Podosome formation in cultured A7r5 vascular smooth muscle cells requires Arp2/3-dependent de-novo actin polymerization at discrete microdomains

Gross alterations in cytoskeletal architecture are tightly linked to increased cell motility, invasion, metastasis and angiogenesis. The remodeling of actin cytoskeleton structures is commonly accompanied by the rearrangement of focal substrate adhesions, and rapid focal adhesion turnover is consistent with elevated migration rates (for a review, see Kaverina et al., 2002b). While most quiescent cultured cells develop predominantly spear-shaped, stable focal contacts, which serve to anchor actin stress fibers to the sites of cell-matrix adhesion (reviewed in Schoenwaelder and Burridge, 1999; Sastry and Burridge, 2000), motile and spreading cells in addition form transient focal complexes in their cell periphery (Rottner et al., 1999; Smilenov et al., 1999). Monocyte-derived cells, like leukocytes, macrophages and osteoclasts, by contrast, form highly dynamic, dot- or donut-shaped adhesion sites that are generally referred to as podosomes. Transformation of cells by Rous sarcoma virus (RSV) induces podosomes in fibroblasts (Provenzano et al., 1998; Hakak et al., 2000; Mizutani et al., 2002) and epithelial cells (Tarone et al., 1985), and causes a dramatic rearrangement of the actin cytoskeleton and associated focal adhesion sites (Tarone et al., 1985; Marchisio et al., 1988). This transformation is mediated by a single tyrosine kinase, Src, which localizes to the actin cytoskeleton at the ends of stress fiber bundles in focal contacts (Fincham et al., 2000; Marchisio et al., 1988). Focal adhesions and podosomes share major structural components including α-actinin, vinculin, zyxin and paxillin (Duong and Rodan, 2000; Chellaiah et al., 2001). However, podosomes also harbor the actin polymerization machinery components Arp2/3 and Wiskott-Aldrich Syndrome protein (WASp), and may thus generally form at sites where rapid polymerization/depolymerization of actin filaments occurs (Schafer et al., 1998).

Differentiation of monocyte-derived precursor cells into osteoclasts by oseteoclast differentiation factor RANKL-ODF (Matsuzaki et al., 1998) induces high levels of integrin αvβ3 expression (Pfaff and Jurdic, 2001). In addition, αvβ3 localizes to the sites of podosome initiation and the interaction with the non-receptor tyrosine kinase Pyk2 is required for the reinforced distribution of podosomes (Pfaff and Jurdic, 2001). Expression of αvβ3 integrins is strongly restricted to a few cell types, namely activated endothelial cells, vascular smooth muscle cells (VSMCs), tumor cells, and osteoclasts (Westlin, 2001). In agreement with these findings, we have shown recently that VSMCs can also form podosomes in vitro (Gimona et al., 2003).

In response to phorbol ester, the pattern of adhesions in cultured rat vascular smooth muscle cells is partially modulated from spear-shaped focal adhesions to rosette-like podosomes. This process can be blocked by specific inhibitors of PKC and Src (Hai et al., 2002; Brandt et al., 2002) and is tightly regulated by smooth muscle-specific proteins of the calponin (CaP) family, which probably suppress podosome formation in normal smooth muscle tissue (Gimona et al., 2003). Owing to their hitherto restricted occurrence in monocytic cells, podosomes have been ascribed a cell lineage-specific function, although their induction by Rous sarcoma virus or active Src indicated the general potential for podosome formation also in other cell types. Formation of podosomes is associated with an increased migratory and invasive potential of cells. This correlation between podosome formation and the enhancement of invasive behavior prompted us to thoroughly investigate in more detail the structure and dynamics of smooth muscle podosomes.

The cDNA of human Arp-C4 (p20) was kindly provided by Dr Laura Machesky (University of Birmingham, UK). The coding region was amplified by polymerase chain reaction using the following primers: 5′-GAAGAAGATCTATGACTGCCACTCTCC-3′ and 5′-GAGAGACTCGAGAAAATTCTTAAGGAACTCTTC-3′ introducing the restriction sites BglII and XhoI. The fragment was cloned into either the EGFP-C1 or EGFP-N3 vector (Clontech, Palo Alto, CA) and verified by DNA sequencing. GFP-α-actinin and DsRed SM22 constructs have been described before (Gimona et al., 2003). Dr Laura Machesky (University of Birmingham, UK) kindly provided myc-tagged SCAR-WA, and myc-tagged SCAR-W expression constructs. For microscopy of live cells they were co-transfected with GFP-β-actin or GFP-zyxin together with SCAR-WA. DNA for co-transfections was mixed at a 1:10 ratio.

A7r5 rat smooth muscle cells (ATCC, Manassas) were grown in low glucose (1000 mg/l) DMEM without Phenol Red, supplemented with 10% fetal bovine serum (FBS; PAA, Austria), and penicillin/streptomycin (Gibco, Austria) at 37°C and 5% CO₂. For transient expression, cells were grown in 60 mm plastic culture dishes and transfected using Superfect (Qiagen, Hilden) at 70% confluence, essentially as described elsewhere (Kranewitter et al., 2001). Expression and stability of the constructs was assessed by western blotting using a monoclonal antibody against GFP (Clontech, Germany).

Cells were replated onto 15 mm coverslips 16 hours post-transfection and prepared for immunofluorescence microscopy after an additional 48 hours. Cells were washed three times in PBS (138 mM NaCl, 26 mM KCl, 84 mM Na₂HPO₄, 14 mM KH₂PO₄, pH 7.4), extracted in 3.7% PFA/0.3% Triton X-100 in PBS for 5 minutes and fixed in 3.7% PFA (Merck, Germany) in PBS for 30 minutes. Fluorescence images were recorded on a Zeiss Axioscop microscope equipped with an Axiocam driven by the manufacturer's software package (all Zeiss, Vienna) using a 63× oil immersion lens.

Cells were observed in an open, heated chamber (Warner Instruments, Reading, UK) at 37°C on a Zeiss Axiovert TV-135 inverted microscope equipped with epifluorescence, phase-contrast and DIC optics. The objectives, 40×/NA 1.3 Plan-Neofluar and 100×/NA 1.4 Plan-Apochromat, were used with or without 1.6 optovar intermediate magnification. 100 W tungsten lamps were used for fluorescence and phase contrast illumination. Images were obtained using a back-illuminated, cooled charge-coupled-device camera (Princeton Research Instruments, New Jersey) driven by a 16 bit controller. The camera controller was driven by IPLabs software (Visitron Systems, Eichenau, Germany), and shutters were used on the illumination ports to minimize photodamage (Anderson et al., 1996). The digital images were analyzed on an Apple Power Macintosh G3, using IPLabs and Adobe Photoshop 2.5 and 5.5 software.

Confocal images were captured on a confocal spinning disk system (QLC100 confocal head from Visitech, UK) mounted on a ZEISS Axiovert 100 M microscope (ZEISS, Germany). A 63× objective (NA 1.4, exposure time 800 mseconds, 488 nm LASER excitation), and a Micromax camera (Princeton Instruments, USA), driven by IPLab version 3.5.5 software (Scanalytics, USA) were used.

For total internal reflection microscopy, the TIRF imaging system from T.I.L.L. Photonics mounted on a Zeiss Axiovert 200 microscope, including Zeiss 100×/1.45oil, Alpha plan fluar TIRF objective lens, TIRF-Dual Port Achromatic UV/epifluorescence condenser no. 1, and an air-cooled argon ion 488 nm Spectra-Physics model 177-G12 laser was used. Images were acquired with an Interline Transfer CCD Camera IMAGO Type VGA and Imaging Software TILLvisION (generously provided by Maria Marosvoelgyi). A Zeiss Axiovert 100 with Zeiss 100× PlanApo NA 1.4 objective lens was used for interference reflection microscopy as described by Geiger (Geiger, 1979).

Monoclonal antibody to Cortactin was from Upstate Biotechnology (Lake Placid, NY), monoclonal anti-phosphotyrosine (clone PY-99) and anti-myc (clone 9E10) were from Santa Cruz. Anti-p16 antibody (clone #323H3 or #323F2) was a kind gift from Kerstin Schilling (GBF, National Research Centre for Biotechnology, Braunschweig) and used as described (Olazabal et al., 2002). Secondary horseradish peroxidase-coupled antibodies were from Amersham, fluorescently labeled secondary antibodies and phalloidin labeled with Alexa 350 (blue), Alexa 488 (green) or Alexa 568 (red) antibodies were from Molecular Probes (Leiden, The Netherlands)

Injections of 5′-TAMRA-labelled vinculin were performed using Femtotips (Eppendorf, Hamburg) held by a Leitz micromanipulator and attached to an Eppendorf Model 5242 microinjector, essentially as described previously (Kaverina et al., 2002a).

PDBu-induced podosomes in A7r5 cells appear as highly dynamic phase-dense formations by phase contrast microscopy (Fig. 1A). Using confocal fluorescence microscopy and green fluorescent protein (GFP) as a cytoplasmic marker podosomes could be seen to arise as cone-shaped cytoplasmic extensions. These cones were devoid of GFP in their central cavity but showed strong enrichment along the sides of a channel in the x-z axis. Confocal cross sectioning along the x-y axis further revealed a ring-shaped concentration of the GFP-signal (Fig. 1, boxed areas 1 and 2). In small sized, presumably newly forming podosomes, the level of resolution was, however, not sufficient to resolve the central cavity (Fig. 1, boxed area 3).

Podosomes were clearly visible by phase contrast microscopy and the GFP-free foci co-localized with the center of these podosomes (Fig. 1B). Cells double transfected with GFP and DsRed-tagged SM22, a calponin-family actin-binding protein that was found associated with the dynamic pool of actin filaments and became readily redistributed into podosomes as they formed (Gimona et al., 2003), showed strong accumulation of SM22 in the center of the podosomes. This finding suggested that dynamic actin remodeling may occur preferentially in the central regions of podosomes. Transfection of the GFP-tagged p20 subunit of the Arp2/3 complex revealed that Arp2/3 also accumulated strongly in the center of the podosomes, indicating that actin polymerization in podosomes was concentrated at these inner regions (Fig. 1D,E). From these data we concluded that smooth muscle podosomes, like their counterparts in osteoclasts, contain a densely packed core of actin filaments, and that the dense packing accounts for the lack of cytoplasmic signal in the central cavity. This claim was supported by the finding that neither p20-(Fig. 1D,E), nor SM22-labeled podosomes (Fig. 1C, boxed area 5) had a detectable inner cavity.

Total internal reflection fluorescence microscopy was applied to evaluate the proximity of GFP-α-actinin and GFP-β-actin-containing structures to the underlying substrate in smooth muscle cells after PDBu treatment. TIRF images (Fig. 2A′) revealed the presence of focal adhesions (arrows) and podosomes (arrowheads), but not stress fibers, in contrast to epifluorescence images (Fig. 2A). Thus, the lower part of the podosome matrix lies within a 100 nm from the substrate (the average penetration depth of the evanescent wave in TIRF microscopy). Consistently, the ventral membrane in the region of podosomes appeared to form adhesive contact with the underlying substrate, visible as dark patches in interference reflection images (Fig. 2B′). We thus conclude that podosomes in smooth muscle cells are functional adhesive structures with obligatory membrane and cytoskeletal elements.

Our previous studies have indicated that the ends of stress fibers in unstimulated A7r5 cells frequently had a zone just adjacent to the insertion point of the focal adhesion that was devoid of CaP (Gimona et al., 2003). These regions display another striking feature: the actin filaments in this region fail to become decorated by phalloidin (Fig. 3A) leaving a bare zone between the incoming stress fiber bundle and the anchorage site of the focal adhesion. We observed that cortactin was significantly enriched in this region (Fig. 3B) and accumulated at the ends of the focal adhesions (Fig. 3B,C). Cortactin itself was rapidly incorporated into podosomes following the addition of PDBu (Fig. 3D-F). From these findings we suggested that the microdomains located at the focal adhesion/stress fiber interface could represent podosome-initiation sites at which Arp2/3-dependent actin polymerization might be induced.

Podosome dynamics involves the turnover of actin filaments and a number of actin-binding proteins, as well as the de-novo polymerization of new actin filaments, but the temporal sequence of events leading to podosome formation remain to be established. We investigated this by using a GFP-p20 cDNA construct to visualize the sites of actin polymerization in A7r5 cells, and followed their in vivo dynamics before and after PDBu stimulation. In untreated cells, p20 localized preferentially to the tips of lamellipodia and several dot-like, transiently appearing foci, which were present throughout the cytoplasm (Fig. 4A; see also Movie 1). These foci were extremely short-lived (with an average life span of not more than 30 seconds compared to the minute-scale lifetime of late podosomes; see below). Notably, SM22 was not enriched at these foci. Following the induction of podosome formation by PDBu a subset of p20-containing foci began to grow rapidly in size, and continued to serve as platforms for podosome initiation (Fig. 4C). These foci corresponded to Arp2/3-enriched microdomains, which arose in close proximity to stress fibers, overlapping with the sites of cortactin accumulation (see above). Within a few minutes after the first appearance of Arp2/3-marked actin polymerization sites, SM22 was recruited to the newly forming podosomes and was co-localized with all prominent p20-containing patches (Fig. 4A′,B). Consistent with the observation in unstimulated cells, short-lived foci remained devoid of SM22 (Fig. 4B). Monitoring the dynamic formation and disruption of podosomes over a prolonged period (15-30 minutes) further revealed that, in the course of podosome disassembly, p20 was removed from the foci after SM22 depletion (Fig. 4C). We thus concluded that actin polymerization was required at all stages of podosome turnover.

Ectopic expression of WA domains of SCAR has been shown previously to sequester the Arp2/3 complex from sites of potential actin polymerization, and therefore inhibits this polymerization pathway (Machesky and Insall, 1998). We have used this approach to investigate the importance of Arp2/3-dependent polymerization in the remodeling of the actin cytoskeleton and the attached focal adhesions, mediated by phorbol ester. SCAR-WA-expressing smooth muscle cells had a normal morphology but there were numerous filopodia at the cell edges (Fig. 5B) instead of the lamellipodia found in control cells (Fig. 5A). The presence of myc-SCAR-WA prevented GFP-p20 from localizing to the cell edge as well as its aggregation in cytoplasmic foci. Instead, the Arp2/3 complex was found evenly distributed throughout the cytoplasm (Fig. 5B). Phorbol ester treatment failed to induce the clustering of Arp2/3 in myc-SCAR-WA cells (Fig. 5B, right panel), in contrast to control cells (see Fig. 5A), and in the absence of Arp2/3 clustering we detected no sign of podosomes. The lack of podosome formation was consistent in 100% of the SCAR-WA-expressing cells stained for filamentous (F)-actin (Fig. 5C, cell no. 1), while control (Fig. 5C, cell no. 2) or SCAR-W-expressing cells (Fig. 5C, cell no. 3) with uncompromised Arp2/3 distribution responded to PDBu with the formation of podosome clusters.

Notably, PDBu still induced cytoskeletal remodeling other than the formation of podosomes and lamellipodia, which appeared independently of Arp2/3 sequestering. Video microscopy analysis of SCAR-WA-expressing cells revealed continuous disassembly of stress fibers (Fig. 5D) and focal adhesions (Fig. 5E), as well as cell spreading (see Fig. 5B). In contrast to the findings of Machesky and Insall (Machesky and Insall, 1998) in serum-deprived Swiss 3T3 cells, the cultured smooth muscle cells used here were able to spread and form filopodia, despite the over-expression of SCAR-WA. Thus, these latter processes might be mediated via an Arp2/3-independent actin polymerization pathway in A7r5 cells.

PDBu not only facilitates actin polymerization and podosome development, but also induces a general alteration in the adhesion system of smooth muscle cells (Hai et al., 2002). Upon closer inspection this process could be divided into two stages (Fig. 6; see also Movies 2 and 3). In the beginning, podosome formation was induced proximal to stress fibers in discrete microdomains at the focal adhesion/stress fiber interface (see above). In A7r5 cells this process was initiated within 2 minutes post PDBu treatment, and disassembling podosomes were replaced rapidly by newly forming ones in their immediate proximity and adjacent to actin stress fibers within the microdomains (Fig. 6A,B boxed areas). As a second step, focal adhesions, as well as the attached stress fibers were reduced in size, and disassembled either partially or completely. Podosomes, which remained in close proximity to focal adhesions continued to be visible as prominent structures and gradually replaced the focal adhesions (Fig. 6D; see also Movie 4). In A7r5 vascular smooth muscle cells this second phase was reached between 30-70 minutes after PDBu addition. Yet, smooth muscle podosomes had an average life span of 3-10 minutes. Thus, the initial assembly of transiently forming podosomes may not require the complete remodeling of focal adhesions and the recruitment of focal adhesion components at this early stage. To test this hypothesis we followed the dynamics of podosome formation in α-actinin and SM22-transfected cells. Intense actin polymerization during podosome formation was accompanied by a loss of the stable connection between the actin stress fibers and the focal adhesions in which they terminated (Fig. 6C; see also Movie 5). This disconnection occurred prior to the above described reduction of focal adhesion and stress fiber mass, suggesting that rapid actin turnover in this zone contributes to the remodeling mechanism.

To investigate the interrelationship of podosome formation and focal adhesion disassembly we followed the redistribution of major focal adhesion proteins during PDBu-induced cytoskeletal rearrangement (Movie 6). Vinculin was not detectable in early SM22-containing podosomes but was in clearly marked focal adhesion plaques that did not display any noticeable changes in comparison with the untreated cell (Fig. 7A, 6 minutes). Only at later stage, when adhesion were partially disassembled and the majority of podosomes had lost their attachment with the actin stress fibers (Fig. 7A, 32 minutes), was vinculin specifically redistributed into podosomes. Simultaneous analysis of phase contrast images and the distribution of YFP-vinculin further revealed that vinculin associated only transiently with unstable podosomes, and was removed from these structures prior to their disassembly (Fig. 7B). Zyxin displayed similar dynamics to vinculin. GFP-zyxin was undetectable in early, stress fiber-associated podosomes (Fig. 7C,D), but co-localized with microinjected 5′-TAMRA vinculin in late podosomes and in focal adhesions of A7r5 cells (Fig. 7C). Alpha-actinin remained as a stable component of podosomes at all stages, and, as a consequence, co-localized with zyxin in late podosomes (Fig. 7D).

Our results shed light on the spatial origin, genesis and dynamics of transient podosomes in cultured vascular smooth muscle cells. We have shown that podosome initiation and maintenance in A7r5 cells requires Arp2/3-dependent actin polymerization. We further determined the site of podosome initiation as a discrete microdomain, positioned at the anchorage point of actin stress fibers with the focal adhesions. In the course of cytoskeletal rearrangements in response to the tumor-promoting phorbol ester PDBu, focal adhesions gradually disassembled and were remodeled into podosomes, which later became the prevalent adhesive structures in the cell periphery. Thus, cytoskeletal rearrangement leading to podosomes in smooth muscle involved both early recycling and de novo synthesis of actin filaments, and the late disassembly of focal adhesions. Together with our previous data on the differential involvement of h1CaP and SM22 we conclude that podosome formation is governed by the controlled recruitment and regulation of smooth muscle-specific cytoskeletal components, which modulate the stability and turnover of the actin cytoskeleton in response to PKC- and Src-dependent signals.

Monitoring the dynamics of podosome formation in osteoclasts, macrophages or RSV-transformed fibroblast cells has helped in our understanding of the processes leading to the formation of these special adhesion sites (Destaing et al., 2003). However, in all of the above cell types podosomes are generated from cytoskeletons lacking a well developed actin stress fiber system. The A7r5 cell system is in total contrast to other systems. The cells display prominent stress fibers and focal contacts, and have been shown to respond to contractile stimuli. Thus, the system enabled us to monitor not only the formation of podosomes, but to follow the entire remodeling process from a stable focal adhesion and well anchored stress fibers into early transient and late stable podosomes. An additional benefit of the present cell system is the fact that in A7r5 cells podosomes form preferentially in the cell periphery, while the central stress fibers undergo contraction in response to PDBu-induced PKC activation (Gimona et al., 2003). We were thus able to describe the spatiotemporal events prior to and during the genesis of podosomes in greater detail. It has been suggested previously that cortactin is involved in the spatial organization of sites of intense actin polymerization via the Arp2/3 complex (Weed et al., 2000). Co-localization of Arp2/3 and cortactin in punctate structures in the cell periphery of Ptk1 cells further led to the suggestion that cortactin's interaction with the Arp2/3 complex modulates cortical actin assembly (Weed et al., 2000). The presence of cortactin-enriched foci within the specialized microdomains at the focal adhesion/stress fiber interface in unstimulated VSM cells now suggests that these sites are designated areas for cytoskeletal remodeling which recruit, stabilize and activate the actin polymerization machinery following the appropriate stimuli. This finding further demonstrates the existence of a second subcellular location for intense Arp2/3-dependent actin polymerization at the inner face of focal contacts. Thus, we provide evidence that actin polymerization via Arp2/3 is not restricted to the tips of protruding lamellipodia. We are currently investigating the molecular events leading to cytoskeletal remodeling and focal adhesion disassembly at specialized microdomains.

Podosomes represent hot spots of high actin turnover in smooth muscle and monocytic cells, and the formation of newly polymerized actin filaments requires the delivery of G-actin subunits to the site of polymerization. In accordance with this model Akisaka et al. (Akisaka et al., 2001) have demonstrated that cone-shaped podosomes of cultured osteoclasts contain a dense core consisting of short actin filaments, which is surrounded by a cloud or ring of primarily monomeric, G-actin. In human macrophages the Arp2/3 complex is a podosome component whose localization depends on functional WASp (Linder et al., 1999; Linder et al., 2000a; Linder et al., 2000b). Actin nucleation by Arp2/3 is a key factor in the dynamic remodeling of the cytoskeleton and triggers the polarized cellular protrusion of migrating cells. We have shown here that smooth muscle podosomes contain Arp2/3 and that the localization of Arp2/3 complex to the sites of podosome formation marks the earliest detectable step in the process. Thus, in agreement with the situation in monocytes, the formation of podosomes in vascular smooth muscle cells not only coincides with the breakdown of the actin cytoskeleton, but also involves the additional controlled synthesis of new actin filaments.

Our time-resolved studies demonstrate that focal adhesions disassemble gradually during the dynamic formation of podosomes. Cytoskeletal remodeling induced by Rho kinase inhibition causes the reduction of focal adhesions via the modulation of actomyosin interactions (Amano et al., 1996). It has also been established that focal adhesion stability is maintained by contractility exerted by the attached actin cytoskeleton and that the loss of contractility by myosin light chain kinase inhibition accounts for these effects (Chrzanowska-Wodnicka and Burridge, 1996; Hirose et al., 1998; Rottner et al., 1999). We show that focal adhesions can turn over in spite of a stable contractile cytoskeleton. From our data we hypothesize that, at least in the case of PDBu-induced podosome formation, the loss of contractility is not applied throughout the entire actomyosin system, but that regional actin polymerization at specialized microdomains causes a local reduction of contractile forces, sufficient to destabilize focal adhesions. Our results demonstrate that recycling of the actin filament system and the recruitment of focal adhesion components does not occur simultaneously, and that actin polymerization is the initiating step in podosome formation. Thus, the signaling cascade(s) initiated by cPKC activation and Src-phosphorylation (Hai et al., 2002; Brandt et al., 2002), which converge at the actin cytoskeleton, should be acting asynchronously. Early F-actin polymerization via Arp2/3, and late focal adhesion disassembly could thus be governed by the balance between two antagonizing processes, which remain to be defined in detail. This may include the remodeling of actomyosin interactions and the concomitant local reduction of actomyosin-based contractility in microdomains, which in turn present the prerequisites for coordinated focal adhesion disassembly. Our results using the SCAR-WA domain to block Arp2/3-dependent actin polymerization further support the concept that in the absence of actin polymerization podosome formation is suppressed while the remodeling of the actin cytoskeleton and focal adhesions progresses independently. Our current investigations are testing the validity of this hypothesis. Interestingly, Bhatt et al. (Bhatt et al., 2002) have demonstrated a direct regulation of focal adhesions by the protease calpain, but the direct substrates of calpain responsible for this regulation have not been identified. It is noteworthy in this context that CaP, which can stabilize the actin cytoskeleton and partially prevent podosome formation (Gimona et al., 2003) is a substrate of calpain in vitro, and that stress fiber-bound calponin has reduced susceptibility to calpain cleavage (Tsunekawa et al., 1989; Yoshimoto et al., 2000). It remains, however, to be established if proteolysis of CaP by calpain proteases indeed stimulates podosome formation in rat VSMCs.

Although we were able to identify all of the hallmark proteins found also in classical podosomes of monocyte-derived cells (Marchisio et al., 1988; Fincham et al., 2000; Duong and Rodan, 2000; Chellaiah et al., 2001) there was an unexpected difference: smooth muscle podosomes did not contain high levels of phosphotyrosine, and one of the major tyrosine phosphorylated proteins of focal adhesions, paxillin, was absent from podosomes in A7r5 cells (data not shown). A possible explanation for this finding is that smooth muscle podosomes remain transient and may thus not mature in the same way as podosomes in osteoclasts and macrophages.

Our work demonstrates that cells of the monocytic lineage are not the only non-transformed cells forming podosomes. Moreover, we extrapolate that many cells with invasive or metastatic potential may have the intrinsic capacity to form podosomes, and that this process is tightly controlled by specific cellular components under normal conditions. High levels of expression for the actin-binding protein h1CaP are known to be required for the maintenance of a normal smooth muscle phenotype (Horiuchi et al., 1998; Horiuchi et al., 1999; Meehan et al., 2002). Downregulation of h1CaP, in contrast, is generally accompanied with smooth muscle dedifferentiation (Gimona et al., 1990; Horiuchi et al., 1999) and malignant alterations of smooth muscle tissue (Takeoka et al., 2002). The molecular mechanism underlying the increased tumor forming potential of CaP-deficient cells remains to be elucidated. The remodeling of the actin cytoskeleton in response to elevated PKC and Src levels, and the resulting formation of transient podosomes could, however, serve as the molecular basis for such events.

Human vascular endothelial cells (HUVECs) display collagenolytic activity in response to co-culture with human monocytic cells. The co-culture also significantly increases Src activity indicating complicated interactions between vascular cells, including smooth muscle cells, and monocytes/macrophages in the regulation of vascular structure and function (Hojo et al., 2000). Several groups have recently identified a correlation between the pathogenesis of atherosclerosis, or the formation of abdominal aortic aneurisms, with the enhanced expression of matrix metalloproteinases MMP-1 and MMP-2 (Hojo et al., 2000; Goodall et al., 2002), indicating a correlation between an increased migration of smooth muscle cells and the active remodeling of extracellular matrix components. Considering this circumstantial evidence we speculate that the potential for podosome formation of cultured vascular smooth muscle cells probably mirrors the continuous remodeling of the actin cytoskeleton, the dense-plaque-associated cell-matrix adhesion sites and possibly (in the situation of malignant dedifferentiation) of the surrounding extracellular matrix of smooth muscle cells in vivo. It is worth noting at this point that A7r5 cells indeed display fibronectinolytic capacity in vitro, as seen from the use of fluorescently labeled substrates (G. Burgstaller and M.G., unpublished). Although the formation of podosomes remains to be demonstrated in visceral smooth muscle cells, our findings indicate a direct correlation between the transient establishment of podosomes as accessory dynamic adhesion structures in stimulated smooth muscle cells, and the development of tumors.

We thank Dr Alexander Bershadsky for YFP-vinculin cDNA, Dr Anna Huttenlocher for DsRed-Zyxin cDNA, and J. V. Small for critical and helpful comments on the manuscript. Mrs Ulrike Tischler and Ms Johanna Prast are gratefully acknowledged for their technical support. T.E.B.S. was in part supported by an HFSPO grant to Juergen Wehland (GBF, Braunschweig). This work was supported by grants from the Austrian Science Foundation (FWF) to M.G.