Three myosins contribute uniquely to the assembly and constriction of the fission yeast cytokinetic contractile ring

Summary

Cytokinesis in fission yeast cells depends on conventional myosin-II (Myo2) to assemble and constrict a contractile ring of actin filaments. Less is known about the functions of an unconventional myosin-II (Myp2) and a myosin-V (Myo51) that are also present in the contractile ring. Myo2 appears in cytokinetic nodes around the equator 10 min before spindle pole body separation (cell cycle time −10 min) independent of actin filaments, followed by Myo51 at time zero and Myp2 at time +20 min, both located between nodes and dependent on actin filaments. We investigated the contributions of these three myosins to cytokinesis using a severely disabled mutation of the essential myosin-II heavy chain gene (myo2-E1) and deletion mutations of the other myosin heavy chain genes. Cells with only Myo2 assemble contractile rings normally. Cells with either Myp2 or Myo51 alone can assemble nodes and actin filaments into contractile rings, but complete assembly later than normal. Both Myp2 and Myo2 contribute to constriction of fully assembled rings at rates 55% of normal in cells relying on Myp2 alone and 25% of normal in cells with Myo2 alone. Myo51 alone cannot constrict rings but increases the constriction rate by Myo2 in Δmyp2 cells or Myp2 in myo2-E1 cells. Three myosins function in a hierarchal, complementary manner to accomplish cytokinesis with Myo2 and Myo51 taking the lead during contractile ring assembly and Myp2 making the greatest contribution to constriction.

Introduction

Myosin-II has been considered to be the main motor for cytokinesis since its discovery in the contractile ring [1] and demonstration that cleavage furrows do not form in either echinoderm embryos microinjected with antibodies to myosin-II [2] or Dictyostelium amoebae with a deletion mutation of the myosin-II gene [3]. However, cytokinesis can proceed in a variety of cells with compromised myosin-II. For example, Dictyostelium can survive without the single myosin-II gene by pulling themselves apart as they move in opposite directions on a surface [3]. Budding yeast can divide without myosin-II [4–6] or with myosin-II lacking the motor domain [7]. The ability to divide without myosin-II motor activity was attributed to the growth of the cell wall or actin filament severing and crosslinking. Mammalian COS-7 cells can complete cytokinesis with myosin-II lacking motor activity [8]. The fission yeast gene for conventional myosin-II, myo2⁺, is essential for viability [9, 10], but cytokinesis is remarkably normal in cells with the temperature sensitive myo2-E1 mutation [11], which has minimal biochemical activity even at 25°C, the permissive temperature [12]. This suggested that either the remaining function of Myo2p-E1 is sufficient to generate the forces for cytokinesis or that one or more other myosins contribute to cytokinesis.

Fission yeast cells are favorable for investigating cytokinesis motors, since a large body of quantitative information is available on its division (for a review see [13]), and the genome encodes just five myosin heavy chains: type I myosin myo1⁺; type II myosins myo2⁺ and myp2⁺; and type V myosins myo51⁺ and myo52⁺ [14]. Myosin molecules consist of heavy chains and light chains, and we refer to each molecule in this work by the name of its heavy chain (e.g. Myo2p is the heavy chain of the Myo2 molecule). On the other hand, we refer to the polypeptide when naming a tagged protein (e.g. mEGFP-Myo2p).

Each of the five myosins has distinctive functions. Myo1 has well-characterized roles in endocytosis and mating but does not participate in cytokinesis other than its role in endocytosis during septum formation [15–17]. Myo2p and Myp2p bind light chains Cdc4p and Rlc1p and both participate in cytokinesis [12, 18–20]. Essential Myo2 is a conventional myosin II, since the heavy chain forms a homodimeric, coiled-coil tail [21]. Myp2p is not essential for viability or cytokinesis under normal growth conditions [19, 21, 22] and is unconventional, since its exceptionally long tail folds upon itself to form an anti-parallel coiled-coil [21]. Myo51 and Myo52 carry cargo along actin filament cables, a crucial contribution to cell polarity [14]. During cell division Myo51 redistributes from actin cables to the equator [23], and Myo52 transports vesicles containing beta-glucan synthetase Bgs1 along actin cables to the forming septum (but does not concentrate in the contractile ring) [24]. Both type V myosins require actin for their localization [23].

Our quantitative analysis of the behavior of live fission yeast cells with combinations of myosin mutations revealed that Myo2, Myp2 and Myo51 each contribute uniquely to contractile ring assembly and constriction. Myo2 is the primary myosin for ring assembly and Myp2 and Myo51 compensate when Myo2 function is compromised. Myp2 is most important for ring constriction. Myo2 and Myp2 localize to separate concentric sub-sections of constricting rings. The presence of Myo51 improves the performance of both Myo2 and Myp2 during ring formation and constriction. Participation of multiple myosins may explain why cytokinesis is successful in other organisms with myosin-II mutations.

Results

Myo2, Myp2 and Myo51 accumulate at the equator at discrete cell cycle times

Myo2p, Myp2p and Myo51p each concentrated in the cytokinesis apparatus at different times (Figures 1A asterisk and 1B). We used spindle pole body separation to define cell cycle time zero. Myo2 concentrated in cytokinetic nodes around the equator between time −10 min and time zero (Figures 1A and 1B) [25]. Formin Cdc12p accumulated in cytokinetic nodes at time zero [25] and initiated actin polymerization from nodes at time +1 min (Figure 1A), coinciding with the onset of node motions [26]. We used GFP-CHD to mark actin filaments [27], because GFP-actin doe not incorporate into the contractile ring [28]. Cytokinetic nodes condensed along with the actin filaments (Figures 1A and 2D) to form a contractile ring between time +1 and +25 min (Figure 2E) [25, 26]. Some actin filaments and small bundles of filaments persisted on the lateral borders of rings throughout constriction (Figure 1A arrow) [26, 29]. Myo51 localized along For3p-dependent actin cables during interphase [23] giving a filamentous appearance in the cytoplasm (Figure 1A arrow). Beginning at time zero Myo51p concentrated in a narrow equatorial band and in filamentous structures between nodes [30] around the middle of the cell (Figures 1A asterisk and 1B), both dependent on actin filaments [23]. Myp2 began to accumulate in fully formed contractile rings at time +20 min (Figures 1A asterisk and 1B) [25]. Myo52 transports cargo including vesicles with Bgs1 β-1,3-glucan synthase along actin filament cables during interphase and mitosis, but does not concentrate in the contractile ring [14, 23, 24, 31–33]. We confirmed previous observations [32, 33] that the Δmyo52 strain grows slowly at both 25°C and 36°C, that the Δmyo52 mutation did not enhance the myo2-E1 growth defect at either temperature, but that Δmyo52 is synthetically lethal with Δmyp2 at 36°C (Figure S1A). Node coalescence was similar to wild-type cells in Δmyo52 Δmyp2 and Δmyo52 myo2-E1 cells (Figure S1C), evidence that Myo52 does not participate in cytokinesis beyond its role in delivering vesicles with Bgs1p to the division site.

Figure 1.

Recruitment of three myosins and actin filaments to the division site in wild-type cells. All times are in min relative to SPB separation at time zero. See also Figure S1. (A) Time series of fluorescence micrographs (maximum intensity projections of 18 confocal sections, inverted grayscale lookup table (LUT)) of single cells expressing two fluorescent fusion proteins: Sad1p-RFP to label spindle pole bodies (SPB); and a second protein fused to mEGFP or 3YFP. One marker is shown in each time series. Row 1: Sad1p-RFP. Row 2: mEGFP-Myo2p. Row 3: mEGFP-Myp2p. Row 4: Myo51p-3YFP. Row 5: GFP-CHD to mark actin filaments in actin patches and the contractile ring. Scale bar: 5µm.

(B) Outcomes plot showing the fraction of cells with each of three myosins at the cleavage site over time: (●) mEGFP-Myo2p; (■) Myo51p-3YFP; and (▲) mEGFP-Myp2p.

(C) Color code of strains with corresponding genotype and functional myosin present used in the following figures.

Figure 2.

Contractile ring formation and constriction in wild-type cells and strains lacking one or two myosins. Confocal fluorescence micrographs are maximum intensity projections of 18 confocal z-sections. See also Figure S2. Scale bars: 5µm.

(A) Confocal fluorescence micrographs (inverted grayscale LUT) of fields of asynchronous cells. Each is a single time point from Movies S1–4. The functional myosin(s) in each strain are at the top of each panel.

(B) Confocal micrographs of newly formed rings in myo2-E1 Δmyo51 cells with the percentage of each type of ring in the population.

(C) Graph of the distribution of instantaneous node velocities for seven genotypes and wild-type cells with 100 µM Latrunculin A (LatA). The active myosins in each strain are indicated. Nodes marked with mEGFP-Myo2p or mEGFP-Myo2p-E1 were tracked at 2 s intervals (n ≥ 20 nodes per mutant genotype except for the LatA treated sample, n = 12 nodes). The two-sample Kolmogorov-Smirnov test showed that with the exception of the Myo2 Myp2 (Δmyo51) strain the node velocity distributions of all the genotypes and LatA treated samples differed significantly (p > 0.05) from the wild-type cells.

(D) Time series of micrographs (inverted grayscale LUT) of strains expressing mEGFP-Myo2p or mEGFP-Myo2p-E1 and Sad1p-RFP. Insets show confocal micrographs of rings in Δmyp2 and Δmyp2 Δmyo51 cells with the mean percentage difference between the ring circumference and the cell circumference.

(E–F) Outcomes plots showing the accumulation over time of cells with (E) their nodes condensed into a contractile ring or other compact structure at the equator or (F) the onset of ring constriction. Time zero is at SPB separation. All cells expressed mEGFP-Myo2p or mEGFP-Myo2p-E1 to mark nodes. The legends list the functional myosins in each strain. A Log-Rank statistical test for outcome plots showed that each strain differed from wild-type cells (black lines) with (*) p < 0.05, (**) p < 0.01 and (***) p < 0.001.

(G) Phenotypes of rings observed in confocal micrographs of myo2-E1 Δmyp2 relying on Myo51 and expressing mEGFP-Myo2p-E1. Left panel: Time series of a binucleate cell that failed to form a septum in the previous cell cycle, with cell cycle time in minutes. Arrow points to a ring detaching from the cell cortex. Dashed outlines highlight the cell. Right panel: Two time points (minutes after first image) of a cell that formed a septum and divided after a long delay. Dashed outlines highlight the cell.

(H) Histogram of average contractile ring constriction rates (± standard deviations, n = 20 to 25 cells per genotype). The active myosins in each strain are indicated within the bars. Asterisks indicate significant statistical difference of separate strains compared to wild-type (black bar) by student t-test (*) p < 0.05 and (**) p < 0.01.

(I) Kymographs at 2 min intervals of contractile rings in horizontally-oriented single cells generated from time series of micrographs (inverted grayscale LUT). Kymographs are aligned on the vertical line at the time nodes completed coalescence. Rings in myo2-E1Δmyp2 cells did not constrict but undulated and disassembled after node coalescence.

Myo2, Myp2 and Myo51 left the contractile ring at characteristic times. The Myo51p fluorescence in the ring declined shortly after the onset of ring constriction at time +35 min and disappeared over 15 min (Figures 1A, 1B and S1F). Both Myo2 and Myp2 were retained in rings during constriction (Figure 1B) and their local concentrations increased until the ring disappeared [28, 34].

Cells with mutations in a single or a pair of myosins are viable

Loss-of-function alleles of myo2⁺ and genetic deletions of myp2⁺ and myo51⁺ allowed us to evaluate how these three myosins contribute to cytokinesis (Figures 2, S2 and S1; Movies S1–S4). Strains with deletion mutations of myp2⁺ or myo51⁺ were viable at both 25°C and 36°C (Figure S1A). Motegi et al. reported that a different Δmyp2 strain was temperature sensitive at 36°C [19]. Since myo2⁺ is essential for viability [9, 10], we used the severely disabled myo2-E1 allele [11] that grows slowly at 36°C (Figure S1A). A point mutation in the catalytic domain gives the Myo2p-E1 protein low ATPase activity and very weak affinity for actin filaments in presence of ATP even at 25°C [12]. Given the minimal motor activity of the Myo2p-E1 protein, its tail may provide functions that account for the viability of the strain. The supplemental materials report experiments on the phenotypes of three additional myo2⁺ mutations (Figure S2). The two strains with point mutations in the catalytic domain (myo2-S1 and myo2-S2) were viable but grew slowly with obvious cytokinetic defects at both 25 and 36°C, whereas the neck deletion strain, myo2-ΔIQ1, encoding a myosin that cannot translocate actin filaments in vitro [12], had no noticeable growth or phenotypic defects at any temperature tested. Like wild-type cells [25] myo2-E1 cells assembled and constricted their contractile ring faster at 36°C than at room temperature, and since their defects were not enhanced at 36°C, we studied all of the strains at 23–25°C.

All three double-mutant strains were viable at 25°C, but combining the myo2-E1 mutation with either the Δmyp2 or Δmyo51 mutations was synthetically lethal at 36°C (Figure S1A), as shown previously for myo2-E1 Δmyp2 [19, 35]. A large fraction of double mutant cells accumulated in cytokinesis at 25°C, indicating that cells depending on single cytokinetic myosins have strong defects (Table 1). None of the strains with single or double myosins mutations differed significantly from wild-type cells in the numbers of the remaining myosin(s) in whole cells or contractile rings (Figure S1D–F).

Table 1.

Contractile ring parameters.

Genotypes	Functional myosins present	Peak node velocities, nm/sa	Mean time coalescence, minb	Mean time maturation, minc	Mean time constriction onset, mind	Mean constriction rate, µm/mine	Percen cytokinetic cell 25°Cf
Wild type	Myo2 Myp2 Myo51	38, 62, 78	18.5 ± 4.2	13 ± 4	31.9 ± 2.9	0.27 ± 0.04	19
myo2-E1	Myp2 Myo51	28	24.2 ± 3.9	9 ± 4	33.4 ± 3.1	0.28 ± 0.05	29
Δmyp2 Δmyo51	Myo2 alone	38, 68	35.5 ± 6.8	8 ± 4	47.1 ± 10.4	0.07 ± 0.03	42
Δmyo51	Myo2 Myp2	38, 62, 72	24.8 ± 4.0	8 ± 4	31.7 ± 4.5	0.30 ± 0.05	22
myo2-E1 Δmyp2	Myo51 alone	22	41.5 ± 11.4	Indefinite	None	None	52
Δmyp2	Myo2 Myo51	38, 58	15.1 ± 3.7	30 ± 10	45.7 ± 8.3	0.20 ± 0.06	22
myo2-E1 Δmyo51	Myp2 alone	25	63.6 ± 10.7	11 ± 14	65.2 ± 13.5	0.07 ± 0.03	59

^aSee Figure 2C.

^bAll cell cycle times are relative to SPB separation at time zero. See Figure 2E.

^cCalculated from the data in Figures 2E and F.

^dSee Figure 2F.

^eSee Figure 2H.

^fPercent of cells with mEGFP-Myo2p in nodes or a contractile ring in populations of asynchronous cells at 25°C (n = 115 to 205 per genotype).

Time-lapse confocal microscopy of live cells with a fluorescent tag on Myo2p provided data for four measurements to evaluate contractile ring assembly and constriction: 1) the cell cycle time when nodes started to move; 2) the distribution of instantaneous velocities of moving nodes; 3) the cell cycle time when nodes coalesced into a contractile ring or other compact structure where individual nodes were no longer resolved by confocal microscopy; and 4) the time course of ring constriction and disassembly. Table 1 lists all of the statistics. Outcomes plots (Figures 2E, 2F and S2A) and the standard deviations of the mean duration for nodes to coalesce (Table 1) illustrate the variability in the process in each cell population.

The distribution of node velocities had a major peak at ~40 nm/s and minor peaks at ~60 and 80 nm/s (Figure 2C). Wild-type cells completed the assembly of a ring between times +15 min and +25 min (Figure 2E) with a mean completion time of +18.5 ± 4.2 min (mean ± standard deviation) on the cell cycle clock. In an unsynchronized population of wild-type cells, 19% of the cells were in cytokinesis meaning that they had some mEGFP-Myo2p labeled features at their equator ranging from a broad band of nodes to a fully constricted and disassembling ring (Table 1).

All six strains with mutations in one or two myosin genes had cytokinesis defects, but the double mutant strains were more informative than strains with a single myosin mutation as they revealed the contributions of the single myosin present. Strains with mutations in one myosin gene revealed how the remaining pair of myosins functions together. To simplify the presentation we refer to strains by the functional myosin(s) they expressed in bold font in addition to their genotype in italic font throughout the text, and we use a color code to identify each strain in the figures (Figure 1C).

Single cytokinesis myosins can support contractile ring assembly

All three double-mutant strains coalesced their nodes into compact structures, but the process took longer and was more variable than in wild-type cells (Figures 2A, B, D and E; Movies S2–S4; Table 1). Nodes began to move at time +1 min in all three double mutant strains as in wild-type cells, but these movements were often ineffective and most rings formed were imperfect. Thus Myo2, Myp2 and Myo51 can each contribute to node coalescence in the absence of other two myosins.

Cells dependent on the motor activity of Myo2 alone (Δmyp2 Δmyo51 strain) completed node coalescence 17 min later than wild-type cells (Figure 2E; Table 1). The main peak of node velocities in these cells was around 40 nm/s like wild-type cells (Figure 2C), but fewer nodes moved at the higher rates observed in wild-type cells. About 80% of rings in these double mutant were bent, because they were larger than the circumference of the cell (Figure 2D inset). Constriction subsequently reduced the circumferences of these large rings suggesting that they were not taut when newly assembled. The Δmyp2 Δmyo51 cells had cytokinetic defects including multiple septa and branched morphology (Figure S1B) [30].

Contractile rings assembled slowly in cells dependent on Myo51 alone (myo2-E1 Δmyp2 strain) (Figure 2E; Table 1), with a slightly longer mean time than cells with only Myo2. However, unlike cells with only Myo2, the distribution of node velocities had a single peak at ~20 nm/s, so Myo51 alone cannot generate fast node movements (Figure 2C, Table 1). The newly assembled rings were often irregular in shape (Figure 2G).

Of the three double-mutant strains, node coalescence was compromised most severely in cells dependent on Myp2 alone (myo2-E1 Δmyo51 strain). Nodes marked with disabled mEGFP-Myo2p-E1 began moving at cell cycle time +1 min with a single peak at ~25 nm/s in the distribution of node velocities (Figure 2C). We do not understand the mechanism of the slow node motions before Myp2 arrived at the equator at time +20 min in these cells. These cells took much longer to coalesce their nodes than wild-type cells. The first cells with complete rings appeared at time +45 min and the last at time +90 min (Figure 2E). Although 35% of these cells assembled a single, normal ring, the remaining cells had 2 or 3 separate rings, lasso-shaped rings, two laterally joined rings or curved bundle of mEGFP-Myo2p-E1 (n = 63 cells) (Figures 2A and 2B).

These observations show that Myo2 is the main myosin for node coalescence, followed by Myo51. Myp2 normally contributes little to node coalescence, in accord with its late arrival (+20 min) in the fully formed ring in wild-type cells (Figure 1A).

Pairs of myosins assemble contractile rings better than single myosins

Strains with any combination of two myosins assembled contractile rings better than single myosins, showing that they normally cooperate in the process. Pairing Myo2 with either Myo51 or Myp2 was more effective than the combination of Myp2 and Myo51.

The pair of Myo2 and Myo51 in Δmyp2 cells was the most successful at ring assembly. Both myosins arrived at their normal cell cycle times, and nodes moved with nearly normal velocities (Figure 2C). Nodes coalesced into rings slightly earlier than in wild-type cells (Figure 2E, Table 1) and the only defect was a slight bowing of the rings in about one fifth of the cells (Figure 2D inset). Our Δmyp2 cells were morphologically normal when grown in rich medium at room temperature and 22% of these cells were in cytokinesis similarly to a wild-type cell population (Table 1). Therefore, Myp2 does not normally contribute to contractile ring assembly.

Cells dependent on Myo2 and Myp2 (Δmyo51 strain) had a very subtle phenotype. The arrival of Myo2 in equatorial nodes at cell cycle time −10 min, and motions of these nodes after cell cycle time +1 min were indistinguishable in wild-type cells (Figure 2C). Nevertheless, Δmyo51 cells completed the coalescence of their nodes ~6 min later than wild-type cells (Figure 2E, Table 1). The Δmyo51 cells had no obvious morphological defects at room temperature and 22% of these cells were in cytokinesis similar to wildtype cells (Table 1) [23, 30]. Thus Myo51 makes a modest contribution to node coalescence.

The combination of Myp2 and Myo51 (myo2-E1 strain) performed only modestly better than cells with Myo51 alone and had more problems with ring assembly than the other strains with two functional myosins. Nodes began moving at the normal time +1 min, but the node velocity distribution had only a single peak at the slow velocity of ~30 nm/s (Figure 2C). The population of myo2-E1 cells began to accumulate compact rings ~10 min late, and the mean time for completion was 6 min late (24.2 ± 3.9 min) (Figure 2E, Table 1).

Close examination of the myo2-E1 cells revealed subtle problems with the distributions of both myosins and actin filaments. Myp2 and Myo51 accumulated at the division site at their normal cell cycle times (asterisks in Figures 3A and 3B), but the equatorial zone of Myo51p-3YFP was broader than normal (Figure 3B bracket, 3D and 3E). Myp2 appeared in spots near the equator rather than in fully formed rings (Figure 3A). Most Myp2 spots were separate from nodes marked with mEGFP-Myo2p-E1 (Figure 3F). Cells treated with Latrunculin A to depolymerize actin filaments did not accumulate these spots of Myp2 even though the localization of mEGFP-Myo2p to nodes was unaffected (Figure 3G). Thus localization of both Myo51 [30] and Myp2 to the equator requires actin filaments.

Figure 3.

Localization of myosins in cells with and without Myo2p function or actin filaments. Confocal fluorescence micrographs are maximum intensity projections of 18 confocal z-sections. Scale bars: 5 µm.

(A) Time series of micrographs (inverted grayscale LUT) of wild-type and myo2-E1 cells expressing mEGFP-Myp2p and Sad1p-RFP (not shown). Myp2p localizes to the equator 20 min after SPB separation in both wild-type and myo2-E1 mutant cells.

(B) Time series of micrographs (inverted grayscale LUT) of myo2-E1 cells expressing Myo51p-3YFP and Sad1p-mEGFP (top) and GFP-CHD (bottom). Myo51p localizes to the equator at t = 0 min at SPB separation in myo2-E1 mutant cells as in wild-type cells (see Figure 1A). The equatorial zones of both Myo51p-3YFP and the GFP-CHD were wider in myo2-E1 cells than wild-type cells (brackets).

(C) Graph of the ratio of equatorial GFP-CHD fluorescence intensity to the total cellular GFP-CHD fluorescence intensity in (■) wild-type cells and (●) myo2-E1 cells. Student t-tests show significant differences (p < 0.05) between the ratios in wild-type and myo2-E1 cells.

(D, E) Graphs of the width of the broad band of nodes, actin network and Myo51 in cells expressing Rlc1p-tdTomato, GFP-CHD or Myo51p-3YFP in (D) wild-type cells (n = 11) and (E) myo2-E1 cells (n = 11). The width of the network of actin was measured until the actin patches returned to the equator masking the network of actin filaments and bundles. Error bars are standard deviations.

(F) Micrographs (inverted grayscale LUT) and color merge of a cell expressing mEGFP-Myo2p-E1 (red) and mCherry-Myp2p (green). Myo2p-E1 and Myp2p do not colocalize.

(G) Effect of 100 µM LatA for 10 min on the localization of mEGFP-Myo2p and mEGFP-Myp2p in cells expressing Sad1p-RFP to determine cell cycle time: (left) mEGFP-Myo2p in a wild-type cell treated with LatA; (middle) mEGFP-Myp2p in a myo2-E1 cell treated with LatA; and (right) mEGFP-Myp2p in a myo2-E1 cell, DMSO. Micrographs (inverted grayscale LUT).

In both wild-type and myo2-E1 mutant cells, actin accumulated steadily around the equator during the 30 min of ring assembly and maturation, peaked at the onset of ring constriction and decreased steadily as the ring disassembled (Figure 3C). However, the GFP-CHD fluorescence at the equator was higher throughout all phases of cytokinesis in myo2-E1 mutant cells (Figure 3C), especially at the end of cytokinesis when the GFPCHD signal persisted for ~15 min longer than normal (Figure 3C, movie S5). Furthermore, the equatorial actin network was wider than normal in myo2-E1 cells (Figures 3B bracket, 3D and 3E) even after the nodes had coalesced.

Short maturation periods compensate for slow ring assembly in all myosin mutant strains except Δmyp2

The maturation phase, the period of time after nodes coalesce into a ring and before the ring starts to constrict, normally lasts about 13 min (compare Figures 2F and 2G, Table 1). In many mutant strains with slow contractile ring formation [36–41], the maturation period is foreshortened by the onset of ring constriction at the normal cell cycle time, +32 min in our experiments. In cells where node coalescence extends well beyond +32 min, constriction starts just a few minutes after a continuous ring forms [36, 39] consistent with a cell cycle clock that delays constriction until a set time of +32 min or until the ring is complete if ring assembly exceeds this limit.

Most of our mutant strains followed these patterns. The maturation periods were shorter myo2-E1 cells (9 ± 4 min), Δmyo51 cells (8 ± 4 min) and Δmyp2 Δmyo51 (8 ± 4 min), which all had delays in ring formation. The myo2-E1 Δmyo51 cells relying on Myp2 took more than an hour to form rings, well beyond the 32 min time point, and then started constricting shortly after nodes finished coalescing (Figures 2E and 2F). The rings in myo2-E1 Δmyp2 cells dependent on only Myo51 never properly constricted, so maturation was indefinitely long (Figures 2G and 2I, Table 1).

The Δmyp2 cells were an exception, since they assembled rings slightly faster than wild-type cells, but had a very prolonged maturation period of 30 ± 10 min (Figures 2E and 2F). So the absence of Myp2 delayed the onset of constriction far beyond the usual time set by the cell cycle clock. As the rings showed no observable defect, the reason for the delay in the onset of ring constriction is unclear.

Myo2 and Myo51 support contractile ring constriction by Myp2

We followed contractile ring constriction in strains with single and double mutations of the three myosins using fluorescent protein tags on either the N-terminus of the Myo2p heavy chain (Figures 2F, 2H and 2I) or the C-terminus of the regulatory light chain Rlc1p (Figure S2). Rings began to constrict in wild-type cells at cell cycle time +32 min and continued at a steady rate of 0.27 µm/min until the ring circumference decreased from 10.5 ±0.3 µm to ~4 µm (Figures 2F, 2H and 2I, Table 1, Movie S1).

In cells relying on Myp2 alone (myo2-E1 Δmyo51 strain) contractile ring constriction started very late (owing to very slow ring formation compensated by a very short maturation time) but proceeded at ~55% of the rate of wild-type cells (Figures 2F, 2H and 2I, Table 1, Movie S3). Although 38 of 63 of these cells formed abnormal rings (Figure 2B), all of these rings constricted. Cells with laterally joined rings (4% of rings observed, Figure 2B) slowly merged the two rings into one before constriction started. In the cells with lasso-shaped rings, the lateral branch disassembled as the ring constricted (Movie S3).

Rings in cells expressing Myo2 alone (Δmyp2 Δmyo51 strain) constricted at 0.07 µm/min, ~25% the rate of wild-type cells (Figures 2H and 2I, Table 1, Movie S2), but 14 of 37 rings regressed after constricting partially (Figure 2I). In those cells starting with rings larger than the circumference of the cell (Figure 2D inset) constriction pulled the rings into taut straight rings. In a few Δmyp2 Δmyo51 cells (6/37 cells) a branch marked with mEGFP-Myo2p separated from the constricting ring and reformed a new ring, similar to Δmyp2 cells.

Of the strains dependent on a single myosin, only the myo2-E1 Δmyp2 strain relying on Myo51 alone failed to constrict their contractile rings (Figures 2G, 2H and 2I, Movie S4). Newly assembled rings were smooth but bent. Rather than constricting these rings fragmented or separated into node-like structures (Figure 2G, Movie S4). A large proportion (77%, n = 77 cells) of the cells failed to separate during time lapse movies lasting >2 h. The strain was viable, because the remaining 23% of the population eventually formed a septum that allowed the daughter cells to separate (Figure 2G). Of the cells that did not separate, 60% grew a septum juxtaposed to the irregular ring visible as a clear indentation. The ring material in the remaining 40% disassembled without the formation of a visible septum (Figure 2G).

The presence of either Myo2 or Myo51 with Myp2 allowed for normal ring constriction, and the presence of Myo51 improved the performance of Myo2 (Figure 2H and 2I). Wang et al. also observed normal constriction rates in Δmyo51 and myo2-E1 cells [30]. Two of these strains had unique features. Rings in myo2-E1 mutant cells disassembled slowly suggesting a role for Myo2 in contractile ring disassembly (Figures 2I arrow, 3C and Movie S5). Constricting rings branched in a few Δmyp2 cells (4/45 cells), and the branch formed a second ring that also constricted. This behavior suggests that Myp2 helps to maintain ring integrity during constriction.

Myo2 and Myp2 localize to separate concentric ring sub-sections during constriction

These experiments show that Myo2 is the most important myosin for ring assembly whereas Myp2 is the most important myosin for ring constriction. Furthermore, the two myosins have different distributions around the equator [35], so we analyzed the distributions of Myo2p and Myp2p in contractile rings. We used microfabricated devices to hold yeast vertically during imaging [42, 43] to improve the resolution in the plane of the ring over digital reconstructions from series of optical sections (Figure 4).

Figure 4.

Two layers of myosin-II in constricting contractile rings.

(A–E) Fluorescence micrographs (inverted grayscale LUT and color merge, maximum intensity projections of 5 confocal z-sections) from time lapse (see also Movie S5) of cells held vertically in microfabricated chambers. Scale bars: 5 µm.

(A) Time series of a wild-type cell expressing mCherry-Myo2p and mEGFP-Myp2p. The Myp2 ring began to constrict before the Myo2p ring resulting in concentric rings with the Myp2p ring inside the Myo2p ring. Time in min from time zero at the onset of constriction.

(B) Wild-type cell expressing mCherry-Myo2p and GFP-CHD to mark actin filaments in a constricting ring showing overlap between the two proteins. GFP-CHD also labels endocytic patches that form on both sides of the cleavage furrow (arrow).

(C) Wild-type cell expressing mCherry-Myp2p and GFP-CHD to mark actin filaments showing overlap between the two proteins in a constricting ring and several punctate actin patches.

(D) Micrographs at two time points of a myo2-E1 mutant cell expressing GFP-CHD and mCherry-Myp2p.

(E) Cleavage furrow of a wild-type cell expressing the plasma membrane marker GFPPsy1p and mCherry-Myp2p.

(F) Cleavage furrow of a wild-type cell expressing the plasma membrane marker GFP-Psy1p and mCherry-Myo2p. Fluorescence micrographs are z-reconstructions of maximum intensity projections of 18 confocal z-sections from a cell imaged lying horizontally.

Time-lapse confocal microscopy (Movie S5) showed that Myo2 and Myp2 colocalized in the contractile ring prior to the start of constriction (Figure 4A, top row) but during constriction the ring of Myp2 was internal to the ring of Myo2 (Figure 4A, bottom three rows). The internal Myp2 ring began constricting ~4 min earlier than the peripheral Myo2 ring but constricted at a slightly slower rate than the Myo2 ring, so the two rings converged at the end of constriction. Rings of both Myo2 and Myp2 were located ahead of the leading edge of the invaginating plasma membrane marked with the SNARE protein GFP-Psy1p [44] (Figure 4E and 4F). During constriction Myo2 remained in a perfect ring, but the distribution of Myp2p around the ring was irregular and varied over time. The ring of actin filaments marked with GFP-CHD overlapped with both mCherry-Myo2p and mCherry-Myp2p suggesting that there are sub-regions of the contractile ring enriched in Myo2 or Myp2 (Figures 4B and 4C, Movie S5). Bundles of actin filaments containing mCherry-Myp2p peeled off from the main ring and moved centripetally (Movie S5) in wild-type cells and more so in myo2-E1 cells (Figure 4D, Movie S5). The Psy1p marker did not associate with these fragments suggesting that plasma membrane does not follow.

Discussion

In spite of evidence that myosin-II motor activity is not essential for cytokinesis in several model organisms [5–8, 11] previous work did not provide decisive evidence that other types of myosin contribute to the assembly or constriction of the contractile ring. We show that two myosins besides conventional myosin-II participate in cytokinesis in fission yeast. Each myosin has different biochemical properties and each makes a unique contribution to the assembly or constriction of the contractile ring. Given the participation of three myosins in a redundant system, mutations of single myosin genes did not give a clear picture of their contributions to cytokinesis. However, their functions and limitations were obvious in strains with pairs of mutations making the cells dependent on just one of the three myosins. Multiple myosins may explain cytokinetic ring constriction in COS-7 cells with compromised myosin-II activity [8].

A hierarchy of myosins contributes to contractile ring assembly

Normally Myo2 and Myo51 cooperated to assemble contractile rings, and either Myo2 or Myo51 can assemble normal contractile rings without the assistance of Myp2. Myo2 is the first myosin to arrive and is a component of cytokinesis nodes, followed at time zero by Myo51, which concentrates in puncta among the broad band of nodes [30] [23, 33]. Localization of Myo51 depends on actin filaments assembled from nodes and on a complex of proteins called Rng8p and Rng9p [30]. As a component of nodes Myo2 is thought to apply force on actin filaments from other nodes [26], whereas Myo51 is more likely to act on oppositely polarized actin filaments between nodes [30]. Myo51 may promote node interactions by pulling antiparallel filaments into bundles that interact with nearby nodes as proposed by Wang et al. [30] as well as attracting actin cables to the forming ring [45].

Node condensation by Myo51 alone has two interesting features. First, although nodes start moving near time +1 min in cells relying on either Myo2 or Myo51 alone, these strains require far more time to condense nodes into a contractile ring than cells with both Myo2 and Myo51. Thus, even though Myo2 alone moves nodes at normal rates, condensation is slow without Myo51. Clearly these two myosins interact productively during contractile ring formation. Second, formation of the narrow equatorial band of Myo51 depends on actin filaments [23, 33], but these initial rings of Myo51 form before nodes condense and the distribution of Myo51 in these rings differs from actin and all other known cytokinesis components (Figure 1A).

Myp2 does not participate in contractile ring assembly in wild-type cells, because it arrives after rings are complete, but in cells lacking both Myo2 and Myo51 function, Myp2 alone can move nodes at ~25 nm/s and these motions slowly form a contractile ring. However, many of these rings are defective. Purified Myo2p-E1 has little activity even in the presence of its chaperone Rng3p [12], so residual Myo2p-E1 motor activity is unlikely to assist Myp2 in these cells. Actin filaments are necessary but not sufficient for Myp2 to concentrate around the equator, since it arrives well after actin. Thus, other factors must limit targeting of Myp2 to the equator until time +20 min.

A different hierarchy of myosins contributes to contractile ring constriction

Judging from the behavior of cells with single functional cytokinetic myosins, Myp2 is the only myosin that reliably completes ring constriction on its own. This is surprising, since Myp2 is dispensable for cytokinesis under most conditions [19, 21, 22]. On the other hand, the essential Myo2 alone can only support constriction at 25% of wild-type rate and partially completed furrows often regressed unless Myo51 was also present. Myo51 cannot support ring constriction in the absence of Myp2 and Myo2, but is an important accessory as it can assist either Myo2 or Myp2 during constriction.

The explanation for these observations is that each of the three myosins makes a unique contribution to constriction and furrow formation, such that pairs of the myosins are more effective than any single myosin. The tails of the myosins may be responsible, since experiments with chimeras showed that the tails of Myo2p and Myp2p each make unique contributions [35].

Computational simulations of a model of cytokinetic ring constriction with myosin clusters and anchored formins polymerizing actin filaments produce forces similar to those measured in protoplasts [29]. This model is agnostic about the type of myosin generating the forces but assumes that myosin forms clusters associated with the plasma membrane that interact with oppositely polarized actin filaments. Myo2 fulfills these criteria in that it aggregates at physiological salt concentrations [21] and multiple Myo2 molecules cluster in nodes associated with the membrane [28]. However, Myo2 is less important than the single headed Myp2 in constricting the ring. Neither Myp2 nor Myo51 is known to associate with nodes, so they may interact with actin filaments in between their Cdc12p anchors. Myo51 forms macromolecular assemblies with Rng8p and Rng9p [30], but nothing is known about the assembly of Myp2 or its interactions with the plasma membrane. Given that Myo2 is closer to the base of the cleavage furrow than Myp2, Myo2 may have stronger physical interactions with transmembrane proteins that anchor the contractile ring than Myp2.

Our evidence for three myosins participating in contractile ring assembly and constriction raises a long list of questions about mechanisms and opens up new areas of research. Addressing these questions will require more detailed information about the physical and enzymatic properties of the three myosins and their organization in the contractile ring.

Experimental Procedures

Strains, growing conditions, and genetic and cellular methods

Table S1 lists the S. pombe strains used in this study. All tagged genes, except for the GFP-CHD construct, were under the control of endogenous promoters and integrated at their native chromosomal loci. N-termini of myo2⁺ and myp2⁺ were tagged by integrating their promoters into the pFA6a fission yeast integration vector [46]. Long PCR products were generated and integrated into the genome [46]. Positive clones were selected on YE5S plates with 0.1 mg/mL G418 and verified by PCR of the insert using appropriate primers, by fluorescence microscopy and DNA sequencing. We tagged genes in the myo2-E1, myo2-S2 and myo2-ΔIQ12 in the same way and verified the specific point mutations or deletions by genomic DNA sequencing. Cells were grown in an exponential phase for 36–48 h before microscopy [47]. To de-repress the expression of the Pnmt41x-GFP-CHD actin marker, cells were shifted to EMM5S 9–12 h prior to imaging.

Microscopy and data analysis

Cells for live-cell microscopy were collected from liquid cultures, centrifuged at 5,000 rpm, and then washed into EMM5S for imaging. Cells were spun down for 30 s at 5,000 rpm, washed in EMM5S and then in EMM5S with 0.1 mM n-propyl-gallate and resuspended in EMM5S with 0.1 mM n-propyl-gallate and placed onto a thin disk of 25% porcine gelatin (Sigma-Aldrich) dissolved in EMM5S + 0.1 mM n-propyl-gallate. Cells were covered with a number 1.5 coverslip, sealed with VALAP and immediately imaged at room temperature (23–25°C).

We imaged cells with an inverted Olympus IX-71 microscope with a 100×/1.40 oil objective (Olympus) on a spinning-disk confocal microscope (Yokogawa CSU-X1) with 488/514, and 568nm argon ion lasers and an electron multiplying cooled charge-coupled device camera (EMCCD IXon 897, Andor). Maximum intensity projections of images, gray scale montages, and other image analyses were made with ImageJ (National Institutes of Health). Images in figures are maximum intensity projections of z-sections spaced at 0.34 µm. Images were analyzed with ImageJ (National Institute of Health). We defined the end of node coalescence as the time when nodes converged into a narrow ring around the equator such that individual nodes were no longer resolved by confocal microscopy.

We measured the rate of ring constriction on kymographs of maximum projection images (18 z confocal planes taken for 6.12 µm) of time-lapse datasets taken at 1–2 min time intervals. The diameter of a ring was measured automatically by fluorescence thresholding and the circumference was calculated.

We counted myosins in whole cells and contractile rings by measuring fluorescence intensities in stacks of 18 optical images separated by 0.34 µm of fields of cells [28]. The images were corrected for the camera noise and uneven illumination. The numbers of molecules were calculated from the fluorescence intensities and standard curves for either YFP or mEGFP.

Node velocity measurements

A single z-plane of cells expressing a node marker was imaged by time-lapse microscopy at intervals of 110 ms, 1 s or 2 s. Nodes were identified manually and selected using a Region Of Interest 7 pixels (0.58 µm) in diameter using ImageJ. Nodes were tracked through time until the object disappeared (perhaps due to photobleaching of the fluorescent tag or moving out of focus) or until it fused with other another node or the assembling ring. At each time point, the node position was obtained with sub-pixel resolution using its fluorescence intensity center of mass, i.e. the brightness-weighted average of the×and y coordinates of all pixels in the selection. Velocities of nodes were calculated using Matlab as the displacement of the particle between two consecutive time points and filtered using a moving average with a weight of 2 for the current time point and a weight of 1 for the previous and following time points. This procedure removed occasional jumps in speeds due to localization errors.

Yeast cell holders

We made a mold on a silicon wafer consisting of regular arrays of pillars, 6 or 7 µm in diameter and 14 µm high. The pillars were made of SU8 negative resist using standard photolithography methods at the School of Engineering and Applied Science clean room (Yale University). We then cross-linked poly(dimethyl siloxane) (PDMS) on the mold, using Sylgard 184 Silicone Elastomer Kit (Dow Corning). We mixed the base and the cross-linker at 10:1 ratio, degassed the mixture, and poured it on top of the mold. The PDMS was cured overnight at 65°C. Yeast holders were cut from the PDMS, placed under vacuum for at least 30 min before being used with cells. A dilute solution of cells prepared as above was poured on top of the PDMS mold, turned onto a circular coverslip number 1.5 of 40 mm in diameter. The mold was pressed against the coverslip to allow the cells to enter the chambers and the cells were imaged immediately. Chambers 6 and 7 µm in diameters allowed space for medium around the cells.