Assembly of normal actomyosin rings in the absence of Mid1p and cortical nodes in fission yeast

In many eukaryotes, physical division of the cell depends on the function of an actomyosin ring (Schroeder, 1968; Fujiwara and Pollard, 1976; Mabuchi and Okuno, 1977). Forces generated upon actomyosin interactions coupled with the delivery of new membranes (and cell walls in some cases) lead to the bisection of the mother cell into two daughters (Fujiwara and Pollard, 1976; Balasubramanian et al., 2004; Glotzer, 2004). The actomyosin rings are assembled in a plane perpendicular to the axis of chromosome segregation. How the actomyosin ring components organize into a ring structure that is perpendicular to the plane of chromosome segregation is only poorly understood.

In recent years, Schizosaccharomyces pombe, which divides using an actomyosin ring, has emerged as an attractive model for the study of cytokinesis (Balasubramanian et al., 2004; Wolfe and Gould, 2005). In fission yeast, an actomyosin ring is assembled early in mitosis and constricts after mitotic exit (Marks et al., 1986; Fankhauser et al., 1995; McCollum et al., 1995; Wu et al., 2006). Filamentous actin (F-actin) in the ring has been shown to assemble from a leading cable (Chang et al., 1997; Chang, 1999; Arai and Mabuchi, 2002). Recent studies have led to the proposal that actomyosin ring assembly instead is initiated from multiple cortical nodes, each containing two dimers of the formin Cdc12p, anillin-related protein Mid1p, type II myosins, the FCH-BAR (Fer/Cip4 homology–Bin-amphiphysin-Rvs) domain protein Cdc15p, and the IQGAP (IQ motif–containing GTPase-activating protein)-related protein Rng2p (Wu et al., 2006; Vavylonis et al., 2008). Nodelike structures have also been detected from several other studies (Bahler et al., 1998; Wong et al., 2002; Motegi et al., 2004). Mid1p plays a key role in the cortical anchoring of these nodes because loss of Mid1p function eliminates the assembly of detectable nodes (Wu et al., 2006). Computational modeling experiments consistent with experiments of Daga and Chang (2005) have revealed that forces generated by type II myosins combined with a search and capture mechanism help compact nodes placed within a 8-μm zone into a proper ring (Daga and Chang, 2005; Vavylonis et al., 2008). Although the node-based ring assembly mechanism is very attractive, a prediction that arises from this model is that cells containing fewer or no nodes should be unable to assemble normal actomyosin rings.

In this study, we have characterized the process of actomyosin ring assembly in cells lacking Mid1p and detectable nodes. We find that cells lacking Mid1p and detectable nodes can assemble normal-looking actomyosin rings, albeit with a delayed kinetics. Thus, a cortical node–independent mechanism is sufficient for the assembly of properly organized actomyosin rings.

In S. pombe, Wu et al. (2006) have shown that Mid1p assembles first into nodes and recruits other proteins (including type II myosin and its light chain Rlc1p, formin Cdc12p, FCH-BAR protein Cdc15p, and IQGAP-Rng2p) into nodes (Motegi et al., 2000, 2004; Paoletti and Chang, 2000). Fig. 1 A shows examples of Rlc1p-3GFP nodes in wild-type cells and their absence in mid1Δ cells, which is consistent with experiments of Wu et al. (2006). Furthermore, as shown by Wu et al. (2006), we have found that Cdc15p organizes into a few nodelike structures (15 ± 7 nodes/cell; n = 17; Fig. 1 A) and that the number and intensity of these structures is increased (45 ± 18 nodes/cell; n = 22; Fig. 1 A) upon treatment with latrunculin A (LatA), which prevents actin polymerization (Ayscough et al., 1997). We have also been able to observe colocalization of Cdc15p-GFP nodes with Rlc1p-mCherry nodes in LatA-treated cells (Fig. 1 B). As shown by Wu et al. (2006), we have found that organization of Cdc15p into nodes in untreated and LatA-treated cells depended on Mid1p function. Although we have been able to detect Rlc1p and Cdc15p in cortical nodes, we were unable to detect the formin Cdc12p in cortical nodes in unperturbed cells, which is consistent with studies by Chang (1999) and Yonetani et al. (2008). Nevertheless, as described by Wu et al. (2006), we did detect Cdc12p nodes upon LatA treatment (Fig. 1 A). Again, organization of Cdc12p nodes in LatA-treated cells depended on Mid1p function. Approximately 51% of Cdc12p nodes (n = 270) generated by LatA treatment colocalized with Rlc1p nodes (Fig. 1 C). We have also found that Rlc1p nodes are not detected in mutants defective in the fission yeast Polo kinase Plo1p (unpublished data), which is a known regulator of fission yeast Mid1p (Bahler et al., 1998). Collectively, these studies established that some actomyosin ring components (such as Rlc1p and Cdc15p) are detected in nodes in unperturbed cells, whereas others (such as Cdc12p) are detected in nodes only upon LatA treatment. Furthermore, our experiments as well as other studies (Motegi et al., 2004; Wu et al., 2006) have established that the organization of cortical nodes depends on Mid1p function.

Cells defective for Mid1p or Plo1p have been shown to assemble abnormal actomyosin rings (Chang et al., 1996; Sohrmann et al., 1996; Bahler et al., 1998), which at first glance is consistent with the predictions of recent studies (Wu et al., 2006; Vavylonis et al., 2008). We have found that the abnormal rings in mid1 and plo1 mutant cells can become stabilized as a result of septum assembly before ring compaction. Fig. 2 A shows examples of F-actin distribution in wild type, cells of three different mutant alleles of mid1 (mid1-18, mid1-ΔNES, and mid1Δ), and plo1-1 cells. Unlike in mitotic wild-type cells, F-actin rings in mitotic mid1 and plo1-1 mutants are detected at various angles (Fig. 2 A; Bahler et al., 1998). To eliminate any confusion in scoring phenotypic effects that might arise from the assembly of division septum along improperly organized rings, we inactivated Cps1p, a 1,3-β-glucan synthase involved in septum assembly (Liu et al., 1999), in cells defective in Mid1p or Plo1p function. Surprisingly, >85% of cps1-191 mid1-18, cps1-191 mid1-ΔNES, cps1-191 mid1Δ, and cps1-191 plo1-1 cells assembled normal and orthogonal actomyosin rings (Fig. 2, B and C). Rings in cps1-191 mid1-18, cps1-191 mid1-ΔNES, cps1-191 mid1Δ, and cps1-191 plo1-1 cells were positioned at abnormal sites, which is consistent with a role for Mid1p and Plo1p in actomyosin ring positioning.

It remained possible that inactivation of Cps1p function led to the restoration of medial nodes by unknown mechanisms. To test this possibility, cps1-191 and cps1-191 mid1Δ cells were fixed and stained with antibodies against tubulin and the actomyosin ring protein Cdc4p, a light chain for Myo2p (Fig. 2 D). Cdc4p was readily detected in medial nodes in 30/54 cps1-191 cells with short spindles, whereas the rest contained fully formed actomyosin rings. Such organization of Cdc4p was not observed in cps1-191 mid1Δ cells (36/53 contained no obvious structures, and 17/53 contained disorganized cables), ruling out the possibility that normal ring assembly in cps1-191 mid1Δ cells is caused by the restoration of medial nodes. These observations established that Mid1p and detectable nodes are not required for assembly of orthogonal myosin and F-actin rings. These observations also established that the nonorthogonal rings become fixed at improper angles caused by septum assembly before ring compaction in mid1 and plo1 mutants.

To gain further insight into the mechanism of ring assembly in the absence of Mid1p (and nodes), we imaged ring assembly in cps1-191 mid1-18 expressing Rlc1p-GFP as a ring marker. As previously reported by others and us (Bahler et al., 1998; Huang et al., 2007), the assembly of myosin II bundles initiated at nonmedial sites (Fig. 2 E). Interestingly, these myosin II bundles eventually organized into orthogonal rings (Fig. 2 E). These experiments established that in the absence of Mid1p (and detectable nodes), myosin II bundles assemble initially from random locations and at various angles and that they eventually compact into orthogonal rings.

Although normal actomyosin rings were assembled in cps1-191 mid1Δ cells, the assembly process appeared to be slower. To characterize ring assembly in cps1-191 mid1Δ cells, we imaged dynamics of myosin II ring assembly and mitosis simultaneously in cells expressing Rlc1p-GFP and Pcp1p-GFP (a spindle pole body [SPB] marker; Flory et al., 2002). In cps1-191 mutants (Fig. 3 A), myosin II ring assembly was initiated soon after separation of SPBs at prometaphase. Organization of a myosin II ring was completed in <10 min (9.7 ± 1.4 min; n = 7 cells) from the time of SPB separation. In contrast, although myosin II cable assembly was initiated soon after SPB separation in cps1-191 mid1Δ cells (Fig. 3 B), organization of normal myosin II rings was accomplished only after full separation of the SPBs to the opposite ends of the cell. Starting at the point of SPB separation, organization of normal myosin II rings took at least twice as long (32.3 ± 19.5 min; n = 8 cells) in cps1-191 mid1Δ cells.

To further investigate the timing of myosin II ring organization in cps1-191 mid1Δ, we fixed and stained cps1-191 and cps1-191 mid1Δ cells with antibodies against tubulin and Cdc4p. Consistent with the time-lapse experiments, fully formed myosin II rings were detected in cps1-191 cells with intermediate length and elongated spindles as well as in cells with postanaphase microtubule arrays (Fig. 3, C and D). In contrast, in the cps1-191 mid1Δ cells, fully formed myosin II rings were observed in <20% of cells with intermediate length spindles and in <35% of cells with elongated spindles (Fig. 3, C and D). However, fully formed myosin II rings were detected in nearly 90% of cps1-191 mid1Δ cells (Fig. 3, C and D). These experiments suggested that Mid1p and cortical nodes play a role in organizing actomyosin rings in early mitosis but that other mechanisms operate in later stages of mitosis to this end.

To independently test whether Mid1p was important for ring assembly in early mitotic cells, we arrested mid1-18 cells expressing Rlc1p-GFP at metaphase by overproduction of the mitotic checkpoint protein Mad2p (He et al., 1997). We performed three different experiments. First, we induced the expression of Mad2p, shifted cells to the restrictive temperature for mid1-18, and fixed and stained these cells with antibodies against Cdc4p, tubulin, and DAPI. In this experiment, we found that >85% of metaphase-arrested mid1-18 cells contained nonorthogonal myosin II cables (Fig. 4, A and B). Second, we imaged the process of myosin II ring assembly in metaphase-arrested mid1-18 cells (Fig. 4, C and D). We found that myosin II cables and bundles assembled efficiently in metaphase-arrested mid1-18 cells. However, these myosin II cables failed to become organized into rings even upon incubation for nearly 90 min (Fig. 4 D). Finally, we induced the expression of Mad2p, allowed myosin II ring assembly at the permissive temperature for mid1-18, and subsequently shifted cells to the restrictive temperature (Fig. 4, C and E). We found that shift of metaphase-arrested mid1-18 cells with a preexisting myosin II ring to the restrictive temperature led to unraveling and disassembly of this ring (Fig. 4 E, red arrows). These observations suggested that Mid1p and possibly the nodes were important for organization of actomyosin rings in early mitosis and that Mid1p function might be continuously required for the medial maintenance of the actomyosin ring until septation. The fact that metaphase-arrested unseptated mid1-18 cells failed to organize orthogonal rings ruled out the possibility that prevention of septation in cps1-191 mid1⁻ mutants was solely responsible for actomyosin ring organization. Furthermore, this result suggested that some aspect of regulation occurring after metaphase is responsible for the organization of orthogonal rings in cps1-191 mid1⁻ cells.

To further assess whether Mid1p and the cortical nodes were important for ring assembly in metaphase, we arrested nda3-KM311 and nda3-KM311 mid1Δ cells at prometaphase by a shift to 18°C. nda3-KM311 is a cold-sensitive mutant allele of the sole gene encoding the fission yeast β-tubulin (Hiraoka et al., 1984). nda3-KM311 cells arrest with an actomyosin ring (Chang et al., 1996). Interestingly, we found that nearly 59% of nda3-KM311 mid1Δ cells contained misoriented rings, whereas the rest contained orthogonal rings (Fig. 5, A–C; Rlc1p and Sid4p, an SPB component [Chang and Gould, 2000], are shown in A, whereas F-actin is shown in C). In contrast, rings in the nda3-KM311 cells were always orthogonal (unpublished data). These results furthered the notion that the function of Mid1p and cortical nodes was important for organization of normal actomyosin rings in early mitosis. The slightly higher proportion of orthogonal rings observed in nda3 mid1Δ cells might be caused by the prolonged block of nda3-KM311 cells at the restrictive temperature.

Our experiments have led to the conclusion that in the absence of Mid1p and nodes, a second pathway can allow organization of actomyosin rings. Because rings in mid1 mutants assembled later in mitosis, it was possible that some aspect of mitotic progression promoted ring assembly in mutants such as the cps1-191 mid1Δ cells. Alternatively, because loss of Cps1p function leads to a prolonged maintenance of the SIN, it was possible that the SIN played a role in organization of actomyosin rings in the absence of Mid1p and the cortical nodes. The SIN is a signaling module that is activated upon passage through anaphase and plays a key role in maintenance and organization of actomyosin rings (Krapp et al., 2004; Wolfe and Gould, 2005). Previous experiments have shown that ectopic activation of the SIN led to the formation of actomyosin rings even in interphase cells (Cerutti and Simanis, 1999). Therefore, we tested whether activation of SIN in interphase-arrested mid1-18 and mid1Δ cells would lead to the formation of orthogonal actomyosin rings. We shifted cdc16-116, cdc16-116 mid1-18, and cdc16-116 mid1Δ cells expressing Rlc1p-GFP and Uch2p-GFP (as a nuclear marker) to the restrictive temperature for cdc16-116. Cdc16p is a GTPase-activating protein for Spg1p-GTPase, a key component of the SIN (Furge et al., 1998). cdc16-116 mutants maintain Spg1p in a GTP-bound form, leading to persistent SIN activation. The use of cdc16-116 also allowed us to distinguish whether SIN or some aspect of anaphase was important for ring assembly in mid1 mutants. Interestingly, myosin II rings in interphase-arrested cdc16-116, cdc16-116 mid1-18, and cdc16-116 mid1Δ were all found to be orthogonal in orientation (Fig. 5 D). These experiments established that activation of SIN led to the organization of orthogonal actomyosin rings in the absence of Mid1p and cortical nodes. mid1-6 mutant cells deleted for par1, the gene encoding a regulatory subunit of phosphatase 2A, assemble orthogonal division septa (Le Goff et al., 2001). Although actomyosin rings were not looked at in this study, the fact that Par1p is a negative regulator of the SIN supports the idea that SIN activation helps establish orthogonal actomyosin rings. Additional genetic experiments (Fig. S1, available) provided evidence that organization of actomyosin bundles in mid1 mutants and their reorganization in the cps1-191 background depended on the function of the SIN.

Our experiments with two mutants (mid1 and plo1) lacking nodes have established that the accumulation of actomyosin ring components into nodes is not essential for organization of orthogonal rings, although Mid1p and the nodes are important for the kinetics of ring assembly. How do Mid1p and cortical nodes participate in actomyosin ring assembly? Previous experiments using electron microscopy of myosin SI decoration of F-actin filaments have shown that in early mitosis, the actomyosin ring is composed of two parallel bundles of F-actin organized together with other components of the ring (Kamasaki et al., 2007). It is possible that such an arrangement is achieved by actin filament cross-linking rather than by myosin II–based contractility, which depends on antiparallel actin filaments. Consistent with the idea that myosin II may not be fully active in early mitosis, Lord and Pollard (2004) have shown that the localization of the UCS (UNC-45/CRO1/SHE4) domain protein Rng3p, a fission yeast myosin II activator, occurs only after the completion of anaphase. Finally, actomyosin ring constriction (possibly via activation of myosin II–based contractility) occurs concomitant with the activation of SIN in anaphase B (Sparks et al., 1999). Collectively, previous work has shown that at the point of actomyosin ring assembly, myosin II might not be fully active (Lord and Pollard, 2004). It is possible that Mid1p and cortical nodes play a role in the organization of moderately or noncontractile actomyosin bundles into ring structures. Activation of myosin II (upon SIN activation and Rng3p localization) might be sufficient for the organization of orthogonal actomyosin rings even in the absence of Mid1p and cortical nodes. Further analysis of actomyosin ring assembly in cps1-191 mid1Δ cells should provide additional insights into the general mechanism of cytokinesis.

The fission yeast strains used in this study are listed in Table I. Strain Rlc1-3GFP-KanMX6 was provided by J.Q. Wu (Ohio State University, Columbus, OH), strain mid1-ΔNES was provided by F. Chang (Columbia University College of Physicians, New York, NY), and strain Cdc15p-GFP was provided by K. Gould (Howard Hughes Medical Institute, Nashville, TN). Strain Pcp1p-GFP was provided by S. Oliferenko (Temasek Life Sciences Laboratory, Singapore). To arrest cells in interphase, cells were first treated with 12 mM hydroxyurea (HU; Sigma-Aldrich) for 4 h and treated with the same amount of HU for an additional 2 h. Conventional fission yeast genetic and molecular biology techniques were performed as described previously (Moreno et al., 1991).

Cells were fixed and stained with DAPI, Alexa Fluor 488–conjugated phalloidin, Tat1 (provided by K. Gull, University of Manchester, England, UK), and Cdc4p primary antibodies and α-rabbit and α-mouse IgG conjugated with either Alexa Fluor 594 or 488 (Invitrogen) secondary antibodies as described previously (Moreno et al., 1991). Live cell imaging and fluorescence microscopy were performed as described previously (Tran et al., 2004). For time-lapse microscopy, cells were concentrated from early log-phase culture and spotted on a glass slide containing a yeast extract with supplements agarose pad. Microscopic images were captured using a fluorescence microscope (IX71; Olympus) equipped with a camera (CoolSNAP ES; Photometrics) and a Plan Apo 100× 1.45 NA objective lens (Olympus). For confocal microscopy and time-lapse microscopy, cells were observed under a confocal microscope (LSM 510; Carl Zeiss, Inc.) equipped with a Plan Apo 100 ×1.3 NA objective lens, a 405-nm diode laser (for DAPI excitation), a 488-nm argon laser (for GFP excitation), and a 543-nm HeNe laser (for Alexa Fluor 594 excitation). Confocal images were captured with LSM 510 software (Carl Zeiss, Inc.). The time-lapse analysis was conducted at 36°C. The temperature was controlled by an objective heater (Bioptechs). All of the images were analyzed with Metamorph software (MDS Analytical Technologies). All images in Figs. 1, 5, and 2 (A and B) were captured on a microscope (Olympus), whereas a confocal microscope (LSM 510) was used for the rest of the imaging experiments. All of the images were collected in 3D mode (0.5–0.6-μm step size) and are displayed as a maximum projection.

Fig. S1 shows actomyosin ring/bundle assembly and maintenance in the absence of Mid1p and that the cortical nodes depend on the SIN.

We wish to especially thank F. Chang, K. Gull, K. Gould, S. Oliferenko, and J.Q. Wu for gifts of strains and reagents. Special thanks are due to all members of the Cell Division Laboratory and in particular to T.G. Chew, W. Ge, M. Mishra, X. Tang, and Graham Wright for discussion and/or useful comments on the manuscript.

This work was supported by research funds from the Temasek Life Sciences Laboratory and the Singapore Millenium Foundation.