Cells use their cytoskeletons to move, polarise, divide and maintain organisation within multicellular tissues. Actin is a highly conserved essential building block of the cytoskeleton that forms cables and struts, which are constantly remodelled by more than 100 different actin-binding proteins. The initiation of new actin filaments and their subsequent organisation is a key step in the development of specialised cellular structures, such as filopodia (spike-like protrusions), lamellipodia (sheet-like protrusions), stress fibers (elastic contractile bundles), microvilli (finger-like surface protrusions) and invadopodia (invasive cell feet) (see Table 1 for a more complete list). Whereas the cytoskeleton is important in normal cellular function, it can be subverted in cancer cells and contributes to changes in cell growth, stiffness, movement and invasiveness. We hereby give an overview of the role of actin-filament bundling in cellular structures and discuss how alterations in the activity or expression patterns of actin-bundling proteins could be linked to cancer initiation or progression.
Actin is one of the most abundant proteins in mammalian cells, and underpins the compartmentalisation of cellular contents and motility. Filaments are assembled into superstructures by actin-filament-bundling proteins. Some bundling proteins (e.g. fascin and α-actinin) form parallel bundles, whereas others (e.g. filamin) form looser orthogonal meshworks (Bartles, 2000). In general, crosslinking proteins have two actin-binding sites, often because they dimerise, and the location of actin-binding sites determines the filament arrangement and type of crosslinked structure formed. Actin filaments are polar, with a fast-growing and a slow-growing end, and this polarity is maintained by a cycle of ATP hydrolysis (see Poster) (Pollard and Cooper, 2009). Bundling proteins can be selective about the orientation with which they bind to the filament, allowing the specific formation of bundles of either mixed or uniform polarity (see Poster). Bundling proteins are often modular and contain repeated actin-filament-binding domains (see Poster). For example, the calponin-homology domain (CH domain), gelsolin domain and spectrin domain are used by many actin bundlers (see Poster).
A role for actin-bundling proteins in metastasis
Metastatic cancer cells use actin bundles to underpin protrusions that allow them to break away from a primary tumour and invade through the surrounding tissue. After travelling in the vasculature or lymphatic system, they exit into a new niche and seed a new tumour, often after lying dormant for months or years (Hanahan and Weinberg, 2011). During metastasis, cells adapt their motility and adhesive capacity to suit their environment, in much the same way as embryonic cells do during morphogenesis (recently reviewed by Hanahan and Weinberg, 2011; Roussos et al., 2011b). The actin cytoskeleton serves as a scaffold for signalling, a connection with the extracellular environment and a mechanosensor. However, there is no general rule whether actin bundling promotes or inhibits cancer metastasis; rather, cancer cells can adjust the extent of actin bundling to alter their signalling, growth, or adhesion and mechanical properties and thus can be selected for survival during various phases of tumour progression and metastatic spread. Typically, mechanical stiffness is positively correlated with invasion and metastatic potential (Narumiya et al., 2009), but exceptions exist (Swaminathan et al., 2011).
The actin cytoskeleton maintains the compartmentalisation of cellular contents and thus is a major determinant of cell polarity. Polarity is essential for normal tissue homeostasis, and when disrupted, can lead to tumour promotion through the breakdown of cell–cell junctions and to epithelial-to-mesenchymal transition (EMT) (reviewed by Royer and Lu, 2011). Cell divisions are also polarised within tissues, so if polarity is lost, tissue integrity can be compromised, resulting in overgrowth, aberrant invasive behaviour and promotion of tumours (reviewed by Royer and Lu, 2011). Actin bundling contributes to the polarity of epithelial cells by maintaining cell–cell adherens junctions, tight junctions and microvilli, and to polarised membrane trafficking (see Poster). However, understanding of the mechanism by which cells subvert actin bundling to succeed in metastasis is still very much emerging and represents an exciting area of future research.
Actin-bundle-containing structures in normal and cancer cells
Below, we summarise a subset of the various cellular structures that rely on actin bundles for their integrity, and the actin bundling proteins, which have been implicated in cancer initiation or progression.
Underneath the plasma membrane lies a meshwork of actin filaments and crosslinking proteins termed the cortex (see Poster). A strong cortical attachment to the plasma membrane that is balanced by contractility between actin and myosin promotes protrusive motility, which is often referred to as ‘mesenchymal’. By contrast, a weaker attachment of the cytoskeleton to the cortex than the force of actin- and myosin-based contractility promotes blebs, which are detachments between the cortex and membrane, leading to bleb-based motility (Friedl and Wolf, 2010). Modulation of cortical stiffness thus changes how cells move in different environments. The cortex also provides a scaffold for the organisation of transmembrane receptors and glycoproteins into networks for effective signal transduction and coupling of mechanical stresses to signals.
Non-muscle myosin IIa and myosin IIb are the main actin-based contractile myosin motors that crosslink actin filaments of the cell cortex and regulate cell stiffness. Phosphorylation of the myosin II light chain triggers the contractile activity of myosin II (Narumiya et al., 2009). High levels of the myosin kinase Rho-associated protein kinase 1 (ROCK1) are associated with poor survival in breast cancer patients (Lane et al., 2008) and correlate with poor tumour differentiation, muscle invasion and lymph node metastasis in bladder cancer (Kamai et al., 2003). One clinical study showed a positive correlation between myosin light chain kinase, which activates myosin II, with disease recurrence and metastasis in non-small cell lung cancer (Minamiya et al., 2005). Better reagents for the detection of active myosin II or specific enrichment of myosin II isoforms are needed for further clinical studies (Vicente-Manzanares et al., 2009).
Tumours can also influence the contractile properties of stromal cells, such as fibroblasts (Gaggioli et al., 2007; Sanz-Moreno et al., 2008; Wyckoff et al., 2006). Increased stromal cell contractility promotes increased matrix stiffness and this has tumour-promoting properties (Samuel et al., 2011). Matrix stiffness promotes increased integrin attachment, and signalling and activation of pro-survival and growth signals such as activation of focal adhesion kinase (FAK) (Frame et al., 2010).
Whereas myosin II forms parallel contractile bundles, filamin proteins are long, hinged actin bundlers that provide mechanical strength and signalling scaffolds close to membranes (Popowicz et al., 2006) (see Poster). Filamins are mechanosensors, and regulate transcription, membrane trafficking, ion channel function, adhesion and receptor-mediated signalling (Popowicz et al., 2006). For example, filamin binds to the androgen receptor in a complex with β1 integrin, and modulates cell-motility responses downstream of androgen signalling, which could drive invasion in prostate cancer (Castoria et al., 2011; Loy et al., 2003) (see Poster). Filamins also form a same complex with a pro-prion protein PrP in pancreatic cancer to give cancer cells a growth advantage (Sy et al., 2010), and this same complex might contribute to progression of melanoma (Li et al., 2010c). Filamin modulates hepatocyte growth factor receptor (HGFR, also known as the proto-oncogene Met) signalling, which is crucial for many epithelial cancers to metastasise (Zhou et al., 2011). Filamins might also form part of the nuclear skeleton, where they interact with DNA repair complexes such as breast cancer type 1 susceptibility protein (BRCA1) (Velkova et al., 2010) and with cell cycle progression proteins such as cyclin D1 (CCND1) (Zhong et al., 2010) (also see Poster). Finally, a secreted variant of filamin has been detected in the blood of patients with advanced metastatic breast cancer and astrocytomas, indicating that filamin might have a prognostic value (Alper et al., 2009) (Table 1).
Spectrins (also called fodrins) are another class of important actin crosslinkers of the cell cortex, which have been implicated in cancer. In colorectal and pancreatic cancers, β2-spectrin binds to and regulates the activity of transcriptional activators SMAD3 and SMAD4 of the transforming growth factor beta (TGFβ) signalling pathway. TGFβ signalling normally acts as a tumour suppressor of colorectal cancer by suppressing growth and promoting apoptosis, but its dysregulation through loss of β2-spectrin inappropriately activates Wnt signalling and promotes tumourigenesis (Jiang et al., 2010; Thenappan et al., 2009) (also see Poster). Embryonic spectrin (also called embryonic liver fodrin, ELF) shows altered expression in some cancers (Table 1) and its loss causes deregulation of cyclin D1 and aberrant cell cycle progression (Kitisin et al., 2007).
Epithelia are held together by adherens junctions, which contain transmembrane cadherin receptors that interact extracellularly with the cadherins of neighbouring cells and intracellularly with the actin cytoskeleton (see Poster) (reviewed by Etienne-Manneville, 2011). Adherens junctions connect to the cell cortex and to actin filament bundles that are held in place by actin-bundling proteins, such as α-actinin and myosin II (Etienne-Manneville, 2011). When epithelial cells become cancerous, adherens junctions break down, which frees β-catenin from cadherins to enter the nucleus and activate transcriptional changes that lead to endothelial–mesenchymal transition (EMT) through the canonical Wnt signalling pathway (Heuberger and Birchmeier, 2010). Junctional breakdown also physically releases tumour cells, allowing them to escape from the primary tumour and invade the surrounding tissue (see Poster).
α-Actinin-1 and α-actinin-4 localise to cell–cell contacts (Gonzalez et al., 2001) where they regulate actin bundling and epithelial integrity. α-Actinin-4 binds to and recruits the tight junction proteins junctional Rab13 binding protein (JRAB) and molecule interacting with CAS-like 2 (MICAL-L2) and thus participates in tight junction formation (Nakatsuji et al., 2008). Tight junctions lie apical to adherens junctions (see Poster) and maintain impermeability of epithelial tissues. Loss of α-actinin-4 disrupts the integrity of tight junctions and has been associated with cancer invasion and metastasis (Nakatsuji et al., 2008). However, in most studies, high levels of α-actinin-4 correlate with poor outcome or advanced disease (Honda et al., 1998; Honda et al., 2005; Honda et al., 2004; Kikuchi et al., 2008; Menez et al., 2004; Patrie et al., 2002; Weins et al., 2007; Welsch et al., 2009; Yamada et al., 2010; Yamamoto et al., 2007; Yamamoto et al., 2009), and the relevance of its role in tight junction assembly for cancer thus remains unclear. Other functions of α-actinin-4, such as in leading edge protrusion (Honda et al., 1998) might contribute to metastasis and further study is warranted.
Microvilli are finger-like projections of the plasma membrane that increase the surface area of cells to enhance absorption and secretion. Intestinal brush-border microvilli contain a parallel actin bundle core made up of about 40 actin filaments of uniform polarity that are crosslinked by at least three different actin-bundling proteins: T-plastin (also named T-fimbrin), villin and small espin (Bartles et al., 1998; Loomis et al., 2003). Microvilli also contain the cortical components spectrin and myosin II in the terminal web (actin meshwork) at their base (see Poster) (Brown and McKnight, 2010) and are bound to the apical surface by brush border myosin I (McConnell and Tyska, 2007). T-plastin is a monomeric protein, highly expressed in the small intestine, that crosslinks F-actin into straight bundles (Brown and McKnight, 2010; Delanote et al., 2005). L-plastin (Table 1, also known as L-fimbrin) is normally only present in haematopoietic cells; however, one study showed that it is expressed in more than half of epithelial carcinomas and non-epithelial mesenchymal tumours (Delanote et al., 2005). L-plastin expression further correlates with stage and severity of colorectal cancers and is considered a potential prognostic indicator (Foran et al., 2006; Yuan et al., 2010). Villin mediates bundling, nucleation (initiation of new filaments), capping and severing of actin filaments in a Ca2+-dependent manner (Friederich et al., 1990), and is highly expressed in adenocarcinomas originating from epithelial cells of the intestinal tract that bear brush border microvilli (Grone et al., 1986; Moll et al., 1987; Suh et al., 2005). Small espin contributes to elongation of microvilli from the barbed end of the actin bundle, but has not yet been implicated in cancer. In malignant cells, an increased number of microvilli with irregular morphology can correlate with metastatic status (Ren, 1991; Ren et al., 1990), but the significance of this is unclear.
Filopodia are long, thin, actin-based protrusions that promote cell migration and contribute to cancer cell invasion (Mattila and Lappalainen, 2008; Nurnberg et al., 2011) (see Poster). The parallel actin-bundling protein fascin is found in filopodia, but is normally expressed in cells derived from mesenchymal and neural sources rather than epithelia (Adams, 2004a; Adams, 2004b; Hashimoto et al., 2011). Fascin expression is often upregulated in epithelial cancers and is associated with invasion and metastasis (Machesky and Li, 2010). Fascin-mediated actin bundle formation strengthens filaments and increases the lifetime of both filopodia and invasive protrusions (Li et al., 2010a). Fascin is highly expressed at the invasive front of tumours, and in vitro reduction of fascin causes reduced motility and invasion (Hashimoto et al., 2007; Hashimoto et al., 2005; Li et al., 2010a; Schoumacher et al., 2010). Formins (including the mDia proteins, see Poster) are also filopodial proteins with both actin-nucleating and actin-bundling activity. The actin-binding FH2 domains of mDia1, mDia2 and mDia3 dimerise and can both nucleate and bundle actin (Machaidze et al., 2010). Not much is known about the role(s) of the mDia proteins or indeed the other 12 mammalian formins in cancer (Table 1) (Nurnberg et al., 2011).
The Ena/VASP proteins (Mena, VASP and Evl in mammals) comprise a family of proteins that promote actin polymerisation and bundling, and associate with filopodia tips, as well as with lamellipodia, cadherin-based cell–cell contacts (Breitsprecher et al., 2008; Breitsprecher et al., 2011; Scott et al., 2006) and focal adhesions (reviewed by Pula and Krause, 2008). Recently a splice variant of Mena, termed MenaINV, was found to be overexpressed in breast and colorectal cancers (Di Modugno et al., 2004). Mena deficiency decreases invasion, metastasis and tumour progression in polyoma middle-T transgenic mouse models and impairs normal breast development (Roussos et al., 2011a; Roussos et al., 2010; Roussos et al., 2011c).
Invadopodia are dynamic actin-rich membrane protrusions found only in invasive cancer cells (Weaver, 2006). They contain a mixture of bundled and branched actin (Schoumacher et al., 2010) and are used for matrix remodeling (see Poster). Podosomes are structurally and functionally similar to invadopodia, but occur in hematopoietic cells, endothelial cells and Src-transformed fibroblasts (Murphy and Courtneidge, 2011). Invadopodia and podosomes contain a number of actin-bundling proteins, including fascin (Li et al., 2010a; Schoumacher et al., 2010), α-actinin, formins and Ena/VASP proteins (see Poster) (Murphy and Courtneidge, 2011). The actin bundles are used for protrusion into matrix and possibly for delivery of endocytic cargo such as matrix metalloproteases (Murphy and Courtneidge, 2011).
Stress fibres are bundles of parallel actin filaments with mixed polarity along their length (Cramer et al., 1997) (see Poster) and myosin II motors that are crosslinked by alternating zones of α-actinin and Ena/VASP proteins and anchored at their ends by focal adhesion proteins (see Poster). Stress fibres connect the cytoskeleton to the extracellular matrix at focal adhesion sites, where integrins span the plasma membrane and cluster to form large macromolecular hubs of signalling and cytoskeletal proteins (Wolfenson et al., 2009). Focal adhesions are mechanosensing signal-transducing assemblies that reflect the interaction of a cell with its stroma and relay survival and growth signals. α-Actinin-1 distributes along stress fibres in a periodic fashion and binds a number of focal adhesion constituents (Edlund et al., 2001), thereby connecting the actin cytoskeleton to the cell membrane. α-Actinin-4 is generally found at the leading edges of motile cells in lamellipodia. α-Actinin (especially α-actinin-4) is implicated in multiple tumours, including breast (Guvakova et al., 2002), ovarian (Yamamoto et al., 2009) (where it is a prognostic indicator of poor outcome), pancreas (Kikuchi et al., 2008) and lung cancers (Menez et al., 2004) (see Table 1).
Epithelial protein lost in neoplasm (Eplin), is another component of stress fibres; it has filament side-binding and bundling activity that is antagonistic toward Arp2/3 complex branching activity (Maul et al., 2003) and is downregulated during cancer progression (Table 1). Eplin contains a central LIM domain flanked by two actin-filament-binding sites. Palladin is an immunoglobulin-repeat-containing protein that binds to actin filaments and serves as a scaffold for other stress-fibre-associated proteins, such as VASP, α-actinin, EPS8 and the ERM (ezrin, moesin and radixin) proteins. It is a substrate of AKT1 kinase and it can promote actin bundling and inhibit breast cancer cell invasion in vitro (Chin and Toker, 2010). However, another study found that knockdown of palladin inhibits invasive migration of breast cancer cells (Goicoechea et al., 2009). Clearly, there is a need for further study and in vivo verification of the role of palladin in metastasis.
Supervillin is a component of focal adhesions, which contains three villin-related actin-binding sites (Wulfkuhle et al., 1999). It also has four predicted nuclear localisation signals and might shuttle in and out of the nucleus. In prostate cancer, it might be associated with the androgen receptor and therefore involved in the control of cell growth and androgen-dependent signalling (Table 1) (Sampson et al., 2001).
Concluding remarks and future perspectives
Cancer metastasis represents the most deadly aspect of most cancers and also arguably one of the most exciting frontiers for modern biomedical investigation. Tackling metastasis is a complex goal and the improvement of technologies to study tumour lineages and metastatic spread are rapidly developing (e.g. Campbell et al., 2010; Yachida et al., 2010). But equally, as we find out more about how tumours evolve, we are also humbled by the staggering complexity of cancer and of the body. The actin cytoskeleton represents a major network of proteins that impinge on motility, invasion, polarity, survival and growth of normal cells, and as such is often subverted by tumour cells. We are just starting to understand how tumours manipulate the cytoskeleton to gain advantage and to uncover those key proteins that might be future targets against invasion and metastasis. It seems unlikely that one particular actin-binding protein will ever rise above the rest as the most important target in metastasis, but rather, like signal transduction networks, we will find hub proteins or key pathways that can promote tumour progression and develop therapies aimed at these.
A high-resolution version of the poster is available for downloading in the online version of this article at jcs.biologists.org.
- © 2012.