An actin–myosin-II interaction is involved in maintaining the contractile ring in fission yeast

After the completion of mitosis, one cell mechanically divides into two daughters, and this process is called cytokinesis. This coordinated cytofission ensures proper segregation of the genome and organelles, and its failure can induce changes in ploidy and oncogenesis or cell death (Ganem et al., 2007). Actin filaments (F-actin) and myosin-II both assemble into a circular structure along the cell equator, called the contractile ring, during cytokinesis in many eukaryotic cells. In addition to large amounts of F-actin and myosin-II, the contractile ring comprises various regulatory proteins such as actin-modulating proteins and signaling molecules, indicating the elaborate coordination of molecular systems (Eggert et al., 2006; Goyal et al., 2011). Similar to during the contraction of muscle sarcomeres, bipolar myosin-II filaments are considered to slide oppositely orientated F-actin filaments closer together in order to drive shrinkage of the contractile ring, which, in turn, generates the inwards force required for cortical ingression during cytokinesis (Wang, 2005; Pollard, 2010). This is called the purse-string model and is consistent with previous findings showing that contractile ring F-actin had mixed polarities (Sanger and Sanger, 1980; Maupin and Pollard, 1986; Kamasaki et al., 2007). The number of sarcomere-like contractile units in the unconstricted contractile ring of Caenorhabditis elegans embryos has been suggested to reflect the scalability of the ring constriction rate: larger rings constrict faster than smaller ones (Carvalho et al., 2009).

Although a previous study has suggested that the motor activity of myosin-II is involved in the constriction of the contractile ring or cytokinesis (Straight et al., 2003), the precise configuration of actin and myosin-II in the ring and the mechanism by which the local actin–myosin interaction is harnessed for global shrinkage have yet to be elucidated. The motor activity of myosin-II has also been shown to promote the turnover of cortical actin during cytokinesis (Guha et al., 2005; Murthy and Wadsworth, 2005), which suggests that it functions in actin disassembly. Myosin-II has previously been shown to induce the disintegration of F-actin bundles in vitro and has been implicated in cellular actin network disassembly (Haviv et al., 2008; Wilson et al., 2010). Moreover, motor-impaired mutant myosin-II proteins have been shown to effectively support cytokinesis (Reichl et al., 2008; Ma et al., 2012), suggesting that myosin-II modulates cortical tension by crosslinking F-actin rather than sliding F-actin in the contractile ring during cytokinesis.

Two myosin-II isoforms in Schizosaccharomyces pombe, Myo2 and Myo3 (also called Myp2), both localize to the contractile ring and have been implicated in cytokinesis. Myo2 is essential for cell proliferation, whereas the function of Myo3 is vital for growth under certain conditions. During metaphase, the anillin-related protein Mid1 recruits Myo2 to the medial cortex as a broad band of nodes in an F-actin-independent manner, possibly through an interaction with the C-terminal tail moiety of Myo2 (Motegi et al., 2004). The motor activity of Myo2 has been suggested to drive the capture and pulling of F-actin elongated from the nodes, leading to the coalescence of these nodes and F-actin into a ring (Vavylonis et al., 2008), which indicates the primary role of Myo2 is in ring assembly. Myo3 is a single-headed myosin-II with a large gap in the coiled-coil region (Fig. 1A) and localizes to the ring in an F-actin-dependent manner (Wu et al., 2003). A double mutation of myo3Δ (a null mutation in myo3) and the myo2-E1 (a temperature-sensitive allele of myo2; Balasubramanian et al., 1998) have previously been shown to cause severe defects in cytokinesis (Motegi et al., 2000). The motor heads of Myo2 and Myo3 are assumed to have similar properties because they have been found to be functionally exchangeable (Bezanilla and Pollard, 2000). Thus, although the two myosin-II proteins have been suggested to play an overlapping role, possibly through their head moieties, the underlying mechanism currently remains unclear. As Myo3 is single-headed, it is currently unclear whether and how it might be involved in contractile ring constriction, whereas Myo2 might assemble into bipolar filaments and contribute to ring constriction through the purse-string mechanism. In addition, previous studies have indicated that Myo3 is necessary for cell division in high-Cl⁻ environments. The molecular bases of the Cl⁻ tolerance of Myo3 and its relationship to the shared function with Myo2 also have not yet been elucidated.

In the present study, we focused on Myo3 in order to examine the role of myosin-II following contractile ring assembly. Myo3 localized to the contractile ring following the assembly of the actin ring, in a manner that was solely dependent on F-actin. A site-directed mutational analysis revealed that the motor properties of Myo3 were required for its localization to the medial region. Under conditions in which the local accumulation of Myo3 and motor activity of Myo2 were both reduced, the contractile ring that assembled was not maintained properly in a proportion of cells, resulting in inefficient septation, which suggests that myosin-II proteins have a collaborative role in maintaining the ring. Moreover, Myo3 appeared to form motile clusters on the ring. We propose that large assemblies, bipolar filaments or clusters of myosin-II proteins consolidate the contractile ring by continuously binding to F-actin in the ring, thereby cooperatively contributing to its maintenance during cytokinesis.

We recently reported that several components of the contractile ring of the fission yeast S. pombe, including Myo2, can localize to this structure through both F-actin-dependent and -independent mechanisms, even after the completion of ring assembly, such that when F-actin is lost a fraction of the component remains at the cortical ring (Takaine et al., 2014a). Consistent with our previous findings, we found that mYFP–Myo2 colocalized with F-actin in the medial region as a node structure at early anaphase and as a ring at late anaphase (Fig. 1B). The localization of Lifeact was homogenous in cells treated with latrunculin-A (Lat-A), indicating that this drug causes the disassembly of cellular F-actin structures into undetectable actin subunits. Upon Lat-A treatment, Myo2 localized as medial cortical dots that could be distinguished from precursor nodes (Fig. 1B, arrowheads). We then attempted to determine whether another myosin-II, Myo3, in the contractile ring could also localize in both an F-actin-dependent and -independent manner. The Lat-A-induced dispersal of F-actin and localization of Myo3 were examined by time-lapse imaging. Myo3 and actin colocalized in the contractile ring at late anaphase. The exposure of cells to a low concentration of DMSO did not affect maintenance or constriction of the ring (Fig. 1C; supplementary material Movie 1). The addition of Lat-A disintegrated the actin–Myo3 ring and other F-actin structures within several minutes, suggesting that Myo3 localization to the ring was completely dependent on F-actin (Fig. 1D; supplementary material Movie 2). The decrease in fluorescence intensity in these rings favorably fitted a single exponential function. The half-decline time of Myo3 was slightly smaller than that of actin (1.6 versus 3.5 min; P=0.001). The disassembly of F-actin in the constricting ring similarly caused dispersion of the Myo3 signal (Fig. 1E). These results suggest that contractile ring F-actin is the primary and essential scaffold for Myo3 throughout cytokinesis.

We simultaneously examined the accumulation kinetics of Myo3 and actin to the equatorial ring. In early anaphase, actin gradually accumulated at the cell middle, forming a thin ring (Fig. 2A; supplementary material Movie 3). As the actin ring thickened at late anaphase, the Myo3 signal began to increase in the medial region (Fig. 2B,C). The actin and Myo3 rings remained together in place for several minutes (known as the ‘dwell time’ of the ring), and then constricted in a synchronized manner. Myo3 localized only very slightly to the F-actin net that is the precursor of the actin ring (Fig. 2D). Deconvolution microscopy revealed that the actin rings were almost continuous, suggesting that the concentration of actin in the contractile ring was constant, whereas the Myo3 rings were more inhomogeneous and discontinuous (Fig. 2E). Hence, Myo3 might preferentially localize to a limited number of sites on the actin ring. We further examined the formation of Myo3 and actin rings in mid1Δ cells, in which the contractile ring is assembled from strands, not from medial nodes (Saha and Pollard, 2012). In these cells, actin appeared as strands and assembled into a ring, and Myo3 formed a ring after the actin ring assembled (Fig. 2F,G). Thus, Myo3 might specifically interact with F-actin, which assembles into a ring that was nearly fully formed independently of Mid1.

The F-actin-dependent localization of Myo3 to the contractile ring implies that Myo3 interacts with F-actin in the ring through its motor domain. In order to investigate the relationship between the motor properties of Myo3 and its localization, we introduced several well-characterized point mutations into the highly conserved regions in the head moiety (Fig. 3A,B; supplementary material Fig. S1), which took advantage of myo3 not being required for viability, in contrast to myo2. An immunoblot analysis demonstrated that these mutants were produced at the same level as wild-type (wt) (Fig. 3C). The E480K mutant corresponds to one of the mutant myosins that shows no detectable ATPase activity and actin filament sliding (Ruppel and Spudich, 1996). Myo3-E480K–3mYFP, which was expected to have no ATPase activity, showed no ring localization at anaphase (Fig. 3D). The R240A mutant Myo3 is predicted to have low-actin-activated ATPase activity and to not be able to drive F-actin sliding given that the original mutant myosin-II had a V_max value one-fifth that of wt, a K_m value 2.5-fold higher than that of wt, and also showed no motility (Shimada et al., 1997). Although Myo3-R240A localized to the contractile ring, its local concentration was 15% that of wt (Fig. 3D,E). The S469V mutant corresponds to a mutant myosin that retains high-actin-activated ATPase activity, but translocates F-actin at only one-tenth the speed of wt (indicating the uncoupling of ATPase activity and motility) (Ruppel and Spudich, 1996; Endow, 2000). The local concentration of Myo3-S469V in the contractile ring was approximately half that of wt (Fig. 3D,E). The R702C and R709C mutants of human non-muscle myosin IIA and myosin IIB show ∼20% of the maximum actin-activated ATPase activity of wt and have a higher affinity for actin in the presence of ATP (its K_m value is less than one-third that of wt) owing to the very slow release of ADP (Kim et al., 2005). Moreover, the motilities of the mutant myosins are severely impaired (Hu et al., 2002; Kim et al., 2005). The local concentration of the corresponding Myo3-R694C mutant in the ring was 1.8-fold higher than that of wt (Fig. 3D,E). The G680V mutant myosin-II of Dictyostelium discoideum strongly binds to actin even in the presence of ATP by possibly forming a stable actin–myosin-II complex carrying ADP and phosphate (Patterson et al., 1997; Uyeda et al., 2002). The bulk of the corresponding Myo3-G688V mutant formed an aggregate in the medial region, whereas the rest localized faintly to the contractile ring (Fig. 3D). The aggregates were distinct from spindle pole bodies (SPBs) and often persisted during interphase (Fig. 3F). These results suggest that (1) an undisturbed ATP hydrolysis cycle is required for the proper distribution of Myo3 over the ring, and (2) that the Myo3 affinity for F-actin in the presence of ATP is relevant to its local concentration. The motility of Myo3 itself does not appear to be essential for its localization.

We also examined the localization of the C-terminal tail (amino acids 827–2104) of Myo3. An mYFP-tagged Myo3tail (Myo3tail–3mYFP) expressed under the control of the nmt41 promoter diffused throughout the cytoplasm, whereas full-length Myo3 expressed under the same promoter localized to the contractile ring (Fig. 3G). This result again suggests that Myo3 localizes to the cytokinetic actin ring in a manner that depends on its head moiety.

Myo3-null cells show a morphological phenotype in the presence of high concentrations of the Cl⁻ ions, possibly because of the failure of cytokinesis (Bezanilla et al., 1997; Motegi et al., 1997). Therefore, we investigated the Cl⁻ ion sensitivity of mutant myo3 strains. Cells from E480K, R240A, G688V and myo3tail mutants were often multi-septated and swollen in 1 M KCl, similar to myo3Δ cells, whereas cells from myo3⁺, S469V and R694C strains showed no morphological defects (Fig. 3H). Therefore, the function of myo3⁺ to maintain cell morphology under high Cl⁻ ion conditions might require its local concentration to the contractile ring to be higher than a minimum level, but does not appear to correlate with its motor activity.

We next examined the S469V and R694C mutants because they did not exhibit the KCl phenotype, but had a lower or higher local concentration to the contractile ring, respectively, than wt. We observed the accumulation of wt or mutant Myo3 together with Rlc1, a light chain of myosin-II, during cytokinesis. Rlc1 colocalized with Myo2 and Myo3, and could be used to visualize the course of contractile ring assembly from a band of cortical nodes in the medial region. The wt Myo3 formed a medial ring ∼15 min after the Rlc1 ring had assembled (Fig. 4A,D; supplementary material Movie 4). Myo3-R694C accumulated on the contractile ring earlier than wt and decelerated constriction of the ring (Fig. 4B,D,E; supplementary material Movie 4). Myo3-S469V accumulated on the ring with kinetics similar to those of wt; however, its local amount was markedly lower, and did not interfere with ring constriction (Fig. 4C–E). Neither of the mutant Myo3 proteins retarded the cell growth (Fig. 4F) or assembly of the contractile ring, as monitored by Rlc1 localization (Fig. 4D) in myo2⁺ strains. These results suggest that binding sites for Myo3 on the ring are available after the completed assembly of the ring, and diffused Myo3 molecules directly bind to the sites rather than migrating towards the ring through ring-associated actin tracks (Arai and Mabuchi, 2002; Huang et al., 2012).

The deceleration of ring constriction by the Myo3-R694C mutant implied the involvement of Myo3 in ring constriction during cytokinesis under normal conditions, as reported recently (Arasada and Pollard, 2014), and not only under severe conditions (Bezanilla et al., 1997; Motegi et al., 1997). Therefore, we re-examined the phenotypes of myo3Δ under normal growth conditions. The percentage of singly septated cells was significantly higher in an asynchronous culture of myo3Δ than in wt, suggesting prolonged cell division (supplementary material Fig. S2A,B). We also observed that the rate of constriction of the contractile ring in myo3Δ cells was significantly lower (80%) than that in wt cells, whereas the time required for ring assembly and the dwell time of the contractile ring were similar in both strains (supplementary material Fig. S2C–E). These results imply that Myo3 is involved in the constriction of the contractile ring and, hence, show that it is required for efficient septation under normal conditions.

In order to understand the function of Myo3 in cytokinesis in more detail, we examined motor-impaired myo3 mutants in the myo2-E1 background, a temperature-sensitive mutant strain of myo2 that shows severe defects in contractile ring assembly at a restrictive temperature (Balasubramanian et al., 1998) and slower assembly and constriction of the ring at a permissive temperature (Stark et al., 2010). Biochemically, the mutant Myo2 has significantly diminished ATPase activities and is unable to drive the gliding of F-actin, but binds to filaments under restricted diffusion conditions (Stark et al., 2013). We generated myo2/myo3 double mutants and observed contractile ring assembly at 25°C. In myo2-E1 myo3⁺ cells, the assembly and constriction of the contractile ring were both slower than those in myo2⁺ cells (compare Fig. 4A,D,E and Fig. 5A,E,F). Prolonged ring assembly also delayed the localization of Myo3 to the contractile ring. In myo2-E1 myo3-R694C cells, constriction of the contractile ring was retarded more than that in myo2-E1 cells, indicating that Myo2-E1 and Myo3-R694C additively affect ring constriction (Fig. 5B,E,F). Myo3-R694C localized to the contractile ring earlier than Myo3, as observed in myo2⁺cells. Although Myo3-S469V–3mYFP localized to the contractile ring in the myo2-E1 background, its signal was very susceptible to photobleaching (supplementary material Fig. S3A), and, thus, we were unable to precisely determine when the Myo3-S469V ring was formed. The contractile ring constricted slower in 44% of myo2-E1 myo3-S469V cells than in myo2-E1 cells (Fig. 5C–F; supplementary material Movie 5). In half of the cells, the ring signal of the Rlc1 gradually degraded after late anaphase, and septation was retarded (Fig. 5C,D,G; supplementary material Movie 6). This was unexpected because more than 95% of cells in the myo2-E1 and myo2-E1 myo3-R694C strains showed constriction of the ring, and no ring degradation was observed. Furthermore, the rings that assembled were deformed in 45% of myo3-S469V cells (arrowheads in Fig. 5C,H,I). Correspondingly, the growth of myo2-E1 myo3-S469V cells was slightly slower (Fig. 4F). The contractile ring assembled with similar kinetics in these three strains. These results suggest that the motor activities of Myo2 and Myo3 are jointly involved in maintaining the contractile ring after anaphase.

We also examined the behavior of the actin ring in myo2-E1 myo3-S469V cells (supplementary material Fig. S3). Most myo2-E1 cells septated after the medial actin ring constricted. The actin ring constricted more slowly in 50% of myo2-E1 myo3-S469V cells than in myo2-E1 cells, and delayed septation occurred. In the remaining cells, the actin ring assembled, but then decondensed into fragments, and septation was also retarded. Collectively, we conclude that actin and myosin-II structures in the contractile ring might both be decomposed in half of the myo2-E1 myo3-S469V cells. Taken together, these results suggest that catalytically active myosin-II motors contribute to the maintenance of actin and myosin structures in the contractile ring.

Cdc8 is the single tropomyosin of fission yeast and is essential for the cytokinesis and organization of the actin cytoskeleton (Balasubramanian et al., 1992; Chang et al., 1996), and has been shown to localize to F-actin cables and the contractile ring (Arai et al., 1998). Biochemically, Cdc8 stabilizes F-actin, enhances the association between Myo2 and F-actin, and promotes the actin-activated ATPase activities of Myo2 (Stark et al., 2010). Therefore, we generated and examined a double mutant bearing both myo3Δ and cdc8-110, a temperature-sensitive mutant of cdc8 (Nurse et al., 1976). Under all permissive temperatures tested, myo3Δ cdc8-110 cells grew more slowly than wt and single mutants (supplementary material Fig. S4A). Moreover, the number of multinucleated cells was significantly higher in the double mutant than in the single mutants (supplementary material Fig. S4B). These results indicate that myo3 and cdc8 are synergistically involved in cytokinesis and cell growth and are consistent with previous findings (Motegi et al., 1997). We also examined the assembly and constriction of the contractile ring in the double mutant under permissive conditions. The time required for assembly of the contractile ring was longer in cdc8-110 and myo3Δ cdc8-110 cells, by ∼7 min, than in wt or myo3Δ cells (Fig. 6A,B,D), suggesting that there is no cooperativity between myo3 and cdc8 for ring assembly. However, there appeared to be significant cooperativity for ring constriction and ring maintenance. The ring constricted very slowly in 32% of the double mutant cells, whereas in most of the wt, myo3Δ and cdc8-110 cells the ring constricted much faster (Fig. 6C,E,F). In 44% of myo3Δ cdc8-110 cells, the ring persisted for an abnormally long time (at least 70 min, the maximum lifetime of the contractile ring in wt) without constriction (Fig. 6C,E,F). This long-lasting ring was also observed in 4% of cdc8-110 cells. The ring disappeared after an unusually prolonged duration without constriction in 20% of the double mutant cells (Fig. 6E,F; supplementary material Movie 7). In addition, 35% of myo3Δ cdc8-110 cells displayed a deformed contractile ring after its assembly (Fig. 6G,H), which was not observed in other strains or resembled myo2-E1 myo3-S469V cells. These results suggest that myo3 cooperates with cdc8 to properly maintain the contractile ring after late anaphase, and are partially supportive of the aforementioned hypothesis that myosin motors contribute to maintaining the contractile ring.

We attempted to elucidate the mechanism responsible for the Myo3-mediated maintenance of the contractile ring. Fluorescence recovery after the photobleaching (FRAP) analysis revealed that Myo3 at the contractile ring was largely immobile and turned over relatively slowly (Fig. 7A,B), which indicates the stable localization of Myo3 and is consistent with its function as a ring stabilizer. Myo3-R694C had a smaller mobile fraction and turned over more slowly than wt, again suggesting that ATP hydrolysis correlated with the dissociation of Myo3 from the ring.

In order to gain further insights into the behavior of Myo3, we investigated the discontinuous signals of Myo3–3mYFP at the contractile ring (as described in Fig. 2) using time-lapse imaging. The signals of Myo3–3mYFP gradually accumulated as dozens of fine dots along the actin ring and formed a discontinuous ring (Fig. 7C). Hereafter, we refer to these dots as Myo3 clusters. These clusters were constantly motile and their fluorescence intensities varied (Fig. 7D; supplementary material Movie 8). A kymographic analysis revealed that the Myo3 clusters continually emerged and disappeared, with an average lifetime of 29 s, and were capable of moving in both directions (Fig. 7F). Myo3-R694C–3mYFP formed slightly more clusters than wt, and they had significantly longer lifetimes (Fig. 7E,F; supplementary material Movie 8), which is consistent with the higher local concentrations and low mobility.

We constructed several deletion forms of Myo3–3mYFP (Fig. 7G) in order to identify the regions involved in the clustering or localization of Myo3. The C-terminal coiled-coil region spanning 500 amino acids was dispensable for localization to the contractile ring and cluster formation (Fig. 7H). Deletions including the central region of the tail moiety (corresponding to amino acids 1248–1615) significantly decreased the medial accumulation of Myo3: Myo3-1048 localized as dim puncta along the equator, and Myo3-1247 and Myo3-827 sometimes accumulated as several dots in the medial region. We also found that the moderate expression of Myo3-1248-1615–3mYFP did not localize to the contractile ring (Fig. 7I). When overexpressed, Myo3-1248-1615 localized weakly to the equator in late anaphase cells. Most of the deletion forms of Myo3–3mYFP localized similarly in myo2-E1 strains (supplementary material Fig. S4D), which means it is unlikely that they are recruited through interactions with Myo2. Taken together, these results suggest that Myo3 molecules form multimeric complexes on the contractile ring and that the medial localization of Myo3 depends on the motor domain and central region of the tail.

We then investigated the biochemical properties of Myo3-1248-1615 using a sedimentation assay. GST–Myo3-1248-1615 alone partially sedimented, whereas GST did not (Fig. 8A). F-actin significantly increased the amount of sedimentable GST–Myo3-1248-1615 (Fig. 8B). These results indicate that Myo3-1248-1615 is more likely to form large aggregates or interacts with F-actin weakly with an apparent K_d of 7±2 µM.

The contribution of the motor domain of myosin-II proteins to their localization to the cleavage furrow or the contractile ring had previously been unclear because the accumulation of myosin-II to the cleavage furrow or contractile ring was typically concurrent with its filamentogenesis and depended heavily on the tail moiety, but not on the motor domain (Dean et al., 2005). Some scaffold proteins might tether the rod of myosin-II filaments to the equatorial cortex in these cells. In the present study, we showed that Myo3 localized to the contractile ring by solely depending on F-actin throughout cytokinesis and that its motor properties were relevant to its accumulation. The local concentrations of mutant Myo3 proteins (R240A, E480K and R694C) correlated with the affinity of the head for F-actin in the presence of ATP, as represented by the K_m value. The reduced local concentration of Myo3-S469V to the contractile ring might also have been due to its lower affinity for F-actin because a similar uncoupling mutant S456L appears to have a higher K_m value for actin than that of wt (Murphy et al., 2001). In addition, the clumping of Myo3-G688V suggests that an uninterrupted ATP hydrolysis cycle is required for uniform distribution along the ring. In contrast, the motility of the mutant Myo3 did not appear to directly correlate with its localization. A cycle of ATP hydrolysis by myosin-II is coupled to a cycle of attachment to and detachment from F-actin. Thus, disassociation by ATP hydrolysis from F-actin rather than motility might be essential for spreading Myo3 over the contractile ring. All of the point mutations introduced into Myo3 were located outside the actin-binding face of the motor domain and, hence, were less likely to directly alter its affinity for F-actin.

The motor domain of Myo2 localizes solely to the contractile ring (Lord et al., 2005), and Myo3-1048 also accumulated along the ring, suggesting the specific association of myosin-II motor domains with contractile ring F-actin. One possible mechanism for this is the preferential binding of myosin-II motor domains to mechanically stretched F-actin (Uyeda et al., 2011). Myo3 was hardly detectable on precursor medial reticular F-actin, whereas it accumulated along the sharp actin ring, which implies that the configuration or tension of the equatorial F-actin affects its affinity for Myo3. The fact there is an inefficient equatorial accumulation of Myo3-827 in the myo2-E1 strain (in which tension could be weakened) might also support this idea. Alternatively, actin-binding proteins associating with contractile ring F-actin might increase the affinity of myosin-II for actin, as previously reported for Cdc8 (Stark et al., 2010).

The motor activity of myosin-II might drive the shrinkage of interdigitated F-actin in the contractile ring (the purse-string model). Our results suggest that myosin-II motor activity was also vital for maintaining the contractile ring, providing a new perspective regarding the functions of myosin-II in cytokinesis (Fig. 8C). Maintenance of the contractile ring evidently underpins its constriction because destabilization, or even disassembly, of the ring either before or during constriction is unfavorable. After being joined to the ring at late anaphase, Myo3 might collaborate with Myo2 in its maintenance and constriction. Based on the functional equivalency of the motor heads between Myo3 and Myo2 (Bezanilla and Pollard, 2000), the two myosin-II proteins might work in parallel using a similar mechanism. Given that half of the myo2-E1 myo3-S469V cells displayed constriction of the contractile ring, whereas the other half did not, a minimum requirement (i.e. a watershed, proportional to their local concentrations and specific ATPase activities) appears to exist for ring-localized myosin-II motor activity that is necessary in order to maintain the ring. The amount of myosin-II activity might be very close to the minimum in these cells. Given that myo3-S469V cells did not exhibit a morphological phenotype in 1 M KCl, maintenance of the contractile ring might be the primary role of Myo3, which could be separate from its role in tolerance of high concentrations of the Cl⁻ ion. Assuming that a signaling cascade is activated under high-Cl⁻ ion conditions, it is possible that Myo3 provides a scaffold for signal transduction in which signaling molecules interact, as has already been speculated (Bezanilla and Pollard, 2000).

We also found that maintenance of the contractile ring was compromised in cdc8-110 myo3Δ cells, in which the activity of myosin-II proteins was presumably reduced. Cdc8 might indirectly enhance the actin–myosin interaction and ATPase activity of Myo2 (and possibly Myo3) in order to properly maintain the contractile ring after late anaphase. Double mutant cells also displayed unusual persistence of the contractile ring. The small fraction of cdc8-110 cells with a long-lasting ring indicates that the initiation of ring constriction is mechanically unfavorable, which might be exaggerated in combination with myo3Δ. We speculate that Cdc8 also stabilizes the actin ring through direct binding in order to maintain the ring to some extent and promote transition from the dwell phase to constriction.

A previous study has demonstrated that the recombinant Myo3 tail is monomeric because it formed an intramolecular coiled-coil and, hence, Myo3 is a single-headed myosin-II (Bezanilla and Pollard, 2000). However, contractile-ring-localized Myo3 might be in large assemblies because most of the Myo3 signal appeared as dozens of motile puncta. We also found that the motor domain of Myo3 and central region of its tail were both weakly localized to the medial cortex at late anaphase, and both were required for sufficient accumulation. Myo3 appeared to interact superficially with the actin ring rather than being integrated into the ring considering it is concentrated at the ring after F-actin. These results prompted us to speculate that Myo3 molecules on the actin ring formed clusters with their tails, which associate with a kernel (Fig. 8D). As the local amount of Myo3 has been estimated to be ∼2000 molecules (Wu and Pollard, 2005), each cluster might comprise several tens of Myo3 molecules. The Myo3 cluster might be able to sustain continuous attachment to the contractile ring F-actin and crosslink the filaments, either circumferentially or transversely, as it has multiple motor domains. The multivalent attachment of Myo3 clusters might contribute to consolidation of the actin ring. A FRAP analysis revealed the slow turnover rate and low mobility of ring-localized Myo3, which might be explained by its preferential incorporation into the clusters. Meanwhile, single-headed Myo3 would not be expected to interact with F-actin as persistently because it is presumed to have a very low duty ratio. Thus, clustering of Myo3 might be a suitable machinery to permit a single-headed myosin-II to maintain the contractile ring by associating stably with the actin ring.

The molecular entity of cluster kernels remains unknown. The transmembrane protein Chs2 localizes to the medial region during septation and co-immunoprecipitates with Myo3 (Martín-García and Valdivieso, 2006). However, given that Myo3 localizes to the contractile ring in chs2Δ cells, Chs2 might not provide a strong anchor for Myo3. Amino acids 1248–1615 of Myo3 might be directly involved in clustering on the actin ring because this moiety is liable to form aggregates, and F-actin might facilitate aggregation. Given that these properties and cellular localization appear to differ from those of the entire tail (Bezanilla and Pollard, 2000; Fig. 3G), this region might be structurally masked by other regions. Posttranslational modifications (e.g. phosphorylation) could relieve this masking.

Myo3-R694C maintained the contractile ring as effectively as Myo3 in the myo2-E1 background, suggesting that the association between myosin-II proteins and F-actin might be more important for maintenance of the ring than their ATPase activity or motility. This is reminiscent of the situation in COS-7 cells (Ma et al., 2012) in which the mutant non-muscle myosin-II, human non-muscle myosin IIB R709C (equivalent to the R694C alternation), was able to support cytokinesis; however, it had significantly reduced ATPase activity and was unable to translocate F-actin. The proposal that bipolar myosin-II filaments are not required for their motor activity, to glide F-actin, but as an active actin crosslinker to generate cortical tension, which drives cytokinesis, appears to be consistent with our results. Moreover, our results are compatible with studies using Dictyostelium and protoplasts of fission yeast showing that the contractile ring maintains local cortical tension, but does not limit the speed of constriction (Reichl et al., 2008; Stachowiak et al., 2014). The greater importance of the local concentration of Myo3 over its motor activity for ring maintenance also supports this idea. One of the common roles of myosin-II in eukaryotic cell cytokinesis might be the relocatable crosslinking of F-actin, which would enable tension to be exerted on cortical F-actin, including the contractile ring, to maintain the ring.

By contrast, Myo3-R694C dominantly reduced the rate of contractile ring constriction, suggesting that Myo3 itself is involved in ring constriction. Excessive consolidation of the ring through Myo3 clusters and Myo2 filaments might decelerate ring constriction. Conversely, the knockdown of α-actinin, a longer actin crosslinker that localizes to the cleavage furrow, accelerates cytokinesis in NRK cells (Mukhina et al., 2007). Therefore, local concentrations of different types of actin crosslinkers might coordinately determine the optimal constriction rate of the contractile ring.

The fission yeast strains used in this study are listed in supplementary material Table S1. These strains were constructed by a PCR-based-gene-targeting method, a PCR sewing technique (Bähler et al., 1998; Matsuyama et al., 2000) and genetic crosses. The plasmids used in this study are listed in supplementary material Table S2. Cells were grown in complete yeast extract medium YES or in minimal medium EMM with auxotrophic supplements. Cells were grown to mid-log phase at 25°C before imaging unless otherwise noted.

In order to precisely determine the localization of Myo3, we integrated three tandem copies of the monomeric yellow fluorescent protein (mYFP, a monomeric variant of YFP with an A206K mutation) at the C-terminus of the myo3 chromosomal locus. A 3mYFP-hphMX6 module flanked by homologous sequences (400–500 bp) was amplified using MTP19 (Takaine et al., 2014a) as a template for this purpose. The DNA fragment was transformed into the wt strain MTY10, and transformants were selected for hygromycin resistance. The Myo3–3mYFP fusion was fully functional. In order to generate tail-truncated mutant strains of myo3, MTY1001, MTY1002, MTY1003, MTY1008 and MTY1004, 3mYFP-hphMX6 modules were amplified using appropriate primers, as described above, and inserted after amino acids 1990, 1615, 1247, 1048 and 827 of endogenous myo3, respectively.

Point mutations were created by oligonucleotide-directed mutagenesis using a TAKARA mutagenesis kit (TaKaRa) and appropriate primers on the plasmid MTP467 carrying the myo3⁺ gene (Motegi et al., 1997), to generate mutant myo3-3mYFP strains. All plasmids were sequenced to confirm PCR fidelity. In a myo3-3mYFP strain (MTY465), a region of the myo3 gene (corresponding to 400–2400 bp) was replaced with a ura4⁺ module, yielding strain MTY672. A myo3 fragment, corresponding to 0–4000 bp, carrying a point mutation was linearized by a PstI/ScaI cut of the mutagenized plasmid (MTP461, 465, 466, 486 or 487) and subjected to the transformation of MTY672. Colonies were selected for uracil auxotrophy on a plate containing 5-fluoroorotic acid. Correct integrations were confirmed by diagnostic PCR and sequencing.

Single time-point and epifluorescence microscopy was performed using an inverted fluorescence microscope (BX51; Olympus) equipped with a Plan-Apo 100×1.40 NA objective lens (Olympus) using U-MNIBA3 and U-MWIG3 filters (Olympus) and a cooled charge-coupled device (CCD) camera (ORCAII-ER-1394; Hamamatsu Photonics) using SimplePCI software (Compix, Inc.) at room temperature. Time-lapse imaging was performed under a confocal microscope (LSM 700; Carl Zeiss, Inc.) equipped with an alpha Plan-Apochromat 100×1.46 NA objective lens (Carl Zeiss, Inc.) or using an inverted fluorescence microscope (BX71; Olympus) equipped with a UPlanSApo 100×1.40 NA objective lens (Olympus), spinning-disk confocal scanner (CSU22; Yokogawa), piezo objective positioner (E-665; Physik Instrumente) and electron multiplying CCD camera (iXon 3 885; Andor) with excitation by 488-nm and 568-nm lasers using GFP and mCherry filters and MetaMorph software (version 7.7.5.0; Molecular Devices) at 25°C. In the latter system, simultaneous two-color imaging was performed using an emission beam splitter (Dual View; Roper Scientific). Cells for time-lapse imaging were typically immobilized on pads made from 20% gelatin and 5% low-melting agarose in medium, or were directly mounted on a slide for single time-point observations. Stacks of 12–17 confocal z-sections spaced by 0.3–0.5 µm were typically collected. Confocal z-sections were sometimes deconvolved using Huygens Essential software (version 4.5; Scientific Volume Imaging) based on the classic maximum likelihood estimation method. In order to distinguish ring signal degradation in myo2-E1 myo3-S469V cells from global fluorescence decay due to photobleaching during imaging, we only scored the cells where the Rlc1–3mCherry ring signal disappeared before the mean fluorescence intensity at each frame of the time-lapse image was significantly decreased due to photobleaching. Throughout this study, mitotic progression was monitored by the separation of spindle pole bodies (SPBs) labeled by Sid4–3mYFP, indicating elongation of the spindle. The images shown are maximum projections of an xy image of z-stacks unless otherwise noted.

In time-lapse imaging of motile Myo3 clusters on the contractile ring, the signals of myo3–3mYFP were imaged by collecting nine confocal z-sections spaced by 0.31 µm through only the lower half of the cell with a 15-s interval using the inverted fluorescence microscope equipped with the CSU22 spinning-disk confocal scanner (see above). Each z-section was an average of two images. The images were deconvolved to explicitly visualize each cluster. In order to compare the mean number of clusters on the ring in myo3-3mYFP and myo3-R694C-3mYFP cells, we semi-automatically detected clusters using the find maxima function of ImageJ. The number of clusters detected in each frame was averaged across all (29–36) frames.

We manually selected tracks in kymographs to measure the lifetimes of clusters and only further analyzed clusters that met the following criteria: (1) they were brighter than the background signals, (2) were able to be discriminated from neighboring signals, (3) were almost spherical, and (4) had an almost constant signal intensity through its duration.

Fluorescence recovery after the photobleaching (FRAP) analysis was performed using LSM 700 at 25°C. Stacks of 13 z-sections spaced by 0.5 µm were collected with a 20-s interval. Each z-section was an average of two images scanned using 0.5% laser power at 488 nm. Photobleaching was performed at 80% laser power by 20 iterative scanning in a circular region (20 pixels diameter) of the ring. Each z-stack was projected from an xy image using Image J and the sum intensity projection procedure. The mean fluorescence intensity of the ring region was measured, the background value was subtracted (an average of two cytoplasmic regions), and the result was normalized to the prebleach fluorescence intensity (an average of two time points before photobleaching). Data were further corrected for acquisition bleaching using the function, I_t=I_∞+(1−I_∞)×exp(−k_decay×t), where t is time, I_t is the normalized intensity at time point t, I_∞ is the normalized intensity at an infinite time point, and k_decay indicates a decay constant. The decay function was calculated by fitting the data from the unbleached ring regions using Kaleidagraph software (Synergy Software). The resultant recovery curve was fitted to a single exponential function, I_t=I_∞+(I₀−I_∞)×exp(−k×t), where I₀ is the normalized intensity just after bleaching, and k is the recovery rate constant closely related to the dissociation rate constant. The mobile fraction and half time (τ_1/2) for recovery were calculated according to the equations mobile fraction=(I_∞−I₀)/(1−I₀) and τ_1/2=ln 2/k, respectively.

Myo3–3mYFP ring intensities were quantified as previously described (Takaine et al., 2014a). In brief, we imaged myo3-3mYFP sid4-3mYFP cells in 13 z-sections at 0.38-µm intervals using the spinning disk confocal microscope averaging two images. The stack of z-sections was projected from an xy image using the sum intensity projection method of ImageJ. We then measured the mean fluorescence intensities of two SPBs and the medial rings in late anaphase cells. Regions of interests were a circle of 0.64 µm (8 pixels) in diameter for one SPB and a rectangle of 0.5–1 µm (6–12 pixels)×2.4–4.0 µm (30–50 pixels) for the ring, respectively. After background subtraction, the total ring intensity was normalized to the average intensity of Sid4–3mYFP at SPBs in the cell.

Cells for time lapse imaging were adsorbed onto the surface of a 35-mm glass-bottomed dish coated with lectin (L-2320, Sigma) for 10 min. The immobilized cells were imaged by collecting stacks of 12 z-sections spaced by 0.42 µm at 0.5- or 1-min intervals. During the interval after the second stack was acquired, the culture was supplemented with 0.36% DMSO alone or 0.36% DMSO and 5 µM Lat-A by the manual addition of an equal volume (approximately 50 µl) of the 2× concentrated solution. These cells were subsequently imaged until the F-actin signal disappeared. The mean fluorescence intensity of the medial region was measured, the background value subtracted, and the result was normalized to the pretreatment fluorescence intensity. Using Kaleidagraph software, the normalized decline curve was fitted to a single exponential function.

Total cell lysates were prepared from mid-log phase cells as described previously (Takaine et al., 2014b). Proteins were resolved by SDS-PAGE on 5–20% precast gels (C520L, ATTO) and transferred onto a polyvinylidene difluoride membrane (Immobilon-P; Millipore). YFP and α-tubulin were probed using an anti-GFP antibody (Roche) and the anti-tubulin antibody TAT-1 (Woods et al., 1989), respectively.

Monomeric rabbit skeletal muscle actin, bacterially expressed recombinant GST, and GST–Myo3-1248-1615 were prepared as described previously (Takaine et al., 2009) and clarified by centrifugation at 200,000 g for 20 min. Samples containing 4 μM actin and 0–6.3 μM GST-Myo3-1248-1615 in KMEI (0.1 M KCl, 2 mM MgCl₂, 0.5 mM EGTA, 0.5 mM DTT, 0.2 mM ATP, and 10 mM imidazole-HCl, pH 7.5) were incubated for 30 min at 25°C, centrifuged at 200,000 g for 15 min, and the supernatants and pellets obtained were then subjected to SDS-PAGE. The gels were stained with Coomassie Blue, and bands were quantified by densitometry using Image J software.

Means, standard deviations and P-values were calculated using Excel software (Microsoft). Error bars denote s.d. unless otherwise noted. Significance between two sets of data was tested using an unpaired one-tailed Welch's t-test. These P-values are in bold type if the value was less than 0.05. Significance is also indicated by an asterisk. Box plots show the 75th and 25th percentiles of the data (interquartile range) as the upper and lower edges of the box, the median as the medial line in the box, the 1.5× interquartile range as whiskers, and outliers as circles.

We are grateful to the Yeast Genetic Resource Center (Osaka City University) for providing yeast strains.