Two Distinct Actin Networks Drive the Protrusion of Migrating Cells

See allHide authors and affiliations

Science  17 Sep 2004:
Vol. 305, Issue 5691, pp. 1782-1786
DOI: 10.1126/science.1100533


Cell migration initiates by extension of the actin cytoskeleton at the leading edge. Computational analysis of fluorescent speckle microscopy movies of migrating epithelial cells revealed this process is mediated by two spatially colocalized but kinematically, kinetically, molecularly, and functionally distinct actin networks. A lamellipodium network assembled at the leading edge but completely disassembled within 1 to 3 micrometers. It was weakly coupled to the rest of the cytoskeleton and promoted the random protrusion and retraction of the leading edge. Productive cell advance was a function of the second colocalized network, the lamella, where actomyosin contraction was integrated with substrate adhesion.

Cell migration involves a coordinated cycle of plasma membrane protrusion at the leading edge, adhesion site formation under the protrusion, disruption of older adhesion sites at the cell rear, and cytoskeleton contraction against adhesions to yield cell body movement (1). Protrusion is thought to result from actin filament (F-actin) polymerization against the plasma membrane (2), with the polymerization rate regulated by the rate of monomer addition to the fast-growing (“barbed”) ends of filaments. This may depend on actin-related protein 2/3 (Arp2/3) complex activation, which creates free barbed ends by branching and de novo nucleation of filaments (dendritic nucleation) (3), and on actin depolymerizing factor (ADF) cofilin, which creates free barbed ends by severing preexisting filaments and promoting depolymerization of free filament “pointed” ends (4). Filament growth is limited by barbed end–capping proteins and depletion of the polymerization-competent pool of actin monomers (5).

We exploited quantitative fluorescent speckle microscopy (qFSM) (6) to study the dynamic organization of F-actin in migrating cells and define its relationship to cell protrusive behavior. In FSM, a nonuniform pattern of fluorophores forms by co-polymerization of F-actin from a pool of actin monomers with a very low ratio between fluorescently labeled and unlabeled monomers. In high-resolution images, fluorophore clusters are detected as local intensity maxima called speckles (fig. S1). We imaged speckles containing 2 to 8 fluorescent monomers. Computational tracking of >105 speckles per FSM movie (7) and statistical processing of their intensity fluctuations (8) allowed us to map F-actin flow (kinematics) and turnover (kinetics) with submicrometer resolution (9).

We quantified F-actin dynamics in newt lung epithelial cells (Fig. 1, A and B, and movie S1) and potoroo kidney (PtK1) epithelial cells (Fig. 1, C and D, and movie S2). In kinematic maps, both cell types exhibited a thin (1 to 3 μm) band along the leading edge where speckles underwent fast flow (300 to 500 nm/min) toward the cell center (Fig. 1, A and C, and movies S3 and S4). The band abutted a much larger area of slower retrograde flow (100 to 250 nm/min). Steady-state maps of F-actin kinetics showed a ∼1-μm-wide band of strong polymerization along the leading edge that transitions into an equally narrow band of depolymerization (Fig. 1, B and D), which corresponded in kinematic maps to the region of negative speed gradients. Animated kinetic maps (movies S5 and S6) revealed that bursts of depolymerization also occurred at the leading edge. Beyond these bands, kinetic maps displayed a random pattern of foci 1 to 2 μm in diameter, where F-actin assembled and disassembled in cycles of ∼40 to 50 s. We refer to the zone with fast retrograde flow and juxtaposed narrow bands of polymerization and depolymerization as the lamellipodium and the zone characterized by slower flow and a punctate pattern of filament turnover as the lamella. The lamellipodium and lamella had distinct molecular signatures, with high concentrations of Arp2/3 and ADF/cofilin in the lamellipodium and myosin II and tropomyosin present exclusively in the lamella (fig. S2).

Fig. 1.

F-actin dynamics characterized by qFSM in (A and B) a newt lung cell (movies S1, S3, and S5) and (C and D) a PtK1 cell (movies S2, S4, and S6). [(A) and (C)] Time-averaged F-actin flow (60 frames); colors encode flow speed, and vectors, flow direction. Inset: Individual speckle trajectories over 200 s; time evolution, dark blue to light green. [(B) and (D)] Time-averaged F-actin turnover (60 frames). Red channel, assembly rate; green, disassembly rate. (E) Net F-actin assembly. Black, negative assembly rates; white, positive assembly rates. The lamellipodium-to-lamella transition (white line) (movie S8) was computed from rate profiles (along the red lines) of net assembly and flow. (F) Rate profiles indicate that the lamellipodium-to-lamella transition coincides with the first local maximum of net assembly (arrowhead) after the first maximum in depolymerization (dashed green line). A, integrated assembly; D, integrated disassembly. (G) The kinetically defined transition (black asterisks) colocalizes with the first local minimum in flow rate (blue asterisks). Examples are calculated from the yellow profiles (1) and (2) in (E).

Given the kinematic, kinetic, and molecular differences, we sought mathematical criteria that would allow us to determine the relationship between the lamellipodium and lamella over time. We tracked the cell border and calculated the rates of network assembly, disassembly, and retrograde flow along regularly spaced profiles perpendicular to the edge (Fig. 1E). In 80 profiles analyzed, the assembly rate peaked at ∼1 μm from the leading edge, followed closely by a peak in depolymerization (Fig. 1F). Before the variation in both rates diminished, a second peak of net assembly (Fig. 1F, arrowhead) indicated the location where lamellipodium kinetic behavior turned into lamella kinetic behavior. This kinetic boundary colocalized with the first local minimum of the retrograde flow speed (Fig. 1G). Together, these criteria permitted robust automatic extraction of the lamellipodium-to-lamella transition (Fig. 1E, white line).

Integration of the net assembly rate (the difference between areas A and D in Fig. 1F) revealed the amount of polymer generated and disassembled within the boundary of the lamellipodium. Most (85 to 90%) of the actin network underwent a complete cycle of assembly and disassembly within the first 1 to 3 μm from the cell edge. Thus, lamellipodium and lamella are composed of materially nearly decoupled networks, suggesting that the lamellipodium does not supply substantial filaments to other actin assemblies in the cell (10). The tight spatial juxtaposition of assembly and disassembly likely generates monomer gradients high enough to maintain an efficient treadmill of actin turnover based on diffusive transport (1113).

Next, we partitioned speckles into two classes (Fig. 2A): fast-moving and short-living (class 1) and slow-moving and long-living (class 2). Class 1 speckles clustered in the lamellipodium, whereas 90% of the speckles in the lamella belonged to class 2 (Fig. 2B). However, 33% of the speckles falling into the area between the leading edge and the lamellipodium-lamella transition had the characteristics of lamella speckles (Fig. 2B, inset). We examined the ensemble behavior of these speckle classes and found that class 1 speckles displayed the kinetic signature of the lamellipodium (Fig. 2C), whereas class 2 speckles displayed the signature of a lamella network (Fig. 2D) that extended all the way to the leading edge. This coexistence of two spatially overlapping yet kinetically and kinematically distinct networks was masked in joint evaluation of all speckles (Fig. 1, B and D), because class 1 speckles outnumbered class 2 speckles near the cell boundary.

Fig. 2.

Distinction of spatially overlapping lamellipodium and lamella by classification of speckle velocity and lifetime. (A) Separation of fast-moving and short-living speckles (class 1, red) and slow-moving and long-living speckles (class 2, black). LTth, lifetime threshold; Vth, velocity threshold. (B) Thresholds are set to minimize the number of class 1 speckles in the lamella while maximizing the number in the lamellipodium. This multi-objective optimization has a unique solution, because class 1 speckles cluster in the lamellipodium with a residual occurrence of ∼10% in the lamella. Class 2 speckles dominate the lamella but expand all the way to the leading edge, contributing 33% to the lamellipodium region. (C) Turnover calculated from class 1 speckles shows the spatial kinetic signature of the lamellipodium. (D) Turnover calculated from class 2 speckles shows the spatial kinetic signature of the lamella.

We probed the molecular characteristics of the two networks in a series of pharmacological perturbation experiments. First, we applied blebbistatin to inhibit nonmuscle myosin II adenosine triphosphatase activity (14). In agreement with myosin II localization data (fig. S2), blebbistatin reduced lamella retrograde flow but not lamellipodium flow (Fig. 3A), with little effect on the rates of filament turnover in either network.

Fig. 3.

Probing molecular characteristics of the two networks by small molecule inhibitors. (A) Perfusion with 100 μM blebbistatin slows lamella F-actin flow exclusively (left two panels), with little effect on F-actin turnover (right two panels). P, polymerization; D, depolymerization. (B) Perfusion with 0.5 μM cytD decreases F-actin flow in the lamellipodium and induces retraction of the leading edge (white lines) (fig. S3). Washout rescues cell edge position and lamellipodium flow rate. (C) Space-time diagrams of F-actin network turnover, edge displacement, and F-actin flow along the leading edge (fig. S3) (9) during perfusion and washout of cytD. White dashed lines highlight periodicities. (D) Correlation between edge displacement (E) and F-actin flow (F); between edge displacement and F-actin turnover (T); and between flow and turnover. (E) Perfusion with 1 μM jasplakinolide selectively removes the kinetic and kinematic signature of the lamellipodium. The white line indicates the lamellipodium-to-lamella transition before drug application.

We next treated cells with 0.5 μM cytochalasin D (cytD) (Fig. 3B) to inhibit polymerization of free barbed ends (15). Space-time diagrams of F-actin turnover, edge displacement, and network flow computed along the leading edge (Fig. 3C) (9) revealed that, before cytD perfusion, periodic bursts of polymerization at the leading edge transformed in equal parts into protrusion and F-actin retrograde flow (Fig. 3C, dashed white lines). Cross-correlation analysis of the three parameters (Fig. 3D) confirmed a positive correlation between edge displacement and network turnover and negative correlations for the two other parameter combinations. This behavior changed after cytD perfusion. The correlation between turnover and both edge displacement and network flow broke down, and the correlation between edge displacement and network flow switched from negative to positive values (Fig. 3D). Thus, cytD selectively removed the lamellipodium network by capping the free barbed ends, and the synchronous retraction of the leading edge and network (movie S7) is associated with lamella contraction (16). Despite cytD treatment, the lamella was in a state of predominant disassembly for only ∼6 min (disassembly:assembly ratio of 1.8 ± 0.2). Subsequently, periodic patterns of assembly and disassembly were re-established (disassembly:assembly ratio of 1.2 ± 0.1 in cytD versus 1.4 ± 0.1 before perfusion), indicating that much of the kinetic activity in the lamella network is insensitive to cytD. However, the maintenance of the lamella position and its advancement required barbed end assembly. Washout of cytD immediately re-established the leading edge position and lamellipodial kinetic and kinematic signatures.

Similarly, inhibiting depolymerization by stabilizing filaments with 1 μM jasplakinolide (11) also affected the lamellipodium selectively. Within minutes of jasplakinolide application, both filament turnover and retrograde flow were stopped in the lamellipodium, whereas the lamella appeared unaffected (Fig. 3E). The loss of the lamellipodium did not alter edge position, and the same region now showed the signatures of the lamella, providing additional evidence for two distinct yet spatially overlapping networks.

Which of the two networks drives cell protrusion? We compared leading edge displacement, network turnover, and retrograde flow in movies of cells that displayed very slow net advancement to those with persistent rapid protrusion. For cells with slow edge advancement, without exception, we found periodic cycles of edge protrusion and retraction (17). They correlated with cycles of polymerization and depolymerization and of faster and slower retrograde flow. F-actin flow in the lamella (measured proximal to the lamellipodium-lamella transition) exhibited a periodicity highly correlated with the lamellipodium flow (Fig. 4A), but the amplitude modulation was damped by 69% (fig. S4), confirming the rather weak material coupling between the two networks.

Fig. 4.

The contribution of lamellipodium and lamella to edge protrusion. (A) Space-time diagrams of F-actin turnover, edge displacement (displ.), and F-actin flow along the leading edge and lamella F-actin flow adjoining the lamellipodium-to-lamella transition (fig. S4). Cycles of assembly and disassembly at the leading edge (period, 0.01 Hz) are transformed in equal parts into oscillatory edge movement and retrograde flow (white dashed lines). The same period is found in lamella flow (there is a high correlation between flow at the leading edge and lamella flow, F versus LF), but the amplitudes are markedly decreased (cf. fig. S4). (B and C) Co-movement of the leading edge and the lamellipodium-to-lamella transition during persistent cell protrusion [the red arrow in (B) and arrows in (C)] (movie S9). (D) Overlay of GFP-vinculin (white), marking focal adhesions on the actin turnover map and showing protrusions of the leading edge (arrowheads) and the position of the lamellipodium-lamella transition (white line).

Next, we examined cells with rapid, persistent edge protrusion (Fig. 4B and movie S2). One possibility was that the balance between forward edge movement and lamellipodium retrograde flow could be tipped toward edge movement. However, when polymerization increased (movie S6), it was still accompanied by acceleration of retrograde flow (movie S4). Alternatively, propulsion of the leading edge by increased polymerization could be accompanied by a widening of the lamellipodium. Instead, we found that the cell boundary and lamellipodium-to-lamella transition advanced in concert (Fig. 4C), creating the impression of a lamellipodium surfing on a forward-growing lamella (movies S8 and S9). Thus, persistent protrusion depends on the local expansion of the lamella network. Lamellipodium polymerization is not sufficient.

To determine if the expansion of the lamella spatially correlated with the establishment of linkages between the cytoskeleton and the extracellular matrix, we filmed actin dynamics by FSM and localized focal adhesions using green fluorescent protein (GFP)-tagged vinculin. Protrusive edge sections (Fig. 4D, arrowheads) were consistently positioned just beyond the distalmost portion of focal adhesions where the adhesions initiated at the lamellipodium-to-lamella transition.

Thus, qFSM of F-actin dynamics in migrating cells reveals two kinetically, kinematically, and molecularly distinct F-actin networks at the leading edge of epithelial cells, which spatially overlap and yet are only weakly coupled, and whose transition may be defined by the initiation of substrate-cytoskeleton linkages (fig. S5). Productive advancement of the leading edge requires expansion of the lamella, where actomyosin contractile forces are coupled to substrate adhesion and which depends on cytD-insensitive filament assembly, possibly mediated by formins (18, 19). In contrast, lamellipodium protrusion and retraction probably serve an exploratory function or could provide rapid responses to extracellular cues. However, persistent advancement of the cell relies on the underlying lamella.

Supporting Online Material

Materials and Methods

Figs. S1 to S5

Movies S1 to S9

References and Notes

Stay Connected to Science

Navigate This Article