What is the mechanism through which hMps1 functions in the mitotic spindle checkpoint? It is known that in higher eukaryotes most mitotic spindle checkpoint components, including hMps1, are recruited to unattached kinetochores in early mitosis (Taylor et al., 1998; Taylor et al., 2001; Chen and Murray, 1997; Chen et al., 1998; Yu, 2002), yet the molecular events that occur once they are at the kinetochore remain largely unknown. To date, direct molecular interactions between checkpoint proteins and kinetochores are poorly understood. Nevertheless, the network of interactions and interdependencies of the checkpoint proteins is much better defined for their kinetochore localization, although the specific dependencies on the hMps1 protein kinase are completely unknown.
Therefore, we eliminated hMps1 by siRNA and examined the localization of several checkpoint components. We found that specifically the checkpoint proteins hMad1 and hMad2 were displaced from the kinetochores in hMps1-depleted cells.
Thus, the recruitment of the hMad1/hMad2 complex to kinetochores depends on hMps1. Similar results were reported from studies in Xenopus egg extracts, where immunodepletion of XMps1 prevented kinetochore-association of XMad1/XMad2 (Abrieu et al., 2001). Interestingly, addition of a recombinant, kinase-dead form of XMps1 to depleted extracts did not allow reassociation of XMad1 and XMad2 to kinetochores, whereas wild-type XMps1 did readily (Abrieu et al., 2001), suggesting that the kinase activity of XMps1 has an essential role in mediating kinetochore localization of these proteins. Whether kinetochore recruitment of hMad1/hMad2 also depends on kinase activity in human cells has to be tested in future studies.
How could hMps1 function on the hMad1/hMad2 complex? Previous reports illustrated that Mad1p is tightly complexed with Mad2p throughout the cell cycle
(Chen et al., 1999; Sironi et al., 2001; Luo et al., 2002; Sironi et al., 2002) and Mad1 is thought to be responsible for targeting Mad2 to unattached kinetochores (Chen et al., 1998). Mad2 is one of the prominent checkpoint components, that directly interacts with the molecular target of the mitotic spindle checkpoint, Cdc20, leading to inhibition of the APC/C and thus to metaphase arrest (Li et al., 1997; Fang et al., 1998b; Hwang et al., 1998; Kim et al., 1998). Importantly, the interaction of Mad2 with Mad1 is absolutely required to form a Mad2-Cdc20 complex in vivo (Hwang et al., 1998; Chung and Chen, 2002; Fraschini et al., 2001b; Hardwick et al., 2000).
Structural studies revealed that formation of the Mad2-Cdc20 interaction involves a large conformational change of Mad2 (Luo et al., 2000; Luo et al., 2002).
Furthermore, because Mad1 and Cdc20 contain similar Mad2-binding regions, binding of Mad1 to Mad2 also triggers the same large conformational change of Mad2, as does Cdc20 (Luo et al., 2002; Sironi et al., 2002). Therefore, it is possible that upon dissociation of Mad2 from Mad1, Mad2 transiently retains a conformation more amenable for binding to Cdc20 (Luo et al., 2002). One intriguing possibility is that other checkpoint proteins weaken the Mad1-Mad2 interaction upon checkpoint activation. In that context it is interesting to note that Mad1 is phosphorylated during checkpoint activation (Hardwick and Murray, 1995; Hardwick et al., 1996; Jin et al., 1998; Chen et al., 1999; Waters et al., 1999), and the candidate kinases are Bub1 and Mps1. In yeast, Mad1p is phosphorylated upon Mps1p overexpression (Hardwick et al., 1996), and human Mad1 can be phosphorylated by Bub1 in vitro (Seeley et al., 1999). So far, we did not observe any binding between hMps1 and hMad1/hMad2 in a yeast two-hybrid assay, suggesting that in vivo interactions between these proteins require additional components or modifications (data not shown). Furthermore, we were unable to detect any hMad1 phosphorylation by hMps1 in vitro (data not shown). Nevertheless, it will be attractive to test whether modifications of Mad1 change its binding affinity toward Mad2.
A recent study demonstrated that hMad2 is phosphorylated and that the phosphorylation state of hMad2 regulates its checkpoint activity by modulating its association between hMad1 and Cdc20 (Wassmann et al., 2003). Human Mad2 is modified through phosphorylation on multiple serine residues in vivo in a cell cycle-dependent manner. Notably, however, only unphosphorylated hMad2 interacts with hMad1 or the APC/C in vivo (Wassmann et al., 2003). Considering the tight association of hMad1 with hMad2 during the cell cycle, it is possible that
phosphorylation of hMad2 is required for dissociation of the complex, and subsequently, dephosphorylation of hMad2 is required to allow binding to Cdc20. The hMad2-specific kinase(s) have not been identified yet. However, recombinant hMps1 does not phosphorylate hMad2 in in vitro kinase assays (data not shown).
hMps1 is not required for kinetochore-association of Hec1, hBub1, hBubR1, CENP-E, CENP-F and CENP-B
Along with hMad1 and hMad2, we also examined the kinetochore localization of the checkpoint proteins Hec1, hBub1, hBubR1, and the centromere proteins CENP-E, CENP-F and CENP-B in hMps1-depleted cells. Elimination of hMps1 by siRNA did not abolish kinetochore-association of these checkpoint proteins. Thus, hMps1 mediates kinetochore localization of only a specific subset of checkpoint proteins.
However, there are some discrepancies with data obtained from X. laevis and fission yeast.
Immunodepletion of XMps1 from Xenopus egg extracts did not only prevent kinetochore-association of XMad1 and XMad2, but also of the kinesin-like motor protein CENP-E (Abrieu et al., 2001). Interestingly, while XMps1 kinase activity was fully required for XMad1 and XMad2 recruitment to the kinetochores, it was only partially required for association of CENP-E to kinetochores, since some CENP-E bound to kinetochores in the absence of Mps1 kinase activity, but not in the absence of Mps1 protein (Abrieu et al., 2001). However, we found in human cells that hMps1 is not necessary for kinetochore localization of CENP-E. This difference might be explained by the different phenotypes associated with inhibition of CENP-E function in mammalian cells and Xenopus. In mammalian cells, injection of antibodies that remove CENP-E from kinetochores (Schaar et al., 1997), antisense oligonucleotide-mediated suppression of CENP-E accumulation (Yao et al., 2000), or siRNA-mediated depletion of CENP-E (Martin-Lluesma et al., 2002) causes a mitotic arrest apparently arising from a constitutively activated mitotic checkpoint. In contrast, in Xenopus egg extracts, CENP-E is required for establishing and maintaining a mitotic checkpoint, and disruption of CENP-E function by immunodepletion leads to a failure in arresting in response to spindle damage. In the Xenopus egg system, removal of CENP-E prevents Mad2 from associating with kinetochores (Abrieu et al., 2000). In
addition, stimulation of kinase activity of BubR1 is dependent on the interaction with CENP-E (Abrieu, personal communication). Conversely, depletion of CENP-E leaves BubR1 and Mad2 associated with unattached kinetochores in human HeLa cells (Yao et al., 2000; Martin-Lluesma et al., 2002), thereby enabling these cells to activate the checkpoint. The reasons for these differences are unclear, but may reflect species-specific distinctions between the meiotic stage of egg extracts versus the mitotic stage of cultured cells. Certainly, more detailed studies will be needed to clarify the relationship between Mps1 and CENP-E in the checkpoint.
In the fission yeast S. pombe, kinetochore localization of Mad3p is dependent on Bub1p, Bub3p and the Mps1 homolog, Mph1p, but not on Mad1p and Mad2p (Millband and Hardwick, 2002). However, our data show that BubR1, the putative Mad3 homolog in higher eukaryotes, still localizes efficiently to kinetochores in hMps1- depleted cells, which was confirmed in a recent study by Liu et al. (2003).
The reason for this disagreement is not clear, but it could be due to different experimental approaches (genetics versus siRNA) and/or differences among species.
It is also important to bear in mind that in both budding yeast and fission yeast the BubR1 homolog Mad3p lacks a carboxyl-terminal kinase domain (Hardwick et al., 2000; Millband and Hardwick, 2002). Recently, it has been reported that hBubR1 kinase activity is not required for its ability to inhibit the APC/C in vitro (Tang et al., 2001). Whether the human/Xenopus protein has evolved to have a second function that is kinase-dependent remains unclear, but is suggested by the observed variations on the interdependencies of the checkpoint proteins for kinetochore localization.
Hec1, hBub1 and CENP-F could still associate with kinetochores in the absence of hMps1. Moreover, CENP-B, a structural component of the inner kinetochore (Kitagawa and Hieter, 2001) localized to kinetochores in hMps1-depleted cells, indicating that the kinetochore structure is not disturbed upon hMps1-siRNA.
These data allow us conclude that the localization of only a subset of checkpoint proteins is dependent on hMps1. Apparently, only the Mad1/2-checkpoint proteins are influenced upon removal of hMps1, indicating that they function downstream of hMps1. In contrast, the BUB-checkpoint proteins and the CENP-E/-F proteins are recruited to kinetochores independently of hMps1, suggesting that either they function upstream of hMps1 or in a second, hMps1-independent branch of the checkpoint pathway regulating kinetochore localization.