Clathrin is required for Scar/Wave-mediated lamellipodium formation

Lamellipodia are protrusions of the plasma membrane that define a leading edge during cell migration (Abercrombie et al., 1970). The membrane is projected through the force provided by actin polymerisation, which can be mainly ascribed to the activity of the Arp2/3 complex that generates branched actin networks (Pollard, 2007). In mammalian cells, the three Wave proteins have emerged as the major activators of the Arp2/3 complex at the leading edge (Insall and Machesky, 2009; Machesky and Insall, 1998; Yamazaki et al., 2003). The small GTPase Rac controls lamellipodium formation through Wave proteins (Miki et al., 1998).

Wave proteins are embedded into multiprotein complexes, the purification of which provided the molecular link between the small GTPase Rac and the Arp2/3 complex (Derivery and Gautreau, 2010; Insall and Machesky, 2009). The canonical Wave complex contains five subunits, namely, Sra, Nap, Abi, Brk1 and Wave (Gautreau et al., 2004; Kobayashi et al., 1998; Machesky et al., 1999). However, like Wave proteins, most subunits of the Wave complex are encoded by paralogous genes, which are variably expressed in different cell types, giving rise to a combinatorial complexity of the Wave complexes that can be assembled (Derivery and Gautreau, 2010). The most ubiquitous Wave complex is composed of Sra1, Nap1, Abi1, Brk1 and Wave2 (Gautreau et al., 2004).

Despite this complexity, a core mechanism regulating Wave complex activity has recently emerged. This mechanism involves Rac and phosphatidylinositol 3,4,5-trisphosphate [PtdIns(3,4,5)P₃]. PtdIns(3,4,5)P₃, the product of phosphoinositide 3-kinase, is produced in response to Rac signalling and further activates Rac in a positive feed-back loop (Weiner et al., 2002; Weiner et al., 2006). PtdIns(3,4,5)P₃ is essential for lamellipodium formation and is involved in recruiting the Wave complex to the plasma membrane through a direct interaction (Oikawa et al., 2004; Sossey-Alaoui et al., 2005). In vitro, the purified Wave complex is inactive (Derivery et al., 2009; Ismail et al., 2009). Addition of GTP-bound Rac at high concentration is sufficient to activate the Wave complex through a conformational exposure of the Arp2/3 activating domain of Wave (Ismail et al., 2009). The membrane, however, plays a crucial role in Wave complex activation, because liposomes containing PtdIns(3,4,5)P₃ and prenylated GTP-bound Rac are much more potent activators than soluble Rac alone (Lebensohn and Kirschner, 2009). In addition to this minimal set of activators, BAR (Bin–amphiphysin–Rvs) domain-containing proteins are involved in Wave complex activation and membrane protrusion (Miki et al., 2000; Suetsugu et al., 2006; Takenawa and Suetsugu, 2007).

In the study reported here we used Drosophila cells to identify regulators of the Scar/Wave complex (SWC; Wave is referred to as Scar in this organism). In Drosophila cells, a single SWC exists because each subunit of the SWC is encoded by a single gene, and the SWC controls lamellipodium formation, as in mammalian cells (Ismail et al., 2009; Kunda et al., 2003; Rogers et al., 2003). Using a screen combining proteomics with functional genomics, we found Clathrin heavy chain (CHC) as a protein that binds to the SWC and that regulates lamellipodium formation. These two properties of CHC are conserved from Drosophila to mammalian cells. Importantly, we show that the novel function of CHC in lamellipodium formation depends on its ability to promote the recruitment of the SWC to the plasma membrane, but appears independent of its classical role in membrane trafficking.

We reasoned that activators of SWC should both bind to the SWC and produce the characteristic phenotype associated with SWC inactivation. Thus we designed a dual screen in Drosophila cells, in which the proteins that bind to the SWC were identified through proteomics and the genes producing the phenotype associated with SWC inactivation were identified upon RNA interference (RNAi)-mediated inactivation. We raised polyclonal peptide antibodies targeting the subunits of the Drosophila SWC (supplementary material Table S1). The antibody targeting the Sra1 subunit was very efficient at immunoprecipitating the SWC from Drosophila S2 cells (Fig. 1A). A large-scale immunoprecipitation was performed with an excess of lysate so as to saturate the antibody with its specific target and thus to minimise the recovery of cross-reactive proteins. The immunoprecipitate lane was cut in 54 gel slices from top to bottom and each slice was analysed by liquid-chromatography–tandem-mass-spectrometry (LC-MS/MS). This analysis resulted in the identification of 135 Drosophila proteins associated with the SWC (supplementary material Table S2). In parallel, we carried out a genome-wide RNAi screen in S2 receptor plus (S2R+) cells for cell shape defects using an established methodology (Kiger et al., 2003). For this purpose, a library of 21,306 distinct double-stranded RNAi (dsRNA) was used. Two immunofluorescences per dsRNA were then individually examined for the morphology of the cell periphery. Drosophila cells depleted of the SWC have a characteristic periphery displaying numerous spikes (Kunda et al., 2003; Rogers et al., 2003). Using our genome-wide screen, 35 genes that produced these characteristic spikes upon RNAi-mediated inactivation were identified (Fig. 1B). As expected, genes encoding Arp2/3 complex subunits and the small GTPases, Rac1, Rac2 and Cdc42 were retrieved (Kunda et al., 2003; Rogers et al., 2003). Twenty-eight genes not previously ascribed to this phenotypic class were identified. These genes include those encoding a formin, a SCF ubiquitin ligase complex, proteins previously involved in intracellular traffic, such as clathrin heavy chain (CHC) and dynein light chain, and nuclear factors involved in transcription and RNA metabolism. The other phenotypic classes are reported elsewhere (Rohn et al., 2011). Importantly, the intersection between candidates identified through proteomics and functional genomics contained two gene products, Hrb27C and CHC (Fig. 1C). Hrb27C contains two RRM domains that bind to RNA molecules. Only ambiguous mammalian orthologues, where the homology does not extend beyond the RNA-recognition motif (RRM) domains, were identified. By contrast, CHC is a conserved protein with a well-established function as a vesicular coat component. We thus decided to focus on CHC. Interestingly, the Clathrin light chain was not identified by either screen. We confirmed the phenotype of CHC-depleted cells using S2 cells and an independent non-overlapping dsRNA (Fig. 1D). Indeed, CHC-depleted cells had a spreading defect with the characteristic spikes associated with SWC inactivation. CHC thus fitted the criteria we had defined for a protein potentially regulating SWC function.

To confirm these results and to identify whether they were repeatable in other organisms, we turned to human HeLa cells, from which the ubiquitous Wave complex has been purified and characterised (Gautreau et al., 2004). Specific co-immunoprecipitation of CHC was detected upon immunoprecipitation of the whole SWC using two independent polyclonal peptide antibodies recognising the Abi1 and Wave2 subunits of the SWC (Fig. 2A). CHC thus co-precipitates with the SWC in both Drosophila and human cell lines. HeLa cells normally form a limited number of lamellipodia. To induce them, transient transfection of active Rac, such as the Q61L mutant that is defective in GTPase activity, is often used. However, this procedure gives highly inhomogeneous results because of variable individual levels of expression, precluding accurate quantification. So we turned to the electroporation of the purified Rac1Q61L protein, an experimental system that we previously characterised for the analysis of a precursor in the assembly of the SWC (Derivery et al., 2008). With this method it is possible to introduce a defined dose of protein homogeneously into the cell population. Rac1Q61L was purified and electroporated into HeLa cells in suspension. Cells were then plated in Petri dishes for biochemical analysis or onto coverslips for immunofluorescence analysis. The purified protein indeed entered the cell, as revealed by the increase in total Rac (Fig. 2B). A pulldown experiment using the CRIB domain of PAK indicated that the Rac1Q61L was GTP bound. As expected, this procedure induced a ‘fried egg’ morphology in the vast majority of HeLa cells, with a continuous lamellipodium covering the entire cell periphery (Fig. 2C). This lamellipodium, best revealed by the actin branch marker cortactin (Cai et al., 2008), was associated with the development of radial focal adhesions (supplementary material Fig. S1) and potentiated cell spreading (Fig. 2C).

We then depleted Nap1, a subunit of the SWC, or CHC by transfecting HeLa cells with shRNAs encoding plasmids (Fig. 2D). Nap1 depletion resulted in a destabilisation of the other subunits, Sra1, Wave2, Abi1 and Brk1, as previously described in several organisms and cell types (Derivery and Gautreau, 2010). As expected, CHC depletion impaired transferrin uptake that depends on clathrin-mediated endocytosis (Fig. 2E). We then compared the Rac response of CHC-depleted cells to cells depleted of Nap1. In both CHC- and Nap1-depleted cells, the lamellipodium induced by active Rac was aberrant (Fig. 2C). Instead of being smooth and continuous over the cell periphery, it was frequently interrupted by spikes, which were reminiscent of the spikes observed in Drosophila cells, although significantly shorter. Both Nap1- and CHC-depleted cells had a reduced area following Rac electroporation, suggesting that their aberrant lamellipodia are impaired in powering cell spreading (Fig. 2C).

Some cells in the population transfected with shRNA targeting CHC managed to spread efficiently and formed continuous lamellipodia. We observed that these cells with normal lamellipodia were usually the ones less efficiently depleted of CHC (Fig. 3A). We thus decided to quantify the extent of CHC depletion, surface area and spike formation on a cell-by-cell basis. Indeed, the majority of CHC shRNA-transfected cells that exhibited a lower CHC intensity than control cells also had a smaller surface area (Fig. 3B, lower left quadrant). Conversely, the majority of cells transfected with control shRNA showed high CHC intensity and large cell area (upper right quadrant). The phenotype of spike formation was less penetrant than the reduced cell area, because one third of cells efficiently depleted of CHC formed no spikes. However, the remaining two thirds of CHC-depleted cells had five or more spikes, a number that is very rarely reached by control cells (upper left quadrant, Fig. 3B). These observations indicate an agreement between the level of depletion of CHC and the occurrence of these two phenotypes. Similar results were observed with two unrelated shRNA constructs that efficiently targeted CHC (Fig. 3C). Together these results indicate that CHC, like the SWC, is required for lamellipodium formation in human cells, as well as in Drosophila cells.

The defect in lamellipodium formation observed upon CHC depletion might be an indirect effect of defective membrane trafficking. Indeed, there are numerous ways in which clathrin depletion could affect lamellipodium formation. For example, signalling growth factor receptors, which trigger lamellipodia, are known to be endocytosed in clathrin-coated pits (Scita and Di Fiore, 2010). This was one of the reasons why we used electroporation of active Rac in the experiments described above rather than growth factor stimulation to analyse lamellipodium formation. Even Rac traffics and the intracellular trafficking of Rac was shown to be important in order to restrict its signalling to the leading edge of migrating cells (Palamidessi et al., 2008). These findings are likely to explain why the giant lamellipodium induced by exogenous Rac in our assay is isotropic, i.e. not restricted to a leading edge. To examine whether the classical role of clathrin in membrane trafficking has an impact on this new role we describe, lamellipodium formation, we sought to specifically inactivate the role of clathrin in trafficking.

To generate vesicles, clathrin is recruited through different adaptor proteins at various cellular locations (Robinson, 2004). We first used brefeldin A (BFA), which blocks recruitment of coat proteins to membranes, including AP1–clathrin and GGA–clathrin to the trans-Golgi network (TGN) and endosomes, and causes disassembly of the Golgi. Treatment of cells with BFA for 20 minutes resulted in the expected redistribution of the Golgi marker GM130, indicating that the drug treatment was effective (Fig. 4A). This BFA treatment, however, had no effect on lamellipodium formation. We then directly depleted the adaptor complexes AP1 and AP2 using shRNA-encoding plasmids (Fig. 4B). As expected, AP1 depletion impaired steady state distribution of the mannose 6-phosphate receptor, and AP2 depletion strongly impaired transferrin uptake. The depletion of AP1, unlike CHC depletion, did not induce defective lamellipodia, as evidenced by the number of spikes and the Rac-dependent increase in cell area (Fig. 4C). It even had a slightly potentiating effect on the increase of cell area. AP2 depletion had also no detrimental effect on lamellipodium formation, even though the AP2 complex is the adaptor complex that recruits clathrin to the plasma membrane. Altogether these experiments indicate that the effect of clathrin depletion on lamellipodium formation is unlikely to be a secondary consequence of defective trafficking.

We sought to colocalise CHC with SWC at the lamellipodium. HeLa cells electroporated with active Rac is not the optimal system for this purpose, because the enrichment of the SWC at lamellipodia is very limited (supplementary material Fig. S1). We thus turned to meningioma cells, which spread very efficiently without exogenous Rac. Simple examination of these cells using epifluorescence microscopy reveals an enrichment of cortactin at the periphery (supplementary material Fig. S2A). This enrichment is not observed when a soluble protein, such as RhoGDI, is stained, indicating that enrichment at the periphery is not an artefact due to an increase in membrane thickness in this region. A detectable localisation of the SWC at these lamellipodia was found using Wave2 polyclonal antibody (pAb; Fig. 5) or Brk1 monoclonal antibody (mAb; supplementary material Fig. S2B). Upon depletion of either Wave2 or Brk1, staining of the plasma membrane was lost (supplementary material Fig. S2B,C). X-22 mAb, a well-characterised antibody recognising CHC, also clearly stained lamellipodia. Using confocal sections, we found that CHC colocalises with Wave2. A polyclonal antibody targeting CHC produced similar staining of the lamellipodia. In addition to the lamellipodial staining there was the expected punctuated staining at the ventral membrane using both CHC antibodies (Fig. 5). As expected, both punctate ventral and peripheral stainings obtained with either antibody disappeared upon CHC depletion (supplementary material Fig. S2D). The CHC punctae at the ventral membrane correspond to classical clathrin-coated pits, as indicated by their colocalisation with the adaptor complex AP2 and the so-called endocytic accessory proteins, Eps15, CALM and epsin (Fig. 5). The endocytic accessory proteins promote the assembly of the clathrin lattice around endocytic cargoes clustered by AP2. Importantly, none of these endocytic proteins was detected in the lamellipodia with CHC, suggesting that CHC is unlikely to be involved in endocytosis at the lamellipodium.

We depleted meningioma cells of the SWC by transfecting siRNAs targeting either Brk1, Nap1 or Wave2 subunits. Each of these transfections induced defective lamellipodia (Fig. 6A), indicating that lamellipodia depend on SWC in this cellular system as well. We also depleted meningioma cells of CHC or AP2. Similarly to Rac electroporated HeLa cells, CHC depletion, but not AP2 depletion, induced defective lamellipodia (Fig. 6B). This confirmed that the role of CHC in lamellipodium formation does not rely on its function in endocytosis. Accordingly, AP2 depletion did not impair the peripheral staining of CHC (Fig. 6C). This peripheral staining of CHC was also observed by total internal reflection fluorescence (TIRF) microscopy, confirming it was associated with flat lamellipodia (supplementary material Fig. S3). This observation indicates that CHC localisation at lamellipodia is independent from AP2, the most established adaptor complex of endocytosis.

Wave2 was no longer enriched at the plasma membrane of CHC-depleted meningioma cells, consistent with their defective lamellipodia (Fig. 6C). This observation raised the possibility that CHC controls the distribution of the SWC between the cytosol and the membranes. This was examined using biochemical fractionation of HeLa cells. The procedure we used efficiently separated membranes (both plasma membrane and endomembranes) from the cytosol, as indicated by the respective markers, transferrin receptor, an integral membrane protein, and soluble RhoGDI (Fig. 7A). Upon CHC depletion, the membrane pool of Wave2 was reduced to approximately a quarter of the level in control cells (Fig. 7B for quantifications; supplementary material Fig. S4A for western blot).

To test whether CHC is able to bring the SWC to the plasma membrane, we fused CHC with an N-terminal myristoylation sequence. Myr–CHC and CHC as a control were tagged with a protein C epitope (PC) to detect their expression. As expected, Myr–CHC was almost completely membrane bound, in contrast to endogenous or exogenous CHC, and indeed, the membrane pool of Wave2 increased upon expression of Myr–CHC (Fig. 7C for quantifications; supplementary material Fig. S4B for western blot). Conversely, the overexpression of wild-type CHC decreased significantly the amount of membrane Wave2. This effect of CHC overexpression was unanticipated, but was even more pronounced than the effect of Myr–CHC. By contrast, the overexpression of CHC or of Myr–CHC had no effect on tranferrin endocytosis (Fig. 7D). Because CHC overexpression decreased the amount of SWC in the membrane fraction, similar to CHC depletion, we expected a defect in lamellipodia formation. Indeed, Rac1Q61L electroporation induced defective lamellipodia in CHC-overexpressing HeLa cells, as evidenced by the increased number of spikes and decreased spreading (Fig. 7E). When overexpressed CHC was targeted to the membrane through myristoylation, these defects were not present. Hence, CHC overexpression uncouples the two functions of CHC. Because CHC overexpression impairs the formation of SWC-dependent lamellipodia without affecting clathrin-mediated endocytosis, we selected this method to assess the role of clathrin in cell migration.

To analyse the role of CHC in migration, we turned to murine fibroblast cells that migrate faster than HeLa cells. Swiss 3T3 cells were transfected with CHC, Myr–CHC or the control plasmid, together with a GFP plasmid to identify transfected cells before time-lapse imaging by DIC microscopy (Fig. 8A). The leading edge activity of these cells was characterised by cycles of protrusion and retraction (supplementary material Movie 1). Retraction occurs when the lamellipodium tip detaches from the substratum and moves backward forming a membrane ruffle. Kymograph analysis of leading edges in transfected cells revealed that CHC overexpression significantly reduced the velocity of lamellipodia, whereas Myr–CHC significantly increased it (Fig. 8B). Myr–CHC expression induced frequent membrane ruffles. By contrast, CHC-overexpressing cells generated significantly shorter protrusions followed by reduced ruffles. The frequency of ruffles in CHC-expressing cells was also half that in Myr–CHC-expressing cells. These observations are, therefore, perfectly in line with the respective effects of these constructs on membrane recruitment of the SWC, which is essential for the generation of a branched actin network that projects the plasma membrane (Derivery and Gautreau, 2010; Suetsugu et al., 2006).

To examine how these defects in lamellipodium protrusion affect cell migration, we used an assay, in which islands of cells were seeded through a microfluidic device perforated with holes of 700 μm in diameter. Swiss 3T3 cells were transfected as above with CHC, Myr–CHC or a control plasmid in combination with a GFP plasmid. The islands were fixed and examined after seeding or after incubation for 24 hours. GFP-expressing cells were then counted. CHC overexpression significantly decreased the number of cells migrating outside the initial island compared with the control, whereas Myr–CHC expression significantly increased the number of migrating cells (Fig. 8C). The ability of these cells to migrate is thus associated with the velocity of protrusions and the amount of SWC at the membrane.

The dual screen we conducted using Drosophila cells was aimed at identifying novel regulators of the SWC involved in lamellipodium formation. Surprisingly, this screen identified CHC, a major coat protein involved in membrane trafficking. A molecular interaction between the Sra1 subunit of the SWC and CHC was recently reported by Hoflack and colleagues (Anitei et al., 2010). However, they found that this interaction played a role in the generation of tubular carriers derived from the TGN, a classical function for clathrin. By contrast, the atypical function of CHC in lamellipodium formation we report here seems to be independent from its well-established role in membrane trafficking. The evidence against an indirect effect on lamellipodium formation through defective membrane trafficking is threefold. (1) We were able to uncouple the two functions of CHC, using adaptor depletions or BFA. These experiments impaired trafficking, but not lamellipodium formation. (2) Conversely, overexpression of CHC impaired lamellipodium formation, but not trafficking. (3) CHC was detected in lamellipodia without the adaptor and accessory proteins mediating endocytosis. The two functions associated with the complex formed by CHC and SWC, generation of carriers from the TGN and lamellipodium formation, thus appear distinct, even though they share components. These shared components provide a simple explanation of why AP1-depleted cells spread slightly more than control cells in response to Rac (Fig. 4C), because the SWC, which is no longer recruited to the TGN in AP1-depleted cells, can then perform its function at lamellipodia.

The novel function of clathrin in controlling lamellipodia adds to the case for unconventional roles of clathrin. Indeed, clathrin was previously shown to stabilise the mitotic spindle (Royle et al., 2005) and to mediate p53-dependent transcription (Enari et al., 2006). This latter role is also independent of the light chains, as is the role of CHC in lamellipodium formation we describe here. CHC promotes lamellipodium formation through membrane recruitment of SWC. Indeed, our experiments using Myr–CHC suggest that CHC is able to bring the SWC to the membrane. In line with this idea, CHC depletion, or overexpression, greatly decreases the amount of SWC in the membrane pool. This role of clathrin in the recruitment of the SWC to the plasma membrane might explain the recent finding that clathrin is required for actin polymerisation at the immunological synapse (Calabia-Linares et al., 2011), a structure that also depends on SWC activity (Billadeau et al., 2007).

The recent success of reconstituting the activation of purified SWC in vitro using prenylated Rac and liposomes containing PtdIns(3,4,5)P₃ suggest that clathrin is not absolutely required to induce and maintain the active conformation of the SWC (Lebensohn and Kirschner, 2009). The situation is analogous to the one described for BAR-domain-containing molecules of the IRSp53 family, which are crucial in vivo to deform the plasma membrane, to recruit and activate the SWC (Miki et al., 2000; Suetsugu et al., 2006; Takenawa and Suetsugu, 2007), but which are similarly dispensable in in vitro assays. These results suggest that, even though actin dynamics have been beautifully reconstituted in vitro, the complexity of a lamellipodium, especially its membrane dynamics, has not yet been fully understood and recapitulated in vitro.

It is striking that a major component of the endocytic machinery such as CHC should also be involved in SWC activation and in the formation of lamellipodia. Protrusion of the plasma membrane through actin polymerisation and membrane retrieval through endocytosis are antagonistic. Indeed, biophysical experiments have revealed that either one of these events, but never an unproductive combination of the two, is triggered by the same stimulus, a decrease in membrane tension (Dai et al., 1997; Raucher and Sheetz, 1999; Raucher and Sheetz, 2000). The interaction between CHC and SWC might thus be involved in this coordination by locally shutting down endocytosis in membrane protrusions.

Drosophila S2 cells were cultured at 26°C in Schneider's medium containing 10% foetal bovine serum (FBS) and penicillin–streptomycin. When S2 cells were grown in suspension culture using Erlenmeyer's flasks under gentle agitation, the medium was supplemented with 1:100 volume of a 10% Pluronic F-68 solution (Invitrogen). Meningioma, HeLa and 293T cells were cultured in Dulbecco's modified Eagle's medium (DMEM) containing 10% FBS and penicillin–streptomycin. The meningioma cells (#SF1335) used are a surgical isolate from a patient affected by a benign meningioma, a slow-growing tumour arising from arachnoidal cells of the meninges (Rempel et al., 1993). Swiss 3T3 cells were incubated in DMEM supplemented with 5% donor bovine serum. Mammalian cells were incubated at 37°C in a humidified atmosphere containing 5% CO₂. Brefeldin A (BFA; Sigma) was used at 5 μg/ml for 20 minutes.

Details about the genome-wide screen using S2R+ cells are described elsewhere (Rohn et al., 2011). For RNAi-mediated inactivation, S2 cells were incubated with in vitro produced dsRNAs, as previously described (Bouslama-Oueghlani et al., 2007). dsRNAs were produced from genomic amplified fragments using primers 5′-CTGTACGTGCTCAAGAAAAACG-3′ and 5′-GCATAGGTTCGGGTAGTATTGG-3′ for Abi (CG9749), or 5′-GTGCTTTTCGAGTTTCTGTGC-3′ and 5′-TATCAAAACCGTGGATACTTCG-3′ for CHC (CG9012). These fragments were from an independent library (Echard et al., 2004). S2 cells treated with dsRNA were assayed by spreading on concanavalin-A-coated coverslips as described previously (Rogers et al., 2003). HeLa and 293T cells were transfected with plasmids using electroporation or calcium phosphate precipitation, respectively. Swiss 3T3 fibroblasts were transfected using Lipofectamine LTX and the Plus reagent (Invitrogen). Meningioma cells were transfected with 40 nM On-Target Plus SMART pool (Dharmacon) using lipofectamine RNAi Max (Invitrogen) and analyzed after 4 days. pSuper-Nap1, pSUPER-CHC, pSUPER-AP1γ and pSUPER-AP2μ2 were described previously (Dugast et al., 2005; Saint-Pol et al., 2004; Steffen et al., 2004). For shRNA-mediated RNAi, cells expressed pSUPER plasmids for 3 days. The two additional CHC shRNAs (#2: 5′-TAATCCAATTCGAAGACCAAT-3′; #3: 5′-AATGGATCTCTTTGAATACGG-3′) were cloned into the psiRNA-h7SKblasti G1 plasmid and compared with the so-called scr control plasmid (Invitrogen). For overexpression, cells expressed plasmids for 1 day.

Full-length human clathrin heavy chain (1–1675) was amplified from the IMAGE:6045540 clone and subcloned into custom-made plasmids, pcDNA5 FRT His-PC-TEV or pcDNA5 FRT Myr-PC-TEV between the FseI and AscI sites. This resulted in the fusion of CHC with the N-Terminal peptides, MHHHHHHEDQVDPRLIDGKGGGDYDIPTTENLYFQGAMGRP, or MGSNKSKPKDEDGVDPRLIDGKGGGDYDIPTTENLYFQGAMGRP, respectively. All amplified inserts were verified by sequencing.

Anti-human Sra1, Nap1, Wave2 and Abi1 pAbs were described previously (Gautreau et al., 2004). Anti-human Brk1 mAb (clone 231H9) was described previously (Derivery et al., 2008). Anti-tubulin mAb (clone E7) was developed by M. Klymkowsky and obtained from Developmental Studies Hybridoma Bank. Anti-mannose-6-phosphate receptor mAb (22D4) was a kind gift from P. Benaroch (Institut Curie, Paris). The following Abs were from commercial sources: anti-CHC (for western blotting, mAb clone 23, BD Biosciences; for immunofluorescence, mAb X-22 and pAb ab21679, Abcam), anti-AP2μ2 mAb (for western blotting, clone 31/AP50, BD Biosciences), anti-AP2α2 mAb (for immunofluorescence, clone AC1-M11, Abcam), anti-AP1γ (clone 100/3, Sigma), anti-Rac1 mAb (clone 102, Transduction Laboratories), anti-cortactin mAb (clone 4F11, Millipore), anti-transferrin receptor mAb (clone H68.4, Invitrogen), anti-RhoGDI pAb (A-20, sc-360, Santa Cruz), anti-VASP mAb (clone 43, BD Biosciences), anti-paxillin mAb (clone 165, BD Biosciences), anti-CALM pAb (C-18, sc-6433, Santa Cruz), anti-Eps15 pAb (K-15, sc-11716, Santa Cruz), anti-epsin pAb (R-20, sc-8673, Santa Cruz). HRP-conjugated antibodies were from Pierce. Fluorescent phalloidin and Alexa-Fluor-conjugated secondary antibodies were from Molecular Probes. For immunofluorescence, cells were fixed using 3% paraformaldehyde then permeabilised with 0.5% Triton X-100. They were then processed using standard techniques. None of the reported stainings was seen in control experiments, where the primary antibody was omitted.

HeLa cells or transfected 293T cells were lysed by nitrogen cavitation (500 psi for 20 minutes; Parr Instruments) in XB buffer (20 mM Hepes, 100 mM KCl, 1 mM MgCl₂, pH 7.7) supplemented with protease inhibitor cocktail (Sigma). Debris and nuclei were pelleted by centrifugation at 600 g for 10 minute at 4°C. The supernatant (total fraction) was then centrifuged for 60 minutes at 150,000 g using a TLA-100.4 rotor. The supernatant was the cytosolic fraction, and the membrane pellet was solubilised for 60 minutes in radioimmunoprecipitation assay (RIPA) buffer (50 mM Hepes, 150 mM NaCl, 5 mM EDTA, 1% NP-40, 0.5% sodium deoxycholate, 0.1% SDS, pH 7.7) supplemented with protease inhibitor cocktail. The membrane fraction was then centrifuged again for 60 minutes at 150,000 g to remove insoluble material.

HeLa cells were co-transfected with the plasmid of interest and a GFP-expressing plasmid. Cells were serum starved for 30 minutes in internalisation medium (DMEM, 0.5% BSA, 10 mM Hepes, pH 7.4) at 37°C, washed in PBS and gently detached with trypsin. Cells were then washed with cold internalisation medium. Cells were incubated with Alexa-Fluor-647-conjugated transferrin (5 μg/ml) for 1 hour at 4°C with gentle agitation. After one wash in cold internalisation medium, cells were resuspended in warm internalisation medium and placed in a 37°C water bath. An aliquot of cells was taken at each time point and quickly placed on ice in an excess of cold Stop medium (PBS with 0.5% BSA). Cells were then acid washed in 50 mM glycine, 150 mM NaCl, pH 3, on ice for 3 minutes and acidity was neutralised by an excess of Stop medium. After two washes in Stop medium, cells were analyzed by flow cytometry (MoFlo ASTRIOS, Beckman-Coulter). 10⁴ GFP-positive cells were analyzed per sample and the geometrical mean fluorescence was determined using Summit software.

For immunoprecipitations, 10¹⁰ S2 cells (2.5 l culture, 14 ml cell pellet) were treated for 5 minutes with 1 mM protease inhibitor, di-isopropyl-fluoro-phosphate (Sigma), with appropriate safety procedures. HeLa and S2 cells were lysed in RIPA buffer supplemented with protease inhibitor cocktail (1:1000; Sigma P8340). Extracts were clarified by centrifugation and ultracentrifugation, then incubated with 2 μg non immune rabbit IgG or affinity purified peptide antibodies, together with 10 μl of protein-A-Affiprep beads (Bio-Rad) for 2 hours at 4°C. Beads were then washed four times with RIPA buffer and resuspended in SDS loading buffer. SDS-PAGE was performed with NUPAGE 4–12% Bis–Tris or 3–8% Tris–acetate gels (Invitrogen). Western blots were revealed using HRP-coupled antibodies, Supersignal kit (Pierce) and Fuji LAS-3000 (Fujifilm). Blots were quantified using ImageJ.

Gel slices were reduced, alkylated and subjected to digestion with trypsin (Roche Diagnostics). Extracted peptides were dried and resolubilised in solvent A [95/5 water/acetonitrile in 0.1% (w/v) formic acid]. The total digestion product of a gel slice was used per LC-MS/MS analysis. The extracted peptides were concentrated and separated on a LC-Packings system (Dionex S.A., Voisins-le-Bretonneux, France) coupled to the nano-electrospray II ionisation interface of a QSTAR Pulsar I (Applied Biosystems) using a PICOTip (10 μm internal diameter, New Objectives). Bound peptides were eluted with a gradient of 5–50% of solvent B [20/80 water/acetonitrile in 0.085% (w/v) formic acid]. Information-dependent acquisition was used to acquire MS/MS data, with experiments designed so that the two most abundant peptides were subject to collision-induced dissociation, using nitrogen as the collision gas, every 5 seconds. Data were searched using MASCOT (Matrix Science) software on the NCBI nr Drosophila melanogaster database. All data were manually verified in order to minimise errors in protein identification and characterisation.

For protein electroporation, GST-Rac1Q61L was purified from Escherichia coli BL21* (Invitrogen) and Rac1Q61L cleaved off the GST using thrombin (Sigma). Rac1Q61L was stored at −80°C in (20 mM Hepes, 100 mM KCl, 5 mM MgCl₂, pH 7.7). HeLa cells were trypsinised and resuspended at 25×10⁶ cells/ml in DMEM containing 10% FCS and 15 mM Hepes pH 7.5. Cells (200 μl) were mixed with 50 μl purified Rac1Q61L (40 μg). Cells were pulsed at 240 V and 950 μF in a 0.4 cm wide cuvette using a BTX ECM630 electroporator (Harvard Apparatus). Cells were then quickly resuspended in the same medium, washed once, and plated for 2 hours onto glass coverslips or Petri dishes.

For pull-down of active Rac, GST fused to the CRIB domain of PAK was purified from E. coli BL21*. Electroporated cells (5×10⁶), spread on Petri dishes, were lysed with 500 μl of a solution containing 20 mM Tris-HCl, 150 mM NaCl, 10 mM MgCl₂, 1% Triton X-100, pH 7.5. The clarified lysate was then incubated with 10 μl of glutathione beads (4B, GE Healthcare) and 5 μg of GST or GST-CRIB for 2 hours at 4°C. Beads were quickly washed four times in the lysis buffer and resuspended in SDS loading buffer.

HeLa and meningioma cells were plated for 2 hours onto glass coverslips coated with 20 μg/ml of collagen type I (BD Biosciences) or 50 μg/ml of fibronectin (Sigma), respectively. Fixed cells were imaged by epifluorescence microscopy on an AxioObserver Z1 microscope (Zeiss) equipped with a 63× NA 1.4 oil immersion objective and an Orca-R² camera (Hamamatsu). Where indicated, confocal optical sections were acquired with either of the following custom spinning disc confocal microscopes. The first one is based on a Nikon TE2000-U inverted microscope, a Yokogawa CSU22 spinning disk head, a 100× NA 1.45 oil immersion objective and a CoolSnap HQ2 camera (Photometrics), operated with Metamorph 7.1.4 software (Molecular Devices). The second one is based on a Nikon Eclipse Ti inverted microscope, a Yokogawa CSU-X1 spinning disk head, a 100× NA 1.40 oil immersion objective and an Evolve camera (Photometrics), operated with Metamorph 7.4 software. A maximum intensity z-projection was calculated between five consecutive planes separated by 200 nm. The TIRF microscope was based on a Nikon Eclipse Ti inverted microscope, a TIRF 60× NA 1.49 PlanAPO oil immersion objective, 491 nm and 561 nm lasers and a CoolSnap HQ2 camera (Photometrics), operated with Metamorph 7.6 software. Binning was always set to 1.

For time-lapse microscopy, Swiss 3T3 cells plated onto glass coverslips were placed into chambers for live-cell imaging. Cells were observed with a Nikon Eclipse-Ti microscope equipped with a Plan-Neofluar 40× NA 1.3 DIC objective at 37°C. Images were acquired with an ORCA-ER camera (Hamamatsu) operated with NIS-Elements AR2.3 software (Nikon). To analyse lamellipodial dynamics at the active cell edge, sequences were acquired for 10 minutes at 3 second intervals.

For the migration assay, we used a custom-made microfluidic device composed of 96 holes of 700 μm diameter (Micronit Microfluidics, PV Enschede, The Netherlands), which was applied for 1 hour onto glass coverslips to seed islands of GFP-transfected cells. Then coverslips were either fixed, or incubated for 24 hours before fixation. Images of the GFP signal and DAPI staining were acquired with a 10× objective. The proportion of transfected cells that had migrated outside the original circle was then calculated.

Analyses of images and movies were performed using ImageJ, Adobe Photoshop and Adobe Premiere software. Gamma settings were not adjusted. For measurements of CHC intensity, images were acquired with the same settings and the intensity was measured using ImageJ after background subtraction. Cell area was measured after manual thresholding of the cortactin staining using ImageJ. Spikes were counted visually. For all quantifications, a minimum of 50 cells of a representative experiment were analysed in each condition. Variations in cell area in independent experiments precluded the averaging of different experiments, but all experiments were repeated between two and six times.

Kymographs were generated along a one-pixel-wide line oriented in the direction of protrusions, using ImageJ. For quantitative analysis, straight lines were drawn on kymographs from the beginning to the end of individual protrusions. Projections along the distance axis and the time axis, gave, respectively, the length and the duration of protrusion. The velocity of protrusion is the ratio between this length and duration. Cycles of protrusion followed by retraction were visually identified. The frequency was calculated by dividing the number of cycles by the duration of acquisition. For each transfection, seven cells were analysed with one to three kymographs per cell. Each kymograph shows several protrusive cycles giving a total ranging from 38 to 47 cycles in the different conditions.

All statistics were performed using SigmaStat software (SPSS). ANOVA was followed by pairwise comparisons using the Student–Newman–Keuls method, with an α factor set to 0.05. When the requirements of normality or equal variance parameters failed to pass, statistics were performed on log₁₀ transformations, which gave normal data sets with equal variance.

We thank Philippe Benaroch, Laurence Goutebroze, Christophe Lamaze, Peter Lowe and Theresia Stradal for providing essential reagents. We thank Vic Small for sharing his unpublished assay for cell migration and Thomas Lecuit for suggesting the myristoylated CHC experiment. We thank Cathy Jackson for critical reading of the manuscript. We acknowledge the Developmental Studies Hybridoma Bank developed under the auspices of the NICHD and maintained by the University of Iowa.