The selective inhibition of protein phosphatase-1 results in mitotic catastrophe and impaired tumor growth Free

Cell cycle progression is tightly controlled through reversible phosphorylation of key regulators by a variety of protein kinases and phosphatases (Novak et al., 2010; Qian et al., 2013a). One of the cell-cycle-regulating enzymes is the serine/threonine protein phosphatase-1 (PP1) a ubiquitously expressed and conserved member of the phosphoprotein phosphatase (PPP) superfamily (Heroes et al., 2013). PP1 forms heterodimeric or heterotrimeric complexes with >200 PP1-interacting proteins (PIPs), which function as substrate-targeting or -specifying subunits and activity regulators. PP1 has established substrates in all phases of the cell cycle, including the retinoblastoma protein in the G1 phase (Ludlow et al., 1993), the helicase MCM4 in S-phase (Hiraga et al., 2014), protein phosphatase Cdc25 in G2 phase (Margolis et al., 2006) and protein kinases Aurora A and Aurora B in M-phase (Bollen et al., 2009).

In invertebrates, the deletion of PP1 or its most abundant isoform causes a mid-mitotic arrest that is associated with an abnormal spindle organization and over-condensed chromosomes (Axton et al., 1990; Chen et al., 2007; Doonan and Morris, 1989; Hisamoto et al., 1994). In addition, the spindle assembly checkpoint (SAC), which prevents the onset of anaphase until all chromosomes are properly aligned and bioriented, is activated for a prolonged time (Ohkura et al., 1989; Sassoon et al., 1999). Moreover, in those cells where the SAC is eventually satisfied, the checkpoint cannot be turned off in the absence of kinetochore-associated PP1 (Meadows et al., 2011; Pinsky et al., 2009; Rosenberg et al., 2011; Vanoosthuyse and Hardwick, 2009). It is not clear whether the deletion or inhibition of PP1 also causes a similarly strong mitotic-arrest phenotype in vertebrate cells. Various PIPs mediate the interaction of PP1 with the centrosomes, spindle and mitotic chromosomes in vertebrates, including CENP-E (Kim et al., 2010), Kif18A (Häfner et al., 2014), KNL1 (Espert et al., 2014; Liu et al., 2010; Nijenhuis et al., 2014), RepoMan (also known as CDCA2) (Qian et al., 2013b; Vagnarelli et al., 2011; Wurzenberger et al., 2012) and Sds22 (Eiteneuer et al., 2014; Wurzenberger et al., 2012). These mitotic pools of PP1 holoenzymes have functions throughout the M-phase. Nevertheless, the overall contribution of PP1 to cell cycle progression in vertebrates, and cell division in particular, remains poorly understood. The injection of inhibitory anti-PP1 antibodies into early mitotic mammalian fibroblasts has been reported to cause a metaphase arrest (Fernandez et al., 1992), but the effects on other stages of the cell cycle were not investigated. The knockdown of PP1γ had no effect on the mitotic spindle but caused a cytokinesis defect and the formation of dikaryons (Cheng et al., 2000).

In general, it has been rather challenging to delineate the contribution of PP1 to cell cycle progression in vertebrates, mainly because of the lack of specific cell-permeable inhibitors. Another problem stems from the functional redundancy between PP1 isoforms, making knockdown experiments of individual isoforms rather uninformative (Cheng et al., 2000; Kirchner et al., 2007; Trinkle-Mulcahy et al., 2006; Varmuza et al., 1999). We have delineated the global role of PP1 in cell cycle progression using cell lines that inducibly overexpress nuclear inhibitor of PP1 (NIPP1; also known as PPP1R8). NIPP1 is a very potent and highly specific inhibitor of PP1, and prevents the dephosphorylation of all tested protein substrates by associated PP1, except for phosphoproteins that are recruited through the NIPP1 N-terminal forkhead-associated (FHA) domain (Beullens et al., 1992; Vulsteke et al., 1997). Hence, when the phosphate-binding loop in the FHA domain is mutated, NIPP1 can be used as a highly selective inhibitor of PP1. Here, we show that the expression of NIPP1 causes spindle-assembly and chromosome-congression defects, and a prolonged SAC that often culminates in cell death. We also demonstrate that this arrest is entirely dependent on the ability of NIPP1 to bind PP1 and is largely explained by the removal of PP1 from other mitotic PIPs. Our data show that the global inhibition of PP1 kills dividing cancer cells and suggest that mitotic PP1 holoenzymes are potential targets for a novel cancer therapy.

To delineate the role of PP1 in cell cycle progression we generated HeLa cell lines that inducibly express EGFP [denoted EGFP (Co)] or fusions of EGFP with the PP1 inhibitor NIPP1, using the Flp-In T-REx system. In addition to a fusion with wild-type NIPP1 (NIPP1-Wt), we engineered NIPP1 fusions with a non-functional substrate-binding FHA domain (NIPP1-Fm), a mutated RVxF-type PP1-binding site (NIPP1-Pm), and the combined mutations (NIPP1-Fm+Pm) (Fig. 1A). The cell lines expressed the fusions in a doxycycline (Dox)-dependent manner and to a similar extent (Figs S1A and S2A). As expected (Tanuma et al., 2008; Van Dessel et al., 2010), EGFP traps of the NIPP1-Wt and NIPP1-Fm fusions showed an association with endogenous PP1, which was not seen with the NIPP1-Pm and NIPP1-Fm+Pm fusions (Figs S1B and S2B). The associated PP1 was inactive with glycogen phosphorylase as substrate but the phosphatase activity could be revealed by trypsinolysis, which destroys NIPP1 but not PP1 (Beullens et al., 1998), confirming that NIPP1-Wt and NIPP1-Fm inhibit PP1 (Fig. S1C). In addition, the NIPP1-Wt and NIPP1-Pm fusions co-precipitated with Sap155, a pre-mRNA splicing factor that is an established NIPP1-FHA ligand (Fig. S1D). Such co-precipitation was not seen with NIPP1-Fm, the FHA-ligand binding mutant. Finally, the targeting of NIPP1 to splicing-factor storage sites or nuclear speckles, which is also mediated by the FHA domain (Jagiello et al., 2000), was not detected with the NIPP1-Fm and NIPP1-Fm+Pm fusions (Figs S1E and S2C). Collectively, these data show that EGFP–NIPP1 has the same ligand-binding properties as endogenous NIPP1 and validate EGFP–NIPP1-Pm and EGFP–NIPP1-Fm as PP1- and substrate-binding mutants, respectively.

FACS analysis of non-synchronized cells revealed that the expression of NIPP1-Wt caused an accumulation of cells with fragmented chromatin (increased sub-G1 fraction), hinting at increased cell death (Fig. 1B). Sulforhodamine B colorimetric assays confirmed that there was a reduced cell survival that depended on the concentration of ectopically expressed NIPP1-Wt (Fig. 1C,D). To examine whether overexpressed NIPP1-Wt induced cell death in a specific phase of the cell cycle, we performed live-cell imaging with differential interference contrast (DIC) microscopy. The time to mitotic entry after thymidine release amounted to 742±36 min in non-induced and 900±42 min in NIPP1-Wt-expressing cells (mean±s.e.m, n=30), suggesting a modest mitotic entry delay in the induced cells (Fig. 1E). However, during this period no DNA damage was detected, as shown by γH2Ax stainings (Fig. S2D), and 92% of the NIPP1-Wt-expressing cells entered mitosis. Thus, the induction of NIPP1-Wt did not have major effects on progression through interphase and mitotic entry.

Time-lapse imaging of selected EGFP-expressing cells after the removal of thymidine did reveal, however, a strong mitotic-arrest phenotype after the induction of either the NIPP1-Wt or NIPP1-Fm fusions (Fig. 2A). In these conditions few cells spent less than 2 h in mitosis, which was considered a normal length of mitosis. Instead, the NIPP1-Wt- or NIPP1-Fm-expressing cells often showed a mitotic arrest that culminated in (1) an apparently normal mitotic exit, (2) the formation of binucleated cells, or (3) cell death (Fig. 2B,C). Cell death was associated with membrane blebbing (Movies 1–5) and the accumulation of caspase-3 (Fig. S1F). Importantly, cell death did not occur when the cells were blocked in the G1/S or G2 phases of the cell cycle with thymidine or RO-3306, respectively (data not shown), indicating that cells underwent cell death during mitosis. In general, the mitotic-arrest phenotypes were somewhat stronger with NIPP1-Wt- than with NIPP1-Fm-expressing cells (Fig. 2A–C), hinting at a contribution of a functional FHA domain. However, the mitotic-arrest phenotype was completely absent after the expression of NIPP1-Pm or NIPP1-Fm+Pm (Fig. 2B,C), showing that it was entirely dependent on the ability of NIPP1 to bind PP1. Finally, a mitotic arrest in the NIPP1-Wt- or NIPP1-Fm-expressing cells was biochemically confirmed by immunoblotting, showing an accumulation of the mitotic regulators cyclin B and phosphorylated (mobility shifted) BubR1 (Fig. 2D,E).

To determine at which mitotic stage the NIPP1-Wt- or NIPP1-Fm-expressing cells were blocked, they were followed by time-lapse imaging after the transient expression of Cherry-tagged histone H2B (Fig. 3A,B; Movies 6–9). This analysis revealed that ∼95% of NIPP1-Wt- and ∼75% of NIPP1-Fm-expressing cells failed to align their chromosomes at the metaphase plate and spent a prolonged time in prometaphase. Some cells aligned their chromosomes at the metaphase plate for a short period, but then regressed to a prometaphase-like stage. Quantification of the number of cells in prometaphase, metaphase and anaphase plus telophase after fixation confirmed a prometaphase arrest after the induction of NIPP1-Wt or NIPP1-Fm expression (Fig. S3).

To delineate the cause(s) of the prometaphase arrest, we first examined spindle organization by staining fixed cells for α-tubulin and the centrosomal marker centrobin. This revealed major spindle defects in NIPP1-Wt- or NIPP1-Fm-expressing cells. For NIPP1-Wt these defects included aberrant and elongated spindles with unfocused spindle poles (Fig. 3C,D). NIPP1-Fm-expressing cells also showed elongated spindles, but the spindle poles were focused. In addition, a significantly longer pole-to-pole distance was measured in NIPP1-Wt- or NIPP1-Fm-expressing cells, indicating that there were microtubule dynamics defects (Fig. 3E). We also often noticed chromosome-congression defects (Fig. 3C). To quantify the latter defects, cells were first blocked in prometaphase with monastrol, resulting in the formation of monopolar spindles, and then allowed to progress to metaphase by the removal of monastrol in the presence of the proteasome inhibitor MG132. In these conditions, ∼95% of the NIPP1-Wt-expressing cells and ∼75% of the NIPP1-Fm-expressing cells showed chromosome misalignments, as compared to only ∼10% without prior induction of the transgenes (Fig. 3F,G).

Problems with spindle organization or chromosome congression are known to cause an activation of the SAC (Lesage et al., 2011). Consistent with SAC activation, we found that NIPP1-Wt- or NIPP1-Fm-expressing cells were enriched for hyperphosphorylated BubR1, a SAC component (Fig. 2D). Moreover, BubR1 was enriched at the kinetochores in these cells (Fig. 4A). In addition, NIPP1-Wt overexpressing cells only showed a partial recovery from a mitotic arrest following the washout of the cyclin dependent kinase 1 (Cdk1) inhibitor R0-3306 (Fig. 4B). However, in these experiments the mitotic index was normalized after the prior knockdown of Mad2, an essential SAC component (Maldonado and Kapoor, 2011), indicating that the overexpression of NIPP1-Wt prevents the progression through mitosis because of a persistently active SAC.

Given that PP1 contributes to SAC inactivation in yeast (London et al., 2012; Meadows et al., 2011; Pinsky et al., 2009; Rosenberg et al., 2011; Tang and Toda, 2015; Vanoosthuyse and Hardwick, 2009), we have also examined whether the NIPP1-Wt or NIPP1-Fm overexpressing cells have a skewed SAC silencing. Therefore, we performed SAC override assays (Santaguida et al., 2011) and measured the time to complete mitosis, as detected by the loss of cell rounding, after the inhibition of SAC activation in prometaphase-arrested cells with the Aurora inhibitor hesperadin or the MPS1 inhibitor reversine (Fig. 4C). With both inhibitors, the override time was significantly increased after the expression of either NIPP1-Wt or NIPP1-Fm, hinting at a problem with SAC silencing.

Full-length NIPP1 binds to PP1 with an extremely high affinity and inhibits the phosphatase towards all tested protein substrates except FHA ligands (Beullens et al., 2000; O'Connell et al., 2012; Tanuma et al., 2008). Given that the mitotic-arrest phenotype induced by the overexpression of NIPP1 was entirely dependent on a functional PP1-binding site, we reasoned that the phenotype could be caused by titrating PP1 away from other PIPs. Consistent with this notion, we found that the typical nucleolar enrichment of PP1γ was largely lost by the induction of NIPP1-Wt or NIPP1-Fm, but not by the expression of NIPP1-Pm (Fig. 5A). This suggests that overexpressed NIPP1, which is itself nuclear but excluded from the nucleoli, successfully competed with nucleolar-targeting subunit(s) for binding to PP1. Likewise, following the expression of NIPP1-Wt or NIPP1-Fm, considerably less PP1 activity co-precipitated with endogenous Ki67, RepoMan and Kif18A in prometaphase-arrested cells (Fig. 5B), indicating that NIPP1-Wt and NIPP1-Fm competitively removed PP1 from these mitotic PIPs (Booth et al., 2014; Häfner et al., 2014; Takagi et al., 2014; Vagnarelli et al., 2011). This was confirmed for RepoMan by immunoblotting (Fig. S4A). Ki67 and Kif18A are themselves substrates for associated PP1 (Häfner et al., 2014; Takagi et al., 2014). We noted that the expression of NIPP1-Wt or NIPP1-Fm, but not NIPP1-Pm, reduced the mobility of Ki67 and Kif18A, during SDS-PAGE (Fig. S4B–D), consistent with hyperphosphorylation, as expected from their reduced association with PP1. Finally, we observed a partial delocalization of Aurora B in mitotically arrested cells that expressed NIPP1-Wt or NIPP1-Fm (Fig. 5C,D). Indeed, in these cells the centromeric localization of Aurora B, which is known to be dependent on PP1 (Qian et al., 2011), was partially lost and showed instead a more diffuse distribution.

If the mitotic arrest induced by the overexpression of PP1-binding NIPP1 variants is caused by the inhibition of PP1 pools that are essential for mitosis, one would expect a (partial) rescue of this phenotype upon the overexpression of PP1. We indeed observed that the ectopic expression of RFP–PP1γ in the NIPP1-Fm-expressing cell line reduced the number of cells with a mitotic arrest (Fig. 6A; Fig. S4E). It also decreased the average duration of mitosis (Fig. 6B), as measured by time-lapse imaging. The rescue was proportional to the amount of expressed RFP–PP1. In addition, when the NIPP1-Fm-expressing cells were first synchronized with thymidine and then released for 18 h, a mitotic arrest was evident from the accumulation of cyclin B, but this phenotype was rescued by the expression of Flag–PP1γ (Fig. 6C,D). Finally, the SAC inactivation defect could also be partially rescued by the expression of RFP–PP1γ (Fig. S4F).

PP1 antagonizes the Aurora A and B kinases, both by inactivating the kinases themselves and by dephosphorylating their substrates (for reviews, see Bollen et al., 2009; Wurzenberger and Gerlich, 2011). Therefore, we speculated that the mitotic arrest and death phenotype induced by the selective inhibition of PP1 should be (partially) reversed by inhibition of the Aurora kinases. To test this hypothesis, we made use of our observation that NIPP1-Fm reduced cell survival, as determined by sulforhodamine B assays (Fig. 6E). As hypothesized, the reduced cell survival in cells that expressed NIPP1-Fm was rescued by the mere addition of the Aurora inhibitor hesperadin (Fig. 6E).

We have previously described stable HeLa Tet-off (HTO) cell lines that express Flag-tagged NIPP1-Wt or NIPP1-Pm in the absence of Dox (Fig. 7A) (Tanuma et al., 2008; Van Dessel et al., 2010). This enabled us to study the effect of the selective inhibition of PP1 on colony formation and tumor growth in the absence of an inducing agent. Consistent with our data obtained with the Flp-In T-REx cell lines (Fig. 2C), the expression of NIPP1-Wt in the HTO cell lines resulted in a gradual accumulation of dead cells, but this effect was not seen after the expression of NIPP1-Pm (Fig. 7B). The expression of NIPP1-Wt also hampered anchorage-dependent (Fig. 7C,D) and anchorage-independent colony formation (Fig. 7E,F). These effects were absent in HTO cells expressing NIPP1-Pm, demonstrating their dependency on PP1. In addition, solid-tumor formation following the subcutaneous injection of the HTO cells in nude mice was strongly reduced by the expression of NIPP1-Wt, but this effect was much smaller following the expression of NIPP1-Pm, as revealed by caliper measurements (Fig. 7G,H) and small-animal PET scan analysis (not illustrated). The tumor take of the control, NIPP1-Wt- and NIPP1-Pm-expressing cells was 100%, 88% and 94%, respectively. At 30 days after injection, the average tumor size amounted to 1.13±0.19 cm³ (Co), 0.05±0.02 cm³ (NIPP1-Wt) and 0.36±0.08 cm³ (NIPP1-Pm) (mean±s.e.m, n=16, i.e. two flanks of eight mice). It was verified that the transgenes were expressed in the solid tumors formed after the injection of the NIPP1-Wt (Fig. 7I) and NIPP1-Pm (data not shown) cell lines. Moreover, the NIPP1-Wt-expressing cells showed increased expression levels of Kif18A, NuSAP and TPX2 (Fig. 7I,J), which have previously been shown to be upregulated in mitotically arrested HeLa cells (Beck et al., 2011). Thus, the selective inhibition of PP1 impairs solid-tumor growth, which is associated with a mitotic arrest.

The main goal of this study was to delineate the contribution of PP1 to cell cycle progression in mammalian cancer cells. A knockdown approach was not an attractive option because of the redundancy between PP1 isoforms, which enables PIPs to form functional complexes with other PP1 isoforms once their preferred isozyme is removed (Gibbons et al., 2007; Kirchner et al., 2007; Trinkle-Mulcahy et al., 2006; Varmuza et al., 1999). In addition, none of the available small-molecule, cell-permeable inhibitors of PPP-type phosphatases is specific for PP1, making a chemical genetics approach unfeasible (De Munter et al., 2013; Gehringer, 2004; Kelker et al., 2009). Therefore, we aimed to inhibit the activity of PP1 using cell lines that inducibly express an inhibitory protein. A similar approach has recently been successfully used to explore the importance of PP2A for cell viability (Pores Fernando et al., 2014). Various protein inhibitors of PP1 have been identified but most of these are not suitable for studying global functions of PP1 in intact cells. Indeed, some PP1 inhibitors (e.g. Inhibitor-1 and CPI-17) are phosphorylation-dependent or only inhibit a subset of PP1 holoenzymes (Endo et al., 1996; Eto, 2009). Inhibitor-2, Inhibitor-3 and Sds22 have been biochemically identified as inhibitors of PP1, but genetic data from yeast indicate that they are actually positive regulators, possibly involved in the biogenesis of PP1 (Cheng and Chen, 2015; Eiteneuer et al., 2014; Heroes et al., 2015). In contrast, substantial biochemical and cell biological evidence indicates that full-length NIPP1 is a very specific and extremely potent inhibitor of PP1 (Beullens et al., 1992, 2000; O'Connell et al., 2012; Vulsteke et al., 1997). This applies to all tested protein substrates of PP1, except for phosphoproteins that are recruited through the FHA domain and probably represent the natural substrates of the PP1–NIPP1 holoenzyme. Hence, NIPP1 with a non-functional FHA domain (NIPP1-Fm) can be used as a general and specific inhibitor of PP1. Accordingly, we have found that both NIPP1-Wt and NIPP1-Fm titrate PP1 away from other PP1 holoenzymes (Fig. 5; Fig. S4A,B,D) and that the retained PP1 is inactive towards an exogenous PP1 substrate (Fig. S1C).

A very mild overexpression of NIPP1 in stable HeLa cell lines (<10%), due to promoter leakage from the non-induced transgenes, causes the expression of mesenchymal genes (Van Dessel et al., 2015). However, this low level of overexpression had no effect on the progression through mitosis (Fig. 2). In contrast, the induced overexpression of NIPP1-Wt or NIPP1-Fm to levels that are several-fold higher than that of endogenous NIPP1 caused a mitotic arrest in a large majority of cells. This effect was not seen with a PP1-binding mutant of NIPP1 (NIPP1-Pm) and could be partially rescued by the overexpression of PP1γ (Fig. 6A–D; Fig. S4E,F), which represents strong independent evidence that this phenotype stems from the binding and inhibition of PP1. Interestingly, the mitotic-arrest phenotype could also be rescued by the inhibition of Aurora B kinase (Fig. 6E), consistent with the notion that PP1 balances or reverses signaling by Aurora kinases in mitosis, by directly inactivating the kinases themselves and/or by dephosphorylating their substrates (Emanuele et al., 2008; Lesage et al., 2011; Liu et al., 2010). In general, our data demonstrate that PP1 is essential for progression through M phase in HeLa cells, which is in accordance with previous findings for fungi and Drosophila (Axton et al., 1990; Chen et al., 2007; Doonan and Morris, 1989; Hisamoto et al., 1994). It is possible that major functions of PP1 in other phases of the cell cycle have been masked in our experiments because NIPP1 is a nuclear protein and might not have been able to titrate PP1 away from cytoplasmic PP1 holoenzymes that are essential in interphase.

The prometaphase arrest that was induced by the expression of NIPP1-Wt or NIPP1-Fm was associated with major spindle-assembly and chromosome-congression defects, and a prolonged SAC (Figs 3 and 4). In addition, SAC inactivation was hampered by the inhibition of PP1. Many NIPP1-Wt- or NIPP1-Fm-expressing cells eventually died in mitosis through apoptosis (Fig. 2B,C), also known as ‘mitotic catastrophe’ (Castedo et al., 2004). Other cells exited mitosis and returned to interphase without undergoing cytokinesis, a process known as ‘slippage’ (Gascoigne and Taylor, 2008; Topham and Taylor, 2013). Slippage occurs because the Cdk1 regulator cyclin B1 is slowly degraded, even when the SAC is active, which eventually causes the Cdk1 activity to drop below the threshold level that is needed for maintaining the mitotic state. The balance between mitotic catastrophe and slippage is dependent on the cell type but also shows considerable cell-to-cell variation (Gascoigne and Taylor, 2008; Topham and Taylor, 2013), explaining why the outcome of the expression of NIPP1-Wt or NIPP1-Fm was not the same in all cells. In any case, the severity and complexity of the observed mitotic-arrest phenotype suggests that PP1 has multiple essential substrates in mitosis, which agrees with the identification of multiple PIPs that mediate the targeting of PP1 to the centrosomes, spindle and chromosomes (Bollen et al., 2009; Wurzenberger and Gerlich, 2011). We speculate that HeLa cell lines that inducibly express NIPP1-Fm can be used in rescue experiments to identify the PIPs that are important for spindle formation, microtubule dynamics, chromosome congression and SAC silencing.

The mitotic-arrest phenotypes were generally somewhat stronger for NIPP1-Wt than for NIPP1-Fm. Indeed, the expression of NIPP1-Wt caused more severe chromosome congression defects, as indicated by an increased number of misaligned chromosomes (Fig. 3F,G). It also caused a spindle-pole focusing defect, which was not seen after the expression of NIPP1-Fm. In addition, NIPP1-Wt-expressing cells showed a higher incidence of mitotic catastrophe (Fig. 2B,C). Taken together, these results hint at a contribution of the FHA domain to the phenotype of overexpression of NIPP1-Wt. Importantly, there was no mitotic phenotype after the expression of NIPP1 that was mutated in both its FHA domain and RVxF-type PP1-binding site, indicating that the contribution of the FHA domain is PP1-dependent and cannot be explained by the mere titration of FHA ligands from other protein complexes. Given that NIPP1 enables the dephosphorylation of FHA ligands by associated PP1 in a cellular context, we speculate that the stronger phenotype associated with the expression of NIPP1-Wt is due to the recruitment and subsequent dephosphorylation of FHA ligand(s). It is known that at least some NIPP1 FHA ligands are dephosphorylated at the end of mitosis (Badouel et al., 2010; Boudrez et al., 2000, 2002; Dulla et al., 2010; Minnebo et al., 2013), but it remains to be investigated whether the overexpression of NIPP1 affects the dephosphorylation of its FHA ligands during mitosis and whether this contributes to the mitotic arrest.

Finally, we have demonstrated that the overexpression of NIPP1 also hampers anchorage-dependent and -independent colony formation and tumor growth in xenograft assays in a PP1-dependent manner (Fig. 7). These data suggest that PP1 is a potential target for the development of novel cancer therapies. We envisage that such therapies will not be based on inhibitors of the catalytic subunit, which are likely to be toxic, unless they can be specifically targeted to cancer cells (De Munter et al., 2013). At present there appears to be a better therapeutic potential for molecules that interfere with the assembly or substrate recruitment of key mitotic PP1 holoenzymes, in particular those that are upregulated in cancer cells.

Anti-EGFP, anti-PP1γ and sulforhodamine B were purchased from Santa Cruz Biotechnology. Anti-α-tubulin and anti-RepoMan antibodies, nocodazole, mineral oil and doxycycline hyclate were obtained from Sigma-Aldrich. Anti-GAPDH and anti-cleaved caspase 3 antibodies were purchased from Cell Signaling Technology, and anti-BubR1 and anti-Kif18A antibodies from Bethyl laboratories. The anti-NIPP1 and anti-PP1 antibodies were made in-house, as previously described (Van Dessel et al., 2010). Anti-Sap155, anti-cyclin B1, anti-γ-H2Ax, anti-Aurora B, anti-Aca and anti-Flag antibodies were purchased from MBL International, BD Pharmingen, Millipore, BD Transduction Laboratories, ImmunoVision (Springdale, AR) and Stratagene, respectively. Anti-Ki67 and anti-centrobin antibodies were purchased from Abcam. Anti-NuSAP antibodies were obtained from ProteinTech (Manchester, UK) and anti-TPX2 antibodies from Novus Biologicals (Abingdon, UK). Secondary HRP-conjugated antibodies and anti-Ki67 antibody were purchased from Dako (Heverlee, BE) and secondary Alexa-Fluor-555- and 633-conjugated antibodies were from Invitrogen. S-Trityl-L-Cysteine (STLC), hesperadin, thymidine, reversine and RO-3306 were purchased from Calbiochem, Selleckchem (Munich, Germany), Acros organics (Geel, Belgium), Cayman Chemical and Tocris (Bristol, UK), respectively.

Hela Tet-off cell lines expressing Flag-tagged NIPP1 variants in the absence of doxycycline (Dox) and Flip-In T-REx cell lines expressing EGFP or EGFP-tagged NIPP1 variants in the presence of Dox were obtained as described previously (Tanuma et al., 2008; Van Dessel et al., 2015). Transfections were carried out using a commercial kit (Jet-Prime, Polyplus), according to the manufacturer's instructions. The used Mad2 small interfering RNA (siRNA) has been described previously (Maldonado and Kapoor, 2011). For immunofluorescence studies, cells grown on coverslips were fixed with 4% formaldehyde in Phem buffer (60 mM PIPES at pH 7, 25 mM Hepes pH 7, 10 mM EGTA and 2 mM MgCl₂), permeabilized with 0.5% Triton X-100 and blocked with 3% BSA in PBS. After incubation with primary and secondary antibodies, and DNA staining with DAPI, the coverslips were mounted in Mowiol onto microscope slides. For some experiments, a pre-extraction was performed. In that case, cells were incubated before fixation for 4 min at 4°C with CSK buffer (10 mM PIPES at pH 6.8, 300 mM sucrose, 100 mM NaCl and 3 mM MgCl₂), supplemented with 0.2% Triton X-100. Confocal images were acquired with a Leica TCS SPE laser scanning confocal system mounted on a Leica DMI 4000B microscope and equipped with a Leica ACS APO 40× or 63×1.30 NA oil objective.

For live-cell imaging, cells were grown in chambered coverglass four-well plates from Thermo Scientific. For DIC imaging, cells were synchronized with a double thymidine arrest (2 mM thymidine for 18 h, 9 h release, 2 mM thymidine for 16 h). Dox was added together with induction of the second thymidine block. After release from this arrest the cells were imaged for 60 h with a Leica TCS SPE laser scanning microscope using a 20× magnification objective. Mineral oil was added to prevent evaporation of the medium. For H2B imaging, cells were transfected with Cherry–H2B, synchronized with a single thymidine arrest (2 mM thymidine for 18–20 h) and after release imaged for 20 h using a 40× magnification objective. The SAC override assay has been described previously (Santaguida et al., 2011). The microscope was equipped with a live-imaging chamber ensuring 37°C and 5% CO₂. Images were taken every 10 min (H2B and SAC override assays) or 15 min (DIC time-lapse imaging). Images and movies were processed with the ImageJ software (National Institutes of Health). For quantifying the amount of RFP–PP1 overexpression (Fig. 6B), z-stack scans were performed with wide-field microscopy (12 sections with 4-µm intervals), before live-cell imaging was started. After processing the signals with the ImageJ software using the ‘sum slices’ feature of z-project, a circular region with fixed diameter was centered on randomly chosen cells and raw integrated intensities were measured. After subtraction of the background signal, two thresholds for PP1 expression were determined. Cells harboring RFP intensities below 350,000 were classified as ‘low overexpressing’, above that value, cells were classified as ‘high overexpressing’.

Cell viability in HTO cells (Fig. 7B) was measured with Trypan Blue. For this purpose, HTO cells were seeded in a 24-well plate (3500 cells/well). After 24, 48 and 96 h, Trypan Blue was added and the percentage of colored cells was taken as a measure of cell death. Cell survival in Flp-In T-REx cells (Figs 1C and 6E) was derived from sulforhodamine B colorimetric assays (Voigt, 2005). For these experiments cells were treated with 143 nM hesperadin or DMSO, and the indicated amounts of Dox for 48 h. After fixation and staining with 0.4% sulforhodamine B the absorbance was measured at 550 nm.

FACS analysis was performed as described previously (Kig et al., 2013). Briefly, asynchronous growing cells were induced or non-induced for 48 h. After fixation in 70% ethanol, cells were stained with 50 µg/ml propidium iodide (Sigma) and the DNA content was analyzed by cytometric analysis (Attune Acoustic Focusing Cytometer), using the Attune FACS software. To examine the variability of EGFP expression of the different transgenes (Fig. S1A), cells were induced or non-induced for 24 h. Following fixation with 70% ethanol for 10 min at room temperature, EGFP-expressing cells were quantified with the Attune cytometer, using the Attune FACS software.

Immunoblotting and EGFP-trapping were performed as described previously (Kig et al., 2013; Van Dessel et al., 2010). For the immunoprecipitation of endogenous Ki67, RepoMan and Kif18A, cells were first synchronized with a single thymidine arrest, released for 5 h and incubated for 16 h with 7.5 µM S-Trityl-L-cysteine (STLC). Cells were then lysed in RIPA lysis buffer (50 mM Tris-HCl pH 8, 150 mM NaCl, 1 mM EDTA, 0.5% sodium deoxycholate, 1% Triton X-100 and 0.1% SDS). After centrifugation (10 min at 3000 g), the supernatants were used for the immunoprecipitation with the appropriate antibodies and protein-A–Sepharose (GE Healthcare). The immunoprecipitates were washed once with TBS complemented with 0.1 M LiCl, and three times with TBS containing 0.1% NP-40. Afterwards, precipitates were assayed for the presence of PP1 by measuring the glycogen phosphorylase phosphatase activity (Beullens et al., 1992).

For anchorage-dependent growth, 500 HTO cells were seeded in 10-cm plates. After 2 weeks, the colonies were fixed, stained with Crystal Violet, photographed and counted with a ColCount (Oxford Optronix, Oxford, UK). For anchorage-independent growth, 40,000 HTO cells were first mixed with 2 ml of culture medium containing 0.35% agarose and plated on top of a basal layer of medium containing 0.6% agarose in a 6-well plate. After 1 week, a fresh layer of culture medium containing 0.35% agarose was added. After a further 2 weeks, pictures were taken.

HTO cells were cultured for 3 days in the absence of Dox. Subsequently, 3×10⁶ cells were subcutaneously injected into both flanks of eight athymic mice. Tumor size was measured at the indicated time points with a caliper. Additionally, after 4 weeks the tumors of the three mice closest to the median of each group were visualized with small animal PET (Focus 220 microPET; Concorde-CTI/Siemens) and analyzed as previously described (Deroose et al., 2007). All animal experiments were performed according to approved guidelines.

The statistical significance between control and experimental groups was calculated using paired or unpaired Student's t-tests and is denoted as ***P<0.001; **P<0.01; *P<0.05.

We thank Annemie Hoogmartens and Fabienne Withof for providing technical assistance.