Many Faces Of Actin

12
1110 NATURE CELL BIOLOGY VOLUME 9 | NUMBER 10 | OCTOBER 2007 REVIEW The many faces of actin: matching assembly factors with cellular structures Ekta Seth Chhabra and Henry N. Higgs Actin filaments are major components of at least 15 distinct structures in metazoan cells. These filaments assemble from a common pool of actin monomers, but do so at different times and places, and in response to different stimuli. All of these structures require actin-filament assembly factors. To date, many assembly factors have been identified, including Arp2/3 complex, multiple formin isoforms and spire. Now, a major task is to figure out which factors assemble which actin-based structures. Here, we focus on structures at the plasma membrane, including both sheet-like protrusive structures (such as lamellipodia and ruffles) and finger-like protrusions (such as filopodia and microvilli). Insights gained from studies of adherens junctions and the immunological synapse are also considered. To the best of our knowledge, actin polymerizes into one type of struc- ture in eukaryotic cells: helical filaments. However, these filaments can be assembled further into a wide variety of higher-order cellular struc- tures ranging from lamellipodia to microvilli (Fig. 1), each of which performs specific functions. We ask one basic question in this review: how are filaments assembled for these structures? Actin filament nucleation is unfavourable; therefore, nucleation factors are required to initiate assembly of any actin-based structure 1 . At present, there are three known classes of nucleation factor: Arp2/3 complex, formin proteins and spire. As many of these factors function in multiple steps during assembly of actin-based structures, we prefer the term ‘assembly factor’. Identifying the assembly factor for a given actin-based structure is a key step in determining assembly mechanisms, and has been most thoroughly investigated in yeast 2 . Here, we focus on actin-based structures in metazoan cells, which are more complex in several respects: first, they contain more known actin-based structures; second, these structures often overlap spatially, making isolated analysis more difficult; and third, metazoans have more known assembly factor isoforms (18 in mammals versus three or four in yeasts). Once nucleated, actin filaments can have several fates. It is the pres- ence of additional regulators that determines which type of actin struc- ture will form. In the absence of other factors, filaments are capped by abundant barbed-end capping proteins, preventing further elongation. Inhibition of capping is essential for structures requiring longer filaments. Furthermore, mature structures often require filament crosslinking into networks, parallel bundling or interaction with membranes. Although we do not discuss in detail the proteins mediating these processes, both Arp2/3 complex and formins can perform these functions (Fig. 2). Instead, we focus on the relationship between specific assembly factors and protrusive actin-based structures at the plasma membrane. Broadly, Ekta Seth Chhabra and Henry N. Higgs are in the Department of Biochemistry, Dartmouth Medical School, Hanover, NH 03755, USA. e-mail: [email protected] we divide these protrusive structures into those that are sheet-like (lamel- lipodia/lamella, ruffles, phagocytic cups, podosomes and invadopodia) and those that are finger-like (filopodia and microvilli). We also discuss two other structures, the immunological synapse and adherens junc- tions, because of unique issues they raise, particularly concerning links to the microtubule cytoskeleton. Space limitations prevent us from dis- cussing the many other metazoan actin-based structures (for example, stress fibres, endocytic pits, peri-organellar actin, the cytokinetic ring, cortical spectrin-actin and nuclear actin). Kickstarting polymerization: actin assembly factors Actin monomers self-assemble into helical filaments (Fig. 2a) 1,3 . The initial ‘nucleation phase’ (assembly of dimeric and trimeric complexes) is highly unfavourable and slow but, once overcome, filament elonga- tion is much more rapid. As all actin subunits face the same direction, the filament is polar, with a ‘barbed’ and a ‘pointed’ end. Monomers can add to or depolymerize from either end, but both processes are more than tenfold faster at the barbed end. In metazoan non-muscle cells, the action of profilin – which adds actin monomers to the barbed end only — enhances this effect and effectively limits elongation to barbed ends. Most actin-based structures maintain appreciable flux of actin monomers through the filament, with monomers adding at barbed ends and depolymerizing at pointed ends 4–7 . The first actin assembly factor to be identified was the Arp2/3 com- plex, composed of seven polypeptides (ARPC 1–5, Arp2 and Arp3). Arp2 and Arp3 are actin-related proteins, and are postulated to mimic an actin dimer that can elongate towards the barbed end. The Arp2–Arp3 dimer thereby serves as a nucleation site, and the complex remains at the pointed end of the filament 8,9 . In addition, the Arp2/3 complex binds to the side of pre-existing actin filaments, and its nucleation activity © 2007 Nature Publishing Group

Transcript of Many Faces Of Actin

Page 1: Many Faces Of Actin

1110 NATURE CELL BIOLOGY VOLUME 9 | NUMBER 10 | OCTOBER 2007

REV IEW

The many faces of actin: matching assembly factors with cellular structuresEkta Seth Chhabra and Henry N. Higgs

Actin filaments are major components of at least 15 distinct structures in metazoan cells. These filaments assemble from a common pool of actin monomers, but do so at different times and places, and in response to different stimuli. All of these structures require actin-filament assembly factors. To date, many assembly factors have been identified, including Arp2/3 complex, multiple formin isoforms and spire. Now, a major task is to figure out which factors assemble which actin-based structures. Here, we focus on structures at the plasma membrane, including both sheet-like protrusive structures (such as lamellipodia and ruffles) and finger-like protrusions (such as filopodia and microvilli). Insights gained from studies of adherens junctions and the immunological synapse are also considered.

To the best of our knowledge, actin polymerizes into one type of struc-ture in eukaryotic cells: helical filaments. However, these filaments can be assembled further into a wide variety of higher-order cellular struc-tures ranging from lamellipodia to microvilli (Fig. 1), each of which performs specific functions. We ask one basic question in this review: how are filaments assembled for these structures?

Actin filament nucleation is unfavourable; therefore, nucleation factors are required to initiate assembly of any actin-based structure1. At present, there are three known classes of nucleation factor: Arp2/3 complex, formin proteins and spire. As many of these factors function in multiple steps during assembly of actin-based structures, we prefer the term ‘assembly factor’. Identifying the assembly factor for a given actin-based structure is a key step in determining assembly mechanisms, and has been most thoroughly investigated in yeast2. Here, we focus on actin-based structures in metazoan cells, which are more complex in several respects: first, they contain more known actin-based structures; second, these structures often overlap spatially, making isolated analysis more difficult; and third, metazoans have more known assembly factor isoforms (18 in mammals versus three or four in yeasts).

Once nucleated, actin filaments can have several fates. It is the pres-ence of additional regulators that determines which type of actin struc-ture will form. In the absence of other factors, filaments are capped by abundant barbed-end capping proteins, preventing further elongation. Inhibition of capping is essential for structures requiring longer filaments. Furthermore, mature structures often require filament crosslinking into networks, parallel bundling or interaction with membranes. Although we do not discuss in detail the proteins mediating these processes, both Arp2/3 complex and formins can perform these functions (Fig. 2).

Instead, we focus on the relationship between specific assembly factors and protrusive actin-based structures at the plasma membrane. Broadly,

Ekta Seth Chhabra and Henry N. Higgs are in the Department of Biochemistry, Dartmouth Medical School, Hanover, NH 03755, USA.e-mail: [email protected]

we divide these protrusive structures into those that are sheet-like (lamel-lipodia/lamella, ruffles, phagocytic cups, podosomes and invadopodia) and those that are finger-like (filopodia and microvilli). We also discuss two other structures, the immunological synapse and adherens junc-tions, because of unique issues they raise, particularly concerning links to the microtubule cytoskeleton. Space limitations prevent us from dis-cussing the many other metazoan actin-based structures (for example, stress fibres, endocytic pits, peri-organellar actin, the cytokinetic ring, cortical spectrin-actin and nuclear actin).

Kickstarting polymerization: actin assembly factorsActin monomers self-assemble into helical filaments (Fig. 2a)1,3. The initial ‘nucleation phase’ (assembly of dimeric and trimeric complexes) is highly unfavourable and slow but, once overcome, filament elonga-tion is much more rapid. As all actin subunits face the same direction, the filament is polar, with a ‘barbed’ and a ‘pointed’ end. Monomers can add to or depolymerize from either end, but both processes are more than tenfold faster at the barbed end. In metazoan non-muscle cells, the action of profilin – which adds actin monomers to the barbed end only — enhances this effect and effectively limits elongation to barbed ends. Most actin-based structures maintain appreciable flux of actin monomers through the filament, with monomers adding at barbed ends and depolymerizing at pointed ends4–7.

The first actin assembly factor to be identified was the Arp2/3 com-plex, composed of seven polypeptides (ARPC 1–5, Arp2 and Arp3). Arp2 and Arp3 are actin-related proteins, and are postulated to mimic an actin dimer that can elongate towards the barbed end. The Arp2–Arp3 dimer thereby serves as a nucleation site, and the complex remains at the pointed end of the filament8,9. In addition, the Arp2/3 complex binds to the side of pre-existing actin filaments, and its nucleation activity

© 2007 Nature Publishing Group

Page 2: Many Faces Of Actin

NATURE CELL BIOLOGY VOLUME 9 | NUMBER 10 | OCTOBER 2007 1111

R E V I E W

thereby produces a branched filament structure with an angle of 70° between the filaments. Repeated branching leads to a ‘dendritic net-work’ (Fig. 2b). In this manner, the Arp2/3 complex functions not only as a nucleation factor, but also provides structure to the network. Most of the Arp2/3 complex subunits are single isoforms in all organisms, with a few exceptions in mammals10–12 and Drosophila13. The Arp2/3 complex requires activation by one of several nucleation promoting fac-tors (NPFs), including WASp, N-WASP, Scar/WAVE1, Scar/WAVE2 and Scar/WAVE3 in mammals8.

Formins were identified more recently as a second family of actin assembly factors. In contrast with the Arp2/3 complex, formins are single polypeptide, multidomain proteins14. All formins studied to date are dimeric, due to dimerization of their formin homology 2 (FH2) domain. The FH2 domain is responsible for driving actin nucleation and formin-nucleated filaments are not branched. After nucleation, the FH2 domain remains bound at the barbed end and moves processively as it elongates, promoting elongation by preventing the access of capping proteins (Fig. 2b). Formin-mediated elongation is further enhanced by

the association of profilin with the FH1 domain of formin15. Although all formins have these basic properties, they vary significantly in their potency14. In addition, some formins also bundle, sever or depolymer-ize actin filaments16–20. Most organisms have multiple formin isoforms (15 in mammals, six each in Drosophila and Caenorhabditis elegans)21. It is becoming increasingly clear that formins have additional roles in microtubule dynamics (see below)22,23. At present, the best understood mechanism of formin regulation is through direct activation by Rho family GTPases24.

Spire proteins are a third class of actin nucleators that were initially identified in Drosophila25. Spire proteins are single polypeptides charac-terized by the presence of four Wasp homology 2 (WH2) motifs, which are responsible for actin nucleation26. As with formins, spire-nucleated filaments are not branched (Fig. 2b). Spire can crosslink microtubules and actin filaments27, and also inhibits actin nucleation by the formin cappuccino in Drosophila and its mammalian homologue, formin1 (M. Quinlan, University of California at San Francisco, San Francisco, CA, and R. D. Mullins, University of California at Los Angeles, Los Angeles,

Lamellipodium

Peripheral ruffleDorsal ruffle

Lamellum

Filopodium

Cortical actin

Phagocytic cup

Endocytic pit

Golgi-associated actin

Podosome

Microvillus

Lamellipodium

Lamellum

Nuclear actin

Endosome

Ruffle

Cadherin-basedadherens junction

Stress fibre

Podosome

N

N

Invadopodium

a

b

Substratum

Figure 1 Actin-based structures in metazoan cells. The helical actin filament is used in a myriad of cellular processes. (a) A hypothetical metazoan cell, migrating upwards and attached to a second cell on the right. Cellular structures known or strongly suspected to contain actin filaments are indicated. (b) A side view of a cell, emphasizing several points: ruffles are dorsal structures; peripheral ruffles and dorsal

ruffles are distinct structures; podosomes and invadopodia are ventral; lamellipodia are weakly adherent to the substratum. As far as we know, all these structures use the same basic helical actin filament. This is by no means a complete inventory of metazoan actin-based structures, and we strongly suspect filaments to be involved in other, as yet unidentified processes. N, nucleus.

© 2007 Nature Publishing Group

Page 3: Many Faces Of Actin

1112 NATURE CELL BIOLOGY VOLUME 9 | NUMBER 10 | OCTOBER 2007

R E V I E W

CA, unpublished observations). This inhibition is through bind-ing between KIND domain of spire and the FH2 domain of formin. Metazoans have two spire genes, whereas yeast have none.

Assembling filaments for diverse cellular functionsA schematic overview of metazoan actin-based structures is shown in Fig. 1. In this section, we review what is known of the actin-filament architectures underlying some of these structures, and how they are assembled.

Sheet-like protrusive structures. Cells extend a variety of sheet-like structures, and we discuss three basic types: lamellipodia/lamella, ruf-fles and phagocytic cups. The adhesive structures known as podosomes and invadopodia are not strictly sheet-like, but require many of the same assembly proteins as lamellipodia.

Lamellipodium and lamellum. We define the lamellipodia/lamella as surface-attached sheet-like membrane protrusions observed dur-ing crawling cell motility and spreading. The first step in crawling cell motility is the forward protrusion of the cell front, generally as a thin sheet of membrane-enclosed cytoplasm1. In an elegant series of papers, Abercrombie et al. defined the key characteristics of two regions within this protrusive front, now known as the lamellipodium and lamellum28–31 (see Timeline). The lamellipodium is more distal, starting at the leading edge and extending several micrometres back. The lamellum then takes over, extending from the lamellipodium to the cell body. The lamellipodium is thinner (100–160 nm thick versus >200 nm), more densely stained in electron microscopy images31, less adherent, seems devoid of organelles and is more dynamic. Although these properties were revealed by Abercrombie’s landmark studies, subsequent work better defined additional adhesive differences, with lamellipodia being weakly adherent, and strong adhesion beginning at the lamellipodia–lamella boundary32–34.

Two breakthroughs have changed our conception of the lamellipo-dium and lamellum dramatically. The first was the discovery that den-dritically branched filaments dominate the actin network at the extreme leading edge35,36; the second was the observation that two distinct actin-based networks constitute the lamellipodium and lamellum37. The evo-lution of this field is a beautiful example of two basic facets of modern biomedical research: the ability of technological advances to change a field; and the power of having multiple laboratories investigating over-lapping subjects.

The first finding, that dendritically branched filaments dominate near the leading edge35,36, represented a major change in our view of cell motility. Before these data, the protrusive region of the cell was generally considered to contain orthogonally crosslinked filaments. The discov-ery of the dendritic network coincided closely with the finding that the Arp2/3 complex could nucleate dendritically branched filaments38. Very rapidly, the Arp2/3 complex was considered the dominant nucleator driving leading-edge protrusion.

The second finding, that two filament populations exist at the leading edge, was a direct result of technological advances and astute observa-tion37. The technological advance was fluorescent speckle microscopy (FSM) — the phenomenon of ‘speckle’ movement on cellular structures because of low levels of fluorophore incorporation39. Using FSM on actin filaments in marsupial kidney epithelial cells, two speckle populations were observed in the protruding region. One population, referred to as

lamellipodial speckles, is observed at the leading-edge plasma mem-brane and disappears abruptly about 1–3 µm back37, seemingly at the lamellipodial–lamellar boundary where the first stable adhesion to the substratum occurs34. A second population appears in a more graded fashion, increasing in frequency with distance from the leading edge and persisting through the lamellum. The two populations differ in kinetics, with lamellipodial speckles exhibiting rapid retrograde flow and lamel-lar speckles moving more slowly; transitions between the two states are extremely rare. Taken together, these data suggest that two sets of actin filaments exist in the lamellipodia and lamella, which are independently nucleated and disassembled.

Similar results have been obtained using Drosophila S2 cells40, and further support for two independent actin networks comes from ret-rospective examination of fixed cells. In highly motile fish or Xenopus keratocytes, the actin network is significantly more dense near the

Nucleation factor

Arp 2/3 Formin Spire

a

b

B P

Figure 2 Filament polymerization by actin assembly factors. (a) During actin-filament assembly from monomers, the dimerization and trimerization steps are unfavourable, whereas subsequent monomer additions are much more favourable. Nucleation factors overcome these unfavourable dimerization and/or trimerization steps. Actin filaments are polar structures, with the barbed end (B) being the sole site of elongation in non-muscle cells. Thus, physiologically relevant nucleation factors must allow barbed-end elongation. (b) The Arp2/3 complex (multi-coloured) nucleates a new filament from the side of an existing filament, causing filament branching at a 70° angle. The Arp2/3 complex remains at the branch point between the pointed end of the new filament and the side of the existing filament. Repeated branching results in the assembly of a dendritic network. The formin FH2 domain (semi-circles) nucleates a filament, and then moves processively with the barbed end as it elongates. Some formins can also bundle filaments. Spire (circles with black connectors) nucleates a filament by stabilizing a longitudinal tetramer. The current model is that spire dissociates from filaments soon after nucleation.

© 2007 Nature Publishing Group

Page 4: Many Faces Of Actin

NATURE CELL BIOLOGY VOLUME 9 | NUMBER 10 | OCTOBER 2007 1113

R E V I E W

leading edge than it is farther back, and a large proportion of these filaments are labile to extraction35,36,41. Electron microscopy of fish keratocytes reveals that filaments within 1 µm of the leading edge are predominantly short and crosslinked in dendritic branches, whereas most filaments farther back are longer and less dendritically branched35. Electron microscopic analysis of the protrusive regions of mouse embryo fibroblasts suggests a similar organization of the two networks42. Where studied, both lamellipodial and lamellar filaments orient with barbed ends toward the leading edge plasma membrane35 and the movement of actin speckles supports this orientation37.

It has also been suggested that the lamellipodium may lie on top of the lamellum43. At the leading edge, a layer of densely packed filaments seems to lie above a second network of less dense filaments that extends further back. The speculation is that the upper layer represents lamel-lipodial filaments, whereas the lower layer is lamellar (Fig. 3b). The rear of the lamellipodium is thought to be specified by myosin-II-based contraction at adhesion sites, which is vital for lamellipodial disassem-bly. Although this model of the two networks being vertically separated needs to be tested further, there may also be support for this idea in the early electron microscopy work by Abercrombie31.

Based on the above findings, we conclude that two autonomous actin networks are present in the protrusive region, with the lamellipodial filaments probably overlaying the lamellar filaments (Fig. 3a). Clearly, the proportions and characteristics of these networks might vary sub-stantially between cell types, especially between slow-moving (for exam-ple, fibroblasts) and fast-moving cells (for example, keratocytes). In this respect, speckle analysis of actin networks in keratocytes would be highly informative, as so much ultrastructural information is available for these cells. The question also remains concerning the relative importance of these networks in motility.

How are lamellipodia and lamella assembled? Most evidence suggests that lamellipodia are Arp2/3 complex dependent: first, there is a strong correlation between Arp2/3-complex localization and lamellipodial dynamics. The lamellipodial network contains predominately dendritic branches in several instances35,36,44 and the Arp2/3 complex localizes to

dendritic branch points in these cells36, as well as to the leading edge in multiple cell types10,45–48. Second, Arp2/3-complex dynamics are confined largely to lamellipodia in Xenopus49 and Drosophila40. In the latter case, treatments that vary the width of the lamellipodium proportionally vary the width of the Arp2/3 complex-rich region. Third, altering the activity of the Arp2/3 complex affects lamellipodial structures. Overexpression of central acidic (CA) or WH2 central acidic (WCA) constructs that inhibit the Arp2/3 complex also block lamellipodia assembly10,48,50, although efficacy of these constructs is not universally accepted (Box 1). Microinjection of antibodies that inhibit dendritic branching by the Arp2/3 complex inhibits lamellipodial protrusion51. Knockdown of an Arp2/3-complex subunit causes significant lamellipodial loss in mouse cells52 and strongly inhibits cell spreading in Drosophila cells53.

Debate continues, however, as to how universal the role of the Arp2/3 complex at the leading edge is. One RNA interference (RNAi) study in mouse embryonic fibroblasts argues that the Arp2/3 complex is not required in lamellipodia54. Another suggests that the Arp2/3 complex does not localize to the leading edge in neuronal growth cones, and that overexpression of CA constructs does not alter growth cone morphology or slow growth cone motility48. Others have found the Arp2/3 complex present in, and necessary for, growth-cone pro-trusion (T. Svitkina, University of Pennsylvania, Philadelphia, PA, unpublished observations).

One issue in these studies may be how efficiently the Arp2/3 complex is inhibited, given that it is highly abundant9. Mathematical models of leading-edge protrusion suggest that the Arp2/3 complex may be in large excess over that needed for ‘normal’ migration on experimental surfaces55. It would, therefore, be interesting to evaluate Arp2/3 com-plex-inhibited cells by FSM to determine whether the narrow band of lamellipodial speckles is completely ablated.

Less is known about the nucleators required for lamellar actin fila-ments. The Arp2/3 complex seems largely absent from this region36,40,49, and inhibiting the Arp2/3 complex or other lamellipodial proteins does not alter lamellar properties40,50. The formin mDia2 is implicated in lamellar filament assembly, as well as in focal adhesion turnover

TIMELINE – History of the terms “Lamellipodia”, “Lamella” and “Ruffles”

1958 1969 1970 1971 1980

RUFFLECoined in 1958 to describe theprotruding leading edge of a motilecell132. Used interchangeablywith “ruffled membrane”.

LAMELLIPODIUM (plural, lamellipodia)Coined in 1970 to differentiate the highlyprotrusive, distal portion at the front ofthe motile cell (first several microns) fromthe rest of the lamellum22. No distinctionmade between protrusions parallel to thesubstratum and those that protrudedvertically, although the term “ruffle” wasused as well to describe these structures.

1980First record of differentiation betweenlamellipodia and ruffles. In Abercrombie’smuch-cited review134, lamellipodia definedas forward-protruding structures that adhereweakly to substratum, with ruffles being ofsimilar morphology, but non-adhered andoften protruding dorsally.

LAMELLUM (plural, lamella)Coined in 1969 to replace “ruffled membrane” for the protrusive region of a motilefibroblast133. Also recognizedthat only the first 3-5 microns ofthe lamellum had ruffling activity.

1971In 1971, differences between lamellipodiaand lamella were defined more fully byEM criteria, in terms of thickness, stainingdensity, presence of organelles, andapparent adhesion31.

© 2007 Nature Publishing Group

Page 5: Many Faces Of Actin

1114 NATURE CELL BIOLOGY VOLUME 9 | NUMBER 10 | OCTOBER 2007

R E V I E W

(C.Waterman-Storer, University of Washington, WA, unpublished observations), suggesting a potential association between lamellar fila-ments and focal adhesions. However, mDia2 is not expressed in sev-eral motile cell types, suggesting it may be functionally redundant with mDia1, mDia3 or other fomins.

Ruffles. We define ruffles as sheet-like membrane protrusions that do not attach at all to the substratum, as opposed to weakly adherent lamellipodia. There are at least two distinct varieties of these transient structures, both with half-lives of minutes. Peripheral ruffles assemble at the leading edge of motile cells, and move rearward29. Circular dorsal ruffles — also called ‘waves’ or ‘actin ribbons’ — assemble on the dorsal surface and constrict into a circular structure before disappearing. These seem to be distinct structures, as rearward-moving peripheral ruffles do not transition to dorsal ruffles56. Peripheral ruffles are associated with crawling-cell motility only, whereas dorsal ruffles also affect receptor internalization and possibly macropinocytosis57. A distinction between peripheral and dorsal ruffles was made by Abercrombie29.

It is tempting to speculate that peripheral and dorsal ruffles arise through the same mechanisms as lamellipodia. Indeed, early live-cell

imaging studies provide convincing evidence that peripheral ruffles and lamellipodia have common origins29, and recent results support this hypothesis43. The Arp2/3 complex seems to be important for assembly of both types of ruffle56,58. Most recently, the presence of dendritically branched filaments in peripheral ruffles was observed59.

Formins may also function in peripheral-ruffle dynamics: first, active RhoA — thought to directly activate the formins mDia1 and mDia2, but not the Arp2/3 complex — is enriched at the leading edge of multiple mammalian cell types, both in lamellipodia and ruffles60–64, and is required for ruffle assembly61; second, mDia1 is enriched at ruffle edges60,61,63; and third, coexpression of constitutively active Rac1 and dominant-negative RhoA allows spreading, but blocks ruffles61. This result hints at potential differences between lamellipodia and peripheral ruffles, by suggesting that Rac1 activation of the Arp2/3 complex (through Scar or WAVE) is sufficient for lamellipodia, but not ruffle, generation.

The proposal that the lamellipodial network lies above the lamel-lar network might also explain the role of mDia1 in peripheral ruffles43: according to this study, the rear of the lamellipodium is specified by assembly of a myosin II-based contractile com-plex at sites of stable adhesion (Fig. 3a), and its contraction pulls

Filopodium

Ruffle

Lamellipodium Lamellum

Stable adhesion siteWeak adhesion site

a

c

b

Substratum

Filopodium

Figure 3 Schematic representations of models of assembly for lamellipodia, lamella, peripheral ruffles and filopodia. (a) Top view of lamellipodia and lamella. Lamellipodial actin filaments (blue) are dendritically branched and Arp2/3-complex-dependent. These filaments assemble at the leading edge plasma membrane (due to assembly factors bound there) and disassemble abruptly at a band corresponding to the first sites of stable adhesion (green), generally 1–3 µm from the leading edge. Lamellar actin filaments (red) are in a less defined network, and are not Arp2/3 complex-dependent. They can initiate throughout the lamellipodium and lamellum, but their frequency increases as one moves back from the leading edge. Lamellar filaments persist many micrometres behind the lamellipodium, and disassembly can occur throughout this region. (b) Side views of lamellipodium and lamellum.

Lamellipodial filaments may lie above lamellar filaments. At the leading edge, lamellipodial filaments predominate, but lamellar filaments increase in abundance further back. A lamellipodium maintaining weak attachment to the substratum is shown in the upper panel. Myosin II activity at the stable adhesion site causes retrograde flow of the lamellipodial actin network. If the lamellipodial attachment with the substrate is broken, myosin II activity causes this membrane sheet to move rearward as a peripheral ruffle, which disassembles at the stable adhesion site (lower panel). (c) Filopodial assembly from the lamellipodium or lamellum. Filopodial assembly from the lamellipodial network through the convergent-elongation mechanism is shown on the left. Possible filopodial assembly from the lamellar network, through an as yet undefined mechanism, is shown on the right.

© 2007 Nature Publishing Group

Page 6: Many Faces Of Actin

NATURE CELL BIOLOGY VOLUME 9 | NUMBER 10 | OCTOBER 2007 1115

R E V I E W

lamellipodial filaments rearward. If the extreme leading edge remains adhered to the substratum, contraction causes retrograde flow. If, however, the connection is broken, the whole leading edge, including actin filaments and plasma membrane, moves rearward as a peripheral ruffle (Fig. 3b). Support for this model can be found in the earlier studies of Abercrombie31. The lamellar network makes up part of the substrate–adhesion complex34 and so may be essen-tial for peripheral ruffles. Perhaps mDia formins mediate lamellar-filament assembly in fibroblasts? Immunofluorescence microscopy illustrates that mDia1 localizes to the edge of spreading cells in several cell types (refs 23, 63; K. Eisenmann and A. Alberts, Van Andel Institute, Grand Rapids, MI, unpublished observations; H.N.H, unpublished observations).

Our current conclusion is that lamellipodia and peripheral ruffles arise from the same assembly mechanism, and differ in substrate adhe-sion. The assembly mechanism for dorsal ruffles is unclear, but probably involves Arp2/3 complex-generated dendritic networks.

Phagocytic cups and pits. Phagocytosis is the cellular uptake of parti-cles larger than 0.5 µm diameter. Although both types of phagocytosis conducted by macrophages — Fc receptor-mediated (FcR) and com-plement receptor-mediated (CR) — require actin polymerization, they differ substantially in morphology64. In FcR-mediated phagocytosis, the macrophage membrane protrudes around the phagocytosed par-ticle. Following membrane fusion at the distal tip, the particle is then pulled into the macrophage. In CR-mediated phagocytosis, the parti-cle ‘sinks’ into the macrophage and the macrophage does not extend toward the particle.

The morphology of the actin filaments is not known for either proc-ess. Based on the protrusive, sheet-like nature of the phagocytic cups in FcR-mediated phagocytosis, we speculate that these cups use lamel-lipodia-like or lamella-like filaments. One difference, however, is that phagocytic-cup assembly requires extensive vesicle fusion65, which has not been shown for lamellipodia or lamella.

The Arp2/3 complex seems to be required for both types of phago-cytosis, based on its localization to these structures and on inhibition of both processes by WCA constructs66. In addition, WASP–GFP local-izes to FcR-mediated phagosomes, and macrophages lacking WASP are compromised in FcR-mediated phagocytosis67. The formins mDia1 and mDia2 seem to function in CR-mediated phagocytosis68.

Podosomes and invadopodia. Podosomes are integrin-containing struc-tures on the basal cell surface, containing an actin-rich core surrounded by a ring of several actin-associated proteins and signalling proteins (Fig. 1)57,69,70. They have been studied in detail in a limited number of cells — Src-transformed fibroblasts and monocyte-derived cells (such as macrophages, dendritic cells and osteoclasts). However, the evidence is mounting that podosomes are also present in other cell types71,72, and may mediate integrin-based adhesion to extracellular matrix.

Invadopodia have a similar architecture and protein composition to podosomes, but are formed by cells migrating over thick extracellular matrix, and mediate matrix degradation69,70. Podosomes — which some have postulated to be ‘invadopodia precursors’ — are smaller (approxi-mately 0.5 µm diameter and 0.5 µm height), whereas invadopodia are wider (2 µm diameter) and protrude several micrometres into the extra-cellular matrix (Fig. 1). Both structures are ‘hot spots’ for membrane dynamics, and invadopodia have extensive plasma membrane invagina-tions. Where examined, both podosomes and invadopodia are enriched in a specific matrix metalloprotease57,69,70.

All evidence points to the Arp2/3 complex as the key nucleator for both podosomes and invadopodia. The Arp2/3 complex and several of its regulators are enriched in macrophage podosomes57,69,70. Microinjection of CA constructs blocks podosome assembly in macrophages73,74. Human Wiskott-Aldrich Syndrome patients, who lack WASp, do not have mac-rophage and dendritic cell podosomes75,76. N-WASP is enriched at podo-somes of non-haematopoietic cells, and is important for their assembly77. N-WASP and the Arp2/3 complex are enriched in, and are required for, invadopodia assembly78.

The ultrastructural details of the actin network are unclear for podo-somes and invadopodia. Clearly, the filaments are densely packed in both, but it is uncertain whether they are dendritically branched, straight or crosslinked into orthogonal networks. In view of the requirement for the Arp2/3 complex in these structures, one would postulate dendritic branching is present, but it will be important to confirm this.

Finger-like protrusive structures. The range of finger-like protru-sions observed in metazoan cells is enormous, and our general view is that multiple assembly mechanisms are likely. We define filopodia and microvilli as thin (less than 200 nm diameter) protrusions, con-taining parallel bundles of 10–30 actin filaments that run the length of the protrusion and orient their barbed ends towards the membrane.

WCA constructs refer to the minimum polypeptide sufficient for in vitro activation of the Arp2/3 complex, located at the carboxyl termini of most nucleation-promoting factors9. WCA binds both the Arp2/3 complex and an actin monomer. Overexpression of WCA constructs has been used to inhibit the Arp2/3 complex, on the basis that these constructs would activate the Arp2/3 complex in a mislocalized fashion, preventing localized activation by endogenous nucleation-promoting factors.

CA constructs are shorter and only bind to the Arp2/3 complex. The CA construct of WASp competes with Arp2/3-complex activators in vitro, blocking nucleation73. In vivo, CA constructs are predicted to sequester the Arp2/3 complex from endogenous nucleation-promoting factors.

Reviews are mixed on the efficacy and specificity of these constructs for Arp2/3-complex inhibition in cells. Many researchers have reported successful cellular inhibition of the Arp2/3 complex with these constructs (see main text). However, one concern with WCA is that high expression might sequester significant amounts of actin monomer, causing additional effects that are not dependent on the Arp2/3 complex. As for CA, some report lack of effect on cellular processes strongly believed to depend on Arp2/3 complex.

BOX 1 TOOLS FOR REGULATING THE ARP2/3 COMPLEX: WCA AND CA CONSTRUCTS

© 2007 Nature Publishing Group

Page 7: Many Faces Of Actin

1116 NATURE CELL BIOLOGY VOLUME 9 | NUMBER 10 | OCTOBER 2007

R E V I E W

We define filopodia as finger-like protrusions that are adhered in some manner to a substratum or another cell. We define microvilli as finger-like protrusions that are not adhered. Of course, grey areas exist between these definitions, in particular when both adherent and non-adherent finger-like protrusions are experimentally induced79. In addition to these simple finger-like protrusions, a wide variety of larger and more complex protrusions (including Drosophila bristles and hair-cell stereocilia) use multiple modules of actin bundles, but these are not discussed here (for a review, see ref. 80).

Filopodia. Filopodia protrude from the leading edge of many motile cells, including fibroblasts and nerve growth cones (Fig. 4)81. In these situations, filopodia emanate from a lamellipodial or lamellar sheet, and are believed to function as directional sensors82 (perhaps through enrichment of acti-vated integrins in filopodial tips83). Viruses often bind at filopodial tips, and are transported back to the cell body before being internalized84,85. A similar process occurs for activated epidermal growth factor (EGF) recep-tor on adenocarcinoma cells86. Finally, filopodia are abundant on neuronal dendrites, and seem to be important for dendritic spine development87.

Electron microscopy analyses show that filopodia contain bundles of long parallel filaments36,88–90. These extend deep into the lamellipo-dium or lamellum, where they eventually splay at their pointed ends. The sites and mechanisms for filopodia assembly are currently under debate. One model is that filopodia assemble by ‘convergent elongation’ of lamellipodial filaments44,91, through re-organization of the Arp2/3 complex-assembled dendritic network into bundles. This re-organiza-tion is initiated when a subset of filament barbed ends associate with a filopodial-tip protein complex at the leading edge plasma membrane. This complex, containing the Ena/VASP actin-binding proteins and pos-sibly formins, protects barbed ends from capping proteins, thus allowing them to elongate preferentially in the network. Contacting filaments would then be bundled by the bundling protein fascin to eventually pro-duce a filopodium91.

The convergent elongation model requires Arp2/3 complex-medi-ated nucleation, followed by rearrangement of the dendritic network. However, recent studies depleting subunits of the Arp2/3 complex by RNAi have provided conflicting results in terms of whether filo-podia or just lamellipodia were affected (T. Svitkina, unpublished

Tip complexEna/Vasp,Formin, myosin X

Fascin (and/or formin?)

Myosin I

ERM protein

Villin

Fimbrin

Epsin

Terminal web

Electron-dense mass

a

b

Figure 4 Schematic representations of models for assembly of filopodia and microvilli. (a) A top view of a filopodium — a finger-like, actin-rich protrusion that is adhered to the substratum. An electron-dense complex at the tip seems to be enriched in Ena/Vasp proteins, myosin X and formins. Fascin seems to be the predominant bundling protein in the most well characterized filopodia systems. One possibility is that formins might contribute to bundling in some cases. Most often, the filopodial base is embedded in the lamellipodium and lamellum. The convergent-elongation model of filopodial

assembly is thought to occur through re-organization of the Arp2/3 complex-dependent lamellipodial network. (b) A side view of a microvillus, which has similar finger-like morphology to filopodia, but is not attached to the substratum. The most well characterized microvilli on epithelial cells (brush-border microvilli) contain parallel bundles of actin filaments, crosslinked by villin, fimbrin and espin. Microvilli contain an electron-dense mass at their tip. At their base, epithelial microvilli seem to be embedded in the terminal web, which is composed of actin filaments and myosin II.

© 2007 Nature Publishing Group

Page 8: Many Faces Of Actin

NATURE CELL BIOLOGY VOLUME 9 | NUMBER 10 | OCTOBER 2007 1117

R E V I E W

observations)52. Both of these studies use B16F1 mouse melanoma cells, but with potentially key variations in the substrates on which the cells were plated and in the medium added to the cells. A third study reports a slight increase in filopodia number in mouse embryonic fibroblasts when Arp3 is depleted by RNAi54. Finally, N-WASP has been shown to trigger filopodia assembly downstream of Cdc42 (ref. 92), but cells lacking N-WASP still possess filopodia93.

Regarding formins, mDia2 has been linked to filopodia assembly79,94,95 (T. Svitkina, unpublished observations). In many of these cases, it is unclear whether the structures hold to our adhesion-based definitions of filopodia, as both adherent and non-adherent finger-like protrusions are visible. It is possible that overexpression of formins or other proteins used in these studies causes assembly of both structures. A formin is also required for filopodia in Dictyostelium96, although the localization of a GFP-fusion of this formin to finger-like protrusions irrespective of substrate-attachment in this study raises questions about our distinction between microvilli and filopodia.

Our opinion is that there may be multiple potential mechanisms for assembling filopodia. Filopodial assembly requires three basic steps: fila-ment nucleation, sustained barbed-end elongation and filament bundling. A variety of molecules may be capable of accomplishing each task. Moreover, bundled structures can be nucleated by the Arp2/3 complex or formins in vitro17,19,97,98. We believe that it is perfectly possible that different mechanisms of filament nucleation are used in distinct cellular environments.

After assembling, leading-edge filopodia elongate at rates between 1–5 µm min–1 in multiple cell types5,52. However, elongation dynam-ics vary, even during the lifetime of a single filopodium, with periods of elongation and retraction. In growth-cone filopodia, the elongation rate depends on a balance between monomer addition at the tip, and retrograde flow combined with depolymerization at the base5. Most commonly, retrograde flow and depolymerization seems fairly constant,

whereas monomer addition rate varies. An interesting possibility is that formins are at the barbed ends of filopodial filaments and their regulation controls elongation rate.

The basic structure of filopodia, and how they relate to lamellipodia and lamella, is still open to surprises, and recent results may impact on possible assembly mechanisms. Although most electron microscopy studies show that filopodia consist of long, parallel filaments, recent cryo-electron microscopy work in Dictyostelium suggests that shorter filaments may also be present that are not necessarily parallel, with very short filaments in a meshwork pattern at the tip99. The suggestion from this work is that these shorter, labile, filaments may have been destroyed by previous electron microscopy preparations. We reserve judgment on this possibility until similar observations are made for other filopodia, as these Dictyostelium filopodia elongate approximately twentyfold faster than most other protrusions (about 1 µmicron s–1) and, thus, may use a fundamentally different protrusive mechanism.

The suggestion that the lamellipodium lies on top of the lamellum43 has implications for filopodia assembly. If filopodia derive from the action of the Arp2/3 complex in the lamellipodium, do they transition at some point to the lower, lamellar layer, as their roots often end up proximal to lamellipodium (Fig. 3c)? Or, do they assemble from the dendritic network at the lamellipodium–lamellum interface? Could filo-podia-like protrusions derive from the lamellum? Clearly, many other mechanisms are possible.

The definition of a filopodium can be quite subjective, if based solely on light microscopy of fixed cells. Care must be taken to ensure that the struc-tures are protrusive, and not ‘retraction fibres’. Some protrusions defined as filopodia in the literature are less finger-like than others, being wider overall and splaying to widths of several microns at their intersection with the lamellipodium–lamellum. In addition, fibroblasts can protrude filopo-dia-like structures from regions containing no obvious lamellipodium or lamellum (S. Nicholson-Dykstra, Dartmouth, Hanover, NH, and H.N.H., unpublished observations). We question whether all such structures are subject to a common assembly mechanism.

Microvilli. Microvilli are most recognizable on the luminal surfaces of intestinal and kidney epithelial cells, where they are densely packed and of uniform length (around 2 µm; ref. 100). However, shorter (<0.5 µm) and more variable length microvilli are present on many cell types, including circulating lymphocytes and the dorsal surfaces of many cultured cells101,102. Although the function of epithelial microvilli is to increase absorptive surface area, lymphocyte microvilli may segregate cell surface proteins103,104 — a property that may aid extravasation from blood to the periphery.

Both epithelial and lymphocyte microvilli contain long actin fila-ments arranged in parallel bundles100,102. Although epithelial microvilli maintain constant length after initial assembly, their actin filaments are still dynamic, with subunits adding at barbed ends and dissociating at pointed ends7,105. Shorter microvilli seem to have dynamic lengths, elon-gating and retracting on a timescale of minutes101,102.

The cellular environment from which microvilli grow seems to be fun-damentally different from that of most filopodia, and does not involve a clear lamellipodial or lamellar surface. The apical surface of epithelial cells contains an actin meshwork called the ‘terminal web’ (Fig. 4) that is rich in myosin II and spectrin, and the actin filaments do not seem to be dendritically branched106–108. The small diameter of blood lymphocytes

Nucleus

T cell

APC

Arp2/3

Formins(mDia1, FRL1)?

Figure 5 Actin assembly factors and immune synapse. The Arp2/3 complex localizes to the actin-rich contact site (red) at the immune synapse in T cells, and is necessary for full actin accumulation at the immune synapse. mDia1 and FRL1 localize to both the actin-rich site and to regions surrounding the MTOC (green). Knockdown of either mDia1 or FRL1 inhibits MTOC re-orientation toward the immune synapse. Effects of formins on actin polymerization at the immune synapse are controversial, with one report indicating that deletion of mDia1 ablates filament accumulation, and another indicating no effect of either mDia1 or FRL1 knockdown on filament accumulation. FRL1 and mDia1 seem to act independently at the MTOC, in view of their differential localization patterns around the MTOC, and the observations that knockdown of either inhibits MTOC re-orientation. APC, antigen-presenting cell.

© 2007 Nature Publishing Group

Page 9: Many Faces Of Actin

1118 NATURE CELL BIOLOGY VOLUME 9 | NUMBER 10 | OCTOBER 2007

R E V I E W

(<10 µm) allows about 2 µm of cytoplasm surrounding the nucleus, and it is unclear how actin filaments here are organized102. It would be useful to characterize actin organization at the base of microvilli in more detail, in light of our new knowledge of the Arp2/3 complex and formins.

We feel that it is unlikely that microvilli require the Arp2/3 complex for assembly. RNAi-mediated knockdown of Arp2 has no effect on micro-villar morphology in a lymphocyte culture line (S. Nicholson-Dykstra, Dartmouth, Hanover, NH, and H.N.H., unpublished observations). However, formins affect microvillar assembly in both Dictyostelium96 and NIH3T3 cells79. It is unclear how the latter structures relate to Cdc42-dependent filopodia in NIH3T3 cells, which also require mDia2 (ref. 94). In addition, all of these structures deviate from our definition of micro-villi, as they include both attached and unattached protrusions.

Novel insights from adhesion structuresRecent studies of adhesion structures have revealed an intricate interplay between multiple actin-based structures, as well as a role for formins in controlling microtubule-based structures.

Immunological synapse. The term ‘immunological synapse’ refers to the extensive interaction surface between a T lymphocyte and an antigen-presenting cell bound to specific antigen, mediated by both integrin-based and T-cell receptor-based interactions109. This structure is important in establishing cell–cell adhesion and T-cell polarity, and is necessary for T-cell activation. An early and requisite step in syn-apse assembly is massive actin polymerization on the T-cell side, which expands the interaction surface. Following this event, the microtubule organizing centre (MTOC) (and also the Golgi) re-orients towards the synapse, and a variety of substances are secreted towards the antigen-presenting cell.

The actin-based structure at the synapse is widely considered to be a lamellipodium110, and this is supported by recent RNAi studies23. Knockdown of either Arp2 or Arp3 in Jurkat T cells inhibits spreading or polymerization of a tight F-actin band at the synapse, depending on the experimental setting. However, polarized actin-rich finger-like protrusions persist at the synapse interface in the near absence of the Arp2/3 complex (Fig. 5). Interestingly, although two formins (mDia1 and FRL1) colocalize with F-actin at the synapse, suppression of either formin alone or in combination affects neither spreading nor the fin-ger-like protrusions. Instead, mDia1 and FRL1 also localize near the MTOC and are important for MTOC repositioning towards the synapse. Another study, using splenic T cells from mDia1 knockout mice, also results in disrupted MTOC polarization111. However, these cells also have an almost complete loss of polymerized actin staining at the synapse, in contrast with the Jurkat cell study.

Clearly, differences between the experiments could explain these inconsistencies in the role of mDia1 in actin assembly at the immuno-logical synapse. One possibility is that the residual mDia1 remaining after RNAi is sufficient for the actin response, but not for the MTOC response. Using quantitative immunoblotting, we found that Jurkat cells express 170,000 molecules of mDia1 per cell (H.N.H., unpublished observations), so that >90% suppression (estimated from westerns in ref. 23), would result in 10,000–20,000 remaining molecules. Another possibility is that mDia1 knockout in T lymphocytes results in second-ary effects on Arp2/3-complex levels or activity. It would be interest-ing to probe the levels of Arp2/3-complex subunits, or the WASp and

WAVE2 in these cells; WAVE2 seems to be the Arp2/3-complex regula-tor primarily responsible for actin assembly at the synapse112. Another consideration is the time allowed for synapse assembly in these different model systems.

The role for these formins in MTOC repositioning is not the first evi-dence of association between formins and microtubule-based structures. mDia regulates microtubule stabilization in fibroblasts22, and mDia2 can interact with microtubule-binding proteins113. In addition, there is evi-dence that mDia1 effects MTOC positioning in migrating fibroblasts, although this result is not universally accepted22,114.

Are the effects of mDia1 and FRL1 on MTOC re-orientation in T cells related to their effects on actin dynamics? Suppression of either formin alone inhibits re-orientation23, and the formins localize in different pat-terns at the MTOC, suggesting distinct roles. One study suggests that the FH2 domain is required for mDia1-mediated effects on microtubules115, which suggests a link to actin. However, the FH2 domain is also sufficient for the interaction of mDia2 with microtubule-binding proteins113, and the FH2 domain of mDia1 can interact directly with microtubules (E.S.C. and H.N.H., unpublished observations; F. Bartolini and G. Gundersen, Columbia University, New York, NY, unpublished observations). It is not currently known whether actin filaments colocalize with mDia1 or FRL1 at the MTOC in T cells, although these may be difficult to detect. One final point is that the two effects of formins on microtubules (stabiliza-tion and MTOC repositioning) may involve distinct mechanisms, as they presumably occur at different ends of microtubules.

Adherens junctions. Adherens junctions are cell–cell adhesions mediated by homophilic interaction of cadherins: their core components are the transmembrane cadherin and two cytoplasmic proteins, !-catenin and "-catenin116,117. The long-accepted model is that mature adherens junctions in epithelial cells are associated with a circumferential band of actin and myosin II that is contractile and runs parallel to the membrane118. This band of actin filaments is relatively stable, and the actin–adherens junc-tions interaction is mediated by "-catenin, bound to !-catenin, which is bound to cadherin. However, new results call this model into question: first, "-catenin cannot bind !-catenin and actin filaments simultane-ously in vitro119; second, actin filaments are much more dynamic than cadherins or catenins at the adherens junctions119; and third, factors other than core adherens junction components seem to link junctions to actin120,121.

Therefore, our image of the mature adherens junction may well change over the next few years. However, it is clear that extensive actin dynamics are required to assemble the mature structure. Cells approach each other by extension of sheet-like membrane protrusions122. On initial contact, the two interacting cells extend actin-rich finger-like protrusions toward each other, with cadherin concentrated at the tips123–125. These protru-sions interdigitate and then shorten as the junction matures, bringing the two cell membranes closer. The initial sheet-like protrusions are likely to be lamellipodia or lamella, although it is less clear whether the finger-like protrusions are filopodia-like124 or more like an acto-myosin contractile structure126. Clearly, myosin II is important for adherens junction assem-bly, but its localization has not been determined at high resolution.

There is evidence that both the Arp2/3 complex and formins partici-pate in adherens junctions assembly. The Arp2/3 complex localizes to sites of junction assembly127 and its inhibition disrupts this process127,128, although it is unclear whether this may be due to general inhibition of

© 2007 Nature Publishing Group

Page 10: Many Faces Of Actin

NATURE CELL BIOLOGY VOLUME 9 | NUMBER 10 | OCTOBER 2007 1119

R E V I E W

lamellipodial protrusion — recent evidence suggests that the Arp2/3 complex accumulates transiently at nascent junctions, but then disperses rapidly as E-cadherin accumulates129. Formin1 also localizes to assem-bling adherens junctions, concentrating at the tips of the finger-like protrusions. Formin1 binds "-catenin and overexpression of a formin1 construct containing the "-catenin-binding region disrupts adherens junction assembly130. Interestingly, "-catenin binding to actin filaments also inhibits the Arp2/3 complex131. Therefore, the interplay between "-catenin and the two actin nucleators might be crucial for adheren junction assembly. An outstanding question is whether "-catenin can bind !-catenin–cadherin and formin1 simultaneously. In addition, the recently identified ability of spire to inhibit formin1 (M. Quinlan & R. D. Mullins, unpublished observations) suggests that spire may have a role in this process.

An intricate web to weaveA simplistic conclusion, on reading this review, is that the Arp2/3 com-plex mediates dendritically branched structures and that formins control unbranched structures. Although there may be some truth to this, the number of qualifications to this statement are large.

One possibility is that formins might function more as elongation factors than nucleation factors in some processes, and so work coordi-nately with the Arp2/3 complex, spire or other formins. In fact, several formins are very poor nucleators in vitro but still bind the barbed ends of filaments strongly, modulating elongation rate and protecting barbed ends from capping16 (E. S. Harris, Dartmouth, Hanover, NH, and H.N.H, unpublished observations). Thus, perhaps the Arp2/3 complex nucleates filaments and formins control their elongation in some circumstances. Recent results suggest that this situation could occur with mDia2 and the Arp2/3 complex in certain filopodia (T. Svitkina, unpublished obser-vations). Similarly, the ability of spire to both nucleate filaments and to regulate formin1 suggests an intricate interplay between these two factors, although spire-dependent actin filaments have not yet been identified in cells.

A related point is that these simple actin-based structures do not act alone, but interface with other actin-based structures, or microtubule-based structures. The interface between lamellipodia and lamella, and between lamellipodia and filopodia, provide clear examples of this. Both immunological synapses and adherens junctions are even more complex cases. Dissecting these interfaces will be important for understanding the complexity of cytoskeletal organization and the control of its assembly.

ACKNOWLEDGEMENTSWe are indebted to many for useful discussions, including A. Alberts, D. Billadeau, J. Burkhardt, J. Condeelis, F. Flures, G. Gundersen, M. McNiven, D. Mullins, S. Nicholson-Dykstra, T. Svitkina and C. Waterman-Storer. We also thank our anonymous reviewers, whose comments improved this work immensely. This work was supported by National Institutes of Health grant GM069818 and by a Pew Biomedical Scholars Award.

1. Pollard, T. D., Blanchoin, L. & Mullins, R. D. Molecular mechanisms controlling actin filament dynamics in nonmuscle cells. Annu. Rev. Biophys. Biomol. Struct. 29, 545–576 (2000).

2. Moseley, J. B. & Goode, B. L. The yeast actin cytoskeleton: from cellular function to biochemical mechanism. Microbiol. Mol. Biol. Rev. 70, 605–645 (2006).

3. Pollard, T. D. & Cooper, J. A. Actin and actin-binding proteins. A critical evaluation of mechanisms and functions. Annu. Rev. Biochem. 55, 987–1035 (1986).

4. Wang, Y. L. Exchange of actin subunits at the leading edge of living fibroblasts: possible role of treadmilling. J. Cell Biol. 101, 597–602 (1985).

5. Mallavarapu, A. & Mitchison, T. Regulated actin cytoskeleton assembly at filopodium tips controls their extension and retraction. J. Cell Biol. 146, 1097–1106 (1999).

6. Theriot, J. A. & Mitchison, T. J. Actin microfilament dynamics in locomoting cells. Nature 352, 126–131 (1991).

7. Rzadzinska, A. K. et al. An actin molecular treadmill and myosins maintain stereocilia functional architecture and self-renewal. J. Cell Biol. 164, 887–897 (2004).

8. Goley, E. D. & Welch, M. D. The ARP2/3 complex: an actin nucleator comes of age. Nature Rev. Mol. Cell Biol. 7, 713–726 (2006).

9. Higgs, H. N. & Pollard, T. D. Regulation of actin filament formation through Arp2/3 complex: Activation by a Diverse Array of Proteins. Annu. Rev. Biochem. 70, 649–676 (2001).

10. Machesky, L. M. et al. Mammalian actin-related protein 2/3 complex localizes to regions of lamellipodial protrusion and is composed of evolutionarily conserved pro-teins. Biochem. J. 328, 105–112 (1997).

11. Jay, P. et al. ARP3!, the gene encoding a new human actin-related protein, is alterna-tively spliced and predominantly expressed in brain neuronal cells. Eur. J. Biochem. 267, 2921–2928 (2000).

12. Millard, T. H. et al. Identification and characterisation of a novel human isoform of Arp2/3 complex subunit p16-ARC/ARPC5. Cell Motil. Cytoskeleton 54, 81–90 (2003).

13. Hudson, A. M. & Cooley, L. A subset of dynamic actin rearrangements in Drosophila requires the Arp2/3 complex. J. Cell Biol. 156, 677–687 (2002).

14. Higgs, H. N. Formin proteins: a domain-based approach. Trends Biochem. Sci. 30, 342–353 (2005).

15. Kovar, D. R. Molecular details of formin-mediated actin assembly. Curr. Opin. Cell Biol. 18, 11–17 (2006).

16. Harris, E. S., Li, F. & Higgs, H. N. The mouse formin, FRLa, slows actin filament barbed end elongation, competes with capping protein, accelerates polymerization from monomers, and severs filaments. J. Biol. Chem. 279, 20076–20087 (2004).

17. Harris, E. S., Rouiller, I., Hanein, D. & Higgs, H. N. Mechanistic differences in actin bundling activity of two mammalian formins, FRL1 and mDia2. J. Biol. Chem. 281, 14383–14392 (2006).

18. Moseley, J. B. & Goode, B. L. Differential activities and regulation of Saccharomyces cerevisiae formin proteins Bni1 and Bnr1 by Bud6. J. Biol. Chem. 280, 28023–28033 (2005).

19. Michelot, A. et al. The formin homology 1 domain modulates the actin nucleation and bundling activity of Arabidopsis FORMIN1. Plant Cell 17, 2296–2313 (2005).

20. Chhabra, E. S. & Higgs, H. N. INF2 is a WH2 motif-containing formin that severs actin filaments and accelerates both polymerization and depolymerization. J. Biol. Chem. 281, 26754–26767 (2006).

21. Higgs, H. N. & Peterson, K. J. Phylogenetic analysis of the formin homology 2 (FH2) domain. Mol. Biol. Cell 16, 1–13 (2005).

22. Eng, C. H., Huckaba, T. M. & Gundersen, G. G. The formin mDia regulates GSK3 through novel PKCs to promote microtubule stabilization but not MTOC reorientation in migrating fibroblasts. Mol. Biol. Cell 17, 5004–5016 (2006).

23. Gomez, T. S. et al. Formins regulate the actin-related protein 2/3 complex-independent polarization of the centrosome to the immunological synapse. Immunity 26, 177–190 (2007).

24. Wallar, B. J. & Alberts, A. S. The formins: active scaffolds that remodel the cytoskel-eton. Trends Cell Biol. 13, 435–446 (2003)

25. Kerkhoff, E. Cellular functions of the Spir actin-nucleation factors. Trends Cell Biol. 16, 477–483 (2006).

26. Quinlan, M. E., Heuser, J. E., Kerkhoff, E. & Mullins, R. D. Drosophila Spire is an actin nucleation factor. Nature 433, 382–388 (2005).

27. Rosales-Nieves, A. E. et al. Coordination of microtubule and microfilament dynam-ics by Drosophila Rho1, Spire and Cappuccino. Nature Cell Biol. 8, 367–376 (2006).

28. Abercrombie, M., Heaysman, J. E. & Pegrum, S. M. The locomotion of fibroblasts in culture. I. Movements of the leading edge. Exp. Cell Res. 59, 393–398 (1970).

29. Abercrombie, M., Heaysman, J. E. & Pegrum, S. M. The locomotion of fibroblasts in culture. II. “Ruffling”. Exp. Cell Res. 60, 437–444 (1970).

30. Abercrombie, M., Heaysman, J. E. & Pegrum, S. M. The locomotion of fibroblasts in culture. 3. Movements of particles on the dorsal surface of the leading lamella. Exp. Cell Res. 62, 389–398 (1970).

31. Abercrombie, M., Heaysman, J. E. & Pegrum, S. M. The locomotion of fibroblasts in culture. IV. Electron microscopy of the leading lamella. Exp. Cell Res. 67, 359–367 (1971).

32. Izzard, C. S. & Lochner, L. R. Cell-to-substrate contacts in living fibroblasts: an inter-ference reflexion study with an evaluation of the technique. J. Cell Sci. 21, 129–159 (1976).

33. Bailly, M. et al. Regulation of protrusion shape and adhesion to the substratum during chemotactic responses of mammalian carcinoma cells. Exp. Cell Res. 241, 285–299 (1998).

34. Gupton, S. L. & Waterman-Storer, C. M. Spatiotemporal feedback between actomyosin and focal-adhesion systems optimizes rapid cell migration. Cell 125, 1361–1374 (2006).

35. Svitkina, T., Verkhovsky, A. B., McQuade, K. M. & Borisy, G. G. Analysis of the actin-myosin II system in fish epidermal keratocytes: mechanism of cell body translocation. J. Cell Biol. 139, 397–415 (1997).

36. Svitkina, T. M. & Borisy, G. G. Arp2/3 complex and actin depolymerizing factor/cofilin in dendritic organization and treadmilling of actin filament array in lamellipodia. J. Cell Biol. 145, 1009–1026 (1999).

37. Ponti, A. et al. Two distinct actin networks drive the protrusion of migrating cells. Science 305, 1782–1786 (2004).

© 2007 Nature Publishing Group

Page 11: Many Faces Of Actin

1120 NATURE CELL BIOLOGY VOLUME 9 | NUMBER 10 | OCTOBER 2007

R E V I E W

38. Mullins, R. D., Heuser, J. A. & Pollard, T. D. The interaction of Arp2/3 complex with actin: nucleation, high affinity pointed end capping, and formation of branching net-works of filaments. Proc. Natl Acad. Sci. USA 95, 6181–6186 (1998).

39. Waterman-Storer, C. M., Desai, A., Bulinski, J. C. & Salmon, E. D. Fluorescent speckle microscopy, a method to visualize the dynamics of protein assemblies in living cells. Curr. Biol. 8, 1227–1230 (1998).

40. Iwasa, J. H. & Mullins, R. D. Spatial and temporal relationships between actin-filament nucleation, capping, and disassembly. Curr. Biol. 17, 395–406 (2007).

41. Small, J. V., Herzog, M. & Anderson, K. Actin filament organization in the fish keratocyte lamellipodium. J. Cell Biol. 129, 1275–1286 (1995).

42. Svitkina, T. M., Shevelev, A. A., Bershadsky, A. D. & Gelfand, V. I. Cytoskeleton of mouse embryo fibroblasts. Electron microscopy of platinum replicas. Eur. J. Cell Biol. 34, 64–74 (1984).

43. Giannone, G. et al. Lamellipodial actin mechanically links myosin activity with adhe-sion-site formation. Cell 128, 561–575 (2007).

44. Svitkina, T. M. et al. Mechanism of filopodia initiation by reorganization of a dendritic network. J. Cell Biol. 160, 409–421 (2003).

45. Welch, M. D. et al. The human Arp2/3 complex is composed of evolutionarily conserved subunits and is localized to cellular regions of dynamic actin filament assembly. J. Cell Biol. 138, 375–384 (1997).

46. Bailly, M. et al. Relationship between Arp2/3 complex and the barbed ends of actin filaments at the leading edge of carcinoma cells after epidermal growth factor stimula-tion. J. Cell Biol. 145, 331–345 (1999).

47. Falet, H. et al. Importance of free actin filament barbed ends for Arp2/3 complex function in platelets and fibroblasts. Proc. Natl Acad. Sci. USA 99, 16782–16787 (2002).

48. Strasser, G. A. et al. Arp2/3 is a negative regulator of growth cone translocation. Neuron 43, 81–94 (2004).

49. Miyoshi, T. et al. Actin turnover-dependent fast dissociation of capping protein in the dendritic nucleation actin network: evidence of frequent filament severing. J. Cell Biol. 175, 947–955 (2006).

50. Gupton, S. L. et al. Cell migration without a lamellipodium: translation of actin dynamics into cell movement mediated by tropomyosin. J. Cell Biol. 168, 619–631 (2005).

51. Bailly, M. et al. The F-actin side binding activity of the Arp2/3 complex is essential for actin nucleation and lamellipod extension. Curr. Biol. 11, 620–625 (2001).

52. Steffen, A. et al. Filopodia formation in the absence of functional WAVE- and Arp2/3-complexes. Mol. Biol. Cell. 17, 2581–2591 (2006).

53. Rogers, S. L., Wiedemann, U., Stuurman, N. & Vale, R. D. Molecular requirements for actin-based lamella formation in Drosophila S2 cells. J. Cell Biol. 162, 1079–1088 (2003).

54. Di Nardo, A. et al. Arp2/3 complex-deficient mouse fibroblasts are viable and have normal leading-edge actin structure and function. Proc. Natl Acad. Sci. USA 102, 16263–16268 (2005).

55. Mogilner, A. & Edelstein-Keshet, L. Regulation of actin dynamics in rapidly moving cells: a quantitative analysis. Biophys. J. 83, 1237–1258 (2002).

56. Suetsugu, S., Yamazaki, D., Kurisu, S. & Takenawa, T. Differential roles of WAVE1 and WAVE2 in dorsal and peripheral ruffle formation for fibroblast cell migration. Dev. Cell 5, 595–609 (2003).

57. Buccione, R., Orth, J. D. & McNiven, M. A. Foot and mouth: podosomes, invadopodia and circular dorsal ruffles. Nature Rev. Mol. Cell Biol. 5, 647–657 (2004).

58. Legg, J. A. et al. N-WASP involvement in dorsal ruffle formation in mouse embryonic fibroblasts. Mol. Biol Cell. 18, 678–687 (2007).

59. Svitkina, T. Electron microscopic analysis of the leading edge in migrating cells. Methods Cell Biol. 79, 295–319 (2007).

60. Goulimari, P. et al. G"12/13 is essential for directed cell migration and localized Rho-Dia1 function. J. Biol. Chem. 280, 42242–42251 (2005).

61. Kurokawa, K. & Matsuda, M. Localized RhoA activation as a requirement for the induc-tion of membrane ruffling. Mol. Biol. Cell 16, 4294–4303 (2005).

62. Pertz, O., Hodgson, L., Klemke, R. L. & Hahn, K. M. Spatiotemporal dynamics of RhoA activity in migrating cells. Nature 440, 1069–1072 (2006).

63. Watanabe, N. et al. p140mDia, a mammalian homolog of Drosophila diaphanous, is a target protein for Rho small GTPase and is a ligand for profilin. EMBO J. 16, 3044–3056 (1997).

64. Aderem, A. & Underhill, D. M. Mechanisms of phagocytosis in macrophages. Annu. Rev. Immunol. 17, 593–623 (1999).

65. Niedergang, F. & Chavrier, P. Signaling and membrane dynamics during phagocytosis: many roads lead to the phagos(R)ome. Curr. Opin. Cell Biol. 16, 422–428 (2004).

66. May, R. C., Caron, E., Hall, A. & Machesky, L. M. Involvement of the Arp2/3 complex in phagocytosis mediated by Fc#R or CR3. Nature Cell Biol. 2, 246–248 (2000).

67. Lorenzi, R. et al. Wiskott-Aldrich syndrome protein is necessary for efficient IgG-medi-ated phagocytosis. Blood 95, 2943–2946 (2000).

68. Colucci-Guyon, E. et al. A role for mammalian diaphanous-related formins in comple-ment receptor (CR3)-mediated phagocytosis in macrophages. Curr. Biol. 15, 2007–2012 (2005).

69. Linder, S. The matrix corroded: podosomes and invadopodia in extracellular matrix degradation. Trends Cell Biol. 17, 107–117 (2007).

70. Yamaguchi, H., Pixley, F. & Condeelis, J. Invadopodia and podosomes in tumor invasion. Eur. J. Cell Biol. 85, 213–218 (2006).

71. Goicoechea, S. et al. Palladin binds to Eps8 and enhances the formation of dorsal ruffles and podosomes in vascular smooth muscle cells. J. Cell Sci. 119, 3316–3324 (2006).

72. Moreau, V. et al. Cdc42-driven podosome formation in endothelial cells. Eur. J. Cell Biol. 85, 319–325 (2006).

73. Hufner, K. et al. The VC region of Wiskott-Aldrich syndrome protein induces Arp2/3 complex-dependent actin nucleation. J. Biol. Chem. 276, 35761–35767 (2001).

74. Linder, S. et al. The polarization defect of Wiskott-Aldrich syndrome macrophages is linked to dislocalization of the Arp2/3 complex. J. Immunol. 165, 221–225 (2000).

75. Burns, S. et al. Configuration of human dendritic cell cytoskeleton by Rho GTPases, the WAS protein, and differentiation. Blood 98, 1142–1149 (2001).

76. Linder, S., Nelson, D., Weiss, M. & Aepfelbacher, M. Wiskott-Aldrich syndrome protein regulates podosomes in primary human macrophages. Proc. Natl Acad. Sci. USA 96, 9648–9653 (1999).

77. Mizutani, K. et al. Essential role of neural Wiskott-Aldrich syndrome protein in podo-some formation and degradation of extracellular matrix in src-transformed fibroblasts. Cancer Res. 62, 669–674 (2002).

78. Yamaguchi, H. et al. Molecular mechanisms of invadopodium formation: the role of the N-WASP-Arp2/3 complex pathway and cofilin. J. Cell Biol. 168, 441–452 (2005).

79. Pellegrin, S. & Mellor, H. The Rho family GTPase Rif induces filopodia through mDia2. Curr. Biol. 15, 129–133 (2005).

80. DeRosier, D. J. & Tilney, L. G. F-actin bundles are derivatives of microvilli: What does this tell us about how bundles might form? J. Cell Biol. 148, 1–6 (2000).

81. Faix, J. & Rottner, K. The making of filopodia. Curr. Opin. Cell Biol. 18, 18–25 (2006).

82. Zheng, J. Q., Wan, J. J. & Poo, M. M. Essential role of filopodia in chemotropic turning of nerve growth cone induced by a glutamate gradient. J. Neurosci. 16, 1140–1149 (1996).

83. Galbraith, C. G., Yamada, K. M. & Galbraith, J. A. Polymerizing actin fibers position integrins primed to probe for adhesion sites. Science 315, 992–995 (2007).

84. Lehmann, M. J. et al. Actin- and myosin-driven movement of viruses along filopodia precedes their entry into cells. J. Cell Biol. 170, 317–325 (2005).

85. Sherer, N. M. et al. Retroviruses can establish filopodial bridges for efficient cell-to-cell transmission. Nature Cell Biol. 9, 310–315 (2007).

86. Lidke, D. S. et al. Reaching out for signals: filopodia sense EGF and respond by directed retrograde transport of activated receptors. J. Cell Biol. 170, 619–626 (2005).

87. Jontes, J. D. & Smith, S. J. Filopodia, spines, and the generation of synaptic diversity. Neuron 27, 11–14 (2000).

88. Lindberg, U., Hoglund, A. S. & Karlsson, R. On the ultrastructural organization of the microfilament system and the possible role of profilactin. Biochimie 63, 307–323 (1981).

89. Small, J. V., Rinnerthaler, G. & Hinssen, H. Organization of actin meshworks in cultured cells: the leading edge. Cold Spring Harb. Symp. Quant. Biol. 46, 599–611 (1982).

90. Lewis, A. K. & Bridgman, P. C. Nerve growth cone lamellipodia contain two populations of actin filaments that differ in organization and polarity. J. Cell Biol. 119, 1219–1243 (1992).

91. Vignjevic, D. et al. Role of fascin in filopodial protrusion. J. Cell Biol. 174, 863–875 (2006).

92. Miki, H., Sasaki, T., Takai, Y. & Takenawa, T. Induction of filopodium formation by a WASP-related actin-depolymerizing protein N-WASP. Nature 391, 93–96 (1998).

93. Snapper, S. B. et al. N-WASP deficiency reveals distinct pathways for cell surface pro-jections and microbial actin-based motility. Nature Cell Biol. 3, 897–904 (2001).

94. Peng, J. et al. Disruption of the Diaphanous-related formin Drf1 gene encoding mDia1 reveals a role for Drf3 as an effector for Cdc42. Curr. Biol. 13, 534–545 (2003).

95. Wallar, B. J. et al. The basic region of the diaphanous-autoregulatory domain (DAD) is required for autoregulatory interactions with the diaphanous-related formin inhibitory domain. J. Biol. Chem. 281, 4300–4307 (2006).

96. Schirenbeck, A. et al. The Diaphanous-related formin dDia2 is required for the forma-tion and maintenance of filopodia. Nature Cell Biol. 7, 619–625 (2005).

97. Brieher, W. M., Coughlin, M. & Mitchison, T. J. Fascin-mediated propulsion of Listeria monocytogenes independent of frequent nucleation by the Arp2/3 complex. J. Cell Biol. 165, 233–242 (2004).

98. Vignjevic, D. et al. Formation of filopodia-like bundles in vitro from a dendritic network. J. Cell Biol. 160, 951–962 (2003).

99. Medalia, O. et al. Organization of actin networks in intact filopodia. Curr. Biol. 17, 79–84 (2007).

100.Mooseker, M. S. & Tilney, L. G. Organization of an actin filament-membrane complex. Filament polarity and membrane attachment in the microvilli of intestinal epithelial cells. J. Cell Biol. 67, 725–743 (1975).

101.Gorelik, J. et al. Dynamic assembly of surface structures in living cells. Proc. Natl Acad. Sci. USA 100, 5819–5822 (2003).

102.Majstoravich, S. et al. Lymphocyte microvilli are dynamic, actin-dependent structures that do not require Wiskott-Aldrich syndrome protein (WASp) for their morphology. Blood 104, 1396–1403 (2004).

103.von Andrian, U. H. et al. A central role for microvillous receptor presentation in leu-kocyte adhesion under flow. Cell 82, 989–999 (1995).

104.Singer, II et al. CCR5, CXCR4, and CD4 are clustered and closely apposed on microvilli of human macrophages and T cells. J. Virol. 75, 3779–3790 (2001).

105.Tyska, M. J. & Mooseker, M. S. MYO1A (brush border myosin I) dynamics in the brush border of LLC–PK1–CL4 cells. Biophys. J. 82, 1869–1883 (2002).

106.Hirokawa, N., Tilney, L. G., Fujiwara, K. & Heuser, J. E. Organization of actin, myosin, and intermediate filaments in the brush border of intestinal epithelial cells. J. Cell Biol. 94, 425–443 (1982).

107.Heintzelman, M. B. & Mooseker, M. S. Assembly of the intestinal brush border cytoskeleton. Curr. Top. Dev. Biol. 26, 93–122 (1992).

© 2007 Nature Publishing Group

Page 12: Many Faces Of Actin

NATURE CELL BIOLOGY VOLUME 9 | NUMBER 10 | OCTOBER 2007 1121

R E V I E W

108.Bretscher, A. Microfilament structure and function in the cortical cytoskeleton. Annu. Rev. Cell Biol. 7, 337–374 (1991).

109.Dustin, M. L. A dynamic view of the immunological synapse. Semin. Immunol. 17, 400–410 (2005).

110.Billadeau, D. D. & Burkhardt, J. K. Regulation of cytoskeletal dynamics at the immune synapse: new stars join the actin troupe. Traffic 7, 1451–1460 (2006).

111.Eisenmann, K. M. et al. T cell responses in mammalian Diaphanous-related formin mDia1 knock-out mice. J. Biol. Chem. 282, 25152–25158 (2007).

112.Nolz, J. C. et al. The WAVE2 complex regulates actin cytoskeletal reorganization and CRAC-mediated calcium entry during T cell activation. Curr. Biol. 16, 24–34 (2006).

113.Wen, Y. et al. EB1 and APC bind to mDia to stabilize microtubules downstream of Rho and promote cell migration. Nature Cell Biol. 6, 820–830 (2004).

114.Yamana, N. et al. The Rho–mDia1 pathway regulates cell polarity and focal adhesion turnover in migrating cells through mobilizing Apc and c-Src. Mol. Cell. Biol. 26, 6844–6858 (2006).

115.Ishizaki, T. et al. Coordination of microtubules and the actin cytoskeleton by the Rho effector mDia1. Nature Cell Biol. 3, 8–14 (2001).

116.Aberle, H. et al. Assembly of the cadherin-catenin complex in vitro with recombinant proteins. J. Cell Sci. 107, 3655–3663 (1994).

117.Nathke, I. S. et al. Defining interactions and distributions of cadherin and catenin complexes in polarized epithelial cells. J. Cell Biol. 125, 1341–1352 (1994).

118.Gates, J. & Peifer, M. Can 1000 reviews be wrong? Actin, "-catenin, and adherens junctions. Cell 123, 769–772 (2005).

119.Yamada, S. et al. Deconstructing the cadherin-catenin-actin complex. Cell 123, 889–901 (2005).

120.Pilot, F., Philippe, J. M., Lemmers, C. & Lecuit, T. Spatial control of actin organization at adherens junctions by a synaptotagmin-like protein Btsz. Nature 442, 580–584 (2006).

121.Tamada, M., Perez, T. D., Nelson, W. J. & Sheetz, M. P. Two distinct modes of myosin assembly and dynamics during epithelial wound closure. J. Cell Biol. 176, 27–33 (2007).

122.Adams, C. L., Nelson, W. J. & Smith, S. J. Quantitative analysis of cadherin–cat-enin–actin reorganization during development of cell–cell adhesion. J. Cell Biol. 135, 1899–1911 (1996).

123.Adams, C. L., Chen, Y. T., Smith, S. J. & Nelson, W. J. Mechanisms of epithelial cell–cell adhesion and cell compaction revealed by high-resolution tracking of E-cad-herin-green fluorescent protein. J. Cell Biol. 142, 1105–1119 (1998).

124.Vasioukhin, V., Bauer, C., Yin, M. & Fuchs, E. Directed actin polymerization is the driving force for epithelial cell–cell adhesion. Cell 100, 209–219 (2000).

125.Yonemura, S., Itoh, M., Nagafuchi, A. & Tsukita, S. Cell-to-cell adherens junc-tion formation and actin filament organization: similarities and differences between non-polarized fibroblasts and polarized epithelial cells. J. Cell Sci. 108, 127–142 (1995).

126.Vaezi, A., Bauer, C., Vasioukhin, V. & Fuchs, E. Actin cable dynamics and Rho/Rock orchestrate a polarized cytoskeletal architecture in the early steps of assembling a stratified epithelium. Dev. Cell 3, 367–381 (2002).

127.Ivanov, A. I. et al. Differential roles for actin polymerization and a myosin II motor in assembly of the epithelial apical junctional complex. Mol. Biol Cell. 16, 2636–2650 (2005).

128.Verma, S. et al. Arp2/3 activity is necessary for efficient formation of E-cadherin adhesive contacts. J. Biol. Chem. 279, 34062–34070 (2004).

129.Yamada, S. & Nelson, W. J. Localized zones of Rho and Rac activities drive initiation and expansion of epithelial cell cell adhesion. J. Cell Biol. 178, 517–527 (2007).

130.Kobielak, A., Pasolli, H. A. & Fuchs, E. Mammalian formin-1 participates in adhe-rens junctions and polymerization of linear actin cables. Nature Cell Biol. 6, 21–30 (2004).

131.Drees, F. et al. "-catenin is a molecular switch that binds E-cadherin–!-catenin and regulates actin-filament assembly. Cell 123, 903–915 (2005).

132.Abercrombie, M. & Ambrose, E. J. Interference microscope studies of cell contacts in tissue culture. Exp Cell Res. 15, 332–345 (1958).

133.Ingram, V. M. A side view of moving fibroblasts. Nature 222, 641–644 (1969).134.Abercrombie, M. The crawling movement of metazoan cells. Proc. R. Soc. Lond. 207,

129–147 (1980).

© 2007 Nature Publishing Group