Essential Function of Dynamin in the Invasive Properties and Actin Architecture of v-Src Induced Podosomes/Invadosomes

The large GTPase dynamin plays a key role in endocytosis but is also localized at numerous actin rich sites. We investigated dynamin functions at podosomes/invadosomes, actin-based cellular adhesion structures implicated in tissue invasion. Podosomes/invadosomes are constituted of long F-actin bundles perpendicular to the substratum (actin cores), connected to randomly arranged F-actin fibers parallel to the substratum (actin cloud). We show here that dynamin depletion in v-Src-transformed fibroblasts triggers a massive disorganization of podosomes/invadosomes (isolated or in rosettes), with a corresponding inhibition of their invasive properties. The action of dynamin at podosomes/invadosomes requires a functional full-length protein, suggesting that the effects of dynamin at these sites and in membrane remodelling during endocytosis are mediated by similar mechanisms. In order to determine direct effect of dynamin depletion on invadosome, an optogenetic approach based on the photosensitizer KillerRed was developed. Acute dynamin photo-inactivation leads to a very rapid disorganization of invadosome without affecting focal adhesions. Dynamin therefore is a key regulator of the architecture of actin in podosomes/invadosomes.


Introduction
The large GTPase dynamin, a member of a family of GTPases implicated in membrane remodeling, plays a critical role in membrane fission during clathrin mediated endocytosis [1,2,3]. When this endocytic reaction also involves actin, the spatial and temporal recruitment of dynamin closely correlates with the recruitment of actin and actin regulatory proteins, such as cortactin and components of the N-WASP and Arp2/3 actinnucleating complex [4]. In addition, dynamin is associated with numerous other actin-dependent structures, primarily those nucleated by N-WASP, cortactin and the Arp2/3 complex. These include: lamellipodia, phagocytic cups, immunological and neuronal synapses, actin comets, circular dorsal ruffles, podosomes and invadopodia [5,6,7,8,9,10]. Moreover, pharmacological agents that disrupt dynamin function impair actin dynamics at lamellipodia and affect cell spreading [11]. In agreement with these findings, a dominant negative mutant of dynamin, dynamin K44A, which blocks clathrin-mediated endocytosis, also reduces the formation of several actin-rich structures including dorsal circular ruffles [10,12], podosomes [13,14] and invadopodia [15]. Similarly, depletion of dynamin 2 by siRNA treatment leads to abnormal organization and distribution of podosomes in osteoclasts, but still without entirely preventing their formation [14,16].
A mechanistic link between the function of dynamin and the actin cytoskeleton is supported by both direct and indirect interactions. The C-terminal proline-rich domain (PRD) of dynamin binds a variety of actin regulatory proteins, such as cortactin, Tuba, intersectin and members of the Toca family [17,18,19,20]. Purified dynamin 2, in the presence of cortactin, binds, bundles and remodels actin fibers upon GTP hydrolysis, leading to fraying of the F-actin bundles [21]. Additionally, dynamin was reported to modulate elongation of short F-actin filaments by competing with the capping protein gelsolin [22].
However, studies of fibroblasts that lack dynamin, i.e. dynamin 1 and 2 double KO (DKO) fibroblasts, revealed defects in the organization of circular dorsal ruffles induced by growth factors (Hongying Shen, Shawn Ferguson and Pietro De Camilli, unpublished observations), but failed to reveal prominent structural changes in several actin-dependent structures such as focal adhesions, lamellipodia and stress fibers [1,23]. The major actinrelated effect observed in a first analysis of these cells was a prominent accumulation of actin and actin-nucleating proteins at the necks of arrested endocytic clathrin coated buds [1,23]. These findings supported a close link between actin and the endocytic function of dynamin and also revealed that at least at these sites the recruitment and polymerization of actin is not dynamindependent. However, the role of dynamin in podosomes or invadopodia remains elusive.
Podosomes and invadopodia are adhesive mechanosensory modules composed of a dense F-actin core surrounded by a ring of adhesion molecules (Albiges-rizo et al., 2009). The distinction between podosomes and invadopodia is still a matter of debate. Since the exact relationship between invadosomes, podosomes and invadopodia is unclear, degradative F-actin-rich structures induced by v-Src will be grouped here under the term invadosomes [24,25,26]. In this report, we have investigated the function of dynamin in invadosomes, focusing in particular on the so-called rosettes, small ring-shaped adhesion structures that result from the spontaneous self-assembly of individual invadosomes and their surrounding F-actin cloud. Such structures offer interesting advantages as a model to study the role of dynamin in actin dynamics. They are very prominent, and thus easily visualized. Their 3D ultrastructure is well characterized and composed of long columns of tightly bundled F-actin filaments perpendicular to the substratum (actin cores representing individual invadosomes), which are connected by an actin cloud composed of radial F-actin filaments parallel to the substratum [27,28]. Rosettes are highly dynamic structures, which form, expand in diameter, often fuse with each other and disappear due to continuous remodeling resulting from the coordinated assembly of new invadosomes at their outer rim and disassembly of older ones at the inner rim [29,30,31]. In osteoclasts their fusion and expansion leads to the formation of a characteristic podosome peripheral belt and sealing zone on bone [29,30,32,33].
To study the role of dynamin at invadosomes, we have explored the formation of rosettes in dynamin 1 and 2 double knockout mouse embryonic fibroblasts (MEF) by the expression of a constitutively active mutant of the tyrosine kinase Src, v-Src, as previously described [13,34,35]. Our results demonstrate that the formation of invadosome rosettes is blocked in the absence of dynamin and that formation of invadosomes requires both a functional GTPase module and the assembly properties of dynamin. Furthermore, minute-scale loss-of-function of dynamin achieved by photo-inactivation rapidly disrupted invadosome rosettes, while leaving the formation of other structures such as focal adhesions, lamellipodia, or stress fibers unaffected. Thus, dynamin plays an important role in defining the architecture and dynamics of the actin cytoskeleton at invadosomes.

Dynamin and actin in the dynamics of invadosomes in MEF-v-Src cells
It was previously shown that a dominant negative mutant of dynamin 2 (Dynamin K44A) impairs invadosome formation [13,15]. However, dominant negative interference functions, at least in part, by sequestering binding partners of the mutant proteins. Determination of the function of a protein additionally requires studies of the effect of its deletion. To identify the role of dynamin in these structures, we analyzed invadosomes in cells genetically depleted of dynamin. Mammalian genomes contain 3 dynamin genes that encode dynamin 1, 2 and 3, respectively. Dynamin 2 is ubiquitously expressed and represents the major isoform expressed in most non-neuronal cells, while dynamin1 and 3 are expressed primarily in the nervous tissue [36,37]. To generate dynamins-depleted cells in our experiments, we used MEFs derived from Dnm1 2/2 ; Dnm2 LoxP/LoxP mouse embryos [1]. Thus, these cells only lacked dynamin-1, i.e. single KO (SKO MEFs, Dnm1 2/2 , Dnm2 LoxP/LoxP ) and were used to generate a dynamin1/dynamin 2 double KO genotype upon Cre recombinase expression (DKO MEFs ie Dnm1 2/2 ;Dnm2 2/2 ). Invadosome rosettes were induced in these cells by expressing v-Src ( [13,35,38], the constitutively active form of the proto-oncogene c-Src in order to obtain SKO-v-Src-MEFs and DKO-v-Src-MEFs (Dnm1 2/2 ;Dnm2 2/2 [13,35,38]. As previously observed for the localization of endogenous and transfected dynamin in v-Src-expressing wild type fibroblasts [13,15], Dyn2-pTRFP expressed in v-Src-Dnm1 2/2 ,Dnm2 LoxP/LoxP cells accumulated in numerous spots. Some of these spots formed a ring structure (Fig. 1A, red dashed circle in the zoom) that colocalized with actin (Fig. 1B). Furthermore, when MEFs were grown on a gelatin-Or.Green-containing substrate, these rings colocalized with areas of digested gelatin and their expansion occurred concomitantly with the increase of the digested area (Fig. 1A). Thus these rings have the typical characteristics of invadosomes (Fig. 1A, movie S1).
Rosette expansion involves assembly of new invadosomes at the outer edge of the ring and depolymerization at its inner edge [30,35]. Dyn2-pTRFP (Fig. 1A), like endogenous dynamin 2 ( Fig. 2A), was observed throughout the ring in Dnm1 2/2 ; Dnm2 LoxP/LoxP cells, but with a sharp accumulation at its inner edge (Fig. 1B). This intensity profile was similar to that of paxillin (paxillin-GFP) (Fig. 1D), a protein known to be involved in invadosome disassembly (Fig. 1D) [30]. The diffuse pool of dynamin colocalized with GFP-actin and GFP-cortactin throughout the ring (Fig. S1B, C, D and movie S2, S9 and S10). Thus, dynamin is associated with dynamic actin networks in invadosomes, particularly at the inner rim of rosettes, an area where disassembly of these structures is taking place.

Dynamin is essential for invadosome formation
Suppression of dynamin expression in SKO-v-Src-MEFs, was achieved by expression of Cre-recombinase through adenoviral transduction. This procedure resulted in a cell population that at 4 days post-transduction had dramatically lower levels of dynamin 2 expression [the residual dynamin 2 most likely reflects cells where recombination did not occur [1] and no detectable expression of dynamin 1. In these DKO-v-Src-MEFs cells, both isolated invadosomes and invadosome rosettes were almost absent ( Fig. 2A, C; movie S3). In contrast, cells that only lacked dynamin 2, SKO-v-Src-MEFs formed invadosomes (data not shown) indicating that the endogenous expression of dynamin 1 expression is sufficient to compensate the functions of dynamin 2, as shown recently for clathrin mediated endocytosis [1]. These results show an essential function of dynamin in the formation and maintenance of these characteristic actin-based structures.
Based on these results, all subsequent experiments aimed at exploring the function of dynamin in invadosomes were carried out exclusively in DKO-v-Src-MEFs. Importantly, even though the DKO-v-Src-MEFs cells did not form invadosomes, they still generated other actin based structures such as stress fibers, lamellipodia, ruffles, and the previously described actin spots at arrested endocytic clathrin coated pits [1], which are clearly evident in Fig. S1B.

Dynamin is essential for invasion
We next sought to determine whether the loss of invadosomes upon dynamin deletion is associated with a defect in the ability of cells to degrade extracellular matrix. To test this, DKO-v-Src-MEFs expressing VASP-RFP to visualize invadosomes, were grown on a digestible layer of gelatin-Or.Green. Live cell imaging revealed that only SKO-v-Src-MEFs cells digested the extracellular matrix at invadosome contacts (indicated by white arrow), whereas DKO-v-Src-MEFs cells did not (Fig. 3A, B), despite the presence of constitutively active Src in these cells. Quantification of the gelatin digested per cell confirmed the importance of dynamin in the ECM degradative capacity of these cells (Fig. 3C).
To determine the importance of dynamin in the general context of invasion, the invasive properties of SKO-v-Src-MEFs and DKO-v-Src-MEFs cells were analyzed. For that purpose, SKO-v-Src-MEFs and DKO-v-Src-MEFs expressing either GFP-actin (as a control of infection) were harvested, seeded above a matrigel layer on the top of a Boyden chamber and allowed to migrate towards the bottom of the transwell containing a chemoattractant for 20 h. Next, cells that did not invade the matrigel layer were removed and cells that passed through the filter of the Boyden chamber were fixed and stained with crystal violet. In agreement with a defect in extracellular matrix degradation, dynamin depletion inhibited the invasive properties of DKO-v-Src-MEFs through the thick layer of matrigel (Fig. 3C). Note that even though the migration of DKO-v-Src-MEFs cells in matrigel was largely reduced, their ability to form lamellipodia and to move was not entirely blocked but simply reduced by less than 50%, as shown by a monolayer wounding test (Fig.S2). Thus, the defective invadosome formation observed in DKO-v-Src-MEFs is associated with an inhibition of their invasive properties. Invadosomes are dependent on dynamin's catalytic and assembly activities It was of interest to determine whether the same properties that are key to the endocytic function of dynamin are also essential for its role in invadosome formation and dynamics. To address this question, we performed rescue experiments by expressing wild type or mutant dynamin 1 or 2 fused to an HA epitope tag or GFP in DKO-v-Src-MEFs and determined the percentage of GFPexpressing cells with invadosomes. Expression of GFP-actin in DKO-v-Src-MEFs was used as a control and did not restore the formation of invadosomes. Expression of either dyn2-GFP ( Fig. 4A.), HA-dyn2 or HA-dyn1 (Fig. 4C) restored the formation of invadosomes in DKO-v-Src-MEFs cells (Fig. 4), while mutant dynamins harbouring mutations that affect its GTPase cycle by decreasing the affinity for GTP -dyn2-GFP S45N and T65A mutants [39] -did not (Fig. 4A, B). The K688A mutation in dynamin 2, which in human dynamin 1 (L694A) greatly impairs self-assembly in vitro and in vivo [39,40,41,42], produced only a minimal rescue of invadosome formation (Fig. 4A) No rescue was produced by the pleckstrin homology (PH) domain K562E mutant of dynamin 2 that corresponds to a centronuclear myopathy patient mutation [43]. Importantly, all these mutants were still recruited to actin spots (Fig. 4B), that represent in part arrested clathrin coated pits (Fig. S1A,B), speaking against a defect due to inefficient targeting to the actin cytoskeleton. In contrast, dyn2-GFP and dyn1-HA deleted of their C-terminal Pro-Rich Domains (PRD region) neither restored invadosomes, nor were localized to cytosolic actin spots, demonstrating the importance of this region, which contains binding sites for cortactin and actin regulatory proteins, for dynamins targeting [21,44]. Beside there is no significant difference between dyn2-GFP and dyn2-HA to rescue invadosome formation, it is interesting to notice that a statistical difference was seen between dyn2-HA and Dyn1-HA in rescue experiments. Analysis of a series of dynamin 1 constructs with progressive C-terminal truncations demonstrated that the region following amino acid 816 of the PRD is critical for the restoration of invadosomes in DKO-v-Src-MEFs (Fig. 4C). We conclude that, as in the case of the function of dynamin in endocytosis, the coordinated actions of all main domains of dynamin are needed for its role in invadosome formation.

Invadosome disruption by photoinactivation of dynamin
The striking effect of the lack of dynamin on the formation of invadosomes could be due to long-term effects and adaptive changes in signalling pathways. However, the analysis of phosphoproteins that reflect the activation state of the main signalling pathways that regulate invadosomes (cortactin and Src activation) did not reveal major differences (except for an increase of Erk activity) between SKO-v-Src-MEFs MEFs, DKO-v-Src-MEFs and DKO-v-Src-MEFs re-expressing Dyn2-GFP (Fig. S3A). The use of Erk inhibitor did not rescue invadosome formation in DKO-v-Src-MEFs, suggesting that the increase of this signalling pathway is not responsible for the observed phenotype (Fig. S3B). To further investigate the role of dynamin in invadosome formation, and to complement experiments in cells that chronically lack dynamin, we developed a strategy to disrupt dynamin function at the minute time-scale by using photoinactivation.
To this end, the GFP moiety of dyn2-GFP was replaced by the photosensitizer Killer Red TM (KR) protein which is able to locally produce reactive oxygen species (most likely singlet oxygen, Roy et al., 2010) in response to red light irradiation, and thus to inactivate the protein fused to KR [35,45]. Dyn2-KR was expressed in DKO-v-Src-MEFs along with paxillin-GFP, a marker of adhesion structures. The ability of Dyn2-KR to functionally replace WT dynamin was confirmed by its localization at invadosomes and by its ability to rescue invadosome formation in DKO-v-Src-MEFs (Fig. 5A).
After 45 s irradiation with red light, invadosomes became disorganized (dashed line) without any change in focal adhesions (red arrows, Fig. 5A and movie S4). After 10 minutes cells resembled DKO-v-Src-MEFs cells due to the absence of invadosomes (Fig. 5A,F). That local OS production disorganizes invadosome structure specifically through inactivation of dyn2-KR was confirmed by the observation that a 45 s irradiation of SKOv-Src-MEFs (i.e. cells that express endogenous dyn2) also expressing GFP-paxillin and large amounts of free KillerRed did not have any effect on invadosomes (Fig. 5B, Fig. S4 and movie S5). Toxicity of the 45 s light irradiation itself was tested by replacing KR with pTRFP, a more photostable fluorescent protein with an identical excitation/emission spectrum as KR, which did not induce invadosome disorganization after irradiation, (Fig. 5C, F and movie S6). Moreover, the KillerRed strategy had no effect on endogenous proteins, more precisely on non-tagged dynamin as irradiation of SKO-v-Src-MEFs expressing dyn2-KR did not induce invadosome disorganization (Fig. 5D, F and movie S7). Confirming this result, photoinactivation of dyn2-KR in DKO-v-Src MEFs expressing dyn2-GFP and dyn2-KR had no effect on invadosomes and on the localization dyn2-GFP at invadosomes (red arrow, Fig. 5E, Fig. S4 and movie S8).
The consequences of acute dynamin inactivation on the actin cytoskeleton were further assessed in DKO-v-Src-MEFs expressing both GFP-actin and dyn2-KR. While the invadosome ring, as detected by the GFP signal, was disrupted a few minutes after dynamin photo-inactivation (Fig. 6A), the photo-inactivated cell was still able to spread, to develop lamellipodia (red arrows, Fig. 6A, C) and to accumulate actin spots, suggesting no drastic interference with the actin cytoskeleton. As shown above, photoinactivation of dyn2-pTRFP, which is expected to produce less reactive oxygen species, has no effect on invadosomes (Fig. 6B). The rapid disruption of invadosomes upon dynamin depletion (few minutes) contrasts with the subsequent formation of lamellipodia (after 15 min) confirming the direct and specific function of dynamin on actin structures at invadosomes. In a separate experiment, dyn2-KR inactivation did not block the formation of filopodia either (green arrow, Fig. 6C) further showing that the organization and architecture of these structures are also independent of dynamin. Thus, dynamin is essential for the minute-to-minute maintenance of invadosomes.

Discussion
Invadosomes and invadopodia, collectively now called invadosomes [24,25,26] are actin-driven structures that coordinate attachment to the extracellular matrix (ECM) with its degradation [24,31,46]. We show here that invadosomes formation is impaired in cells that lack dynamin. Furthermore, acute disruption of dynamin function by photo-inactivation led to the rapid disappearance of pre-existing invadosomes. Thus, the GTPase dynamin plays a critical importance both in the formation and maintenance of invadosomes.
Based i) on the presence of dynamin at such structures, ii) on the frequent occurrence of tubular plasma membrane invaginations within invadosome/invadosome rosette and iii) on the property of dynamin to tubulate membranes, it had originally been proposed that dynamin was responsible for the tubular invaginations, possibly linking them to their actin shell [13]. Supporting this interpretation was the similar diameter of these tubules to the diameters of tubules generated by purified dynamin in the presence of lipid bilayers in vitro [47,48]. However, it was subsequently found that dynamin is not restricted to the membrane interface at invadosomes but that is localized throughout the actin core and the surrounding actin cloud [14,49]. Furthermore, tubules such as those found in invadosomes can be generated by several other proteins [50,51,52], including dynamin binding proteins concentrated at invadosomes [53,54,55,56].
Dynamin is now known to interact directly or indirectly with a variety of actin regulatory proteins thus explaining its presence in and around invadosomes [13,14,15], as well as in a variety of actin-rich structures, primarily those nucleated by the Arp2/3 complex and WASP/Wave family proteins [8,9,10,18,57]. Interestingly, several of the numerous proteins that bind the prolinerich C-terminal region of dynamin also bind proteins that have a role in actin nucleation. Examples include, but are not limited to, cortactin, Tuba and proteins of the FBP17/Toca/CIP4 family [19,20]. Some of these proteins were first identified for their function as actin regulatory factors [58,59] and only subsequently were found to bind dynamin and to have a potential role in endocytosis. Others were first identified as endocytic factors and later reported to have a role in actin regulation [11,50,60]. A direct interaction of dynamin with actin has also been recently reported [22]. While our findings add new evidence for a close link between dynamin and actin and also conclusively establish the importance of dynamin for invadosome/invadosome function, the mechanism(s) that mediate dynamin's action at these sites remain an open question.
One possibility is that the importance of dynamin for invadosome dynamics may be related to its endocytic role, in spite of lack of clear evidence for intense endocytic activity occurring at invadosomes. For example, since actin has been implicated in several forms of endocytosis, the binding of dynamin to actin and actin regulators may help to concentrate dynamin at endocytic sites and thus to generate the focal concentration of dynamin that is required for its polymerization once the neck of an endocytic pit is generated. In this scenario, which is supported by the concept that membrane remodeling is the key function shared by all members of the dynamin superfamily [61,62,63], the interaction of dynamin with actin would simply play an ancillary role to its endocytic function. However, dynamin depletion led to a general disassembly of actin at invadosome sites whereas the absence of this GTPase is associated with an intense actin polymerization at arrested clathrin-coated pits [1]. This speaks against the possibility that dynamin inactivation may produce the arrest of endocytic intermediates at invadosomes. Furthermore, if the importance of dynamin in invadosome dynamics was simply an indirect consequence of its function in endocytosis, one would expect a delay between dynamin inactivation and disappearance of the invadosome. While conditional Cre-recombinase-dependent KO experiments, which require days for the disappearance of the protein, are compatible with such a delay, the dramatic and very rapid (tens of seconds/minutes) effects of the acute photoinactivation of dynamin speaks against this possibility and favor a direct role of dynamin in the 3D architecture of the actin meshwork at invadosomes.
A related possibility is that the role of dynamin in invadosome dynamics may reflect the effect that endocytosis of receptors has on their signaling output. Lack of dynamin may disrupt normal signaling by preventing the internalization of receptors (including integrins) with an impact on signaling that may occur both at the cell surface and on endosomes. While our results do not provide evidence for this mechanism, this remains a plausible scenario that may contribute to other dynamin functions at invadosomes.
Yet another possibility is that dynamin may have independent actions in endocytosis and in actin dynamics. Its function at invadosomes may be independent of its endocytic function since, based on current knowledge, invadosomes do not appear to have a predominant endocytic role. One could speculate that the proline-rich domain of dynamin, i.e. the portion of the protein that links it to most actin regulatory proteins, may have a function at least partially independent from the portions of the protein involved in endocytosis. However, our experiments with mutant dynamin show that all dynamin domains, including the GTPase domain and it catalytic activity, are needed for the action of dynamin at invadosomes. Recent crystallographic and cryo-EM studies [64,65,66,67] have shown how the functions of the different domains of dynamin are interdependent on each other. Such studies have shown that the polymeric helical organization of dynamin around a tubular template and the interaction of different rings of the spiral with each other are needed to promote the GTPase activity that is responsible for dynamin's action. While such a helical organization has clear relevance for the membrane fission activity of dynamin at the tubular neck of an endocytic pit, its significance in other contexts, such as in the regulation of actin function, remains unclear. Wild type dynamin can oligomerize into a spiral around microtubules [68] and actin bundles [21] in vitro, but it remains to be seen whether this property has a function in living cells. No transient hot spots of dynamin that may correspond to transient formation of ring-like structures, similar to the hots spot of dynamin visible at endocytic sites, can be seen in the actin clouds that surround invadosome cores.
Based on all these considerations, a direct role of dynamin in the regulation of actin at invadosomes remains plausible. However, the precise mechanism through which dynamin may achieve its effects at these structures remains a most interesting priority for future studies. Elucidating such mechanisms will have important implications because the recent developments of drugs that block dynamin activity opens the possibility of targeting its function in pathological conditions where invadosomes/invadopodia are implicated [11,69].

Cell culture and infection
Dynamin 1 2/2 , dynamin 2 flox/flox fibroblasts were isolated from mouse embryos (embryonic day 13) and immortalized by serial passaging as described recently [1]. Yale Cancer Center Animal Genomics Shared Resource approved this study allowing to generate these cells that we used lately. Indeed, animal care and use was carried out in accordance with our institutional guidelines (Yale Cancer Center Animal Genomics Shared Resource).
Cells were grown in DMEM (containing glutamine)+10% fetal bovine serum+1% penicillin/streptomycin supplement. Cre recombinase expression was achieved via adenovirus trandusction and was purchased from the University of Iowa Gene Transfer Vector Core (Iowa City, IA). Cells were generally used for experiments between 4 and 6 days after Cre recombinase delivery. For rescue experiments, the cells were firstly transduced with exogenous dynamin constructs before addition of the CRE recombinase to remove the endogenous form of the GTPase. In most experiments, cDNAs were delivered via a retroviral transduction following packaging in Phoenix-Eco cells or Phoenix-Ampho cells (ATCC, Manassas, VA). Supernatant containing viral particles from such cells was harvested, 0.45 mm filtered and following addition of 8 mg/ml polybrene (Sigma, St Louis, MO) was used to transduce fibroblasts. The cells were serum starved for 14 h prior to experiments in order to help maximize the contribution of v-Src to overall cellular signalling.

Degradation and invasion test
Coverslips were coated with gelatin-OregonGreen (1 mg/ml), fixed with PFA 4% (g/vol)+0.5% glutaraldehyde for 30 min at 4uC. Then, coverslips were washed with PBS and sodium borohydride (30 mg/ml), sterilized with ethanol 70% and rinsed with PBS 16. Cells were seeded on these coverslips in culture medium and imaged for 14 h before fixation. Degradation was measured by subtracting the fluorescence intensity of the images of the gelatin layer at the beginning and end of the movies, and then dividing the total amount of pixel intensity (Intensity AU, corresponding to digested area) per the number of cells present in the field. Our invasion assay is based on the addition of a layer of BD Matrigel TM (Basement Membrane Matrix, High Concentration (HC), LDEV-Free, BD, Franklin Lakes, NJ) on a transwell permeable support (8 mm pores) (Corning Costar Corp., Corning, NY). 100 000 cells were harvested, seeded on top of the Matrigel layer in DMEM without serum and attracted towards the bottom of the transwell for 20 h by the addition of DMEM supplemented with 1% serum in the lower chamber. Cells that did not pass through the matrigel layer and through the upper part of the insert were removed. Only cells that passed through the filters were fixed and stained with crystal violet.
For live imaging, cells were seeded at sub-confluent densities into serum-coated 35 mm glass bottom (thickness = 0.17 mm) dishes (Mattek, Ashland, MA) and allowed to grow for 12 to 48 hours prior to imaging. The dishes were then transferred to CO2 independent medium (Invitrogen), placed on a 37uC heated stage (Carl Zeiss Microimaging, Inc., Thornwood, NY) and imaged with the same Zeiss Axiovert 200M microscope set-up previously described. For photoinactivation experiments, we used Zeiss Axiovert 200M with a 636 (NA 1.4) Plan Apochromat to illuminate with a triple filter set 25E with individual excitation filters (405, 495, or 575, HE), beam splitter (TFT 435+510+600, HE), emission (TBP 460+530+625, HE) (Carl Zeiss microimaging GmbH, Gottingen, Germany). Cells were imaged as for live imaging before imaging was paused to allow illumination for 40 to 50 s (45 s in average) with a 100W HBO mercury lamp (100% power). Imaging Series 7.0 (Universal Imaging, Downington, PA) was used to mount .avi movies from image stacks. Extracted images from stacks were processed with Adobe Photoshop CS2 and Adobe Illustrator CS (Adobe Systems, San Jose, CA) and Image J (http://rsb.info.nih.gov/ij/). Significance of the differences between standard deviations was analyzed in Excel with an F-test. Figure S1 Actin structures induced by dynamin depletion in DKO-v-Src-MEFs cells colocalize partially with clathrin light chain and present different cortactin dynamics. A) Chlathrin light chain staining (in green) revealed that clathrin coated pits are formed around invadosomes (stained by F-actin, red, and phosphotyrosine, blue) in SKO-v-Src-MEFs. B) Dynamin depletion in DKO-v-Src-MEFs cells increased the number of arrested endocytic clathrin coated pits (green arrows) which accumulates F-actin and numerous phosphorylated proteins on tyrosine as revealed by phalloidin (red) staining and antiphosphotyrosine (blue) in higher magnifications of the areas in the white squares. On the contrary, larger actin spots are also containing phosphorylated proteins on tyrosine but are not associated with clathrin light chains (red arrows). Scale bar = 3 mm