Targeting Aurora B to the Equatorial Cortex by MKlp2 Is Required for Cytokinesis

Although Aurora B is important in cleavage furrow ingression and completion during cytokinesis, the mechanism by which kinase activity is targeted to the cleavage furrow and the molecule(s) responsible for this process have remained elusive. Here, we demonstrate that an essential mitotic kinesin MKlp2 requires myosin-II for its localization to the equatorial cortex, and this event is required to recruit Aurora B to the equatorial cortex in mammalian cells. This recruitment event is also required to promote the highly focused accumulation of active RhoA at the equatorial cortex and stable ingression of the cleavage furrow in bipolar cytokinesis. Specifically, in drug-induced monopolar cytokinesis, targeting Aurora B to the cell cortex by MKlp2 is essential for cell polarization and furrow formation. Once the furrow has formed, MKlp2 further recruits Aurora B to the growing furrow. This process together with continuous Aurora B kinase activity at the growing furrow is essential for stable furrow propagation and completion. In contrast, a MKlp2 mutant defective in binding myosin-II does not recruit Aurora B to the cell cortex and does not promote furrow formation during monopolar cytokinesis. This mutant is also defective in maintaining the ingressing furrow during bipolar cytokinesis. Together, these findings reveal that targeting Aurora B to the cell cortex (or the equatorial cortex) by MKlp2 is essential for the maintenance of the ingressing furrow for successful cytokinesis.


Introduction
Cytokinesis is the final event of cell division that results in the irreversible partitioning of a mother cell into two daughter cells. It requires the localized activities of mitotic spindles and the actin cytoskeleton to activate the small GTPase RhoA at the equatorial cortex to promote the formation and ingression of the cleavage furrow [1]. This local activation of RhoA is thought to be controlled by centralspindlin, which is composed of the kinesin-6 family member MKlp1 and the Rho family GTPase RacGAP1/ MgcRacGAP [2].
Another key player in cleavage furrow ingression and completion is Aurora B, the kinase component of the chromosome passenger complex (CPC) [3]. Aurora B is found at the spindle midzone and at the equatorial cortex during the meta-to-anaphase transition [4]. At the spindle midzone, Aurora B is thought to generate an anaphase phosphorylation gradient toward the cell cortex, which provides spatial information to position the cleavage furrow [5]. In contrast, the importance of cortically localized Aurora B for cytokinesis has remained elusive. Interestingly, in HeLa cells undergoing drug-synchronized monopolar cytokinesis that lack the spindle midzone [6], Aurora B but not centralspindlin localizes to the actomyosin filaments in a gap region between the end of polarized monopolar spindles and the furrowing cortical cap [7]. However, the mechanism(s) responsible for Aurora B targeting to the actomyosin filaments of the gap region as well as to the cell cortex (or the equatorial cortex in bipolar cytokinesis) is unknown. Moreover, whether this cortically targeted Aurora B is required for successful cytokinesis in mammalian cells has not been directly tested.
We show here that MKlp2, an essential mitotic kinesin for cytokinesis [8,9], targets Aurora B to the equatorial cortex (or the cell cortex and the growing furrow in monopolar cytokinesis). Mechanistically, the cortical accumulation of MKlp2-Aurora B is accomplished by the ability of MKlp2 to bind myosin-II and actomyosin filaments. This event is required for the highly focused accumulation of active RhoA at the equatorial cortex and for efficient maintenance of the ingressing furrow in bipolar cytokinesis. Specifically, in drug-induced monopolar cytokinesis, targeting Aurora B to the cell cortex by MKlp2 is essential for cell polarization and furrow formation. Supporting this hypothesis, a MKlp2 mutant that is selectively defective in binding myosin-II does not recruit Aurora B to the cell cortex (or the equatorial cortex in bipolar cytokinesis) and does not promote cortical polarization and furrow formation during monopolar cytokinesis. Stable ingression of the cleavage furrow in bipolar cytokinesis also fails in this mutant, although the ability of MKlp2 to target Aurora B to the spindle midzone remains intact. We further demonstrate that continuous Aurora B kinase activity at the growing furrow is required for furrow propagation and completion during monopolar cytokinesis. Together, we propose that MKlp2 is an essential factor for cytokinesis that links Aurora B to the equatorial cortex (or the cell cortex and the growing furrow in monopolar cytokinesis) in mammalian cells.

MKlp2 is Essential for the Maintenance of the Ingressing Furrow in a Partially Redundant Manner with MKlp1
Although MKlp2 is essential for cytokinesis, it is still unclear how MKlp2 contributes to cytokinesis in mammalian cells. To determine the phase(s) of cytokinesis in which MKlp2 is essential, HeLa cells transfected with either control or MKlp2 siRNAs were subjected to time-lapse live-cell imaging analysis in comparison with MKlp1-depleted cells. In control cells transfected with nonsilencing siRNA, the ingressed furrow was maintained until the completion of cytokinesis ( Figure 1A, panel a; Figure 1B). In contrast to control cells, the furrow formed and ingressed but subsequently regressed in MKlp1-depleted cells ( Figure 1A, panel b; Figure 1B, top graph). Although the duration of ingression in MKlp2-depleted cells was longer than in MKlp1-depleted cells ( Figure 1B, bottom graph), the furrow formed and ingressed but was followed by furrow regression in MKlp2-depleted cells ( Figure 1A, panels c, d; Figure 1B, top graph). This finding suggests that MKlp2 is required for the maintenance of the ingressing furrow at a later stage of cytokinesis compared with MKlp1 in mammalian cells. Remarkably, however, co-depletion of MKlp1 and MKlp2 largely inhibited furrow ingression ( Figure 1A, panel e; Figure 1B), suggesting that MKlp2 acts during an early stage of furrow ingression in the absence of MKlp1 in mammalian cells. This result was not due to incomplete depletion of MKlp1 or MKlp2 proteins as determined by immunoblot analysis (Figure 1C), indicating that MKlp1 and MKlp2 function in partially redundant pathways for furrow ingression.
RhoA is required for furrow formation and stable ingression. Notably, the RhoA zone was tightly focused at the equatorial cortex in control cells, whereas the zone was more diffuse in MKlp2-depleted cells ( Figure 1D). Moreover, the maximum intensity projection of serial optical sections through the equatorial cortex revealed that the RhoA zone became diffuse and more unevenly distributed at the equatorial cortex in MKlp2-depleted cells compared with control cells ( Figure S1A). As co-depletion of MKlp1 and MKlp2 largely inhibited furrow ingression ( Figure 1A, panel e), it also eliminated the RhoA zone from the equatorial cortex ( Figure 1D). This result indicates that MKlp2 is responsible for focusing active RhoA at the equatorial cortex. Specifically, the depletion of either MKlp alone did not significantly affect other MKlp levels ( Figure 1C), and the depletion of MKlp2 using different siRNAs did not significantly affect the levels of centralspindlin or CPC components ( Figure S1B; siRNA #3 was used in rescue experiments). Moreover, in HeLa cell lines engineered to express Flag-tagged siRNA-resistant MKlp2 at endogenous levels upon doxycycline (Dox)-treatment, the RhoA zone was focused more tightly at the equatorial cortex compared with non-induced cells ( Figure S1C, S1D). Notably, the total levels of RhoA within the equatorial cortex were similar between control and MKlp2-depleted cells (data not shown), although the RhoA zone was less focused, indicating the unlikelihood that MKlp2 is involved in RhoA activation. Together, our data suggest that MKlp2 promotes the polarized high accumulation of RhoA at the equatorial cortex, which is required for maintaining stable furrow ingression.

MKlp2
Localizes to the Equatorial Cortex via its Ability to Bind Myosin-II and Actomyosin Filaments and is Required for Maintaining the Ingressing Furrow Endogenous ( Figure 1D, panel a) and Dox-induced Flag-MKlp2 ( Figure S1D) accumulated at the equatorial cortex in addition to the spindle midzone, suggesting that MKlp2 may function in furrow ingression at the equatorial cortex. To determine the potential MKlp2-mediated mechanisms(s) involved in furrow ingression at the equatorial cortex, we searched for binding partner(s) of MKlp2 by performing affinity purification of stably expressed Flag-MKlp2 using the HEK293 cell line. Using mass spectrometry analysis, non-muscle myosin-II-A (24 unique peptides) and myosin-II-B (30 unique peptides), herein referred to as myosin-II, were identified in immunoprecipitates from Flag-MKlp2 but not in control cells (data not shown). Indeed, using immunoprecipitation analysis, endogenous MKlp2 and myosin-II were precipitated together in a reciprocal manner ( Figure 2A). Notably, endogenous myosin-II was co-precipitated with HAtagged MKlp2 but not MKlp1 ( Figure 2B). Moreover, compared with full-length HA-MKlp2(1-890), HA-MKlp2(1-842) failed to bind GFP-tagged myosin-II ( Figure 2C). Conversely, HA-MKlp2(1-890) bound strongly to the neck domain (a.a. 779-1087) and weakly to the tail domain (a.a. 1088-1961) of myosin-II ( Figure 2D). Notably, the head domain (a.a. 1-778) of myosin-I, which is responsible for binding filamentous actin, was not found to interact with MKlp2, suggesting that the interaction between MKlp2 and myosin-II was not due to the ability of myosin-II to bind filamentous actin. Specifically, HA-MKlp2(1-842) did not bind myosin-II ( Figure 2D); however, the ability of HA-MKlp2(1-842) to bind microtubules, Aurora B and Plk1 was intact and comparable to HA-MKlp2(1-890) ( Figure S2). Furthermore, the in vitro-translated neck domain of Myc-tagged myosin-II bound recombinant GST-MKlp2(1-890) but not GST-MKlp2(1-842) ( Figure 2E), suggesting that MKlp2 is a binding partner of myosin-II in vivo and in vitro. Consistent with its ability to bind myosin-II, the majority of HA-MKlp2(1-890) and HA-MKlp2 (1-870) co-sedimented with in vitro polymerized F-actin but not MKlp2(1-842) ( Figure 2F), suggesting that MKlp2 forms a complex with actomyosin filaments.
Next, we determined whether these results obtained by the biochemical analysis explained MKlp2 localization at the equatorial cortex. Endogenous MKlp2 localized to the equatorial cortex ( Figure 3A) together with RhoA (panel b) and myosin-II ( Figure 3B) prior to cleavage furrow ingression as shown by immunofluorescence analysis using HeLa cells. Dox-induced Flag-MKlp2(1-890) completely co-localized with Aurora B at the cell equator ( Figure S3A, S3B), supporting the hypothesis that Aurora B is the mitotic cargo of MKlp2 [9]. Moreover, similar to endogenous MKlp2, Flag-MKlp2(1-890) accumulated at the equatorial cortex and the spindle midzone ( Figure 3C, panel a). In contrast, Flag-MKlp2(1-842) selectively failed to accumulate at the equatorial cortex, while it efficiently localized to the spindle midzone ( Figure 3C, panel b). This result is consistent with the finding that Flag-MKlp2(1-842) was selectively defective in binding myosin-II but not other known interacting partners of MKlp2 ( Figure S2). Interestingly, the RacGAP1 centralspindlin component comparably localized to the spindle midzone in cells expressing Flag-MKlp2(1-890) and Flag-MKlp2(1-842) ( Figure  S3C, S3D). Together, these results suggest that the MKlp2:myosin-II interaction is likely required for the accumulation of MKlp2 at the equatorial cortex.
Flag-MKlp2(1-842) was selectively defective in accumulating at the equatorial cortex; therefore, we tested the importance of MKlp2 at the equatorial cortex to maintain furrow ingression. To address this idea, endogenous MKlp1 and MKlp2 were codepleted to inhibit furrow ingression as shown in Figure 1A (panel e). Subsequently, these cells were treated with Dox to induce Flag-MKlp2 and were subjected to time-lapse live-cell imaging. In either Flag-MKlp2(1-890)-or Flag-MKlp2(1-842)-induced HeLa cells transfected with non-silencing control siRNA, the furrow was efficiently ingressed and maintained until the completion of cytokinesis ( Figure 3D  and MKlp2 cells largely failed to initiate and maintain the ingressing furrow; however, Flag-MKlp2(1-842) partially rescued the mutant phenotype compared with co-depleted cells without Flag-MKlp2(1-842). This finding suggests that MKlp2 may have functions independent of the Myosin II interaction, which may possibly occur at the central spindles. Nevertheless, these results reveal that MKlp2 contributes to maintaining the ingressing furrow at the equatorial cortex during bipolar cytokinesis.

MKlp2 is Essential for Aurora B Localization at the Cell Cortex and for Cell Polarization during Monopolar Cytokinesis
The close proximity of the spindle midzone and the equatorial cortex in the ingressing furrow during bipolar cytokinesis creates difficultly in precisely determining the role of MKlp2 within this interface. To better address the issue, we induced drug-synchronized monopolar cytokinesis [6] where Aurora B but not centralspindlin localizes to the actomyosin filaments in a gap region between the end of polarized monopolar spindles and the furrowing cortical cap [7]. In monastrol-arrested monopolar HeLa cells, the majority of MKlp2 and Aurora B was present in the cytoplasm and at the centromeres, respectively ( Figure 4A, panel a). However, when these cells were forced to undergo monopolar cytokinesis caused by treatment with the potent Cdk1 inhibitor purvalanol A (Purv A), endogenous MKlp2 efficiently co-localized with Aurora B at the polarized furrowing cortex ( Figure 4A, panel b). Moreover, MKlp2 co-localized with myosin-II adjacent to the polarized cortex ( Figure 4A, panel d), but it was clearly distinctive from the polarized monopolar spindles (panel c). Using HeLa cells stably expressing GFP-a-tubulin and super-resolution fluorescence microscopy, MKlp2 was shown to be localized to the ends of polarized monopolar spindles and extended to the polarized actomyosin filaments at the growing furrow ( Figure 4B, panel a), indicating that MKlp2 may be involved in linking actomyosin filaments at the growing furrow with polarized monopolar spindles. In contrast, siRNA-mediated depletion of myosin-II ( Figure S4A) markedly suppressed the localization of MKlp2 at the cell cortex but not at the ends of monopolar spindles ( Figure 4B, panel b).Of note, it is possible that since no polarized cortical structures might form in myosin-II-depleted cells, which may prevent MKlp2 from localizing to the cell cortex. Nevertheless, our results indicate that myosin-II is required for MKlp2 to localize to the actomyosin filaments at the growing furrow.
Conversely, depletion of MKlp2 did not affect the cortical localization of myosin-II as well as its protein levels ( Figure S4B, S4C). The cortical localization of Aurora B and INCENP was completely inhibited in MKlp2-depleted cells; however, they were efficiently recruited to the growing furrow in control cells ( Figure 4C). Instead, Aurora B and INCENP largely remained on chromosomes in MKlp2-depleted cells, confirming that MKlp2 is required for removing the CPC from anaphase chromatin during bipolar cytokinesis [9]. Furthermore, as determined by immunofluorescence analysis, depletion of MKlp2 blocked cortical polarization and inhibited subsequent furrow formation ( Figure 4D). This result was also confirmed using time-lapse livecell imaging analysis showing that MKlp2-depleted cells neither stably polarized the cell cortex nor formed a furrow, while control cells efficiently underwent these processes ( Figure 4E, 4F). Interestingly, in these MKlp2-depleted cells, the tip of the monopolar spindle, which was visualized using spindle endtracking GFP-EB1 in MKlp2-depleted cells, was able to contact the cell cortex ( Figure S4C). This result suggests that the failure in polarization was not simply due to a defect in the polymerization . However, both proteins were comparably localized to the spindle midzone. For panels A and C, images were acquired using confocal microscopy. For panel B, images were acquired using 3D-SIM. Insets represent the boxed areas. White bars represent 5 mm. (D) HeLa cells were transfected with non-silencing control siRNA or co-transfected with MKlp1 and MKlp2 siRNAs after the first thymidine block and were treated with Dox during the second thymidine block (see Materials and Methods). The cells were synchronously released from the G 1 /S boundary and subjected to time-lapse live-cell imaging. The cytokinesis progression of cells that were selected at random, and the time they spent in ingression before regression is plotted (top graph). The average duration of furrow in ingression is based on three independent experiments (total n.100 per condition, +/2 standard deviation) (bottom graph). To determine the statistical significance of the duration of furrow in ingression, Student's t-test was performed. P values are indicated. doi:10.1371/journal.pone.0064826.g003 of the monopolar spindles that ultimately prevents contact with the cell cortex. Together, our data reveal that MKlp2 is required for cell polarization and subsequent furrow formation in monopolar cytokinesis.

MKlp2 Recruits Aurora B to the Cell Cortex in Order to Promote Cell Polarization and Furrow Formation during Monopolar Cytokinesis
Next, we determined whether MKlp2 actually targets Aurora B to the cell cortex and subsequently to the growing furrow during monopolar cytokinesis. Furthermore, we wished to determine whether this event is important for polarization and furrow formation. However, siRNA-mediated depletion of MKlp2 alone resulted in the retention of Aurora B on chromosomes, making it impossible to directly evaluate the role of MKlp2 and Aurora B at the cell cortex. Moreover, inhibiting the kinase activity of Aurora B prior to its recruitment to the cell cortex was not an ideal method to determine this issue because such untimely inhibition also causes the mislocalization of other essential factors that are involved in furrow formation (e.g., centralspindlin) from the monopolar spindles as well as the cell cortex [7,10,11].
Therefore, to avoid these complications, we again utilized the rescue assay system consisting of Dox-inducible siRNA-resistant Flag-MKlp2 after knockdown of endogenous MKlp2 in HeLa cells. Similar to the case of endogenous MKlp2 ( Figure 4A), upon induction of monopolar cytokinesis by Purv A treatment, Doxinduced Flag-MKlp2(1-890) relocated to the polarized furrow together with Aurora B ( Figure 5A, panels b-d; Figure S5A, panel a). When these cells were fixed at different time points after Purv A treatment, Flag-MKlp2(1-890) gradually moved to the cell cortex ( Figure S6A, panels b, c) and subsequently to the gap region at the growing furrow (panel d). In contrast, Flag-MKlp2(1-842) defective in binding myosin-II was only able to accumulate at the ends of either non-polarized or polarized monopolar spindles in MKlp2-depleted HeLa cells and failed to accumulate at the cell cortex ( Figure 5A, panels f, g; Figure S6A, panels e-h). This result is consistent with the inability of Flag-MKlp2(1-842) to localize to the equatorial cortex but not the spindle midzone in bipolar cells ( Figure 3C). Moreover, Aurora B was also found at the ends of monopolar spindles together with Flag-MKlp2(1-842) but not at the cell cortex ( Figure 5A, panel h; Figure S5A, panel b). Similar to Aurora B, Flag-MKlp2(1-890) moved together with INCENP to the gap region at the growing furrow ( Figure 5B, panels a, b). In contrast, Flag-MKlp2(1-842) was only able to localize INCENP at the ends of monopolar spindles in MKlp2-depleted HeLa cells ( Figure 5B, panels c, d), indicating that MKlp2 is responsible for localizing both Aurora B and INCENP (thus, most likely the CPC) to the cell cortex. Specifically, in our rescue assay, other essential factors for furrow formation such as centralspindlin and PRC1 were correctly localized to the ends of monopolar spindles towards the cell cortex ( Figure 5C). Together, these results strongly suggest that Flag-MKlp2(1-842) is selectively defective in targeting the CPC from the monopolar spindles to the cell cortex and that this deficiency is likely specific to the CPC without affecting the localization of other essential furrow-inducing factors during monopolar cytokinesis.
Therefore, using this system, we determined whether MKlp2-Aurora B at the cell cortex is important for polarization and furrow formation during monopolar cytokinesis. Quantification of the number of cells undergoing monopolar cytokinesis showed that Flag-MKlp2(1-890) efficiently rescued polarization and furrow formation in MKlp2-depleted cells ( Figure 5D). In contrast, the number of polarized cells with a furrowing cortex was markedly decreased in MKlp2-depleted cells expressing Flag-MKlp2(1-842) ( Figure 5D). Consistent with this result, RhoA was also inefficiently polarized in cells expressing Flag-MKlp2(1-842) compared with Flag-MKlp2(1-890) ( Figure S6B). This result is similar to the diffused RhoA zone at the equatorial cortex observed in MKlp2depleted cells ( Figure 1D; Figure S1A). Moreover, time-lapse livecell analysis using MKlp2-depleted HeLa cells expressing GFP-atubulin and siRNA-resistant mCherry-MKlp2 showed that mCherry-MKlp2(1-890) localized to the growing furrow leading to furrow completion ( Figure S7A, S7D). In contrast, mCherry-MKlp2(1-842) only localized to the ends of monopolar spindles and failed to support monopolar cytokinesis ( Figure S7B, S7D). Notably, it is unlikely that furrowing in general is required for MKlp2 and Aurora B to localize to the cell cortex and to the growing furrow because blocking furrow activity by brief treatment with the myosin-II inhibitor (-)-Blebbistatin prior to Purv A treatment did not prevent Flag-MKlp2(1-890) and Aurora B from localizing to the cell cortex ( Figure S8, panels b, c). Under this condition, Flag-MKlp2(1-842) and Aurora B were also observed at the ends of monopolar spindles ( Figure S8, panels d, e). Together, we conclude that a major role of MKlp2-Aurora B at the cell cortex involves promoting efficient cell polarization and subsequent furrow formation during monopolar cytokinesis.

Continuous Aurora B kinase Activity at the Growing Furrow is Required for Furrow Propagation and Completion
Although the kinase activity of Aurora B is required for stable furrow ingression [12], whether it functions directly at the furrow has remained elusive. MKlp2 is responsible for targeting Aurora B to the growing furrow; therefore, we tested this issue using timelapse live-cell imaging of cells undergoing monopolar cytokinesis. First, as determined by immunofluorescence analysis, mCherrytagged wild-type MKlp2 translocated the majority of Aurora B to the growing furrow ( Figure S5B), suggesting that mCherry-MKlp2 was as functional as endogenous MKlp2. Next, using time-lapse live-cell imaging analysis, ectopically expressed mCherry-MKlp2 efficiently accumulated at the growing furrow upon Purv A treatment, and the furrow propagated and reached completion with similar kinetics compared with normal monopolar cytokinesis ( Figure 6A). In contrast, when the cells were concurrently treated with Purv A and the Aurora B inhibitor ZM447439, cell polarization and furrow formation was inhibited ( Figure 6B, 6E), consistent with a previous report [7]. However, this concurrent inhibition also abolished the cortical localization of mCherry-MKlp2 ( Figure 6B); therefore, no cortical localization of Aurora B was observed, which suggests that MKlp2 and Aurora B are interdependently required for their relocation from anaphase chromatin to the cell cortex. This concurrent treatment also mislocalized MKlp1 from the monopolar spindles ( Figure S9, panel a), creating difficulty in determining the importance of MKlp2-Aurora B at the growing furrow. Therefore, to avoid this complication, ZM447439 was sequentially added after polarizing HeLa cells with Purv A treatment ( Figure 6C, 6E). Using timelapse live-cell analysis, we first determined that mCherry-MKlp2 efficiently accumulated at the growing furrow at 10 min after Purv A treatment ( Figure 6C). Then, ZM447439 was added, and the fate of the same cell was continuously monitored. Notably, upon addition of ZM447439, the furrowing activity immediately ceased, and the furrow regressed, although a considerable amount of mCherry-MKlp2 remained at the regressing furrow ( Figure 6C,  6E). A similar result was also obtained using the Aurora B selective inhibitor hesperadin ( Figure 6D, 6E), excluding a potential off- target effect of ZM447439. Using immunofluorescence analysis, endogenous Aurora B and MKlp2 remained at the regressing furrow under this sequential treatment condition ( Figure S9, panels c & d). Also, this sequential treatment did not affect MKlp1 localization to the ends of monopolar spindles ( Figure S9, panel b). Together, these results suggest that the kinase activity of Aurora B at the growing furrow is continuously required for furrow propagation and completion. Interestingly, this sequential inhibition of Aurora B also depolarized the mitotic spindle (Movie S1, S2), suggesting that the Aurora B kinase activity at the polarized cortex may be essential for maintaining the polarization of monopolar spindles. Taken together, we conclude that targeting Aurora B to the cell cortex by MKlp2 in an interdependent manner may play a role in monopolar cytokinesis by promoting polarization and efficiently maintaining the cytokinesis furrow for completion ( Figure 6F). Moreover, targeting Aurora B to the equatorial cortex by MKlp2 is required for the efficient maintenance of the ingressing furrow in a partially redundant manner with MKlp1 during bipolar cytokinesis.

Discussion
In this study, we propose that MKlp2 is an essential factor for cytokinesis that links Aurora B (and likely as the CPC) to the equatorial cortex (or the cell cortex and the growing furrow in monopolar cytokinesis). We also show that this event may contribute to the efficient maintenance of the ingressing furrow, which is necessary for successful cytokinesis. We also demonstrate that MKlp2 requires myosin-II for its localization to the equatorial cortex, and this event is necessary for recruiting Aurora B to the equatorial cortex in mammalian cells.
Although Aurora B is important for furrow ingression and completion, the role of Aurora B in promoting cytokinesis at the equatorial cortex was unclear. However, one of the major difficulties in resolving this issue involves differentiating Aurora B at the equatorial cortex from Aurora B located at the spindle midzone in mammalian cells. In this sense, we isolated a MKlp2 mutant selectively defective in localizing to the equatorial cortex, and therefore defective in targeting Aurora B and INCENP to the equatorial cortex and not the spindle midzone. Moreover, using monopolar cytokinesis, we visualized that MKlp2 localized to the ends of polarized monopolar spindles and extended to the polarized cortical actomyosin filaments in a myosin-II-dependent manner. This phenomenon may have occurred because no polarized structure formed in myosin-II depleted cells; therefore, MKlp2 was not able to localize to the cell cortex. However, unlike siRNA-mediated depletion of myosin-II, MKlp2 mutants defective in myosin-II binding were also unable to localize to the cell cortex and the growing furrow, although this mutant efficiently localized to the ends of the monopolar spindles both in polarized and nonpolarized cells. Thus, our data strongly suggest that MKlp2 localizes to the cell cortex in a myosin-II-dependent manner. In turn, MKlp2 recruits Aurora B and INCENP (thus, likely as the CPC) to the cell cortex (or the equatorial cortex). Interestingly, MKlp2 localized to chromosomes upon inhibition of Cdk1 activity, and depletion of MKlp2 retained Aurora B and INCENP on the chromosomes; therefore, we speculate that the initial interaction between MKlp2 and Aurora B (or via other components of the CPC) might occur on chromosomes (and potentially in the cytoplasm), which warrants further investigation. Importantly, this MKlp2 mutant was also unable to efficiently promote cell polarization and furrow formation during monopolar cytokinesis. This mutant also failed to maintain stable ingression of the cleavage furrow during bipolar cytokinesis, although its abilities to target Aurora B to the monopolar spindles and the spindle midzone were intact. Together, our results reveal an essential role of a MKlp2-Aurora B complex at the equatorial cortex (or the cell cortex and the growing furrow in monopolar cytokinesis) for cytokinesis, and we propose that the key role of MKlp2 in cytokinesis involves connecting Aurora B at the midzone spindles with the equatorial cortex. We also show here that kinase activity of Aurora B should be present persistently at the furrow for propagation and completion. Given that myosin-II activity is essential for furrow formation and progression, it is likely that kinase activity of Aurora B continuously targets a cortical factor promoting myosin-II activity at the growing furrow, which warrants further investigation.
In this study, we also show that MKlp2 is required for efficient maintenance of the ingressing furrow in mammalian cells. Furthermore, in the absence of MKlp1, MKlp2 becomes essential for initiating furrow ingression during bipolar cytokinesis, demonstrating the role of MKlp2 at an early stage of cytokinesis where it may function in partially redundant pathways with MKlp1. Similar to MKlp1 and MKlp2 co-depletion described in the present work, the combined effect of Plk1 and Aurora B inhibition on furrow ingression is much greater than the effected caused by Plk1 or Aurora B inhibition alone [13]. Thus, the attenuated effects of MKlp2 knockdown alone are likely a result of redundant mechanisms involved in furrow ingression during bipolar cytokinesis. Nevertheless, the cortical localization of MKlp2 becomes essential for furrow ingression in the absence of MKlp1; therefore, we favor a model that MKlp2 targets Aurora B to the equatorial cortex in order to support furrow ingression in a parallel pathway with centralspindlin during bipolar cytokinesis. Notably, the total levels of RhoA at the equatorial cortex did not change significantly in MKlp2-depleted cells, although the RhoA zone measurably expanded. Thus, our result supports the idea that MKlp2 functions in promoting highly polarized accumulation of active RhoA to the equatorial cortex but not by activating RhoA. We speculate that the failure in completing the ingressing furrow in MKlp2-depleted cells may be due to incomplete constriction of the contractile ring or cleavage furrow instability (e.g., cortical oscillation) as also observed in cells depleted of anillin, the assembly scaffold for contractile ring components [14]. Interestingly, co-depletion of anillin and MKlp1 also leads to a complete block of furrow ingression as observed in MKlp1 and MKlp2 codepleted cells [14]. Thus, we further speculate that centralspindlin is required for RhoA activation and myosin-II accumulation at the equatorial cortex; however, MKlp2-Aurora B is recruited to the equatorial cortex through myosin-II, and it may function in focusing (or tightly organizing) the furrowing machinery. This model also supports the recently suggested role of Aurora B in remodeling the contractile ring in Caenorhabditis elegans [15]. However, MKlp2 is evolutionary well-conserved from humans to Xenopus laevis, but it is poorly conserved in Drosophila melanogaster and not found in Caenorhabditis elegans. Thus, the essential role of MKlp2-Aurora B in cytokinesis has apparently evolved specifically in higher organisms, and their different requirements in cytokinesis among species are interesting topics for future study.
Unlike bipolar cytokinesis, we also showed that MKlp2 alone is essential for stable polarization and furrow formation in monopolar cytokinesis, specifically by targeting Aurora B to the cell cortex. Initial events for stable polarization and breaking the symmetry of monopolar spindles are required for monopolar cytokinesis, whereas the symmetry is already broken in bipolar cytokinesis by positioning the mitotic spindles relative to the chromosomes aligned in the metaphase plate. Thus, although the original event triggering the asymmetry of monopolar spindles is unclear, we For panels C and D, ZM447439 or hesperadin was added sequentially after purvalanol A treatment for 10 min. The fate of the same cell was continuously monitored. White bars represent 5 mm. (E) Monopolar cells (from A-D) were scored as cells that completed monopolar cytokinesis (polarized MPC), formed ectopic furrows without a polarized cortex (non-polarized with ectopic furrowing), no significant furrowing activity without a polarized cortex (non-polarized without furrowing), or formed polarized furrows but subsequently regressed (regressed after polarization). The average percentages based on three independent experiments (total n.100 per condition, +/2 standard deviation) are shown. (F) Proposed model. (a) In monopolar cytokinesis, inhibition of Cdk1 triggers a symmetry-breaking reaction that initiates polarization of the monopolar spindles and the cell cortex. Although monopolar spindles are symmetrically positioned, they become asymmetric upon Cdk1 inhibition. Thus, the original event triggering the asymmetry of monopolar spindles remains unclear. (b) During this reaction, MKlp2 mediates Aurora B translocation from the monopolar spindles (or the spindle midzone) towards the cell cortex (or the equatorial cortex) in an interdependent manner. At the ends of monopolar spindles that contact the cell cortex, MKlp2 may act as the bridge between the asymmetrically polarizing monopolar spindles and the actomyosin filaments at the cell cortex. This event may stabilize the polarizing monopolar spindles towards the cell cortex. In turn, it facilitates a rapid accumulation of MKlp2-Aurora B to the cell cortex in a polarized manner. This cortical accumulation of MKlp2-Aurora B promotes furrow formation (or might constrict contractile ring in bipolar cytokinesis) via the kinase activity of Aurora B towards currently unidentified target(s). (c) Once the furrow is formed, MKlp2 targets Aurora B to actomyosin filaments in the gap region between the stably polarized monopolar spindles and the furrowing propose that MKlp2-Aurora B may be important for symmetry breaking in monopolar cytokinesis, and we hypothesize that MKlp2-Aurora B may also contribute to these aspects in bipolar cytokinesis.
Positive feedback between the cell cortex and the mitotic spindles involving Aurora B has been suggested to stabilize polarized microtubule towards the cell cortex [7]. This notion is consistent with our observations involving monopolar cytokinesis that suggest that (i) MKlp2 localizes between the polarizing monopolar spindles and the cortical actomyosin filaments, which may stabilize the monopolar spindles toward the cell cortex in a polarized manner, (ii) MKlp2 recruits Aurora B in an interdependent manner to actomyosin filaments at the growing furrow to provide persistent Aurora B activity for furrow propagation and completion, and (iii) this Aurora B activity is also required for maintaining the monopolar spindles polarized towards the growing furrow, while inhibiting its kinase activity depolarized the monopolar spindles ( Figure 6C-E). However, monopolar spindles are symmetrically positioned, yet they become asymmetric upon Cdk1 inhibition; therefore, the original event triggering the asymmetry of monopolar spindles remains unknown. Nevertheless, our data presented here support the idea that MKlp2 may mediate the proposed positive feedback loops that promotes microtubule polarization and furrowing by connecting and recruiting Aurora B from the polarizing mitotic spindles to the cell cortex, and the initiating event of symmetry breaking warrants further investigation.

Cell Synchronization and Monopolar Cytokinesis
For time-lapse live-cell analysis ( Figure 1A, 3D), the cells were synchronized at the G 1 /S boundary by exposure to 2 mM thymidine for 16 h and incubated in fresh medium for 10 h. During this time, the cells were transfected with siRNAs (100 nM) with Lipofectamine 2000 (Invitrogen) and re-exposed to 2 mM thymidine for 14 h. For Figure 3D, doxycycline (5 mg/ml) was added to the second thymidine block. For monopolar cytokinesis analysis, HeLa cells were treated with the Eg5 inhibitor monastrol (100 mM; Santa Cruz Biotechnology) for 6 h and with the Cdk1 inhibitor purvalanol A (30 mM; Sigma) at 37uC for 15 min ( Figure 4A-C) or 20 min ( Figure 5A-C) before fixation with icecold methanol for 3 min. For Figure 6B-D, Aurora B inhibitor was added [ZM447439 at 2 mM (BIOMOL International); hesperadin at 200 nM (Selleck Chemicals)] either concurrently or 10 min after purvalanol A treatment.

Actin Co-sedimentation Assays
For the actin co-sedimentation assay ( Figure 2F), HeLa cells were grown in 6-well plates and transfected with 4 mg of HA-MKlp2 constructs. Next, 24 h after transfection, the cells were harvested in 0.1% NP-40 cell lysis buffer. Nocodazole (200 ng/ml) was added to the cell lysates to disrupt polymerized microtubules. Overall, 100 mg of total cell lysates were subjected to the actin cosedimentation assay using an actin binding spin-down assay kit containing non-muscle actin (BK013, Cytoskeleton).   Figure 4B and 4C). Overall, 20 mg of total cell lysates was used for immunoblot analysis using the indicated antibodies. Note that depletion of myosin-II did not change the levels of MKlp2 (panel A), but it disrupted the cortical localization of MKlp2 ( Figure 4B, panel b). Additionally, depletion of MKlp2 did not measurably change the levels of myosin-II and MKlp1 (panel B). Relative band intensities to control siRNA are shown. (C) Immunofluorescence analysis of HeLa cells stably expressing GFP-EB1 in monopolar cytokinesis. Overall, 24 h after transfection with indicated siRNAs, the cells were subjected to monopolar cytokinesis by treatment with the Eg5 inhibitor Monastrol (100 mM; Santa Cruz Biotechnology) for 6 h and with the Cdk1 inhibitor purvalanol A (30 mM; Sigma) at 37uC for 20 min. Note that the mitotic spindles and the cell cortex (determined by myosin-II staining) are polarized (white asterisk in merged image; myosin-II is shown in purple) in control cells but not in MKlp2-depleted cells, although the tips of the mitotic spindles (determined by GFP-EB1) were able to contact the cell cortex (panel b, inset). (TIF) Figure S5 MKlp2(1-842) is selectively defective in recruiting Aurora B to the furrow but not to the end of monopolar spindles. (A) Co-localization of Dox-induced Flag-MKlp2(1-890) and Flag-MKlp2(1-842) with Aurora B in HeLa cells undergoing monopolar cytokinesis were determined by immunofluorescence analysis. Purvalanol A (30 mM) were added for 20 min, and the cells were fixed with ice-cold methanol. Note that Flag-MKlp2(1-890) localized together with Aurora B to the cell cortex and to the furrow (top panels, white asterisk). In contrast, Flag-MKlp2(1-842) only co-localized with Aurora B in a punctate staining pattern, representing the end of monopolar spindles (bottom panels) as shown in Figure 5A (panels f-h). (B) mCherry-MKlp2 localizes to the cell cortex together with endogenous Aurora B. Co-localization of ectopically expressed mCherry-MKlp2 with endogenous Aurora B to the cell cortex in HeLa cells undergoing monopolar cytokinesis upon Purvalanol A treatment (as indicated in Figure 6A) was determined by immunofluorescence analysis. The cells were fixed with ice-cold methanol. White bars represent 5 mm. Immunofluorescence analysis. HeLa cells were transfected with MKlp2 siRNAs (50 nM) to deplete endogenous MKlp2 and treated with Dox (5 mg/ml) to express the indicated Flag-MKlp2. Next, 20 h later, cells were treated with Monastrol (100 mM) for 6 h and briefly treated with myosin-II inhibitor (-)-Blebbistatin (100 mM) for 5 min to block furrowing activity, which was then followed by purvalanol A (30 mM) treatment for 15 min. Subsequently, cells were fixed in ice-cold methanol and evaluated for the cortical localization of MKlp2 using the indicated antibodies. Note that (-)-Blebbistatin did not prevent Flag-MKlp2(1-890) and Aurora B from localizing to the cell cortex (panels b, c). Furrowing activity was inhibited; therefore, MKlp2 and Aurora B localized to the cell cortex in a non-polarized manner (panels b, c). The polarization of microtubules and the cell cortex was also completely inhibited (panels b, c). In contrast, Flag-MKlp2(1-842) together with Aurora B failed to localize to the cell cortex, but they were accumulated at the end of monopolar spindles (panels d, e). Furthermore, Aurora B in MKlp2-depleted cells did not localize to the monopolar spindles or the cell cortex but retained the chromosomes (panel a), which confirms the specificity of our imaging analysis. (TIF) Figure S9 Sequential inhibition of Aurora B after polarizing cells does not alter the localization of centralspindlin at the ends of monopolar spindles. Immunofluorescence analysis. HeLa cells were subjected to monopolar cytokinesis. For panel a, purvalanol A (30 mM) and ZM447439 (2 mM) were added concurrently for 15 min. For panels b-d, ZM447439 was added sequentially after purvalanol A treatment for 10 min, and the cells were fixed 10 min after ZM447439 treatment in ice-cold methanol. Arrows (panel b) indicate MKlp1 localization at the ends of microtubules. Although the concurrent treatment of ZM447439 and purvalanol A suppressed the microtubule localization of MKlp1 (panel a) as previously shown [7], when Aurora B was sequentially inhibited after polarizing cells, MKlp1 still localized at the ends of microtubules (panel b). Moreover, Aurora B and MKlp2 largely localized to the polarized cortex. Under this condition, furrowing activity immediately ceased, and the furrows subsequently regressed ( Figure 6C), suggesting that continuous Aurora B activity at the growing furrow is necessary for furrow completion. (TIF)

Supporting Information
Movie S1 Time-lapse video microscopy recordings of HeLa cells stably expressing GFP-a-tubulin in monopolar cytokinesis. The images were captured every 5 min upon purvalanol A treatment (30 mM) using 2D-SIM. (AVI) Movie S2 The cells were treated as Movie S1 except included the addition of ZM447439 (2 mM) at 10 min after purvalanol A treatment (30 mM). Purvalanol A initially induced the polarization of mitotic spindles growing toward the growing furrow. The images were captured every 5 min upon ZM447439 addition using 2D-SIM. Sequential Aurora B inhibition efficiently depolarized the monopolar spindles and inhibited furrowing activity followed by furrow regression. (AVI)