Diffusion, capture and recycling of SCAR/WAVE and Arp2/3 complexes observed in cells by single-molecule imaging

Cells use actin nucleators such as the Arp2 and Arp3 (Arp2/3) complex to control the spatial and temporal dynamics of actin assembly. Activation of the Arp2/3 complex requires its association with actin filaments as well as nucleation-promoting factors such as the SCAR/WAVE complex (Mullins et al., 1998; Machesky and Insall, 1998; Machesky et al., 1999; Rohatgi et al., 1999). Similarly, the SCAR/WAVE complex requires multiple upstream activators such as plasma-membrane-associated phosphoinositides and Rac (Miki et al., 1998; Steffen et al., 2004; Oikawa et al., 2004; Lebensohn and Kirschner, 2009; Chen et al., 2010). Proteins in the SCAR/WAVE family are essential for actin polymerization and movement in many different cell contexts (Bear et al., 1998; Miki et al., 1998; Zallen et al., 2002; Yan et al., 2003; Rogers et al., 2003; Weiner et al., 2006; Ibarra et al., 2006; Patel et al., 2008).

The Arp2/3 complex and other actin regulators associate with the actin network, which is pushed away from the leading edge of a cell in a process called retrograde flow. Retrograde flow was first visualized with techniques such as FRAP and photoactivation (Wang, 1985; Theriot and Mitchison, 1991; Theriot and Mitchison, 1992; Lin and Forscher, 1995; Cramer and Mitchison, 1997). Because data from these techniques can hide local differences in actin assembly, researchers have recently used additional methods such as single-molecule and speckle microscopy to analyze the fine-grained spatial and temporal dynamics of actin rearrangements (Watanabe and Mitchison, 2002; Ponti et al., 2004). Single-molecule imaging has also been used to study actin-associated proteins such as the Arp2/3 complex and capping protein, which have different lifetimes in the actin network (Ponti et al., 2005; Miyoshi et al., 2006; Iwasa and Mullins, 2007). However, single-molecule imaging has not been applied to nucleation-promoting factors such as the SCAR/WAVE complex, and it is unknown whether the SCAR/WAVE complex incorporates into the actin network.

Imaging in other non-single-molecule modes shows that the SCAR/WAVE complex is recruited to the edge of the lamellipodium in fibroblasts (Stradal et al., 2001; Lai et al., 2008) and propagates in ‘waves’ of actin assembly at the leading edge of neutrophils and Dictyostelium (Weiner et al., 2007; King et al., 2010; Xiong et al., 2010). However, these imaging modes might obscure molecules whose dynamics differ from the global population. To elucidate mechanisms of recruitment of the SCAR/WAVE complex to the plasma membrane, convergence with the Arp2/3 complex on a membrane-apposed actin filament, and removal from the membrane, we analyzed the single-molecule dynamics of WAVE2 in Xenopus tissue culture (XTC) cells. In addition to its peripheral association with the leading edge, we show that WAVE2 molecules incorporate into the growing lamellipodial actin network. WAVE2 undergoes retrograde flow at similar speeds, sites of initiation, and lifetimes to that of actin and the p40 subunit of the Arp2/3 complex. Using a drug cocktail that stabilizes the existing cytoskeleton while blocking new assembly, we demonstrate that ongoing actin polymerization is not required for WAVE2 association and dissociation with the membrane, but is required for removal of WAVE2 from the cortex. Finally, we show that p40 and WAVE2 laterally diffuse in the membrane and capture the transition of p40 from lateral diffusion to network incorporation. Based on these data, we propose that the SCAR/WAVE and Arp2/3 complexes locally search the membrane before converging on sites of actin nucleation, and are removed from the actin network through the force of retrograde flow.

We used a crippled CMV promoter to express the low concentration of WAVE2–GFP that is required for single-molecule imaging. Because XTC cells have flat protrusions, single molecules can be visualized with epifluorescence microscopy. This enabled us to image thicker sections than can be achieved with TIRF and with less photodamaging light than needed for confocal imaging. Long camera exposures enabled us to view stabilized fluorescent probes attached to the membrane or cytoskeleton while blurring fast diffusing molecules (Watanabe and Mitchison, 2002). Under these imaging conditions, we observed WAVE2–EGFP molecules in the lamellipodium, filopodia, and regions near the lamellipodium interior (Fig. 1A, left), which is consistent with the known overall distribution of WAVE2 in non-single-molecule imaging mode (Hahne et al., 2001; Stradal et al., 2001; Lai et al., 2008). Surprisingly, we observed persistent movement of WAVE2 molecules away from the leading edge (Fig. 1 and supplementary material Movie 1). Kymograph analysis revealed that retrograde motion of WAVE2 was smooth and continuous (Fig. 1A, left, inset). WAVE2 retrograde movement can be visualized with a maximum intensity projection over the course of the epifluorescence acquisition (Fig. 1A, middle). Here, retrograde motion appeared as linear streaks, as indicated by the arrows. WAVE2–EGFP molecules with retrograde motion had a unimodal distribution of intensities that was similar to p40–EGFP and GFP–actin single molecules and photobleach in a single step (supplementary material Fig. S1). Therefore, these retrograde movement events are likely to represent single molecules. Retrograde flow of WAVE2 was more difficult to observe with shorter exposures in TIRF microscopy (supplementary material Fig. S2A), because transient recruitment of WAVE2 obfuscates stabilized pools of WAVE2 on the plasma membrane and/or cytoskeleton. We also observed retrograde movement with another subunit of the WAVE complex, Abi–EGFP (supplementary material Fig. S2B and Movie 2), supporting the idea that retrograde movement of WAVE2 and Abi reflect the movement of the SCAR/WAVE complex as a whole. Most of our experiments were performed on polylysine, but we also observed WAVE2 retrograde motion on more physiological substrates such as fibronectin, which reduced the retrograde flow speed of WAVE2 (supplementary material Fig. S2C,D). Under these conditions, integrin coupling to fibronectin probably slows actin retrograde flow velocity, which is consistent with earlier reports (Renkawitz et al., 2009).

Cells spatially bias actin polymerization to control shape and movement. The appearance of both actin and p40 is strongly biased toward the leading edge (Watanabe and Mitchison, 2002; Miyoshi et al., 2006). Similarly, the initial position of appearance for retrograde moving molecules of WAVE2 was biased to within 1 μm of the leading edge (Fig. 1A, right and Fig. 1B), suggesting that most actin assembly mediated by the SCAR/WAVE complex occurs in this region. Molecules of WAVE2 that had undergone retrograde flow had a mean speed of 31 nm/second (Fig. 1C). The distribution of WAVE2 lifetimes was well fit by a single exponential curve with a mean lifetime (τ) of 14.9±2.1 seconds (Fig. 1D and inset). We could not measure speed and lifetime for WAVE2 molecules that stayed on the cell edge because individual molecules were not resolvable. Our single-molecule lifetime measurements were roughly similar to lifetimes measured with population techniques (Fig. 4F) (Weiner et al., 2007; Lai et al., 2008). We observed no spatial bias in the lifetime of WAVE2 molecules as a function of the distance from the leading edge where they first appeared (Fig. 1E), which was similar to a lifetime versus position plot for Arp2/3 (Miyoshi et al., 2006).

Because the site of appearance and speed of retrograde flow for WAVE2 molecules are similar to actin retrograde flow, we hypothesized that the SCAR/WAVE complex might be hitchhiking with the Arp2/3 complex on the retrograde-flowing network. To gain greater insight into the relative dynamics of the SCAR/WAVE complex, the Arp2/3 complex, and actin, we compared their rates of retrograde flow and lifetimes in the actin network. Observation of the GFP-tagged p40 subunit of the Arp2/3 complex (Fig. 2A and supplementary material Movie 3) demonstrated that p40 had a remarkably similar mean retrograde flow speed of 30 nm/second (Fig. 2C) to that of WAVE2 (31 nm/second, Fig. 1C). Similarly, the lifetime of p40 during retrograde flow was 15.2±0.5 seconds (Fig. 2E, compared with 14.9±2.1 seconds for WAVE2, Fig. 1D) and did not depend on where molecules initially appeared (supplementary material Fig. S2D). When we observed cells expressing low amounts of EGFP–actin (Fig. 2B and supplementary material Movie 4), we also found that actin had a similar mean retrograde speed of 25 nm/second (Fig. 2D) and a mean lifetime in the network of 18.4±0.5 seconds (Fig. 2F). This is comparable to previous reported values of actin retrograde flow in XTC cells (Watanabe and Mitchison, 2002) but is greater than the mean speed of actin retrograde flow reported in microinjected PtK1 cells using speckle microscopy (Ponti et al., 2004). Similar to WAVE2, the appearance of p40 and actin molecules was heavily biased to within 1 μm of the leading edge (Fig. 2G,H). The distribution of p40 might be more restricted than either actin or WAVE2 for several reasons: other Arp2/3 complex activators such as WASP or N-WASP in combination with WAVE2 might further bias Arp2/3 to the cell periphery, or undersampling of WAVE2 particles at the periphery might occur because WAVE2 accumulation in that region masks the initial appearance of WAVE2 particles with retrograde motion. These data support a model where SCAR/WAVE and Arp2/3 complexes both incorporate into the actin network, maintain their interaction during retrograde flow, and are dissociated from the network together.

In cells expressing both p40–EGFP and EGFP–actin, all observable molecules underwent retrograde flow. Therefore, a large number of tracks could be collected with few cells. By contrast, WAVE2–EGFP retrograde flow was a comparatively rare event (compare Fig. 2A,B middle panels to Fig. 1A middle panel and quantified in Fig. 1F). We hypothesize that this is because only a subset of SCAR/WAVE complexes participate in Arp2/3 complex recruitment and actin nucleation. This is consistent with the observation that the SCAR/WAVE complex shows a much stronger peripheral association with the lamellipodium than does the Arp2/3 complex.

Previously, actin polymer has been shown to be involved in the recycling of the WAVE complex (Weiner et al., 2007) and other nucleation-promoting factors (Weisswange et al., 2009). Furthermore, actin polymer helps to sculpt the spatial distribution of the WAVE complex in response to external stimuli (Millius et al., 2009). However, these experiments relied on treating cells with single pharmacological inhibitors of actin assembly such as the actin-sequestering drug latrunculin. These drugs not only inhibit new polymerisation, but also result in the disassembly of the existing actin network. These dual effects make it difficult to determine whether basal actin polymer or ongoing actin assembly is more relevant for SCAR/WAVE complex recycling. To address this question, we used a three-drug cocktail to stabilize the existing network while preventing new actin polymerization (Peng et al., 2011).

This cocktail involves incubating cells in a mixture of 10 μM Rho kinase inhibitor Y27632 (Y) for 30 minutes to inhibit myosin-II-based actin contraction, then simultaneously adding 20 μM latrunculin (L) to prevent new actin assembly and 20 μM jasplakinolide (J) to stabilize existing actin structures (Fig. 3A). We verified that this JLY drug combination blocks actin nucleation (visualized by retrograde flow of p40–EGFP) and disassembly of the actin network (visualized by the lifetime of p40–EGFP in the actin network). Untreated cells expressing p40–EGFP had significant retrograde flow (trajectories shown in Fig. 3B, left), which is apparent by long streaks in a maximum intensity projection (Fig. 3B, middle) and diagonal lines in the kymograph (Fig. 3B, right). By contrast, p40–EGFP molecules in cells treated with the JLY cocktail remained stably associated with the network; note the lack of long trajectories in drug-treated cells (Fig. 3C, left) and fixed points in the maximum intensity projection (Fig. 3C, middle). The lack of retrograde movement was also apparent as vertical instead of diagonal streaks in the kymograph (Fig. 3C, right). Analysis of trajectories confirmed that retrograde speed decreased (Fig. 3D) and lifetime increased (Fig. 3E) with drug treatment. To test whether our single-molecule measurements reflected the global behavior of the actin network, we examined the morphology and bulk p40–EGFP fluorescence of a cell before and after JLY drug application (Fig. 3F and supplementary material Movie 5). Before JLY drug treatment, cell protrusion and retraction led to a difference of ∼8% in the overlapping cell area. By contrast, after drug treatment, the overlapping area was nearly identical (only 1% difference) because the cell had ceased to protrude and retract. Fluorescence of p40–EGFP also became stabilized after drug addition (Fig. 3F, bottom). We examined p40–EGFP fluorescence recovery after photobleaching (FRAP) to show that new p40–EGFP incorporation in the actin network did not occur (Fig. 4E). Together, these data indicate that the JLY drug cocktail inhibits new actin assembly while stabilizing existing actin structures.

To examine how actin regulators depend on ongoing actin polymerization and depolymerization, we analyzed the effect of JLY in cells expressing low levels of p40–EGFP or WAVE2–EGFP, imaged with TIRF microscopy. To visualize the results of this experiment, we created a kymograph across the whole cell using the ‘Reslice’ function in ImageJ, which shows how molecules change before and after drug treatment. Then we applied a maximum intensity projection (‘Z Project’ function in ImageJ) across all kymograph slices to visualize every region of the cell in one frame. After JLY drug treatment, Arp2/3

complexes became stabilized (Fig. 4A, long vertical streaks; arrow indicates drug addition). Next, we examined WAVE2–EGFP dynamics after acute inhibition of actin polymerization and depolymerization (Fig. 4B,C). Before drug treatment, most WAVE2 molecules transiently associated with the cell membrane and/or cytoskeleton (Fig. 4B, punctate spots above arrow). After actin dynamics cease, a few WAVE2 molecules remained associated with the membrane and/or cytoskeleton (Fig. 4B, long streaks of WAVE2 below arrow in the kymograph; WAVE2 lifetime is quantified in Fig. 4C). To test for a general perturbation of plasma membrane interactions, we assayed the dynamics of the association of Rac2–GFP with the plasma membrane. Rac2 membrane association was equivalent in the presence and absence of JLY (Fig. 4D) and resembled WAVE2 interaction with the membrane before drug treatment (Fig. 4B, above arrow). These data suggest that a subset of SCAR/WAVE molecules (probably those associated with the cytoskeleton) require ongoing actin polymerization for recycling, which is consistent with the observed WAVE2 retrograde flow. To test whether this actin dependence was similar for all WAVE2 molecules, we assayed the global pool of WAVE2 for actin-dependent recycling using FRAP. Here, WAVE2–EGFP molecules recovered similarly in both untreated and JLY-treated experiments (Fig. 4F) in contrast to p40–EGFP (Fig. 4E), which suggests that most WAVE2 molecules freely associate and dissociate with the plasma membrane and do not require an actin-nucleation event for membrane dissociation. Together, our data suggest that most WAVE2 molecules interact with the membrane in an actin-independent manner, but once WAVE2 becomes integrated into the actin network (presumably through Arp2/3 activation), actin-dependent retrograde flow is required for WAVE2 turnover.

We next sought to address how SCAR/WAVE and Arp2/3 complexes arrive at sites of actin nucleation. Arp2/3 complex activation involves its association with nucleation-promoting factors (such as the SCAR/WAVE complex) as well as an actin filament (Mullins et al., 1998; Machesky et al., 1999). Furthermore, the SCAR/WAVE complex is recruited and activated at membranes by association with phospholipids and GTP-bound Rac (Lebensohn and Kirschner, 2009; Padrick and Rosen, 2010). How do all of these components converge to one point in space and time for actin nucleation? We envisioned two possibilities: (1) SCAR/WAVE and Arp2/3 complexes are recruited directly from the cytosol to bind a membrane-associated actin filament, or (2) SCAR/WAVE and Arp2/3 complexes diffuse in the plane of the membrane in search of an actin filament. Consistent with the second hypothesis, when we imaged cells expressing low amounts of WAVE2–EGFP in TIRF microscopy at high-temporal resolution, we observed laterally diffusing molecules (supplementary material Movie 6).

We hypothesized that lateral diffusion of WAVE2 reflects its association with membrane-bound components such as Rac-GTP or phosphoinositides. We used mean-squared displacement (MSD) versus time-lag plots to analyze the diffusion of WAVE2 molecules (a representative plot is shown in Fig. 5A). A histogram of lateral diffusion coefficients gave a mean diffusion constant for WAVE2 of 0.41±0.04 μm²/second (Fig. 5B). This is similar to previous reported values for lateral diffusion of Ras–GFP in the membrane (Murakoshi et al., 2004), and is similar to our own analysis of Rac2–GFP lateral diffusion (supplementary material Fig. S3A,B and Movie 7). Lateral diffusion of WAVE2 was not prevented by the presence of dominant-negative RacN17 (supplementary material Fig. S3C). This is consistent with previous data that the SCAR/WAVE complex has multiple modes of binding to the plasma membrane (Miki et al., 1998; Steffen et al., 2004; Oikawa et al., 2004;Lebensohn and Kirschner, 2009; Chen et al., 2010). We could measure a slight, but statistically significant increase in the mean lateral diffusion coefficient of WAVE2 in the presence of RacN17 (0.54±0.02 μm²/second, supplementary material Fig. S3C compared with 0.41±0.04 μm²/second in control cells, Fig. 5B), which suggests that Rac-GTP might limit the diffusion of the SCAR/WAVE complex through direct binding or through mediating the interaction of the SCAR/WAVE complex with downstream signaling events such as association with the actin cytoskeleton. In some cases, we observed WAVE2 transition from lateral diffusion to immobilization (a representative cell is shown in Fig. 5D, and supplementary material Movie 8), possibly representing a precursor to retrograde flow. We also observed p40–EGFP diffusion (Fig. 6 and supplementary material Movie 9) with a mean diffusion constant of 0.63±0.06 μm²/second (Fig. 6A). WAVE2 and p40 also have similar residence times on the membrane (τ=287±43 mseconds Fig. 5C and τ=325±17 mseconds Fig. 6B, respectively). Our data support a model where the Arp2/3 complex binds the SCAR/WAVE complex and diffuses along the membrane in search of an actin filament for Arp2/3 complex activation.

Are laterally diffusing Arp2/3 complex molecules precursors to actin nucleation? To address this question, we performed extended high-speed imaging of cells that contained both stationary and diffusing p40–EGFP molecules. We observed individual particle diffusion before retrograde flow (an example is shown in Fig. 6C and supplementary material Movie 10), although not all retrograde flow events were preceded by detectable lateral diffusion (Fig. 6D and supplementary material Movie 11). In this high-intensity rapid imaging mode, molecules photobleached rapidly, which made it difficult to discern retrograde motion from an immobile particle. To better capture the transition from lateral diffusion to retrograde flow, we switched imaging frequencies during the acquisition. We first imaged molecules in a high-frequency mode to capture lateral diffusion and then switched to a low-frequency, low-light mode that is more optimal to observe retrograde flow for long periods of time. With this approach, we could capture molecules transitioning from rapid diffusion to an extended period of retrograde movement. In Fig. 6E and supplementary material Movie 12, we show a molecule of p40–EGFP, which underwent 400 mseconds of lateral diffusion with a diffusion coefficient of 0.12 μm²/second (Fig. 6F, right) before its incorporation into the cytoskeletal network. We then followed the molecule during its retrograde motion in a lower-frequency imaging mode. This method enabled us to visualize retrograde flow after particle diffusion (Fig. 6F, left). We found 13 p40 molecules that showed clear retrograde flow. Four of these molecules were preceded by at least four frames (400 mseconds) of lateral diffusion. The other molecules underwent retrograde flow without detectable lateral diffusion, or diffusion was too brief for us to observe. Together, these data show that some Arp2/3 complexes locally search the membrane by 2D diffusion for short periods whereas other molecules are recruited directly from the cytosol before incorporation into the cytoskeleton.

We uncovered new behavior of SCAR/WAVE and Arp2/3 complexes with single-molecule imaging before, during and after nucleation in cells. These data are summarized in our model (Fig. 7). Our first insight relates to the order of recruitment of SCAR/WAVE and Arp2/3 complexes to the plasma membrane. Actin nucleation is a highly regulated event that requires the convergence of several intracellular signals and cytoskeletal structures. Activation of the Arp2/3 complex requires its interaction with nucleation-promoting factors such as the SCAR/WAVE complex, as well as actin filaments (Mullins et al., 1998; Machesky et al., 1999). Similarly, SCAR/WAVE complex activation requires the combinatorial input of Rac-GTP, phosphoinositides and other signals (Oikawa et al., 2004; Lebensohn and Kirschner, 2009; Chen et al., 2010). Two models have been proposed for how the SCAR/WAVE and Arp2/3 complexes first associate with the membrane. In the first model, the SCAR/WAVE complex is recruited and activated by plasma-membrane-bound components such as Rac-GTP and phosphoinositides, and then the Arp2/3 complex is recruited from the cytosol to this membrane-bound SCAR/WAVE complex (Lebensohn and Kirschner, 2009; Chen et al., 2010). In the second model, the Arp2/3 complex is pre-associated with the SCAR/WAVE complex in the cytosol, and both of these complexes are recruited together to membranes (Stradal et al., 2004). However, WAVE2 and p40 have different distributions on the membrane (Figs 1, 2), which suggests that not every SCAR/WAVE complex is pre-associated with an Arp2/3 complex.

What happens after the SCAR/WAVE complex initially associates with the plasma membrane by binding to Rac-GTP or phosphoinositides? The SCAR/WAVE complex probably recruits the Arp2/3 complex to the plasma membrane, as reflected in the similar lifetime of p40 and WAVE2 on the membrane (Fig. 5C, Fig. 6B). Both WAVE2 and p40 laterally diffuse upon recruitment to the plasma membrane, probably because of Rac-GTP or phospholipid binding. Rac2 has a diffusion coefficient of 0.54±0.04 μm²/second (supplementary material Fig. S3A,B); the diffusion coefficient for WAVE2 is 0.41±0.04 μm²/second (Fig. 5B); and the diffusion coefficient for p40 is 0.63± 0.05 μm²/second (Fig. 6A). These values are within the expected range of diffusion coefficients for membrane-associated proteins (Murakoshi et al., 2004; Lommerse et al., 2004; Douglass and Vale, 2005; Matsuoka et al., 2006). Lateral diffusion of SCAR/WAVE and Arp2/3 complexes could help these proteins perform a constrained 2D search for nucleation binding sites, including actin barbed ends or sides of filaments, which are necessary to co-activate the Arp2/3 complex. Limited dimensionality searches have been reported in other biological contexts to decrease search time. For instance, the LacI repressor executes short 1D searches along DNA to find its lacO target 100 times faster than can be accomplished with 3D diffusion in the cytosol (Riggs et al., 1970; von Hippel and Berg, 1989; Wang et al., 2006; Elf et al., 2007). These constrained diffusion events need not be extensive to dramatically decrease search time; the LacI repressor spends 90% of the time diffusing on DNA with a short residence time of 5 mseconds (Elf et al., 2007). Both WAVE2 and p40 had similar residence times of ∼300 mseconds during lateral diffusion (Fig. 5C and Fig. 6B), and more rapid imaging as described previously (Elf et al., 2007) might be needed to capture very short-lived 2D searches.

What happens after SCAR/WAVE and Arp2/3 complexes trigger an actin-nucleation event? We show that the Arp2/3 complex incorporates into the actin network and undergoes retrograde flow, which is consistent with previous reports (Ponti et al., 2004; Miyoshi et al., 2006; Iwasa and Mullins, 2007). Surprisingly, we also observed retrograde flow of the SCAR/WAVE complex (Fig. 1). Because WAVE2 retrograde speed and lifetime matches that of actin and p40 (Fig. 2), we hypothesize that WAVE2 incorporates into the actin meshwork after a nucleation event. Alternatively, WAVE2 might bind the side of an actin filament without first activating Arp2/3 to undergo retrograde motion. In support of the former model, most WAVE2 molecules with retrograde motion first attach to the actin network near the leading edge, where the SCAR/WAVE complex activates the Arp2/3 complex. However, the frequency of WAVE2 retrograde flow is significantly lower than either actin or p40 (Fig. 1F). Several possibilities could account for the discrepancy between the SCAR/WAVE and Arp2/3 complexes. The simplest hypothesis is that only a subset of membrane-bound SCAR/WAVE complex is involved in Arp2/3 complex recruitment and actin nucleation. Consistent with this hypothesis, the overall distributions of WAVE2 and p40 differ, with the cell periphery more enriched for the SCAR/WAVE complex (Fig. 1A) compared with the Arp2/3 complex (Fig. 2A). Alternatively, the SCAR/WAVE complex might activate multiple Arp2/3 molecules before being recycled through retrograde flow. In this scenario, most SCAR/WAVE complexes remain associated at the cell periphery after nucleation and only occasionally track with an Arp2/3 complex during retrograde flow. Finally, all SCAR/WAVE complexes that induce actin nucleation might be removed from the membrane by incorporating into the actin network, but some molecules might be removed more rapidly than our imaging conditions can capture. Future studies are needed to distinguish between these possibilities.

Our work sheds new light on the role of actin polymerization in SCAR/WAVE complex recycling. Both SCAR/WAVE complex removal from the actin cortex (Weiner et al., 2007; Millius et al., 2009) (this study) and N-WASP removal from the surface of vaccinia viruses (Weisswange et al., 2009) depend on actin polymer. Similarly, regulators of actin assembly in budding yeast require actin polymer for rapid recycling (Kaksonen et al., 2003). For these experiments, actin drugs such as latrunculin have been used to uncover the role of actin polymer in recycling. However, latrunculin not only stops new assembly but also results in disassembly of the existing actin network. This makes it difficult to discern whether ongoing actin assembly or recycling proteins associated with the actin network is more relevant for recycling nucleation-promoting factors. There is evidence that both of these mechanisms participate in recycling endocytic components in yeast. For instance, the actin network is known to serve as a scaffold for proteins such as phosphatidylinositol (4,5)-bisphosphate phosphatases, which catabolize signals required for endocytosis (Sun et al., 2007). By contrast, recycling has also been shown to involve the force of actin polymerization to physically dissociate components from the membrane (Kaksonen et al., 2003). To discriminate which of these mechanisms is responsible for SCAR/WAVE complex recycling, we used a combination of actin drugs to stop new actin assembly while also stabilizing the existing actin cytoskeleton (Peng et al., 2011). This pharmacological cocktail increases the lifetime of the cortical WAVE2 complex (Fig. 4B,C), which suggests that ongoing assembly, and not an actin-network-associated effector protein involved in SCAR/WAVE turnover, is more relevant for SCAR/WAVE recycling. However, inhibition of actin dynamics had little effect on the global pool of WAVE2 (assayed by FRAP, Fig. 4F). This suggests that SCAR/WAVE molecules exist in two distinct states. In the first state, SCAR/WAVE molecules freely associate and dissociate with the plasma membrane independently of actin polymerization. This probably represents a low-affinity interaction of the SCAR/WAVE complex with factors such as Rac and phosphoinositides. In the second state, which represents a small fraction of the total pool (Fig. 1F), molecules undergo actin-dependent retrograde flow, and require actin polymerization for recycling. This probably represents SCAR/WAVE molecules that are associated with the actin cytoskeleton. We favor a model where the SCAR/WAVE complex performs multiple diffusive membrane searches until it integrates into the actin meshwork through Arp2/3 activation, and subsequently recycles through actin-based retrograde flow.

Plasmids containing cDNA encoding human β-actin subcloned into pEGFP-C1-derived vector (Clontech Laboratories) or the Xenopus p40 subunit of the Arp2/3 complex subcloned into the pEGFP-N1-derived vector containing a defective cytomegalovirus (delCMV) have been previously published (Watanabe and Mitchison, 2002; Miyoshi et al., 2006). EST clone of Xenopus WAVE2 was obtained from the IMAGE consortium and subcloned into the pEGFP-N1 delCMV vector. Human Rac2–GFP (Michaelson et al., 2001) and RacN17 (Srinivasan et al., 2003) have been described previously. Y27632 (Calbiochem) was incubated with cells at final concentration of 10 μM for at least 30 minutes before imaging. Jasplakinolide (Calbiochem) and latrunculin (Calbiochem) were prepared at 2× final concentration (40 μM) and acutely added during imaging. XTC cells were cultured and transfected as described previously (Watanabe and Mitchison, 2002).

Single-molecule imaging was performed as previously described (Watanabe and Mitchison, 2002). Images were acquired in Japan and in the USA on a variety of different microscopes (BX52 or IX71; Olympus; or TI-2000, Nikon) with different EM-CCD cameras (MMX 1300-YHS, CoolSNAP HQ, Photometrics Evolve, Photometrics Cascade II:512, or Andor), acquisition software (Metamorph or NIS-Elements), lamps (100 W mercury or 75 W Xenon), lasers (20 mW, BCD1 Melles Griot or 100 mW, LMM5 Spectral Research) and objectives (planApo 100× objective; NA 1.40 oil, Olympus or NA 1.49 oil, Nikon). Data from all microscope set-ups were comparable and indicate the robustness of the observed phenomena. XTC cells were plated on polylysine-coated coverglass for 20 minutes in 70% Leibovitz’s L-15 medium (Invitrogen) without serum. Time-lapse imaging was performed at 21–23°C up to 3 hours after cells were seeded. Epifluorescence microscopy imaging of retrograde flow was performed with a 2× lens inserted for cells expressing a low concentration of EGFP-tagged proteins with 2–3 second exposures every 3 seconds. Neutral density filters were used to reduce illumination intensity so that fluorescence decay caused by photobleaching was minimized. The field diaphragm was closed to restrict illumination to the cell edge and diminish out-of-focus light from the cell body. Diffusing molecules were observed by illuminating for 40–100 mseconds in total internal reflection fluorescence (TIRF) microscopy for ∼100 frames.

Plated cells were imaged with a 488 laser line in TIRF microscopy for 400–600 mseconds every 4–6 seconds for 30–60 seconds, a 1 second 488 FRAP pulse with full laser power was spatially controlled and applied using a mosaic digital micromirror device (Andor), and then cells were imaged for 60 seconds to observe fluorescence recovery. Fluorescence intensity was normalized to the fluorescence intensity immediately before (intensity =1) and immediately after FRAP (intensity =0).

Retrograde motion of WAVE2 molecules was analyzed using the two-dimensional Gaussian fit method over a 5×5 pixel region of interest with GTrack software (G-Angstrom, Sendai, Japan). Retrograde motion of p40 and actin molecules was analyzed using GMimPro software (Mashanov and Molloy, 2007). Lifetime measurements were normalized for photobleaching as described (Watanabe and Mitchison, 2002). Mean lifetime (τ) was determined by taking 1-cumulative frequency (CF) and fitting to the curve Cexp(–λt), where λ is the first-order dissociation rate and C is a constant with Prism 5.0 software (Graphpad, San Diego, CA) (Hansen and Mullins, 2010; Bieling et al., 2007). Particle diffusion was measured in Matlab (Mathworks, Natick, MA) with Utrack (Jaqaman et al., 2008). Error bars for mean diffusion coefficients and mean lifetimes are 95% confidence intervals. Diffusion constants did not vary significantly between the cell periphery and the cell interior (supplementary material Fig. S3D,E). Maximum intensity projections and kymographs were created using ImageJ. We used the image calculator from ImageJ to capture molecules transitioning from diffusion to retrograde flow. We created a maximum intensity projection of the last five frames in the fast imaging mode before switching to the slower imaging mode. This maximum intensity projection highlights stabilized particles that might subsequently undergo retrograde flow and eliminates diffusing particles. Then we created a maximum intensity projection from five frames earlier in the imaging acquisition. We subtracted these two maximum intensity projections. Most stabilized molecules appear in both maximum intensity projections and thus are eliminated in the subtracted image. However, a few spots remain, which suggests that they were stabilized particles in the maximum intensity projection just prior to switching imaging modes, but not a stabilized particle earlier in the imaging acquisition. These molecules were subsequently analyzed for diffusion and retrograde flow with Utrack in Matlab.

Single-molecule imaging can only be performed at low concentrations of fluorescent protein where the individual fluorophores are well separated. We compared two different strategies for achieving low fluorescent protein concentration. One approach uses high-intensity light to bleach cells expressing moderate or high amounts of fluorescent proteins down to concentrations that are compatible with single-molecule imaging (de Keijzer et al., 2008; van Hemert et al., 2010). When we analyzed XTC cells expressing low levels of the p40 subunit of the Arp2/3 complex under both high- and low-intensity light (supplementary material Fig. S4, 0.11 and 1.8 mW, respectively) we found that, as expected, the higher-intensity light exponentially increased photobleaching (supplementary material Fig. S4C). However, the brighter light also dramatically reduced mean retrograde speed of p40 (supplementary material Fig. S4B,D), suggesting that bleaching cells down to single-molecule levels is unsuitable for analyzing actin dynamics. Instead, we used a crippled CMV promoter to transcriptionally limit expression of fluorescent proteins, thereby achieving low concentrations of fluorescent proteins without the need for photobleaching (Watanabe and Mitchison, 2002). This avoids potential artifacts from overexpression (Riedl et al., 2008) and eliminates the photodamaging effects of photobleaching on actin dynamics.

We thank members of the Weiner and Watanabe labs for helpful discussions, constructive criticisms, and technical support. Rac2 and RacN17 plasmids were kind gifts from Henry Bourne. Purified monomeric GFP was a kind gift from Scott Hansen.