Many cell types can generate thin actin-based protrusive structures, which are often classified under the general term of ‘filopodia’. However, a range of filopodia-like structures exists that differ both morphologically and functionally. In this brief review, we discuss the different types of filopodial structures, together with the actin-binding proteins and signalling pathways involved in their formation. Specifically, we highlight the differences between the filopodial extensions induced by the Rho GTPases Cdc42 and Rif.
What and where are filopodia?
Filopodia are highly dynamic protrusions made of actin filaments aligned in tight parallel bundles with their fast-growing barbed end pushing against the plasma membrane [1,2]. A distinction is sometimes made between ‘microspikes’, which are short and appear to be embedded deep within the cell cortex, and filopodia, which are longer and extend out from the cell surface . The most extensively studied are the filopodia arising from the sheet-like membrane projections (lamellipodia) at the leading edge of migrating fibroblasts and mouse melanoma B16F1 cells (Figure 1a) [2,3]. Neurite growth cones also extend lamellipodia and filopodia , as shown in Figure 1(b). In both instances, filopodia have been proposed to function in sensing environmental cues to guide cell migration or axon extension [1,4].
Filopodia and other thin, actin-rich cellular protrusions
Filopodia have also been observed in epithelial cells during wound healing and morphogenetic processes such as dorsal closure in Drosophila [1,5]. Filopodia extend beyond the actin–myosin cable at the leading edge of sheets of migrating epithelial cells (Figure 1c) and are proposed to play a role in helping opposing epithelial cells find their partner cell and form adherens junctions. As opposing epithelial sheets meet, the filopodia interdigitate allowing E-cadherins to interact at sites of contact. These sites then mature into adherens junctions through recruitment of catenins and alignment of the plasma membranes, thereby ‘zipping’ the two sheets together [1,5].
Small highly dynamic actin-rich projections can also be seen on the apical surface of a variety of cells in culture . Although the function of these structures is unknown, apical filopodia of certain cell types have been proposed to be precursors of more specialized protrusions such as brush-border microvilli (Figure 1d) or the sensory stereocilia of the inner ear hair cells (Figure 1e) . Thus cells produce a range of thin projections from their surfaces with seeming morphological similarities, but also apparent differences that may be linked to specialized functions. How these different types of filopodia relate to one another is a major question in this research area.
Mechanisms of filopodia formation
The regulation of filopodia formation is complex and involves a large number of actin-binding and signalling proteins. Nucleation of new actin filaments is required for filopodia formation and one way by which this occurs is through activation of the Arp2/Arp3 complex. The Arp2/Arp3 actin nucleation complex generates new actin filaments by binding to their slow-growing pointed ends. It does so by attaching to the sides of existing actin filaments and nucleating new filaments at an angle, thereby generating a branched actin network . Branched networks can be reorganized to form filopodial bundles and a model describing how this might occur has been proposed by Svitkina et al. . The Rho GTPase Cdc42 was the first signalling protein to be shown to induce filopodia [8,9] and Cdc42 is important for filopodia formation at the leading edge of migrating fibroblasts  and Drosophila epithelial cells . Cdc42 can generate filopodia through Arp2/Arp3 by interacting with the WASP/N-WASP (Wiskott–Aldrich syndrome protein/neuronal WASP), which in turn recruits and activates Arp2/Arp3 .
Despite the existence of the Cdc42–WASP–Arp2/Arp3 pathway to filopodia formation, cells deficient in WASP/N-WASP are still capable of forming filopodia , suggesting that there are alternative pathways to filopodia formation which do not require WASP/N-WASP.
Actin nucleation can also occur without Arp2/Arp3 involvement. Formins are known to nucleate straight actin filaments by binding to their fast-growing barbed ends  and Cdc42 can also induce filopodia formation by interacting with and activating the Diaphanous-related formin p134mDia2/Drf3 (mouse Diaphanous 2/Diaphanous related formin 3) .
In addition to actin nucleation, actin-bundling proteins such as fascin are needed for filopodia formation to cross-link adjacent actin filaments into parallel bundles . The adaptor protein IRSp53 (insulin receptor tyrosine kinase substrate p53) has recently been shown to contain an evolutionarily conserved F-actin bundling region at its N-terminus, which can induce filopodia formation . Cdc42 can interact with IRSp53 through a partial CRIB (Cdc42/Rac interactive-binding) motif, allowing induction of filopodia formation through IRSp53 .
Actin filaments are normally capped at their fast-growing barbed ends by capping proteins to prevent further elongation . However, the filopodial tip complex contains proteins such as Ena/VASP (enabled/vasodilator-stimulated phosphoprotein), which bind to barbed ends, protecting them from capping and allowing continued elongation of filopodial actin filaments . IRSp53 has been shown to interact with Mena (mammalian enabled), a member of the Ena/VASP family, and to act synergistically in the formation of filopodia, and Cdc42 can regulate this process through its interaction with IRSp53 .
It is thus clear that routes to filopodia formation are complex and highly regulated. The Rho GTPase Cdc42 is a key regulator of filopodia formation and appears to be involved in a number of steps along these pathways. However, given the diversity of filopodial structures, it is still unclear whether all are dependent on Cdc42 for their formation.
Cdc42 versus Rif: starfish versus hedgehog
Filopodia induced by active Cdc42 are short and thick, and emanate from the cell periphery, making cells adopt a ‘starfish’ appearance (Figure 2). Cdc42 is predominantly cytoplasmic and does not appear to localize to filopodia. Moreover, Cdc42-dependent filopodia contain focal complexes at their tips, which are rich in vinculin and paxillin . The Rho-GTPase Rif (Rho in filopodia) has also been shown to induce filopodia when overexpressed in HeLa cells . However, Rif-induced filopodia are morphologically different from those induced by Cdc42 (Figure 2). Rif induces numerous long, thin apical filopodia more reminiscent of a ‘hedgehog’ than a ‘starfish’. Unlike Cdc42, Rif localizes to the plasma membrane and along the lengths of filopodia, and Rif-induced filopodia lack focal complexes at their tips . Rif has been shown to induce filopodia independent of Cdc42, WASP/N-WASP and Arp2/Arp3 (, and S. Pellegrin and H. Mellor, unpublished work). The pathway linking Rif to filopodia is yet to be elucidated, pointing to another alternative route to the generation of filopodial protrusions.
Cdc42 and Rif-induced filopodia in HeLa cells
Numerous cells possess thin actin-rich surface protrusions, which differ in length, thickness, dynamics, localization (apical versus peripheral) and origin (lamellipodial or non-lamellipodial). One can imagine that by varying these properties, the cell can produce a range of discrete tools to perform specialized cellular functions. One obvious way to generate this diversity would be through the recruitment of specific sets of actin-binding and signalling proteins. Cdc42 and Rif represent signalling proteins that control the formation of protrusions with very different properties. It is by understanding the function of these and other key regulators of filopodia that we can begin to categorize the wide range of specialized filopodia and to understand fully their cellular functions.
Research Colloquia: Research Colloquia at BioScience2004, held at SECC Glasgow, U.K., 18–22 July 2004. Edited by M. Bouvier (Montreal, Canada), G. Milligan (Glasgow, U.K.), V. O'Donnell (Cardiff, U.K.), M. Brand (MRC-Dunn Human Nutrition Unit, Cambridge, U.K.), M. Schweizer (Heriot-Watt University, Edinburgh, U.K.), R. Insall (Birmingham, U.K.), A. Ridley (Ludwig Institute for Cancer Research, London, U.K.) and M. Sutcliffe (Leicester, U.K.). The first eight papers featured in this Section were presented as a part of the GPCR Regulation and Signalling Research Colloquium, incorporating the GPCR–Ion Channel Interactions Pfizer-Sponsored Research Colloquium.