Assembling an actin cytoskeleton for cell attachment and movement

Review

Assembling an actin cytoskeleton for cell attachment and movement

J. Victor Small *, K. Rottner, I. Kaverina, K.I. AndersonInstitute of Molecular Biology, Austrian Academy of Sciences, BillrothstraMe 11, A-5020 Salzburg, Austria

Received 15 April 1998; revised 18 June 1998; accepted 30 June 1998

Keywords: Cytoskeleton; Rho; Actin; Microtubule; Motility

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 271

2. The ¢broblast . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2722.1. Lamellipodia as ¢lament factories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2722.2. Ventral stress ¢bre assembly . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2732.3. Arcs and dorsal stress ¢bres . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2752.4. Concave cell edges . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 275

3. The keratocyte . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2753.1. Generating and recycling an actin cytoskeleton . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 275

4. The neuronal growth cone . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2764.1. Lamellipodia and ¢lopodia in partnership . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 276

5. Microtubules as modulators of the actin cytoskeleton . . . . . . . . . . . . . . . . . . . . . . . . . . . . 278

6. Conclusions and perspectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 279

Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 280

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 280

1. Introduction

The crawling movements of metazoan cells resultfrom coordinated and polarised changes in cellshape, orchestrated via a continuous remodelling of

the actin cytoskeleton. The signalling pathways thatlead to remodelling involve the rho-family of smallGTPases [27,52,56,70,72] whose individual membersinduce the expression of di¡erent actin ¢lament sub-compartments, generally categorised as lamellipodia,¢lopodia, stress ¢bres and arcs ([32,61]; Fig. 1). De-pending on the cell type and conditions, the relativeexpression of these di¡erent actin ¢lament assemblies

0167-4889 / 98 / $ ^ see front matter ß 1998 Elsevier Science B.V. All rights reserved.PII: S 0 1 6 7 - 4 8 8 9 ( 9 8 ) 0 0 0 8 0 - 9

* Corresponding author. Fax: +43 (662) 6396129.

BBAMCR 14355 8-9-98

Biochimica et Biophysica Acta 1404 (1998) 271^281

is highly variable and leads to corresponding di¡er-ences in cell form and motility. In this short reviewwe present some ideas and observations about thegeneration of the actin cytoskeleton and the alterna-tive strategies that some selected cells adopt tospread and move over a substrate.

We will begin by considering the ¢broblast sincethis cell type is rich in di¡erent actin assemblies (Fig.1). Attention will then be turned to a fast moving celltype, the ¢sh keratocyte, and then to the tetheredneuronal growth cone. Two underlying hypotheseson which we will elaborate are, ¢rst, that the ¢la-ments of the di¡erent subcompartments of the actincytoskeleton all ¢nd their origin in lamellipodia, andsecond, that the form of the cytoskeleton is largelydetermined by the pattern of substrate contacts ini-tiated by this actin rich, protrusive organelle. Theway that microtubules may interface with the contactforming machinery and thereby modulate the assem-bly of the actin cytoskeleton will also be discussed.

2. The ¢broblast

2.1. Lamellipodia as ¢lament factories

When £uorescently conjugated G-actin is microin-jected into ¢broblasts, it ¢rst becomes incorporated

within a few minutes in the peripheral lamellipodia[23,40,48]. Shortly afterwards, it appears in focalcontacts close to the cell front and then, much later,along the stress ¢bre bundles and in the arc-shapedactin arrays observed on the dorsal cell surface[23,32,40,81]. The incorporation of actin in lamelli-podia is intimately linked with the protrusion ofthese membrane veils, whose upfolding motionsgive rise to `ru¥es' [1]. Di¡erent lines of evidencehave established that actin monomers are incorpo-rated in the membrane bordering the front of thelamellipodium [21,54,71,82]. At this leading boun-dary, actin ¢laments are nucleated with their fastgrowing, plus ends oriented outwards [63,65,69]and monomers £ux through the ¢laments from frontto rear, as in a treadmill [82]. We propose that ¢la-ments generated here are directly used in the forma-tion of other actin structures or at least seed theirassembly.

To illustrate this idea, it is useful to consider howa cell spreads, since in this process the di¡erent actinarrays are ¢rst formed. Fig. 2 is a schematic illustra-tion of an idealised spreading cell in which the actincytoskeleton is highlighted. Spreading is mediated bypredominantly the Rac-dependent outward growthof lamellipodia (for the general morphology ofspreading ¢broblastic cells see [78]). Depending onthe cell type or the substrate conditions, the lamelli-

Fig. 1. Schematic illustration of di¡erent actin ¢lament `subcompartments' in the actin cytoskeleton of a spread ¢broblast. LAM, la-mellipodium; MS, microspike; FIL, ¢lopodium; P.B., peripheral bundle; P.B.L., peripheral bundle with lamellipodium; v.S.F., ventralstress ¢bre; d.S.F., dorsal stress ¢bre; ARC, arc; N, nucleus.

BBAMCR 14355 8-9-98

J.V. Small et al. / Biochimica et Biophysica Acta 1404 (1998) 271^281272

podium itself may consist of a homogeneous actin¢lament network, or one that is adorned with varia-ble numbers of radially arranged actin bundles (seee.g. [2,28]). If these bundles extend signi¢cantly be-yond the edge of the lamellipodium they are gener-ally referred to as `¢lopodia' and, when only margin-ally, as `microspikes'. We have previously proposed[64] that these bundles form by the convergence ofactin ¢laments in the lamellipodium meshwork (Fig.3), mediated by a lateral ¢lament £ow [62] and theengagement of actin bundling proteins, such as fascin[2,50]; the pathway signalling their induction in-volves Cdc42 [39,53].

2.2. Ventral stress ¢bre assembly

Now let us consider how the meshworks and ¢la-ment bundles of the lamellipodium could give rise tothe other components of the actin cytoskeleton,starting with the stress ¢bre bundles on the ventralcell surface. In a spreading ¢broblast, the ventralstress ¢bres commonly extend from an anchoragesite, a focal adhesion [11], close to the cell edge, toone with a perinuclear location. Stress ¢bre assemblyinvolves two major steps: (1) the establishment offocal adhesion sites and (2) the recruitment of actinand associated proteins into a contractile bundle be-tween them. These processes are initiated via the Rhosignalling pathway [27].

Izzard and Lochner [36] showed for ¢broblaststhat new focal adhesions are formed within and atthe base of the lamellipodium. The sites where thesecontacts are established are also marked by localisedru¥ing activity, associated with upfolding of the la-mellipodium or with small focalised increases inphase density (mini-ru¥es [57]). At least some ofthese sites are occupied temporarily by microspikebundles [17], which appear in these cases to play arole in the initiation of stress ¢bre formation (Fig. 2).A stub of actin ¢laments formed via a microspike isno doubt a favourable template for stress ¢bre as-sembly; however, localised bundling of actin ¢la-ments at nascent contact sites within the lamellipo-dium would also su¤ce. The point to be made here issimply that by one route or another ¢laments pro-duced in the lamellipodium undergo localised bun-dling and that such precursor bundles serve to nucle-ate stress ¢bre assembly.

How the other end of a ventral stress ¢bre mayform is at ¢rst less obvious. Heath and Dunn [30]have suggested that a centrally situated anchoragesite may be provided by a perinuclear network ofactin ¢laments. But this would not explain how stress¢bres in spreading cells commonly traverse beneaththe nucleus, or how the perinuclear end would beanchored to the substrate. We propose that the an-chorage sites in the inner regions of the cell are pro-duced in lamellipodia, just as above, but during theearly stages of spreading. At this time, short stress¢bres (less than 10 Wm long) can be found that spanbeneath the cell body (unpublished observations). Itis reasonable to assume that these ¢bres form be-tween two nascent adhesion sites established in la-mellipodia on opposite sides of the freshly attachedcell. In view of the relatively small separation in-volved, ¢laments polymerised from one adhesionsite could be long enough to overlap with thosefrom the other site so as to initiate the assembly ofa bundle between them (Fig. 2).

Two recent ¢ndings may be called upon to explainbundle formation. First, it has recently been shownthat actomyosin-based contractility is involved in theformation and maintenance of stress ¢bre bundles[12]. And second, myosin assemblies appear at avery early stage of stress ¢bre formation and mayact to recruit actin ¢laments into them [80]. Sincethe polarity of actin ¢laments changes from unipolarin the focal adhesion to mixed polar along the stress¢bre [5,15,59] it has been argued [81] that alreadyformed actin ¢laments are added to stress ¢bres dur-ing assembly, rather than simply arising throughelongation from the contact site. This would be con-sistent with the slow incorporation of microinjectedG-actin into stress ¢bres [81]. In the present schemewe surmise that free actin ¢laments are severed fromthe base of the lamellipodium and that these ¢la-ments contribute to a cytoplasmic pool that is re-cruited for stress ¢bre assembly. The idea that re-cruitment via myosin may facilitate the polaritysorting and registration of actin ¢laments in thestress ¢bres [80] is in this regard an attractive possi-bility.

Returning to the problem above, how are longerstress ¢bres generated during spreading? This couldoccur in a stepwise fashion as illustrated in Fig. 2. Asthe lamellipodium extends the cell border outwards,

BBAMCR 14355 8-9-98

J.V. Small et al. / Biochimica et Biophysica Acta 1404 (1998) 271^281 273

Fig. 2. Schematic diagram of an idealised ¢broblast actin cytoskeleton illustrating the proposed origin of the di¡erent actin subcom-partments. The lamellipodium (LAM) is the primary site of actin polymerisation. Lateral £ow of ¢laments in the lamellipodium cangive rise to microspikes (MS), which can extend to form ¢lopodia (FIL). The lamellipodium ¢laments can lay down their distal por-tions at the base of the lamellipodium to form convex arcs or straight or concave peripheral bundles (P.B.), both in collaborationwith myosin. Filaments from the lamellipodium may also extend deeper into the cytoplasm and become cleaved to contribute to a cy-toplasmic pool of single ¢laments. Focal contacts in association with ventral or dorsal stress ¢bres (v.S.F. and d.S.F.) develop fromprecursor contacts (a,b) formed in the lamellipodium. For these sites to develop, actin ¢laments must be recruited into bundles (b,c)and the bundles dissociated from the advancing lamellipodium (d). This would go hand in hand with stress ¢bre formation. How astress ¢bre may initially form and elongate is shown by steps 1, 2 and 3. Two focal contacts (1 and 2) are formed at the base of la-mellipodia at an initial stage of spreading (not shown). These extend actin ¢laments to overlap, and via the recruitment of myosinand the activation of contractility a primary bundle is formed. As the cell spreads, the contact at `1' remains intact but the contact at`2' is overlapped by ¢laments from a new, more peripheral contact (3). Filaments from this new contact, with the aid of myosin re-cruitment, fuse with the bundle 1, 2 and the contact at 2 eventually dissociates, allowing the stress ¢bre to elongate. Dorsal stress ¢-bres (d.S.F.) extend between a contact at the base of the lamellipodium and an arc (ARC). They are envisaged to arise via the myosinaided recruitment of ¢laments from the arc and the cytoplasmic ¢lament pool. The three small G-proteins Rho, Rac and Cdc42 areplaced so as to indicate which actin subcompartments they induce. m, myosin; act, actin; f. complex, focal complex; f. contact, focalcontact. For further details, see text.

BBAMCR 14355 8-9-98


new substrate contact sites are formed. If the actinarray emanating from one of these new sites extendsover an older, more central adhesion it could, forexample in collaboration with myosin, fuse with theend of the stress ¢bre associated with this adhesionsite (Fig. 2). Longitudinal and o¡-axis fusion ofstress ¢bres has actually been documented in livingcells [81]. If more tension is now generated with thenew contact than with the old, the latter may disso-ciate and thus allow fusion and elongation of thebundle (Fig. 2). Note that the other end of the stress¢bre in this example is the original adhesion formedat the onset of spreading; it is assumed that this focaladhesion on the opposite side of the cell was notoverlapped and replaced by a more peripheral con-tact assembly with its associated ¢laments.

2.3. Arcs and dorsal stress ¢bres

Convex, arc-shaped arrays of actin ¢laments canbe observed on the dorsal surface of spreading andmigrating cells [29,67,81]. These arcs are not an-chored at focal adhesions but are contractile andhave been shown to drive the centripetal £ow ofcortical receptors [32]. Arcs form at the base of con-vex lamellipodia [32] and must have a mixed polarityof actin ¢laments to be contractile. We propose thatthe actin ¢laments of the arcs derive from a sheddingof the trailing ends of ¢laments formed in lamellipo-dia (Fig. 2).

Using arguments based on the diagonal geometryof the actin meshworks of lamellipodia [60] and theestablished treadmilling of actin ¢laments withinthem [81], we have suggested that the actin ¢lamentscan £ow laterally, the rate of lateral £ow relative tothe substrate being inversely related to the rate ofprotrusion [62]. Support for this idea comes fromthe observed lateral translation of microspikes inslowly advancing or stationary ¢broblast lamellipo-dia [17,20,66]. Thus, delivery of ¢laments of mixedpolarity into the arcs would not be from static ¢la-ments, but from ¢laments sweeping in both direc-tions along the lamellipodium and laying down theirtrailing ends parallel to its base. Breakage or cleav-age of these ¢laments from the lamellipodium wouldrelease the sheath of ¢laments so formed and allowits rearward £ow, driven by acto-myosin interactions(ARC, Fig. 2).

In an interesting treatment of actin cytoskeletonorganisation, Heath and Holi¢eld [32] have shownthat the assembly of stress ¢bres on the dorsal cellsurface goes hand in hand with the formation of thearcs. More speci¢cally, stress ¢bres were shown togrow between the base of the lamellipodium and anarc as it moved dorsally towards the perinuclear re-gion. How this process may take place is also illus-trated in Fig. 2. The contact site at the base of thelamellipodium is generated as for ventral stress ¢-bres, via bundling of lamellipodia ¢laments. Fila-ments from the arc become recruited to this site,again utilising myosin assemblies [80], and are thencontinuously fed into the ¢bre from the arc as itretracts centripetally. Filaments from the cytoplasmicpool may be required to a small or large extent tocomplement those from the arc. In this way, arraysof ¢laments with the desired mixed polarity would bedelivered into the dorsal stress ¢bre (d.S.F., Fig. 2).

2.4. Concave cell edges

The non-motile edges of non-con£uent cells arecharacteristically delineated by a concave bundle ofactin ¢laments (P.B., Fig. 2) that contains the samecomponents as stress ¢bres. The ends of these bun-dles are anchored into peripheral focal adhesions. Itis common for the same concave bundles to mark thebase of straight or concave lamellipodia (e.g. [32];Fig. 2). We suppose that these concave bundlesform in much the same way as arcs, by the recruit-ment of ¢laments from a lamellipodium. The transi-tion from a concave bundle with an associated lamel-lipodium to one without a lamellipodium isapparently accompanied by changes in adhesion tothe substrate. When a lamellipodium is present, theconcave bundle is anchored to the substrate viapunctate focal complexes [53]; Rottner and Small,unpublished; Fig. 2). When a lamellipodium is ab-sent, the concave bundles are anchored only at theirends, through focal adhesions.

3. The keratocyte

3.1. Generating and recycling an actin cytoskeleton

The idea that the lamellipodium can generate the

BBAMCR 14355 8-9-98


¢laments required to form an actin cytoskeletonfunctional in cell shape determination and motilityis well illustrated by the example of the ¢sh kerato-cyte. This cell type is the fastest of cells that use actinto move and it manages this feat using an actin cy-toskeleton noted for its simplicity [19,31,65,69]. Theadvancing cell front is marked by a crescent-shapedlamellipodium composed of a diagonal meshwork ofactin ¢laments, devoid of ¢lopodia, that dominatesthe cell form. Straddled transversely behind the la-mellipodium is the cell body, whose spindle shape ismaintained by lateral tension [3,41] exerted across itby contractile bundles of actin ¢laments [3].

Theriot and Mitchison [75] showed that the actin¢laments of the keratocyte lamellipodium do notslide relative to the substrate, so that the rate ofprotrusion of the lamellipodium re£ects directly therate of actin polymerisation. If the ¢laments of thelamellipodium are stationary, the result of their diag-onal arrangement is that their growing, anterior endsmust translate laterally along the front edge of thelamellipodium as the cell moves forward ([3] ; Fig. 4).We have suggested that such a ¢lament £ow, in theframework of the cell, could contribute to the accu-mulation of ¢laments into the bundles at the lateral£anks of the lamellipodium, that extend beneath andaround the cell body. Myosin accumulates in thesebundles and, in combination with actin, is thought toprovide the force required for cell body traction

[3,69]. The main point to be made here is that theassembly or initiation of these posterior bundles isreadily explained by a recruitment of ¢laments fromthe lamellipodium by a process analogous to thatalready put forward for the formation of arcs andbundles at the base of lamellipodia in ¢broblasts.

The continuous delivery of ¢laments to the rear ofthe keratocyte must be balanced by a continuousdisassembly, coupled with the return of actin mono-mers to the front. We have suggested that this recy-cling involves a breakdown of the ¢laments in thecell body after they have served in driving its trans-location. A special feature of the keratocyte is thatthe cell body rolls behind the lamellipodium and thisrolling motion may serve to translate ¢laments fromthe rear to the front of the cell body, where factorsactive in actin ¢lament disassembly [13] could con-vert them to monomers. Free actin monomers pro-vided in this way would then complement those re-leased from the disassembling minus ends of¢laments in the body of the lamellipodium, thus sup-plementing the pool required to maintain lamellipo-dium protrusion (Fig. 4).

4. The neuronal growth cone

4.1. Lamellipodia and ¢lopodia in partnership

The extension of a neurite (neurone or dendrite)from a neuronal cell relies on the motile activity of aso-called `growth cone' at its tip. Growth cones fea-ture ¢lopodia and lamellipodia [42], but in very var-iable proportions, depending on origin and condi-tions [25]. These protrusive outgrowths possess nostress ¢bre bundles, but the axonal process that joinsthem to the cell body bears a thin cortical sheath ofactin ¢laments [10,34,55] that is associated with my-osin [43]. The outgrowth of neurites is inhibited bydominant-negative mutants of Rac or Cdc42, but isstimulated by C3 transferase, which inhibits Rho[39,76]. Focal contacts and stress ¢bres are thus notemployed in neurite outgrowth.

Growth cone translocation in vitro involves thealternating or combined extension of ¢lopodia andlamellipodia [9,24]. In this process, lamellipodiumveils extend between neighbouring ¢lopodia or alongsingle ¢lopodia. The advance of a growth cone over

Fig. 3. Schematic illustration of the formation of a ¢lopodium(FIL.) from a narrow lamellipodium, as may occur in a growthcone. Actin ¢laments formed in the lamellipodium are recruited(R) by lateral £ow, bundled (B) by bundling proteins and fur-ther polymerised (P) to extend a ¢lopodium.

BBAMCR 14355 8-9-98


a substrate has been suggested to depend on the con-tractile activity of ¢lopodia [9,33]; however, neuriteelongation also occurs in the absence of visible ¢lo-podia [25] and embryonic growth cones exhibiting¢lopodia-free lamellipodia move twice as fast aspostnatal ones that bear ¢lopodia [38]. Nevertheless,¢lopodia are present on growth cones in vivo andwhile indispensable for protrusion are likely essentialfor axonal guidance [47,68].

In Aplysia growth cones anchored to the substratewith polylysine, there is a retrograde £ow of actinassociated with polymerisation at the anterior edge[21], as shown by Wang [82] for ¢broblasts. And therate of actin retrograde £ow, relative to the sub-

strate, is inversely proportional to the rate of growthcone advance [45]. This result supports the proposal[51] that the degree of slippage on the substrate, con-trolled by some kind of molecular clutch, determinesthe rate of productive forward movement. We haveearlier suggested [62] that the retrograde £ow of actin¢laments in a stationary lamellipodium is accompa-nied by their lateral £ow and that this contributes tothe formation of microspikes and ¢lopodia. Consis-tent with this idea are the observations that the num-ber of ¢lopodia increase as the rate of growth coneadvance decreases [38] and that ¢lopodia (or micro-spikes) are particularly prevalent in growth conesimmobilised on polylysine [21]. We therefore return

Fig. 4. Proposed model of actin ¢lament dynamics and cell movement in the keratocyte. Actin ¢laments are nucleated and polymeriseat the front edge of the lamellipodium. Owing to their diagonal orientation, ¢lament growth leads to their displacement laterally asthe cell moves (dotted lines), giving rise to a ¢lament £ow towards the lateral £anks of the lamellipodium. Filaments that reach thelateral £anks form bundles at the ends of the cell body and are retracted into the cell body cortex. The interaction of myosin and ac-tin around the cell body leads to tension development, which maintains cell body shape, and to a component of force (F) that drivesthe translocation of the cell body, involving its rotation. Force diagram indicates that the lateral components of F cancel each other,and a net forward component remains. At the base of the lamellipodium, the depolymerisation of the trailing ends of lamellipodium¢laments and of ¢laments within the cell body replenishes the actin monomer pool. Reprinted from Anderson et al. [3] with permis-sion of Rockefeller University Press.

BBAMCR 14355 8-9-98


to the theme that a lamellipodium supplies the ¢la-ments for forming or seeding bundled assemblies ofactin ¢laments (Fig. 3). The primary role of lamelli-podia in growth cone migration is further underlinedby the demonstration that the activation of Rac inneuroblastoma cells induces a dramatic outgrowth ofneurites which is blocked by dominant-negativeN17Rac, but not by dominant-negative N17Cdc42[76].

5. Microtubules as modulators of the actincytoskeleton

The disassembly of microtubules is accompaniedin ¢broblasts by a loss in cell polarity [79] and areduced spreading rate [8,35] and in neurites by theinability to grow [4,83]. Microtubule disruption alsoabolishes the directional locomotion of leucocytes ina chemotactic gradient [49]. Microtubules have thusa profound in£uence on cell polarity and migration,processes primarily dependent on the actin system.Notably, the complete disassembly of microtubulesis not necessary to induce these e¡ects; it su¤cesto inhibit only their dynamic instability. At concen-trations of microtubule inhibitors that only block thedynamic excursions of microtubule ends, the net ad-vance of ¢broblasts and neuronal growth cones ismarkedly reduced [44,73], and growth cones wanderinstead of steer at substrate boundaries [73,14].

How then do microtubules interface with the actincytoskeleton to modulate its assembly and polarity?Collected ¢ndings with ¢broblasts suggest that mi-crotubules exert their in£uence by modulating sub-strate contact formation. The ¢rst indications of sucha link were provided by the ¢nding that an increasein stress ¢bre size and contractility accompanies mi-crotubule disassembly [16,46]. This result was dra-matically con¢rmed in starved ¢broblasts, whichlack stress ¢bres; microtubule disruption in this sys-tem caused the massive formation of focal contactsand stress ¢bres, an e¡ect that was shown to bemediated through the activation of Rho [18,7].

In a di¡erent line of investigations, a common co-localisation was noted in motile ¢broblasts betweenmicrotubule ends and vinculin-containing contactsites formed at the cell front [57]. This ¢nding wasvery recently pursued by studying living cells that

had been co-injected with £uorescent tubulin andvinculin. In these studies it could be clearly demon-strated that such co-localisations are by no meansfortuitous, but re£ect a direct and deliberate target-ing of new contact sites by microtubule ends [37](Fig. 5A). That this spatial overlap of microtubuleswith contact sites re£ects a direct interaction wasindicated by the further ¢nding that contacts in£u-ence the dynamics of microtubule ends that pass overthem and can capture microtubules and stabilisethem against depolymerisation by nocodazole [37].

The targeting of early contact sites by microtu-bules cannot be steered by microtubules alone.Rather, it likely represents a cross-talk between themicrotubule and actin cytoskeletons. Accordingly,we have postulated that single actin ¢laments ema-nating from early contact sites into the cytoplasmmay become tethered to a nearby microtubule via across-linking protein, which would bind the micro-tubule and actin ¢laments in parallel. Several candi-date proteins that bind to both microtubules andactin have already been described [22]. Growth ofthe microtubule would then occur in the contactsite from which the actin ¢lament derived. We mightexpect that the microtubule then delivers a molecularcargo to the contact that in£uences its furtherdevelopment, a cargo most likely carried by micro-tubule motor molecules. This contact targeting inter-action could then provide a means whereby micro-tubules exert their control on the actin cytoskeleton.By delivering components that regulate the stabilityand lifetime of contact sites, microtubules could de-termine the development of contact patterns andthereby the geometry of the advancing front of acell.

For neurites, the modulation of contact sites bytargeting could readily explain the mode of involve-ment of microtubules in growth cone steering [6,74].In this process, ¢lopodia and lamellipodia that leadthe way become dominant over those that eventuallyretract (Fig. 5B). Microtubules accumulate behindthese leading processes [58,73] and can penetrateinto the base of ¢lopodia, or alongside them [26].We contend that microtubule recruitment determineswhich ¢lopodia make longer term associations withthe substrate and therefore become dominant. Inagreement with this general idea, Varnum-Finneyand Reichardt [77] have shown that the down-regu-

BBAMCR 14355 8-9-98


lation of vinculin expression leads to a decrease inthe numbers of stable ¢lopodia and a reduced rate ofgrowth cone advance.

In terms of microtubule dependent polarity andguidance, the keratocyte is an enigma. This cellundergoes directional locomotion when its microtu-bules are completely depolymerised [19] or whentheir dynamics are blocked by taxol (Kaverina, un-published observations). In rapidly moving cells, themicrotubules do not penetrate into the lamellipodi-um (Kaverina, unpublished) and instead are tightlywrapped around the cell body ([19]; Fig. 5C). This ishardly surprising in view of the fact that the cellbody rolls continuously behind the lamellipodiumduring cell locomotion 3]. Only in regions where alamellipodium undergoes retraction is microtubule

penetration observed (Kaverina, unpublished). Thesame regions of retraction exhibit elongated adhesionsites populated with vinculin (Anderson and Rottner,unpublished), indicating that there may be a link,even in these cells, between microtubules and sub-strate contacts. Prominent adhesion sites do occurin keratocytes at the £anks of the cell body in asso-ciation with its asymmetry and these sites could beinvaded and modulated by microtubules.

6. Conclusions and perspectives

Eukaryotic cells have the capacity to assemble ac-tin ¢laments in di¡erent ways, to form cytoskeletalsubcompartments with speci¢c functions. Lamellipo-

Fig. 5. Contact targeting by microtubules in the control of axonal guidance and cell polarity. (A) Fibroblast : the ends of microtubulesin motile ¢broblasts are oriented more or less radially towards regions of lamellipodium protrusion. This directed orientation corre-lates with a targeting by microtubules of new substrate contact sites situated within and behind the lamellipodium (elipses). The specif-ic modulation of contact sites is necessary to sustain protrusion in one direction and is proposed to form the basis of polarity deter-mination. (B) Neuronal growth cone: steering of the growth cone is proposed to be determined by the modulation, via microtubuletargeting, of the stability of the substrate contacts formed beneath ¢lopodia. Those contacts that are targeted persistently (cross-hatched) exist long enough to support the extension of lamellipodia and ¢lopodia in the same direction. (C) Keratocyte: microtubulesin the moving keratocyte are mainly con¢ned to the rotating cell body. However, microtubule ends project from the £anks of the cellbody into the region where contacts associated with the transverse bundles of actin are found. These microtubules may serve a func-tion in regulating contact stability, but the keratocyte is also polar when microtubules are destroyed. See text for further details.LAM, lamellipodium; FIL, ¢lopodium; MT, microtubule; SF, stress ¢bre; CB, cell body; C, substrate contact. Open arrows indicatedirections of movement.

BBAMCR 14355 8-9-98


dia and ¢lopodia are required for motility, and stress¢bres for anchorage. We here have presented ideasabout how these subcompartments are generated.The lamellipodium is the primary site of actin poly-merisation and of the generation of ¢lament bundles,microspikes or ¢lopodia. Microspikes, in their turn,can provide the foci required for the initiation offocal adhesions, leading to stress ¢bre formation.Thus, there is an apparent hierarchy in the assemblyof subcompartments, from lamellipodia to ¢lopodiato stress ¢bres. In highly motile cells like keratocytes,only lamellipodia are expressed. Filopodia and mi-crospikes are found in less motile cells, such as ¢bro-blasts and the neuronal growth cone, and are in-volved in the development of substrate anchorage.In this anchorage process, microtubules serve a mod-ulatory function by in£uencing the stability of indi-vidual adhesions and, thereby, the polarity of thecell.

We have here neglected many aspects of actin cy-toskeleton dynamics that deserve attention. For ex-ample, what determines the switch from protrusionto ru¥ing and to what extent is one or the other ofthese activities important for invasion and metasta-sis? What sort of contact structures are used in dif-ferent situations and how is their formation regu-lated? And if microtubules play a role in contactgenesis, what components do they deliver or removefrom contact sites to modulate their stability?Clearly, signalling molecules of the rho-family playdecisive roles. But rather than acting separately,these molecules likely synergise in di¡erent combina-tions to e¡ect the subtle changes in the actin cyto-skeleton that make the di¡erence between static andmotile as well as between normal and transformedphenotypes. The challenge for the future will be tosort out the ¢ne tuning of these processes.

Acknowledgements

The authors acknowledge the generous support ofthe Austrian Science Research Council, the SeegenStiftung of the Austrian Academy of Sciences andthe Austrian National Bank for funding the facilitiesthat provided the experimental background of thiswork. We also thank Mrs. Elisabeth Eppacher fortyping.

References

[1] M. Abercrombie, J.E.M. Heaysman, S.M. Pegrum, Exp. CellRes. 60 (1970) 437^444.

[2] J.C. Adams, Mol. Biol. Cell 8 (1997) 2345^2363.[3] K.I. Anderson, Y.-L. Wang, J.V. Small, J. Cell Biol. 134

(1996) 1209^1218.[4] J.R. Bamburg, D. Bray, K. Chapman, Nature 321 (1986)

788^790.[5] D.A. Begg, R. Rodewald, L.I. Rebhuhn, J. Cell Biol. 79

(1978) 846^852.[6] D. Bentley, T.P. O'Connor, Curr. Opin. Neurobiol. 4 (1994)

43^48.[7] A. Bershadsky, A. Chausovsky, E. Becker, A. Lyubimova,

B. Geiger, Curr. Biol. 6 (1996) 1279^1289.[8] A.D. Bershadsky, E.A. Vaisberg, J.M. Vasiliev, Cell Motil.

Cytoskeleton 19 (1991) 152^158.[9] D. Bray, K. Chapman, J. Neurosci. 5 (1985) 3204^3213.

[10] P.C. Bridgman, M.E. Dailey, J. Cell Biol. 108 (1989) 95^109.[11] K. Burridge, Cancer Rev. 4 (1986) 18^78.[12] K. Burridge, M. Chrzanowska-Wodnicka, C. Zhong, Trends

Cell Biol. 7 (1997) 342^347.[13] M.-F. Carlier, Curr. Opin. Cell Biol. 10 (1998) 45^51.[14] J.F. Challacombe, D.M. Snow, P.C. Letourneau, J. Cell Sci.

109 (1996) 2031^2040.[15] L.P. Cramer, M. Siebert, T.J. Mitchison, J. Cell Biol. 136

(1997) 1287^1305.[16] B.A. Danowska, J. Cell Sci. 93 (1989) 255^266.[17] J.A. DePasquale, C.S. Izzard, J. Cell Biol. 105 (1987) 2803^

2809.[18] T. Enomoto, Cell Struct. Funct. 21 (1996) 317^326.[19] U. Euteneuer, M. Schliwa, Nature 310 (1984) 58^61.[20] G.W. Fisher, P.A. Conrad, R.L. De Biasio, D.L. Taylor,

Cell Motil. Cytoskeleton 11 (1988) 235^247.[21] P. Forscher, S.J. Smith, J. Cell Biol. 107 (1988) 1505^1516.[22] R.H. Gavin, Int. Rev. Cytol. 173 (1997) 207^242.[23] S.D. Glacy, J. Cell Biol. 97 (1993) 1207^1213.[24] D.J. Goldberg, D.W. Burmeister, J. Cell Biol. 103 (1986)

1921^1931.[25] D.J. Goldberg, D.W. Burmeister, Trends Neurosci. 12 (1989)

503^506.[26] P.R. Gordon-Weeks, NeuroReport 2 (1991) 573^576.[27] A. Hall, Science 279 (1998) 509^514.[28] T. Hasegawa, J. Cell Biol. 120 (1993) 1439^1448.[29] J.P. Heath, J. Cell Sci. 60 (1983) 331^354.[30] J.P. Heath, G.A. Dunn, J. Cell Sci. 29 (1978) 197^212.[31] J.P. Heath, B. Holi¢eld, Nature 352 (1991) 107^108.[32] J.P. Heath, B.F. Holi¢eld, in: G. Jones, C. Wigley, R. Warn

(Eds.), Cell Behaviour: Adhesion and Motility, The Com-pany of Biologists, Cambridge, 1993, pp. 35^56.

[33] S.R. Heidemann, P. Lamoureux, R.E. Buxbaum, J. CellBiol. 111 (1990) 1949^1957.

[34] G. Isenberg, J.V. Small, Eur. J. Cell Biol. 16 (1978) 326^344.[35] O.Y. Ivanova, L.B. Margolis, J.M. Vasiliev, I.M. Gelfand,

Exp. Cell Res. 101 (1976) 207^219.[36] C.S. Izzard, R. Lochner, J. Cell Sci. 42 (1980) 81^116.

BBAMCR 14355 8-9-98


[37] I. Kaverina, K. Rottner, J.V. Small, J. Cell Biol. 142 (1998)181^190.

[38] N. Kleitman, M.I. Johnson, Cell Motil. Cytoskeleton 13(1989) 288^300.

[39] R. Kozma, S. Sarner, S. Ahmed, L. Lim, Mol. Cell. Biol. 17(1997) 1201^1211.

[40] T. Kreis, B. Geiger, J. Schlessinger, Cell 29 (1982) 835^845.[41] J. Lee, M. Leonard, T. Oliver, A. Ishihara, K. Jacobson,

J. Cell Biol. 127 (1994) 1957^1964.[42] P.C. Letourneau, Dev. Biol. 85 (1981) 113^122.[43] P.C. Letourneau, T.A. Shattuck, Development 105 (1989)

505^519.[44] G. Liao, T. Nagasaki, G.G. Gundersen, J. Cell Sci. 108

(1995) 3473^3483.[45] C.-H. Lin, P. Forscher, Neuron 13 (1995) 763^771.[46] C.W. Lloyd, C.G. Smith, A. Woods, D.A. Rees, Exp. Cell

Res. 110 (1977) 427^437.[47] D.J.G. Mackay, C.D. Nobes, A. Hall, Trends Neurosci. 18

(1995) 496^502.[48] L.M. Machesky, A. Hall, J. Cell Biol. 138 (1997) 913^926.[49] M.M. Mareel, M. De Mets, Int. Rev. Cytol. 90 (1984) 125^

168.[50] P. Matsudaira, Semin. Cell Biol. 5 (1994) 165^174.[51] T. Mitchison, M. Kirschner, Neuron 1 (1988) 761^772.[52] C.D. Nobes, A. Hall, Curr. Opin. Genet. Dev. 4 (1994) 77^

81.[53] C.D. Nobes, A. Hall, Cell 81 (1995) 53^62.[54] S. Okabe, N. Hirokawa, J. Cell Biol. 109 (1989) 1581^1595.[55] S. Okabe, N. Hirokawa, Nature 343 (1990) 479^482.[56] A.J. Ridley, Curr. Biol. 6 (1996) 1256^1264.[57] G. Rinnerthaler, B. Geiger, J.V. Small, J. Cell Biol. 106

(1988) 747^760.[58] J.H. Sabry, T.P. O'Connor, L. Evans, A. Toroian-Raymond,

M. Kirschner, D. Bentley, J. Cell Biol. 115 (1991) 381^395.[59] J.M. Sanger, J.W. Sanger, J. Cell Biol. 86 (1980) 568^575.[60] J.V. Small, J. Cell Biol. 91 (1981) 695^705.[61] J.V. Small, Electron. Microsc. Rev. 1 (1988) 155^174.[62] J.V. Small, Semin. Cell Biol. 5 (1994) 157^163.[63] J.V. Small, G. Isenberg, J.V. Celis, Nature 272 (1978) 638^

639.

[64] J.V. Small, G. Rinnerthaler, H. Hinssen, Cold Spring Harb.Symp. Quant. Biol. 46 (1982) 599^611.

[65] J.V. Small, M. Herzog, K. Anderson, J. Cell Biol. 129 (1995)1275^1286.

[66] J.V. Small, K.I. Anderson, K. Rottner, Biosci. Rep. 16(1996) 351^368.

[67] T. Soranno, E. Bell, J. Cell Biol. 95 (1982) 127^136.[68] R.V. Stirling, S.A. Dunlop, Trends Neurosci. 18 (1995) 111^

115.[69] T.M. Svitkina, A.B. Verkhovsky, K.M. McQuade, G.G.

Borisy, J. Cell Biol. 139 (1997) 397^415.[70] M. Symons, Trends Biochem. Sci. 21 (1996) 178^181.[71] M.H. Symons, T.J. Mitchison, J. Cell Biol. 114 (1991) 503^

513.[72] Y. Takai, T. Sasaki, K. Tanaka, H. Nakanishi, Trends Bio-

chem. Sci. 20 (1995) 227^231.[73] E. Tanaka, T. Ho, M.W. Kirschner, J. Cell Biol. 128 (1995)

139^155.[74] E. Tanaka, M.W. Kirschner, J. Cell Biol. 128 (1995) 127^

137.[75] J.A. Theriot, T.J. Mitchison, Nature 352 (1991) 126^131.[76] F.N. Van Leeuwen, H.E.T. Kain, R.A. van der Kammen, F.

Michiels, O.W. Kranenburg, J.G. Collard, J. Cell Biol. 139(1997) 797^807.

[77] B. Varnum-Finney, L.F. Reichardt, J. Cell Biol. 127 (1994)1071^1084.

[78] J.M. Vasiliev, I.M. Gelfand, Neoplastic and Normal Cells inCulture, Cambridge University Press, Cambridge, 1981.

[79] J.M. Vasiliev, I.M. Gelfand, in: R. Goldman, T. Pollard,J. Rosenbaum (Eds.), Cell Motility, Cold Spring HarborLaboratory Press, Cold Spring Harbor, NY, 1976, pp.279^304.

[80] A.B. Verkhovsky, T.M. Svitkina, G.G. Borisy, J. Cell Biol.131 (1995) 989^1002.

[81] Y.-L. Wang, J. Cell Biol. 99 (1984) 1478^1485.[82] Y.-L. Wang, J. Cell Biol. 101 (1985) 597^602.[83] K.M. Yamada, B.S. Spooner, N.K. Wessells, Proc. Natl.

Acad. Sci. USA 66 (1970) 1206^1212.

BBAMCR 14355 8-9-98


Documents

Assembling an actin cytoskeleton for cell attachment and movement