A signalling cascade involving PKC, Src and Cdc42 regulates podosome assembly in cultured endothelial cells in response to phorbol ester

Spatial and temporal coordination of actin polymerisation plays a crucial role in many fundamental cellular functions such as cell adhesion, migration and morphogenesis. In most cells, adhesion occurs through focal adhesions and focal complexes. However, adhesion can also occur through the podosome, an uncommon actin-rich structure which is described as a columnar actin-rich core that stands perpendicular to the plane of the ventral plasma membrane. Many proteins are associated with the F-actin core, including structural proteins, integrins and regulatory proteins such as Ser/Thr kinases, tyrosine kinases and RhoGTPases (Linder and Aepfelbacher, 2003). In contrast to focal adhesions, podosomes are found in a restricted number of cell types. They form spontaneously in monocyte-derived hematopoietic cells including osteoclasts, macrophages and immature dendritic cells. These cells share an intrinsic ability to cross anatomical boundaries and podosomes are therefore thought to support a specialised mode of motility. However, podosomes have now been discovered in non-hematopoietic normal cells, in smooth-muscle cells (Hai et al., 2002) and endothelial cells (Moreau et al., 2003; Osiak et al., 2005). In these latter situations, podosomes are not induced by adhesion, but appear in response to sustained activation of certain signalling components (see below) or by serum factors. More recently, podosome-like structures have been found in epithelial cells (Spinardi et al., 2004). These results underscore the previously unrecognised complexity of these actin-based structures in terms of molecular composition and regulation and raise the question of the function that podosomes may have in these non-hematopoietic cells.

Several signals have been described as stimulating podosome assembly in different models that do not spontaneously form podosomes. (1) The first pathway leading to podosome assembly involves direct activation of Src kinase. Podosomes were initially described in fibroblasts transformed by the Rous-sarcoma virus encoding the v-Src oncogene (Tarone et al., 1985). (2) A second pathway leading to podosome assembly involves direct activation of RhoGTPases. A constitutively active mutant of RhoA (V14RhoA) induces the de novo formation of podosomes in osteoclasts (Chellaiah et al., 2000), whereas V12Cdc42 induces podosomes in HeLa epithelial cells or aortic endothelial cells (Dutartre et al., 1996; Moreau et al., 2003). (3) Phorbol esters represent another means of inducing podosomes as reported in U937 promyelocytic cells and smooth-muscle cells (Gaidano et al., 1990; Hai et al., 2002). This agent mimics diacylglycerol (DAG), the physiological activator of most protein kinase C (PKC) family members. PKCs are Ser/Thr kinases present in a primed yet inactive conformation in the cytosol, which are translocated to the membrane upon activation. The family splits into three classes, among which two are activated by DAG or phorbol esters: the `conventional' PKCs (PKCα, β and γ), which also require Ca<sup>2+</sup> signals, and the `new' PKCs (PKCδ, ϵ, η and θ), which are calcium-independent. It has been proposed that PKCα is involved in podosome formation in murine embryonic fibroblasts and in smooth-muscle cells (Gatesman et al., 2004; Hai et al., 2002). Whatever the model, the minimal essential components in the signal transduction pathway leading to podosome formation are members of Src tyrosine kinase family and members of the Rho family of GTPases. Yet, the role and positioning of PKC in these pathways remains largely unknown.

Among the numerous components present in podosomes, matrix metalloproteases (MMPs) are of interest for podosome function. This zinc-dependent family of enzymes can be divided into two structurally distinct groups, secreted MMPs and membrane-type MMPs (MT-MMP). The membrane-type-1 MMP (MT1-MMP, also known as MMP-14) and the soluble gelatinase MMP-2 have been associated with podosome structures in osteoclasts (Delaisse et al., 2000; Sato et al., 1997). Besides its own degradation activity (Chun et al., 2004), MT1-MMP can process inactive pro-MMP2 into active MMP2 at the external plasma membrane (Sato et al., 1994). Accordingly, the presence of podosomes in osteoclasts has been associated with extracellular-matrix (ECM)-degradation and -invasion.

We recently reported that a constitutively active form of Cdc42 induces the formation of podosomes in endothelial cells (Moreau et al., 2003). Given the established role of PKCα in the induction of podosomes in smooth-muscle cells, we designed experiments to explore whether PKC may regulate podosome assembly in endothelial cells. A constitutively active form of PKCα could mimic PMA in podosome induction. We also show here for the first time a regulatory role of PKCδ in podosome assembly. We describe that Src and Cdc42 act downstream of PKCs in the process of podosome formation in endothelial cells. PMA-induced podosomes are sites of matrix metalloprotease accumulation and ECM degradation. Our data show that MT1-MMP, which is necessary for MMP2 activation, confers major invasive activity at podosomes in endothelial cells. In the vessels, the endothelium is critically involved in the regulation of multiple physiological processes. Most of them involve cytoskeletal dynamics in the control of adhesive interactions with neighbouring cells or with ECM. We hypothesise that podosomes are involved in degradation of the subendothelial matrix thus contributing to endothelium remodelling during pathophysiological processes.

We previously reported that a constitutively active mutant of Cdc42 induced podosome assembly in the model of porcine aortic endothelial (PAE) cells (Moreau et al., 2003). To investigate this finding further, we examined the ability of endothelial cells from either aortic or veinous origin to form podosomes in response to signals known to induce their formation in other models. PAE cells or human umbilical vein endothelial cells (HUVECs) were transfected with a constitutively active form of either Src, Cdc42 or RhoA, or stimulated with PMA. Cells were then fixed and processed for immunofluorescence. Fig. 1A shows that expressing V12Cdc42 triggered podosome formation in HUVECs. Podosomes, identified by the F-actin-vinculin arrangement, were found well individualised and scattered over the ventral membrane, a pattern similar to that obtained with PAE cells (Moreau et al., 2003). By sharp contrast, V14RhoA promoted the formation of stress fibres (Fig. 1A). HUVECs and PAE cells transfected with v-Src also showed podosomes, presenting a vinculin ring around the F-actin core and a spatial distribution similar to those formed with V12Cdc42. It should be noticed that in both cases cells lose their stress fibres and gain podosomes (Fig. 1B). When PAE cells and HUVECs were treated with PMA, podosomes were also induced. In this situation, podosomes were found either as individual podosomes similar to those induced by active Cdc42 or Src, or clustered in rosettes as described in Rous-sarcoma-virus-infected cells (Gavazzi et al., 1989; Tarone et al., 1985) (Fig. 1C).

In contrast to transfection protocols with v-Src or V12Cdc42, PMA treatment provides an opportunity to study cytoskeleton rearrangements leading to podosome formation in a time-course experiment. HUVECs were treated with PMA for various periods of time, fixed and labelled with rhodamine-phalloidin to visualise F-actin (Fig. 2A). Untreated HUVECs showed typical actin stress-fibres going across the cell body. Actin remodelling was detected shortly after PMA addition. Strong actin staining was seen at the leading edges of some cells 15 minutes after PMA exposure and increased in a time-dependent manner (Fig. 2A). After 30 minutes of treatment, most cells showed increased membrane ruffling activity, whereas stress fibres were found strongly reduced. Concomitantly, actin dots became visible either behind the leading edges or clustered within the cell body (Fig. 2A,B). Over time, dots continued to form and at 1 hour, the response was maximal with about 25% of cells showing podosomes and rosettes (Fig. 2C). Actin dots then progressively disappeared and at 3 hours of treatment, stress fibres had returned. When quantification of cells showing either one of these two kinds of patterns was made, the number of rosettes was found to be at a maximum in cells after 1 hour (peak of the response) (Fig. 2C). Extending incubation time revealed that cells displayed no more rosettes after 3 hours of treatment, whereas the percentage of cells with podosomes remained unchanged at that time, indicating that podosome organisation into rosettes is a transient phenomenon.

To characterise the actin dots and rosettes formed in PMA-treated HUVECs, the presence of structural and signalling components associated with podosomes in other models was investigated. Structural proteins, such as scaffold proteins, actin-binding proteins, actin-regulating proteins or integrins, and regulatory proteins, such as Cdc42 or protein kinases, were all found in HUVEC-podosomes (Table 1). No difference could be detected in terms of protein composition between individual podosomes and rosettes. Taken together, these data confirm that, in addition to aortic endothelial cells (Moreau et al., 2003), bona fide podosomes can also form in endothelial cells from veinous origin in response to various inducers, and also establish that podosomes are not restricted to cells from monocyte lineage.

The primary targets of PMA are PKCs, and PKCs regulate podosome formation in response to phorbol esters in smooth-muscle cells (Hai et al., 2002). Based on this work carried out in vascular cells, we explored PKCα and PKCδ regulation in HUVECs at the PMA concentration that induces podosomes. PKCα and PKCδ were found translocated from cytosol to membrane after minutes of treatment (Fig. 3). PKCα and PKCδ associated with the membrane fraction were found phosphorylated at their autophosphorylation sites Ser657 and Ser643, respectively, reflecting kinase activity (Fig. 3). In addition, a decrease in the total amount of PKCα was visible after 1 hour of PMA treatment, consistent with PKCα degradation that follows its activation. For PKCδ, degradation occured after 3 hours of treatment. Because `non-kinase' phorbol ester receptors such as chimaerins (a family of RacGTPase activating protein) or Ras GRPs (exchange factors for Ras/Rap1) are also linked to the actin cytoskeleton (Brose and Rosenmund, 2002), we investigated whether PKC kinase activity was required for PMA-induced podosome formation in HUVECs. To this end, isotype-restricted or isotype-selective inhibitors of PKC were used. HUVECs were pre-treated with either 50 nM Gö6976 or 3 μM rottlerin, which specifically inhibit conventional PKCs or the novel PKCδ isotype, respectively. As shown in Fig. 4A, both treatments suppressed podosome assembly, suggesting that at least two classes of PKC were involved in podosome formation in HUVECs. In addition, pre-treatment with Gö6976 abrogated not only podosome formation, but also stress-fibre disassembly in response to PMA in HUVECs (Fig. 4A). Since stress fibres also dissolve upon inhibition of PKCδ by rottlerin, we concluded that stress-fibre disassembly required conventional PKCs but not PKCδ activities. By contrast, podosome- and rosette-formation required both activities as revealed by quantification of isotype-selective inhibitor effects (Fig. 4B). To confirm the data obtained by using the pharmalogical approach and to analyse the individual role of PKCα and PKCδ in podosome assembly, we used specific small interfering RNA (siRNA) to inhibit PKCα and PKCδ expression (Fig. 4C). As shown in Fig. 4D, the two siRNA PKCα and the two siRNA PKCδ inhibited podosome formation in response to PMA. Moreover, PKCα or PKCδ depleted cells were phenotypically indistinguishable from those obtained with Gö6976 and rottlerin treatment, respectively (data not shown).

To confirm the involvement of PKCα and PKCδ in podosome formation, we expressed constitutively active mutants of these isoforms and examined changes in F-actin organisation. Exogenously expressed wild-type PKCα or PKCδ did not alter the phenotype. Their localisation was found to be mainly cytosolic or associated with ruffles containing F-actin (Fig. 5A). Treatment of such transfected cells with PMA did not induce PKC translocation to podosomes (data not shown). When the constitutively active form of PKCα (PKCαA25E) was expressed, cells changed their morphology (Fig. 5B). They tended to spread and loose stress fibres, and formed lamellipodia and podosomes in a manner similar to that observed under PMA treatment. It should be noticed that podosomes formed upon expression of the active form of PKCα do not contain detectable GFP-PKCα (Fig. 5B). The same experiment performed with active PKCδ (PKCδA147E) produced small elongated cells, devoid of stress fibres and with a strong F-actin staining at the cell cortex (Fig. 5B). In contrast to PKCαA25E, PKCδA147E did not induce podosomes, suggesting that PKCδ activation is necessary but not sufficient to trigger podosome assembly in HUVECs. Taken together, these results confirm that both conventional and novel PKCs are involved in podosome assembly in endothelial cells and, for the first time, establish a role for PKCδ in podosome formation.

Since podosome formation was observed in response to PKC- and Cdc42-activation, we explored a possible link between these pathways in HUVECs. Using the inhibitors Gö6976 and rottlerin, we examined whether a functional PKC was required downstream of Cdc42 for podosome assembly. V12Cdc42-induced podosome induction was found to be insensitive to conventional PKC- and PKCδ-inhibitors, although Gö6976 did not suppress stress fibres (Fig. 6A). By doing the reverse experiment, we examined the requirement for GTPases downstream of PKCs. To assess the individual role of each GTPase, we took advantage of RNA interference by using siRNAs against the three major RhoGTPases RhoA, Rac1 or Cdc42. The ability of siRNA to inhibit GTPase expression in HUVECs was first verified by western blot (Fig. 6B). In the same experiment, the actin cytoskeleton of PMA-treated cells was analysed and cells showing podosomes were quantified (Fig. 6C). Inhibition of Cdc42 synthesis with siRNA prevented podosome appearance in response to PMA. Similar inhibition was obtained when Cdc42 activation was prevented by expression of plasmids encoding GFP-tagged dominant-negative Cdc42 (data not show). Inhibition of RhoA expression resulted in a significant reduction in the number of cells showing podosomes, indicating that, in addition to Cdc42, RhoA was also required for podosome assembly in these cells. However, podosome formation in HUVE cells seems to be independent of Rac1 as absence of Rac1 protein did not affect the amount of podosomes per cell nor the number of cells showing podosomes. These experiments show that RhoA and Cdc42 are involved in PMA-induced actin reorganisation and podosome formation in HUVECs.

Podosome formation in HUVECs was also observed in response to increased Src activity, and we therefore explored a possible link between the PKC and Src pathways. The Src inhibitor PP2 inhibited PMA-induced podosomes, showing that Src is activated downstream of PKC in HUVECs cells forming podosomes in response to PMA (Fig. 7A). In addition, PP2 failed to inhibit V12Cdc42-induced podosomes in HUVECs (Fig. 7A) and in PAE cells (data not shown), suggesting that Src activation is either upstream of Cdc42 or in a Cdc42-independent but parallel pathway in the cascade leading to podosome formation. Since PAE cells show much higher transfection efficiency than primary HUVECs, the involvement of Cdc42 in the v-Src cascade leading to podosome assembly was assessed in PAE cells. v-Src was expressed in PAE cells, in which Cdc42 activation was prevented by co-expressing a dominant-negative form of Cdc42. Under these conditions, no podosome was detected in the majority of co-transfected cells (Fig. 7B). This experiment demonstrated a requirement for Cdc42 in podosome formation downstream of Src. These results suggest that, a linear cascade occurs where PKC activates Src, which in turn activates Cdc42 leading to podosome assembly in endothelial cells.

To explore this hypothesis further, activation of RhoGTPases in response to active Src was measured in a pull-down assay (Fig. 7C). Given that transfection of v-Src into HUVECs led to a very high level of cellular mortality, the experiment was performed with v-Src-transfected PAE cells. GST-fusion proteins containing the GTPase-binding domains of effectors - rhotekin for RhoA and the Cdc42/Rac-interactive-binding domain (CRIB) of PAK for Cdc42 - were coupled to agarose beads and used to precipitate GTP-bound GTPases from v-Src-transfected or control-PAE-cell lysates. Affinity-purified proteins were then analysed by electrophoresis, followed by western blotting with GTPase-specific antibodies. We found that PAE cells transformed with v-Src led to an activation of Cdc42 (Fig. 7D). Concomitantly, a small but reproducible downregulation of RhoA could be detected (Fig. 7D). From this experiment, we confirmed that Cdc42 is activated downstream of Src, in the process of podosome formation in response to PMA.

In v-Src-transformed fibroblasts and in some tumour cells, proteolytic activity has been found associated with podosomes (Chen, 1989; Mizutani et al., 2002). To determine whether endothelial podosomes are able to locally degrade ECM, we performed an in vitro matrix degradation assay as described by Bowden et al. (Bowden et al., 2001). HUVECs were seeded onto glass coverslips pre-coated with crosslinked gelatin that had been conjugated with FITC and treated for 1 hour with PMA. PMA-treated cells still displayed podosomes indistinguishable from those formed on glass coverslips, as judged by rhodamine-phalloidin staining (Fig. 8A). After PMA treatment, ECM degradation was visualised as dark areas in the fluorescent matrix (Fig. 8A). Most of the non-fluorescent patches colocalized with cells showing podosomes or rosettes. The same experiment was then performed on HUVECs transfected with v-Src. Again, gelatin degradation occurred in a punctate manner and most areas of proteolytic activity colocalized with v-Src-induced podosomes (Fig. 8A). As some sites of matrix degradation were not superimposable on actin dots, we believe that the cells moved, disassembled and reformed podosomes throughout the duration of the experiment. By contrast, V12Cdc42-transfected cells did not induce the formation of dark areas in the gelatin underneath podosomes, indicating that V12Cdc42-induced podosomes were not able to degrade the ECM in HUVECs (data not shown). These results show that Src-induced, but not Cdc42-induced, endothelial podosomes are endowed with ECM-degradation activity.

MMPs are the main actors in matrix degradation. To explore their contribution to ECM degradation, we performed the same experiment in the presence of synthetic metalloproteinase inhibitors. GM6001 was used as a broad MMP inhibitor, whereas TSRI265 was employed to target MMP2 activity more specifically. As shown in Fig. 8B, GM6001 and TSRI265 decreased the amount of degradation areas in the gelatin by about 95% and 50%, respectively, leaving actin cytoskeleton and podosome structures intact (data not shown). These results suggest the involvement of MMP2 together with another proteinase in the proteolytic function associated with podosomes. Gelatin zymogram assays were performed next to examine the regulation of MMPs by PMA in cells forming podosomes. Culture supernatants and cell extracts were run in gelatin-containing polyacrylamide gels. A clear band corresponding to gelatin degradation was visualised at the molecular weight of MMP2 after staining with Coomassie Blue (Fig. 9A). An additional band of lower molecular weight, corresponding to active MMP2 cleaved from pro-MMP2, was detected when cells were treated with PMA for 30 minutes. Immunofluorescence experiments revealed that MMP2 colocalised with actin at PMA-induced podosomes and rosettes (Fig. 9B). MT1-MMP, which can process pro-MMP-2 into active MMP-2, was also detected at these sites (Fig. 9B). To confirm the role of MT1-MMP in matrix degradation, we analysed the effect of MT1-MMP depletion in zymogram- and fluorescent-matrix-degradation assays. MT1-MMP siRNA efficiency on MT1-MMP knockdown expression was visualised by western blot (Fig. 10A). Under these conditions, cleavage of pro-MMP-2 into active MMP-2 did not occur when cells were treated with PMA (Fig. 10B). Moreover, transfection of MT1-MMP siRNA decreased by 97.3% ± 3.9%, the number of black holes in the matrix showing that MT1-MMP knockdown strongly inhibited ECM degradation. Simultaneously, we detected a 50% increase in the number of cells showing podosomes (Fig. 10C). Thus MT1-MMP is essential for proteolytic activity at PMA-induced podosomes and seems to play an important role in podosome function.

In this study, we show that podosomes form in primary human endothelial cells from veinous origin when treated with PMA. Recently, Osiak and co-workers have described the spontaneous formation of podosomes in subconfluent HUVECs (Osiak et al., 2005). In our experimental settings, untreated HUVECs did not present podosomes, even when grown subconfluently. This difference could be owing to either primary-cell preparation protocols or culture conditions. We describe podosome formation in HUVECs upon a short exposure to PMA. Upon this treatment, we show that 25% of HUVECs display podosomes whereas the remaining 75% do not do so. In all models where podosomes are inducible, only a fraction of the population exhibit podosomes at a given time. Hypotheses have been raised to explain this phenomenon, for example the presence of this kind of structure could be linked to the cell cycle as suggested for invadopodia (McHugh et al., 2004). Indirect data such as those obtained from matrix-degradation assays (cells showing no rosette but matrix-degradation areas underneath) or preliminary video-microscopy experiments suggest that most HUVECs can form podosomes in response to treatment with PMA. Phorbol-ester-induced changes of the cytoskeleton have been described in numerous models but few of them reported podosome induction. The first description was made in 1984 in BSC-1 cells (Schliwa et al., 1984). More recently, podosomes have been induced in smooth-muscle cells upon PDBu treatment (Hai et al., 2002). Both conventional and novel PKC isoforms are activated by phorbol esters. Using a pharmacological approach together with RNA interference, we found that both PKCα and PKCδ are required for PMA-induced podosome formation in HUVECs, whereas only conventional PKCs were found to be involved in the A7r5 smooth-muscle-cell model (Hai et al., 2002). Another difference with A7r5 cells was that we were unable to localise PKCα (nor PKCδ) at podosomes in HUVECs. These data suggest that, in HUVECs and smooth-muscle cells, phorbol esters do not regulate the formation of podosomes in the same way.

Moreover, we found that active PKCα but not PKCδ can trigger podosome formation in HUVECs. Since PMA induces podosomes whereas PKCδ activation does not do, podosome formation might require cross-talk between PKCδ and a member of the extended phorbol-ester-binding protein family, such as PKCα, chimearins or RasGRPs. Moreover, compared with the entire family of PKC enzymes, tyrosine phosphorylation is an important regulatory mechanism of PKCδ. Thus, another explanation could be that PKCδ needs to be tyrosine phosphorylated to be active in terms of podosome formation. Following this hypothesis, PMA has been shown to trigger PKCδ phosphorylation on Tyr187 (Li et al., 1996). The role of this alteration, which does not influence PKCδ kinase activity, is not yet known. Our results thus establish for the first time that PKCδ plays a role in podosome formation. However, its exact contribution remains to be elucidated. Importantly, the role of PKCδ in other podosome models needs to be investigated.

In this respect, cytoskeleton-associated proteins, such as vinculin and dynamin, that contribute to cytoskeletal reorganisation and to podosome formation, are substrates of PKCs (Parker and Murray-Rust, 2004). In HUVECs, activation of PKCs by PMA or expression of v-Src had similar effects on actin organisation, i.e. loss of actin stress-fibres and formation of podosomes. Although several studies have shown that activation of Src results in the subsequent activation of PKCα and/or PKCδ (Gschwendt et al., 1994; Zang et al., 1995), other studies have demonstrated that PKC can direct the activation of Src (Brandt et al., 2002; Brandt et al., 2003; Levi et al., 1998). Our data indicate that, in the process of podosome assembly, PKCs function upstream of Src (Fig. 11). Even if PKC has been shown to directly activate Src by phosphorylation of its Ser12 and Ser48 (Gould et al., 1985;Moyers et al., 1993), some reports have indicated that other proteins play a role in the ability of PKC to stimulate Src activation. Thus, protein tyrosine phosphatase PTPα and actin-filament associated protein AFAP-110 have been shown to be responsible for relaying signals from PKC to Src kinase in two independent studies (Brandt et al., 2003; Gatesman et al., 2004). Moreover, upon PKC activation, RACK1, the founding member of the family of receptors for activated C kinase, colocalises with Src at the plasma membrane and functions as a substrate, binding partner and inhibitor of Src (Chang et al., 2002). Among those proteins, both AFAP-110 and RACK1 have been shown to be involved in the regulation of podosome formation (Gatesman et al., 2004; Mamidipudi et al., 2004).

Downstream of Src, we found that RhoGTPases are involved in PMA- and v-Src-induced changes of actin-filament-integrity in endothelial cells. RhoA seems to play a role in this process because RhoA-depleted cells are much less responsive to PMA in terms of podosome assembly. Recently, Berdeaux and co-workers have demonstrated that RhoA[GTP] levels increase after transformation of NIH 3T3 fibroblasts with activated Src (Berdeaux et al., 2004). However, in our model, global RhoA activity, as detected by the pull-down approach, did not increase upon podosome formation. One probable explanation is that the observed response integrates RhoA inhibition, regulating stress-fibre-disassembly and local RhoA activation at podosomes. In fact, in endothelial cells, stress-fibre-disassembly occurs concomitantly with podosome formation. This balance between stress fibres and podosomes seems to be a feature of endothelial cells and some other cells that form podosomes in response to external stimuli, such as smooth-muscle cells. However, in this study, we show that V12Cdc42-transfected cells treated with the inhibitor Gö6976 exhibit podosomes together with stress fibres. This result indicates that generalised stress-fibre-dissolution in the whole cell is not a prerequisite in the process of podosome formation. In fact, it seems that a locally restricted stress-fibre-remodelling process occurs during podosome formation.

Using RNA interference or the dominant-negative form of the GTPase, we demonstrated that Cdc42 is required for PMA and v-Src-podosome induction. In addition, in HUVECs and PAE cells, V12Cdc42 and v-Src expression induced the formation of podosomes. Thus, we showed that Src activation leads to Cdc42 activation and to complete podosome assembly. Taken together, our results favour a model where PKC activates Src, which in turn activates Cdc42 leading to podosome assembly in endothelial cells (Fig. 11).

How Src kinase regulates Cdc42 activity is not known yet, but it might depend on several mechanisms involving adaptator molecules and nucleotide-exchange factors. A possible link between Src and RhoGTPase activation is p120-catenin, which is tyrosine phosphorylated by Src (Mariner et al., 2001) and has been shown to bind the Rho family exchange factor Vav2 to elevate Cdc42 activity in cells (Noren et al., 2000). Another study also described the Src-mediated phosphorylation of Vav2 and the downstream activation of Cdc42 in response to epidermal growth factor (EGF) (Tu et al., 2003). In the same study, EGF signalling through Src resulted in the phosphorylation of Cdc42, which in turn stimulated its binding to RhoGDI (Rho-GDP dissociation inhibitor). Preliminary data failed to show Cdc42 tyrosine phophorylation in response to PMA in HUVECs, suggesting that another pathway is involved in the context of podosomes (F.T., unpublished data). Moreover, v-Src induces tyrosine phosphorylation of substrates including FAK, p190RhoGAP and cortactin that could modulate cytoskeletal organisation in podosomal structures.

Endothelial cells are the main actors of angiogenesis, a process involving remodelling of the extracellular matrix. In the present study, we demonstrated that endothelial podosomes are associated with proteolytic activity. PMA- and v-Src-induced podosomes were found to be competent in degrading ECM in FITC-gelatin degradation assays. Interestingly, V12Cdc42-induced podosomes in endothelial cells failed to degrade the ECM. Since Cdc42 occupies a distal position in this cascade (Fig. 11), some proteins that are required to obtain fully functional podosomes might not be activated upon Cdc42 activation alone. The identity of these components is not yet known but FAK, which is phosphorylated by Src, has been recently shown to be involved in the cell-surface expression of MT1-MMP and the subsequent cellular degradation of the extracellular matrix (Wu et al., 2005). One relevant catalytic activity of MT1-MMP is the processing of pro-MMP2 into MMP-2 (Sato et al., 1994), allowing amplification of the degradation process. MMP-2 is localised and activated at podosomes in PMA-induced HUVECs. Activation of MMP-2 occurs at the plasma membrane, a localisation that explains why we detected most MMP-2 activity in cell extracts from PMA-treated HUVECs. As previously described for osteoclasts (Sato et al., 1997), we found that MT1-MMP localised at podosomes in HUVECs. Interestingly, we also showed that inhibition of MT1-MMP synthesis by RNA interference leads to an increased number of cells with podosomes, suggesting a role of MT1-MMP and matrix degradation in podosome turnover, which is consistent with a previous study in osteoclasts (Goto et al., 2002). The cooperative role of MT1-MMP and αvβ3 integrin in activating pro-MMP2 has also been reported (Deryugina et al., 2001); both molecules localised to endothelial podosomes (data not shown), in activating pro-MMP-2. The involvement of αvβ3 also becomes more probable by the fact that the organic molecule TSRI265, which upon binding to αvβ3 blocks its interaction with MMP2, reduces podosome-associated matrix degradation.

The importance of MMP-2-MT1-MMP cooperativity during angiogenesis has been well studied and reviewed (Haas and Madri, 1999). In the same way, TSRI265 or antagonists of αvβ3 integrin block angiogenesis in multiple cell models (Brooks et al., 1994; Brummelkamp et al., 2002; Gutheil et al., 2000; Silletti et al., 2001). A crucial event during angiogenesis is the invasion of the perivascular extracellular matrix by endothelial sprouts arising from the wall of existing vessels. We propose that podosomes are structures that help protease delivery and subsequent matrix degradation at this stage. After 24 hours, PMA treatment of HUVECs has been shown to induce expression of MT1-MMP and pro-MMP2-processing (Foda et al., 1996; Galvez et al., 2001; Lewalle et al., 1995). Most of the experiments described have discussed the effect PMA after 24 hours of treatment. In this study, we demonstrate that events such as activation of MMPs and degradation of the extracellular matrix occur very early after the addition of PMA and are associated with podosome formation. Thus, it is tempting to speculate that PMA initiates PKC activation at the plasma membrane, thereby transiently reorganising the actin cytoskeleton into podosomes to initiate the angiogenic process. Accordingly, vascular endothelial growth factor (VEGF), the most potent angiogenic factor in vivo (Leung et al., 1989) induces podosomes in HUVECs (Osiak et al., 2005). In addition, it is interesting to correlate the podosome studies in endothelial cells with those made in smooth-muscle cells because both cell types are subjected to vascular remodelling. In vitro, smooth-muscle cell podosomes endowed with an ECM degradation potential can also be induced by stimulating a PKC cascade (Burgstaller and Gimona, 2005). We also have described endothelial podosomes in pathological settings (Moreau et al., 2003). Further experiments are required to provide new insights into podosome function in endothelium pathophysiology.

Phorbol-12-myristate-13-acetate (PMA), glutathione sepharose beads, DMSO, monoclonal anti-vinculin antibodies (hVIN-1), gelatin and various chemicals were from Sigma. The inhibitors, Gö6976, rottlerin, PP2, GM 6001 and also TSRI265 and Mowiol 4-88 were from Calbiochem. Rhodamine-phalloidin, Alexa-Fluor-633-phalloidin, Alexa-Fluor-568- and FITC-labelled secondary antibodies were purchased from Molecular Probes. Anti-myc (9E10) and anti-phosphotyrosine (4G10) were kind gifts from Doreen Cantrell (Wellcome Trust Biocentre, University of Dundee, UK). Antibodies against RhoA and PKCδ were from Santa Cruz, those against avian Src and phosphoSer657-PKCα from Upstate Biotechnologies. Cdc42 and PKCα antibodies were from BD Biosciences, phosphoSer643-PKCδ from Cell Signaling, MT1-MMP from Biomol, MMP-2 from Interchim and αvβ3 from Chemicon.

Human umbilical vein endothelial cells (HUVECs) and endothelial growth medium were obtained from Promocell. HUVECs were cultured in 100-mm dishes coated with 0.2% gelatin in endothelial cell basal medium supplemented with `supplement pack' from Promocell. Porcine endothelial cells (PAE) obtained from Saklatvala (Kennedy Institute of Rheumatology, London, UK) were maintained in F12 medium (Ham F12; Gibco BRL) supplemented with 10% heat-inactivated FCS (Globepharm) and antibiotics. Cells were maintained at 37°C in a humidified atmosphere of 5% CO₂ and 95% air. In all experiments, HUVECs were used between passages two and five. Cells were stimulated with PMA at 50 ng/ml in all experiments.

Cells were plated onto 140-mm culture dishes. After two washes with PBS (Gibco, Invitrogen), cells were starved 1 hour before PMA treatment. Cells were resuspended in lysis buffer containing 5 mM Tris-HCl pH 7, 5 mM NaCl, 1 mM CaCl₂, 1 mM MgCl₂, 2 mM DTT, 1 mM Na₃VO₄, 10 mM NaF, 0.1 mM PMSF and protease inhibitors for 30 minutes on ice. Aliquots were then centrifuged 1 hour at 100,000 g at 4°C. Supernatants were collected as cytosol fractions and pellets as membrane fractions. Protein concentration was subsequently determined.

HUVECs were transfected by electroporation as described by (Ear et al., 2001). PAE cells were also transfected by electroporation according to the following protocol. 20 μg of DNA were mixed with 5 millions of PAE cells. The cell suspension was placed in a 4-mm-gap electroporation cuvette. Cells were electroporated at fixed capacitance of 950 μF and at 250 V using a Bio-Rad Gene Pulser instrument (Bio-Rad). The pEF-v-Src construct was a generous gift from Chris Marshall (Cancer research UK, Institute of Cancer Research, London UK). Constructs encoding active or dominant negative mutants of GFP-RhoA or Cdc42 were kindly provided by Philippe Fort (CNRS-UPR1086, Montpellier, France). pRK5-V12Cdc42 construct encoding myc-tagged active Cdc42 was kindly provided by Alan Hall (Cancer Research UK, London). The pEGFP-PKCα, pEGFP-PKCαA25E and pEGFP-PKCδ constructs were generous gifts from Peter Parker (Cancer Research UK, London) and have already been described (Mostafavi-Pour et al., 2003). The constitutively active PKCδ mutant (pEGFP-PKCδA147E) was made using the mutagenese quickchange Stratagene kit where Ala147 is exchanged for Glu as already described (Schonwasser et al., 1998). Oligonucleotides used are PKCδA1 5′-CTATGAACCGCCGTGAAGAGATTAAACAGGCCAAG-3′ and PKCδA2 5′-CTTGGCCTGTTTAATCTCTCCACGGCGGTTCATAG-3′.

Small interference RNAs (siRNAs) were chemically synthesised (Qiagen) and introduced into HUVECs (200 pmol) using a calcium phosphate precipitation transfection protocol. The antisense strand siRNA was targeted against GTPase using 21-nucleotide sequences (5′-AAGAAGTCAGCATTTCTGTC-3′) for hRhoA, (5′-AAGTTCTTAATTTGCTTTTCC-3′) for hRac1, and (5′-AAGATAACTCACCACTGTCCA-3′) for hCdc42 according to published sequences (Deroanne et al., 2003). hPKCδ1 was targeted using 5′-AAGATGAAGGAGGCGCTCAG-3′ oligoribonucleotide as published in (Yoshida et al., 2003). For hPKCα1, hPKCδ2 and hMT1-MMP, we designed and used 5′-AAGGCTTCCAGTGCCAAGTTT-3′, 5′-AAGGCTACAAATGCAGGCAAT-3′ and 5′AAGGCCAATGTTCGAAGGAGG-3′ oligoribonucleotides, respectively. hPKCα2 was targeted using the siRNA sequence validated by Qiagen (number SI00301308). As control, we used control siRNA Alexa Fluor 488 from Qiagen.

Cells plated onto glass coverslips were prepared for immunofluorescence microscopy as previously described (Moreau et al., 2003). Fluorescent images were recorded on an Eclipse Nikon microscope using a 63× oil immersion lens. Confocal images were captured on a Zeiss confocal microscope. The images were processed using Adobe Photoshop 5.5 (Adobe Systems). Quantification of cells showing podosomes was assessed in three independent experiments, in which at least 200 cells were counted.

pGEX-2T constructs containing rhotekin-Rho binding domain and Cdc42/Rac interactive binding domain of PAK were kindly provided by Martin Schwartz (University of Virginia, Charlotteville, USA) and John Collard (Netherlands Cancer Institute, Amsterdam, The Netherlands), respectively. PAE and HUVECs were grown in 140-mm dishes, stimulated for the indicated times with PMA, lysed and protein extracts were used for pull-down assays as previously described (Ren et al., 1999; Sander et al., 1998).

HUVECs were seeded on FITC-gelatin-coated coverslips and prepared as described earlier (Bowden et al., 2001). Colocalization of dark areas and podosomes was visualised after merging FITC and rhodamine-phalloidin images.

HUVEC cells were seeded at 150,000 cells per well in six-well plates and were stimulated for the indicated times with PMA. Metalloproteinase activity was detected in cell supernatants and extracts. Gelatinolytic activity was assayed by SDS PAGE, in 10% polyacrylamide gels containing 1 mg/ml gelatin as described (Stetler-Stevenson et al., 1997). For cell extracts, cell lysates were obtained by treating HUVECs with 20 μl of sample buffer (50 mM Tris-HCl, 2% SDS, 0.1% bromophenol blue, 10% glycerol, pH 6.8) and run by SDS PAGE at 20 mA/gel. For secreted MMPs, 40 μl aliquots of supernatant were used. Gels were then incubated in 2.5% Triton X-100 for 60 minutes to remove SDS followed by overnight incubation in developing buffer (50 mM Tris-HCl, 0.2 M NaCl, 5 mM CaCl₂, 0.02%Brij-35 pH 7.6). Gels were stained for 30 minutes in 30% methanol, 10% glacial acetic acid, 0.5% Coomassie Blue G-250, and destained for 1 hour in 30% methanol, 10% glacial acetic acid.

We thank C. Chaponnier (Geneva, Switzerland), J. Collard (Amsterdam, The Netherlands), P. Fort (Montpellier, France), A. Hall (Cancer Research UK, London), C. Marshall (London, UK), J. Saklatvala (London, UK), D. Cantrell (London, UK), P. J. Parker (London, UK), M. Schwartz (University of Virginia, Charlotteville, USA), D. Stewart (Bethesda, USA), M. Way (London, UK) and M. Welch (Berkeley, USA) for providing cell lines, constructs and antibodies. We are grateful to K. Mizutani and T. Takenawa (Institute of Medical Science, University of Tokyo, Japan) for help with degradation assays. We thank C. Savona-Baron for critical reading of the manuscript. F. Tatin was supported by a predoctoral fellowship from the Région Aquitaine. This work was supported by grants from INSERM, the Association pour la Recherche contre le Cancer (contract no. 4787) and La Ligue Nationale contre le Cancer (Comité régional Dordogne).