Organization of Point Contacts in NeuronalGrowth ConesA. Renaudin,1 M. Lehmann,1 J.-A. Girault, 2 and L. McKerracher 1*1Departement de Pathologie et Biologie Cellulaire, Universite´ de Montreal, Montreal, Quebec, Canada2Chaire de Neuropharmacologie, INSERM U114, Colle`ge de France, Paris, France

Growth cones from rat dorsal root ganglia plated onlaminin contain integrin clusters over the entire growthcone surface, and growth cones make transient adhe-sions at sites called point contacts. We examined, byimmunocytochemistry and confocal microscopy, thecomposition and distribution of point contacts inneuronal growth cones. Vinculin was concentrated inthe central domain of growth cones and at the tips offilopodia. Vinculin was specifically associated withintegrin clusters at the membrane–substrate interfaceand thus marked point contacts. The cytoskeletalproteins paxillin and talin colocalized with b1 integrinin a subpopulation of clusters restricted to the centraldomain of the growth cone and to the tips of filopodia.The neuron-specific kinase, FAK1 also distributedwith the vinculin-positive clusters. The Rho familyproteins RhoA, RhoB, and Cdc42 were present ingrowth cones, and a few Rho clusters were colocalizedwith vinculin. Examination of proteins resistant todetergent extraction in PC12 cells confirmed theretention of b1 integrin, paxillin, talin, and vinculinwith the cytoskeleton. Moreover, we detected FAK1and RhoA in the detergent-resistant cytoskeleton,supporting their distribution to point contacts. Ourobservations indicate that two types of integrin clus-ters are present in growth cones: those associated withvinculin at the cell substratum interface, and those notassociated with vinculin. Point contacts are matureadhesion sites defined by the presence of bothb1integrin and vinculin, and they are associated withsignaling proteins. J. Neurosci. Res. 55:458–471, 1999.r 1999 Wiley-Liss, Inc.

Key words: b1 integrin; p125FAK; vinculin; cytoskel-eton; Rho; dorsal root ganglion

INTRODUCTIONDuring axonal elongation in development or regen-

eration, the growth cone forms transient adhesions withthe substrate, and the cytoskeletal machinery translatesthese adhesions into a pulling force to move the growthcone forward. Structures called point contacts are theadhesive sites found on neuronal growth cones during

integrin-mediated growth on laminin (Arregui et al.,1994; Gomez et al., 1996). Integrin clusters cover boththe upper and lower surfaces of the growth cone. Only thesmall integrin clusters at the membrane–substrate inter-face are strongly associated with the cytoskeleton andform functional contact sites (Arregui et al., 1994).Recent studies have indicated that nonfocal contact typesof adhesive contacts exist (Hall, 1998) but the functionaland structural relationships between these different con-tacts are not well understood. The role, if any, of the otherintegrin clusters remains unknown. Point contacts aresmall integrin-containing adhesion sites that differ in sizeand distribution from the focal contact type of adhesions(Tawil et al., 1993). Focal contacts represent sites ofstrong anchorage to the extracellular matrix and mayreduce the rate of cell migration (Luna and Hitt, 1992).Point contacts are abundant in highly motile cells; aftercell transformation withRous sarcomavirus, the numberof focal contacts decreases and numerous point contactsappear over the cell surface, where they become func-tional sites of anchorage (Bershadsky et al., 1985;Nermut et al., 1991). These structures, alternativelynamed dot contacts or podosomes, were first assumed toconstitute an immature form of focal contact (Burridge etal., 1988). More recent studies with antibodies specific toa phosphorylated form of the cytoplasmic domain ofb1integrin have shown that only point contacts, not focalcontacts, have the phosphorylatedb1 epitope (Johanssonet al., 1994). Therefore, it is likely that there arefunctional and structural differences between focal con-tacts and point contacts.

Little is known about how integrins in point con-tacts engage the cytoskeletal response in growth cones.Growth cones exhibit a retrograde flow of F-actin, and

Contract grant sponsor: MRC (Canada); Contract grant sponsor:FCAR; Contract grant sponsor: MRC.

*Correspondence to: Dr. L. McKerracher, De´partement de Pathologieet Biologie Cellulaire, Universite´ de Montreal, CP 6128, SuccursaleCentre-ville, Montre´al, Quebec, H3C 3J7 Canada. E-mail:mckerral@ere.umontreal.ca

Received 14 August 1998; Revised 30 October 1998; Accepted 30October 1998

Journal of Neuroscience Research 55:458–471 (1999)

r 1999 Wiley-Liss, Inc.

cytoskeletal proteins may act as a ‘‘clutch’’ linking theextracellular matrix to the intracellular actin network (Linand Forscher, 1995). Integrin-associated cytoskeletal pro-teins may be the clutches mediating the response ofgrowth cones to laminin, but the cytoskeletal proteins thatunderlie point contacts are not well characterized. In focalcontacts, cytoskeletal proteins that cluster at adhesionsites include talin, paxillin, and vinculin, all of which aredetected in growth cones, but their relationship to pointcontact structure is not clear (Letourneau and Shattuck,1989; Arregui et al., 1994; Sydor et al., 1996; Gomez andLetourneau, 1994).

Tyrosine kinases might also be expected to associ-ate with point contacts, as is the case for focal contacts.Moreover, phosphotyrosine immunoreactivity is associ-ated with b1 integrin aggregates at the tips of growthcone filopodia (Wu et al., 1996). In previous studies, thefocal adhesion kinase p125FAK was not detected in pointcontacts of PC12 cells (Arregui et al., 1994). However, aneuronal variant of FAK, called FAK1, which has aninsertion in the region responsible for focal adhesiontargeting, has since been reported (Burgaya and Girault,1996). FAK1 is highly expressed in brain, and immuno-reactivity to FAK1 has been detected in primary neuro-nal cultures (Burgaya et al., 1995; Stevens et al., 1996);thus, FAK1 may colocalize at site of point contact. Inneurons, another prominent signaling protein is thetyrosine kinase pp59fyn, which is expressed at high levelsduring development (Bixby and Jhabvala, 1993) and isassociated with the cytoskeletal fraction of growth coneparticles (Helmke and Pfenninger, 1995). Therefore, fyncould be associated with point contacts.

There is increasing evidence that Rho family GTP-ases are important in axonal and dendritic growth.Mutations in Rho-related family members block theextension of axons inDrosophila(Luo et al., 1994) anddisrupt axonal pathfinding inCaenorhabditis(Zipkin etal., 1997). The introduction of mutated Rac, Rho, orCdc42 into cortical neurons affects dendritic morphology(Threadgill et al., 1997), and in transgenic mice thatexpress constitutively active Rac in Purkinje cells, thereare alterations in the development of axon terminals anddendritic arborizations (Luo et al., 1996). Recently,Mackay et al. (1996) hypothesized that Rho may regulategrowth cone spreading or collapse morphology and thatCdc42 may regulate filopodial formation. However, thelocalization of Rho and Cdc42 within growth cones hasnot been examined.

Anatomically, growth cones possess a central and aperipheral domain. The central domain, or base of thegrowth cone, contains diverging microtubules and vari-ous organelles and is the site of membrane addition forneurite elongation (Craig et al., 1995). The peripheraldomain is devoid of organelles and is made up of a

dynamic actin network of lamellipodia and filopodia.Growth cone motility depends on coordinated adhesion,detachment, and readhesion, and one might expect thatpoint contacts regulate this process. To characterize thestructural distribution of point contacts, we studied ratdorsal root ganglion neurons, which are known to expressa1b1, a3b1, and a6b1 integrins as laminin receptors(see McKerracher et al., 1996). We examined the colocal-ization of cytoskeletal and signaling proteins withb1integrin in growth cones of these sensory neurons byimmunocytochemistry and confocal microscopy. We re-port that there are two populations of integrin clusters thatcorrelate with the presence or absence of vinculin and thatpoint contacts present on the lower cell surface aredefined by the presence of vinculin. Further, we confirmin PC12 cells the retention of candidate point contactproteins with the detergent-resistant cytoskeleton.

MATERIALS AND METHODSDissection and Cell Culture

Glass coverslips were coated with laminin byincubation in a 20-µg/ml solution for 2 hr at 37°C. Dorsalroot ganglia (DRG) were dissected from early postnatal(days 0–8) rats, and small explants were placed on thecoverslips and cultured overnight in Dulbecco’s ModifiedEagle’s Medium, 10% fetal bovine serum, 1% vitamins(Gibco BRL, Gaithersburg, MD), 1% penicillin/strepto-mycin, and 100 ng/ml nerve growth factor. The removalof all connective tissue surrounding the DRG reduced thenumber of nonneuronal cells and ensured that neuritesextended on the laminin substrate with distinct, well-spread growth cones.

ImmunocytochemistryCoverslips were washed in phosphate buffered

saline (PBS) and fixed for 30 min in 4% paraformalde-hyde and 0.1 M PO4 buffer, both at room temperature. Insome experiments, growth cones fixed in ice-cold metha-nol were also examined. After fixation, the coverslipswere washed in PBS and blocked for 30 min in PBS, 3%bovine serum albumin, and 1% Triton X-100. Primaryantibodies were incubated in blocking solution overnight,and the next day biotinylation was used to enhance thesignal produced by the primary antibodies. In single-labeling studies, the primary antibody was washed away,and the coverslips were incubated for 1 hr at roomtemperature with the appropriate biotin-SP–conjugateddonkey anti-mouse or anti-rabbit F(ab) fragments (Jack-son ImmunoResearch Laboratories, Bio/Can Scientific,Mississauga, Ontario, Canada), diluted 1:500 in blockingsolution. The coverslips were then washed again andincubated for 1 hr with dichlorotriazinylamino fluores-cein (DTAF)-conjugated streptavidin (Jackson Immuno-

Point Contacts in Neuronal Growth Cones 459

Research Laboratories), diluted 1:500 in blocking solu-tion. For double-labeling studies, the primary mousemonoclonal and rabbit polyclonal antibodies were addedtogether, and after washing, one of the antibodies wasbiotinylated and incubated with DTAF-conjugated strep-tavidin, and the other was treated with the appropriategoat anti-mouse or anti-rabbit IgG RITC (Calbiochem,San Diego, CA). Coverslips were mounted on glass slideswith glycerol antifade mounting medium (MolecularProbes, Eugene, OR), and epifluorescence was visualizedwith a Zeiss Plan-Apochromat 633/1.40 objective or aPL Fluotar 1003/1.32 objective. Micrographs were takenwith Kodak TMAX 400 film at 800 ASA.

Confocal microscopy was performed with a dual-channel BioRad 600 laser scanning confocal microscope.Optical sections of 0.3 µm were taken between the topand the bottom surfaces of the growth cones, in a planeparallel to the substrate. Merging of the green and redchannels generated double-labeled images, with yellowareas corresponding to colocalization between the twomolecules under study.

b1 Polyclonal AntibodyRabbits were injected with a synthetic peptide

corresponding to the cytoplasmic domain of theb1integrin subunit. The sequence CREFAKFEKEKMNAK-WDTGENPIYKSAVTT was coupled with KLH peptide,and 1 mg was injected into rabbits in Freund’s completeadjuvant. Subsequent injections were done with Freund’sincomplete adjuvant at 4 weeks and at 7 weeks. Immuneserum was collected, clarified, diluted with one volume ofPBS, and applied to a 2-ml affinity-purification columncontaining the synthetic peptide. After washing, thebound antibody was eluted with 100 mM glycine (pH2.8). Fractions of 1 ml were collected and neutralizedwith 50 µl of 1 M Tris at pH 9.5. Another antibody tob1integrin used in some experiments was a gift from Dr. S.Carbonetto (McGill University, Montreal, Canada). TheFAK1 polyclonal antibody (Burgaya et al., 1995) wasused at a 1:50 dilution. The commercial primary antibod-ies and dilution used were as follows: vinculin clonehVIN-1, 1:50 (Sigma, Oakville, Ontario, Canada), talinclone 8d4 1:50 (Sigma), monoclonalbIII tubulin (Sigma),paxillin clone 349 1:50 (Transduction Labs, BioConScientific, Mississauga, Ontario, Canada), p59fyn 1:500(Upstate Biotechnology, Lake Placid, NY), monoclonalRhoA 1:500, polyclonal Rho B 1:500, and polyclonal Cdc421:500 (all from Santa Cruz Biotechnology, BioCon Scientific,Mississauga, Ontario, Canada). Specificity of the antibodieswas confirmed by Western blot analysis (see Fig. 8).

Western BlotsTo characterize the polyclonal anti-b1 integrin

antibody, rat brain homogenates and Chinese hamster

ovary (CHO) cell membranes were dissolved in samplebuffer, and the proteins were separated on 7% acrylamidegels, transferred to nitrocellulose, and probed with anti-body. Some samples of CHO membranes were deglyco-sylated with PNGase F according to the manufacturer’sinstructions (New England Nuclear, Beverly, MA). Immu-noreactivity was detected with horseradish peroxidasesecond antibody and enhanced chemiluminesence (Amer-sham Life Sciences, Oakville, Ontario, Canada) or by alkalinephosphase–conjugated second antibody and detection withBCIP and NBT (Gibco BRL, Burlington, Ontario, Canada).

PC12 cells were grown on laminin, and cytoskeletaland soluble fractions of differentiated PC12 cells wereprepared by extracting cells with a cytoskeleton-stabilizing buffer (0.01 M PIPES, pH 6.8, 0.05 M KCl,0.01 M EGTA, 0.003 M MgCl2, 2 M glycerol, 0.001 Mphenyl methyl sulfonyl fluoride, 50 µg/ml leupeptin, 50µg/ml aprotinin, and 1% Triton X-100), as described byArregui et al. (1994). Equivalent volumes of the solubleand cytoskeleton fractions were separated on acrylamidegels and detected on Western blots, as described above.

RESULTSCharacterization of b1 Integrin and VinculinImmunoreactivity in Point Contacts

DRG explants were placed in culture on lamininsubstrates, and after fixation most neuronal growth conesretained a large, well-spread lamella, with several filopo-dial extensions. These spread growth cones contained acentral domain rich in microtubules, some of whichextended into filopodia (Fig. 1). The preservation offixation-labile microtubules indicates that our fixationconditions were appropriate to preserve subcellular structure.

To investigateb1 integrin immunolocalization ingrowth cones, we used an affinity-purified polyclonalantibody raised against peptide to the cytoplasmic do-main ofb1. On Western blots this antibody reacted withtwo bands in CHO cell membrane preparations; afterdeglycosylation of the membranes, the immunoreactionwas to a single band (Fig. 2), showing that the antibodyrecognized both glycosylated and deglycosylatedb1integrin. In samples of whole brain homogenate, theantibody recognized mainly a single band intermediate insize between the fully glycosylated and deglycosylatedforms ofb1 integrin (Fig. 2).

Immunocytochemical observations of growth coneslabeled with our b1 polyclonal antibody showed apunctate immunoreactivity in both neurites and growthcones (Fig. 3a). A different anti-b1 peptide antibody usedin some experiments showed the same pattern of immuno-reactivity (Fig. 3b). The punctateb1 integrin sitescovered the entire growth cone, and immunoreactivitywas also observed to extend along the length of most

460 Renaudin et al.

filopodia (Fig. 3a,b). Therefore, in primary cultures ofDRG neurons grown on laminin substrates, growth coneshave a uniform distribution of small integrin clusters.Whereas integrin clusters are located on both the upperand lower plasma membranes, only those associated withthe lower cell surface are linked to the detergent-resistantcytoskeleton (Arregui et al., 1994).

Glial cells have been reported to have both focalcontacts and point contacts as their integrin-containingadhesion sites (Tawil et al., 1993). Schwann cells repre-sented most of the nonneuronal cells in our cultures, andwe observed with anti-vinculin antibody that the nonneu-ronal cells had both focal contacts (Fig. 3e) and pointcontacts (Fig. 3f, solid arrows). The focal contacts (Fig.3f, open arrow) are recognized as elongated adhesionsites 1–5 µm in length, and point contacts are,10 timessmaller (Bershadsky et al., 1985) and less conspicuous(Fig. 3f). Therefore, although both types of adhesive struc-tures were detected in the primary nonneuronal cells in ourcultures, clear examples of focal contact were not observed inneuronal growth cones growing on laminin (Fig. 3c,d).

Anti-vinculin immunoreactivity was consistentlypunctate in neurons. In contrast to the observations onb1immunoreactivity, vinculin had a more restricted distribu-tion in growth cones. Whereas integrin clusters coveredthe whole growth cone, the vinculin-positive clusterswere often sparse (Fig. 3c) and generally restricted to thecentral domain and to the tips of filopodia (Fig. 3d). Thedifferential immunoreactive patterns of vinculin andb1integrin suggested that there may be two different popula-

tions of integrin clusters: those associated with vinculin,and those that were vinculin negative.

To investigate further the integrin and vinculinclusters, confocal microscopy was used to analyze growthcones double labeled with the anti-b1 polyclonal and themonoclonal anti-vinculin. Confocal microscopy showedcolocalization of b1 integrin and vinculin in only asubpopulation ofb1 integrin punctate sites (Fig. 4a). Theassociations betweenb1 integrin and vinculin were mostprominent in the central domain and in certain filopodia,areas of the growth cone that adhere strongly to thesubstrate (Zheng et al., 1994). These results suggestedthat vinculin may associate with integrins at the level ofthe lower membrane of the growth cone. To determinewhether integrin-vinculin interactions were restricted tothe growth cone surface that interacts with laminin, weperformed a z-series of sections taken parallel to thesubstrate. Figure 4a represents the summation of all theimages taken throughout a growth cone, in whichb1integrin and vinculin were partly colocalized (yellowareas). Figure 4b shows a single section at the level of theupper membrane, with almost no yellow areas, whichindicated thatb1 integrin and vinculin were not colocal-ized at planes above the substrate. Separate clusters of

Fig. 1. Growth cones retain a well-preserved microtubulenetwork after fixation. Anti-bIII tubulin antibody immunoreac-tion of a growth cone fixed with paraformaldehyde. Individualmicrotubules are seen to project into the peripheral domain ofthe growth cone, including filopodia. Bar, 10 µm.

Fig. 2. Specificity of the anti-b1 integrin polyclonal antibodyon Western blots. Affinity-purified antibody recognized twobands in Chinese hamster ovary (CHO) cell membranes (lane 1)and a single band after deglycosylation of the CHO cellmembranes (lane 2). In whole brain homogenates, the antibodyrecognized a major band, with an approximate apparent molecu-lar mass of 120 kDa (lane 3).

Point Contacts in Neuronal Growth Cones 461

Fig. 3. Comparison ofb1 integrin and vinculin immunoreac-tion in growth cones and primary nonneuronal cells.a,b: Thepattern of immunoreactivity was the same for two differentanti-b1 antibodies. Point contact clusters entirely coveredneurites and growth cones and were also found at the tips offilopodia (arrows).c,d: Vinculin immunoreactivity was more

restricted to the central domain than wasb1 integrin. Vinculinclusters were also observed at the tips of some filopodia(d, arrows).e,f: Primary nonneuronal cells possessed prominentfocal contacts immunoreactive to vinculin. Both focal contacts(open arrow) and point contacts (solid arrows) were observed inprimary nonneuronal cells (f). Bars, 10 µm.

462 Renaudin et al.

Fig. 4. b1 Integrin and vinculin are colocalized at the level ofthe lower membrane of growth cones.a: Confocal image ofsummed 0.3-µm optical sections of a growth cone labeled withb1 integrin (red) and vinculin (green), with yellow areas beingregions of colocalization.b: A single 0.3-µm section taken atthe surface of the growth cone above the substrate shows that

b1 integrin and vinculin were not colocalized at the upper cellsurface.c: A single 0.3-µm section at the level of the substrateshows that the association ofb1 integrin and vinculin waswhere the lower membrane of the growth cone was in contactwith laminin.

vinculin and integrin exist on the upper growth conesurface. Figure 4c was taken at the level of the lowermembrane in contact with the substrate and showscolocalization ofb1 integrin and vinculin. These immuno-reactive patterns ofb1 and vinculin were conserved whenthe red and green fluorochromes were switched. There-fore, DRG growth cones plated on laminin appeared topossess two populations of integrin clusters: (1) widelyspread integrin clusters devoid of vinculin, and (2)vinculin-positive clusters restricted mainly to the centraldomain and certain filopodia. Therefore, colocalization ofb1 integrin and vinculin immunoreactivity in growthcones likely marks adhesive point contacts, and vinculinandb1 integrin are also present as separate clusters on theupper surface of the growth cone.

Cytoskeletal Proteins Associated With Point ContactsTo characterize further the difference between the

vinculin-positive and vinculin-negative integrin clusters,

the distribution of other proteins known to form part ofthe submembrane cytoskeleton was studied in DRGgrowth cones. Talin is a 270-kDa cytoskeletal protein thatinteracts with integrins (Luna and Hitt, 1992; Samuelssonet al., 1993). In the primary DRG cultures, talin wasdistributed in punctate clusters that were often moreprominent in the growth cone central domain, and someimmunoreactivity extended into filopodia (Fig. 5A).These results are consistent with those of a recent studyperformed in chick DRG neurons showing punctate talinimmunolocalization in growth cones (Sydor et al., 1996).Summed confocal images of growth cones double labeledwith anti-b1 integrin and anti-talin showed that colocal-ization was generally present in the central domain and onfilopodia (Fig. 6A, arrows). Therefore, only a subpopula-tion of talin clusters are likely to be associated withadhesive contacts.

Paxillin is another protein concentrated at sites offocal contact, and it is also present in punctate clusters in

Fig. 5. Other adhesion-related proteins show punctate immunoreactivity in growth cones.A: Talin immunoreactivity was widely distributed in growth cones and prominent in the centraldomain.B: Paxillin accumulated in punctate clusters in growth cones.C: The Fyn kinase alsodistributed to punctate clusters that covered the surface of the growth cones, up to the tips offilopodia (arrows).D: Anti-FAK1 immunoreactivity was punctate, prominent in the centraldomain and in filopodia, and more sparse in the peripheral lamella. Bar, 10 µm.

464 Renaudin et al.

neuronal growth cones (Gomez et al., 1996). In DRGneurons, paxillin was present in distinct clusters visible inboth neurites and growth cones, and these clusterstypically covered the whole surface of the growth cone(Fig. 5B). Confocal microscopy on double-labeled ex-plants showed that paxillin was colocalized withb1integrin point contacts in the central domain of neuronalgrowth cones (Fig. 6B, yellow areas). Therefore, some ofthe paxillin clusters are likely to associate with thepopulation of vinculin-positive point contacts. Indeed,paxillin possesses a vinculin binding site (Turner et al.,1990).

Tyrosine Kinases Associated With Point ContactsTo determine whether point contacts colocalize

with signaling molecules, we examined the localizationof nonreceptor tyrosine kinases in DRG growth cones.

Fyn is a nonreceptor tyrosine kinase of the c-src family,which is highly expressed early in developing neurons(Bixby and Jhabvala, 1993; Helmke and Pfenninger,1995). In growth cones, we observed that fyn waslocalized to punctate sites distributed over the wholegrowth cone, and immunoreactivity frequently extendedto the tips of filopodia (Fig. 5C).

The tyrosine kinase p125FAK is a signaling proteinthat becomes phosphorylated with integrin clustering andligand binding in focal contacts (Schaller et al., 1992).Recently, multiple transcripts of the protein have beenidentified in rat brain, including a form called FAK1,which is particularly enriched in neurons (Wigley andBerry, 1988; Burgaya and Girault, 1996). Immunoreactiv-ity with FAK1 antibodies produced an intense, punctatesignal in the central domain of the growth cone andfilopodia, with lighter punctate staining in the lamellipo-

Fig. 6. Confocal images of double-labeled growth cones.A: Immunoreactivity to talin (red) andb1 integrin (green)showed sites of colocalization (yellow) in the central domainand at the tips of filopodia (arrows).B: Double labeling withanti-paxillin (red) and anti-b1 integrin (green) showed somesites of colocalization in the central domain.C: Immunolabel-

ing of FAK1 (red) and vinculin (green) showed sites ofcolocalization in the central domain, but FAK1 immunoreactiv-ity was much more widely distributed than vinculin.D:Labeling of RhoB (green) and vinculin (red) showed someprominent colocalization at the growth cone margin (arrows).

Point Contacts in Neuronal Growth Cones 465

dia (Fig. 5D). In confocal images of DRG growth conesdouble labeled with FAK1 and vinculin, FAK1 wasassociated with some vinculin-positive sites in the centraldomain of DRG growth cones, whereas peripheral lamel-lipodial regions labeled mainly with FAK1 alone(Fig. 6C). Therefore, FAK1 likely associates with thepopulation of integrin point contacts that contain vincu-lin.

Neuronal Distribution of Rho GTPasesThe small GTPases of the Rho family, including

RhoA, RhoB, and Cdc42, are molecular switches whosefunctions include regulating the actin cytoskeleton inadhesion and motility (Mackay et al., 1996). In growthcones the distribution of RhoA and RhoB differed fromthat of Cdc42. RhoA was distributed in punctate clustersover the entire surface of neurites and growth cones, up to

the tips of filopodia (Fig. 7C, arrows). A similar patternwas observed with RhoB (Fig. 7A), which was clearlypunctate in DRG neurites and growth cones. In contrast toRho, Cdc42 immunofluorescence was diffuse (Fig. 7D),and filopodia were brightly filled. Therefore, the localiza-tion of Cdc42 was consistent with a possible rolein regulating filopodia (Mackay et al., 1996), andCdc42 was not obviously associated with adhesion sites.Therefore, the Rho GTPases were present in growthcones.

To determine whether the small GTPases wereassociated with vinculin-positive point contacts, double-labeling studies were performed on the DRG explantswith RhoB polyclonal and vinculin monoclonal antibod-ies. RhoB showed a much more intense labeling than didvinculin (Fig. 7A,B), and the RhoB sites were much morenumerous than were the spots of vinculin immunoreactiv-

Fig. 7. Distribution of Rho GTPases in growth cones. Doublelabeling of RhoB (A) and vinculin (B) showed that RhoBimmunoreactivity was more widely distributed than that ofvinculin and that RhoB was enriched in the central domain ofgrowth cones, with immunoreactivity also at the tips of

filopodia (arrows).C: Two different growth cones immuno-stained with RhoA showed that RhoA was present as granularimmunoreaction over the entire growth cone.D: In contrast,Cdc42 was consistently diffuse in growth cones, with strongimmunoreactivity in filopodia. Bar, 10 µm.

466 Renaudin et al.

ity. By confocal microscopy, some colocalization ofimmunoreactivity was observed in the central domain,although much of this appeared to be in the thickerregions of the growth cone (Fig. 6D), where it might beattributed to cytoplasmic bulk. A striking colocalizationof Rho with some vinculin spots was observed at thegrowth cone margin or filopodial tips (Fig. 6D, arrows),but the sites of colocalization were rare. The prominenceof the Rho proteins in growth cones and the strikinglocalization with some vinculin spots remain consistentwith a possible role in the dynamic regulation of transientadhesions.

Immunoreactivity to Cdc42, RhoA, or RhoB in theprimary nonneuronal cells was light and had a grainyappearance (data not shown). Immunoreactivity of theseGTPases was not observed to colocalize with the focalcontacts in these cells. Therefore, although the Rhoproteins may regulate focal contact formation (Ridley andHall, 1992), they are not obviously associated with focalcontacts.

Fractionation of Point Contact Proteins With theDetergent-Resistant Cytoskeleton

To investigate further the proteins associated withpoint contacts, we analyzed biochemically the retentionof candidate proteins with the detergent-resistant cytoskel-eton. For these studies we turned to PC12 cells becausewe previously developed extraction conditions that retainintegrins as well as tubulin and actin (Arregui et al.,1994). Moreover, like DRG growth cones, PC12 cellshave point contacts as their onlyb1 integrin–containingadhesion contact when plated on laminin (Arregui et al.,1994).

To confirm that the extraction procedure efficientlyretained point contacts, we first examined the distributionof b1 integrin after detergent extraction. When weprepared cytoskeletal ghosts, about half of theb1 integrinwas retained in the cytoskeletal fraction (Fig. 8). Simi-larly, about half of the tubulin and actin proteins wereinsoluble, in agreement with previous results (Arregui et

Fig. 8. Analysis of proteins associated with the detergent-resistant cytoskeleton in PC12 cells.Cytoskeletal ghosts (C) and soluble proteins (S) were prepared from differentiated PC12 cells,separated by electrophoresis, and transfered to nitrocellulose membrane for immunoblotting.Arrows show the band for the indicated protein.

Point Contacts in Neuronal Growth Cones 467

al., 1994). Next, we examined the distribution of vinculin,talin, and paxillin after detergent extraction. We foundthat all of these point contact proteins were present in thecytoskeletal fraction, although more than half of theprotein was found in the soluble fraction. Therefore,although these proteins are associated with the cyto-skeleton, a significant amount is present in the solublepool.

We then examined if Fyn and FAK were retainedwith the cytoskeletal ghosts under our extraction condi-tions. Fyn forms a detergent insoluble complex afterpalmitoylation, either because of its interaction withglycosyl phosphatydyl inositol–anchored proteins or be-cause of its association with cytoskeletal proteins (Wol-ven et al., 1997). After extraction, some of the Fyn wasobserved to partition with the cytoskeleton. These resultsshow that in PC12 cells Fyn exists in an insoluble form,an observation consistent with the immunocytochemicallocalization of Fyn in clusters on DRG growth cones.Similarly, FAK was observed to partition in part withcytoskeletal ghosts (Fig. 8).

In additional experiments we examined RhoA andCdc42 to determine whether we could detect associationof these proteins with the detergent-insoluble fraction.We consistently observed the retention of a small amountof RhoA in the cytoskeletal ghosts. By contrast, Cdc42was never retained in the detergent-insoluble fractions, afinding consistent with the diffuse immunocytochemicallocalization of this protein. In addition, we examined thepartitioning of RhoA and Cdc42 in 3T3 fibroblasts, a celltype that predominantly has focal contacts, not pointcontacts. In the 3T3 fibroblasts RhoA and Cdc42 werenever observed to partition with the Triton X-100–insoluble fraction (not shown). Together, these datasuggest that RhoA found in the cytoskeletal fraction maybe associated with a small subpopulation of point con-tacts.

DISCUSSIONThe aim of the current work was to characterize the

distribution of cytoplasmic proteins in growth cones ofmammalian DRGs plated on laminin. The use of immuno-cytochemistry and confocal microscopy enabled us tolocate proteins to discrete punctate sites in the growthcone and to determine the extent of colocalization withb1 integrin clusters. Extraction of cells with Triton X-100allowed us to confirm the association of specific proteinswith the detergent-resistant cytoskeleton. Several conclu-sions can be drawn from our results. (1) Punctate integrinclusters are the onlyb1 integrin structure on DRG growthcones plated on laminin. Moreover, there are at least twodistinct populations of integrin clusters, depending onwhether the receptor clusters are associated with vinculin.

(2) The vinculin clusters define the point contacts at thelevel of cell–laminin interactions. (3) Point contacts havemany cytoskeletal and signaling proteins in common withfocal contacts. However, some point contacts may beassociated with Rho proteins, and Rho was detected in thedetergent-resistant cytoskeleton. (4) Point contacts arepresent together with focal contacts in migrating primarynonneuronal cells, which in our cultures are presumablyfibroblasts and Schwann cells. Point contacts or podo-somes have previously been identified in migratingastrocytes (Tawil et al., 1993), PC12 cells (Arregui et al.,1994), and a variety of tumor cells (Nermut et al., 1991;Johansson et al., 1994). Neuronal growth cones onlaminin are also highly motile structures, and pointcontacts constitute their only adhesion structure. There-fore, we suggest that point contacts represent an adhesionstructure specialized for transient attachment duringmotility, which can be dynamically and rapidly regulatedin response to environmental influences.

Integrin Clusters Associated With Vinculin FormPoint Contacts

In single-labeling studies, the distribution ofb1integrin, paxillin, and talin was similar, with punctateimmunoreactivity over the entire growth cone, includingthe filopodia. In contrast, vinculin expression was gener-ally restricted to the central domain of the growth coneand to the tips of filopodia, in keeping with results of aprevious report (Sydor et al., 1996). Moreover, proteincolocalization by confocal microscopy showed that vincu-lin was restricted to the cell–substratum interface (Fig. 4),where integrins are firmly linked to the cytoskeleton inthe form of point contact (Arregui et al., 1994). Clustersof integrins were more widely distributed. Similarly,paxillin and talin, both components of the membranecytoskeleton, may cluster in growth cones. Our observa-tions suggest that paxillin and talin clusters are associatedonly with b1 integrin in the central domain, where theb1integrin is associated mainly with vinculin. Therefore, thepresence of vinculin withb1 integrin at the cell–substratum interface defines a point contact, and thelocalization of vinculin in the central domain and the tipsof filopodia is at the area of strongest adhesion of growthcones to the substrate (Zheng et al., 1994). Other clustersof b1 integrin, paxillin, and talin exist that are notassociated with vinculin. Although these clusters are notlikely to be functional in cell adhesion, they may haveother roles in growth cone motility. Therefore, vinculinappears to be a key molecule in stabilizing transmem-brane linkages between integrins and the cytoskeletonand may act as a ‘‘clutch’’ linking the extracellular matrixto the intracellular actin network (Lin and Forscher, 1995).

468 Renaudin et al.

Other evidence implicates vinculin in stabilizingtransmembrane linkages between the extracellular matrixand the cytoskeleton in neurons. A deficiency in theamount of vinculin in PC12 cells induces filopodial andlamellar instability, which hinders motility (Varnum-Finney and Reichardt, 1997). Also, inactivation of vincu-lin in the filopodia of DRG growth cones inducesfilopodial buckling and bending (Sydor et al., 1996).Taken together, current evidence suggests that vinculin isrecruited to integrin clusters in regions where the growthcone pulls against the substrate, perhaps as a key step inthe consolidation phase of growth cone advance.

Our results provide some insight into the molecularmechanisms that govern growth cone adhesion andmotility. Growth cones exhibit a retrograde F-actin flow,generated by actin polymerization and myosin motors,with a flow rate inversely proportional to growth coneadvance (Lin and Forscher, 1995; Lin et al., 1996).Vinculin-linking laminin-bound integrins to the actinnetwork may provide an anchor against which the forcegenerated by the retrograde F-actin flow would betransformed into filopodial and lamellar advance.

Integrin clusters containing theb1 subunit undergodirected transport toward the leading edge through tran-sient interactions with the cytoskeleton (Schmidt et al.,1995). The localization of talin and paxillin in growthcones suggests that they may participate in this process.Talin has been implicated in cellular motility becauseantibodies to talin decrease cell migration in fibroblasts(Nuckolls et al., 1992), and in growth cones inactivationof talin inhibits the dynamic extension and retraction offilopodia (Sydor et al., 1996). Moreover, tyrosine phos-phorylation ofb1 integrin may regulate interactions ofintegrins with the cytoskeleton because integrins phos-phorylated by pp60src have a decreased capacity to bindtalin (Tapley et al., 1989), and integrin clusters intransformed cells are phosphorylated (Johansson et al.,1994). Therefore, in contrast to vinculin, talin or paxillinmay participate in dynamic motile events that are notnecessarily linked to adhesion, such as filopodial exten-sion.

The presence of paxillin in point contacts provides amechanism for the recruitment of second-messengercascades that regulate motility and cytoskeletal dynam-ics. Studies in nonneuronal cells have shown that paxillinpossesses a binding site for SH2 domain proteins (Schallerand Parsons, 1995). In neurons, nerve growth factorstimulates tyrosine phosphorylation of paxillin (Choi etal., 1994) and causesb1 integrin to accumulate at the tipsof filopodia (Grabham and Goldberg, 1997). These find-ings demonstrate a link between stimulation of neuriteoutgrowth and adhesion and suggest that paxillin plays akey role in such convergent signaling. In addition,paxillin has a vinculin binding site (Turner et al., 1990),

suggesting that it may also participate in binding vinculinin point contacts at the tips of filopodia and in the centraldomain.

Point Contacts as Sites of Kinase ActivityWe investigated whether point contacts are mature

adhesion contacts with a signaling function by examiningthe distribution of two tyrosine kinases in DRG growthcones, pp125FAK and pp59fyn. In a previous study, weobserved immunoreactivity to FAK in focal contacts inastrocytes but not in PC12 cell point contacts, a finding atodds with the detection of FAK in the detergent-resistantcytoskeleton by Western blot in PC12 cells (Fig. 8). Wereexamined the immunoreactivity of FAK by using anantibody to a neuron-specific splice variant of FAK,called FAK1 (Burgaya and Girault, 1996; Craig andJohnson, 1996). In our cultures, FAK1 was concentratedin the central regions of growth cones, and colocalizationwas associated with vinculin in punctate clusters locatedin the central domain of the growth cones. This findingsuggested that FAK1 accumulates in the subpopulationof substrate-bound point contacts, in addition to a morewidespread localization. Our findings are also in agree-ment with those of a recent study reporting that FAK andvinculin are colocalized in hippocampal neurons (Stevenset al., 1996) and further indicate that the form of FAKfound in DRG growth cones is FAK1.

We also examined the immunolocalization ofpp59fyn in DRG neurons. Fyn is a nonreceptor tyrosinekinase that is highly enriched in growth cones (Helmkeand Pfenninger, 1995), and it can form stable links withboth pp125FAK and pp60src (Cobb et al., 1994). In ourcultures, fyn was generally punctate all over the growthcones, up to the tips of filopodia, which suggests aninvolvement in point contact signaling.

Rho Protein Family in Growth ConesBecause growth cones have such rapid morphologi-

cal responses to environmental influences and extracellu-lar cues, point contacts must be dynamically regulated toallow the transient adhesion of the growth cone and itsfilopodia as they sample the surrounding substrate. Innonneuronal cells, the recently identified family of smallRho GTPases appears to regulate cell adhesion andmigration (Mackay et al., 1996). We demonstrated thatRhoA, RhoB, and Cdc42 are present in primary growthcones. Cdc42 had a bright, diffuse appearance, particu-larly strong in filopodia, whereas RhoA and RhoBshowed more granular immunoreaction. Confocal analy-sis of explants double labeled with anti-RhoB/anti-vinculin demonstrated that the Rho GTPases were associ-ated with vinculin-containing point contacts located atcell margins or tips of filopodia. This finding is supported

Point Contacts in Neuronal Growth Cones 469

by the presence of a small amount of Rho, but not Cdc42,in the detergent-resistant cytoskeleton of PC12 cells (Fig.8). Also, aggregation of integrins by treatment withligand-coated beads induces accumulation of Rho clus-ters (Miyamoto et al., 1995). Together these findingssuggest that Rho GTPases may contribute to the dynamicregulation of growth cone morphology in response toenvironmental signals by transient associate with pointcontacts. There are indications that Rho may regulatetension in growth cones by acting on the actomyosinnetwork (Jalink et al., 1994). The finding that activationof Rho causes growth cone collapse (Tigyi et al., 1996)leads us to speculate that Rho regulates attachment anddetachment of adhesive contacts as the growth conesextend in axonal growth.

CONCLUSIONSWe have shown that integrin clusters form point

contacts in association with vinculin when rat DRGgrowth cones are plated on laminin. Other clusters ofintegrins that are not associated with vinculin also exist,and their function is not known. Adhesive point contactsshare most of the common cytoskeletal and signalingcomponents of focal contacts, although novel pointcontact-associated proteins may remain to be discovered.Association of integrin and vinculin could provide aclutch allowing the forward motion of growth cones withrespect to the substrate (Lin et al., 1996). We have alsoshown that the Rho GTPases are present in growth cones,which suggests that they may contribute to the regulationof integrin clusters in motility and pathfinding. Differ-ences in adhesion of individual filopodia at the vinculin-containing point contacts at filopodia tips could signal thedirection for cytoplasmic engorgement and consolidationof the extending axon. We speculate that point contactsare specialized to function as adhesive structures inhighly motile cells and growth cones.

ACKNOWLEDGMENTSWe thank Charles Essagian and Ester Yu for excel-

lent technical help, and Sal Carbonetto for providinganti-b1 integrin antibodies. A.R. was supported by anFCAR studentship and M.L. by an MRC postdoctoralfellowship.

REFERENCES

Arregui CO, Carbonetto S, McKerracher L. 1994. Characterization ofneural cell adhesion sites: point contacts are the sites ofinteraction between integrins and the cytoskeleton in PC12cells. J Neurosci 14:6967–6977.

Bershadsky AD, Tint IS, Neyfakh AAJ, Vasiliev JM. 1985. Focalcontacts of normal and RSV-transformed quail cells. Exp CellRes 158:433–444.

Bixby JL, Jhabvala P. 1993. Tyrosine phosphorylation in early embry-onic growth cones. J Neurosci 13:3421–3432.

Burgaya F, Girault J-A. 1996. Cloning of focal adhesion kinase,pp125FAK, from rat brain reveals multiple transcripts withdifferent patterns of expression. Mol Brain Res 37:63–73.

Burgaya F, Menegon A, Menegoz M, Valtorta F, Girault J-A. 1995.Focal adhesion kinase in rat central nervous system. Eur JNeurosci 7:1810–1821.

Burridge K, Fath K, Kelly T, Nuckolls G, Turner DC. 1988. Focaladhesions: transmembrane junctions between the extracellularmatrix and the cytoskeleton. Annu Rev Cell Biol 4:487–525.

Choi E-H, Rettig WJ, Wayner EA, Srour ML, Clegg DO. 1994.Functional identification of integrin laminin receptors thatmediate process outgrowth by human SY5Y neuroblastomacells. J Neurosci Res 37:475–488.

Cobb BS, Schaller MD, Leu T-Z. 1994. Stable association of pp60srcand pp59fyn with the focal adhesion-associated protein tyrosinekinase pp125FAK. Mol Cell Biol 14:147–155.

Craig SW, Johnson RP. 1996. Assembly of focal adhesions: progress,paradigms, and portents. Curr Opin Cell Biol 8:74–85.

Craig AM, Wyborski RJ, Banker G. 1995. Preferential addition ofnewly synthesized membrane protein at axonal growth cones.Nature 375:592–594.

Gomez TM, Letourneau PC. 1994. Filopodia initiate choices made bysensory neuron growth cones at laminin–fibronectin bordersinvitro. J Neurosci 14:5959–5972.

Gomez TM, Roche FK, Letourneau PC. 1996. Chick sensory neuronalgrowth cones distinguish fibronectin from laminin by makingsubstratum contacts that resemble focal contacts. J Neurobiol29:18–24.

Grabham PW, Goldberg DJ. 1997. Nerve growth factor stimulates theaccumulation ofb1 integrin at the tips of filopodia in the growthcones of sympathetic neurons. J Neurosci 17:5455–5465.

Hall A. 1998. Rho GTPases and the actin cytoskeleton. Science279:509–514.

Helmke S, Pfenninger KH. 1995. Growth cone enrichment andcytoskeletal association of non-receptor tyrosine kinase. CellMotil Cytoskeleton 30:194–207.

Jalink K, van Corven EJ, Hengeveld T, Morii N, Narumiya S,Moolenaar WH. 1994. Inhibition of lysophosphatidate- andthrombin-induced neurite retraction and neuronal cell roundingby ADP ribosylation of the small GTP-binding protein Rho. JCell Biol 126:801–810.

Johansson MW, Larsson E, Luning B, Pasquale E, Ruoslahti E. 1994.Altered localization and cytoplasmic domain-binding propertiesof tyrosine-phosphorylatedb1 integrin. J Cell Biol 126:1299–1309.

Letourneau PC, Shattuck TA. 1989. Distribution and possible interac-tions of actin-associated proteins and cell adhesion molecules ofnerve growth cones. Development 105:505–519.

Lin CH, Forscher P. 1995. Growth cone advance is inversely propor-tional to retrograde F-actin flow. Neuron 14:763–771.

Lin CH, Espreaficao EM, Mooseker MS, Forscher P. 1996. Myosindrives retrograde F-actin flow in neuronal growth cones. Neuron16:769–782.

Luna EJ, Hitt AL. 1992. Cytoskeleton–plasma membrane interactions.Science 258:955–964.

Luo L, Liao J, Jan LY, Jan YN. 1994. Distinct morphogenetic fucntionsof similar small GTPases: Drosophila Drac1 is involved inaxonal outgrowth and myoblast fusion. Genes Dev 8:1787–1802.

470 Renaudin et al.

Luo L, Hensch TK, Ackerman L, Barbel S, Jan YN. 1996. Differentialeffects of the Rac GTPase on Purkinje cell axons and dendriticspines. Nature 379:837–840.

Mackay DJG, Nobes CD, Hall A. 1996. The Rho’s progress: a potentialrole during neuritogenesis for the Rho family GTPases. TrendsNeurosci 18:496–501

McKerracher L, Chamoux M, Arregui CO. 1996. Role of laminin andintegrin interactions in growth cone guidance. Mol Neurobiol12:95–116.

Miyamoto S, Teramoto H, Coso OA, Gutkind JS, Burbelo PD,Akiyama SK, Yamada KM. 1995. Integrin function: molecularhierarchies of cytoskeletal and signaling molecules. J Cell Biol131:791–805.

Nermut MV, Eason P, Hirst EMA, Kellie S. 1991.Cell/substratumadhesion in RSV-transformed rat fibroblasts. Exp Cell Res193:382–397.

Nuckolls GH, Romer LH, Burridge K. 1992. Microinjection ofantibodies against talin inhibits the spreading and migration offibroblasts. J Cell Sci 102:753–762.

Ridley AJ, Hall A. 1992. The small GTP-binding protein Rho regulatedthe assembly of focal adhesions and actin stress fibres inresponse to growth factors. Cell 70:389–399.

Samuelsson SJ, Luther PW, Pumplin DW, Bloch RJ. 1993. Structureslinking microfilament bundles to the membrane at focal con-tacts. J Cell Biol 122:485–496.

Schaller MD, Parsons JT. 1995. pp125FAK-dependent tyrosine phos-phorylation of paxillin creates a high-affinity binding site forCrk. Mol Cell Biol 15:2635–2645.

Schaller MD, Borgman CA, Cobb BS, Vines RR, Reynolds AB,Parsons JT. 1992. pp125FAK, a structurally distinctive protein-tyrosine kinase associated with focal adhesions. Proc Natl AcadSci USA 89:5192–5196.

Schmidt CE, Dai J, Lauffenburger DA, Sheetz MP, Horwitz AF. 1995.Integrin–cytoskeleton interactions in neuronal growth cones. JNeurosci 15:3400–3407.

Stevens GR, Zhang C, Berg MM, Lambert MP, Barber K, Cantallops I,Routtenberg A, Klein WL. 1996. CNS neuronal focal adhesion

kinase forms clusters that co-localize with vinculin. J NeurosciRes 37:445–455.

Sydor AM, Su AL, Fen-Song W. 1996. Talin and vinculin play distinctroles in filopodial motility in the neuronal growth cone. J CellBiol 134:1197–1207.

Tapley P, Horwitz A, Buck C, Duggan K, Rohrschneider L. 1989.Integrins isolated from Rous sarcoma virus-transformed chickenembryo fibroblasts. Oncogene 4:325–333.

Tawil N, Wilson P, Carbonetto S. 1993. Integrins in point contactmediate cell spreading: factors that regulate integrin accumula-tion in point contacts vs focal contacts. J Cell Biol 120:261–271.

Threadgill R, Bobb K, Ghosh A. 1997. Regulation of dendritic growthand remodeling by Rho, Rac, Cdc42. Neuron 19:625–634.

Tigyi G, Fischer DJ, Sebok A, Yang C, Dyer DL, Miledi R. 1996.Lysophosphatidic acid-kinduced neurite retraction in PC12cells: control by phosphinositide2 Ca12 signaling and Rho. JNeurochem 66:537–548.

Turner CE, Glenney JR, Burridge K. 1990. Paxillin: a new vinculin-binding protein in focal adhesions. J Cell Biol 111:1059–1068.

Varnum-Finney B, Reichardt L. 1997. Vinculin deficient PC12 celllines extend unstable lamellipodia and filopodia and have areduced rate of neurite outgrowth. J Cell Biol 127:1071–1084.

Wigley CB, Berry M. 1988. Regeneration of adult rat retinal ganglioncell processes in monolayer culture: comparisons betweencultures of adult and neonatal neurons. Brain Res 470:85–98.

Wolven A, Okamura H, Rossenblatt Y, Resh RD. 1997. Palmitoylationof p59fyn is reversible and sufficient for plasma membraneassociation. Mol Biol Cell 8:1159–1173.

Wu D, Wang L, Mason CA, Goldberg DJ. 1996. Association ofb1integrin with phosphotyrosine in growth cone filopodia. JNeurosci 16:1470–1478.

Zheng J, Buxbaum R, Heidemann SR. 1994. Measurements of growthcone adhesion to culture surfaces by micromanipulation. J CellBiol 127:2049–2060.

Zipkin IL, Kindt RM, Kenyon CJ. 1997. Role of a new rho familymember in cell migration and axon guidance inC. elegans. Cell90:883–894.

Point Contacts in Neuronal Growth Cones 471