Assembly Mechanism of the Contractile Ring for Cytokinesis by Fission Yeast

See allHide authors and affiliations

Science  04 Jan 2008:
Vol. 319, Issue 5859, pp. 97-100
DOI: 10.1126/science.1151086


Animals and fungi assemble a contractile ring of actin filaments and the motor protein myosin to separate into individual daughter cells during cytokinesis. We used fluorescence microscopy of live fission yeast cells to observe that membrane-bound nodes containing myosin were broadly distributed around the cell equator and assembled into a contractile ring through stochastic motions, after a meshwork of dynamic actin filaments appeared. Analysis of node motions and numerical simulations supported a mechanism whereby transient connections are established when myosins in one node capture and exert force on actin filaments growing from other nodes.

Cytokinesis is the final step in cell division, when a dividing cell physically separates into two. Assembly and constriction of a contractile ring of actin, myosin, and associated proteins cleaves cells in many organisms, including animals and yeast. Analysis of budding and fission yeast identified ∼50 participating proteins (1), the temporal sequence of events (2), and local concentrations of many proteins (3). A major challenge is to understand how the network of proteins functions, ultimately at the level of a testable quantitative model of the entire system.

Fission yeast assembled a cytokinetic contractile ring by condensation of a broad band of membrane-associated nodes (Fig. 1, A and B, and Movie S1). The nucleus determines the initial location of the zone with these nodes by releasing the anillin-like protein Mid1p (4), which recruits to nodes the motor protein myosin-II, followed by other proteins, including formin Cdc12p (2). Actin filaments nucleated by Cdc12p (5, 6) are proposed to establish a network between the nodes, allowing myosin-II to pull the nodes together (710) into a ring (Fig. 1C).

Fig. 1.

Formation of contractile rings from bands of nodes. (A) Confocal maximum intensity projection of cells expressing Rlc1p-3GFP. Bands of nodes (yellow arrow) condense into rings (white arrow). (B) 3D reconstructions showing cells expressing Rlc1p-3GFP at progressive stages of ring formation (rows). Columns show successive 45° y-axis rotations. (C) Model forring formation via actin connections between nodes (s, arc length; R, radius). (D to H) Search, capture, pull, and release mechanism. (D) Actin filaments polymerize out of nodes with pointed ends leading at rate vpol. (E) Pointed ends passing within rc of another node form connections. (F) Each connection produces an attractive force F. Nodes move in the direction of the total force with velocity v = Ftot/ζ. (G) The lifetime of connections is τbreak, and (H) the lifetime of un-connected actin filaments is τturn; in both cases, a new filament grows immediately. (I and J) Simulated images of condensing bands (11). The y axis is the arc length s. (I) A model with permanent connections between each two nearest neighbors generates clumps. (J) Model with reactions in (D) to (H) (vpol = 0.2 μm/s; F/ζ = 20 nm/s; τturn = τbreak = 20 s; rc = 100 nm) condenses nodes into rings over 550 s (Movies S3 and S4). Scale bars, 5 μm.

Numerical simulations of a simple model, with nodes pulling on their nearest neighbors in a network of fixed connections, moved nodes along linear trajectories into clusters (Fig. 1I) rather than in a continuous ring around the equator (11). Thus, additional features must be present, which we explored using confocal microscopy to follow myosin-II and actin filaments with high temporal and spatial resolution in live cells. Cells expressed the myosin-II light chain Rlc1p fused to three copies of green fluorescent protein (GFP) or tdTomato from its native promoter (Fig. 2). We used the calponin homology domain of S. pombe IQGAP Rng2p tagged with GFP calponin homology domain (GFP-CHD) (12) to mark actin filaments (Fig. 3) because actin tagged with yellow fluorescent protein (YFP) did not incorporate into contractile rings (3, 10).

Fig. 2.

Node formation and motions in cells expressing Rlc1p-3GFP. (A) Time course (seconds) of node formation (Movie S5). (B) Position of a newly formed node at 1-s intervals. (C) Time course (seconds) of movements as nodes condense into a ring (Movie S2). (D) Details of node motions [1-s intervals from the series shown in (C)]. Node 1 (filled arrowhead) moves toward a node in the upper right corner and stops. Node 2 (open arrowhead) moves toward node 1. (E) MSD of newly formed nodes. Black curves, MSD of three nodes in the same cell versus time; gray lines, relative MSD between the three nodes. (F) Distribution of node velocities during condensation (data from 10 cells). Unresolved node motions in regions of high node density are not included. (G) Durations of node movements. (Inset) Directions of node motions with respect to the long axis of the cell (absolute values). In (A), (C), and (D), each frame is a moving average of six successive frames at 1-s intervals. Scale bars, 5 μm.

Fig. 3.

Actin filament and node dynamics in cells expressing GFP-CHD to mark actin filaments (A to D) and Rlc1p-mRFP1 (B) or Rlc1p-tdTomato (C and D) to mark nodes. Time in minutes in (A) and seconds in (C) and (D). (A) Time series showing actin patches at both ends of the cell and assembly of a contractile ring from a meshwork of actin filaments (Movie S8). (B) Six cells from the same field arranged according to cell-cycle stage showing networks of actin filaments connecting nodes. Movie S9 shows 3D projections of cells b to d. (C) Time series showing actin filaments transiently connecting two nodes (arrow) and regrowing from this node (arrowhead) (Movie S15). (D) Time series of actin filaments in the boxed region capturing a lagging node and pulling it toward the contractile ring until the filament apparently breaks (arrow). Kymograph of the moving node with time horizontal. See Movies S11 to S14. Scale bars, 5 μm [(A) and (B)], 1 μm [(C) and (D)].

We discovered that nodes make many starts, stops, and changes of direction as they condense into a contractile ring (Fig. 2, C and D, and Movie S2). The highly stochastic pattern of node motions suggested that the connections between nodes were unstable (Fig. 1, G and H). A model with traction between nodes depending on transient connections established by stochastic search-and-capture produced simulations in which nodes condense into a continuous ring through lifelike stop-go movements (Fig. 1J and Movies S3 and S4) (11). The success of this model depends on its parameter values, so we observed live cells to measure parameters and test assumptions and predictions.

Wild-type cells assembled 63 ± 10 (n = 22) nodes located close to the plasma membrane in a band ∼1.8 μm wide around the middle of the cell (Fig. 1A and Movie S1). Nodes were distributed randomly around the equator and in a Gaussian manner along the long axis of the cell (fig. S1A). Rlc1p-3GFP accumulated in individual nodes over ∼1 min (Fig. 2A, fig. S1B, and Movie S5). The Rlc1p-3GFP fluorescence intensities of nodes varied ∼fourfold within a cell (fig. S1C), fluctuated over time (Fig. 2B), and recovered with a half-time of about 30 s after photobleaching (11), indicating that the protein exchanges dynamically with the cytoplasmic pool.

When first formed, nodes moved relatively little for ∼10 min (Movies S5 and S6), but the mean square displacement (MSD) of the center of each node increased over time (Fig. 2E, black), consistent with 2-dimensional diffusion: MSD(t) = MSD(0) + 4Dt. The diffusion constants D distributed around 20 nm2/s in control cells and cells treated with Latrunculin A to depolymerize actin (fig. S2). Node motions appeared uncorrelated, because the slopes of relative MSD between pairs of nodes (Fig. 2E, gray, and fig. S2B) were on average equal to the sum of the slopes of individual node MSD curves. Diffusive behavior implies that any anchors or cross-links of nodes were dynamic or very soft on minute time scales. The large mass [>22,000 kD (10)] of nodes may contribute to D being orders of magnitude smaller than D of transmembrane proteins (13).

Nodes began to move about the time a dynamic meshwork of actin filaments appeared around the equator, 2.3 ± 1.9 min (n = 29 cells) after spindle pole bodies separated and Cdc12p joined the nodes (2) (fig. S3 and Movies S7 to S9). Nodes moved in 20-s bursts (Fig. 2G) at velocities of ∼30 nm/s (Fig. 2F and Movie S2). Node movements terminated in pauses or by merger with another node. Many motions could be clearly attributed to pairwise attractions, and a few nodes switched partners during these directed motions (Fig. 2D and fig. S4). Although most node motions were toward the final location of the contractile ring, the distribution of orientations was broad (Fig. 2G, inset) and the velocities and durations of movements were uncorrelated within our resolution (fig. S4J).

We used fluorescence microscopy to verify actin filament connections between nodes. Linear elements marked with GFP-CHD extended in all directions from Rlc1p nodes and established connections among them (Fig. 3, B to D, fig. S5, and Movies S10 to S15). The GFP-CHD fluorescent intensity of mature contractile rings [with ≥20 filaments in wild-type cells (3, 14)] was 10 to 20 times as high as the least intense linear elements in the broad band; thus, the faintest elements probably represent small bundles of actin filaments or perhaps even single actin filaments. Contractile ring actin filaments grew normally from nodes in vegetative cells lacking the formin For3p that nucleates interphase actin cables (12) (Movies S12, S13, and S16).

The dense actin filament network among the nodes obscured details of their relationships, but in favorable cases (e.g., during early stages of node condensation) we observed linear elements elongating at ∼0.2 μm/s (75 actin subunits/s) (Fig. 3C, fig. S5, and Movie S15). Some growing linear structures were aimed directly at a target node; others appeared to move laterally to associate with a node (Movies S11 and S13). When growing actin filaments buckled [as observed in vitro (15)], their contour lengths also increased at ∼0.2 μm/s (Movie S17). Some nodes connected by actin filaments moved along the direction of the linear element (Movies S12 to S14).

Filaments growing at 0.2 μm/s generally remained in a zone of full width ∼4 μm on both margins of contractile rings (Fig. 3, A and B), suggesting filament lifetimes of <20 s. We observed two processes that contribute to filament turnover. Long-growing filaments could break and short filaments could disappear rapidly (Movies S10, S15, and S16), which resembled severing of growing actin filaments by cofilin in vitro (16). Figure 3D (Movie S14) shows a node stopping when its actin connection apparently broke, and Movie S18 shows the ends of a broken filament recoiling as if under tension.

Monte Carlo simulations (11) of a model with search, capture, pull, and release (Fig. 1, D to H) reproduced the start-stop motion of nodes and over ∼10 min assembled a continuous ring with nonuniform fluorescence intensities (Fig. 1J and Movies S3 and S4) as observed in cells (Fig. 1B). We assume that nodes nucleate actin filaments that grow in random directions parallel to the membrane. Myosin-II binds filaments passing within 100 nm and exerts constant tensile force on the filament, tending to reduce node separation. The total force Ftot from all connections results in a node velocity v = Ftot/ζ, where the friction coefficient ζ = kBT/D ≈ 0.2 pN s/nm from the Einstein relation, using D ∼20 nm2/s. Because nodes move at 30 nm/s, Ftot ∼6 pN.A force of 4 pN per connection reproduced the observed distribution of node speeds (fig. S6). This force value is a lower bound, because sources other than thermal motion may contribute to the observed translational diffusion of nodes. An average of 40 Myo2p heads per node (3), with each producing ∼2 pN force (17), should produce force in this range. Node velocities are low compared with the ability of S. pombe Myo2 to move unloaded actin filaments at 400 nm/s (18), so we expect that node-bound myosin-II operates near stall, consistent with our lower estimate. We assume a short-range (150 nm) repulsive radial force to prevent nodes from overlapping. We assume that nodes pause when they lose connections with neighboring nodes after mean time τbreak. Connections fail when the filament breaks or the filament dissociates from myosin-II, as expected for motors with low duty cycles (17). Filaments searching for nodes also break randomly after τturn. After breaks, a new filament grows in a new direction. We assume that tension switches off elongation of filaments anchored by Cdc12p and that a filament connecting two nodes does not accrete additional nodes (11).

The model was robust over a large range of parameter values (Fig. 4, A to D, and figs. S7 to S16). Using experimental parameter values, simulations generated rings with few gaps (Fig. 4, B and D) in the same time (Fig. 4, A and C) as live cells. Reliable avoidance of large gaps among disconnected clumps of nodes requires >50 nodes distributed in a zone <2.8 μm wide and τbreak < 80 s. Clumps tend to form in a search-and-capture mechanism, because nodes find close neighbors more efficiently than distant ones and collapse locally. Condensation of the nodes in 10 min also depended on 1 to 4 actin elongating from nodes and active formins in >50% of nodes. Nodes had time to find their neighbors within τturn and complete condensation in ∼10 min if they grew >20 subunits/s. In agreement with this, inhibiting actin polymerization with Latrunculin A or by depleting profilin results in clumps of nodes (10, 19).

Fig. 4.

Simulations of contractile ring assembly by the model in Fig. 1, D to H, using parameters in Fig. 1J unless noted otherwise. Graphs plot averages of 50 simulations ±SD. Gray regions indicate physiological parameter range. (A) Time (τconv) for the band of nodes to condense 50% versus number of nodes. (B) Largest circumferential gap between nodes 500 s after the onset of condensation versus number of nodes. (C) As (A), but as a function of the initial half-width of the broad band. (D) As (B), but as a function of the lifetime of connections. (E) Snapshots of simulations of ring formation in cdc25-22 cells (100 nodes, initial half-width 1.6 μm). (F) Simulations of attraction between two condensed rings (70 nodes each), as in (20). (Top) Rings separated by 4 μm fuse within 20 min. (Bottom) Rings separated by 10 μm fail to converge.

The search, capture, pull, and release model reproduces two classic experiments. When cdc25-22 cells are arrested in the G2 phase of the cell cycle for 4 hours at the restrictive temperature, they continue to grow in length and assemble 97 ± 17 nodes (n = 17) in a band ∼3.5 μm wide. When released into mitosis at the permissive temperature, they form a contractile ring in ∼25 min. In simulations using the standard parameters but arrays of 100 nodes in a band 3.5 μm wide, the nodes condense into somewhat irregular rings in 15 to 30 min (Fig. 4E and fig. S17). The model with standard parameters also reproduces the fusion of pairs of contractile rings (20) separated by up to 8 μm at velocities comparable to experiments (Fig. 4F and fig. S18).

The success of this random search-and-capture mechanism without feedback suggests that its inherent randomness may endow the mechanism with an advantageous plasticity, aiding navigation of the complex landscape of kinetic barriers and traps separating a broad band of nodes from a completed contractile ring. The model attempts to capture the essential components of the mechanism, but additional features such as spatial, temporal, and orientational control of actin polymerization and myosin activity cannot be excluded. Thus, the precise mechanistic role of many regulatory proteins such as tropomyosin and cofilin remains to be established. Most important, formation of actin filament bundles clearly contributes to ring assembly (Fig. 3, A and B, and Movie S19) (9, 20). Our observations suggest that transient bundles stabilize tenuous initial connections between nodes rather than leading ring assembly (9, 14). Cytokinesis by animal cells also involves formins (4) and mobile clusters of myosin-II (21), so our model may represent a conserved module used in contractile networks.

Supporting Online Material

Materials and Methods

Figs. S1 to S18

Table S1


Movies S1 to S19

References and Notes

View Abstract

Stay Connected to Science

Navigate This Article