Regulated reconstitution of spindle checkpoint arrest and silencing through chemically induced dimerisation in vivo

Spindle checkpoint signalling was initially reconstituted in Xenopus egg extracts (Kulukian et al., 2009; Minshull et al., 1994) and most recently using recombinant complexes of human checkpoint proteins (Faesen et al., 2017). Major advantages of such in vitro assays are that complex systems can be simplified through biochemical fractionation and manipulated through immunodepletion. They also enable the regulated addition of specific components, whereby the timing, concentration and activity of these can all be varied.

In parallel, yeast genetics has driven the identification of most of the molecular components of this pathway, such as the mitotic arrest deficient (Mad) and budding uninhibited by benzimidazoles (Bub) proteins (Hoyt et al., 1991; Li and Murray, 1991) and their Cdc20 effector (Hwang et al., 1998; Kim et al., 1998). This combination of yeast genetics and in vitro reconstitution has proven invaluable in dissecting the molecular mechanism of action of spindle checkpoint signals and inhibition of the downstream effector Cdc20-APC/C (London and Biggins, 2014; Musacchio, 2015).

Here, we have employed a hybrid approach, using yeast genetics and partial reconstitution of the pathway in vivo. We used synthetic biology to re-wire and simplify the upstream part of the checkpoint signalling pathway and chemically induced dimerisation (CID) to add an extra level of regulation that can be easily controlled experimentally in intact cells. Employing this strategy, we were able to achieve the following outcomes. (1) We simplified the system through regulated, ectopic activation of the spindle checkpoint, enabling kinetochore-independent studies. (2) We used yeast genetics to enable rapid iterative analyses. (3) We employed synthetic biology and CID to generate specific complexes in an experimentally controlled fashion. (4) We used abscisic acid (ABA) addition and wash-out to provide tight temporal control of the initiation and termination of checkpoint signalling.

More specifically, we generated a synthetic checkpoint arrest ectopically by inducing heterodimerisation of the checkpoint proteins Mph1 (the fission yeast homologue of Mps1) and Spc7 (the fission yeast homologue of KNL1)  in fission yeast. This led to checkpoint arrest in a synchronous population of cells within 30 min of addition of the plant phytohormone ABA. As expected, this checkpoint response required the downstream Mad and Bub factors. To analyse silencing of the checkpoint, we simply washed out the ABA from arrested cells and analysed mitotic exit. We found that the kinetics of release was critically dependent on recruitment of protein phosphatase 1 (PP1) to the Mph1-Spc7 signalling platform.

We previously published a synthetic checkpoint arrest assay (SynCheck) in which we activated the spindle checkpoint in fission yeast using heterodimers of TetR-Spc7 and TetR-Mph1 kinase (Yuan et al., 2017). However, in those experiments, dimerisation was constitutive, being driven by formation of Tet repressor dimers (TetR). Thus, checkpoint signalling was challenging to regulate, both in terms of initiation and termination. We controlled checkpoint arrest at the transcriptional level using an nmt promoter to drive expression of the TetR-Mph1 fusion protein. Unfortunately, the fission yeast nmt1 promoter requires induction in medium lacking thiamine for several hours. As a consequence, the peak of arrest was observed ∼14 h after induction and was not as synchronous as hoped. To improve both timing and control, we modified our approach by employing CID to give tight temporal control over the initiation and termination of checkpoint signalling.

Following the strategy of Crabtree and colleagues (Liang et al., 2011), we fused the PYL domain (residues 33-209) of the ABA receptor after the N-terminal 666 amino acids of fission yeast Spc7. By deleting the C-terminal half of Spc7, this protein is unable to be targeted to kinetochores because it lacks the Mis12-interacting region (Petrovic et al., 2016; Petrovic et al., 2014). This fusion protein was expressed from the constitutive adh21 promoter (Tanaka et al., 2009). The ABI domain (residues 126-423) of ABI1 was fused to the C-terminus of the Mph1 spindle checkpoint kinase. We also deleted the first 301 amino acids of Mph1 to prevent it going to kinetochores (Heinrich et al., 2012). This Mph1-ABI fusion protein was expressed from the adh41 promoter (Tanaka et al., 2009). In the presence of ABA, the PYL and ABI domains are sufficient to form a tight complex (Miyazono et al., 2009), thus forming complexes of Mph1-ABI and Spc7-PYL (Fig. 1A). We combined these constructs in a strain that also had the cdc25-22 mutation, enabling synchronisation in G2, the Bub1 checkpoint protein tagged with GFP and microtubules labelled with mCherry-Atb2 (α-tubulin).

Cells were synchronised in G2 using a temperature-sensitive cdc25-22 mutant that blocks cells in G2 after 3.5 h at 36°C. When cells were shifted to 25°C, they were ‘released’ from the block, enabling progression through the cell cycle. After 5 min, ABA was added to activate the spindle checkpoint through the formation of Spc7-PYL and Mph1-ABI heterodimers (Fig. 1B). We observed that 60 min after ABA addition to the synchronous population of cells, over 70% of cells had short metaphase spindles (Fig. 1C,D). The metaphase arrest could be sustained for at least 4 h (Fig. 1D). We tested a range of ABA concentrations (0-500 µM) and found that 250 µM was optimal for reproducible, robust arrests (Fig. S1A). The ABA can be added later (e.g. 20 min after cdc25 release) and cells arrest with similar efficiency to that observed after anti-microtubule drug treatment with carbendazim (see Fig. S1B). Without pre-synchronisation in G2, the mitotic index increases over time and reaches a peak 4 h after ABA addition (Fig. S1C). In our previous SynCheck studies, cells arrested for several hours but then died (Yuan et al., 2017). We wanted to determine whether the ABA arrest also had a significant effect on cell viability or whether our ability to release this arrest (through ABA wash-out) meant that viability was maintained. After ABA treatment, we found a gradual drop in cell viability (see Fig. 2E), which was similar to that observed upon anti-microtubule drug treatment (data not shown).

In the arrested cells, we observed Bub1 enrichment at the spindle poles (Fig. 1C). This is consistent with our previous SynCheck assay, where movement of all spindle checkpoint proteins to spindle poles was reported to be Mad1-Cut7 kinesin driven (Yuan et al., 2017). As expected, deleting the first N-terminal coiled coil (136 amino acids) of Mad1, required for its interaction with Cut7 (Akera et al., 2015), prevented Bub1 accumulation at spindle poles. This de-localisation of checkpoint proteins from spindle poles did not affect the efficiency of the arrest (Fig. S1D), as found in SynCheck (Yuan et al., 2017).

ABA-induced metaphase arrest is dependent on heterodimerisation of Spc7-PYL and Mph1-ABI. Strains lacking either the Mph1-ABI component or the Spc7-PYL component failed to arrest in the presence of ABA (Fig. 1E). Deleting the downstream checkpoint protein Mad1 abolished the arrest (Fig. 1E), showing that ABA-induced arrest is checkpoint dependent. In these constructs, Spc7 and Mph1 lack their kinetochore-binding domains, making initiation of this arrest ectopic and independent of the complexities of the kinetochore. The Mph1-ABI, Spc7-PYL strain used above lacks endogenous mph1, which prevents all Mad and Bub checkpoint proteins from targeting to kinetochores (Heinrich et al., 2012). As an additional measure, to confirm kinetochore independence, we employed a strain containing the spc7-12A MELT mutant allele (Mora-Santos et al., 2016; Yamagishi et al., 2012). This mutant Spc7 kinetochore component cannot be phosphorylated by Mph1, preventing recruitment of Bub3-Bub1, and thereby Mad1-Mad2 complexes, to kinetochores. The spc7-12A mutant arrested with very similar efficiency to spc7+ cells under ABA control (Fig. S1F), indicating that the Spc7wt-PYL Mph1-ABI heterodimer does not need to be aided by endogenous kinetochore-based checkpoint signalling to generate a checkpoint arrest. Importantly, spc7-12A-PYL fusion protein was unable to generate an arrest in combination with Mph1-ABI, demonstrating that the ectopic signalling scaffold needs to be phosphorylated on conserved Spc7 MELT motifs to recruit Bub3-Bub1 complexes for active signalling (Fig. S1G).

Crucial consequences of checkpoint action are the stabilisation of cyclin B and securin. Using a modified strain, we analysed cyclin B (Cdc13) levels in the ABA-induced arrest. Fig. 1F shows that Cdc13-GFP accumulated on short metaphase spindles and was enriched at mitotic spindle poles, as expected. As a technical aside, we found that different tags can affect the efficiency of the ABA-induced arrest. For example, this Cdc13-GFP strain reproducibly arrests more efficiently than the strain containing Bub1-GFP (Fig. 1G). This is probably a result of a partial loss of function when C-terminally tagging the Bub1 checkpoint protein. The Cdc13-GFP strain also contains the endogenous wild-type Mph1 gene, but we found that this did not significantly impact the efficiency of arrest (see Fig. S1E).

Thus, we have reconstituted a robust, kinetochore-independent checkpoint arrest that can be initiated very simply in vivo through ABA addition to culture media. This works efficiently in both minimal (PMG) and rich (YES) fission yeast growth media. Hereafter, we refer to this assay as SynCheckABA.

A significant advantage of SynCheckABA is the ability to reverse the effects of ABA by simply washing cells with fresh medium lacking ABA and thereby releasing them from metaphase arrest (Fig. 2A,B). We can use this to study spindle checkpoint silencing, which has proven to be technically challenging in the past. Fig. 2C,D demonstrates that washing out the ABA results in rapid cyclin degradation and spindle elongation (see also Fig. S2A).

Previous work has shown that PP1 (Dis2) is a key spindle checkpoint silencing factor in yeasts (Meadows et al., 2011; Pinsky et al., 2009; Vanoosthuyse and Hardwick, 2009). The N-terminus of Spc7 has two conserved motifs (SILK and RRVSF, also referred to as the A and B motifs) that mediate stable PP1 association (Fig. 3A). Mutation of both binding sites leads to a lethal metaphase block in Saccharomyces cerevisiae and Schizosaccharomyces pombe (Meadows et al., 2011; Rosenberg et al., 2011). There are additional kinetochore-binding sites for PP1 such as Klp5 and Klp6 (Meadows et al., 2011) and these are relevant to checkpoint silencing, although binding to Spc7 appears to be the major player. In human cells, similar motifs are found at the N-terminus of KNL1; PP1 binding is regulated by Aurora B activity as this kinase can directly phosphorylate the B motif, disrupting PP1 association (Liu et al., 2010).

Employing SynCheckABA, we tested mutations of the A and B motifs at the N-terminus of Spc7 and removal of the Klp6 kinesin. For these experiments (Figs 3 and 4), all strains contained endogenous wild-type Mph1 kinase and thus were able to recruit checkpoint proteins to their kinetochores. These include the Mph1 and Bub1 kinases, which are also thought to also have ‘error correction’ functions. Thus, silencing probably needs to take place not only at the ectopic Mph1-Spc7 signalling scaffold, but also at kinetochores. Strains were pre-synchronised in G2 using cdc25, released and arrested at metaphase using ABA, and then washed to terminate checkpoint signalling. Progression through anaphase was scored through analysis of spindle elongation and/or cyclin B degradation (using Cdc13-GFP) over a 90 min time course. Mutation of the A motif delayed spindle elongation by 30 min and the A/B double mutant was delayed even more profoundly (Fig. 3B,C). This indicates that PP1 activity on or near the Spc7 protein (previously phosphorylated by Mph1 kinase) is a limiting factor in checkpoint silencing. This system will prove useful for dissecting the regulation of PP1 binding to Spc7 in more detail, and for analysis of putative regulators of PP1 activity.

Mutation of fission yeast kinesin 8 (either Klp5 or Klp6) leads to stabilisation of microtubules, aberrant chromosome movements and long metaphase spindles (Gergely et al., 2016; Klemm et al., 2018; Meadows et al., 2011; West et al., 2002). In these mutants, checkpoint silencing defects cannot simply be analysed through spindle elongation. Instead, we imaged Cdc13-GFP and used the decrease in the number of cells with cyclin B enriched at their spindle poles as a measure of checkpoint silencing. Fig. 4A,B demonstrates that deletion of Klp6 significantly reduces the efficiency of silencing and cyclin B degradation. Finally, we analysed the silencing defect upon deletion of PP1 phosphatase (dis2Δ). In the dis2Δ strain, which is rather sick, checkpoint silencing was extremely defective with no significant drop in Cdc13-GFP levels over the 90 min time course (Fig. 4C). It should be noted that these dis2Δ strains display significant mitotic delays, even in the absence of ABA addition, presumably because the lack of this mitotic phosphatase leads to pleiotropic mitotic defects (see Fig. S4 for images of these cells).

Thus, SynCheckABA neatly recapitulates the balance of opposing kinase and phosphatase activities between Mph1-dependent checkpoint activation and PP1-driven checkpoint silencing on Spc7 and kinesin 8-dependent pathways (see general model in Fig. 4D).

Here, we employed CID to generate a rapid, controlled spindle checkpoint arrest. Addition of ABA to SynCheckABA strains induces the heterodimerisation of Mph1-ABI and Spc7-PYL fusion proteins and this is sufficient to generate an activated signalling scaffold and metaphase arrest within minutes. Like our original SynCheck assay, which was driven by constitutive TetR homodimers (Yuan et al., 2017), this arrest acts independently of spindle checkpoint signalling at endogenous kinetochores, but is dependent on downstream checkpoint components such as Mad1.

A significant advantage of SynCheckABA is that we can wash out the ABA and study the kinetics and mechanism of spindle checkpoint silencing. This was not possible with the original SynCheck strain as we were unable to control TetR dimerisation and thus unable to dissociate the Mph1-TetR–Spc7-TetR signalling scaffold.

Using this new assay, we confirmed that PP1 is crucial for silencing the Mph1-Spc7 scaffold (Figs 3 and 4). PP1 binds to the N-terminus of Spc7, not far from the conserved MELT motifs that, once phosphorylated by Mph1, bind Bub3-Bub1 complexes to initiate generation of the mitotic checkpoint complex (MCC) (Shepperd et al., 2012). Thus, Spc7 acts as the platform for both checkpoint activation and silencing and appears to be a major site of action for both checkpoint activation kinases and silencing phosphatases (Meadows et al., 2011). It is important to note that not all aspects of silencing are recapitulated in our ectopic assay, as some of these relate to specific kinetochore processes that are not captured.

Kinesin 8 is also confirmed as a PP1 recruitment site relevant for checkpoint silencing in SynCheckABA. The phenotypes of the klp6Δ mutant suggest that targeting of PP1 to spindle microtubules and kinetochores is also relevant to mitotic exit from an ABA-induced arrest, even though the arrest is initiated away from the kinetochore (see Fig. 4D).

We believe that all forms of spindle checkpoint reconstitution are useful for mechanistic dissection of this dynamic signalling pathway, whether this be in vitro within cytoplasmic extracts (Minshull et al., 1994), in vitro with purified recombinant proteins (Faesen et al., 2017) or in vivo with synthetically re-wired and simplified signalling pathways (SynCheckABA). The advantages of the latter system are as follows.

(1) The signalling pathway downstream of Spc7 and the downstream effectors are present at normal physiological levels and there are simple, quantitative physiological read-outs (cyclin B degradation, sister chromatid separation and/or anaphase spindle elongation).

(2) Checkpoint arrest is induced in the absence of additional stresses; simple addition of ABA (low toxicity) to the growth media is sufficient for checkpoint activation. There is no need for cold shock (to depolymerise tubulin, nda3 arrest), heat shock (to perturb temperature-sensitive kinetochore mutants) or overexpression of checkpoint activators.

(3) The PYL and ABI domains have limited cross-reaction in yeast as they are derived from plant proteins. Although we have not compared them directly, we believe that ABA has certain advantages over the use of rapamycin, a very popular CID. To use rapamycin in fission yeast one needs to engineer strains to remove endogenous rapamycin-binding proteins, such as by deleting the fkh1+ gene that encodes a native FKBP12 domain (Ding et al., 2014). Importantly, because ABA does not bind tightly to the PYL domain, we can wash ABA out easily to initiate checkpoint silencing. By comparison, rapamycin is very difficult to wash out, making efficient release experiments unrealistic.

(4) Compared to in vitro studies with large, recombinant complexes, these fission yeast experiments are simple, cheap and fast. The system also enables rapid iterative studies, because of the ease of further genetic manipulation in yeast.

(5) Importantly, we can easily test candidate regulators (e.g. silencing factors) without needing to know what complexes they are part of, purifying them and worrying about their relevant concentration and post-translational modifications.

(6) Compared with our transcriptionally controlled SynCheck (which employs nmt-tetR-Mph1), the ABI-PYL system is less leaky, enabling sick strains (such as the dis2 mutant analysed in Fig. 4) to be constructed. Previously, we were unable to isolate nmt-tetR-Mph1, dis2Δ strains because of leaky expression from the weak nmt81 promoter.

Our ongoing studies with SynCheckABA will enable a detailed mechanistic dissection of PP1-mediated spindle checkpoint silencing in fission yeast. We believe that ABA holds promise as an alternative CID to rapamycin and that it has significant advantages.

Mph1 (residues 303-678) was amplified from a pDONR 201 plasmid containing Mph1 (303-678) (Yuan et al., 2017). 3×HA was amplified from a plasmid from the Allshire laboratory (University of Edinburgh) containing codon-optimised PYL-3×HA. ABI was amplified from a pMT_CID_ABI_VS_H vector from the Patrick Heun laboratory (University of Edinburgh). These PCR fragments were treated with Dpn1 and assembled into a Sma1-digested and antarctic phosphatase-treated gel-purified pRad41 yeast expression vector by Gibson assembly.

The yeast expression vector pIY03 (Yuan et al., 2017) was digested with Nhe1 and Xho1 and gel purified. The insert (mCherry-2×FLAG-Spc7_1-666) was used as a template to amplify Spc7_1-666. PYL was amplified from a bVNI-221 vector from the Heun laboratory. The fragments were then assembled into the digested pIY03 vector backbone using Gibson assembly.

Plasmids containing full-length Spc7 PP1-binding mutants (ΔA, deletion of residues 136–150; ΔAB, deletion of residues 136-150 and residues 331–345) (provided by the Millar laboratory, University of Warwick) were used as templates to amplify mutant versions of Spc7_1-666. NheI-NLS and PacI sites were introduced during amplification, allowing Spc7 constructs to be digested and ligated into digested pIY03-derived vector backbone, which also contained a C-terminal mCherry-2×FLAG-PYL tag.

The fission yeast strains used are listed in Table S1.

Cells were grown at 25°C for 1-2 days on YES (rich yeast media, with additional leucine, arginine, lysine, histidine and uracil) plates. They were then pre-cultured in 10 ml of liquid YES containing amino acid supplements at 25°C over the day and inoculated into a larger culture of YES overnight. The following day, log phase cultures were shifted to 36°C for 3.5 h to block in G2. After this, cultures were cooled quickly in iced water to rapidly shift them back to 25°C and release them from the G2 block.

Following a cdc25-22 block, 250 mM ABA stock (Sigma Aldrich A1049) was added to cultures 5 min after release (20 min if comparing to a carbendazim arrest) to achieve a final concentration of 250 μM (unless otherwise stated).

Following an ABA-induced synthetic arrest, the cells were washed three times with 50 ml YES.

Culture (1-1.5 ml) was centrifuged for 1 min at 6000 rpm. The cell pellet was fixed in 200–500 μl of 100% ice-cold methanol. To image cells, 8 μl of the cell suspension in methanol was added to a glass slide; when the methanol evaporated, 1-2 μl DAPI (0.4 μg/ml) was added to the sample and a glass cover slip was placed on top.

Cells were imaged immediately using a 100× oil immersion lens and a Zeiss Axiovert 200M microscope (Carl Zeiss), equipped with a CoolSnap CCD camera (Photometrics) and Slidebook 5.0 software (3i, Intelligent Imaging Innovations). Typical acquisition settings were 300 ms exposure (FITC and TRITC) and 100 ms exposure (DAPI), 2× binning, Z-series over 3 μm range in 0.5 μm steps (seven planes).

Following a cdc25-22 block, 3.75 mg/ml stock of carbendazim (Sigma Aldrich) was added to cultures 20 min after release to achieve a final concentration of 100 μg/ml.

Following a synthetic arrest assay, cells from 1 ml of culture were harvested by centrifugation at 6000 rpm for 1 min and re-suspended in 1 ml of distilled water. Tenfold serial dilutions were made in distilled water. Cells were diluted by factors of 100 and 1000, and 0.1 ml plated in triplicate. Colony forming units (cfu) per millilitre of culture was calculated and cell viability over time was plotted as a percentage relative to that at time zero.

We would like to thank Patrick Heun and Eftychia Kyriacou for providing constructs containing the ABA-binding heterodimerisation domains of PYL and ABI; Jonathan Millar for the PP1-binding spc7 mutants and the klp6 mutants; and Ken Sawin for the mCherry-Atb2 strain.