The interplay between the proteolytic, invasive, and adhesive domains of invadopodia and their roles in cancer invasion

Abstract

Invadopodia are actin-based protrusions of the plasma membrane that penetrate into the extracellular matrix (ECM), and enzymatically degrade it. Invadopodia and podosomes, often referred to, collectively, as “invadosomes,” are actin-based membrane protrusions that facilitate matrix remodeling and cell invasion across tissues, processes that occur under specific physiological conditions such as bone remodeling, as well as under pathological states such as bone, immune disorders, and cancer metastasis. In this review, we specifically focus on the functional architecture of invadopodia in cancer cells; we discuss here three functional domains of invadopodia responsible for the metalloproteinase-based degradation of the ECM, the cytoskeleton-based mechanical penetration into the matrix, and the integrin adhesome-based adhesion to the ECM. We will describe the structural and molecular organization of each domain and the cross-talk between them during the invasion process.

Preface

Invadopodia are actin-based protrusions of the plasma membrane that penetrate into the extracellular matrix (ECM) and enzymatically degrade it.^1-3 They belong to a family of structures, called invadosomes,⁴ which facilitate cell invasion through tissues, a process that occurs under specific physiological conditions such as wound repair, pathogen infection, embryogenesis, and cell differentiation, as well as under pathological conditions such as cancer metastasis.⁵

The first indication of invadosome family structures came in 1980, when David-Pfeuty and Singer observed, in chicken embryo fibroblasts, the relocalization of vinculin and α-actinin from focal adhesions into circular “rosettes,” following transformation with Rous sarcoma virus (RSV).⁶ Five years later, these v-src induced structures were defined by Tarone and Marchisio as “podosomes”⁷, due to their foot-like morphology. Later, more of their structural and signaling components, such as actin,⁸ fimbrin,⁹ and the oncogenic v-src itself,^10,11 were identified, and RSV-induced podosomes were shown by Parsons et al. to be ECM contacts that also serve as sites of local matrix degradation.¹¹ Their capacity to invade and degrade the ECM led Chen, in 1989, to name these structures “invadopodia.”¹² Chen and colleagues continued to characterize these structures, and demonstrated their dependence on metalloproteinase degradation activity,¹² tyrosine phosphorylation,¹² and matrix adhesion.^13-16 Similar adhesive and degradative structures were found in normal cells such as macrophages,¹⁷ osteoclasts,^14,18 dendritic cells,¹⁹ vascular smooth muscle cells,²⁰ and endothelial cells,²¹ as well as in a wide variety of metastatic cancer cell lines such as breast carcinoma,²² melanoma,¹⁶ bladder carcinoma,²³ head and neck carcinoma,²⁴ prostate cancer,²⁵ and a variety of primary tumors.²⁶

Despite the overall similarity between invadopodia, podosomes, and src-induced invadosomes, they do vary in their size, shape, density, and stability.^1,4 Under a fluorescent microscope, invadopodia of cultured cancer cells usually appear as dot-like structures, with an F-actin core, containing several actin-binding proteins and signaling molecules.^2,5 The cores can be scattered or clustered at the cell center, usually in close proximity to the nucleus.¹ Podosomes, on the other hand, can be distributed in various fashions. They can be located throughout the cell in marcrophages, located at the cell periphery in smooth muscle cells, forming rosette-shaped clusters in endothelial cells, and in unique super-structures, such as sealing zones, in osteoclasts.⁴ Src-induced invadosomes in fibroblasts display rosette-shaped peripheral invadosomes.⁴

Invadopodia are long-lived, stable structures that last up to a few hours,^27,28 as opposed to podosomes, whose lifespans is in the order of few minutes.⁵ The podosome actin core is surrounded by an adhesion ring containing plaque proteins such as vinculin, paxilin, and talin,^4,29 whereas the adhesion component of invadopodia is still poorly characterized. Invadopodia and podosomes also differ in their degree of penetration into the ECM: invadopodia were shown to penetrate deep into the matrix³⁰ and form filament-like processes, while podosomes display shallow appearance⁴ and, fail to form long membrane protrusions.³¹ Both podosomes and invadopodia degrade the ECM by local secretion of proteolytic enzymes; as indicated, podosomes tend to degrade the exposed surface of the matrix, while invadopodia, usually, penetrate deeper, into the ECM.^1,4 Invadopodia, podosomes, and src-induced invadosome also display different activation pathways for core components such as the arp 2/3 complex, whose activation in invadopodia require NCK1, while in src-induced invadosomes the activation is mediated by Grb2. In podosomes of macrophages, the activation is not mediated by either one of the proteins.³² This emphasizes the fact that while these structures are molecularly similar, and perform similar functions there might act via different pathways.

In this review, we will focus on the functional architecture of invadopodia in cancer cells, addressing the mechanisms underlying the proteolytic and mechanical remodeling of the ECM by them. We will describe the structural and molecular organization of invadopodia and the roles of the proteolytic domain, the invasive domain and the adhesive domain (Fig. 1A and B; proteins presented in 1A are bolded in the text), and discuss the cross-talk between them during the invasion process.

Figure 1. Mapping of the signaling cascades associated with the proteolytic, invasive, and adhesive domains of invadopodia. The proposed signaling and protein interaction cascades associated with the formation of invadopodia, their adhesion to the matrix, and their penetration into it. Key components involved in these processes, their upstream regulators and downstream targets are shown. Proteolytic domain: ECM degradation by invadopodia is attributed to the activation and function of matrix metalloproteinases (MMPs). The membrane-bound form MT1-MMP is the master regulator of the degradation process, together with the secreted proteases MMP9 and MMP2. The synthesis and secretion of proteases is mediated by the ER and Golgi systems that deliver MMP-containing vesicles along microtubules. One of the central regulators of this process is src tyrosine kinase, which is activated upon receptor tyrosine kinase (RTK) activation. Src phosphorylates MT1-MMP and promotes its clathrin-mediated endocytosis and recycling, a process that promotes the transport of active MT1-MMP to the plasma membrane. Other downstream effectors of the RTK signaling cascades are the small GTPases, Rho A and CDC42, which can activate IQGAP interaction with sec8, a component of the exocytosis machinery that further promotes the delivery of MT1-MMP to the plasma membrane. Another mechanism for MT1-MMP recruitment to invadopodia involves phosphorylated cortactin, in a mechanism that is src- and Arg kinase-dependent. As a master regulator, MT1-MMP can also cleave and activate the secreted MMP9 and MMP2. Invasive domain: Assembly of an invasive actin network is induced by activation of RTKs (e.g., the EGF receptor), which leads to src and PKC kinase activation, as well as recruitment of adaptor molecules such as NCK1. NCK1 binds to WIP (WASP interacting protein) and block its inhibitory interaction with N-WASP, thus enabling N-WASP binding to Arp 2/3 and activating actin nucleation. The Arp 2/3 complex can nucleate actin polymerization on existing filaments, forming a branched actin network. The Arp 2/3 actin nucleation activity is promoted by the serving activity of cofillin, which forms free barbed ends on existing actin filaments. In response to PKC kinase activation, phosphoinositides (PI) are generated at the inner aspect of the plasma membrane of newly forming invadopodia. The adaptor protein TKS5 is recruited to the plasma membrane by binding to PIs, and is phosphorylated by src. The phosphorylated and membrane-associated TKS5 recruits cortactin that is also phosphorylated by src. In its unphosphorylated form, cortactin binds to cofilin, and inhibits it. Upon phosphorylation, cortactin detaches from cofilin, enabling it to sever actin filaments. Active cortactin facilitates both actin filament bundling and recruitment of MT1-MMP (see “proteolytic domain”). PKC can also activate the actin bundling protein fascin, which stabilizes the actin bundle. Adhesive domain: The adhesive domain of invadopodia is a ring-shaped structure associated with the periphery of the interface between the actin core bundle and the ventral plasma membrane (see “invasive domain”). This domain assembles within seconds to minutes following core formation. The assumption is that the adhesion ring is molecularly similar to other types of integrin-mediated adhesions (e.g., focal adhesions; see also Fig. 2, white arrow). It is suggested that activation of integrins, upon binding to ECM ligands in invadopodia sites, induces the local recruitment of adhesion plaque proteins such as vinculin, paxillin, and Hic-5. IQGAP also localizes to the adhesion domain, and possibly interacts with the plaque proteins and facilitates the exocytosis of MT1-MMP (see “proteolytic domain”). (B) Organization of the different functional domains of invadopodia. A schematic representation of the structural organization of invadopodia domains (cross-section) is shown. Each domain is denoted in different color. The proteolytic domain (in orange) includes the ER and Golgi systems, which produce protease-containing vesicles and transfer them to the invadopod. Anchoring and secretion of MMP-containing vesicles are apparent in close proximity to the membrane (see “proteolytic domain,” Fig. 1A), as are recycling and exocytosis of the membrane-bound MT1-MMP. The invasive domain (in black) is localized inside invadopodial protrusions into the ECM. This domain is composed of actin and actin-associated proteins (see invasive domain Fig. 1A). The protein interactions within this domain lead to actin polymerization in a branched and parallel fashion. The polymerizing actin bundles generate forces that are applied on the ECM and push the ventral cell membrane toward it. Around the core, at the protrusion’s edges the adhesive domain is localized, anchored to the ECM via integrins, and associated adhesome cytoplasmic plaque proteins. This adhesion can mechanically support the invasive process. Most of the cross-talk between the three domains occurs upon recruitment of the various components to the plasma membrane, at the site of forming invadopodia. This cross-talk orchestrates the force generation mechanism which is applied to the matrix, together with mechanical stabilization by adhesion and matrix degradation, ultimately driving the invasion and migration of the cell into the matrix.

The Proteolytic Domain

(Fig. 1A [middle panel] B). The matrix degradation capacity of invadopodia is attributed to a variety of proteases, belonging to the metalloproteinase (MMPs), ADAM, and serine protease families.^1,5,33,34 Among these, the MMPs were shown to be the most functionally significant proteases associated with invadopodia activity in cancer cells. MMPs comprise a large family of zinc-dependent enzymes, both membrane-bound and soluble,¹which are capable of cleaving multiple ECM proteins³⁵. The membrane-bound form, MT1-MMP (or MMP14), was shown to be not just an invasion-promoting component, but also a master regulator of invadopodia development,^33,36 whose recruitment promotes invadopodia maturation.² MT1-MMP displays intrinsic collagenolytic activity, and can also cleave and activate some of the soluble MMPs.^34,37 Soluble MMP9 and MMP2 are secreted around invadopodia, and promote local ECM degradation.^16,35-41 Blocking MMP activity by inhibitors, antibodies, or siRNA impairs invadopodia function and matrix degradation.^1,5 Furthermore, MMP inhibition also inhibits invadopodia formation, suggesting that MMPs regulate and fine-tune invadopodia function, at early stages of their formation.^2,5

The mechanisms where by proteases are recruited to sites of invadopodia remain unclear, and likely involve several recruitment pathways (Fig. 1A and B; proteolytic domain). These include vesicle trafficking along microtubules, which was demonstrated for MMP2 and MMP9 in melanoma cells.⁴² MT1-MMP is also regulated by clathrin-mediated and caveolar endocytosis;⁴⁰ its recycling is suggested as a means of generating a membrane-active enzyme.⁴³ Polarized exocytosis and transport of vesicles of MT1-MMP to sites of invadopodia is regulated by Rab8 GTPase in breast cell carcinoma.⁴⁴ In macrophages Rab5a, Rab8a, and Rab14 were shown to be important for MT1-MMP polarized transport.⁴⁵ MT1-MMP can be phosphorylated in a src-depended manner;⁴⁶ and such phosphorylation was shown to promote clathrin-mediated endocytosis.⁴⁰ Exocytosis was also shown to be involved in MT1-MMP regulation via IQGAP, a key regulator of cell polarity, which connects the actin cytoskeleton to microtubules.⁴⁷ Upon activation of RhoA and CDC42, IQGAP is localized to invadopodia together with sec8, a component of the exocytotic machinery; together, they regulate the docking of MT1-MMP vesicles to the plasma membrane.⁴⁸ Furthermore, the actin polymerization regulator cortactin was shown to participate in the recruitment of MMPs to invadopodia, and regulate their secretion.^36,39,49 The adhesion machinery is also thought to play a role in invadopodia maturation and matrix degradation: β1 integrin co-clusters with MT1-MMP at sites of interaction with collagen fibers along the cell’s leading edge.^50,51 In addition, inhibition of integrin-mediated adhesion reduces recruitment of IQGAP to invadopodia, as well as matrix degradation.⁵² A possible mechanism for protease recruitment by β1 integrin involves Arg kinase, which can phosphorylate cortactin on tyrosin 421 and promote matrix degradation.⁵³ Finally, dynamin 2 was shown to regulate the local matrix degradation in invadopodia, through proper localization and its GTPase activity.^54,55

Cancer cells are commonly associated with elevated levels of MMPs, whose expression is correlated with cancer aggressiveness and poor prognosis.^40,56 Thus, studies of their activation and of mechanisms that could inhibit their activity are of great medical interest. Treating cancer with MMP inhibitors was considered to be a very promising approach and was studied in a variety of clinical trials as therapy for various types of cancers.^57-59 Unfortunately, those trials were largely unsuccessful, most likely due to the development of drug resistance by the tumor cells,⁶⁰ secretion of MMP from stroma cells around the tumor,⁶¹ lack of sufficient specificity of the inhibitors, and changes in the cancer cell migration and invasion mechanism, capable of switching from proteolysis-dependent migration to amoeboid migration (a so-called “mesenchymal–amoeboid transition”).^50,62

The failure of MMP inhibition to block cancer invasion indicates that the remodeling of the ECM per se is not sufficient for preventing cancer invasion and metastasis, and suggests that cell invasion into the matrix requires an active migratory process, whereby cancer cells “force their way” into the matrix. This mechanical penetration is primarily attributed to the invasive domain of invadopodia.

The Invasive Domain

General overview

(Fig. 1A [right panel] and B). Local matrix invasion, driven by invadopodia, is a highly coordinated process, whereby enzymatic ECM degradation and mechanical cytoskeletal “pushing” occur at approximately the same time and in the same area, ultimately leading to effective penetration of the cell protrusion into the matrix. Considerable efforts were invested in characterizing the mechanical component of this process, which is believed to be driven primarily by actin polymerization, in a manner similar to the protrusive phase observed in the lamellipodia of migratory cells.^1-3,5 In this section, we will introduce the overall structure and assembly of the cytoskeletal machinery of invadopodia, and then discuss the specific contributions of each component to the overall process.

The temporal sequence of events that eventually leads to invadopodia formation is not fully understood; yet it has been proposed that this process is initiated by ligand-induced activation of receptor tyrosine kinases, such as the EGF receptor ^28,63 (Fig. 1A and B). This signaling event triggers the activation of c-src,^7,64-66PKC isoforms,^67-70 and eventually, the adaptor protein TKS5,⁷¹ which plays a key role in early stages of the assembly of the actin machinery, TKS5 can then bind to Ptdlns(3,4)P₂ at the plasma membrane, promoting the recruitment of actin regulators such as the SH2/SH3-containing, signaling adaptor protein NCK1 to the membrane.²

The exact temporal sequence of these events is still not elucidated. A new approach of high-resolution spatiotemporal live cell imaging, published recently, shows that cortactin, N-WASP, cofilin, and actin arrive simultaneously to form the invadopodia precursor, while TKS5 is recruited later and is required for the structure stabilization.⁷²

Actin polymerization in the core of invadopodia is regulated by small Rho-family GTPases, mostly, CDC42,^1,28 which can directly activate the N-WASP-WIP complex, which, in turn, drives actin polymerization by the Arp2/3 complex.^73-76 Arp 2/3 actin nucleation is promoted by the severing activity of cofillin.^77-79 Additional reports indicate that actin polymerization in invadopodia of certain cancer cells (e.g., MDA-231 breast cancer cell line) can also be induced by specific Diaphanous-related formins, whose cell type specificity, and mode of action and regulation, are still unclear.⁸⁰ The actin bundle, thus formed, is further stabilized and mechanically reinforced by cortactin⁵ and fascin,^73,81,82 producing a stable “invasive protrusion” that pushes against the ventral cell membrane, promoting its penetration into the ECM.^{5,27,28,36,39,68,83}

To get a closer look at the constituents of the invasive domain (Fig. 1A and B; invasive domain) we summarized below the modes of action of the major proteins that are believed to participate in invadopodia nanomechanics:

(1) The adaptor protein N-WASP and its regulators were previously shown to be an essential component of invadopodia.^28,84 N-WASP elimination reduced the ability of mammary adenocarcinoma cells to form invadopodia, and metastasize.²⁶ N-WASP, together with WASP in hematopoietic cells and WAVE 1–3, are major regulators of actin cytoskeleton dynamics.²⁶ Mutations in WASP are known to cause an X-linked inherited disease, Wiskott–Aldrich syndrome, characterized by immunodeficiency. WASP is negatively regulated by WASP-interacting protein (WIP), which is removed upon stimulation by upstream regulators such as NCK1.⁸⁵ In invadopodia, N-WASP is a direct effector of CDC42,⁵ and is regulated by upstream signals from NCK, src, and phosphoinositides.²⁶ N-WASP is characterized by a C-terminal catalytic verprolin homology and a central acidic (VCA) domain, which mediate activation of the Arp2/3 complex. Activation of Arp 2/3 leads to actin polymerization, which generates the mechanical force that enables invadopodia to penetrate the ECM.^86,87 N-WASP was recently suggested to bind to barbed ends and accelerate actin assembly and polymerization in an Arp 2/3-independent manner.⁸⁸

(2) The Arp2/3 complex drives the nucleation of a new actin filament that branches off an existing filament.⁵ Arp2/3 is a 7-subunit complex composed of Arp2, Arp3, and five additional subunits (ARPC1–5⁸⁹). Arp2/3 is associated with invadopodia, and its knockdown by siRNA, or preventing N-WASP from binding to it, inhibits invadopodia formation.^27,28 The Arp2/3 complex, together with N-WASP, is upregulated in tumors and invading cells.^90,91 Stabilization of the actin polymerization capacity of the Arp 2/3 complex is attributed to the binding of Arp 2/3 to both N-WASP and cortactin, the latter also binds to actin, and stabilizes the filament.^76,92,93

(3) Cortactin is a cytoskeletal, multi-domain protein that is important for the induction of actin polymerization.^30,63,94 Cortactin is ubiquitously expressed, and can be phosphorylated by src kinase.^95,96 Cortactin phosphorylation was shown to be an Arg kinase-dependent process, triggered by EGFR activation, and leading to invadopodia maturation.⁹⁷ Tyrosin phosphorylation by src facilitates the response of cortactin to upstream signaling, including dynamic changes in the actin cytoskeleton.⁹⁸ Phosphorylation of Y421, Y466, and Y482 in the proline-rich domain is necessary for metastatic dissemination of breast carcinoma cells,⁹⁹ while Y421 and Y466, but not Y482, are required for the generation of free barbed ends of actin filaments of invadopodia.¹⁰⁰ When phosphorylated, cortactin releases cofilin, and enables it to sever actin filaments, in order to form barbed ends that promote Arp 2/3 actin polymerization.^77,92,101 Cortactin can also bridge between the Arp 2/3 complex and the actin bundle, due to its Arp 2/3 binding N-terminal acidic domain, and thus, serves as an actin stabilizer.¹⁰² Blocking cortactin activity by means of inhibitory antibodies, inhibition of src phosphorylation, or cortactin knockdown by siRNA,⁶⁸ block both the formation of invadopodia and matrix degradation. Interestingly, besides its role in the regulation of the actin machinery, cortactin is believed to directly connect the cytoskeletal components of invadopodia to the proteolytic machinery.^36,39,103

(4) Cofilin was shown to promote actin polymerization by its severing activity, which generates free barbed ends that induce the nucleation of actin polymerization by the Arp 2/3 complex.^5,77-79 Cofilin was shown to be important for invadopodia stabilization, most likely due to its interaction with cortactin (see above) and in its absence invadopodia become small and short-lived.²⁸ Cofilin severing activity is regulated and confined to invadopodia by RhoC and its upstream regulators p190RhoGEF and p190RhoGAP.¹⁰⁴ RhoC regulation on cofilin is mediated by regulation on cofilin phosphorylation through Rho-associated protein kinase and LIM-domain kinase.^79,104

(5) Fascin is an actin-bundling protein that cross-links actin filaments into a tightly packed bundle.^105,106 In humans, there are three members of the fascin family (1–3), of which fascin 1 is the most widely expressed in tissues.¹⁰⁷ The actin bundling capacity of fascin is attributed to the actin-binding sites at its N- and C-termini,^108,109 and the bundling activity in invadopodia is specifically attributed to a PKC phosphrylation site at serine 39.⁸¹ Fascin is upregulated in cells localized at the invasive front of various invasive cancers.^110,111 Fascin is important for invadopodia elongation and actin filament stability, and its depletion was shown to reduce the ability of invadopodia to degrade the ECM.^81,83

(6) Other cytoskeletal systems, besides the actin network, namely, microtubules, and intermediate filaments, are believed to play either direct or indirect roles in invadopodia formation and function.

Little is known about the specific mechanisms underlying the roles of microtubules and intermediate filaments in invadosomes in general, and particularly, in invadopodia of cancer cells. Nevertheless, intact microtubules are required for the function of other invadosomes, such as podosomes of monocytes, macrophages, and osteoclasts^1,112,113. In breast cancer cells, intact microtubules are not necessary for invadopodia formation,^114,115 but their destruction affects invadopodia elongation and maturation.⁸³ In melanoma cells, disruption of microtubules by inhibitory drugs causes invadopodia disassembly, inhibits their de novo formation, and consequently, reduces matrix degradation.¹¹⁶ Previous studies also show that transport of proteins toward the leading edge of migrating fibroblasts requires intact, functioning microtubules,¹¹⁷ a process that is thought to be important for the polarized transport of MMPs to invadopodia.¹¹⁶

Intermediate filaments were shown to interact with actin filaments via fimbrin during the formation of podosomes in macrophages.¹¹⁸ Vimentin intermediate filaments are localized to newly formed podosomes of vascular smooth muscle cells, but are not directly necessary for their formation. In invadopodia, vimentin, but not cytokeratin, was shown to play a role in the invadopodia elongation.⁸³

To summarize, various proteins of the actin machinery work together in order to form a dense web of a dendritic actin network and long parallel bundles, which, together, apply mechanical force that pushes the plasma membrane toward and into the ECM. Other cytoskeletal elements; namely, microtubules and intermediate filaments, are believed to affect the elongation stage of invadopodia; but their specific roles are yet to be explored.

The mechanical force generated by actin polymerization plays a key role in invadopodia protrusive action, but several mechanistic questions still remain open: Is inavdopodia protrusive activity sufficient for driving the entire process of cell invasion? What prevents a protrusion from retracting? What mechanisms coordinate, spatially, tumor cell “proteolytic invasion” and migration? What keeps the actin network from extending deep into the cytoplasm rather than pushing the plasma membrane?

The Adhesive Domain

(Fig. 1A [left panel] and B). Tissue cells grow and function in tight association with diverse components of the ECM. These adhesive interactions play a central role in tissue morphogenesis and remodeling, as well as in physiological and pathological cell migration. Cells connect their cytoskeleton to ECM components via a variety of specialized transmembranal receptors, the most common of which are the heterodimeric integrins.¹¹⁹ Integrins interact with the ECM components via their extracellular domains, and are connected to the actin cytoskeleton via their cytoplasmic tails, and a multi-protein complex of adaptor, scaffold, and signaling proteins, known collectively as the integrin adhesome.^120,121 In addition to their mechanical role, integrin adhesions to the ECM act as “multi-sensory” sites, enabling cells to “scan” the ECM, and respond to its chemical-molecular properties, as well its physical characteristics, including rigidity, stretching,^122-125 and topography of the ECM.^126-130

It is commonly accepted that the spatiotemporally regulated coupling between protrusive processes and the establishment of nearby adhesions is essential for cell locomotion and invasion.^63,131,132 During the migration process, for example, the forward extension of lamellipodia or filopodia, driven by local actin polymerization, is guaranteed by the presence of focal adhesions or focal complexes at the posterior aspects of these protrusions.^131,133

Do matrix adhesions play an essential role in ECM invasion by invadopodia, or is this process primarily driven by MMPs that locally degrade the matrix? Over the years, researchers have addressed this question, yet information about the adhesive domain of invadopodia, its molecular composition, and its structure, remains scarce. Previous studies suggest that invadopodia of cancer cells lack a distinct adhesive capacity, unlike podosomes, which contain a conspicuous adhesive ring.^3,4 It is noteworthy that podosomes, have typical lifetimes on the order of a few minutes, while invadopodia can be stable for hours.¹ More recently, these claims were challenged by the detection of vinculin, paxillin, and Hic-5 in rings located at the periphery of newly formed invadopodia in various cultured cancer cell lines^52,134 (and our unpublished data Fig. 2). It is likely that this apparent discrepancy is attributable to differences between the cancer cell lines, or to variations in the persistence of the adhesion rings during the lifespan of the invadopodia. In fact, it was reported that the adhesive domain of invadopodia is formed soon after the formation of the core F-actin bundle, showing an oscillatory behavior, and occasionally, disappears at later stages, after stabilization of the structure^52,72 (and our unpublished data). The mechanisms underlying the interplay between the core and the adhesion ring, and their significance for ECM remodeling, are still poorly understood.

Figure 2. The actin core and adhesion ring of invadopodia: A375 cells were cultured on gelatin-coated glass bottomed dishes for 2h. Cells were fixed and stained for TKS5, an invadopodia core marker, and vinculin, a marker for integrin adhesions. The adhesion ring (shown enlarged in the upper-right corner) is associated with the invadopodia core at early stages (within minutes) of invadopodia formation, and often disappears at later stages (hours). Stable focal adhesions (FAs), located mainly at the cell’s periphery, are also formed in these cells.

Nevertheless, it was shown that integrin-mediated adhesion regulates invadopodia formation by activation of Rho GTPase family members and various kinases, together with recruitment of adhesion molecules^53,135,136 (Fig. 1A and B; adhesive domain). Several integrins were shown to be associated with invadopodia, including α6β1 in melanoma^135,136 and α3β1 in breast cancer cells,^53,137 suggesting that the association of integrins with invadopodia is tissue-specific. Within the adhesion sites, integrins are believed to bind and activate metalloproteases, besides their “classical” signaling and scaffolding functions. Blocking of integrin-mediated adhesion by soluble RGD peptides or inhibitory antibodies results in reduced matrix degradation, and less localization of MT1–MMP to invadopodia sites.⁵²

Integrins are not the only players in the adhesive domain of invadopodia. In order to form stable adhesions with the ECM and link them to the cytoskeleton, various components of the integrin adhesome, including scaffolding and signaling molecules, are recruited to the adhesion site.^138-141 For example, the adaptor protein paxillin¹⁴² is tyrosine-phosphorylated, and can form a complex with PKCμ and cortactin in invadopodia.⁶⁸ Hic-5, a member of the paxillin family, localizes to invadopodia rings, and upon TGFβ stimulation, promotes ECM degradation in a src phosphorylation-dependent manner.¹³⁴ The role of Focal Adhesion Kinase in invadopodia regulation is rather controversial; it was shown to function as a negative regulator of invadopodia through the spatial control of src in MTLn3 mammary adenocarcinoma cells,¹⁴³ While other reports indicate that it is localized to invadopodia, and its increased expression promotes the formation of src-induced invadopodia.¹⁴⁴ VASP, another adhesome component, accumulates at the degradation sites of invadopodia in MDA-231 breast cancer cells⁸³ and its phosphorylation on Ser239 was reported to suppress invadopodia formation, and inhibit the formation of cancer metastasis.¹⁴⁵

Like other adhesion structures, invadopodia are highly regulated by the chemical and physical properties of the ECM. Matrix rigidity directly increases both the number and activity of invadopodia. This mechanosensing is regulated by the contractile machinery of the cells (e.g, myosin II and Rho kinase).¹⁴⁶ Matrix organization can also affect invadopodia shape. For example, tumor cells cultured on a fibrillar collagen I matrix form linear F-actin structures, which share features with conventional invadosomes, and are capable of degrading the matrix.¹⁴⁷

To date, it is not clear which adhesion proteins participate in formation of the ring in periphery of the invadopodia core. Are they all co localized in the ring, both spatially and temporally, or are there differential kinetics and localization for each protein, as reported for focal adhesions?^148,149 Blocking of integrin-mediated adhesion didn’t block invadopodia formation; yet, it reduced the extent of matrix degradation.⁵² These and other results raise many questions about the role of the adhesive domain in recruitment of the degradation machinery, and the cross-talk between them. Which of the adhesion components are important for invadopodia formation and function? Why do they appear primarily at the early stages of the invadopodia formation? Are there other roles for the adhesion molecules in the invasion process? Does adhesion to the ECM stabilize invadopodia mechanically? Are there differences between the adhesion mechanism, structure, and dynamics in 2D vs. 3D matrices? What is the interplay between the adhesive domain, and other cytoskeletal components?

An Integrated View of Invadopodia’s Functional Architecture

In this article, we attempted to review the structure–function relationships in invadopodia, focusing on structures that carry out specific functions in driving cell invasion, which distinguish them from other types of cellular protrusions (e.g., filopodia or lamellipodia) or other adhesion sites (e.g., focal adhesions or podosomes). Unique to invadopodia is the spatial and temporal coordination of invasion, adhesion, and matrix degradation. Each of these three processes is regulated and executed by a set of multiple proteins that can be define as a “functional domain,” the spatial organization of which is schematically depicted in Figure 1B.

The overall “mission” of invadopodia is the promotion of cell migration across connective tissue barriers (commonly practiced by metastatic cancer cells), as opposed to locomotion along specific, pre-existing migration tracks (as physiologically occurs during organogenesis and embryogenesis). The special property of these three functional domains of invadopodia is their spatial organization and temporal coordination, which enables them to act in concert, in what might be described as a coherent “grip-soften-push” mechanism. Such tactics require fine-tuned cross-talk between the proteolytic (enzymatic), invasive (cytoskeletal), and adhesive domains.

According to this view, the cell utilizes its integrin adhesion machinery to probe the ECM and bind to it (the “grip” step). Upon appropriate stimulation, global or local, actin nucleation is initiated at the cytoplasmic aspects of these adhesions, forming an initial actin core. Subsequently, the proteolytic domain is activated, presenting MT1-MMP on the ventral cell surface, and releasing MMP2 and MMP9. Focal degradation of the matrix by MT1-MMP, and broader degradation by the soluble proteases, reduces matrix barriers under and around invadopodia (the “softening” step), conditioning these areas for the penetration step. Naturally, ECM degradation can also affect nearby adhesion to the matrix by remodeling the ECM, which may explain the observation that not all invadopodia are associated with adhesion rings. Next, rigorous actin polymerization in the core of the invasive domain pushes against the plasma membrane, driving the penetration into the “softened” matrix (the “push” step). Repeated cycles of this “grip-soften-push” by multiple invadopodia could effectively constitute the mechanism underlying invadopodia-mediated cell invasion. Recent evidence for repeating cycles of protrusion and retraction of invadopodia was indeed described.¹⁵⁰

The three functional domains of invadopodia, which are discussed above, are in fact, packed into one coherent structure presented in the transmission electron micrograph, shown in Figure 3. Transmission electron micrographs of invadopodia were shown previously in several cell types.^33,54,83 In the ultrathin section shown here, cut perpendicular to the plane of the gelatin matrix, a single invadopod is seen, pushing against the gelatin layer and apparently compressing or degrading it. The invadopod itself is packed with F-actin fibers, and is devoid of cytoplasmic organelles. Microtubules are often detected at the periphery of the core actin bundle (indicated by the arrow in Fig. 3). Interestingly, the gelatin layer at the peripheral aspect of the invadopod is slightly thicker, in line with the possibility that this region correspond to the adhesion ring. Naturally, invadopodia formed by cultured cells, growing on thin ECM, are rather shallow structures compared with their in vivo counterparts, yet the general distributions of their functional/structural domains might be quite similar.

Figure 3. Imaging of invadopodia by electron microscopy: A375, human metastatic melanoma cells were cultured for 2h on glass-bottom dishes, coated with cross-linked gelatin. Cells were fixed (3% paraformaldehyde, 2% glutaraldehyde, 5 mM CaCl₂) in 0.1 M cacodylate buffer (pH 7.4) for 1h, washed, and post-fixed with 1% OsO_4. Samples were washed again with 0.1 M cacodylate buffer and stained with 2% aqueous uranyl acetate (EMS) for 1h, dehydrated, and embedded in Epon. Ultrathin sections were then cut and examined in an FEI Tecnai™ Spirit T12 transmission electron microscope. In this electron micrograph actin filaments are apparent throughout the core of the invadopodial protrusion, excluding other cytoplasmic structures (e.g., ribosomes) from the core area. The ventral membrane at the invadopode area is in close contact with the glass and the gelatin layer is cleared.

It should be acknowledged that despite our vast information about invadopodia, their structure, and their physiological roles, we are still left with many basic, as yet unanswered questions concerning their coordinated action. To mention a few: are there special features of the ECM (composition, rigidity, topography) that contribute to the assembly and activation of the different domains of invadopodia? How does MMP enzymatic activity affect invadopodia formation, adhesion, and mechanics? How is the timing of the different phases in the “invasion cycle” controlled and coordinated? Answers to these and other open questions may provide a clearer sense of the mechanisms underlying cancer invasion and metastasis and, hopefully, pave the way toward novel and effective therapies.

10.4161/cam.27842