Here we have shown that the SAC maintains mitotic arrest by degrading Cdc20 as an APC/C substrate. To do this the SAC requires Mad2 but, in contrast to most current models of the checkpoint, we find that in human cells the majority of Mad2 does not form a stable complex with Cdc20. Instead, Cdc20 mostly accumulates in a complex with BubR1 and Bub3, and this complex binds to the APC/C. In the absence of Mad2, however, Cdc20 does not bind to BubR1. Thus, our working model () is that Mad2 acts catalytically to promote Cdc20 binding to BubR1, which in turn presents Cdc20 to the APC/C as a substrate, targeting it for degradation. Once the checkpoint is inactivated, Cdc20 no longer binds BubR1 and is free to activate the APC/C. This agrees with data from budding yeast that show that Cdc20 is degraded during the spindle checkpoint 13, 14
, that the BubR1 homologue Mad3 has the greatest effect on Cdc20 stability, and that Mad2 is required for Cdc20 to bind to Mad3 27
. Furthermore, overexpressing Cdc20 in budding yeast to 3 fold more than the endogenous level allows cells initially to arrest but not to remain in mitosis in the presence of spindle poisons 13
. Thus, maintaining the SAC through degrading Cdc20 may be conserved through evolution.
Why Cdc20 has to be degraded to maintain the checkpoint is not clear. The simplest explanation is that it prevents Cdc20 exceeding the level of the BubR1-Bub3 complex. This is feasible because Cdc20 and BubR1 are present at similar levels in the cell, and free Cdc20 appears when we add MG132 to checkpoint-arrested cells. It may not be as simple as this, however, because overexpressing wild-type Cdc20 to ~10 fold endogenous levels does not override the checkpoint, and co-expressing BubR1 did not prevent the K-less mutant driving cells out of mitosis (although BubR1 might require Bub3 and/or another checkpoint component to act). Since even small amounts of K-less Cdc20 that are unlikely to exceed the amount of Bub3-BubR1 do overcome the checkpoint, this might indicate that ubiquitination directly inactivates Cdc20. (Although we cannot exclude the possibility that mutating all the lysines also altered other regulatory post-translational modifications such as phosphorylation, a mutant version of Cdc20 in which all the Cdk consensus phosphorylation sites were altered to alanine did not override the SAC, not shown). An alternative explanation might relate to our observation that metaphase substrates of the APC/C can only be degraded if they are able to localise to specific places in the cell (F.Cooke, A. Hagting and JP, in preparation). Therefore, overexpressing Cdc20 throughout the cell may not saturate the ability of the degradation machinery if it is only required to keep the levels of Cdc20 low in a specific location, whereas the K-less mutant could locally exceed the level of available checkpoint complexes. We have previously shown that the APC/C itself is recruited to improperly attached kinetochores by the checkpoint proteins 29
, and the data we present here could indicate that it is at unattached kinetochores that BubR1 presents Cdc20 to the APC/C as a substrate.
Our model has the advantage that it makes the checkpoint a dynamic system. Previous work by ourselves and others has shown that the SAC and APC/C activity are very tightly coupled: cyclin B1 destruction begins almost immediately after the checkpoint is turned off, and destruction stops almost immediately if the checkpoint is re-imposed (see ref 15
). Since Cdc20 is constantly synthesised during the SAC, this means that activating or inactivating Mad2 will very rapidly alter APC/C activity by determining whether Cdc20 binds to BubR1 and becomes inactive, or remains free to activate the APC/C.
We propose that Mad2 catalyses the binding between Cdc20 and BubR1. (A similar conclusion was reached by Davenport et al. but they used overexpressed protein in asynchronous cells 30
). However, Mad2 does not have a recognisable catalytic domain and the crystal structures of Mad2 do not give immediate clues to what this catalytic activity might be. The counter evidence, however, that Mad2 alone is a Cdc20 inhibitor, or forms a stoichiometric part of the Cdc20-inhibitory complex, is not strong. Mad2 is a very poor inhibitor in vitro
, and in fission yeast, Mad3/BubR1 is essential for Mad2 to block cells in mitosis 31
. There are also caveats to the method that defined the stoichiometry of the Mitotic Checkpoint Complex 24
as 1:1:1:1 Mad2:BubR1:Bub3:Cdc20, where cells were only labelled with 35
S-methionine for 6 hours, which is insufficient to label long-lived proteins to equilibrium.
Since the APC/C requires an activator to recognise its substrates, it is an interesting question whether the same molecule of Cdc20 acts as both activator and substrate during the SAC, or whether Cdc20 bound to BubR1 requires another APC/C activator to be degraded. We cannot yet distinguish between these possibilities but favour the idea that Cdc20 acts as both activator and substrate on the following evidence. Firstly, siRNA treatment shows the other APC/C activator, Cdh1, is not required for the SAC; indeed the mouse knock-out shows that Cdh1 is not required for embryonic or somatic cell division 32
. Secondly, those motifs required for Cdc20 degradation during the checkpoint map to two classes (): the first are motifs in the Mad2-binding region (R132), and the second are the C-box and the C-terminal ‘IR” motifs that are required for Cdc20 to bind and activate the APC/C.
We propose that ubiquitination and degradation are used to inactivate human Cdc20 and maintain the SAC, and not to turn off the checkpoint as recently proposed 11, 12
. Although we note that the K-less Cdc20 binds more Mad2, this is probably because we increased its affinity for Mad2 by changing a conserved lysine in the Mad2 binding site of Cdc20, since an arginine was preferred in this position in a phage-display screen for Mad2-binding peptides 8
. We find no evidence that Cdc20 ubiquitination is important to inactivate the checkpoint because the K-less form of Cdc20 that cannot be ubiquitinated is able to substitute for wild-type Cdc20 to promote mitotic exit with apparently normal timing, and is not impaired in its ability to be released from either Mad2 or BubR1 when the checkpoint is turned off.
Ubiquitination and degradation cannot be the only means by which Cdc20 is inactivated since cells expressing the K-less mutant initially delay in mitosis in the presence of microtubule poisons, and they progress at a similar rate to cells with wild-type Cdc20 through unchallenged mitosis, whereas cells without a SAC are greatly accelerated through mitosis 33
. Thus, Cdc20 may initially be inactivated when it is bound by Mad2 and incorporated into the BubR1 complex, but to maintain the arrest Cdc20 must be ubiquitinated and degraded. The inactivation may be related to the way in which BubR1 binds and presents Cdc20 to the APC/C such that Cdc20 targets itself for destruction: it is unlikely that in this state Cdc20 could activate the APC/C against another substrate.
Lastly, we find that Cdc20 is continuously synthesised in mitosis to replace the protein that is degraded. Cdc20 synthesis is insensitive to rapamycin, indicating that it might be translated from an IRES. There is considerable variation in the ability of tumour cell lines to maintain a checkpoint arrest 34
and it will be interesting to see whether this variability arises from differences in Cdc20 synthesis or degradation.