We study how mechanical forces integrate spatially and temporally with regulatory signals at the leading edge of migrating cells. To probe the dynamics of this system, we developed quantitative fluorescent speckle microscopy, which maps out actin cytoskeleton transport, assembly and disassembly with high spatial resolution. Statistical processing of single speckle properties revealed two kinetically, kinematically and molecularly distinct, yet spatially overlapping, actin arrays at the leading edge of migrating epithelial cells. The first network, referred to as the lamellipodium, polymerizes and depolymerizes 1–2 μm from the edge in an Arp2/3 (actin-related protein 2/3)- and cofilin-dependent fashion. The second network, referred to as the lamella, exhibits Arp2/3-independent polymerization. To elucidate the dynamic relationship between the two networks, we have begun to examine how assembly and flow are temporally modulated with respect to a protrusion event. In control cells we found bursts of protrusion preceding bursts of F-actin assembly. The time lag disappears in cells where Arp2/3-function is impaired. This and other results allowed us to propose a model in which tropomyosin protects lamella filaments from branching and severing, and to conjecture that Arp2/3-mediated lamellipodium assembly is a natural consequence of lamella expansion, but not the initiator of cell protrusion.
The mechanics of actin-mediated cell protrusion
Cell migration results from several force-generating machineries that are spatially and temporally co-ordinated via precisely tuned signals. Since Abercrombie , cell migration has been described as a recurring cycle of (i) leading edge protrusion; (ii) formation of new adhesion sites; (iii) disruption of older adhesion sites at the cell rear; and (iv) development of contractile forces balanced between strong adhesion bonds at the cell front and weaker bonds at the cell rear yielding net advancement of the cell body . The functionality of the actin cytoskeleton and its associated proteins is sufficient to execute this cycle. However, most cell types require also an interaction of the actin and microtubule cytoskeletons for directed movements .
As the initial event in the cycle, leading edge protrusion sets the stage for cell migration . Protrusion is generally associated with the assembly of a dendritically branched network of actin filaments (F-actin) next to the leading edge plasma membrane . Biophysical models have posited that filament elongation acts as a ratchet to rectify the thermal fluctuations of plasma membrane and actin filaments in forward direction . Although this theory still awaits direct experimental verification in the context of cell migration, in vitro studies and studies of pathogen motility have provided ample evidence that actin polymerization could produce protruding forces [7,8]. These and many other studies of actin assembly/disassembly in migrating cells have given rise to the dendritic nucleation model. The model suggests a treadmilling of actin, where network growth is regulated by a host of actin sequestering, capping, severing and depolymerizing proteins . A key player in this machinery is the Arp2/3 (actin-related protein 2/3) complex, which increases network polymerization by nucleation of new filaments at the ends and sides of existing filaments.
F-actin polymerization alone is not sufficient to explain cell protrusion. To convert the work of polymerization into leading edge advancement, the growing network needs to couple to the extracellular domain. In many cell systems, F-actin polymerization is indeed only partially converted into protrusion and partially into retrograde network flow away from the leading edge . The ratio between network growths and rearward flow is modulated by transient reconfiguration of cytoskeleton adhesion.
Cytoskeleton coupling is also required for motor-dependent contraction to be effective in pulling the cell body and tail forward. As with polymerization-induced forces, contractile forces are partially transformed into a forward pull of the cell body and partially generate retrograde F-actin flow at the cell front . Thus the spatial distribution of contraction forces defines an additional regulator of protrusion.
Finally, the assembly, adhesion and contraction machineries are tightly coupled via regulatory signals, primarily mediated by the Rho-family GTPases Rac and Rho . Thus cell protrusion is by itself a highly integrated molecular process where mechanical and chemical signals are coupled in a complex regulatory network. Whether a cell protrudes or retracts depends on the spatial and temporal distribution of F-actin network assembly, contraction and adhesion, as well as the relative timing and transport of the various signalling pathways connecting them.
To dissect the mechanism of cell protrusion, we developed qFSM (quantitative fluorescent speckle microscopy). qFSM maps F-actin turnover and contraction with high spatial and temporal resolution and allows us to correlate F-actin dynamics with cell boundary movements. Combined with systematic molecular perturbation of nodes in the regulatory network, it will thus become possible to determine the timing and coupling of the various mechanochemical components involved in cell protrusion.
qFSM of F-actin dynamics
FSM capitalizes on the method of fluorescent analogue cytochemistry, in which purified protein is covalently linked to a fluorophore and microinjected into living cells or expressed as a GFP (green fluorescent protein) fusion. Speckle formation occurs when the ratio of unlabelled to labelled protein is less than approx. 2–3% . When applied to the visualization of the actin cytoskeleton, incorporation of labelled actin monomers together with endogenous, unlabelled monomers yields a non-uniform distribution of fluorophores throughout the network of filaments. Upon imaging with a high-resolution fluorescence light microscope, areas with a high local density of fluorophores appear as diffraction-limited intensity peaks, called speckles, on a dim background (Figure 1A). Speckles act as fiduciary marks of network movement, and their intensity changes reflect local net assembly and disassembly of the network . Both relationships are defined only in a stochastic sense. To extract quantitative information about F-actin dynamics, hundreds of thousands of speckles per movie have to be computationally tracked from appearance to disappearance and then analysed in a statistical framework that relates fluctuations in speckle position and intensity to network flow and turnover .
F-actin retrograde flow and turnover in a migrating epithelial cell
We developed single-particle tracking methods to follow 105–106 speckles per movie. The movement of a speckle comprises a coherent component associated with the network flow and a random component associated with thermal fluctuations and local contraction of the network (Figure 1B). To suppress the second component, we filter and interpolate speckle trajectories with kernels whose correlation length matches the persistence length of actin filaments (Figure 1C). Importantly, such interpolated maps are generated for every time point of the movie to reveal non-steady-state flow dynamics. Figure 1(D) displays a scalar map of the vector magnitudes. Together with Figure 1(C) this analysis shows the organization of F-actin flow in four kinematically distinct zones: (i) a lamellipodium (Lp) with, on average, the highest rate of retrograde flow; (ii) a lamella (La) with slower retrograde flow; (iii) a convergence zone (CZ), including a pole (P); and (iv) a zone of slow, anterograde flow associated with the ventral actin cortex of the cell body (CB).
Figure 1(E) displays a two-channel colour overlay of F-actin turnover, produced with an algorithm by which intensity changes at every event of speckle appearance and disappearance are statistically processed to compute a kinetic score proportional to the local net exchange of monomers (i.e. monomers associated minus monomers dissociated) . The red channel presents association scores (polymerization) and the green channel presents dissociation scores (depolymerization). Areas with equally strong polymerization and depolymerization appear in yellow. We found that the four kinematically defined zones also have distinct turnover behaviours. The lamellipodium polymerizes next to the leading edge (arrowhead), juxtaposed to a band of strong disassembly at the lamellipodium–lamella junction (white arrow). Regional analysis of the scores reveals that 80–90% of F-actin polymerized at the leading edge is disassembled within the lamellipodium, suggesting that lamellipodium and lamella are essentially separate structures. Strong depolymerization is also present in the convergence zone (black arrows), suggesting a coupling between F-actin disassembly and contraction . The lamella exhibits a punctate pattern of polymerization and depolymerization loci, corresponding in time-resolved maps to oscillatory network assembly and disassembly.
Two kinetically, kinematically and molecularly distinct, yet spatially overlapping, F-actin arrays mediate cell protrusion
We investigated the co-ordination of lamellipodium and lamella at the level of single speckles. Given the fast retrograde flow and narrow width of the lamellipodium as compared with the flow and size of the lamella, we speculated that speckles with a short lifetime and fast movement (class 1) would preferentially localize in the lamellipodium, while speckles with a long lifetime and slow movement (class 2) would spread over the lamella. We clustered speckles accordingly and found indeed that class 1 speckles were mainly localized in the lamellipodium and class 2 speckles were localized in the lamella. Yet, 33% of the speckles falling into the area covered by the lamellipodium belonged to class 2, suggesting that speckles with different kinematic and kinetic behaviour can spatially coexist . We then mapped the assembly and disassembly separately for both classes and compared it with the turnover computed from all speckles. The analysis revealed that the treadmilling F-actin network subadjacent to the leading edge plasma membrane (cf. Figure 1E, red–green band parallel to the leading edge) is formed by class 1 speckles only, while class 2 speckles represent a network with punctate patterns of assembly and disassembly, characteristic of the lamella (Figure 1E). The co-localization of two speckle classes with distinct flow velocities implies that the two networks are materially only transiently integrated.
The networks also have distinct molecular properties. Low doses of the barbed end capping drug cytochalasin D and filament stabilizing drug jasplakinolide altered the F-actin/monomer balance of the treadmill, selectively eliminating the lamellipodium network. By contrast, the lamella was sensitive to blebbistatin, which blocks retrograde flow by inhibition of myosin II activity, while neither flow nor turnover of the lamellipodium network changed. Consistent with these pharmacological studies, we measured high concentrations of Arp2/3 and cofilin in the lamellipodium, and much lower ones in the lamella, while myosin II was spread over the lamella but excluded from the lamellipodium. Also, time correlation analysis between the formation of Arp2/3 clusters and network assembly underlined that polymerization in the lamellipodium is mediated by Arp2/3 whereas lamella assembly is independent of Arp2/3. Together, these analyses suggest an organization of the actin cytoskeleton in two networks where the contractile lamella array reaches all the way to the leading edge and is superimposed by a narrow lamellipodium undergoing fast treadmilling (Figure 2).
Model of the cell protrusion mediated by Arp2/3-independent lamella expansion
Persistent cell protrusion is initiated by lamella expansion
Analysis of the timing of events at the leading edge indicated that protrusion is synchronized with transiently reduced retrograde flow, while maximal assembly lags by 15–20 s. The specific dependence of lamellipodium formation on Arp2/3 permitted selective removal of this network by Arp2/3 inhibition . Cells without a lamellipodium protruded as efficiently as control cells, but the time lag between protrusion and assembly vanished. We concluded that in the absence of a lamellipodium, cells protrude because of forminmediated lamella expansion. One of the methods to inhibit Arp2/3 function was by microinjection of skeletal muscle tropomyosin that decorated actin filaments and thus prevented their branching. Alternatively, lamellipodium formation was impaired by expression of the CA (cofilin homology and acidic) domain of N-WASP (neuronal Wiskott–Aldrich syndrome protein). Analysis of the ratio between tropomyosin and F-actin in control cells indicated that approx. 30–35% of actin filaments at the leading edge were decorated by tropomyosin, in good agreement with the relative population of speckles assigned to the lamella network by cluster analysis (see above).
Together, these results suggest the following mechanism for cell protrusion (Figure 2). (i) Two mechanically and materially unintegrated networks are superimposed at the leading edge. (ii) Extracellular or intracellular signals initiate lamella expansion resulting in edge advancement. Some lamella filaments escape the decoration by tropomyosin, giving way to Arp2/3-mediated branching of filaments so that a new lamellipodium assembles in an autocatalytic process. Our measurements suggest that lamellipodium formation takes 15–20 s, explaining the delay in maximal assembly relative to the protrusion event. (iii) Lamellipodia and lamella networks decouple gradually due to the severing activity of cofilin. Consequently, continued assembly of the lamellipodium against the cell wall is increasingly transformed into contraction-independent retrograde flow. At the same time, the force increase that accompanies lamellipodium assembly induces the maturation of focal contacts .
The controversial issue of this model is that persistent cell protrusion requires mainly Rho instead of Rac activity at the leading edge. On the other hand, many extracellular signals have biochemically been identified to converge in the Rac pathway. How these signals are directly or indirectly targeted to the Rho pathway remains to be answered in order to consolidate the proposed model.
Cell Architecture: from Structure to Function: A Focus Topic at BioScience2005, held at SECC Glasgow, U.K., 17–21 July 2005. Edited by S. Cockroft (University College London, U.K.), Y. Goda (University College London, U.K.), R. Insall (Birmingham, U.K.) and M. Wakelam (Birmingham, U.K.).
This work has been accomplished in close collaboration with Dr Clare Waterman-Storer at The Scripps Research Institute and is funded by a joint grant from the National Institutes of Health (GM67230).