Rapid Actin Transport During Cell Protrusion

See allHide authors and affiliations

Science  04 Apr 2003:
Vol. 300, Issue 5616, pp. 142-145
DOI: 10.1126/science.1082026


Transformed rat fibroblasts expressing two variants of green fluorescent protein, each fused to β-actin, were used to study actin dynamics during cell protrusion. The recently developed FLAP (fluorescence localization after photobleaching) method permits the tracking of one fluorophore after localized photobleaching by using the other as a colocalized reference. Here, by visualizing the ratio of bleached to total molecules, we found that actin was delivered to protruding zones of the leading edge of the cell at speeds that exceeded 5 micrometers per second. Monte Carlo modeling confirmed that this flow cannot be explained by diffusion and may involve active transport.

Understanding actin dynamics in the leading lamellae of cultured fibroblasts has long been considered the key to understanding cell locomotion. Recent studies have unraveled an intricate underlying biochemistry, including the role of the Arp2/3 complex in actin polymerization and the role of cofilin in depolymerization (1, 2). They have provided mechanistic explanations for the polymerization of actin during the protrusion of thin veil-like lamellipodia (ruffles) at the leading margin of the cell and for the constant rearward flow of polymerized actin (F-actin) at a rate of 1 to 2 μm/min away from the margin (1,3). Advances in single-molecule speckle fluorescence have allowed the lifetimes of actin molecules in filaments to be studied and have revealed that, although the rate of actin polymerization is promoted within 1 μm of the leading edge, both polymerization and depolymerization occur at a high rate during the rearward flow of F-actin (3).

In contrast, the equally crucial question of how actin monomer (G-actin) returns to the leading edge has received little attention beyond the assumption that it happens by diffusion (1). We report an extension of the fluorescence localization after photobleaching (FLAP) method that may shed light on the mechanism of this process. Highly malignant transformed rat fibroblasts of the T15 line were chosen for this study because they show very active lamellipodial protrusion but few high-order actin structures such as stress fibers (Fig. 1, B and F). The yellow fluorescent protein (YFP) target fluorophore was photobleached in a narrow strip across the leading lamella of a T15 cell and analyzed with both absolute and ratio FLAP methods (Fig. 1). Absolute FLAP images (Fig. 1, C and G) were derived from the difference between the images of the target fluorophore and the reference fluorophore, cyan fluorescent protein (CFP) (4). FLAP ratio images (Fig. 1, D and H) were calculated as 1 minus the ratio of target to reference fluorophore intensity (5). Although the absolute FLAP images give the expected number of bleach-labeled YFP-actin molecules at each pixel location, the FLAP ratio images express this number as a fraction of the total YFP-actin molecules at that location. Thus, the FLAP ratio signal enabled us to track the dispersal of the bleach-labeled actin through regions of differing actin density (Fig. 1, C and D).

Figure 1

(A to H) Eight scanning confocal images of the leading lamella of a T15 cell scanned from top to bottom immediately after a strip (white box) was photobleached for 4 s (A to D) and scanned again 20 s later (E to H). (A) and (E), CFP fluorescence images; (B) and (F), phase contrast images; (C) and (G), absolute FLAP images; (D) and (H), FLAP ratio images. (I and J) Traces of CFP- and YFP-actin (cyan and yellow, respectively) and the FLAP ratio signal (red) integrated across a narrow central band [gray lines in (D) and (H)] immediately after bleaching (I) and 20 s later (J). Vertical parallel lines indicate the bleach zone, and the horizontal axis is calibrated for both distance and time of scan progress after the end of bleaching. (K andL) Diffusion of G-actin along the y axis of the cell immediately after bleaching (K) and 20 s later (L) is modeled with a diffusion coefficient of 6 μm2/s, an F-actin/G-actin ratio of 0.1, and a half-life of F-actin of 20 s. Composite images simulate the progress of scanning. Details of laser scanning microscopy with a Zeiss LSM 510 and of the cDNA constructs used for nuclear microinjection are as given in (4,5).

In images scanned immediately after the 4-s bleach period, the central region of the bleach strip showed a rapid dispersal of the FLAP ratio signal into the surrounding lamella, whereas the lateral regions showed a much more persistent signal (Fig. 1D). We used a Monte Carlo model of diffusion during photobleaching (Fig. 1, K and L) to estimate that the bleach-labeled G-actin in the central region (Fig. 1, I and J) had a diffusion coefficient of 5.65 μm2/s (supporting online text). This is consistent with a published estimate of 5.8 μm2/s for G-actin in cytoplasm derived from fluorescence recovery after photobleaching (FRAP) measurements (6). Our best-fit model incorporated nondiffusible F-actin with an F-actin/G-actin ratio of 0.1. An F-actin half-life of less than 20 s was required to mimic the dispersal of the signal (Fig. 1, J and L), which is comparable with an estimate of 30 s for the half-life of actin filaments in the broad leading lamellipodium of Xenopus fibroblasts (3). When the same cell was bleached closer to the leading edge a few minutes later (Fig. 2), the dispersal of the FLAP ratio signal was slow (Fig. 2E) and comparable to that in the lateral regions of the bleach zone inFig. 1. This indicates that a continuous band of relatively stable actin, about 5 to 7 μm wide, extended around the peripheral zone of the leading lamella.

Figure 2

Six images of the same T15 cell as in Fig. 1, scanned from top to bottom immediately after a strip (white brackets) was photobleached for 4 s, showing (A) CFP-actin, (B) YFP-actin, (C) phase contrast, (D) absolute FLAP, (E) FLAP ratio, and (F) a diagram of protrusion (green) and retraction (red) of the cell margin constructed from two thresholded CFP images acquired immediately before and after bleaching. The CFP and YFP images are shown in gray scale for comparison with each other. White arrows in the FLAP ratio image (E) indicate regions of strongly labeled actin at the leading edge.

The most noticeable feature, however, was the sudden appearance of strongly labeled actin at discrete regions of the leading edge (Fig. 2E, arrows). The maximum time available for these YFP-actin molecules to become photobleached and then to migrate to the leading edge was 5 s. The high signal also indicates that these molecules probably polymerized at the leading edge. A diagram of protrusion and retraction of the leading margin during a 15-s interval (Fig. 2F) revealed that the discrete zones of highly labeled actin were confined to a region in which the margin was actively protruding.

To analyze whether this rapid translocation of actin to the leading edge during protrusion could be due solely to molecular diffusion, we assumed that the diffusion coefficient of G-actin was again 5.65 μm2/s, although there is much more F-actin obstructing diffusion in the frontal zone. However, the effect of obstruction is probably slight: It has been estimated that a reticular network must occlude 90 to 97% of diffusion space in order to reduce diffusion by 39 to 60% (7).

In an example of the furthest distance over which we have observed rapid translocation, highly labeled actin (Fig. 3D, arrows) reached the edge 2 s after bleaching a strip 12 μm away for 1.8 s, giving an estimated speed of translocation of around 6 μm/s. Again, the protrusion-retraction diagram (Fig. 3C, arrows) showed a good correlation between the appearance of actin at the edge and active protrusion. A diffusion model required an F-actin/G-actin ratio of 0.8 and a half-life of F-actin of 4 s in order to model the dispersion of actin within the lamella, although this model showed no elevated FLAP ratio signal at the leading edge (supporting online text). We obtained a slightly elevated signal at the leading edge (Fig. 3F), although not nearly as high as in the real cell (Fig. 3E), by incorporating protrusion into the model and by simulating the conversion of G-actin to F-actin within 0.4 μm of the edge. Only 15.9% of the YFP molecules in the whole cell were bleached, and in the model the edge signal never rose much above 20%, no matter how long it was run, indicating that molecular diffusion could not account for the observed signal of 40% at the leading edge regardless of diffusion coefficient. In order to obtain an adequate fit to observed data, we incorporated active transport of actin into the model (Fig. 3G). This was done by biasing the diffusion of a subpopulation of G-actin (6% of total G-actin) to give a forward drift velocity of 6 μm/s superimposed on diffusion (supporting online text). A further experiment showed that the labeled actin newly incorporated into the leading edge migrated rearward (Fig. 3, J and K). The fact that this actin did not disperse rapidly indicates that it was polymerized.

Figure 3

Images of two T15 cells scanned immediately after a strip (white box) was photobleached for 1.8 s (A to D) or 1.7 s (H toJ) and of the last cell scanned again 10.8 s later (K), showing CFP-actin (A and H), phase contrast (B and I), protrusion and retraction (C), and FLAP ratio (D, J, and K). White arrows indicate coincidence between regions of profusion (C) and regions of highly labled actin (D). (E) Signal traces as inFig. 1 are integrated between the gray lines in (D) immediately after bleaching. (F) Diffusion of G-actin along the y axis of the cell is modeled with a diffusion coefficient of 5.65 μm2/s, an F-actin/G-actin ratio of 0.8, and an F-actin half-life of 4 s together with protrusion and trapping of G-actin at the leading edge. (G) The final model further incorporates a forward drift at 6 μm/s of 6% of total G-actin; all other parameters are as in (F).

A biochemical analysis revealed that the rapid recruitment of actin to the leading edge was abolished by treatment with 5 μM of the actin filament–stabilizing drug jasplakinolide for 20 min before bleaching (Fig. 4, A to C). After treatment, the bleach zone showed a highly persistent FLAP ratio signal with little decay after nearly 12 min (Fig. 4C), whereas normally the signal was uniformly dispersed throughout the cell after 2 to 3 min. Depletion of adenosine triphosphate with 6 mM 2- deoxyglucose plus 1 μM antimycin for 20 to 40 min suppressed ruffling activity and also abolished the rapid actin recruitment (Fig. 4, D to F). In contrast, treatment with the myosin light chain kinase (MLCK) inhibitor ML-7 (1 μM) (Fig. 4, G to I) or with the Rho-associated kinase (ROCK) inhibitor Y-27632 (10 μM) for 1 to 2 hours (Fig. 4, J to L) did not inhibit ruffling. Even so, there was no sign of rapid actin translocation to the leading edge in either case (Fig. 4, I and L). The ML-7 treatment but not the Y-27632 treatment significantly suppressed the speed of cell motility, measured as 5-min displacements, in analysis of variance tests versus the control [control: 1.32 ± 0.08 μm (mean ± SEM), n = 41 cells; ML-7 treatment: 1.11 ± 0.07 μm, n = 42 cells, P < 0.05; Y-27632 treatment: 1.49 ± 0.07 μm, n = 41 cells, not significant] (5). Finally, treatment with 500 nM nocodazole for 30 min (Fig. 4, M to O) led to the complete disruption of cytoplasmic microtubules and, although ruffling activity was changed in character (Fig. 4N), there were indications that labeled actin was still rapidly recruited to the leading edge (Fig. 4O).

Figure 4

CFP-actin (left), phase contrast (center), and FLAP ratio images (right) under different treatments with metabolic inhibitors: (A to C) 5 μM jasplakinolide; (D to F) 6 mM 2-deoxyglucose and 1 μM antimycin; (G to I) 1 μM ML-7; (J toL) 10 μM Y-27632; and (M to O) 500 nM nocodazole. Each FLAP ratio image is marked with the time after the end of bleaching that scanning was started.

In summary, we found that actin polymerized during protrusion of the leading edge can be recruited within 2 to 3 s from a region that extends up to 15 μm behind the edge. This process is too rapid to be explained by diffusion of G-actin and must involve some form of active transport. It requires G-actin, adenosine triphosphate, and active protrusion but not intact microtubules. The effects of MLCK inhibition indicate that myosin function is required for rapid actin transport, although not for ruffling. The similar effects of ROCK inhibition further suggest that the required myosin is activated by Rho and is probably nonmuscle myosin II.

We conclude that G-actin is not carried rapidly to the leading edge by microtubule-based motors or by motor proteins of the myosin superfamily. Myosin II is unlikely to transport actin directly, but it is needed for contraction of the cell body. This could generate a pressure gradient leading to a hydrodynamic flow that carries G-actin with it. The flow would be directed toward sites of reduced pressure that may occur where the cell margin is expanding rapidly. Such rapid expansion might occur passively at sites of cortical weakness or actively where polymerizing actin is pushing the membrane outward (1). Pressure-driven flow through gels can result in channeling, and there is some evidence for this in the figures. Retraction of the cell margin may occur by Rho-dependent contraction (8), in which case the mechanism we propose could account for a reported positive correlation between rates of protrusion and retraction (9). Rapid actin transport is not required for cell locomotion or for ruffling, which is thought to be regulated by Rac (10), but it may determine sites of protrusion and thus participate in signal-mediated cell polarity and directed locomotion. Our view is that hydrodynamic flow supplies actin from sites of high depolymerization to sites of rapid protrusion where diffusion alone is inadequate.

Supporting Online Material

Materials and Methods

Supporting Text

Figs. S1 to S8

References and Notes


Stay Connected to Science

Navigate This Article