Nek2A destruction marks APC/C activation at the prophase‐to‐prometaphase transition by spindle‐checkpoint‐restricted Cdc20

The anaphase‐promoting complex/cyclosome (APC/C) is an E3 ubiquitin ligase that, together with either one of its regulatory co‐activators, Cdc20 or Cdh1, targets multiple mitotic regulators for proteasomal degradation. These include cyclin B1, securin and geminin, making APC/C^Cdc20 a major factor in directing cell division, sister chromatid separation and DNA replication licensing (Clijsters et al., 2013; Peters, 2006; Pines, 2011). Several questions remain about how the activity of APC/C^Cdc20 is controlled in mitosis. Phosphorylation of the APC/C by mitotic kinases at the end of prophase leads to an increased affinity for Cdc20 (Kramer et al., 2000; Yudkovsky et al., 2000). The formation of the complex between the APC/C and co‐activator probably induces a conformational change that activates the APC/C (Dube et al., 2005; Kimata et al., 2008), perhaps by facilitating the recruitment of the E2 enzyme UbcH10 (Chang et al., 2014; Van Voorhis and Morgan, 2014). Cdc20 also acts as an APC/C substrate recruitment factor that binds directly to degradation motifs in APC/C substrates, such as the D‐box and the KEN box (da Fonseca et al., 2011; Kraft et al., 2005). At the point in the cell cycle when APC/C^Cdc20 complexes are formed, however, the spindle checkpoint also becomes active and blocks Cdc20. Spindle checkpoint proteins, including Mad2 and BubR1, capture Cdc20 into the inhibitory mitotic checkpoint complex (MCC) (Chao et al., 2012). Cdc20 remains inhibited by the spindle checkpoint until the chromosomes are bi‐oriented on the mitotic spindle (Foley and Kapoor, 2013; Kim and Yu, 2011; Lara‐Gonzalez et al., 2012). Once the spindle checkpoint is satisfied, APC/C^Cdc20 becomes active and sends cyclin B1, securin and geminin for proteasomal degradation (Clijsters et al., 2013; Clute and Pines, 1999; Hagting et al., 2002). Interestingly, however, the APC/C^Cdc20 substrate cyclin A2 disappears shortly after nuclear envelope breakdown, regardless of the inhibitory effect of the spindle checkpoint (den Elzen and Pines, 2001; Geley et al., 2001).

The mechanism by which cyclin A destruction evades the spindle checkpoint has largely been solved. The N‐terminus of cyclin A associates strongly with Cdc20 and thereby competes off the spindle checkpoint proteins (Di Fiore and Pines, 2010; van Zon et al., 2010; Wolthuis et al., 2008). Thus, cyclin A, by its N‐terminus, binds a specific fraction of Cdc20 that cannot be blocked by Mad2 and BubR1. In addition, the Cdc20–cyclin‐A complex, bound to Cdk1 and Cks, is exclusively recruited to the APC/C in prometaphase, when the APC/C becomes phosphorylated (Wolthuis et al., 2008). Recently, it has been shown that cyclin A destruction early in mitosis is necessary for progressive stabilization of the mitotic spindle, promoting proper attachments to kinetochores and formation of the metaphase plate (Kabeche and Compton, 2013).

Nek2 isoform A (Nek2A) is a centrosomal kinase that is highly expressed in G2 phase but rapidly disappears in prometaphase. Nek2A phosphorylates, for instance, C‐Nap and Rootletin, which are involved in centrosome separation and bipolar spindle formation (Bahe et al., 2005; Bahmanyar et al., 2008; Fry et al., 1998a; Fry et al., 1998b), but more recently it has also been implicated in the Hippo signalling pathway (Mardin et al., 2010). Although Nek2A is an APC/C substrate, conclusive evidence that its destruction in mammalian cells depends only on APC/C^Cdc20, or that a different proteasomal targeting pathway contributes to its degradation, too, is lacking. Furthermore, the role of Cdc20 in directing APC/C‐mediated Nek2A degradation is under debate (Kimata et al., 2008; Sedgwick et al., 2013). In contrast to cyclin A, even at high levels Nek2A, has not been found to interfere with the ability of BubR1 to bind Cdc20 (Sedgwick et al., 2013), indicating that Nek2A and cyclin A might differ in the way their destruction escapes control by the spindle checkpoint. Because the spindle checkpoint might block the recruitment of substrates to the APC/C by Cdc20, an attractive model explaining the timing of Nek2A degradation is that its destruction depends only on the APC/C, not on Cdc20. An observation in support of this model is that Nek2A has a C‐terminal MR tail that binds directly to the TPR motifs of APC/C subunits (Hames et al., 2001; Hayes et al., 2006; Sedgwick et al., 2013). However, in such a model, Nek2A binding to the APC/C would be expected to be regulated by the cell cycle, to explain its timely destruction. Furthermore, a TPR‐binding tail is, for instance, also present in the stable APC/C component APC10, showing that this motif alone is insufficient to turn a protein into an APC/C substrate (Wendt et al., 2001; Vodermaier et al., 2003; Matyskiela and Morgan, 2009). Nek2A forms dimers that facilitate Nek2A binding to the APC/C (Sedgwick et al., 2013), but dimerization is also not regulated by the cell cycle (Fry et al., 1999). Taken together, therefore, it is unclear which mechanism ensures that Nek2A is degraded at the right time in mitosis and what the role of Cdc20 is in this process. Here, we tried to address this by asking whether Nek2A turnover relies exclusively on the APC/C and Cdc20, and, how Nek2A degradation escapes control by the spindle checkpoint. We analyzed Nek2A degradation in live cells in relation to two well‐characterized APC/C^Cdc20 substrates, geminin, which is stabilized in response to the spindle checkpoint, and cyclin A, which is degraded independently of the spindle checkpoint.

As detected by western blotting, Nek2A is degraded when cells are arrested in mitosis by taxol treatment (Fig. 1A). We wanted to know whether, as was reported recently for cyclin A, Nek2A might be partially stabilized by increasing the formation of the Cdc20‐inhibitory MCC, a consequence of treating mitotic cells with spindle poisons (Collin et al., 2013; Westhorpe et al., 2011). To follow detailed changes in protein stability over time, we used time‐lapse fluorescence microscopy of U2OS cells expressing geminin–Cherry, a validated checkpoint‐controlled APC/C^Cdc20 substrate (Clijsters et al., 2013), together with an N‐terminally tagged Venus–Nek2A fusion, during G2 phase and mitosis (Fig. 1B). Upon nocodazole treatment, geminin–Cherry remained stable as long as cells delayed in mitosis (Fig. 1C). However, fluorescent Nek2A was destroyed right at the prophase‐to‐prometaphase transition, regardless of whether cells were left untreated or were blocked by either nocodazole or taxol (Fig. 1C,D; supplementary material Fig. S1A shows the expression levels of the fluorescent Nek2; also see Fig. 3B, below). We conclude therefore that Nek2A differs from other APC/C^Cdc20 substrates, including cyclin A, in that its degradation is not delayed at all by increasing spindle checkpoint activity. These results indicate that Nek2A is either not exclusively degraded by APC/C^Cdc20, or that Nek2A destruction occurs regardless of whether Cdc20 is blocked by spindle checkpoint proteins or not.

To resolve this matter, we first investigated whether Nek2A degradation exclusively depended on Cdc20. We used time‐lapse fluorescence microscopy to follow cells stably expressing both geminin–Cherry and Venus–Nek2A, after treatment with small interfering RNA (siRNA) directed against Cdc20. In control cells, Venus–Nek2A destruction started right at nuclear envelope breakdown (NEBD) [Fig. 2A, top panel, Fig. 2B; fluorescent Nek2A protein levels reached 50% of their NEBD levels within 15.6±6.1 min (mean±s.d.), n = 25, in three independent experiments]. Although the mitotic delay after Cdc20 siRNA varied between cells, we found that cells arresting in mitosis for 2 h or more did not degrade geminin–Cherry. Remarkably however, fluorescent Nek2A was degraded only slightly more slowly (Fig. 2A, middle panel, Fig. 2C). In these cells, the point when 50% of fluorescent Nek2A had disappeared was delayed to 28.5±12.0 min (mean±s.d., n = 47, five independent experiments, Fig. 2C). We then investigated the effect of the APC/C inhibitor ProTAME, which blocks normal binding of Cdc20 to the APC/C (Zeng and King, 2012; Zeng et al., 2010). Although treatment with 20 µM of ProTAME almost completely stabilized geminin–Cherry, we observed only modest stabilization of Venus–Nek2A, roughly similar to the partial stabilization of Venus–Nek2A following Cdc20 siRNA (half‐life 41±12.6 min; mean±s.d., n = 15 from two independent experiments, Fig. 2A, bottom panel; Fig. 2D). When depleting the cullin‐like subunit APC2, NEBD to anaphase lasted 489.1±200 min (n = 13 from three independent experiments). In addition, in these cells, geminin–Cherry was clearly stabilized, confirming efficient depletion of the APC2 subunit (Fig. 2E, western blot inset). Nevertheless, Venus–Nek2A was still degraded effectively (Fig. 2E; time to 50% of the Venus–Nek2A levels at NEBD was 17.8±3.6 min, mean±s.d.). Significant stabilization of Venus–Nek2A was not observed in APC3 siRNA cells or after the combined knockdown of Ube2S and UbcH10, even though geminin–Cherry was largely stable in these cells (supplementary material Fig. S1B,C, respectively). Endogenous Nek2A also disappeared despite depletion of APC subunits or the APC/C‐directed E2 enzyme Ube2S (supplementary material Fig. S1D). Hence, Nek2A degradation proceeds even when the function of APC/C^Cdc20 is significantly impaired. This indicates that a second, APC/C^Cdc20‐independent, pathway targets Nek2A under these conditions. Alternatively, a catalytic amount of APC/C^Cdc20, remaining after either APC subunit or Cdc20 depletion by RNAi, or after pharmacological inhibition of APC/C^Cdc20, is sufficient to effectively process Nek2A.

We then directly compared the degradation of Nek2A to that of the spindle‐checkpoint‐independent APC/C^Cdc20 substrate cyclin A in live cells. We made use of tetracyclin‐inducible cyclin‐A–Venus U2OS cells stably expressing Cherry–Nek2A. During unperturbed mitosis, or after nocodazole treatment, Nek2A degradation started several minutes before that of cyclin A, exactly at the point of NEBD, as determined by the abrupt appearance of cytoplasmic cyclin A (Fig. 3A, top panel, Fig. 3A,B, destruction plots). Nek2A degradation progressed more rapidly than that of cyclin A (Fig. 3A, destruction plot). Typically, we had found that Cdc20 needs to be depleted to below 5% of its normal cellular levels for cyclin A stabilization (Wolthuis et al., 2008). In a Cdc20 RNAi experiment where the time from NEBD to anaphase was 183 min ±72.9 s.d., and 50% of cyclin‐A–Venus remained after 74 min of mitotic delay, Cherry–Nek2A was only minimally stabilized (Fig. 3C, supplementary material Fig. S2A). In another experiment that led to more severe Cdc20 depletion, NEBD to anaphase lasted more than 12 h (766.5±152.3 min, mean±s.d., supplementary material Fig. S2B). However, although these Cdc20 siRNA cells failed to degrade cyclin A for the first 120 min of mitosis, Nek2A still declined rapidly (Fig. 3D). Apparently, depleting Cdc20 affects cyclin A destruction much more than Nek2A destruction (Fig. 3C,D; supplementary material Fig. S2A,B). Nevertheless, the degree to which Nek2A was stabilized correlated with the degree of cyclin A stabilization upon Cdc20 depletion. Similar results were obtained when we followed cells depleted for either APC2 or the combination of E2 enzymes, Ube2S and UbcH10 (supplementary material Fig. S2C,D). These results again suggest that Nek2A can either be processed independently of the APC/C or Cdc20, or that a very small amount of Cdc20, remaining after siRNA treatment, is sufficient to support Nek2A degradation. To fully block the function of Cdc20, we then combined Cdc20 siRNA with proTAME, which act synergistically (Zeng et al., 2010). Interestingly, both cyclin‐A–Venus and Cherry–Nek2A became completely stable during prometaphase (compare Fig. 3E, proTAME alone, with Fig. 3F, proTAME plus Cdc20 siRNA). ProTAME, a cell‐permeable compound that resembles an IR tail, did not interfere with the recruitment of Nek2A to the APC/C (supplementary material Fig. S3A,B, and see below). This shows that Nek2A destruction in mitosis fully depends on binding of Cdc20 to the APC/C. We propose that, although processing of cyclin A by the APC/C requires stoichiometric cyclin‐A–Cdc20 complexes, Nek2A degradation is directed by a catalytic effect of Cdc20 on the APC/C that immediately springs into action at the prophase‐to‐prometaphase transition.

Previous in vitro work has shown that Nek2A can bind directly to the APC/C even in the absence of Cdc20 (Hayes et al., 2006). To explain the sudden disappearance of Nek2A when cells enter mitosis, we hypothesized that Nek2A recruitment to the APC/C might be regulated by the cell cycle. We compared binding of Nek2A to the APC/C in extracts from cells synchronized in G2 phase or in mitosis. To stabilize Nek2A, we arrested cells in nocodazole and added the proteasome inhibitor MG132. Surprisingly, Nek2A bound strongly to the APC/C in G2 phase, as well as in mitosis (Fig. 4A,B, APC4 immunoprecipitations; supplementary material Fig. S3C shows validation of the specificity of the detected Nek2A and Nek2B bands). Low levels of Nek2A protein were also detected in Cdc20 immunoprecipitations, together with the APC/C (Fig. 4B, Cdc20 immunoprecipitations). Whereas Mad2 bound predominantly to mitotic APC/C^Cdc20, Nek2A similarly bound G2‐phase or mitotic APC/C^Cdc20 (Fig. 4B). Apparently, and in contrast to cyclin A, Nek2A is recruited to the APC/C in interphase, before it gets degraded in mitosis. Furthermore, APC/C‐binding of Nek2A occurred independently of Cdc20 or Cdh1 (Fig. 4C; Hayes et al., 2006; Kimata et al., 2008). Cdc20 or Cdh1 depletion did not affect binding of Nek2A to the G2‐phase APC/C, indicating there is no competition between co‐activators and Nek2A for APC/C binding, nor is there a clear stimulatory effect of the co‐activators on recruitment of Nek2A to the APC/C (Fig. 4C). This shows that degradation of Nek2A is not initiated by its increased binding to APC/C^Cdc20 and implies that the start of Nek2A degradation, which we show is entirely APC/C^Cdc20 dependent, reflects the exact moment when Cdc20 activates the APC/C.

For its timely degradation, cyclin A needs to compete Mad2 and/or BubR1 away from Cdc20 (Wolthuis et al., 2008; van Zon et al., 2010; Di Fiore and Pines, 2010). However, in nocodazole‐arrested cells treated with MG132 after mitotic shake‐off, we found that Nek2A is in complex with Cdc20 as well as Mad2 (Fig. 4D). Furthermore, in BubR1 immunoprecipitations of mitotic cells treated with nocodazole and, treated with MG132 after mitotic shake‐off, we detected APC/C, Cdc20 and Nek2A (Fig. 4E,F). This is in agreement with earlier in vitro experiments showing that Nek2A does not interfere with BubR1–Cdc20 complex formation (Sedgwick et al., 2013). Although both Nek2A and checkpoint proteins bound to the APC/C, only a small amount of Nek2A re‐accumulated on the MCC‐bound APC/C during the course of MG132 treatment. Nek2A will probably also bind apo‐APC/C (compare Fig. 4E, APC4 immunoprecipitation versus BubR1 immunoprecipitation).

We conclude that the mechanisms by which cyclin A and Nek2A escape stabilization by the spindle checkpoint are most likely different, for the following reasons: (1) Nek2A starts to be degraded exactly at the prophase‐to‐prometaphase transition, which in most experiments, is detectable several min before cyclin A starts to decline – this difference might be explained by the special dependence of cyclin A destruction on competition between spindle checkpoint proteins and cyclin A for Cdc20 binding; (2) in contrast to that of cyclin A, Nek2A destruction is completely insensitive to increased MCC formation induced by nocodazole treatment; (3) Nek2A destruction, but not cyclin A destruction, proceeds effectively under conditions of ∼95% Cdc20 depletion, or after 20 µM proTAME treatment; (4) although cyclin A degradation was found to depend on a competition mechanism between cyclin A and BubR1, required to liberate Cdc20 (Di Fiore and Pines, 2010), Nek2A, even at high concentrations, does not compete for BubR1 binding to Cdc20 in vitro (Sedgwick et al., 2013); indeed, here we show that Nek2A can form complexes with BubR1‐inhibited APC/C; and (5) Nek2A bound to the APC/C in G2 phase, prior to its destruction in mitosis, whereas cyclin A, in complex with Cdc20, was only recruited to the APC/C from prometaphase onwards, when it is also degraded (Wolthuis et al., 2008).

Interestingly, in in vitro APC/C ubiquitylation assays, Cdc20 that is part of the MCC also has a small positive effect on APC/C activity (Herzog et al., 2009; Izawa and Pines, 2015; Kelly et al., 2014). In addition, autoubiquitylation of Cdc20 occurs while the checkpoint is actively inhibiting the APC/C, showing that the APC/C, in principle, can target its substrates regardless of being bound to the MCC (Foster and Morgan, 2012; Ma and Poon, 2011; Nilsson et al., 2008; Uzunova et al., 2012; Visconti et al., 2010). Taking observations together, we therefore propose that, in cells, binding of spindle checkpoint proteins does not completely prevent the ability of Cdc20 to activate the APC/C to the minimal level that is required to efficiently process Nek2A. APC/C^MCC (or APC/C^MCC‐CDC20, see Izawa and Pines 2015) probably has a catalytic activity that is slightly higher than that of late prophase APC/C, when Emi1 is degraded but Cdc20 is not yet bound. This activity forms right at NEBD, by the binding of MCC to phosphorylated APC/C. Formation of APC/C^MCC alone does not lead to cyclin A turnover, but might just be sufficient to catalyze the Cdc20‐dependent degradation of Nek2A, immediately at the prophase‐to‐prometaphase transition. Nevertheless, we cannot fully rule out that, in cells, a small amount of Cdc20 will never be incorporated into the MCC, but still binds to the APC/C at the start of mitosis and is responsible for Nek2A destruction.

Mutation of the TPR‐binding MR tail of Nek2A prevents its binding to the APC/C, including in G2 phase (Fig. 5A), and delays, but does not prevent, Nek2A degradation in mitosis (Hayes et al., 2006; Sedgwick et al., 2013). Because we found that Nek2A destruction is entirely Cdc20 dependent, we reasoned that in the absence of APC/C binding by its MR tail, Nek2A should turn into a spindle‐checkpoint‐controlled substrate. To test this, we generated cell lines stably expressing a mutant of Venus–Nek2A lacking the MR tail (Venus–Nek2AΔMR) together with the spindle‐checkpoint‐target geminin–Cherry (Clijsters et al., 2013). When comparing these two substrates in single cells, we found a complete overlap of their destruction curves (Fig. 5B, note that here the graphs are synchronized around anaphase onset; supplementary material Fig. S1A). Importantly, degradation of Nek2AΔMR became highly sensitive to Cdc20 depletion, similar to that of geminin (Fig. 5C). We also compared the timing of Nek2AΔMR destruction to that of Aurora‐A–eCFP, a known APC/C^Cdh1 substrate (Floyd et al., 2008; Honda et al., 2000), and found that Venus–Nek2AΔMR was degraded well before Aurora A (supplementary material Fig. S4A) and independently of Cdh1 (supplementary material Fig. S4B). Nek2AΔMR remained largely stable during a taxol‐induced mitotic delay (supplementary material Fig. S4C), similar to cyclin B1 (Brito and Rieder, 2006; Clute and Pines, 1999; Gascoigne and Taylor, 2008). When the spindle checkpoint was silenced by the Mps1 inhibitor reversine, destruction of geminin–Venus and Cherry–Nek2AΔMR began at NEBD (Fig. 5D). We conclude that abolishing the binding of Nek2A to the APC/C makes Nek2A destruction strictly dependent on an activity of Cdc20 that can only be released by passing the spindle checkpoint.

The model emerging from our study is that spindle‐checkpoint‐restricted APC/C^Cdc20 has sufficient catalytic activity to initiate destruction of Nek2A, provided that Nek2A is constitutively recruited towards the APC/C. Spindle checkpoint proteins typically prevent the binding of Cdc20 to a destruction motif such as the D‐box or the KEN box. The Nek2 gene is spliced into three different isoforms of which Nek2A is the longest (Wu et al., 2007). This is the only isoform to contain an evolutionary conserved KEN box. In line with the spindle‐checkpoint independence of Nek2A destruction, mutating the KEN box did not stabilize Nek2A (Cherry–Nek2A‐AEN, Fig. 6A; supplementary material Fig. S1A; also see Sedgwick et al., 2013). Then, we investigated whether the KEN box could contribute to the spindle‐checkpoint‐dependent destruction of Nek2AΔMR. Interestingly, a double Nek2A mutant, lacking both the APC/C‐binding tail and the KEN destruction motif, remained fully stable throughout mitosis (Fig. 6B; supplementary material Fig. S1A). First, this result confirms that Nek2A degradation is indeed entirely dependent on the APC/C. Second, it shows that the spindle checkpoint very effectively blocks the recognition of the Nek2A KEN box by APC/C^Cdc20. Normally, this does not already occur in mitosis because Nek2A has largely disappeared when cells reach anaphase. The Nek2A KEN box did not play a role in binding Nek2A to the APC/C in G2 phase (Fig. 6C). These results imply that preventing binding of destruction motifs, like the KEN box, to Cdc20 is the main mechanism by which the spindle checkpoint stabilizes APC/C substrates in prometaphase. Indeed, the MCC complex inhibits APC/C^Cdc20 by binding to the KEN box and D‐box receptor (Izawa and Pines, 2015). Nek2A destruction normally does not depend on a destruction motif, so it can start in the presence of an active spindle checkpoint as soon as Cdc20 activates the APC/C. Only after satisfaction of the spindle checkpoint, does APC/C^Cdc20 start to recognize APC/C destruction motifs such as the cyclin B1 D‐box, or the Nek2A KEN box.

Although the spindle checkpoint inhibits the binding of Cdc20 to destruction motifs (Chao et al., 2012), it is possible that the checkpoint also impairs, at least to some extent, the ability of Cdc20 to promote the catalytic activity of the APC/C (Izawa and Pines, 2012). Indeed, in vitro APC/C^MCC, although not completely inactive, is less active than checkpoint‐free APC/C^Cdc20 (Fang, 2002; Herzog et al., 2009; Tang et al., 2001). This is in line with several other studies showing that the checkpoint inhibits APC/C catalytic activity (Maciejowski et al., 2010; Mansfeld et al., 2011; Sliedrecht et al., 2010).

Next, we tested whether inability to activate the spindle checkpoint at the prophase‐to‐prometaphase transition would further increase APC/C^Cdc20 activity towards Nek2A, as also shown for cyclin A (Collin et al., 2013). Therefore, we abolished the spindle checkpoint by treating G2 phase cells with the Mps1 inhibitor reversine (Kwiatkowski et al., 2010; Schmidt et al., 2005) or with Mad2 siRNA, and measured the degradation of geminin–Cherry and Venus–Nek2A as cells entered mitosis (Fig. 7A, upper panel and lower panel, respectively). The average time from NEBD to anaphase in control cells was 25.4 min (±4.8 s.d.) (Fig. 7B,C), whereas bypass of the checkpoint reduced the duration of mitosis to 14.3 min (±2.5 s.d.; reversine) and 12.6 min (±1.9 s.d.; Mad2‐RNAi) (Fig. 7B,C). Geminin–Cherry levels reached 50% of their maximal fluorescence within 31.3 min in controls (±8.1 s.d.), which was accelerated approximately twofold by either of the two means of checkpoint inhibition, to 15.8 min (±2.1 s.d.) in reversine‐treated cells and 15.4 min (±3.0 s.d.) for Mad2 siRNA cells (Fig. 7B,C). Remarkably, Nek2A was also degraded approximately twofold faster after silencing the spindle checkpoint: we found a decrease in half‐life from 14.36 min in controls to 8.0 min (±1.5 s.d.) for reversine‐treated cells and 8.7 min (±1.8 s.d.) for Mad2‐depleted cells (Fig. 7A–C). Under both experimental conditions, the order of substrate degradation was unaltered, which is in line with the idea that direct Nek2A binding to the APC/C makes it a uniquely effective substrate. We then investigated whether the KEN box played a crucial role in accelerating degradation of Nek2A in the absence of the spindle checkpoint. Although a single point mutant in the KEN box was enough to prevent binding in the absence of the MR tail (Fig. 6C), we now mutated the complete KEN box. Importantly, the destruction of a complete alanine‐substitution mutant of the Nek2A KEN box, Venus–Nek2A‐KEN‐AAA, also occurred faster upon reversine treatment (Fig. 7D). Taken together, these results therefore indicate that removal of the spindle checkpoint, independently from facilitating the recognition of a KEN box by Cdc20, slightly increases APC/C^Cdc20 activity from the start of prometaphase onwards. In conclusion, the spindle checkpoint predominantly blocks binding of Cdc20 to destruction motifs in APC/C substrates, but also slightly attenuates the catalytic activity of APC/C^Cdc20 in prometaphase. The latter inhibitory effect of the spindle checkpoint is insufficient to prevent Nek2A destruction and is also not enforced by spindle poisons.

The stability of every APC/C substrate might be governed in a unique way to ensure its degradation occurs at a specific point in the cell cycle (e.g. Lu et al., 2014). Cyclin A and cyclin B1 are both APC/C substrates that similarly depend on Cdc20 for their destruction, but they are degraded at different times in mitosis (Wolthuis et al., 2008). Whereas cyclin A gets degraded in prometaphase, cyclin B1 is stabilized by the spindle checkpoint until metaphase. Previously, we and others have shown that the N‐terminus of cyclin A binds to Cdc20 in such a way that it competitively inhibits the ability of checkpoint proteins to bind Cdc20. These ‘checkpoint‐free’ cyclin‐A–Cdc20 complexes are then recruited to the phosphorylated APC/C in mitosis (Fig. 8). Here, we show that another mitotic regulator that disappears rapidly in prometaphase, Nek2A, requires only very limited amounts of Cdc20 to be degraded effectively. Nek2A destruction also does not detectably depend on binding to Cdc20, or on a canonical KEN box or D‐box destruction motif. Inhibiting the ability of APC/C^Cdc20 to bind to destruction boxes, by treatment with the recently discovered APC/C inhibitor APCin, did not stabilize Nek2A [Sackton et al., 2014; and unpublished data]. Nevertheless, Nek2A relies exclusively on APC/C^Cdc20 to be degraded in mitosis: simultaneously reducing Cdc20 levels by siRNA, combined with inhibiting Cdc20 binding to the APC/C by proTAME, completely blocks Nek2A destruction. Reducing the levels of APC/C subunits by siRNA had surprisingly little effect on Nek2A degradation, especially when compared to spindle‐checkpoint‐dependent APC/C^Cdc20 substrates. This most likely reflects the fact that Nek2A is a very effective APC/C substrate: Nek2A is constantly targeted to the APC/C, possibly to the TPR motif containing APC8 [although a role for other subunits has not been excluded (see also, Sedgwick et al., 2013)], and this renders Nek2A highly sensitive for efficient Cdc20‐dependent degradation.

By treating mitotic cells with nocodazole, more checkpoint signal is generated, as Mad2 is bound threefold as effectively to Cdc20 (Collin et al., 2013). This slows down cyclin A degradation but not Nek2A degradation. An attenuating effect of increasing spindle checkpoint strength on cyclin A disappearance fits with the unique requirement for competition between cyclin A and spindle checkpoint proteins for Cdc20 binding.

The time when Nek2A destruction begins in mitosis is not set by increased Nek2A recruitment to Cdc20 or the APC/C, but marks the point when Cdc20 starts to catalytically activate the APC/C (Fig. 8). In contrast, cyclin A destruction requires the prior formation of stable complexes between cyclin A and Cdc20, their timely recruitment to the prometaphase APC/C, and finally a function of Cdc20 that is sensitive to the spindle checkpoint, possibly the positioning of the cyclin A N‐terminus towards the active site of the APC/C (Fig. 8).

Although we recently discovered that Nek2A is very slowly degraded by an APC/C‐dependent mechanism during S and G2 phases (Hames et al., 2005; and our unpublished data), Nek2A disappears only in mitosis. This not due to decreased translation of Nek2A at mitotic entry, because Nek2A is still rapidly synthesized in mitosis (e.g. see Fig. 4B). We think that the simplest model explaining our data is that rapid Nek2A disappearance marks the point when Cdc20, regardless of its incorporation into or inhibition by the MCC, activates the APC/C at the prophase‐to‐prometaphase transition.

At mitotic entry, Cdc20 binds the APC/C by means of its C‐terminal tail (Vodermaier et al., 2003), the KILR motif (Izawa and Pines, 2012) and by its N‐terminal C‐box (Kimata et al., 2008). The binding of Cdc20 to the APC/C is enforced by mitotic phosphorylation of the APC/C, but also by the spindle checkpoint: the Cdc20 C‐box might be involved in stabilizing complexes between the APC/C and spindle checkpoint proteins (Hein and Nilsson, 2014). Hence, Cdc20, when incorporated in the MCC, effectively forms complexes with the APC/C at the start of prometaphase. Recent in vitro data showed that Nek2A ubiquitylation might be refractory to increasing levels of checkpoint proteins (Kelly et al., 2014). Moreover, activation of the E2 enzyme Ube2S does not seem to be hindered by the checkpoint proteins in its ability to elongate Nek2A monoubiquitin chains in vitro (Kelly et al., 2014). Intriguingly, the MCC has been shown to bind two Cdc20 molecules (Izawa and Pines, 2015), as has been hypothesized previously (Primorac and Musacchio, 2013). BubR1 blocks the substrate recognition domain of the Cdc20 molecule bound to the APC/C (Kraft et al., 2005; Lara‐Gonzalez et al., 2011). We also find that the spindle checkpoint predominantly acts to prevent Cdc20 from binding to destruction motifs (Figs 5, 6). Nek2A destruction is only dependent on an APC/C‐activating step that coincides with the association of Cdc20 to the APC/C in mitosis. However, a mutant of Nek2A, Nek2AΔMR, strictly needs a KEN box for its destruction and is easily stabilized by Cdc20 depletion, as well as being strictly controlled by the spindle checkpoint. This fits with the concept that the spindle checkpoint particularly blocks stoichiometric complex formation between Cdc20 and APC/C substrate destruction motifs.

Virtually no BubR1‐free APC/C^Cdc20 is detected in spindle‐poison‐arrested cells, unless these cells are treated with proteasome inhibitor (Herzog et al., 2009). This indicates that any Cdc20 free of checkpoint proteins is rapidly degraded in nocodazole‐arrested cells (e.g. see Fig. 4E). This, combined with the observation that enforcing the spindle checkpoint does not delay Nek2A degradation, supports our hypothesis that even spindle‐checkpoint‐inhibited Cdc20 is able to partially activate the APC/C at prometaphase (Fig. 8). Alternatively, however, a very small fraction of APC/C^Cdc20 could remain completely uninhibited during prometaphase and be sufficient for Nek2A destruction to proceed entirely regardless of the spindle checkpoint.

Our results reveal a paradoxical role of the spindle checkpoint in Nek2A degradation. Drug‐induced enforcement of the spindle checkpoint cannot delay the time when Nek2A degradation starts but ablating the spindle checkpoint, by treating cells with Mps1 inhibitor or depleting Mad2, increases the rate by which Nek2A disappears. This can be explained by assuming that, whereas the spindle checkpoint blocks recognition of destruction motifs very effectively, it only moderately impairs the catalytic activity of the APC/C^Cdc20 during prometaphase.

Interestingly, this would also imply that, in cells that pass through mitosis normally, APC/C^Cdc20 gains further activity after spindle checkpoint release during metaphase and anaphase (Lindon and Pines, 2004). The nature of the increased APC/C^Cdc20 activity could be twofold: either the C‐box of Cdc20 becomes unrestricted by checkpoint release and triggers a catalytic activation of the APC/C complex to which it is already bound, or the total number of APC/C^Cdc20 complexes in cells increases at metaphase, because the spindle checkpoint prevents accumulation of Cdc20 onto the APC/C in prometaphase (Mansfeld et al., 2011; Nilsson et al., 2008; Uzunova et al., 2012). While we were preparing this manuscript, the Barford laboratory published that binding of the N‐terminal part of Cdh1 containing the C box, which is identical to the C box in Cdc20, is responsible for interaction with APC1 (Chang et al., 2014), and allows for conformational change of the APC/C catalytic module, APC2–APC11. Their work also implies that release of the spindle checkpoint enhances binding of ubiquitin‐bound UbcH10, boosting the activity of the APC/C. Likely, checkpoint silencing will not only permit increased UbcH10 binding, but also increased Ube2S binding and thus higher APC/C catalytic activity (Van Voorhis and Morgan, 2014). Recently, we and others proposed that enhanced APC/C^Cdc20 activity upon spindle checkpoint release might help to avoid the ‘anaphase problem’: the risk that separating sister chromatids when losing tension could re‐impose the spindle checkpoint in case cyclin B1 is not completely degraded when cells reach anaphase (Clijsters et al., 2014; Kamenz and Hauf, 2014; Rattani et al., 2014; Vázquez‐Novelle et al., 2014). The implications of these findings require further analysis of the way changes in APC/C^Cdc20 influence mitotic exit.

Human osteosarcoma cells were grown in Dulbecco's modified Eagle's medium (DMEM; Gibco) containing FCS (Sigma), penicillin, streptomycin and cultured at 37°C in 5% CO₂. At 24 or 48 h before synchronization or transfection, cells were plated on 9‐cm Falcon dishes or, for time‐lapse fluorescence microscopy on 3.5‐mm glass‐bottomed dishes (Wilco Wells) or four‐well glass‐bottomed dishes (Labtek II). For enrichment of cells in G2 phase, cells were treated for 24 h with thymidine (Sigma, 2,5 mM final concentration) and incubated for 8 h after release.

Other drugs, used as indicated, were: Mps1 inhibitor reversine (catalog number 10004412, 50 nM final concentration; Cayman Chemicals); proteasome inhibitor MG132 (catalog number 13697, 5 µM final concentration; Cayman Chemicals); translation inhibitor cycloheximide (catalog number C6255, 5 or 10 µM final concentration; Sigma‐Aldrich), RO‐3306 (catalog number 217699, 3 µM final concentration; Calbiochem), ProTAME (I‐440, 12 µM final concentration or as noted; R&D systems).

Nek2A was cloned from cDNA into a Clontech C1 vector, encoding either a Cherry or Venus fluophore, and subsequently cloned into Clontech pLib vectors. To create stable cell lines, Phoenix‐ecotropic cell lines were transfected in six‐well plates with 4 µg of pLIB‐vector containing the insert of choice, using standard calcium phosphate transfection. Viral supernatant was collected three times, 40, 48 and 64 h after transfection. The supernatant was cleared through a 0.45‐µm filter (EMD Millipore). U2OS cells expressing the ecotropic receptor (from Johan Kuiken, NKI, Amsterdam, Netherlands) were infected twice in the presence of polybrene.

Cells were transfected with 40 nM siRNA oligonucleotide pools (ON‐TARGET‐plus oligonucleotides, Dharmacon) using Lipofectamine 2000 (Invitrogen) according to the manufacturer's protocol. siRNAs (Thermo Fisher Scientific as ON‐TARGET plus SMART pools) used were against: Nek2 (targeting both Nek2A and Nek2B), 5′‐GGAUCUGGCUAGUGUAAUU‐3′, 5′‐GCAGACAGAUCCUGGGCAU‐3′, 5′‐ GGCAAUACUUAGAUGAAGA‐3′ and 5′‐ GCUAGAAUAUUAAACCAUG‐3′; Cdc20 (CDC20), 5′‐CGGAAGACCUGCCGUUACA‐3′, 5′‐GGGCCGAACUCCUGGCAAA‐3′, 5′‐GAUCAAAGAGGGCAACUAC‐3′ and 5′‐CAGAACAGACUGAAAGUAC‐3′; Mad2 (MAD2L1), 5′‐UUACUCGAGUGCAGAAAUA‐3′, 5′‐CUACUGAUCUUGAGCUCAU‐3′, 5′‐GGUUGUAGUUAUCUCAAAU‐3′ and 5′‐GAAAUCCGUUCAGUGAUCA‐3′; Cdh1 (FZR), 5′‐CCACAGGAUUAACGAGAAU‐3′, 5′‐GGAACACGCUGACAGGACA‐3′, 5′‐GCAACGAUGUGUCUCCCUA‐3′ and 5′‐GAAGAAGGGUCUGUUCACG‐3′; APC2 (ANAPC2), 5′‐GAGAUGAUCCAGCGUCUGU‐3′, 5′‐GACAUCAUCACCCUCUAUA‐3′, 5′‐GAUCGUAUCUACAACAUGC‐3′ and 5′‐GAGAAGAAGUCCACACUAU‐3′; Apc10 (ANAPC10), 5′‐GAGCUCCAUUGGUAAAUUU‐3′, 5′‐GAAAUUGGGUCACAAGCUG‐3′, 5′‐GCAAUCAGAUGGUUCCCAG‐3′ and 5′‐CAUGAUGUAUCGUUCAAUA‐3′; APC3 (Cdc27), 5′‐GGAAAUAGCCGAGAGGUAA‐3′, 5′‐CAAAAGAGCCUUAGUUUAA‐3′, 5′‐AAUGAUAGCCUGGAAAUUA‐3′ and 5′‐GCAUAUAGACUCUUGAAAG‐3′; Ube2S (UBE2S), 5′‐ACAAGGAGGUGACGACACU‐3′, 5′‐GGAGGUCUGUUCCGCAUGA‐3′, 5′‐GCAUCAAGGUCUUUCCCAA‐3′ and 5′‐CCAAGAAGCAUGCUGGCGA‐3′; UbcH10 (Ube2C) 5′‐GAACCCAACAUUGAUAGUC‐3′, 5′‐UAAAUUAAGCCUCGGUUGA‐3′, 5′‐GUAUAGGACUCUUUAUCUU‐3′ and 5′‐GCAAGAAACCUACUCAAAG‐3′.

Antibodies used were: goat anti‐actin (SC‐1616, Santa Cruz Biotechnology), rabbit anti‐APC2 (provided by Jonathon Pines, Department of Zoology, Gurdon Institute, Cambridge, UK), mouse anti‐APC3 (BD Transduction), goat anti‐APC4 (SC21414, Santa Cruz Biotechnology), goat anti‐Cdc16/APC6 (1∶1000; SC‐6395, Santa Cruz Biotechnology), rabbit anti‐APC10 (611501/2, Biolegend), mouse anti‐BubR1 (Chemicon MAB3612; 1∶500), rabbit anti‐BubR1 (A300‐386a, Bethyl), mouse anti‐Cdc20/p55 (sc‐13162, Santa Cruz Biotechnology), mouse anti‐Nek2 (1∶500; BD 610593), mouse anti‐Cdk1 (Cell Signaling), mouse anti‐Mad2 (MBL K0167‐3), mouse anti‐Cdh1 (MS1116‐p1, Neomarkers), goat anti‐Cdk4 (sc‐260, Santa Cruz Biotechnology), rabbit anti‐cyclin A2 (H‐432, Santa Cruz Biotechnology), rabbit anti‐TopoIIα (A300‐054A‐1, Bethyl), rabbit anti‐PTTG‐1/Securin (1∶500; 34‐1500, Zymed) and custom rabbit anti‐GFP ‘2C’ antibodies.

Cells were lysed in ELB+ (150 mM NaCl, 50 mM HEPES pH 7.5, 5 mM EDTA, 0.3% NP‐40, 10 mM β‐glycerophosphate, 6% glycerol, 5 mM NaF, 1 mM Na₃VO₄ and Roche protease inhibitor cocktail). Lysates were cleared by centrifugation (13,000 g, 12 min at 4°C). Protein levels were equalized by using Bradford analysis. For immunoprecipitations, 2 µg antibodies were precoupled for 4–12 h to 20 µl of protein‐G–Sepharose (Amersham Biosciences) and washed with ELB+. Precoupled beads and lysates were incubated overnight at 4°C and washed three times with 1.0 ml of ice‐cold ELB+. All remaining buffer was removed and beads were resuspended in 60 µl sample buffer; 25 µl was separated by SDS‐PAGE and blotted on nitrocellulose (0.4 µm pore). Immunoprecipitations of GFP were performed with GFP–Trap_A beads (Chromotek), according to the manufacturer's protocol. Membranes were blocked with 4% milk solution made from ELK powder (Campina Friesland, Amersfoort, The Netherlands) in PBS containing 0.1% Tween. Development of blots was achieved with either with silver film and scanning, or by using the Chemidoc Imaging System (Bio‐Rad Laboratories) and quantification was performed with the Image Lab (Bio‐Rad Laboratories) software.

U2OS or RPE1‐TERT cells transfected with siRNA and the indicated plasmids were followed by fluorescence time‐lapse microscopy. Acquisition of differential interference contrast (DIC) and fluorescence images started 24 or 48 h after transfection on a microscope (Axio Observer Z1; Carl Zeiss) in a heated culture chamber (5% CO₂ at 37°C) using DMEM with 8% FCS and antibiotics. The microscope was equipped with an LD 0.55 condenser and 40× NA 1.40 Plan Apochromat oil DIC objective and CFP/YFP and GFP/HcRed filter blocks (Carl Zeiss) to select specific fluorescence. Images were taken using AxioVision Rel. 4.8.1 software (Carl Zeiss) with a charge‐coupled device camera [ORCA R2 Black and White CCD (Hamamatsu Photonics) or Roper HQ (Roper Scientific)] at 100‐ms exposure times. Alternatively imaging was performed on a Deltavision Elite system, using L15 Leibovits medium (Gibco), in a 37°C culture chamber, without the need of supplying CO₂.

For quantitative analysis of degradation, MetaMorph software (Universal Imaging), ImageJ (National Insitute of Health) and Excel (Microsoft) were used. Captured images were processed using Photoshop and Illustrator software (Adobe).

We thank Daisuke Izawa and Jon Pines for providing the APC2 antibody, Arne Lindqvist for sharing the U2OS Tet inducible Cyclin A2‐Venus cell line, the Hyman lab for the LAP‐BubR1 HeLA celline and Katarzyna Kedziora for assistance with Image J macro writing. We thank Erik Voets and other division members for fruitful discussion and critically reading the manuscript.