Interactions between myosin and actin crosslinkers control cytokinesis contractility dynamics and mechanics

Summary

Introduction:

Contractile networks are fundamental to many cellular functions, particularly cytokinesis and cell motility. Contractile networks depend on myosin-II mechanochemistry to generate sliding force on the actin polymers. However, to be contractile, the networks must also be crosslinked by crosslinking proteins and to change the shape of the cell, the network must be linked to the plasma membrane. Discerning how this integrated network operates is essential for understanding cytokinesis contractility and shape control. Here, we analyzed the cytoskeletal network that drives furrow ingression in Dictyostelium.

Results:

We establish that the actin polymers are assembled into a meshwork and that myosin-II does not assemble into a discrete ring in the Dictyostelium cleavage furrow of adherent cells. We show that myosin-II generates regional mechanics by increasing cleavage furrow stiffness and slows furrow ingression during late cytokinesis as compared to myoII nulls. Actin crosslinkers dynacortin and fimbrin similarly slow furrow ingression and contribute to cell mechanics in a myosin-II-dependent manner. Using FRAP, we show that the actin crosslinkers have slower kinetics in the cleavage furrow cortex than in the pole, that their kinetics differ between wild type and myoII null cells, and that the protein dynamics of each crosslinker correlate with its impact on cortical mechanics.

Conclusions:

These observations suggest that myosin-II along with actin crosslinkers establish local cortical tension and elasticity, allowing for contractility independent of a circumferential cytoskeletal array. Furthermore, myosin-II and actin crosslinkers may influence each other as they modulate the dynamics and mechanics of cell shape change.

Introduction

Cytokinesis is one of the most elegant cellular shape changes, as a mother cell reforms into two daughter cells in as little as five minutes. Fundamentally mechanical, cytokinesis is driven by myosin-II and actin filaments, and signaling pathways emanating from the mitotic spindle ultimately lead to their accumulation along the equatorial region of the cell [1]. Myosin-II is a mechanoenzyme that uses the energy of ATP hydrolysis to move actin filaments. The actin filaments in combination with crosslinkers give cells their shape and mechanical properties. The actin crosslinking proteins stabilize actin filament interactions and tune the mechanical (rheological) properties of the actin network. Thus, the actin crosslinkers define the passive material properties of the network whereas myosin-II uses energy to modify this network, creating an active network [2, 3]. Because of their central importance to cell function, how actin networks and myosin-II motors control the mechanical properties of cells is of considerable interest. Reconstitution systems have been developed to explore how actin networks respond to mechanical stress (force per area, pressure) and deformation (response of the network to mechanical stress) and how myosin-II contributes to active and passive properties of these actin networks [4-6]. However, it is unclear what the relevant mechanical properties of dividing cells are, how they are generated molecularly, and how they contribute to cytokinesis cell shape change.

Classically, myosin-II is considered the primary force generator of cytokinesis, generating long-distance forces that deform the network. In this most general case, myosin-II pulls on dynamic actin filaments that are either crosslinked to other actin filaments or to the membrane, constricting the cleavage furrow cortex. In many, but not all, cell-types, these actin networks are further organized into concentric antiparallel arrays, allowing the myosin motors to pull the filaments, contracting the membrane in a purse-string fashion. However, neither Dictyostelium nor mammalian tissue culture cells require myosin-II for mitosis-coupled cell division if the cells are adherent, and recent studies have suggested other roles for myosin-II such as in removing actin filaments from the equatorial region during furrow constriction [7-9]. The actin crosslinking proteins link the filaments together so that when myosin-II pulls against the filament, tension on the filament can propagate into the crosslinked network. Even with this basic framework, it is not understood in any system how myosin-II and actin crosslinkers interact to contract the network nor how these factors control the dynamic features of furrow ingression. Also, because myosin-II pulls on filaments bound by the crosslinkers, the crosslinkers and myosin-II may modulate each other's activities. Finally, it is not understood how the cleavage furrows of wild type cells constrict in such a stereotypical fashion nor how cleavage furrow ingression can occur without myosin-II.

To address these questions, we use the model system Dictyostelium to study cytokinesis cell shape change. This organism performs cytokinesis in a similar fashion to many types of mammalian cell culture cells and is readily amenable to mechanical and genetic interaction studies. Using this system, we have discovered and are studying a two-module system of equatorial (myosin-II and the actin crosslinker cortexillin) and global/polar (RacE small GTPase and actin crosslinkers dynacortin, coronin, enlazin, and fimbrin) proteins that form the genetic basis of the shape control system that regulates cytokinesis contractility (this paper) [10-13].

Here, we establish that the actin network in the contractile zone of Dictyostelium cells is a meshwork, rather than the circumferential actin ring found in many cell-types. We then set out to uncover how actin and myosin-II interact to control cytokinesis contractility, using a variety of mechanical and dynamical approaches to study the contractile system. We conclude that during Dictyostelium cytokinesis, myosin-II generates a tension and stiffness differential between the furrow and polar cortex, the dynamics of actin crosslinkers vary spatially during cytokinesis, and that these crosslinker dynamics are altered in myoII null cells. Because changes in cell mechanics are the logical output of regulatory pathways that provide the spatiotemporal control of cytokinesis, this analysis offers an analytical framework for ultimately linking these regulatory pathways to the mechanical changes that drive cytokinesis shape change.

Results

The actin network is a meshwork in the Dictyostelium cleavage furrow

Using transmission electron microscopy (Pt-TEM) and scanning electron microscopy (Pt-SEM), we determined that the actin polymer network is assembled into a meshwork in the furrow region of dividing Dictyostelium cells (Fig. 1A; Fig. S1A-C; see Supplementary Methods). In images of wild type (n=73) and myoII null (n=6) dividing cells, we never observed a circumferential orientation of actin filaments at the equator; instead the filaments were assembled into a meshwork, similar to the cytoskeletons of interphase cells and polar cortices of dividing cells (Fig. S1A, D, E). Throughout cytokinesis the actin filament density at the furrow appeared to be nearly constant by Pt-TEM. We confirmed this by quantitative fluorescence imaging of rhodamine-phalloidin-stained cytoskeletons (Fig S1F). To complement the Pt-TEM and Pt-SEM, we used 3D-EM tomography (3D-EM) of plastic sections of plunge-frozen, freeze-substituted dividing cells, which also revealed a highly disordered actin filament network near the plasma membrane (Fig. 1B; Fig. S1G; Movies S1-S4; Supplementary Methods) [14]. Because the IMOD software package used to build the 3D-EM tomograms allows one to recover the dimensions of each structure drawn, we were able to analyze the actin filament length distribution, which gave an average filament length of 94 ± 57 nm (Mean ± SD; Fig. S1H). This compares well with the 200-nm mean length estimated previously for vegetative cells using kinetic criteria [15]. From the filament length and number, and the bridge volume, we determined the concentration of F-actin in the late stage intercellular bridge (Fig. S1H) to be 150 μM, which is reasonable considering the average polymeric actin concentration for vegetative cells is 70 μM [16].

Figure 1.

Actin filaments are organized into a meshwork and actin and myosin-II are not enriched in a uniform ring as revealed by platinum-shadowed transmission electron microscopy (Pt-TEM; panel A), 3D-electron tomography (3D-EM; panel B), 3D-deconvolution (3D-decon, panel C-E), and total internal reflection fluorescence (TIRF; panel F, lower images) microscopy. A. The actin network is observed at the furrow of a wild type cell by Pt-TEM. Scale bars, 2 μm and 500 nm. B. Rotated 3D-EM images of a model of a 0.5-μm section (derived from combining two adjacent sections) of the lower surface of a cleavage furrow reveal disordered actin filaments. Mitochondria are green, vesicles are cyan, plasma membrane is blue, and the actin filaments are yellow. The first 2 panels show the furrow model viewed from top and bottom, respectively. The third panel is the furrow viewed down the long axis of the furrow. Scale bar, 2μm, applies to all panels. The z-series of the raw EM data can be found in Sup. Movie S1. The corresponding movie of the model can be found at Sup. Movie S2. C. Non-uniform cleavage furrow cortical actin. TRITC-phalloidin staining of filamentous actin in a wild type cell. Inset shows a cross-section of the furrow actin where the actin is enriched along the lateral surface. Scale bar, 10 μm applies to all images. Equatorial localization of GFP-myosin-II and binucleation (DAPI) confirms cell was undergoing cytokinesis prior to fixation. D, E. Non-circumferential distribution of myosin-II thick filaments. Wild type cells expressing GFP-tubulin and GFP-myosin-II reveal that, like actin, myosin-II does not form a continuous ring at the cleavage furrow. D. Early stage dividing cell. E. Late stage dividing cell. The C-S images show the cross-sectional fluorescence intensities of the furrow. F. Epi-fluorescence and TIRF images indicate that myosin-II is not circumferentially-oriented at the basal region of the furrow. Note that at later stages of cytokinesis, the furrow lifts from the surface.

For independent approaches, we used 3D-deconvolution (3D-decon) of rhodamine-phalloidin-stained fixed cells to view the actin distribution and 3D-decon and total-internal reflection fluorescence (TIRF) imaging of live cells expressing GFP-myosin-II to observe the myosin-II distribution (Fig. 1C-F). Neither actin nor myosin-II had a continuous ring distribution at the equator by any method or at any stage of furrow ingression. Instead, the intensity of the actin and myosin-II was greatest at the lateral cortex of the furrow. This is especially apparent in the TIRF images where the disorganized network of the myosin-II thick filaments had the highest concentration along the lateral edge of the cleavage furrow cortex. In later stages of furrow-thinning, the bridge lifts from the surface and is not observable by TIRF microscopy. Birefringence imaging of dividing Dictyostelium cells compressed by a sheet of agarose had also revealed cytoskeletal filaments oriented both perpendicularly and parallel to the long axis of the furrow region [17]. Overall, there is no apparent uniform contractile ring of actin or myosin-II in the equator of dividing Dictyostelium cells grown on surfaces. Though many organisms use a concentric array of actin and myosin-II filaments (for example, [18-20]), some cell-types, such as mammalian NRK and Swiss 3T3 cells, also have a more disordered actin and myosin-II network [21, 22], similar to what we observe in Dictyostelium.

More generally than a particular polymer organization, the deformation of living actin networks depends on two key features: actively generated forces from myosin-II and from actin polymer dynamics coupled with cell traction and the resistance to deformation (stiffness) from actin-crosslinking proteins. Because the actin filaments are very short in Dictyostelium, crosslinking proteins likely play an important role in linking the filaments together to form an integrated network. Cytokinesis cell shape evolution may also depend on fluid dynamical features (similar to Laplace-like pressures that originate from surface tension and which minimize the surface area to volume ratios in liquids) from cortical tension through the crosslinked actin network that can help promote furrow ingression [13]. Therefore, to understand how the actin-myosin-II network contracts during cytokinesis, we dissected how myosin-II and actin crosslinkers control cytokinesis contractility dynamics and mechanics.

Role of myosin-II during cytokinesis

To assess how myosin-II impacts the dynamics and mechanics of cytokinesis contractility, we examined the morphology, dynamics and mechanics of interphase and dividing cells with altered myosin-II mechanochemistry (Table S1). We previously predicted that wild type furrowing occurred ∼50-fold more slowly than expected based solely on fluid dynamical considerations. Consistent with these predictions, we demonstrated that by removing the global proteins RacE and dynacortin, thereby reducing the proposed resistive stresses, the furrow-thinning rates increased ∼30-fold compared to wild type [13]. Additionally, a five-fold slower myosin (ΔBLCBS) that lacks light chain binding sites could rescue myoII null cytokinesis nearly to wild type levels [23]. In this study, we examined cytokinesis of myoII null cells complemented with 10-fold slower motor, S456L, which moves slower due to a 3-fold longer ADP-bound time and ¼ productive step-size [24]. This S456L motor expresses at wild type levels but only rescues the growth rates in suspension culture to 1/3 of wild type levels. However, the slower motor rescued the cytokinesis morphology and furrowing dynamics to wild type levels on surfaces (Fig. S2; Movies S5-S7). Remarkably, the cleavage furrows from cells expressing wild type or 10-fold slower myosin-IIs had indistinguishable furrow-thinning trajectories whereas the myoII null cleavage furrows contracted faster late in furrow ingression than either of the strains with myosin-II, indicating that myosin-II actually slows late stage furrow ingression (Fig. 2A; Table S2). Thus, the velocity of the motor is clearly not rate-limiting, and other processes must govern the furrow ingression rate (see below).

Figure 2.

Wild type myosin-II mechanochemistry is required for wild type interphase mechanics, but not cytokinesis mechanics and kinetics. A. Comparison of furrow-thinning trajectories in cells expressing wild type myosin-II or myosin-II S456L to myoII null cells shows that S456L is able to fully restore the uniform furrow-thinning kinetics of wild type dividing cells. Note that myoII null cells have a faster furrow-thinning rate at later stages of division. B.-D. Expression of myosin-II S456L in myoII cells does not fully recover wild type cellular mechanics during interphase. B. MyoII cells have lower viscoelasticity (|G*|) than wild type cells as measured by LTM. C, D. MyoII cells are more deformable (panel C) and have a lower effective cortical tension (panel D) as measured by MPA. The S456L cells have cortical mechanics similar to wild type cells at longer time-scales (10 rad/s, panel B) and smaller deformations (lower L_p/R_p, panel C) but are more like myoII cells at shorter time-scales (>10² rad/s, panel B) and larger deformations (larger L_p/R_p, panel C). E. Representative micrographs showing cells aspirated at metaphase and during cytokinesis at the pole and furrow. For the polar cortex, we aspirated at angles ranging from parallel to perpendicular to the spindle axis with no detectable differences in the level of deformability. F. The degree of deformability of wild type interphase and metaphase cells was not significantly different. During anaphase, the furrow was slightly less deformable than during metaphase, while the pole was more deformable than the furrow or metaphase cortices. G. Conversely, the furrow and pole of myoII cells were not significantly different from each other and both regions were much more deformable than the polar region of wild type cells. H. S456L reduces the level of deformability of the furrow and polar regions to wild type levels. Error bars indicate standard error of the mean. Sample sizes for panel D are shown on the histograms. Samples sizes for A are provided in Table S2; sample sizes for panel B are shown in the histograms in Fig. S3, and the calculated E values and sample sizes for C and F-H are provided in Table S3.

To determine how myosin-II contributes to cell deformability (a stiffer material is less deformable), we used two methods, laser-tracking microrheology (LTM) and micropipette aspiration (MPA). These methods draw upon very different principles and assumptions, but used in combination, they allow for cross-comparison and for different features of cortex mechanics to be assessed. In LTM, the surface-attached bead particles serve as non-invasive reporters of cortical stiffness (measured as complex viscoelastic moduli (|G*|); see Methods), which is important because myosin-II has load-dependent actin-binding that can alter the enzyme's duty ratio [25-27]. From LTM, phase angle information can also be extracted to relate the solid- to liquid–like properties of the cortex (see Methods). However, slow Brownian motions (on timescales > 200 ms) of particles are obscured by active force generation in the cell, limiting LTM's usefulness for measuring viscoelastic moduli to time-scales ≤100 ms [3]. Furthermore, it has not been feasible to apply LTM to dividing cells in a statistically rigorous manner. In contrast, MPA measures mechanics on longer time-scales but requires relatively large mechanical strains for the measurements to be made. MPA offers the ability to position micropipettes so that spatial mechanics can be assessed for cytokinetic cells (Fig. 2E). On the longest 100-ms-time-scale (10 rad/s) measured by LTM, the interphase cortex of wild type cells have a viscoelastic modulus (|G*|) of 0.1 nN/μm² (100 Pa), which from the power-law behavior might extrapolate to ∼0.07 nN/μm² at 1 s (Fig 2B). Because the beads are surface attached and may not be fully immersed in the network, this could be an underestimate of the viscoelastic modulus of the cortex. Yet, these LTM values agree well with the elastic modulus (E) of 0.1 nN/μm² obtained from the ΔP vs. L_p/R_p relationship measured by MPA (Fig 2C; Methods; Table S3). When the cells are more spherical as during interphase, MPA can also be used in a different way to measure an effective cortical tension (T_eff) (measured at L_p/R_p=1), which was 1 nN/μm for wild type cells (Fig. 2D; Supplementary Methods) [10]. Thus, the combination of these methods allows three parameters to be assessed: a frequency-dependent viscoelastic modulus with its phase angle (LTM), an effective cortical tension (MPA), and an elastic modulus (MPA). However, for some of the different genetic mutants and in some of the different cell cycle phases, the plots of ΔP vs. L_p/R_p have similar slopes but are offset (for example, wild type vs. myoII in Fig. 2C). The offsets are likely due to non-linearities of the cells' responses to small versus large deformations. As a result, we primarily interpret these data in terms of how the cell deforms in response to applied pressure (greater L_p/R_p at a given pressure implies greater deformability). Nevertheless, the calculated elastic moduli for each case are presented in Table S3.

We first analyzed interphase wild type, myoII null, and S456L cells using LTM and MPA (Fig. 2B-D; Fig. S3). By LTM, the myoII null cells had slightly lower viscoelastic moduli than the wild type cells across all frequencies while the S456L mutant cells were lower than wild type at high frequencies and comparable to wild type at low frequencies (Fig. 2B, Fig. S3). In the frequency range of the LTM measurements, all strains measured here were more solid-like (phase angle values at 100 rad/s were 13°-15° for all strains except for S456L, which had a phase angle of 11°; see Methods). By MPA, the myoII nulls were more deformable than the wild type cells and again the S456L mutant was intermediate between the two strains (Fig. 2C, Table S3). The effective cortical tension was ∼20% reduced for myoII null (which is similar to the 30% reduction observed by needle poking [28]) and S456L cells when compared to wild type control cells (Fig. 2D). Thus, consistent with its significant (10-fold) motility defect, S456L only partially rescues the myoII null interphase mechanical defect.

During mitosis (Fig. 2E,F; Table S3), wild type metaphase cells were indistinguishable from interphase cells. However, as predicted by classical models, the polar cortex became much more deformable (polar relaxation [29, 30]) while the equator stiffened slightly (equatorial stimulation [31, 32]) during telophase as compared to interphase cells (Fig. 2F). In myoII null cells, the equatorial and polar cortices were not significantly different from each other, but were significantly more deformable overall than wild type cells (Fig. 2G; Table S3). Perhaps explaining the similarity between wild type and S456L furrowing dynamics, S456L rescued the deformability of the equatorial and polar cortices to wild type levels (Fig. 2H; Table S3). Therefore, the slowly decreasing furrow diameters of wild type and S456L cells correlate with decreased deformability of the cleavage furrow cortex (Fig. 2A vs. Fig. 2F-H). S456L may provide more wild type function in the context of the cleavage furrow cortex where myosin-II becomes enriched and where the cortex is actively deforming (straining), which may put the myosin-II under greater mechanical load. It should be noted that we recently documented mechanosensory responses in dividing cells, which occur on timescales of ∼40 s after mechanical perturbation [33]. For the analysis presented here, we measured all genotypes within ∼20 s of manipulation, and we also followed GFP-myosin-II in the rescued myoII null (myoII: GFPmyoII) cells to make sure that we did not trigger mechanosensory responses in the timeframe of the experiment. Therefore, these data reflect the level of deformability of the cortex, not mechanosensory responses.

Complex Interaction of Myosin-II and Crosslinkers in the Actin Network

To prevent nonproductive filament sliding, myosin-II requires crosslinking proteins to couple short filaments to each other so that the network deforms as a whole. Therefore, the actin crosslinkers represent the other half of the contractile system. Two classes of crosslinkers (global and equatorial) have been uncovered through genetic interaction screens [10, 11]. Cortexillin-I is an equatorially enriched actin crosslinker with membrane binding sites that is involved in Dictyostelium cytokinesis contractility and cortical mechanics [25, 34]. Globally distributed dynacortin has emerged as an important component that contributes to cytokinesis furrowing dynamics by acting as a brake to slow furrow ingression [11, 13]. Because dynacortin overexpression produces enlarged, multinucleated cells (Fig. S4A) [11], we used this phenotype to screen for other factors that may act in the global pathway. We expressed a cDNA library in Dictyostelium cells and isolated the actin crosslinker fimbrin as a factor that produced enlarged, multinucleated cells that saturated at lower cell densities when overexpressed (Fig. S4A-C). Similar to dynacortin, fimbrin is a globally distributed actin crosslinking and bundling protein (Fig. S4D) [11]. Both dynacortin and fimbrin have similar apparent affinities for actin (K_d∼1μM) and cellular concentrations (∼1μM), and increase cortical tension when overexpressed (Fig. S5) [25, 35, 36].

Using LTM and MPA, we measured and compared the frequency spectra of viscoelastic moduli and cortical tensions of wild type and myoII null cells devoid of these crosslinkers. Interphase cells lacking myosin-II or dynacortin have lower viscoelastic moduli and cortical tension than control cells (Fig. 3A,B; Fig. S3). In contrast, fimbrin mutant cells have lower viscoelastic moduli than wild type cells when measured by LTM, which measures fast time-scales (ST: p< 0.02 at 10 and 100 rad/s; Fig. 3C; Fig S3), but have similar cortical tension, which is measured on longer time-scales, to wild type cells (ST: p= 0.5; Fig. 3D). However, removal of fimbrin from myoII null cells did lead to a significant reduction in cortical tension, suggesting that myosin-II might modulate fimbrin's contribution to long time-scale cortical mechanics (MPA) (ST: p=0.002) (Fig. 3D). Similarly, the differing contributions of dynacortin and fimbrin to cellular-scale mechanics may be observed in the furrow-thinning rates: myoII null cells lacking dynacortin thin faster than myoII null cells lacking fimbrin (Fig 3E; Table S2).

Figure 3.

Dynacortin has a greater contribution to cortical mechanics and furrow-thinning kinetics compared to fimbrin. A. Removal of dynacortin and/or myosin-II reduces the viscoelasticity (|G*|) of interphase cells as measured by LTM. B. Likewise, myosin-II and dynacortin contribute to the effective tension (T_eff) as measured by MPA. C. Removal of fimbrin reduces the viscoelasticity (|G*|) of interphase cells as measured by LTM. D. In wild type cells, fimbrin does not contribute to cortical tension as measured by MPA, but it does contribute significantly in a myoII null background. Both the fimbrin knockout strain (fimbrin) and fimbrin RNAi (fimhp) strains have the same effects. The control for fimbrin is the parental strain and the control cells for the fimhp expressing cells are wild type cells carrying the empty vector. E. In a myoII null background where fimbrin has a significant mechanical contribution on longer time-scales (MPA), reduction of fimbrin or dynacortin increases the rate of furrow ingression compared to control cytokinesis. However, dynacortin has a greater contribution to the furrow-thinning kinetics than fimbrin has. Error bars, standard error of the mean.

The mechanical properties of crosslinked actin networks are derived from the complex organization of the actin polymers and the kinetic properties of the crosslinking proteins [4, 37]. The organizational features lead to the structures and level of filament entanglements that define the mechanics of the network. The crosslinking proteins organize the structures and stabilize the entanglements - slower (longer lived) crosslinkers maintain stable associations between the polymers while faster crosslinkers release quickly, allowing the filaments to slide past one another (see Discussion). At the whole-cell level, discerning quantitatively the subtle differences in network organization is not yet feasible. However, we could begin to discern some of the kinetic features of the crosslinkers in wild type and myoII null cells using fluorescence recovery after photobleaching (FRAP) analysis.

First, we measured the dynamics of the crosslinkers and myosin-II in wild type interphase cells. The fluorescence recovery rate of GFP-fimbrin (τ_rec=0.26 s) was faster than the rate for GFP-dynacortin (τ_rec=0.45 s; MW: p=1×10⁻⁵). Overall, both were slower than soluble GFP in the cortical region of the cell (Table 1; Fig. S6). GFP-fimbrin and GFP-dynacortin also showed similar immobile fractions (Table 1; Fig. S7). In contrast, cortexillin-I (τ_rec=3.3 s) and myosin-II (τ_rec=8-10 s) had similar immobile fractions but had significantly longer median recovery times than either fimbrin or dynacortin (Table 1; Fig. S6, S7). Thus, a simple paradigm of fast global proteins and slow equatorial proteins is suggested from the interphase protein dynamics.

Table 1.

Median recovery times, τ_rec, and median immobile fractions, F_i, for global class and equatorial class proteins in interphase and during cytokinesis: FRAP analysis.

	τrec, Fi (n)
	Interphase		Cytokinesis
	Wt	myoII	equator	pole
Global class
GFP-dynacortin	0.45 s, 26% (24)	0.29 s, 40% (30)	0.98 s, 31% (21)	0.51 s, 38% (19)
GFP-fimbrin	0.26 s, 33% (30)	0.68 s, 41% (46)	0.58 s, 2.9% (13)	0.31 s, 21% (16)
Equatorial class
GFP-cortexillin-I	3.3 s, 32% (33)	5.4 s, 49% (42)	5.4 s, 57% (14)	4.5 s, 53% (13)
GFP-myosin-II	11 s*	7.6 s, 24% (8)	10 s*
Soluble
GFP (cortex)	0.15 s, 13% (14)

^*Values from [48]. Mean ± SEM are shown on the histograms in Figs. S6 and S7.

Next, we compared the dynamics of these crosslinkers in interphase wild type and myoII null cells. Because fimbrin had a detectable impact on the cortical tension of myoII null cells but not wild type cells while dynacortin had its greatest impact on the viscoelastic moduli and cortical tension of wild type cells, we wondered if these two crosslinkers would have differential dynamics in wild type and myoII null cells. Indeed, fimbrin's τ_rec was faster in wild type cells (0.26 s) than in myoII null cells (0.68 s) (MW: p =1×10⁻⁷; Table 1; Fig. S6), correlating with its impact on myoII null cortical tension. However, dynacortin was slower in wild type cells (0.45 s) than in myoII null cells (0.29 s) (MW: p=0.005). In contrast, cortexillin-I showed only a weakly significant increase in τ_rec (MW: p=0.04) but a larger immobile fraction without myosin-II (MW: p=0.006; Table 1; Fig. S6, S7).

Finally, we compared the dynamics of these proteins during wild type cytokinesis. The τ_rec of both GFP-fimbrin and GFP-dynacortin increased at the equatorial region to 0.58 s (MW: p=0.006) and 0.98 s (MW: p=0.005), respectively; whereas the recovery times at the polar cortices remained at interphase levels (Table 1; Fig. S6). GFP-cortexillin-I had a recovery time that was similar between the equator (τ_rec=5.4 s) and pole (τ_rec=4.5 s) (MW: p=0.7) (Table 1; Fig. S6), while its immobile fraction increased during cytokinesis (MW: p=0.038) (Table 1; Fig. S7). Overall, each of the crosslinkers in the equatorial region has a longer lifetime, while those in the polar region have shorter lifetimes. In sum, for these proteins, a simple paradigm of slow equatorial and fast global crosslinking proteins appears to control cytokinesis shape change.

Discussion

Because cytokinesis is an inherently mechanical process, mechanical studies have played an important role in cytokinesis research for many decades (for example, [29, 32, 38, 39]). From these studies, a diversity of mechanical scenarios for cytokinesis contractility has been observed across a wide range of organisms. However, more fundamentally, the contractility of a cytoskeletal network results from the integrated behavior of actin crosslinkers and myosin-II. Our data demonstrate that myosin-II and global and equatorial actin crosslinking control spatial mechanics in the absence of a clear concentric ring of actin polymers (Fig. 4). In dividing wild type cells, the polar cortex is more deformable than the equatorial cortex, and the globally distributed actin crosslinkers have much shorter recovery times than the equatorial crosslinkers. The correlation between crosslinker lifetimes and cortex deformability suggests that myosin-II and equatorial crosslinkers primarily increase the local cortical tension and elasticity in the furrow region. This increased equatorial cortical tension may generate surface stresses that lead to Laplace-like pressures, which help push cytoplasm out of the furrow region. The Laplace-like pressures may originate from stresses in the actin network that are actively generated by pulling forces from myosin-II and/or from pushing forces from actin assembly at the poles. Because the cells are highly elastic (phase angle ∼15° at 100 rad/s), mechanical stresses may propagate through the crosslinked network. However, the cell cortex and cytoplasm have enough viscous character that as the pole-to-pole length increases and the furrow radius decreases, the surfaces stresses squeeze cytoplasm from the midzone, driving furrow ingression in a manner analogous to how a fluid droplet breaks up. Wild type furrows may also constrict more slowly than myoII null furrows during late cytokinesis stages because of their increased elasticity, which may lead to a longer elastic relaxation time (previously estimated to be ∼30 s for wild type cells [13]). In contrast, the higher level of deformability of dividing myoII null cells may facilitate their ability to divide with the aid of traction. In the myoII nulls, force generation from actin assembly at the poles likely allows the emerging daughter cells to crawl apart enough to form the appropriate geometry for equatorial cortical stresses to promote the Laplace-like pressures in the furrow region [10, 13, 40, 41]. Importantly, myoII null cells do not elongate enough to simply crawl apart, rather cytoplasm flows from the midzone and the furrowing dynamics of the myoII null cells can be modeled as a surface tension-driven process, further supporting a role for the Laplace-like pressures [13].

Figure 4.

Model for cytokinesis cell shape change through the contraction of an actin meshwork: myosin-II and actin crosslinkers interact to control furrow ingression dynamics, equatorial and polar cortical tension, and crosslinker lifetimes. Here, the equatorial cortex is principally controlled by myosin-II and cortexillin. The global/polar cortex is modulated by dynacortin, fimbrin, and myosin-II. The local increased cortical tension (S_cf>S_cp) by myosin-II generates equatorial stresses (orange arrows) that help squeeze cytoplasm from the furrow region whereas the globally distributed crosslinkers generate resistive stresses (green arrows) that slow furrow ingression (this paper and [13]). The equatorial crosslinker cortexillin-I and equatorial populations of fimbrin and dynacortin (represented as red ellipses) persist much longer at the cortex, perhaps contributing to the increased tension in this region. Conversely, polar actin crosslinkers (fimbrin and dynacortin represented as blue ellipses) release from the network on fast time-scales, making the global cortex more deformable. This system of myosin-II and equatorial and global actin crosslinkers generates the stress differential that drives and controls the dynamics of cytokinesis cell shape change.

While many types of cells perform cytokinesis through the constriction of a purse-string of actin and myosin-II filaments, mammalian NRK cells, Swiss 3T3 cells and Dictyostelium cells do not have such a highly organized structure of concentric actin filaments (this paper) [21, 22]. Despite these structural variations, how myosin-II, actin polymers, and actin crosslinkers interact is likely to be a fundamental principle, governing cytokinesis contractility dynamics. Indeed, alpha-actinin functions in mammalian NRK cells analogously to the actin crosslinkers studied here: NRK cells devoid of alpha-actinin have accelerated furrow ingression, similar to dynacortin and fimbrin mutants, and myosin-II inhibition leads to slower alpha-actinin dynamics, similar to fimbrin [42].

To deform the cytoskeletal network, myosin-II motor proteins must pull on actin filaments that are crosslinked by crosslinking proteins, and without crosslinkers, myosin-II would simply slide filaments past one another without deforming the network [43]. Therefore, interactions between the crosslinkers and motors are an essential feature of a contractile system, and the crosslinkers and motors are poised to influence each others' activities through crosstalk across the filaments. This crosstalk may occur in principally three ways: (1) through structural organization of the filament network, (2) through binding interactions (allostery) through the actin filament, and/or (3) through tension effects across the filaments. Structurally, the crosslinker or myosin-II may organize filaments into network structures that affect the binding of the other. By binding, the motor or crosslinker could alter the conformation of the actin filament so that it modulates the binding of the other. Finally, because myosin-II pulls on the actin filament, tension through the filament may stabilize or destabilize the binding of the crosslinker. Whether tension directly influences the crosslinker binding dynamics, slow crosslinkers would not only make the network less deformable (more elastic, increased stiffness) but would provide the mechanical load needed for myosin-II to generate tension on the network.

Cortexillin-I appears to have an important role in assisting myosin-II in tension generation. Cortexillin-I is enriched in the furrow cortex and with its slow recovery dynamics (∼10-fold slower than those of dynacortin and fimbrin), it is likely to be the major crosslinker that myosin-II pulls against to generate increased cortical tension and elasticity in the furrow region. From other work, cortexillin-I and myosin-II relocate to sites of cell deformation in dividing cells as part of a mechanosensory cell shape control system, whereas dynacortin and fimbrin do not [33]. Cortexillin-I mutants also have a similar reduction in cortical stiffness and have a slower initial phase of furrow ingression similar to myoII null cells [10, 13, 25]. All of these observations together identify cortexillin-I and myosin-II as core components of a contractile module that controls cytokinesis cell shape change.

In contrast, fimbrin and dynacortin might antagonize myosin-II. Both proteins provide a braking function, slowing furrow ingression kinetics, and both have fast (sub-second) recovery dynamics in the cortex. In vitro, myosin-II can extract actin filaments from fimbrin crosslinked networks, suggesting that myosin-II pulling might release fimbrin crosslinking [36]. Similarly, in our study, removal of fimbrin from a myoII null, but not from wild type, led to a significant reduction in cortical tension, suggesting that myosin-II may antagonize fimbrin's crosslinking activity. Alternatively, dynacortin has slower recovery dynamics in wild type cells than in myoII null cells and has a bigger impact on the viscoelastic moduli and cortical tension of wild type cells than myoII null cells. Precedence exists for crosslinkers to have slower dynamics in response to myosin-II-generated forces, as has been suggested for zyxin at the focal adhesion [44].

Overall, complex interactions between actin-associated proteins control cytokinesis dynamics and mechanics, and to decipher this complexity requires quantitative analysis of single and double mutant combinations of cytoskeletal and regulatory factors. As judged from latrunculin-treated cells, the cortical cytoskeleton contributes ∼90% of the cortical stiffness [25]; yet myosin-II and each individual crosslinking protein only contributes ∼20-30% to cortical tension (this paper)[10]. However, single versus double mutant combinations of myosin-II or cortexillin-I with global crosslinkers do not necessarily lead to additive reductions in cortical stiffness and tension (this paper) [25]. Therefore, the molecular determinants of cortical mechanics interact in a highly complex fashion, leading to nonlinear effects. While one can only speculate as to how many crosslinkers have to be removed to reduce tension to the latrunculin level, the small GTPase RacE may provide part of the clue. RacE nulls have a 70-80% reduction in cortical tension [45], and RacE is known to be required for the accumulation of crosslinkers dynacortin and coronin, but not fimbrin or enlazin, at the cortex [11]. Thus, cytokinesis shape change is the result of a complex system of interacting regulatory and cytoskeletal proteins that control cell mechanics.

In sum, with our current data sets (this paper) [10, 13, 25], we are building an analytical framework that relates contractility dynamics, cell mechanics, and crosslinker recovery dynamics. This framework provides a number of quantitative outputs that can be assessed to see how cytokinesis regulatory pathways modulate cytokinesis cell shape change. Ultimately, it will be important to develop the computational tools to test this analytical framework quantitatively. Additionally, whole-cell measurements always have the caveat that unknown proteins may impact the system. Therefore, the development of reconstitution systems that allow the interface between mechanical strain, crosslinked actin network structure, and crosslinker dynamics to be directly analyzed and contrasted with these in vivo data will be essential.

Methods

Details of the cell strains, genetic screening, analysis of growth rates, molecular biology techniques, Pt-TEM, 3D-EM tomography, 3D-deconvolution and TIRF fluorescence imaging, and FRAP and FLIP analyses may be found online in the Supplementary Data.

Furrow-Thinning Dynamics

Time-lapse DIC images were taken at 2-s intervals using a 40 × (N.A. 1.3) objective with 1.6 × optivar. Minimal furrow diameters and lengths were measured with 4-s time resolution, and furrow-thinning dynamics were analyzed using a previously described rescaling strategy [13].

Laser-tracking microrheology

Laser-tracking microrheology (LTM) of interphase cells was performed using previously published methods [25]. In short, beads were tracked for 11 1-s iterations. The generalized Stokes-Einstein relationship is used to convert bead fluctuations into cell viscoelasticity spectra:

$$\left|G^{\ast }\right|=\frac{2K_{B}T}{6\pi r\langle \mathit{msd}\rangle }$$ (1)

where r equals the bead radius, which is 0.35 μm. |G*| is a complex modulus that is a combination of elastic (storage) and viscous (loss) moduli so that G*= G′ + iG″. The phase angle (δ) relates these two components so that G′=|G*|cosδ and G″=|G*|sinδ. Lognormal means and standard error of the means of |G*| values are presented after transformation back into real space values.

Micropipette aspiration

Micropipette aspiration (MPA) of logarithmically growing cells was performed using glass pipettes with inner radii of 2.5-5 μm. DIC images were taken every 5 s using a 60 × (NA 1.45) objective with 1.6 × optivar. With MPA, different cell mechanical models such as solid-body deformation and the cortical shell-liquid core are used to convert the pressure-deformation relationships into mechanical parameters [46, 47]. The effective tension of interphase cells was measured at a pressure that induced a hemispherical deformation (L_p/R_p=1) of the cell into the pipette. The tension was calculated using the Law of Laplace [47]:

$$\Delta P=2T_{\mathit{eff}}\left(\frac{1}{R_p}-\frac{1}{R_c}\right)$$ (2)

To determine the relative stiffness of mitotic cells during metaphase and at the equator and poles during anaphase, pressure jumps using MPA were applied to the cells. The measurements were taken within 20 s of aspiration and GFP-myosin-II was monitored to ensure that pressures were recorded prior to the mechanosensory response [33]. From the slope (m) of the ΔP vs. L_p/R_p curves, the elastic modulus E could be estimated using the following equation:

$$E=\frac{3m}{2\pi \varphi }$$ (3)

where ϕ=2.1 [46].

Statistical Analyses

For each comparison, either a Mann-Whitney (MW) or two-tailed Student's t-test (ST) was performed, and each p-value has either MW or ST to designate the test used.

Abbreviations

ΔP

aspiration pressure

S_cf

cortical tension furrow

S_cp

cortical tension polar cortex

D_x

cross-over distance

T_eff

effective cortical tension

FLIP

fluorescence loss in photobleaching

FRAP

fluorescence recovery after photo-bleaching

FTD

furrow-thinning dynamic

F_i

immobile fraction

LTM

laser-tracking microrheology

L_p

length of cell inside pipette

MW

Mann-Whitney test

MPA

micropipette aspiration

Pt-SEM

platinum-shadowed scanning electron microscopy

Pt-TEM

platinum-shadowed transmission electron microscopy

R_c

radius of the cell

R_f

radius of the furrow

R_p

radius of pipette

τ_rec

recovery time

ST

Student's t-test

TIRF

total internal reflection fluorescence microscopy

3D-decon

3D-deconvolution

3D-EM

3D-electron tomography