The segregation of sister chromatids during mitosis is one of the most easily visualized, yet most remarkable, events during the life cycle of a cell. The accuracy of this process is essential to maintain ploidy during cell duplication. Over the past 20 years, substantial progress has been made in identifying components of both the kinetochore and the mitotic spindle that generate the force to move mitotic chromosomes. Additionally, we now have a reasonable, albeit incomplete, understanding of the molecular and biochemical events that are involved in establishing and dissolving sister-chromatid cohesion. However, it is less well-understood how this dissolution of cohesion occurs synchronously on all chromosomes at the onset of anaphase. At the centre of the action is the anaphase-promoting complex/cyclosome (APC/C), an E3 ubiquitin ligase that, in association with its activator cell-division cycle protein 20 homologue (Cdc20), is responsible for the destruction of securin. This leads to the activation of separase, a specialized protease that cleaves the kleisin-subunit of the cohesin complex, to relieve cohesion between sister chromatids. APC/C–Cdc20 is also responsible for the destruction of cyclin B and therefore inactivation of the cyclin B–cyclin-dependent kinase 1 (Cdk1). This latter event induces a change in the microtubule dynamics that results in the movement of sister chromatids to spindle poles (anaphase A), spindle elongation (anaphase B) and the onset of cytokinesis. In the present paper, we review the emerging evidence that multiple, spatially and temporally regulated feedback loops ensure anaphase onset is rapid, co-ordinated and irreversible.
Chromosomes bi-orient autonomously
Chromosome bi-orientation, which is the process whereby sister chromatids are attached to dynamic microtubules from opposite spindle poles, ensures equal segregation of genetic material to the daughter cells. Microtubules attach to a multi-protein structure, known as a kinetochore, built directly on to the centromeric DNA of each chromosome. Accurate chromosome bi-orientation requires that only tension bearing kinetochore–microtubule connections are stabilized, whereas those that do not generate tension are destabilized . The interaction of microtubules with kinetochores is regulated by dynamic changes in the phosphorylation of key proteins at the microtubule–kinetochore interface. These phosphorylation events are largely catalysed by Aurora B kinase, a component of the chromosome passenger complex (CPC) that also contains Survivin, inner centromere protein (INCENP) and Borealin and localizes at the inner centromere region during early mitosis . Microtubule–kinetochore attachments that do not generate tension, fail to separate Aurora B kinase from its substrates at the outer kinetochore. This promotes phosphorylation of several components of the KMN complex, including Ndc80, KNL1 and Dsn1, which inhibits microtubule binding to the kinetochore . When tension is applied across the sister chromatids, the outer kinetochore is stretched away from the inner centromere, thereby separating Aurora B kinase from its substrates [4,5]. The application of tension also promotes interaction of type 1 phosphatase (PP1) with the N-terminus of KNL1. Association of PP1 with KNL1 is promoted by dephosphorylation of a residue within the PP1-binding site in KNL1. In this manner, a positive-feedback loop is established, which ensures that Aurora kinase-dependent phosphorylation of proteins at the outer kinetochore is reversed to stabilize tension bearing microtubule–kinetochore attachments . In this manner, kinetochore dynamics can be controlled in a chromosome-autonomous manner.
Dynamic SAC signalling prevents anaphase onset until all chromosomes are bi-oriented
It is generally accepted that the timing of anaphase-promoting complex/cyclosome–cell-division cycle protein 20 homologue (APC/CCdc20) activation is largely dictated by the spindle-assembly checkpoint (SAC), which ensures that anaphase is delayed until all chromosomes are correctly bi-oriented . Components of the checkpoint include the mitotic arrest deficient-like 1 (Mad1), Mad2, Mad3(BubR1) and Bub3 proteins and the Bub1, Mph1(Mps1) and Aurora B kinases. Checkpoint proteins are recruited to kinetochores that are not bound to the spindle microtubules or which are not under tension. This binding causes Mad2 to undergo a conformational change which triggers its association with Mad3(BubR1) and Cdc20 to form the mitotic checkpoint complex (MCC), a potent inhibitor of the APC/CCdc20 . The spindle checkpoint is satisfied when all sister kinetochores are attached and bi-oriented, which usually follows chromosome congression to the metaphase-plate. This leads to disassembly of the MCC followed by APC/C activation and anaphase onset. Although it was initially thought that the SAC acts as a binary switch, more recent analysis has suggested that the SAC acts as a rheostat, in that the length of the time-cells delay in metaphase is proportional to the number of unattached kinetochores [8–10]. However, this raises a conundrum: if the SAC acts as a rheostat, how can local SAC satisfaction at the final attached and bi-oriented kinetochore lead to a rapid and concerted loss of sister cohesion between all chromosomes? Clearly, MCC generation from a single kinetochore must first be sufficient to inhibit APC/C throughout the cytoplasm and, secondly, that this signal must also be terminated in a concerted and rapid manner so that the cohesion is lost simultaneously on all chromosomes. Crucially, during early mitosis, the MCC is rapidly turned over by APC/C itself. Specifically, when Cdc20 is presented to APC/C in a complex with Mad2, Mad3(BubR1) and Bub3, Cdc20 is ubiquitinylated by APC/C, leading to either its dissociation from APC/C or its proteolytic destruction [11,12]. When kinetochores are unattached fresh free MCC is produced, which replenishes the pool of MCC bound to APC/C. Thus the rate of MCC production at a single unattached kinetochore must exceed the rate of MCC turnover by APC/C in an order that APC/C does not prematurely ubiquitinate cyclin B or securin. Ubiquitination of Cdc20 by APC/C requires APC15, a small subunit of the APC/C, which is dispensible for the ubiqutinylation of securin and cyclin B [12–14]. In the mammalian cells, disassembly of either free MCC or MCC that is bound to APC/C is stimulated by coupling of ubiquitin conjugation to endoplasmic reticulum degradation domain-containing protein 2 (CUEDC2) , through an unknown mechanism and by p31Comet, a structural mimic of Mad2, in conjunction with the thyroid-hormone receptor interacting protein 13 (TRIP13) ATPase associated with various cellular activities (AAA)-type ATPase [16,17]. Rapid APC/C-dependent removal or disassembly of the MCC throughout prometaphase and metaphase ensures that APC/C activity is exquisitely sensitive to the status of kinetochore attachment but may also ensure a switch-like activation of APC/C towards securin and cyclin B when MCC generation falls below a critical threshold. At present, there is no indication that the rate of MCC disassembly increases upon chromosome bi-orientation, but this remains a formal possibility.
Local SAC silencing at the final attached kinetochore triggers global activation of the APC/C
The mechanisms governing SAC silencing at the kinetochore are not well understood. Most components of the spindle checkpoint [including Mad1, Mad2, Mad3(BubR1), Bub1, Bub3 and Mps1] associate to the kinetochore when the spindle checkpoint is active and dissociate or decrease in level when the checkpoint is satisfied . This suggests that either the microtubule engagement with the kinetochore or the deformation of the kinetochore upon the application of tension triggers dissociation of checkpoint proteins from the kinetochore and, by inference, the rate of MCC generation. Recent studies reveal a potential mechanism by which this might occur. Mps1-dependent phosphorylation of the KNL1 protein on the threonine residues within a repetitive motif, called MELT [after the one letter amino acid code for methione (M), glutamic acid (E), leucine (L) and threonine (T)], creates a binding-site for the Bub3–Bub1–BubR1 complex at kinetochores [19–22]. Since mutation of the PP1-binding site in KNL1 increases the kinetochore binding of Bub3 and Bub1 , it has been suggested that KNL1–PP1 may be responsible for MELT dephosphorylation and thus checkpoint silencing. Since Aurora B kinase enhances association of Mps1 to the kinetochore  and Bub1 phosphorylates histone H2A to create a binding site for shugoshin–CPC complex at inner centromere , dephosphorylation of MELT repeats would also reduce Aurora B levels at the inner centromere, thereby further increasing the concentration of PP1 on KNL1 and decreasing Mps1 levels and so on. However, in fission yeast, anaphase onset is not blocked in spc7-12TE cells, in which the threonines (T) of 12-MELT repeats of the KNL1 protein (called Spc7) have been replaced with glutamic acid (E). In these cells, the Bub1, Bub3 and Mad3 proteins bind constitutively to the kinetochore in spc7-12TE cells . This strongly suggests that dissociation of Bub3, Bub1 and Mad3 from the kinetochore may contribute to, but is not strictly necessary for, checkpoint silencing.
Importantly, Mad1 and Mad2 completely dissociate from the kinetochore when the checkpoint is satisfied, whereas low levels of the Bub1 and Bub3 can still be detected on kinetochores during the early anaphase . Crucially, ectopic fusion of Mad1, Mad2 or Mps1 to the kinetochore or induced re-association of Mad1, results in prolonged delay in metaphase [22,26–28]. These results suggest dissociation of Mad1, Mad2 and Mps1 from the kinetochore is critical events for spindle checkpoint silencing. London and Biggins  have recently found that phosphorylation of the N-terminal non-catalytic domain of Bub1 by Mps1 kinase creates a binding site for the Mad1–Mad2 complex on kinetochores. Dissociation of Mad1–Mad2 complex from Bub1 could, in principle, be catalysed by PP1 bound to KNL1. However, constitutive association of PP1 to Spc105 (KNL1) in budding yeast does not override the spindle checkpoint, indicating that regulated association of PP1 to Spc105 (KNL1) may not be the sole trigger of the silencing process . In the fission yeast, checkpoint silencing requires association of PP1 not only to Spc7 (KNL1) but also to Klp5/Klp6 (kinesin-8) motors, suggesting that a more complex process is involved . In mammalian cells, but not yeast, the Mad1 and Mad2 proteins are removed from kinetochores via the dynein motor protein. Dynein interacts with a kinetochore protein, spindly, in a phosphorylation-dependent manner . The establishment of chromosome bi-orientation may trigger dephosphorylation of both Bub1 and/or spindly to trigger dynein-dependent stripping of Mad1 and Mad2 from the kinetochore. However, the phosphatase(s) responsible for these events in mammalian cells are not clear, as association of PP1 to KNL1 is not critical for spindle chec-kpoint silencing . An alternative possibility is that a pool of type-2A phosphatase (PP2A), which binds the BubR1 protein in the mammalian cells, but not in yeast, is important for triggering the SAC-silencing process [33,34]. As highlighted above, the Mps1 kinase is required for the association of all checkpoint proteins to the kinetochore with the exception of Aurora B kinase . Mps1-dissociation from the kinetochore is controlled by microtubule engagement, however, its precise kinetochore-binding site and the mechanism involved remains elusive. It will be important to resolve these issues to fully understand how the mechanical events at the microtubule–kinetochore interface result in rapid and co-ordinated activation of the APC/C.
Inactivation of Cdk1 sharpens the metaphase–anaphase transition
Increasing evidence suggests that once initiated, the destruction of securin and cyclin B are inter-connected events that further sharpen the anaphase switch. In the budding yeast, cyclin-dependent kinase 1 (Cdk1)-dependent phosphorylation of securin near its destruction-box motif inhibits securin ubiquitination by the APC/CCdc20 . This phosphorylation is reversed by the Cdc14 phosphatase, which is released from the nucleolus by the fourteen early-anaphase release (FEAR) pathway, triggered by separase activation [37,38]. Dephosphorylation of securin by Cdc14 increases the rate of securin ubiquitination and the abruptness of separase activation and therefore anaphase. In fission yeast, which lacks a FEAR pathway, the activity of Clp1 (Cdc14) phosphatase is instead inhibited by Cdk1-dependent phosphorylation in early mitosis . As cyclin B levels and therefore Cdk1 activity drop at anaphase onset Clp1 catalyses its own dephosphorylation and activation. At the recent Dynamic Cell conference in Cambridge we reported that Clp1-mediated dephosphorylation and interaction of CPC and Klp9 kinesin [a homologue of human mitotic kinesin-like protein 2 (MKLP2)] is important for dictating the timing of anaphase onset (Meadows and Millar, unpublished data). In cells lacking Klp9, chromosome bi-orientation is completed but anaphase is delayed, suggesting that phospho-dependent relocalization of CPC to the spindle mid-zone helps to sharpen the anaphase switch in fission yeast. It has been thought previously that Mad2 catalyses the interaction between BubR1(Mad3) and Cdc20 and is only removed from the MCC as a consequence of MCC disassembly [11,40–42]. Surprisingly, however, we find that when phospho-dependent interaction of CPC and Klp9 is disrupted, APC/C activation is delayed by a mechanism that requires the Mad3, but not the Mad1 or Mad2, checkpoint proteins. This indicates that the Mad2–Mad3–Cdc20 complex (MCC) is not the sole inhibitor of APC/C that dictates the timing of anaphase onset. In the human cells, Cdc14-like phosphatases play a less prominent role at the metaphase to anaphase transition. Instead, separase becomes active shortly before anaphase onset on chromosomes, dependent on the properties of its binding proteins securin and cyclin B to both inhibit its protease activity and to target it to chromosomes respectively . Following its catalytic activation at anaphase onset, separase binds to cyclin B to inhibit Cdk1 activity. In this manner, separase acts sequentially, first as a protease to cleave cohesin and then as a Cdk1 inhibitor to ensure that anaphase onset occurs rapidly and irreversibly. In summary, eukaryotic cells have evolved multiple feedback pathways to ensure that the loss of sister chromatid cohesion and poleward movement of sister chromatids in anaphase occurs both with a high fidelity and in a highly spatially and temporally co-ordinated manner.
J.C.M. is supported by a Global Research Fellowship from the Institute of Advanced Study, University of Warwick. J.B.A.M. is supported by a programme grant from the Medical Research Council.
The Dynamic Cell: held at Robinson College, Cambridge University, Cambridge, U.K., 4–7 September 2014.