|Home | About | Journals | Submit | Contact Us | Français|
The mitotic spindle is generally considered the initiator of furrow ingression. However, recent studies suggest that furrows can form without spindles, particularly during asymmetric cell division. In Dictyostelium, the mechanoenzyme myosin II and the actin cross-linker cortexillin I form a mechanosensor that responds to mechanical stress, which could account for spindle-independent contractile protein recruitment. Here we show that the regulatory and contractility network composed of myosin II, cortexillin I, IQGAP2, kinesin-6 (kif12), and inner centromeric protein (INCENP) is a mechanical stress–responsive system. Myosin II and cortexillin I form the core mechanosensor, and mechanotransduction is mediated by IQGAP2 to kif12 and INCENP. In addition, IQGAP2 is antagonized by IQGAP1 to modulate the mechanoresponsiveness of the system, suggesting a possible mechanism for discriminating between mechanical and biochemical inputs. Furthermore, IQGAP2 is important for maintaining spindle morphology and kif12 and myosin II cleavage furrow recruitment. Cortexillin II is not directly involved in myosin II mechanosensitive accumulation, but without cortexillin I, cortexillin II's role in membrane–cortex attachment is revealed. Finally, the mitotic spindle is dispensable for the system. Overall, this mechanosensory system is structured like a control system characterized by mechanochemical feedback loops that regulate myosin II localization at sites of mechanical stress and the cleavage furrow.
More than a century of research has indicated that cytokinesis contractility is generally regulated by the mitotic spindle (Wolpert, 1966 ; Hiramoto, 1990 ; Rappaport, 1996 ; Burgess and Chang, 2005 ). Components of these regulatory pathways include the kinesin-6 proteins that form complexes with signaling proteins such as MgcRacGap, which regulates the Rho pathway, or the chromosomal passenger complex proteins (inner centromeric protein [INCENP] and aurora kinase; Cooke et al., 1987 ; Glotzer, 2005 ; Li et al., 2008 ). The Dictyostelium kinesin-6 (also known as kif12), INCENP, and aurora kinase localize to the central spindle and the cleavage furrow at later stages of cytokinesis (Chen et al., 2007 ; Li et al., 2008 ). Although these pathways are believed to modulate the cortex mechanics that promote cytokinesis shape change, disruption or removal of the spindle after chromosome separation does not affect cytokinesis in at least some cell types (Hiramoto, 1956 ; von Dassow et al., 2009 ). In addition, the mitotic spindle may not be the primary director of contractility during asymmetric cell divisions (Cabernard et al., 2010 ; Ou et al., 2010 ). Thus cytokinesis contractility is regulated by multiple mechanochemical pathways (Surcel et al., 2010 ), which undoubtedly promote cytokinesis fidelity and versatility in order to maintain genomic stability (Fujiwara et al., 2005 ).
One critical component of the cytokinesis machinery is the force-sensitive, actin-binding mechanoenzyme myosin II. Since the classic experiments of Fenn and Huxley (Fenn, 1923 ; Huxley and Simmons, 1971 ), myosin II has been well understood to be load sensitive, and this load sensitivity is at the crux of how various muscle tissues function across diverse physiological processes. However, given that muscle is a late evolutionary arrival, it has been unclear how this force sensitivity might contribute to cytokinesis dynamics and perhaps promote contractility in the absence of obvious spindle cues. To begin to address this question, we previously tested whether dividing Dictyostelium cells respond to applied mechanical stress and whether this sensitivity relies on the load-dependent nature of the myosin II motor domain. Indeed, dividing cells are exquisitely sensitive to applied mechanical stress. The mechanosensor consists of at least three parts: the myosin II motor itself with force amplification through the lever arm, the dynamics of myosin II bipolar thick filament assembly/disassembly, and actin filament anchoring through cortexillin I (Effler et al., 2006 ; Ren et al., 2009 ). A physical model confirms that the mechanical stress–sensitive cooperative binding of myosin II to actin is sufficient to mediate myosin's cellular mechanosensitive accumulation (Luo et al., 2012 ). In particular, the cellular mechanosensory response can be accounted for quantitatively based on the assumptions that myosin II cooperatively binds actin filaments in the isometric state (Orlova and Egelman, 1997 ; Tokuraku et al., 2009 ) and that myosin II binding to actin filaments enhances bipolar thick filament assembly (Mahajan et al., 1989 ).
Dictyostelium and mammalian bipolar thick filament assembly/disassembly dynamics is regulated by heavy chain phosphorylation (Egelhoff et al., 1993 ; Bosgraaf and van Haastert, 2006 ; Breckenridge et al., 2009 ). In contrast, cortexillin I regulation at the molecular level is much less well understood. Cortexillin I may interact with rac1 through IQGAP1 (also known as DGap1) and IQGAP2 (also known as GapA; Faix et al., 1998 , 2001 ; Lee et al., 2010 ; Mondal et al., 2010 ). Cortexillin II, which is 60% identical in amino acid sequence to cortexillin I, does not appear to interact with IQGAP proteins. In contrast to cortexillin I (cortI)-null cells, cortexillin II (cortII)-null cells also do not exhibit an apparent cytokinesis phenotype (Faix et al., 1996 ; Mondal et al., 2010 ). On the other hand, IQGAP2 is essential for cytokinesis, as cells depleted of IQGAP2 fail to divide and are highly multinucleated (Adachi et al., 1997 ). The role of IQGAP1 in cytokinesis remains elusive, but it is believed to regulate the actin cytoskeleton (Faix et al., 1998 ). In mammals, IQGAP proteins interact with the microtubule plus-end–binding protein CLIP-170 upon activation of rac1 and cdc42 and links microtubules to the actin cytoskeleton network in fibroblasts (Fukata et al., 2002 ). However, whether IQGAP regulates microtubule-based signaling pathways—specifically the ones emanating from the mitotic spindle in dividing cells—is less clear. Nevertheless, all of these observations point to different key components that play roles in cytokinesis. Integration of these pieces of information is required to understand how cytokinesis is regulated with high fidelity.
In this study, we characterized a mechanosensory system that fine-tunes the level of myosin II and cortexillin I at the cleavage furrow to facilitate division under various mechanical contexts. This mechanosensory system consists of the mechanical stress–induced, self-amplifying recruitment mechanism of myosin II and cortexillin I, which form the core mechanosensor module of the system. Mechanical stress is detected by the mechanosensor, which then triggers the accumulation of mitotic spindle signaling proteins kif12 and INCENP through IQGAP2. However, IQGAP2, kif12, and INCENP are not required for myosin II mechanosensitive localization. Therefore the IQGAP2, kif12, and INCENP accumulation reflects the conversion of the mechanical stress sensed by myosin II into their redistribution, and we refer to this form of mechanosensitive accumulation as mechanotransduction. We also uncover a complex relationship between IQGAP1 and IQGAP2, where IQGAP1 acts as an inhibitor of the mechanosensor. IQGAP2 is needed to relieve this inhibition. Cortexillin II is not required for mechanosensing, although cortexillin I/II (cortI/II) double mutants have severe membrane–cortex attachment defects, implying a role for cortexillin II in maintaining membrane–cortex attachment. Overall, IQGAP2, kif12, INCENP, and the mechanosensor proteins concentrate myosin II at the cleavage furrow. Thus this mechanosensory system is composed of multiple feedback loops, which may ensure the flexibility and high fidelity of cytokinesis.
Although spindle signals direct the contractile machinery (CM; specifically cortexillin I and myosin II) to initiate furrow formation (Cooke et al., 1987 ; Glotzer, 2005 ; Li et al., 2008 ), mechanical stress can also direct the recruitment of the CM (Effler et al., 2006 ; Ren et al., 2009 ). Therefore we tested whether the mechanical stress has any effect on spindle signaling molecules. Because Dictyostelium cells do not have MgcRacGAP or Rho kinase, we focused on kinesin 6 (kif12), INCENP, and aurora kinase (Lakshmikanth et al., 2004 ; Chen et al., 2007 ; Li et al., 2008 ). We expressed green fluorescent protein (GFP)–tagged kif12, INCENP, and aurora kinase in WT cells. On application of mechanical stress using micropipette aspiration (MPA), kif12 and INCENP, but not aurora kinase, accumulated at the deformation site in a similar manner to the myosin II mechanosensitive response (Figure 1A). The accumulation of kif12 and INCENP was significantly higher than the intensity ratio (Ip/Io = 0.82) of soluble GFP or mCherry, which reflects the volume (thickness) ratio between the portion of the cell inside the micropipette as compared with the opposing cortex (Figure 1A, dot plot). The recruitment of kif12 to the mechanically stressed region is myosin II dependent because in myosin II (myoII)-null cells, kif12 failed to accumulate at the micropipette (Figure 1B). Kif12 accumulation is also independent of the mitotic spindle. After ~5 min of 10 μM nocodazole treatment, which disrupts the spindle, kif12 still accumulated to the micropipette (Figure 1B). Because kif12 and INCENP, as well as myosin II and cortexillin I, can be recruited to the mechanically deformed cortex, we examined whether kif12 and INCENP are required for myosin II mechanosensitive localization. In the absence of either kif12 or INCENP, myosin II was recruited to the deformation site at wild-type (WT) levels (Figure 1C). Furthermore, the kif12 and incenp mutants had nearly normal cortical tension (Figure 1D). These results imply that although kif12 and INCENP undergo mechanosensitive accumulation, they are neither required for mechanosensing nor crucial for cortical tension maintenance. Therefore mechanical stress is detected by the mechanosensory module formed by myosin II and cortexillin I and transmitted downstream to recruit kif12 and INCENP as part of a mechanotransduction pathway (Figure 1E).
Cortexillin I is one central component of the mechanosensory module. Cortexillin I is believed to localize to the cleavage furrow by forming complexes with rac1 (encoded by three nearly identical genes—rac1A, B, and C), IQGAP1, and IQGAP2 (Faix et al., 1998 , 2001 ; Lee et al., 2010 ; Mondal et al., 2010 ). To see how cortexillin I might be regulated in mechanosensing, we examined the regulators of cortexillin I, as well as of cortexillin II, to see whether they are recruited in response to mechanical stress. We expressed GFP-tagged rac1A in WT cells and GFP-cortexillin II, GFP-IQGAP1, and GFP-IQGAP2 in cortII-, iqgap1-, and iqgap2-null cells, respectively. Performing MPA on dividing cells, we found that IQGAP2, but not cortexillin II, IQGAP1, or rac1A, accumulated at the deformation site (Figure 2A). We also obtained similar results when we expressed GFP-IQGAP1 and GFP-IQGAP2 in WT cells (Supplemental Figure S1A and Figure 2B). Both myosin II and cortexillin I were required for the recruitment of IQGAP2 to the micropipette (Figure 2B). Moreover, we observed that IQGAP2 strongly accumulated at blebs (arrowhead in Figure 2B) in cortI-null cells, which have a less stable cortex.
We then tested whether cortexillin II, rac1, IQGAP1, and/or IQGAP2 are required for myosin II mechanosensitive accumulation by examining GFP-myosin II in cortII-, rac1A/C double-mutant (the rac1A/B/C triple mutants appear to be inviable), and iqgap1- and iqgap2-null cells. Myosin II mechanosensitivity was defective in iqgap2-null cells but remained normal in cortII, rac1A/C, and iqgap1 mutants (Figure 3A, images and dot plot). When the iqgap2 mutant was complemented with GFP-IQGAP2, myosin II mechanosensing was restored to WT levels (Figure 3B). Initially, one might conclude that IQGAP2 is required for myosin II mechanosensing. However, it is also possible that IQGAP1 and IQGAP2 act antagonistically to regulate mechanosensitivity but are not part of the mechanosensory module. Therefore we measured myosin II mechanosensitive localization in iqgap1/2 double-mutant cells and found that the myosin II accumulation was WT like in this strain (Figure 3A). IQGAP2 then is not an integral part of the mechanosensory module but instead suppresses IQGAP1-mediated inhibition of myosin II mechanosensitive localization. Consistently, overexpression of GFP-IQGAP1 in WT cells, which still express endogenous IQGAP1, suppressed myosin II–mediated mechanosensing (Figure 3B). Furthermore, iqgap1/2 double-mutant cells expressing GFP-IQGAP2 had normal myosin II mechanosensitive localization. On the other hand, expression of GFP-IQGAP1 in the double mutant inhibited myosin II mechanosensitive localization (Supplemental Figure S1B). In addition, cortexillin I responded similarly to myosin II in rac1A/C, iqgap1, iqgap2, and iqgap1/2 mutant cells (Figure 3C). Overall, these results are consistent with the notion that cortexillin I and myosin II constitute the mechanosensory module and are both regulated by the IQGAPs.
When we checked myosin II recruitment in cortI/II double-mutant cells, the Ip/Io intensity ratio showed no difference as compared with WT (Figure 3A, dot plot, Student's t test: p = 0.41). However, cortI/II cells have a less stable cortex–membrane attachment, resulting in recruitment of myosin II to blebs and dramatic membrane–cortex rupture sites inside or outside the pipette (Supplemental Figure S2A). Because IQGAP2 localized strongly to blebs in cortI null cells (Figure 2B), we checked whether IQGAP2 also accumulates at the blebs of cortI/II null cells. We found that indeed IQGAP2 localizes to blebs in the cortI/II mutant cells, and the Ip/Io intensity ratio did not show a significant difference from WT (Supplemental Figure S2B). Given the level of membrane–cortex rupture observed in nearly every cortI/II mutant cell analyzed, this mode of myosin II mechanosensitive localization is different from that in the rest of the mutant cell lines we characterized. In addition, cortexillin II is not required for cortexillin I mechanosensitive localization, which was verified by analyzing GFP-cortexillin I expressed in cortI/II mutant cells (Figure 3C, dot plot). Thus cortexillin II does not appear to play a direct role in the myosin II– and cortexillin I–mediated mechanosensitive localization.
Finally, we measured the cortical tension of all of these mutant cell lines (Figure 3D). The IQGAPs are important for cortical tension maintenance: cortical tension was reduced by 35, 60, and 80% in iqgap1, iqgap2, and iqpgap1/2 nulls, respectively. The cortII cells had a slight 16% increase in cortical tension. The cortical tension was reduced by 35 and 60% in cortI and cortI/II, respectively. These results demonstrate that the changes in mechanosensitivity do not simply correlate with changes in cortical tension levels.
In the preceding sections, we found that kif12 accumulation at the mechanically stressed cortex is myosin II dependent, and IQGAP2 maintains myosin II and cortexillin I mechanosensing by suppressing IQGAP1. Therefore IQGAP2 could mediate the mechanosensitive accumulation of kif12. Indeed, the level of GFP-kif12 recruitment to the micropipette is significantly lower in the iqgap2 mutants than in the WT parental cells (Figure 4A). Because myosin II mechanosensitive accumulation in the iqgap2 mutant is suppressed but in the iqgap1/2 double-mutant cells is active, we tested whether having active myosin II mechanosensing is sufficient to trigger the recruitment of kif12. In the absence of both IQGAP1 and IQGAP2, kif12 failed to be recruited to the micropipette despite having active myosin II mechanosensitive accumulation (Figure 4A). By coexpressing GFP-kif12 and mCherry-myosin II in the iqgap1/2 mutants so that both proteins could be tracked simultaneously, we further confirmed that myosin II, but not kif12, showed active mechanosensitive localization in these double-mutant cells (Supplemental Figure S1C).
Because kif12 in iqgap2 and iqgap1/2 mutants failed to be recruited by mechanical stress, we then determined whether the kif12 levels at the cleavage furrow depended on IQGAP2. We compared the mean intensity ratio of GFP-kif12 at the furrow to that at the polar cytoplasm (If/Ip cyto) in WT, iqgap1, iqgap2, and iqgap1/2 cells (Figure 4B). WT and iqgap1-mutant cells showed higher GFP-kif12 levels at the furrow than at the polar cytoplasm (Figure 4, B and C). Conversely, iqgap2 and iqgap1/2 mutants showed little GFP-kif12 at the furrow (Figure 4, B and C). Thus IQGAP2 is required for kif12 accumulation at the mechanically stressed cortex and at the cleavage furrow.
We observed other cell division defects in the iqgap2, iqgap1/2, and kif12 mutants. The iqgap2 and iqgap1/2 cells had more severe spindle morphological defects than did WT and iqgap1 cells (Figure 4, B and D). The percentage of cells that completed cytokinesis was also much lower in iqgap2 and iqgap1/2 mutant cells than in WT and iqgap1 cells (Figure 4E). In contrast, we did not see spindle morphological defects in the kif12 mutant (n = 9) or its WT parental cell line (n = 22), although kif12 mutant cells did show severe defects in cytokinesis completion (with 2 of 14 cells vs. 32 of 32 WT cells completing cytokinesis). Cytokinesis in kif12 cells was usually arrested at the stage at which a thin intercellular bridge connected the two daughter cells, and this bridge often fused back to form binucleated cells. Thus spindle morphology by itself does not simply correlate with cytokinesis success in these strains. Overall, these results highlight the central role of IQGAP2 in maintaining normal mitotic spindle morphology and serving as a mediator that transmits mechanical signals detected by the CM mechanosensor to kif12 (Figure 4F).
Because our goal is to define the network that regulates myosin II accumulation at the cleavage furrow, we quantified the mean furrow-to-pole intensity ratio of GFP-myosin II in WT and mutant cells undergoing cytokinesis unperturbed on surfaces. To simplify the analysis, we focused on cells in which the furrows had ingressed to around the crossover distance Dx, which is where the furrow length and diameter are equal and the myosin II amounts have plateaued (Zhang and Robinson, 2005 ; Ren et al., 2009 ). In kif12- and incenp-null cells, the myosin II furrow-to-pole intensity ratio (If:Ip) is comparable to the If:Ip ratio of both WT and the incenp rescued cells (Figure 5, A and B, graphs). Thus kif12 and INCENP are not required for myosin II accumulation when cells divide unperturbed on surfaces.
We then analyzed cortI, cortII, iqgap1, and iqgap2 single- and double-mutant cells to determine whether they show alterations in myosin II furrow accumulation. Only the iqgap2-null mutant cells showed a statistically significant reduction in myosin II accumulation at the cleavage furrow when they divided unperturbed on surfaces (Figure 5C, graph). Of interest, compared with WT cells, the cortI-null cells had slightly reduced but statistically indistinguishable myosin II levels at the cleavage furrow. This could be due to the existence of cortexillin II. Therefore we checked the furrow localization of cortexillin II in the absence of cortexillin I by expressing GFP-cortexillin II in cortI/II null cells. Indeed, cortexillin II localized to the cleavage furrow in these cells, as well as in the rac1A/C mutants (Supplemental Figure S3). However, the important role of the IQGAPs and the distinct regulation of cortexillin II were further highlighted by the observation that cortexillin II did not show significant cleavage furrow enrichment in iqgap1/2-null cells, even though these cells did localize myosin II (Supplemental Figure S3 and Figure 5C).
Because only the iqgap2 mutants showed a significant reduction in myosin II levels at the cleavage furrow, simply examining unperturbed cells dividing on surfaces may not reveal the full richness of the systems that regulate myosin II accumulation. Therefore another assay is required to mechanically challenge the dividing cells to determine how these pathways synergize to regulate myosin II cleavage furrow accumulation.
Yumura et al. (1984 ) demonstrated that dividing cells gently compressed under a thin layer of agarose showed a significant increase in myosin II localization at the cleavage furrow We used this method of applying a uniform two-dimensional global mechanical stress to cells and analyzed the If:Ip ratios. First, we assayed soluble GFP, which was uniformly distributed throughout the cytoplasm and had an unchanged If:Ip ratio under agarose overlay (Supplemental Figure S4). This confirms that the changes from compression by agarose overlay detailed later are not simply due to the imaging of flattened cells. After this test, we examined fluorescent protein–labeled myosin II in WT cells and each of the various mutant cell types characterized earlier.
In WT cells, GFP-kif12 rescued kif12-null cells and GFP-INCENP rescued incenp-null cells, the myosin II If:Ip ratio increased threefold to sevenfold over uncompressed cells. The kif12 and incenp mutants showed little or no increase in this ratio under agarose overlay compared with the unperturbed state (Figure 6, A and B). This ratio could increase if If goes up, Ip goes down, or a combination of the two occurs. Therefore we measured the polar-to-cytoplasm intensity ratio (Ip/Ic) and found that these were significantly different between these strains (e.g., WT Ip/Ic = 0.93 ± 0.06 vs. kif12 Ip/Ic = 1.2 ± 0.1; Student's t test: p = 0.02). This rise of polar myosin II in the kif12 and incenp mutants was expected because the myosin II localization in these cells is still mechanoresponsive (Figure 1C). However, this amount of polar myosin II increase was not sufficient to account for the decrease in the If/Ip ratio, indicating the kif12 and INCENP focus myosin II at the furrow in response to mechanical stress. These results support the role of spindle signaling proteins kif12 and INCENP in directing myosin II localization to the cleavage furrow.
Because kif12 recruitment to the deformation site induced by MPA is mitotic spindle independent (Figure 1B) and previously we found that myosin II mechanosensing does not require the mitotic spindle (Effler et al., 2006 ), we tested whether the strong myosin II enhancement at the cleavage furrow under agarose overlay is also spindle independent. We labeled the WT cells with GFP–myosin II and red fluorescent protein (RFP)–tubulin so that the spindle could be tracked. Dividing cells were first treated with 10 μM nocodazole and monitored until the spindles were disassembled. Cells were then compressed by agarose overlay. These dividing cells had a similar level of GFP–myosin II If:Ip ratio as the dimethyl sulfoxide (DMSO)–treated control cells (Figure 6C). Therefore the mechanical stress–induced recruitment of myosin II to the cleavage furrow is mitotic spindle independent.
Next we quantified the GFP–myosin II levels at the cleavage furrow in the WT, single and double cortexillin mutant, and single and double iqgap mutant cells under agarose overlay. In contrast to cytokinesis on surfaces where only the iqgap2-null cells showed a statistically significant reduction in cleavage furrow myosin II, a much richer set of differences and trends was observed with agarose overlay. Without IQGAP2 (analysis of variance [ANOVA]: p < 0.0001) or cortexillin I (p = 0.0002), the GFP–myosin II If:Ip ratio is about one-third of WT levels (Figure 6D). In the iqgap1/2-null cells, the GFP–myosin II If:Ip ratio was slightly higher than in the iqgap2 and cortI single mutants (Figure 6D). This ratio for iqgap1/2-null cells, however, is still much lower than that of WT (p = 0.0006) and iqgap2 cells rescued with GFP-IQGAP2 (p = 0.03; Figure 6D). The If:Ip ratio in cortII nulls is significantly lower than in WT (70% of WT; p = 0.008) but is not significantly different from that in iqgap2-rescued cells (Figure 6D, p = 0.13). In the absence of both cortexillin I and II, the If:Ip ratio is about half of the WT level (p = 0.0008). Finally, the iqgap1 single-mutant cells had a GFP–myosin II If:Ip ratio comparable to that of WT and iqgap2-rescued cells (Figure 6D). Thus, by perturbing components that control myosin II mechanosensing and kif12/INCENP mechanosensitive accumulation, we observed predictable changes in the levels of myosin II recruitment to the cleavage furrow. These results suggest that the mechanosensory system helps tune the myosin II levels at the cleavage furrow cortex.
Finally, we used one more approach to test whether myosin II mechanosensitivity contributes to cleavage furrow amplification due to global mechanical stress. One of the central ideas about the myosin II–based mechanosensory system is that myosin II binds actin cooperatively in the isometric state and accumulates at sites of mechanical stress (Ren et al., 2009 ; Tokuraku et al., 2009 ; Uyeda et al., 2011 ; Luo et al., 2012 ). Previous studies demonstrated that this principle applies to myosin II mechanosensing by analyzing the myosin II recruitment in myoII-null cells rescued with either WT myosin II (with the normal essential light-chain- and regulatory light-chain–binding sites), a long-lever-arm myosin II with an extra essential light-chain–binding site (2xELC), or a short-lever-arm myosin II with both light-chain–binding sites deleted (ΔBLCBS). The pressure dependence of the mechanical stress–induced accumulation correlated inversely with the motors' lever arm lengths (i.e., the longer the lever arm, the lower is the stress required to drive accumulation; Ren et al., 2009 ). This observation indicates that myosin accumulates because the motor stalls on the actin in response to stress.
To verify whether the mechanosensitive property of myosin II is important for the stress-induced amplification of the myosin levels at the cleavage furrow, we explored how ΔBLCBS myosin II responds to the uniform mechanical stress from agarose overlay. We found that accumulation of citrine (CIT)-labeled ΔBLCBS in the furrow region is not significantly different from that of CIT-WT myosin II under agarose compression (both myosins were expressed in myoII-null cells labeled with GFP-tubulin; Figure 7, A and B). However, ΔBLCBS did not concentrate at the lateral edges of the cleavage furrow cortex as strongly as WT myosin II did. This observation is reflected in the line-scan graphs and the intensity ratios of the furrow edges relative to the furrow center (If/If center; Figure 7, A and C) and was verified by confocal microscopy (Supplemental Figure S5A). Thus the WT myosin II lever arm is required for the lateral cortex concentration at the cleavage furrow, which is predicted to be the region with the highest cortical tension (Liu et al., 1996 ). These observations are also consistent with an earlier agarose overlay study showing that the nonhydrolyzer myosin II mutant (without ATPase and motor activity) localized at the cleavage furrow region but failed to be concentrated at the lateral edges of the cleavage furrow cortex (Yumura and Uyeda, 1997 ). Thus full myosin II function and mechanosensing correlate with myosin II's ability to incorporate strongly at the lateral cleavage furrow cortex.
Finally, we returned to unperturbed cells to test whether the lever arm affects cleavage furrow ingression dynamics (Zhang and Robinson, 2005 ; Reichl et al., 2008 ). We measured the furrow ingression dynamics of myoII-null cells and myoII-null cells rescued with CIT-labeled WT, 2xELC, or ΔBLCBS myosin II proteins. We found an inverse relationship between the rates of furrow ingression and the lever arm length. Furrows of cells expressing WT and the 2xELC were slower than the myoII-null furrows during late stages of furrow ingression (Figure 7D and Supplemental Figure S5, B–D). Cells expressing ΔBLCBS myosin II showed myoII null–type dynamics, but these furrow-thinning trajectories did not collapse onto a single universal curve (Figure 7E). Instead, they showed a broad range of times at which the furrow ingression dynamics transitioned to the fast, final furrow-thinning phase. These observations support the notion that the WT and 2xELC myosin II motors are experiencing mechanical stress and operating near stall during furrow ingression (Zhang and Robinson, 2005 ; Reichl et al., 2008 ). This mechanical stress may then trigger a mechanosensory control system to tune the myosin II levels (Figure 8).
Many biological tasks, ranging from allosteric protein–protein interactions and immune function to information flow in the brain that governs human behavior, rely on a control system that automatically ensures proper function in the presence of disturbances or perturbations (Monod et al., 1963 ; Tucker and Williamson, 1984 ; Takahashi and Yamada, 1998 ). These control systems rely on feedback to maintain quasi-equilibrium or homeostasis. Such a feedback control system has also been characterized in motility regulation in mammalian cells, which requires the synergistic coupling of F-actin, myosin II, and focal adhesion dynamics (Gupton and Waterman-Storer, 2006 ).
The network described here also has the hallmark of a control system (Figure 8). As the mitotic cell enters anaphase, the spindle elongates and delivers initial cues, which stimulate cleavage furrow formation (Figure 8A). Cells respond to applied mechanical stresses (similar in magnitude to those generated at the cleavage furrow; Zhang and Robinson, 2005 ) by recruiting CM to generate counteracting contractile stress, thereby modulating the localization of spindle signaling proteins (Figure 8B). In addition, this mechanosensory system appears to function at the cleavage furrow. After breaking the symmetry of the dividing cells, spindle microtubules, along with contractile stress in the cortex, drive CM recruitment to the cleavage furrow cortex (Figure 8C). The control system is sensitive to mechanical perturbations, such as the intrinsic stress that a cell normally experiences at the cleavage furrow or from external stresses imposed by the environment, which significantly amplify myosin II levels (Figure 8D). The amplification occurs at two places: myosin II/cortexillin I–mediated mechanosensing and then mechanosensitive accumulation of kif12 and INCENP mediated by IQGAP2. The baseline of cleavage furrow myosin II in both unperturbed and compressed states is found in the iqgap2 single-mutant cells. These iqgap2 mutants also failed to shows myosin II mechanosensing and kif12/INCENP mechanosensitive accumulation. The addition of mechanosensation (iqgap1/2 double mutants) increases myosin II cleavage furrow levels ~1.5-fold when applying compressive stress (as compared with the iqgap2-mutant baseline under compression). Overall, the intact WT system amplifies myosin II levels approximately fivefold in the presence mechanical stress. This amplification, along with the continuously increasing rates of myosin II accumulation in response to mechanical stress, suggest the presence of a feedback loop, a hallmark of many control systems (Ren et al., 2009 ; Luo et al., 2012 ).
This control system works as part of, or alongside, other modes of CM targeting, which include diffusion, active transport, regulatory factors, and/or cortical receptors (Zang and Spudich, 1998 ; Yumura et al., 2008 ; Fang et al., 2010 ; Uehara et al., 2010 ; Laporte et al., 2011 ). If the spindle microtubules are disrupted after the dividing cell elongates, the contractile stress at the furrow, along with the mechanosensory control system, are sufficient to ensure adequate recruitment of CM. The spindle independence suggests that the control system might account for how cells divide under diverse mechanical constraints and in the absence of a mitotic spindle (Hiramoto, 1956 ; Cabernard et al., 2010 ; Ou et al., 2010 ).
We discovered an intricate relationship between cortexillin I–interacting proteins IQGAP1 and IQGAP2. Neither IQGAP is essential for mechanosensing; however, IQGAP2 is required for kif12/INCENP mechanosensitive recruitment and to counteract the inhibition by IQGAP1, allowing mechanosensing to occur. These observations suggest a mechanism by which different signals—biochemical and mechanical—may be discriminated by a similar set of cytoskeletal proteins. The inhibition by IQGAP1 may also dampen the system, preventing it from being overly sensitive to mechanical inputs. Moreover, IQGAPs are found to be important for maintaining normal cell mechanics, and IQGAP2 specifically is crucial for cytokinesis. It is also known that IQGAP1 and 2 interact with the C and N-termini of cortexillin I, respectively (Faix et al., 1996 ; Mondal et al. 2010 ). The N-terminus of cortexillin I consists of a calponin-homology actin-binding domain, whereas the C-terminus is important for actin bundling and lipid binding (Stock et al., 1999 ). Previously, we found that the N-terminus of cortexillin I is dispensable, but the C-terminus including the coiled-coil domain is essential, for myosin II mechanosensitive localization (Ren et al., 2009 ). Cortexillin I may be trapped in a nonmechanosensitive conformation when IQGAP1 interacts with its C-terminus. On the other hand, IQGAP2 may enable cortexillin I mechanosensitivity by interacting with its N-terminus, freeing up the C-terminus for myosin II–mediated mechanosensing and actin and lipid binding. Perhaps in this conformation, cortexillin I is more effective at transmitting mechanical stress from the plasma membrane to the actin cortex, which may promote myosin II's mechanosensitivity. It is significant that IQGAPs are becoming central players in cell division in a variety of organisms. For example, budding and fission yeast IQGAP proteins are crucial for anchoring myosin II at the division site (Fang et al., 2010 ; Laporte et al., 2011 ). IQGAP2 clearly plays a central role in the control system described here and may constitute one of the myosin II cortical anchors in Dictyostelium as well (Robinson, 2010 ).
Unlike cortexillin I, cortexillin II does not seem to be directly involved in the mechanosensory system, as it does not localize in response to mechanical stress. Cortexillin II does contribute to cortical tension and stability, as the cortI/II double mutants display a much reduced cortical tension and higher levels of membrane–cortex rupture than does either single mutant. Cortexillin II helps maintain cortex–membrane connections and a wild type–like actin architecture (Shu et al., 2012 ). In the context of a more-WT-like architecture, cortexillin I is essential for myosin II–mediated mechanosensitive accumulation.
Our results indicate that mechanical stress recruits kif12 and INCENP but not aurora kinase. On the basis of previous studies by De Lozanne's group, kif12 is required for the localization of INCENP at the cleavage furrow (Chen et al., 2007 ). Both kif12 and INCENP are important for aurora kinase to target to the central spindle, which puts aurora at the bottom of this pathway (Li et al., 2008 ). Of interest, myosin II is required for normal localization of aurora and INCENP during cytokinesis (Chen et al., 2007 ; Li et al., 2008 ). Consistently, we did not observe kif12 localization at the cleavage furrow of myoII-null cells (unpublished data). Therefore myosin II accumulation from mechanical stress, in addition to signals emanating from the spindle, can localize kif12 and INCENP as part of the mechanosensitive control system. Similarly, the actomyosin system, microtubules, and aurora B kinase ensure positive feedback of symmetry breaking in dividing mammalian cells (Hu et al., 2008 ).
Kif12 and INCENP also play multiple roles in cytokinesis, particularly the abscission step and in recruiting myosin II when cells are challenged with mechanical stress. The kif12-null cells were previously reported to have problems accumulating myosin II at the cleavage furrow in unperturbed cells (Lakshmikanth et al., 2004 ). However, we detected normal myosin II cleavage furrow accumulation in the two kif12-null cell lines (same lines used in Lakshmikanth et al., 2004) , as well as in kif12 RNA interference cells when the cells were dividing unperturbed on surfaces (unpublished data). Because kif12 mutants have severe cytokinesis defects, including a longer time to cytokinesis completion/failure, the interpretation of myosin II accumulation may be sensitive to image sampling frequency (i.e., phototoxicity may become a problem unless the time to complete cytokinesis is corrected for, which was not done in the work of Lakshmikanth et al., 2004 ). Nevertheless, our observations that kif12 is part of a control system that tunes myosin II accumulation is in agreement with the main idea from Lakshmikanth et al. (2004 ), which is that kif12 participates in myosin II cleavage furrow accumulation.
Finally, the mechanosensory system has a natural shut-off mechanism. Because the myosin II/cortexillin I sensor depends on myosin II lever arm length (Ren et al., 2009 ), this suggests that it is the isometric, cooperative actin-binding state of the myosin II motor that senses the stress (Tokuraku et al., 2009 ; Luo et al., 2012 ). As more myosin II heads accumulate, the average force/head decreases, allowing the heads to exit the cooperative actin-binding state. The myosin II heavy chain kinases then reset the level of myosin II assembly, maintaining the free pool of myosin II monomers (Yumura et al., 2005 ; Luo et al., 2012 ). Thus, implicit in the myosin II mechanochemical and regulatory systems is the shut-off valve for the control system. Myosin II–based feedback control may be fundamental for other contractile systems such as those that drive tissue morphogenesis (Fernandez-Gonzalez et al., 2009 ; He et al., 2010 ) and focal adhesion maturation (Kuo et al., 2011 ).
Dictyostelium discoideum strains used in this study are listed in Supplemental Table S1. All cells were cultured in enriched HL-5 media (1.4× HL-5 enriched with 8% FM) with penicillin and streptomycin at 22°C on 10-cm Petri dishes (Robinson and Spudich, 2000 ). Wild-type strains include KAx3, Ax3:Rep orf+ (HS1000), and rescued mutant strains (Ruppel et al., 1994 ; Robinson and Spudich, 2000 ; Lee et al., 2010 ). The myoII cells (Ruppel et al., 1994 ); iqgap1, iqgap2, and iqgap1/2 cells (Lee et al., 2010 ); cortI, cortII, and cortI/II cells (Robinson and Spudich, 2000 ; Lee et al., 2010 ); kif12 cells (Lakshmikanth et al., 2004 ); and incenp cells (Chen et al., 2007 ) have been described previously. Where possible, strains were confirmed to have WT endogenous levels of myosin II, cortexillin I, and cortexillin II proteins (as appropriate). Wild-type and mutant strains were transformed with plasmids carrying fluorescently labeled genes of interest. All plasmids, RFP-tubulin, GFP-myosin II, citrine-WT myosin II, citrine-ΔBLCBS myosin II, GFP-IQGAP1, GFP-IQGAP2, GFP-cortexillin I, GFP-kif12, GFP-INCENP, and GFP-aurora, have also been described previously (Lakshmikanth et al., 2004 ; Effler et al., 2006 ; Chen et al., 2007 ; Li et al., 2008 ; Ren et al., 2009 ; Lee et al., 2010 ). GFP-rac1A and GFP-cortexillin II were prepared by amplifying both genes from a cDNA library and cloning into a modified version of the pDM181 vector, which includes a GFP-fusion insert. For control strains, cells were transformed with the empty vector and/or the DMSO carrier (for pharmacological experiments) as appropriate. In addition, only the matched parental strain and the rescued strain were considered as WT controls for each mutant cell line. Transformation was achieved by electroporation using a Genepulser-II electroporator (Bio-Rad, Hercules, CA). Transformed cells were selected with either 10–15 μg/ml G418 or 15-50 μg/ml hygromycin, or both drugs when two plasmids were transformed at once. Fluorescent reporters were confirmed to have similar expression levels (for the same reporter) across strains by comparing cytoplasmic fluorescence intensities.
The instrumental and experimental setups were previously described in detail (Effler et al., 2006 ). For the mechanosensing experiments, 0.2–0.5 nN/μm2 of pressure was applied to the cell cortex with a micropipette (2–3 μm in radius, Rp). For cortical tension measurements, the aspiration pressure was increased to the equilibrium pressure (ΔP) in which the length of the cell inside the pipette (Lp) was equal to Rp. The effective cortical tension (Teff) was determined by applying the Young–Laplace equation: ΔP = 2Teff(1/Rp − 1/Rc), where Rc is the radius of the cell and ΔP is the equilibrium pressure when Lp = Rp (Derganc et al., 2000 ; Octtaviani et al., 2006 ). For nocodazole experiments using MPA, cells were incubated overnight in HL-5 media with 0.2% DMSO (the nocodazole carrier) to eliminate the effects from DMSO treatment. Then a dividing cell was located and 10 μM nocodazole was added. The spindle marked with RFP-tubulin was monitored for about 5 min until the spindle disassembled, and then the dividing cell was aspirated. Nocodazole-treated cells were compared with DMSO-treated control cells.
For imaging, cells were transferred from Petri dishes to imaging chambers and allowed to adhere for 15 min in growth media. After the cells adhered, the growth media was gently replaced with 2-(N-morpholino)ethanesulfonic acid (MES) starvation buffer (50 mM MES, pH 6.8, 2 mM MgCl2, 0.2 mM CaCl2). For confocal imaging, a Zeiss 510 Meta with a 63× (numerical aperture [NA] 1.4) objective was used (Carl Zeiss, Jena, Germany). For all other studies, cell imaging was performed in a temperature-controlled room at 22°C with a motorized Olympus IX81 microscope using a 40× (NA 1.3) objective and a 1.6× optovar (Olympus, Center Valley, PA), as described previously (Effler et al., 2006 ). ImageJ (National Institutes of Health, Bethesda, MD) was used for image analysis. Most cell strains were labeled with RFP- or GFP-tubulin to mark the mitotic spindle of the dividing cells. For the mechanosensing analysis, dividing cells were deformed by micropipette aspiration. For quantification, the logarithm of the ratio of the background-corrected mean pixel intensity of the cortex inside the pipette (Ip) to the opposite cortex outside the pipette (Io) was determined, that is, log(Ip/Io). For quantifying the myosin II furrow-to-pole intensity ratio, the average background-corrected pixel intensities of both sides of the furrow (If) and both of the poles (Ip) were measured. The furrow-to-pole intensity ratio (If/Ip) was calculated. The average background-corrected pixel intensities of the cytoplasm (Ic) were also measured so that ratios relative to cytoplasm could be determined. Please note that the term Ip is used in two different contexts, intensity in the pipette and intensity of the polar cortex.
For spindle morphology analysis, cells expressing GFP-kif12, GFP-tubulin, or RFP-tubulin were imaged and scored qualitatively for their spindle morphology. Cells with straight and symmetrically and centrally aligned spindles were considered to have normal morphology. Cells with curvy or asymmetrically displaced spindles were considered to have abnormal morphology.
The imaging and quantification methods were previously described in detail (Zhang and Robinson, 2005 ). In short, differential interference contrast movies of dividing cells were captured at 2-s intervals. Furrow diameter and length at each time point were measured using ImageJ. The point at which the furrow diameter (Df) equaled the furrow length (Lf) yielded the crossover distance (Dx). For rescaling, the Df was normalized by Dx and plotted against the shifted time axis, by which the time at which Dx was achieved was reset to 0 s.
Thin sheets of 2% agarose gel in sterile MES starvation buffer were prepared following the protocol developed by Fukui et al. (1986 , 1987 ), which is available from dictyBase (http://dictybase.org). After cells settled in the imaging chambers for 15 min, the media in the imaging chamber was gently removed and replaced by MES starvation buffer. Thin agarose sheets were carefully placed on the surface of MES buffer. MES buffer was slowly and almost completely removed to allow the agarose sheet to directly press on the cells in the imaging chamber. To avoid potential complications of the imaging chamber drying due to evaporation, fresh imaging chambers were replaced every 15 min to ensure optimum conditions for cell behaviors. We confirmed the success of cytokinesis of the WT cells under these conditions. We found that 80% (22/28) of WT cells completed cytokinesis, which is only a little lower than the success rate (95–98%) for WT cells dividing unperturbed on surfaces (Effler et al., 2006 ; Octtaviani et al., 2006 ). For nocodazole treatment with agarose overlay, cells were cultured in HL-5 media with 0.2% DMSO overnight. During the experiment, a dividing cell was located, and 10 μM nocodazole was treated. The spindle was monitored until it disassembled, after which an agarose sheet was applied.
Data sets were collected and analyzed using KaleidaGraph (Synergy Software, Reading, PA). Analysis of variance (ANOVA) or Student's t tests were performed using KaleidaGraph. Comparison of proportions was performed with SE = √[f(1 − f)/n], where f is the fraction of the population showing a behavior and n is the sample size. Only p values <0.05 were considered significant, and the calculated p values are included on the graphs, in the text, and/or in the figure legends.
We thank Jim Spudich for the kif12-null cell lines, the GFP-kif12 expression construct, and discussions regarding myosin II localization the kif12-null cells. We thank Cathryn Kabacoff for help generating cell lines and the Robinson lab for helpful feedback on the manuscript. This work is supported by National Institutes of Health Grants GM066817 (to D.N.R.) and GM86704 (to D.N.R. and P.A.I.), American Cancer Society Grant RSG CCG-114122 (to D.N.R.), and an American Heart Association predoctoral fellowship (to Y.-S.K).
This article was published online ahead of print in MBoC in Press (http://www.molbiolcell.org/cgi/doi/10.1091/mbc.E11-07-0601) on February 29, 2012.