MKlp2 was previously copurified with tandem affinity purification (TAP)–wild type (wt)–Mad2, but not with the nonfunctional deletion mutant, by mass spectrometry analysis after TAP of HEK293 cell lysates expressing TAP-tagged Mad2 (Lee et al., 2008
). Indeed, endogenous MKlp2 was coimmunoprecipitated with Myc epitope–tagged Mad2, but not the nonfunctional mutant of Myc-Mad2(ΔC20), using HEK293 cells (, lanes 1–3). Similarly, using HeLa (, lanes 4–6) and HEK293 cells (, lanes 7–9), HA-tagged MKlp2 was coimmunoprecipitated with Myc-Mad2 but not Myc-Mad2(ΔC10), in which the minimal functional region was deleted (Luo et al., 2000
). Using a series of deletion mutants of MKlp2 (), we found that the C-terminal region of MKlp2 encompassing amino acids 871–880 was required for Mad2 binding. Similar results were obtained using HeLa cells (unpublished data). Mad1 and Cdc20 possess similar Mad2-binding motifs that conform to the consensus sequence (K/R)ϕϕXXXXXP (ϕ, an aliphatic residue; X: any residue; Luo et al., 2002
; Sironi et al., 2002
). Amino acid residues 871–879 of MKlp2 conform to this consensus sequence, which is conserved from human to Xenopus laevis
but not found in its Kinesin-6 family member, MKlp1 (; for review see Verhey and Hammond, 2009
). Indeed, an MKlp2 mutant in which residues 871–874 were replaced with alanine (MKlp2(871A4)) failed to bind Mad2 (). Furthermore, although in vitro–translated HA-MKlp2(wt) and HA-MKlp2(1–880) bound recombinant GST-Mad2, HA-MKlp2(1–870) and HA-MKlp2(871A4) failed to do so (). Thus, we conclude that MKlp2 is a direct binding partner of Mad2.
Figure 1. MKlp2 is a novel binding partner of Mad2. (A, B, and D) Immunoblot analysis with the indicated antibodies, and 10% of the input is shown as total lysates. The positions of molecular mass markers (kilodaltons) are indicated. (A) Lysates of HEK293 cells (more ...)
As the C-terminal region of MKlp1 does not have sequence homology with MKlp2 (), Myc-Mad2 bound GST-tagged MKlp2 but not GST-MKlp1 (). Furthermore, treating with the microtubule destabilizer nocodazole, which activates the mitotic checkpoint, increased the levels of Myc-Mad2 bound to GST-MKlp2 () and the levels of endogenous MKlp2 bound to Mad2 using HeLa cells (). Because Mad2 is essential for the mitotic checkpoint, we examined whether formation of the Mad2–MKlp2 complex is regulated by mitotic checkpoint signaling by depleting Mad1 to inactivate Mad2 (Chen et al., 1998
). To avoid the complication of using nocodazole, which increased the levels of MKlp2, HeLa cells were released from the G1
–S boundary to mitosis. Endogenous MKlp2 bound Mad2 in the absence of nocodazole, whereas depleting Mad1 decreased the interaction (). Likewise, inactivating the mitotic checkpoint by depleting BubR1 decreased the levels of Mad2 bound to MKlp2, where the comparable levels of mitotic marker phospho–Histone H3 (p-Histone H3) were found (). Moreover, Mad2 coprecipitated with MKlp2 in early mitotic lysates (), whereas Aurora B did so in late mitotic lysates (Gruneberg et al., 2004
; Hümmer and Mayer, 2009
), suggesting that Mad2 binds MKlp2 in early mitosis and that the mitotic checkpoint also promotes this complex formation. Notably, MKlp2 did not bind Mad1 (not depicted), and depleting MKlp2 had no effect on the levels of Mad2 bound to Mad1, Cdc20, and the APC (), suggesting that endogenous MKlp2 does not compete with the binding partners of Mad2.
Figure 2. Mad2 binding to MKlp2 depends on the mitotic checkpoint. Immunoblot analysis with the indicated antibodies, and 10% of the input is shown as total lysates. The positions of molecular mass markers (kilodaltons) are indicated. (A) Lysates of HEK293 cells (more ...)
As MKlp2 is a motor that relocates the CPC to the central spindle at the metaphase to anaphase transition (Gruneberg et al., 2004
), the motorless MKlp2(500–890) failed to relocate Aurora B (Fig. S1
). Given that the interaction of kinesins with microtubules is an important regulatory step (for review see Verhey and Hammond, 2009
), we examined whether Mad2 controls the microtubule-binding ability of MKlp2. We chose HeLa cells because of their suitability for immunofluorescence analysis over HEK293 cells. Overexpressed kinesins often show a filamentous staining pattern in interphase (Verhey et al., 1998
). Although endogenous MKlp2, which is periodically expressed in mitosis (Hill et al., 2000
), was not detected in the cytoplasm of interphasic cells, ectopically expressed HA-MKlp2 revealed such staining and colocalized with microtubules (). Furthermore, when HEK293 cells expressing HA-MKlp2 with or without Myc-Mad2 were subjected to microtubule cosedimentation assay using taxol-stabilized microtubules (), ~90% of HA-MKlp2 cosedimented with taxol-stabilized microtubules. In contrast, only ~30% of HA-MKlp2 coexpressed with Myc-Mad2 did so (, lanes 1–4), as quantified by phosphoimager analysis, suggesting that Mad2 inhibits the MKlp2-binding microtubule. Notably, HA-MKlp2(871A4) cosedimented with taxol-stabilized microtubules independent of Myc-Mad2 (, lanes 5–8), whereas it did not without taxol-stabilized microtubules (, lanes 9 and 10).
Figure 3. Mad2 inhibits MKlp2 loading onto the mitotic spindle. (A) HeLa cells expressing HA-MKlp2 were subjected to immunofluorescence analysis with the indicated antibodies. Inset shows enlarged view of the boxed area. (B) Lysates of HEK293 cells expressing the (more ...)
Next, we investigated whether Mad2 also inhibits MKlp2 binding the mitotic spindle. Because the levels of ectopically expressed HA-MKlp2 were typically higher than endogenous MKlp2 by >10-fold (unpublished data), we coexpressed Myc-Mad2 with HA-MKlp2 in HeLa cells. Subsequently, cells were exposed to K858, an inhibitor of the mitotic kinesin Eg5 (Nakai et al., 2009
), to induce the formation of monoasters, a proteasome inhibitor MG132 to maintain Cdk1 active by inhibiting cyclin B1 degradation, and paclitaxel to stabilize the spindles (). HA-MKlp2 largely localized in the mitotic cytoplasm (), whereas HA-MKlp2(871A4) localized in the mitotic spindle (). Moreover, endogenous MKlp2 localized to the mitotic spindle in Mad2-depleted cells (), whereas it did not in control (), suggesting Mad2 inhibits MKlp2 loading onto the mitotic spindle.
As relocating the CPC by MKlp2 to the central spindle is negatively controlled by Cdk1 (Hümmer and Mayer, 2009
), we determined whether Mad2 together with Cdk1 coordinately regulates this event. To address this issue, HeLa cells coexpressing HA-MKlp2 and Myc-Mad2 were forced to form monoasters (), and the localizations of Aurora B and INCENP were determined (). HA-MKlp2 localized in the mitotic cytoplasm, whereas Aurora B and INCENP showed punctate centromere staining patterns (). In contrast, HA-MKlp2(871A4) localized to the mitotic spindle independent of Aurora B and INCENP (). However, when these cells were treated with a Cdk1 inhibitor purvalanol A, HA-MKlp2(871A4) relocated with Aurora B and INCENP to the mitotic spindle (), suggesting that loading MKlp2 onto the mitotic spindle is a temporally separated step from the MKlp2-mediated relocation of the CPC, the event that Cdk1 negatively controls. Moreover, when HeLa cells were treated with paclitaxel to stabilize the mitotic spindle and to activate the mitotic checkpoint, MKlp2 showed a staining pattern of the paclitaxel-stabilized mitotic spindle in Mad2-depleted cells but not in control (). Indeed, MKlp2 localized to the mitotic spindle in Mad2-depleted cells (). Furthermore, as determined with the centromere marker CREST, MKlp2 was not found with Aurora B localized at centromeres in control cells (). In contrast, MKlp2 colocalized with Aurora B and INCENP in Mad2-depleted cells independent of purvalanol A treatment (). Depleting Mad2 prematurely activates the APCCdc20
complex, which inactivates Cdk1 by degrading cyclin B1 (for review see Musacchio and Salmon, 2007
). Indeed, when Mad2-depleted cells were treated with MG132 to stabilize cyclin B1, Aurora B and INCENP were at centromeres, whereas MKlp2 showed the mitotic spindle staining patterns (). Together, these results suggest that endogenous Mad2 prevents MKlp2 from loading onto the mitotic spindle when the mitotic checkpoint is active. Furthermore, this regulatory step is temporally separated from the Cdk1-mediated regulation of MKlp2 on relocating the CPC.
Figure 4. Mad2 and Cdk1 coordinate the mitotic kinesin function of MKlp2 in a spatiotemporal manner. Immunofluorescence analysis using the indicated antibodies is shown. (A) HeLa cells were transfected with expression vectors encoding indicated HA-MKlp2 together (more ...)
To determine whether Mad2 inhibits MKlp2 loading onto the mitotic spindle without spindle damage, we treated control or Mad2-depleted HeLa cells with MG132 to keep Cdk1 active and subsequently extracted the soluble cytosolic proteins to directly visualize MKlp2 bound to the mitotic spindle. Indeed, depleting Mad2 increased the levels of MKlp2 localized at the mitotic spindle compared with control (). Similar results were also observed in the absence of MG132 and Mad1-depleted HeLa cells (Fig. S2
). Furthermore, ectopically expressed HA-MKlp2 in metaphasic HeLa cells localized at the mitotic spindle (), whereas coexpressing Myc-Mad2 abolished it (). Although a Mad2-independnet effect of MKlp2(871A4) might exist, HA-MKlp2(871A4) localized at the mitotic spindle in all metaphasic cells expressing Myc-Mad2 () independent of MG132 treatment (). Moreover, inhibiting Cdk1 by purvalanol A failed to induce the mitotic spindle localization of HA-MKlp2 in cells coexpressing Myc-Mad2 (), whereas HA-MKlp2(871A4) localized together with Aurora B and INCENP (), further suggesting that Mad2 inhibits MKlp2 binding the mitotic spindle independent of Cdk1 activity. Notably, ~20% of cells coexpressing HA-MKlp2(871A4) with Myc-Mad2 became binucleated () where HA-MKlp2(871A4) often localized at the entire mitotic cytoskeleton (). Furthermore, endogenous MKlp2 and Aurora B specifically accumulated in the midbody of control cells undergoing cytokinesis (), whereas both proteins were also found at the mitotic cytoskeleton of Mad2-depleted cells (), which were evident in Mad2-depleted cells treated with paclitaxel.
Figure 5. Controlling MKlp2 by Mad2 is important for proper mitotic progression and cytokinesis. (A, B, and D–F) Immunofluorescence analysis with the indicated antibodies is shown. (A) HeLa cells transfected with control or Mad2 siRNA for 48 h were treated (more ...)
Given that Mad2 inhibited MKlp2 binding of the mitotic spindles, we determined whether ectopically expressed Mad2 inhibits the ability of MKlp2 to relocate the CPC to the central spindle in dividing cells. Because overexpressing Mad2 arrested cells in metaphase, HeLa cells expressing Myc-Mad2 were exposed to purvalanol A for 2 h to induce exit from metaphase (). Determined by immunofluorescence analysis, MKlp2, Aurora B, and INCENP localized in the midbody of control cells (n > 50 each; ) but not in cells expressing Myc-Mad2 (n > 50 each; ), where MKlp2 localized with Myc-Mad2 in the mitotic cytoplasm (). Determined by the punctate staining patterns, Aurora B and INCENP appeared to stay at centromeres of dividing cells expressing Myc-Mad2 (), suggesting that Mad2 negatively regulates the ability of MKlp2 to relocate the CPC for cytokinesis.
In this study, we propose that MKlp2 is a mitotic target of Mad2 and provide several lines of evidence that the mitotic checkpoint promotes complex formation between Mad2 and MKlp2 (). MKlp2 bound Mad2 but not Mad1 in early mitosis, and the mitotic checkpoint increased the levels of Mad2 bound to MKlp2. Conversely, inactivating the mitotic checkpoint by depleting Mad1 or BubR1 decreased the levels of Mad2 bound to MKlp2. Depleting MKlp2 had no effect on the mitotic checkpoint induced by nocodazole treatment (unpublished data), suggesting that it does not regulate the Mad2-mediated checkpoint. Furthermore, depleting MKlp2 had no effect on the complex formations of Mad1–Mad2 and Cdc20–Mad2, suggesting that endogenous MKlp2 does not compete with the binding partners of Mad2. Notably, as kinetochore localization of Mad2 by Mad1 facilitates formation of the Mad2–Cdc20 complex (for review see Nasmyth, 2005
), inhibiting MKlp2 by Mad2 might occur in a kinetochore-dependent manner, which remains to be tested. Furthermore, as relocating the CPC from centromeres is important to inhibit the mitotic checkpoint at anaphase (Vázquez-Novelle and Petronczki, 2010
), MKlp2 might control the mitotic checkpoint by regulating the CPC.
Together with that inhibiting Cdk1 activity promoted MKlp2 and CPC to the mitotic spindle (Hümmer and Mayer, 2009
), we propose that at least two temporally and biochemically separable steps are involved in orchestrating the function of MKlp2. First, Mad2 negatively controls the step of loading MKlp2 onto the mitotic spindle independent of the CPC. Second, Cdk1 activity determines the timing of MKlp2 relocating the CPC. These results do not necessarily contradict that T59 phosphorylation of INCENP by Cdk1 controls the translocation of the CPC and MKlp2 to the midzone during the metaphase to anaphase transition. Instead, we report that the active mutant of MKlp2 constitutively bound the mitotic spindle independent of the CPC, suggesting that loading MKlp2 onto the mitotic spindle does not require the CPC in early mitosis. Completion of bipolar spindle attachment results in the termination of the mitotic checkpoint. Thus, this two-step mechanism of regulating MKlp2 by Mad2 and Cdk1 might orchestrate timing of bipolar spindle attachment, which terminates the Mad2-mediated mitotic checkpoint, allowing MKlp2 to load onto the mitotic spindle. Subsequently, segregation of sister chromatids upon Cdk1 inactivation relocates the CPC by MKlp2 to the central spindle for cytokinesis (). Furthermore, as depleting Mad2 mislocalized MKlp2 to the mitotic cytoskeleton, Mad2 might be a part of a mechanism controlling “free” MKlp2 in postmetaphase cells, suggesting a potential role of Mad2 in late mitotic phases.
We also report that Mad2 inhibits microtubule association of MKlp2 through the C-terminal region of MKlp2, which locates outside of its microtubule-binding motor domain. Although kinesins often function as dimers through their nonmotor regions, an excess amount of recombinant Mad2 did not compromise the self-multimerizing ability of MKlp2 (unpublished data). Instead, as folding the tail and motor domains of kinesins together is suggested as a general mechanism for inhibiting kinesin motors (for review see Verhey and Hammond, 2009
), Mad2 might keep MKlp2 in a folded inactive state in early mitosis. Furthermore, as the C-terminal region of MKlp2 is evolutionary conserved from Humans to Xenopus
, the generality of mechanism controlling MKlp2 by Mad2 also warrants further investigation. Notably, although Mad2 did not bind MKlp1 (), Cdk1/cyclin B directly phosphorylates the motor domain of MKlp1 on an evolutionary conserved site within a basic N-terminal region that inhibits its binding to microtubules (Mishima et al., 2004
), suggesting a distinctive role of Mad2 in controlling MKlp2.
In summary, MKlp2 is an important mitotic target of Mad2 that controls CPC-mediated cytokinesis. Mad2 is overexpressed in many human cancers, which is thought to be a direct cause of chromosome instability and tumorigenesis (Hernando et al., 2004
; Sotillo et al., 2007
), raising the possibility that deregulated Mad2 causes cytokinesis failure by misregulating MKlp2 and the CPC, and so contributes to chromosome instability and tumorigenesis.