nuclear envelop breakdown (NEBD) is indicated.(E) The time from NEBD to anaphase was measured from Figure EV1.(A) YFP IP from cells stably expressing YFP-Sgo2 and transfected with FLAG-B56a constructs.(B) Validation of the B56 RNAi and rescue system.Endogenous B56a was efficiently depleted 48h after the RNAi treatment.The RNAi resistant YFP-B56a rescue constructs were expressed approximately at the endogenous level.(C-D) Localization of Sgo1(C) and Sgo2 (D) in cells depleted of B56 and expressing the indicated B56a variants.Representative images from 3 independent experiments are shown.(E) Experimental protocol of the live cell imaging shown in (F).(F) B56 RNAi and rescue with the indicated B56a RNAi-resistant constructs were performed.Time (min) from nuclear envelop breakdown (NEBD) is indicated.(G) The time from NEBD to anaphase was measured from 2 independent live cell imaging experiments.Each circle represents an individual cell.Blue circles indicate the cells that were still arrested at the end of filming, and red circles indicate the cells that died during mitosis.The median is indicated with the red horizontal bars.
Figure EV2.(A) Validation of the Sgo1 antibody and the Sgo1 RNAi by immunoblotting.While the Sgo1 antibody detects unspecific bands in the whole cell lysates (see input), it is specific for Sgo1 after B56 IP, as the treatment with Sgo1 RNAi completely abolishes Sgo1 signal after 48h.(B) Validation of the Sgo1 antibody and the Sgo1 RNAi by immunofluorescence.Representative immunofluorescent images are shown.(C) The conservation of the Sgo1 coiled-coil region.The residues mutated in Sgo1 3A and 4A are indicated.
Figure EV3.(A-E) Mitotic U-2 OS LacO cells expressing Sgo1 1-130 -LacI-GFP variants or LacI-GFP (control) were stained for PP2A-C (A) or CPC components, Aurora B (B) and Borealin (not shown).PP2A-C (C), AuroraB (D) and Borealin (E) signal intensity was quantified, normalized to GFP, and plotted.Each circle represents an individual cell, and the mean is indicated.Representative of at least 3 experiments.(F) Blot of stable, doxycycline inducible Figure EV4.(A) Experimental protocol of the live cell imaging with Sgo1 complementation with and without partial B56 depletion shown in (C).(B) WB showing the partial KD of all B56 isoforms.(C) Sgo ± partial B56 KD and rescue with the indicated Sgo1 RNAi-resistant constructs were performed.The time from nuclear envelop breakdown (NEBD) to anaphase was measured from the live cell imaging.Each circle represents an individual cell.Blue circles indicate the cells that were still arrested at the end of filming, and red circles indicate the cells that died.The median is indicated with the red horizontal bars.Representative of 2 independent experiments.