the f3 neuronal glycosylphosphatidylinositol-linked molecule is localized to glycolipid-enriched...

Journal of NeurochemisttyLippincott-Raven Publishers, Philadelphia

1995 International Society for Neurochemistry

The F3 Neuronal Glycosylphosphatidylinositol-LinkedMolecule Is Localized to Glycolipid-EnrichedMembrane Subdomains and Interacts with

L l and Fyn Kinase in Cerebellum

Sylviane Olive, *Catherine Dubois, tMelitta Schachner, and Geneviève Rougon

Laboratoire de Génétique et Physiologie du Développement, UMR 9943 CNRS-Université Aix-Marseille 11, Marseille;*Laboratoire de Biologie Cellulaire, Faculté de Médecine Saint-Antoine; Paris, France ;

and tNeurobiologie, ETH Hönggerberg, Zürich, Switzerland

Abstract : The F3 molecule is a member of the immuno-globulin superfamily anchored to plasma membranes bya glycosylphosphatidylinositol group. In adult mousecerebellum, F3 is predominantly expressed on a subsetof axons, the parallel fibers, and at their synapses . In vitrostudies established that it is a plurifunctional moleculethat, depending on the cellular context and the ligandwith which it interacts, either mediates repulsive interac-tions or promotes neurite outgrowth. In the present study,we report the isolation of two fractions of F3-containingmicrodomains from adult cerebellum on the basis of theirresistance to solubilization by Triton X-100 at 4°C. Bothfractions were composed of vesicles, ranging from 100to 200 nm in diameter. Lipid composition analysis indi-cated that the lighter fraction was enriched in cerebro-sides and sulfatides . F3 sensitivity to phosphatidylinositolphospholipase C differed between the two fractions, pos-sibly reflecting structural differences in the lipid anchorof the F3 molecule . Both fractions were highly enrichedin other glycosylphosphatidylinositol-anchored proteinssuch as NCAM 120 and Thy-1 . It is interesting that thesevesicles were devoid of the transmembrane forms (NCAM180 and NCAM 140), which were recovered in Triton X-100-soluble fractions, but contained the L1 transmem-brane adhesion molecule that is coexpressed with F3 onparallel fibers and the fyn tyrosine kinase . Immunoprecipi-tation experiments indicated that F3, but not NCAM 120or Thy-1, was physically associated in a complex withboth L1 and fyn tyrosine kinase . This strongly suggeststhat the interaction between L1 and F3, already describedto occur with isolated molecules, is present in neural tis-sue. More important is that our study provides informa-tion on the molecular machinery likely to be involved inF3 signaling . Key Words: Neuronal cell adhesion-Im-munoglobulin superfamily-Glycosylphosphatidylinosi-tol linkage-Triton X-100-insoluble complexes-Signaltransduction .J. Neurochem. 65 2307-2317 (1995) .

In the nervous system specific synaptic connectionsare established during development by extensions of

2307

axons toward their targets along restricted pathways(Dodd and Jessell, 1988 ; Schachner, 1991, 1995 ;Hynes and Lander, 1992; Goodman, 1994) . Severalmajor classes of surface molecules have been impli-cated in the guidance of axons or maintenance of syn-apses (Rathjen, 1991) . Among them, immunoglobulin(1g)-like proteins, such as Ll or NCAM, can serveboth as receptors on the surface of the growth coneand as substrate for growing axons (Doherty et al .,1989 ; Lemmon et al ., 1989) . Several other glycopro-teins from this family, which are localized predomi-nantly on axons of vertebrate neurons, have also beencharacterized (Brümmendorf and Rathjen, 1993) .Among them, BIG-1 (Yoshihara et al ., 1994), ratTAG] (Furley et al ., 1990) and its chick homologueaxonin-1 (Zuellig et al ., 1992), and mouse F3 (Gennar-ini et al ., 1989a,b) and its chick homologue Fl l(Brümmendorf et al ., 1989) have a glycosylphosphati-dylinositol (GPI) membrane anchor .

Several in vitro studies aimed at deciphering F3function showed that it is a multifunctional moleculeinvolved in different types of heterophilic interactions,depending on its physical form (anchored to the mem-brane or soluble) and on the availability of differentreceptors (multiple ligands) (Gennarini et al ., 1991 ;Durbec et al ., 1992, 1994; Pesheva et al ., 1993) . F3-mediated neurite extension and repulsion are likely to

Received March 17, 1995 ; revised manuscript received May 15,1995 ; accepted May 15, 1995 .

Address correspondence and reprint requests to Dr . G . Rougon atLaboratoire de Génétique et Physiologic du Développement, CNRS9943, Faculté des Sciences de Luminy-Case 907, F-13288 MarseilleCedex 9, France .

Abbrevialions used: GPI, glycosylphosphatidylinositol ; GSL, gly-cosphingolipid ; HP-TLC, high-performance TLC ; Ig, immunoglobu-lin ; MBS, MES-buffered saline ; MDCK, Madin-Darby canine kid-ney; PAGE, polyacrylamide gel electrophoresis ; PBS, phosphate-buffered saline; Pl-PLC, phosphatidylinositol phospholipase C ; SDS,sodium dodecyl sulfate ; Tx-100, Triton X-100 .

2308

involve at least two steps : first, a recognition (adhe-sive) event with a ligand expressed on a neighboringcell or in the extracellular matrix, and second, trans-membrane signaling that leads eventually to reorgani-zation of the cytoskeleton and to a cellular response .Hence, cellular components linking F3 on the cell sur-face with structural elements in the cytoplasm mustexist. How such GPI-linked molecules transduce sig-nals is far from clear . Experiments made with GPI-anchored molecules expressed by cells other than neu-rons, such as lymphocytes, have shown that antibody-mediated cross-linking of these proteins induces rapidphosphorylation of several substrates on tyrosine resi-dues (Hsi et al ., 1989) . In agreement with this, aninteraction between src-like kinases and GPI moleculeshad also been evidenced (Stefanova et al ., 1991) . Thepattern of expression of scr, fyn, and yes protein ki-nases throughout the developing brain would be con-sistent with a potential role in neurite outgrowth (Ma-ness, 1992) . However, it is not clear how GPI-an-chored molecules and such kinases can associate .Presumably, a transmembrane "linker" protein medi-ates the interaction between the two . Cinek and Horejsf(1992) showed the existence of specialized membranemicrodomains that contained several GPI-anchoredproteins and glycolipids as well as p561 ck kinase . Insupport of this, Brown and Rose (1992) recoveredGPI-anchored molecules from epithelial cells from Tri-ton X-100 (Tx-100) lysates in detergent-resistantmembrane vesicles that are rich in glycolipids . Morerecently, it was shown that these complexes might pro-vide a model for studying the interactions betweenlipids and proteins organized in possibly functionalmicrodomains (Sargiacomo et al ., 1993 ; Lisanti et al .,1994) .Although to our knowledge this approach has never

been applied to neural cells, we set out to isolate simi-lar microdomains enriched in GPI-linked moleculesfrom cerebella tissue-a source highly enriched inF3-expressing axons (Faivre-Sarrailh et al ., 1992) .

In this article we report the isolation of two popula-tions of F3-containing membrane patches insoluble inTx-100 at 4 °C . They were enriched in other GPI-linkedmolecules such as Thy-l and NCAM 120 but devoid ofthe transmembrane forms of NCAM. The Ll moleculedistributed in both Tx-100-insoluble and -soluble frac-tions . Furthermore, analyses of immunoprecipitatesobtained with anti-F3 and anti-L1 antibodies, respec-tively, indicated a physical association between thesetwo molecules . The immunoprecipitates were alsoshown to contain the fyn tyrosine kinase . The assemblyof these proteins in a complex could be at the basis ofF3-triggered events occurring in neurons .

MATERIALS AND METHODS

ChemicalsPhenylmethylsulfonyl fluoride was from Boehringer

Mannheim (Germany), and protein A-Sepharose was from

J. Neurochem., Vol . 65, No . 5, 1995

S. OLIVE ET AL.

Pharmacia . Unisil was from Clarkson Chemical (Williams-port, PA, U.S.A .) . High-performance TLC (HP-TLC) andSilica gel 60 plates were from Merck (Darmstadt, Ger-many) . Sep-Pak (C18 cartridges) was from Water Associ-ates (Milford, MA, U.S.A .) . All chemicals unless otherwiseindicated were from Merck or Sigma (Paris, France) .

Phosphatidylinositol phospholipase C (PI-PLC) from Ba-cillus thurengiensis was purified in our laboratory (Théve-niau et al ., 1990) .

AntibodiesAnti-F3 was a rabbit polyclonal antibody prepared against

a fusion protein comprising the F3 Ig-like domains (Genna-rini et al ., 1991) . Anti-NCAM was a rabbit polyclonal raisedagainst purified mouse NCAM and recognizing all NCAMisoforms . Thy-l was detected with a rat monoclonal antibody(clone 154-177), a kind gift from Dr . H.-T . Hé (CIML,Marseille, France) . Anti-LI was a rat monoclonal antibody(Rathjen and Schachner, 1984) . Irrelevant antibodies usedas control in immunoprecipitation experiments were a rabbitpolyclonal anti-placental alkaline phosphatase prepared inour laboratory and a rat IgG monoclonal antibody directedagainst P31 /HSA (NMelec et al ., 1992) . The mouse IgGImonoclonal antibody anti-caveolin (also recognizing mousecaveolin) (no . C13620) was distributed by Affiniti (GPTBusiness Park, Nottingham, U.K.) . The rabbit polyclonalanti-c-src, reacting with mouse src r6°, Yes p6z , andfyn °59 , wasdistributed by Santa Cruz Biotechnology (U.S.A .), and therabbit anti-fyn tyrosine kinase (no . 06-133 ; batch 12109)was from UBI (U.S .A .) .

Isolation of low-density Tx-100-insolublecomplexesOne cerebellum freshly dissected from a 5-month-old

mouse (Swiss strain), anesthetized and killed according tothe experimental protocols approved by the European legis-lation, was homogenized at 4°C by 12 strokes of a Douncehomogenizer equiped with a Teflon pestle in 2 ml of MES-buffered saline [MBS ; 2 mM morpholinoethanesulfonic acid(pH 6.5) and 0.15 M NaCl ] containing I% Tx-100 and 1mM phenylmethylsulfonyl fluoride as a protease inhibitor .The extract was adjusted to 40% sucrose (with 2 ml of MBSbuffer containing 80% sucrose) and placed at the bottom ofa 13-ml SW41 ultracentrifuge tube . A 5-30% linear sucrosegradient was formed above the lysate in MBS lacking Tx-100 (6 ml) and centrifuged at 4°C for 16-22 h at 39,000rpm in an SW41 Beckman rotor . Fractions (1 ml) wereharvested from the top . An opaque band (migrating at 10-20% sucrose density) was easily discernible and corres-ponded to the low-density Tx-100-insoluble complexes .Each fraction was immediately centrifuged at 100,000 g for60 min, and pellets and supernatants were separated andsubsequently analyzed .

ImmunoblotPellets and supernatants of each fraction of the gradient

were mixed with an equal volume of reducing or nonreduc-ing (wherever indicated) electrophoresis sample buffer andwere boiled for 3 min . Samples were separated by 7 or 12%polyacrylamide gel electrophoresis (PAGE) . After blottingonto nitrocellulose membranes (Amersham) and saturationin phosphate-buffered saline (PBS) containing 5% defattedmilk, proteins were detected by incubation with primary anti-bodies (overnight at 4°C) . Bound antibodies were revealedby incubation with immunopurified rabbit anti-species anti-

CHARACTERIZATION OF F3-CONTAINING MICRODOMAINS

bodies (Immunotech, France ; 4 l-tg/ml in PBS containing5% defatted milk) when needed, i .e ., mouse anti-caveolin,rat anti-L1, and anti-Thy-1 monoclonal antibodies, for 4 hat 4°C . Then rabbit bound antibodies were revealed with 1251_

protein A (0.5 X 10 6 cpm/ml) for l h .In some instances, several proteins were sequentially de-

tected on the same blot . Between uses bound antibodieswere removed from the blot by desaturation with 2% sodiumdodecyl sulfate (SDS) and 100 mM 2-mercaptoethanol in50 mM Tris-HCI (pH 6.8) for 30 min at 70°C followed byextensive rinsing in PBS and saturation with PBS containing5% milk .The relative intensities of bands were determined by den-

sitometry .

ImmunoprecipitationHigh-speed pellet fractions obtained from Tx-100-insolu-

ble low-density material were resuspended and homogenizedin 1 % Nonidet P-40 and 0.5% deoxycholate in 50 mM Trisand 150 mM NaCl for 2 h at 4°C . After preclearing (8 h at4°C) with protein A-Sepharose with shaking, samples wereimmunoprecipitated with the corresponding antibody on pro-tein A preformed complexes for 12 h at 4°C . The beads werewashed four times sequentially, first with 50 in M Tris and150 mM NaCl, then twice with 50 mM Tris, 150 mM NaCl,and 1 % Nonidet P-40, and again with the first buffer, all atpH 7.4 .When a coimmunoprecipitated protein was searched for

in the immunoprecipitates, they were analyzed as describedabove by immunoblot with the corresponding antibody .Then, the blot was desaturated and subsequently analyzedwith the antibody used for the immunoprecipitation .

PI-PLC sensitivityHarvested fractions containing the low-density Tx-100-

insoluble complexes resuspended in PBS were incubatedwith PI-PLC (1 U/ml) for 1 h at 37°C before the separationof soluble and insoluble fractions by centrifugation at100,000 g (4°C, l h) . In some instances pellets were sub-jected to a second round of PI-PLC treatment under thesame conditions . Pellets and supernatants were separated andanalyzed by SDS-PAGE and immunoblot. In some experi-ments, the treatment was conducted on cerebellar homoge-nates (1 h at 37°C), and then the temperature was loweredto 4°C overnight before separation of the Tx-100-insolublecomplexes .

Lipid analysisTotal lipids were extracted as described by Dubois et al .

(1990) . In brief, sucrose gradient fractions were first sub-jected to ultracentrifugation (100,000 g as described above),and pellets were rinsed with MBS buffer to remove most ofthe sucrose. Lipids from pellets and from one cerebellarhomogenate in MBS buffer were extracted with chloroform/methanol (2 :1 vol/vol) and chloroform/methanol/water(4:8 :3 by volume) . The combined extracts were dried andapplied to a 2-cm-long silicic acid (Unisil) column in chloro-form (Irwin and Irwin, 1979) . The neutral lipids and gan-gliosides were eluted with two different mixtures of chloro-form/methanol/water (65 :25 :4, then 5:5 :1 by volume) .

The neutral lipids were evaporated to dryness and thenexamined by HP-TLC using chloroform/methanol/water(65 :25 :4 by volume) as a developing solvent . They werevisualized by spraying with orcinol reagent .The ganglioside fractions were desalted on a Sephadex LH

20 column and then separated by HP-TLC using chloroform/methanol/0.25% KCl (5 :4 :1 by volume) and detected byresorcinol .

Electron microscopyFractions 3 and 5 of the sucrose gradient were centrifuged

at 100,000 g and incubated for 1 h at 4°C in a fixative solutioncontaining 2% paraformaldehyde and 3% glutaraldehyde in0.1 M phosphate buffer, pH 7.4 . After two washes with 5%sucrose in phosphate buffer, pellets were postfixed for 40min in 2% osmium tetroxide in phosphate buffer, stained for30 min with uranyl acetate, dehydrated, and embedded inEpon . Ultrathin sections were cut and examined on a Hitachielectron microscope.

RESULTS

2309

Isolation of F3-rich membrane domainsSeveral reports in the literature have suggested that

GPI-anchored proteins are not fully solubilized by Tx-100 (reviewed by Low, 1989) . To test if this were thecase for F3, we used adult mouse cerebellum, an F3-rich source in which it is known that the vast majorityof F3 present in this structure is expressed on axonsof the granule cells (parallel fibers) in the molecularlayer and at synapses they establish with Purkinje celldendrites (Faivre-Sarrailh et al ., 1992) . To this tissuewe applied the recently described procedure for isolat-ing caveolin-rich membrane domains from culturedMadin-Darby canine kidney (MDCK) epithelial cells(Sargiacomo et al ., 1993), which depends on (a) resis-tance to solubilization by Tx-100 at 4°C and (b) buoy-ancy at specific density in sucrose gradients .

For each experiment one adult cerebellum was ho-mogenized in 1% Tx-100, and the homogenate wassubjected to an equilibrium density gradient centrifu-gation according to the procedure described in Tech-niques .Each fraction (1 ml) was spun at 100,000 g, and

pellets and supernatants were analyzed for their proteinconcentration. Most of the proteins were found in theTx-100-soluble fractions recovered at the bottom ofthe gradient, and only small amounts were found inthe upper fractions (Fig . I ) . The presence of F3, Thy-1, and NCAM proteins was investigated by immu-noblot in both the resulting pellets and supernatants .F3, Thy-1, and NCAM 120 isoforms were present infractions 3-7, and >80% of them were recovered inthe 100,000-g pellets (shown for F3 and NCAM 120in Fig . I ) . This indicated that these GPI-anchored mol-ecules behaved similarly during the treatment and thatthey are associated with low-density lipid-rich com-plexes, insoluble in Tx-100 . Only small amounts ofthe tested GPI molecules were found in the pellet offraction 12 at the bottom of the gradient (data notshown) .The gradient separation was highly reproducible be-

tween experiments . It is interesting that GPI proteinswere constantly recovered in two areas of the gradientwith different sucrose density . These were fraction 3

J. Neurochem., Vol. 65, No. 5, 1995

2310

FIG . 1 . Sucrose density gradient analysis of F3, NCAM, andThy-1 from cerebellum lysates . After separation on a gradient of5-30% sucrose of a lysate corresponding to one adult mousecerebellum, 10 fractions were collected from the top . All fractionswere pelleted at 100,000 g, and resulting pellets (P) and superna-tants (SN) were diluted in electrophoresis buffer and separatedinto three aliquots to analyze their F3, NCAM, and Thy-1 contentby immunoblot . For F3 (upper panels) and NCAM (middle pan-els) data are shown for both the soluble fraction (SN) and pellets(P) after high-speed centrifugation ; for Thy-1 only pellet analysisis shown (lower right panel) . Note that GPI-anchored proteins(F3, Thy-1, and NCAM 120) were recovered in the low-densityTx-100-insoluble complexes corresponding to pellets of frac-tions 3-7, whereas NCAM transmembrane proteins (NCAM 140and 180) were recovered in the high-density Tx-100-soluble frac-tions (SN, fractions 7-10) . GPI-anchored proteins were concen-trated in two separate areas of the gradient, corresponding tofraction 3 and fractions 5 and 6, respectively . Protein content ineach fraction (lower left panel) was estimated by a Bio-Radassay performed before centrifugation . Fractions 2-6, corre-sponding to Tx-100-insoluble material, contained only -500 pgof proteins, whereas most of them (6 .5 mg) were recovered inTx-100-soluble fractions .

and fractions 5 and 6, respectively . This likely reflectsa different composition of the fractions, especially intheir lipid content .By contrast, the transmembrane isoforms of NCAM,

i.e ., NCAM 140 and 180, were found in fractions 7-10, and >90% were recovered in the supernatant afterhigh-speed centrifugation in agreement with their solu-bility in Tx-100 (Fig . 1) .

Electron microscopic analysis of low-densityTx-100-insoluble materialTo examine in more detail the Tx-100-insoluble

fractions that we had isolated on sucrose gradients, weprepared the material from fractions 3 and 5, respec-tively, for electron microscopy . Figure 2 shows thatthe major difference between the two fractions residedin their content in myelin sheaths . About 80% of thetotal myelin sheaths were recovered in fraction 3, andit is very likely that myelin determined its high buoy-ancy on sucrose gradient. In both fractions, most ofthe vesicles ranged from 100 to 200 nm in diameterand apparently contained membrane bilagers . These

J. Nerro,hem ., Vol. 65, No. 5, 1995

S. OLIVE ET AL.

were more numerous in fraction 5 than in fraction 3,although indistinguishable in their size. Pellets ob-tained from fraction 3 were consistently of larger vol-ume than those of fraction 5 . When equivalent amountsofproteins from fractions 3 and 5 were electrophoresed

FIG . 2. Morphological and biochemical comparison of Tx-100-insoluble complexes in fractions (A) 3 and (B) 5, respectively .High-speed centrifugation pellets from fractions 3 and 5 wereprepared as described in Materials and Methods and examinedby transmission electron microscopy . Note that the fraction 3was highly enriched in myelin sheaths (arrowheads) . No clearmorphological difference could be seen among vesicular struc-tures (100-200 nm) observed in the two fractions . These weremore numerous in fraction 5 ; however, the size of the pellet forfraction 3 was larger than for fraction 5 . Inset: Proteins in frac-tions 3 and 5 were electrophoresed on SDS-PAGE and revealedby Coomassie Brilliant Blue staining .


FIG . 3. Glycolipid composition ofTx-100-insoluble fractions and totalcerebellum . Lipids were extractedand analyzed for each fraction as de-scribed in Materials and Methods .For fractions 3 and 5 half of the lipidscorresponding to the fraction werecharged, whereas it was 1 :4 for thewhole membrane extract (T) . A: Neu-tral glycolipids visualized with orci-nol . Lanes 1-3 (ST) represent mixedglycolipid standards . CTH, ceramidetrihexoside ; CDH, ceramide dihexo-side ; CMH, ceramide monohexoside(cerebrosides) ; Sulf, sulfatide ; GM3,ganglioside GM3 . Membrane majorcomponent cerebrosides (CMH) andsulfatides were mainly recovered infraction 3 . B : Gangliosides visualizedwith resorcinol . Lanes 1-3 represent mixed ganglioside stan-dards . Their position is indicated in the left margin . Amounts ofgangliosides in fraction 5 were very low and could hardly beseen on the photograph . Nonspecific staining due to sucrose isindicated (*) .

on SDS-PAGE followed by Coomassie Brilliant Bluestaining, no difference could be detected in either qual-ity or quantity, and the most abundant of them werein the range of 100-200 kDa molecular size (Fig . 2) .

The vesicles we isolated were larger and more het-erogeneous in size than intracellular transport vesicles .For this reason, we believe that most of these deter-gent-resistant complexes do not exist in the cells asintact vesicles before detergent lysis . It is more likelythat in the cell, the complexes are found as microdo-mains of associated proteins and lipids within the bi-layer .Lipid composition of the Tx-100-insoluble vesicles

After purification of lipids, the glycolipid composi-tion of fractions 3 and 5 was analyzed by HP-TLCby comparison with glycolipid composition of totalcerebelumr homogenate . The HP-TLC plate shown inFig . 3 displays the neutral glycolipids and sulfatidesof the Tx-100-insoluble fractions and total membranes .The cerebrosides and the sulfatides were the main com-ponents of total membranes . Other components, suchas di- and trihexoside ceramides, were also identified,although the faint bands were barely detectable on thephotograph of the plate (Fig . 3) . Most of the glyco-sphingolipids (GSLs) found in total membranes wererecovered in fraction 3 . In addition, this fraction wasenriched in one component, very likely a complex gly-colipid migrating below GM3, barely detectable in to-tal membranes . Fraction 5 also contained cerebrosidesand sulfatides but in much lower quantity than fraction3 . The presence of GSLs in Tx-100-insoluble fractionsis in agreement with the observations of Brown andRose (1992) on epithelial cells . A major difference,however, between their data and ours is that in cerebel-lum these components were also major components ofmembranes, whereas in epithelial cells they are minorcomponents .

231 1

Purified gangliosides were also analyzed (Fig . 3) .In the whole membrane extract they were mainly GMI,GD I a, GD3, and GTI b, in agreement with gangliosidecomposition of adult brain (Sonnino et al ., 1983) . Asfor the GSLs, most of them were recovered in fraction3 . GTIb appeared enriched in fraction 5 by comparisonwith the others .

Sensitivity of F3 in microdomains to PI-PLCWe investigated whether F3 found in the two distinct

sucrose fractions could be distinguished on the basisof its sensitivity to PI-PLC (Fig . 4) . This was doneby first treating each F3-containing fraction individu-ally by B. thuringiensis PI-PLC enzyme and monitor-ing the quantity of F3 that can be recovered in solubleform after the treatment . Figure 4B shows the resultsobtained for fractions 3 and 5, respectively . It wasquite clear that F3 in fraction 5 was fully solubilizedby the action of the enzyme, whereas, under the sameconditions, -40% of F3 in fraction 3 remained associ-ated with the pellet . Moreover, the level of cleavagewas not improved after a second performance of PI-PLC treatment (data not shown) . To check that thisresistance to PI-PLC cleavage did not result from inter-actions that F3 contracts during microdomain separa-tion or to an inhibition by myelin components highlyrepresented in fraction 3, we subjected the tissue ho-mogenate to treatment with PI-PLC before the micro-domain separation procedure (Fig . 4A) . Only F3 con-tained in the higher part of the gradient (fraction 3)was recovered in the pellet after high-speed centrifuga-tion of the fractions . This indicates that the resistanceto PI-PLC was intrinsic to that portion of F3 in thecerebellum . This could be explained by the structureof the GPI anchor, which might differ in the two frac-tions . It is known that GPI anchors share a common

FIG. 4 . PI-PLC sensitivity of F3 in low-density Tx-100-insolublecomplexes . For both experiments shown samples were incu-bated for 1 h at 37°C with (+PIPLC) or without (-PIPLC) PI-PLC enzyme . A: PI-PLC treatment was performed before theseparation on sucrose gradient . Note that the treatment led tothe disappearance of F3 in high-speed pellets of fractions 5 and6 but not of fraction 3 . B : PI-PLC treatment was performed onfractions 3 and 5, respectively . After treatment the samples werecentrifuged at 100,000 g, and pellets (P) and supernatants (SN)were analyzed by immunoblot for their F3 content . PI-PLC wasable to solubilize >95% of the F3 molecules in fraction 5 butonly 55% in fraction 3 as analyzed by densitometry .

1. Neurochem., Vol, 65, No. 5, 1995

2312

backbone, but several distinct variations such as theattachment of galactosyl side chains and additionalphosphoethanolamine or mannosyl residues have beendescribed . Moreover, there are considerable variationsin the type of fatty acyl or alkyl groups linked to theglycerol . Finally, in some instances the inositol phos-phate residue can be acylated . Some of these variationsmight be responsible for the lack of susceptibility toPI-PLC observed for certain subsets of GPI-anchoredmolecules (reviewed by Low, 1989; Ferguson, 1992) .Alternatively, F3 could be associated with anothermolecule that may retain the cleaved form in the insol-uble fraction . This would imply that the interaction isTx-100 resistant. It also shows that the presence of theGPI anchor is necessary for F3 to be recovered inthat fraction because the solubilized forms distributedrandomly in the Tx-100-soluble fractions in the lowerpart of the gradient (data not shown) .

The GPI molecule-containing microdomains aredevoid of caveolin

Several recent reports indicate that a large fractionof GPI-anchored molecules of epithelial cells (Sargia-como et al ., 1993) or smooth muscle cells (Chang etal ., 1994) could be recovered in subcompartments ofthe plasma membrane called caveolae (also known asplasmalemmal vesicles) (Anderson, 1993a,b) . Caveo-lin, a 22-kDa transmembrane phosphoprotein, is animportant structural component of caveolae (Glenneyand Soppet, 1992) . As several caveolar membranecomponents (GPI-linked proteins, GSLs, and caveo-lin) are selectively resistant to solubilization by Tx-100at 4°C, we searched for the presence of this molecule incerebellar homogenates and in our sucrose fractions .An antibody recognizing caveolin in extracts ofMDCK cells (taken as a positive control) was used(Fig . 5 ) . Immunoreactivity could not be detected eitherin cerebellum homogenate or in the sucrose fractions(fractions 2-7, corresponding to low-density Tx-100-insoluble complexes, are shown in Fig . 5) and in otherparts of the brain tested such as hippocampus, hypoph-ysis, or hypothalamus (data not shown) . These dataare in accordance with the findings of Glenney andSoppet (1992), who reported that caveolin was absentfrom brain tissue .

The Ll transmembrane adhesion molecule ispresent in both Tx-100-soluble and -insolublefractions

L1 is a transmembrane adhesion molecule of the Igsuperfamily highly expressed on granule cell axons inthe adult cerebellum (Persohn and Schachner, 1987) .Several mechanisms have been identified by which L1interacts : homophilic binding between like moleculeson opposing cell surfaces (Lemmon et al ., 1989) butalso heterophilic binding between L1 (or Ng-CAM, aputative chick homologue of mouse L1) and axonin-1 /TAG1 (Kuhn et al ., 1991), F11 (chick F3 homologue)(Brümmendorf et al ., 1993), and yet unidentified ex-tracellular matrix molecules (Appel et al ., 1993) .

J. Neurochem., Vol. 65, No . 5, 1995

S. OLIVE ET AL.

FIG. 5. Analysis of caveolin and L1 molecules in cerebellumlysates. Cerebellum lysate preparation and gradient separationwere done as described in Materials and Methods. A: Immu-noblot analysis of caveolin . Lanes on both sides correspond toanalysis of MDCK cell extracts used as a positive control forcaveolin (22 kDa) . Caveolin could not be detected in high-speedpellets of fractions 2-7 corresponding to one cerebellum lysate .B: Sucrose density gradient analysis of L1 from cerebellum ly-sates. Fractions (1-6) corresponding to the low-density Tx-100-insoluble proteins were pelleted at 100,000 g, and resulting pel-lets were diluted in electrophoresis buffer (analysis of 1 :3 of thetotal) ; other fractions (7-10) were directly diluted in sample buffer(analysis of 1 :50 of the total) . Each fraction was then analyzedby immunoblot for its Li content using anti-L1 monoclonal anti-body . L1 molecules were recovered both in the low-density Tx-100-insoluble complexes corresponding to pellets (15%) of frac-tions 2-5 (with a maximum in fractions 2 and 3) and in solublefractions (7-10) (85%). Note that L1 was essentially found in its200-kDa form in soluble fractions, whereas both 200- and 140-kDaforms were revealed in Tx-100-insoluble fractions.

To demonstrate a possible functional association be-tween F3 and L1 by a different approach, we firstanalyzed the distribution of L1 in our gradient fraction-ation . The data shown (Fig . 6) indicate that -85% ofthe molecule was recovered at the bottom of the gradi-ent (fractions 7-10), as expected for a transmembranemolecule, and colocalized with transmembrane iso-forms of NCAM. However, the remaining part (15-20%, depending on the experiment) was found colocal-ized with F3 and other GPI-anchored molecules mainlyin fractions 2 and 3 . It is interesting that the 140-kDaform of L1, described as arising from a proteolyticcleavage of the 200-kDa form (Faissner et al ., 1985 ;Sadoul et al ., 1988), was essentially found in the upperfractions .

Ll and F3 molecules coprecipitateIn a second series of experiments, F3 and L1 were

first immunoisolated from fractions 3 and 5 of thegradient using their respective specific antibodies .Then, after separation of the proteins contained in theimmunoprecipitates by SDS-PAGE and transfer to anitrocellulose membrane, the presence of L l was in-vestigated by immunoblot in the F3 immunoprecipi-tate, and conversely . In whatever combination, bothL1 and F3 molecules could be found in the immuno-precipitates as an indication that they were interacting


FIG. 6 . Analyses of L1 and F3 immunoprecipitates (IPP) ob-tained from fractions 3 and 5, containing the Tx-100-insolublecomplexes, and fraction 10, containing Tx-100-insoluble pro-teins . Immunoprecipitations were performed with anti-L1 andanti-F3 antibodies on fractions 3, 5, and 10, respectively, ac-cording to procedures described in Materials and Methods. Im-munoprecipitates were electrophoresed on 7% SDS-PAGE, andproteins were transferred to nitrocellulose membranes . Thesewere first probed for the presence of L1 and F3 ; then, afterremoval of bound antibodies, they were sequentially analyzedfor the presence of F3 (in L1 immunoprecipitates), L1 (in F3immunoprecipitates), and NCAM in both . We can see that L1and F3 coprecipitate in fractions 3 and 5 containing the Tx-100-insoluble proteins, but not in soluble fractions . NCAM could notbe detected in these immunoprecipitates . Note that only the140-kDa forms of L1 were found in the immunoprecipitates inour conditions .

in the microdomains . The specificity of this interactionwas asserted by the following controls : (a) LI wasabsent. when the precipitation was done with irrelevantantibodies . (b) Neither NCAM 120 nor Thy-1 couldbe revealed in the immunoprecipitates . Thus, we werenot evidencing an interaction between L1 and any kindof GPI-linked molecule . The molecular size of the Llmolecule coprecipitated with F3 was consistentlyfound (six independent experiments) to be 140 kDa,whereas the 200-kDa form, which can also be foundin the microdomains, was absent . Identical results werefound (data not shown) when in one experiment weused an anti-L1 polyclonal antibody . At this stage itis impossible to decide whether the 140-kDa form wasgenuinely associated with F3 in the tissue or whether itderived nonphysiologically from proteolytic cleavagetaking place during our isolation procedure .

Fyn protein kinase, LI, and F3 can becoprecipitated as a complex from themicrodomainsThe same approach as described above was applied

to investigate the formation of a complex between F3and protein kinases of the src family . The rationaleof the experiment was based on the observation thatinteractions between GPI molecules and src kinasesoccur in lymphocytes (Stefanova et al ., 1991 ; Thomasand Samelson, 1992) . We also observed that when

expressed in Chinese hamster ovary cells, F3 coprecip-itated with fyn tyrosine kinase (S . Gomez, personalcommunication) . Moreover, src-related protein ki-nases in nerve growth cones (p59' - 'Y", pp60c-s,

c, andpp62`-yes ) are potential intracellular signaling mole-cules for cell adhesion molecule-directed axonalgrowth (see Atashi et al ., 1992 ; Beggs et al ., 1994 ;Ignelzi et al ., 1994) .

First, by using a polyclonal antibody directed againstc-src kinases we observed that immunoreactivity wasfound both coloealizing with F3 in the sucrose gradientand in Tx-100-soluble fractions (data not shown) . Sec-ond, an antibody specific for fyn kinase gave a similardistribution pattern as c-src kinases (Fig . 7) . Third,when the F3 or L 1 immunoprecipitates were analyzedfor the presence of fyn kinase, this molecule was foundin both cases . The specificity of the interaction wasdemonstrated by the observation that NCAM 120 im-munoprecipitates did not contain fyn kinase (data notshown) .

DISCUSSION

2313

The assembly of membrane proteins is a key stepin the formation of functionally special membrane do-mains such as coated pits, synapses, or cell adhesionstructures (Edidin, 1990) . In the present study, we

FIG. 7 . A: Sucrose density gradient analysis of the fyn proteintyrosine kinase in cerebellum lysate . Tx-100-insoluble fractions(1-6) were pelleted and resuspended in nonreducing buffer ; theother fractions (7-10) were diluted 1 :2 in the same buffer . Allfractions were then analyzed by 12% SDS-PAGE and immu-noblotted for their content in the fyn protein tyrosine kinase .For fractions 7-10, only 1 :50 of the total fraction was analyzed,whereas 1 :3 of the pellets were analyzed for fractions 1-6 . B :Presence of fyn protein tyrosine kinase in F3 and L1 immunopre-cipitates . Immunoprecipitations were performed on fractions 3and 5 of the sucrose gradient with anti-F3 (F3+ lane), anti-L1(L1+ lane), or irrelevant antibodies (see Materials and Methods)(F3- and L1- lanes), according to conditions described in Mate-rials and Methods. Proteins in the immunoprecipitates were elec-trophoresed on 12% SDS-PAGE, transferred to nitrocellulosemembranes, and analyzed by immunoblot for their content offyn protein tyrosine kinase. Fyn kinase (60 kDa) was detectedboth in F3+ and L1 + immunoprecipitates but not with irrelevantantibodies .

J. Neurochem., Vol. 65, No . 5, 1995

2314

have shown that the F3 GPI-linked adhesion moleculeexpressed in adult mouse cerebellum is present in largenoncovalent complexes resistant to dissociation by thecommonly used nonionic detergent Tx-100 . Inagreement with their buoyancy properties on sucrosegradient, these complexes contained high amounts ofGSLs, which are also major components of cerebellummembranes . They also contained several other GPI-anchored molecules, part of the transmembrane adhe-sion molecule L1, and intracellular proteins, includingthe fyn and src tyrosine protein kinases . Immunopre-cipitation experiments indicated that F3, but not otherGPI-linked molecules, such as NCAM 120 and Thy-1, could be coisolated with Ll and fyn from thesemicrodomains . If such associations preexisted in thecerebellum before the Tx-100 treatment, it might repre-sent a functional complex active in mediating signalsF3 received at the cell surface .Are low-density Tx-100-insoluble complexescaveolae?

Clues to the possible organization of GPI-linked pro-teins into membrane come from studies on lympho-cytes, endothelial, smooth muscle, and polarized epi-thelial cells . Several studies appear to indicate thatthese proteins segregate in 50-100-nm membrane mi-crodomains called caveolae . Caveolae are morphologi-cally defined as a subcompartment of the plasma mem-brane ; they appear as spherical structures with a stri-ated cytoplasmic coat consisting of concentric rings ofgranular subunits linked together as strands (Anderson,1993a,ó ; Rothberg et al ., 1992) . GPI-linked proteinsbut also glycolipids, cholesterol, and caveolin cluster inthese microinvaginations (Rothberg et al ., 1990, 1992)and are resistant to Tx-100 extraction . When appliedto MDCK cells (Sargiacomo et al ., 1993) and lungtissue (Lisanti et al ., 1994), isolated microdomains fitcriteria applied to defined caveolae . Although their sizeis somewhat larger, the composition of Tx-100-insolu-ble complexes we isolated from adult cerebellum islargely similar to that reported for MDCK microdo-mains (Sargiacomo et al ., 1993), as they contain GPI-linked proteins, glycolipids, and signaling molecules .They differ in one major point, however : the absenceof caveolin in our fractions . Indeed, caveolin was ab-sent from all the fractions of the gradient as well asfrom the cerebellum homogenate and other areas ofthe brain examined . Our study therefore raises the pos-sibility that in the brain, caveolin function is accom-plished by a similar but immunologically unrelatedmolecule .

Do the two F3-containing fractions representdifferent cell subcompartments?

In polarized epithelial cells, GPI-linked proteins areselectively transported to the apical surface and ex-cluded from basolateral domains, with GPI acting asa dominant signal for apical transport (Lisanti et al .,1989) . This mechanism could also operate in granulecell neurons from the cerebellum if a parallel is made


S. OLIVE ET AL.

between axonal compartment and apical surface (Dottiet al ., 1991) . In these neurons, F3 is in fact expressedexclusively on their unmyelinated axons and at thesynapses they have established (Faivre-Sarrailh et al .,1992) . In epithelial cells, apical recognition of the GPIoccurs intracellularly via the trans-Golgi network (LeBivic et al ., 1990), and GPI-linked proteins are sortedvia coclustering with glycolipids . In support of thismodel, GPI-linked proteins become resistant to Tx-100 solubilization during apical transport (Brown andRose, 1992) . In contrast to the data reported fromother types of cells or tissues to which this separationtechnique has been applied, we constantly observedpartitioning of cerebellar GPI-linked proteins in twofractions discernible by their buoyancy on sucrose gra-dient. Vesicles in these two fractions were indistin-guishable by their size or their composition in majorGSLs. The only apparent difference was the quantityof myelin sheaths they contained . We believe, how-ever, that colocalization of myelin sheaths and F3-containing vesicles in the lighter fractions is fortuitousbecause F3-expressing parallel fibers are nonmyelin-ated ; moreover, Ll is not found on myelinated axonsin cerebellum (Persohn and Schachner, 1987) . We donot exclude the possibility that this separation is anartifact resulting from properties of proteins and lipidsin Tx-100 . There were, however, some observationsthat were in favor of the physiological existence of twoGPI-linked protein-containing compartments . First, thesensitivity of F3 to PI-PLC treatment was different forthe two fractions, with the upper fraction being par-tially resistant to cleavage even when the treatmentwas repeated or done before separation . Second, theLl protein was found essentially associated with theupper fraction . Hence, it is difficult to decide at presentwhether these two fractions represented microdomainsresulting from different subpopulations of axons, forexample . Further additional experiments aimed at ana-lyzing the composition of these fractions are requiredto determine their origin .

Biological significance of F3/L1 associationPart of the transmembrane molecule L1 was always

detected in the Tx-100 microdomains . It is interestingthat the transmembrane isoforms of NCAM, for exam-ple, are never found in the Tx-100-insoluble fractions .We also noticed some variations from one experimentto another in the ratio of L1 Tx-100 soluble/microdo-main associated . This might reflect the fact that thelocalization in these domains is a regulated affiliation,possibly ligand induced . We hypothesized that thepresence of L1 in the Tx-100-insoluble fraction wasthe result of its interactions with other GPI-linked mol-ecules of the Ig superfamily present in this fraction .Analyses of L1 or F3 immunoprecipitates providedevidence for a physical interaction between these twomolecules .

Several data of the literature support direct interac-tions and functional links between proteins of the Ig


superfamily . They concern Ig-like molecules expressedby cells of the immune system (for review, see Dustinand Springer, 1991 ; Janeway, 1992) but also axonalmembers of this family . This is supported by the find-ing that antibodies to individual Ig/FNIII-like mole-cules interfere with the bundling of retinal axons andwith the outgrowth of sympathetic axons on other ax-ons to a similar extent, suggesting that these proteinsmight be functionally linked (Chang et al ., 1987) .Consistent with this assumption is the finding that thedistribution of these proteins in vivo is overlapping .This is also the case for F3 (Faivre-Sarrailh et al .,1992) and Ll in cerebellum (Persohn and Schachner,1987), both present in parallel fibers . Furthermore,Kuhn et al . (1991) found that Ng-CAM (the putativechick homologue of L1) and axonin-l (the chick ho-mologue of TAG-1) undergo a specific interaction tosupport and regulate axonal growth . Binding studiesusing these purified molecules have confirmed a directinteraction between the two proteins . Particularly rele-vant to our present data is the finding by Brümmendorfet al . (1993), using binding and neurite outgrowthassays, of an interaction between Ng-CAM and F11 .In that study, binding of wild-type Fl l blocked bydomain-specific antibodies and Fl 1 mutants expressedin COS cells revealed that the interaction between F11and Ng-CAM resides in the first two Ig-like domainsof F11 . Our data, obtained using a completely differentapproach, support this observation and indicate thatsuch an interaction might also be functional in vivo .We have not been able to decide, however, whetherthe binding between F3 and Ll occurs directly or viaintermediate proteins . The NHZ-terminal domains ofLl are apparently involved, because in our experimentsonly the 140-kDa C-terminal truncated form of L1(Faissner et al ., 1985 ; Sadoul et al ., 1988) was recov-ered in the immunoprecipitates .

It is worth mentioning here that the data from ourimmunoprecipitation studies differ somewhat fromthose of Cinek and Horejsi (1992) on lymphocytes .These authors reported that preclearing of their deter-gent lysates with different antibodies directed to GPI-linked proteins or lipids precipitated all the compo-nents of the Tx-100-insoluble fraction jointly in a sin-gle type of complex . In our case, immunoprecipitatesobtained with anti-NCAM, for example, did not con-tain Ll or F3 . The explanation for these differencescould be methodological . Our conditions for solubiliz-ing microdomains (Nonidet P-40 and deoxycholate)and washing the immunoprecipitates were more strin-gent than theirs . In our case only interactions resistantto these detergents could be detected. In addition, it islikely that F3-L1 complexes segregate in vesicles thatare physically different from those containing NCAM120, which is known to be mainly expressed byastrocytes and not by axons .Is the complex F3-LI fyn kinase functionallyrelevant?To our knowledge, we have demonstrated for the

first time the coprecipitation of a protein tyrosine ki-

2315

nase of the src family with GPI-linked proteins ex-pressed in neurons . This observation gives support tothe idea that association of GPI proteins with tyrosineprotein kinases is rather general (Brown, 1993) . Forexample, the Thy-1 molecule had been found associ-ated with fyn in murine T-cell hybridoma and in mu-rine thymocytes (Thomas and Samelson, 1992 ; seealso Lisanti et al ., 1994) . The present data are sup-ported by independent experiments on F3-transfectedChinese hamster ovary cells . We have shown that inthese cells, F3 also coprecipitated with the fyn kinaseand that F3 cross-linking by antibodies induced tyro-sine phosphorylation of several intracellular proteins(S . Gomez, personal communication) .

Considering their cellular localization, F3 and fynkinase cannot interact directly, and the existence of alinker molecule should be considered. L1 would havebeen a likely candidate because it spans the membrane,but in the immunoprecipitates only the 140-kDa form,without the intracellular part, was recovered. Thismakes unlikely that L1 interacts directly with fyn ki-nase . Other hitherto undiscovered candidates shouldexist in the immunoprecipitated complex . It is interest-ing that it had been previously observed that the L1molecule immunoisolated from brain extracts copuri-fied with at least two kinase activities (Sadoul et al .,1989), although it is not known whether one of themcould represent the fyn kinase .The observation of the existence of a complex

among F3, L1, and fyn kinase also supports a func-tional interaction between these three molecules . Fyn isexpressed in brain and involved in neuronal functionsbecause the fyn knock-out mouse shows subtle neuro-logical defects such as impaired long-term potentiationand abnormal hippocampal development (Grant et al .,1992) . Also relevant to our observation are the recentfindings that pp60-src (Ignelzi et al ., 1994) and fyn(Beggs et al ., 1994) are components of the intracellularsignaling pathway in L1- and NCAM-mediated axonalgrowth, respectively . Further experiments are neededto elucidate the molecular components of the complexmaking up the two recognition molecules L1 and F3and the fyn kinase and to determine the physiologicaland functional relevance of our observations .

Acknowledgment : We thank Dr. Catherine Faivre-Sar-railh for help with electron microscopy and for useful sug-gestions, Sandrine Martin for technical help with lipid analy-sis, and Drs . Hai-Tao Hé, André Le Bivic, and Christo Gor-idis for comments on the manuscript . This work wassupported by institutional grants from the CNRS and theUniversity of Aix-Marseille 11, by grants from INSERM(CRE 92012) to G.R . and the Association Française contreles Myopathies to G.R . and C.D ., and by an EEC ConcertedAction (PL 92 0012) to M.S . and G.R .

REFERENCESAnderson R . G . (1993x) Plasmalemmal caveolae and GPI-anchored

membrane proteins . Curr. Opin. Neurobiol . 5, 647-652.

J. Neurochern ., Vol. 65, No . 5, 1995

2316

Anderson R . G . (1993h) Caveolae-where incoming and outgoingmessengers meet. Proc . Nall. Acad. Sci. USA 90, 10909-10913 .

Appel F ., Holm J ., Conscience J . F ., and Schachner M . (1993)Several extracellular domains of the neural cell adhesion mole-cule L1 are involved in neurite outgrowth and cell body adhe-sion . J. Neurosci . 13, 4767-4775 .

Atashi J ., Klinz S ., Ingraham C ., Matten W., Schachner M ., andManess P . (1992) Neural cell adhesion molecules modulatetyrosine phosphorylation of tubulin in nerve growth cone mem-branes . Neuron 8, 831-842 .

Beggs H ., Soriano .P., and Maness P . (1994) NCAM-dependentneurite outgrowth is inhibited in neurons from fyn-minus mice .J. Cell Biol. 127, 825-833 .

Brown D . (1993) The tyrosine kinase connection : how GPI-an-chored proteins activate T cells . Curr. Opin. Immunol . 5, 349-354 .

Brown D . and Rose K . (1992) Sorting of GPI-anchored proteins toglycolipid-enriched membrane subdomains during transport tothe apical cell surface . Cell 68, 533-544 .

Brümmendorf T . and Rathjen F . G . (1993) Axonal glycoproteinswith immunoglobulin- and fibronectin type III-related domainsin vertebrates : structural features, binding activities, and signaltransduction . J. Neurochem . 61, 1207-1219 .

Brümmendorf T ., WolffJ ., Franck R ., and Rathjen F . (1989) Neuralcell recognition molecule Fl 1 . Homology with fibronectin typeIII and immunoglobulin type C domains . Neuron 2, 1351-1361 .

Brümmendorf T ., Hubert M., Treubert U ., Leuschner R ., Tarnok A .,and Rathjen F . (1993) The axonal recognition molecule F1 I isa multifunctional protein : specific domains mediate interactionswith NgCAM and restrictin . Neuron 10, 711 -727 .

Chang S ., Rathjen F . G ., and Raper J . A . (1987) Extension ofneurites on axons is impaired by antibodies against specificneural cell surface glycoproteins . ,I. Cell Biol . 104, 355-362 .

Chang W.-J ., Ying Y.-S ., Rothberg K . G ., Hooper N . M ., TurnerA . J ., Gambliel H . A ., De Gunzburg J ., Mumby S . M ., GilmanA . G ., and Anderson R . G . W . (1994) Purification and charac-terization of smooth muscle cell caveolae. J. Cell Biol. 126,127-138 .

Cinek T . and Horejsí V . (1992) The nature of large non covalentcomplexes containing glycosylphosphatidyl inositol membraneanchored glycoproteins and protein tyrosine kinases . J. Immu-nol. 149, 2262-2270 .

Dodd J . and Jessell T . (1988) Axon guidance and patterning ofneuronal projections in vertebrates . Science 242, 692-699 .

Doherty P ., Barton C ., Dickson G ., Seaton P ., Rowett L ., Moore S .,Gower H ., and Walsh F. (1989) Neuronal process outgrowthof human sensory neurons on monolayers of cells transfectedwith cDNAs for five NCAM isoforms . J. Cell Biol. 109, 789-798 .

Dotti C., Parton G . R., and Simons K . (1991) Polarized sorting ofglypiated proteins in hippocampal neurons . Nature 349, 158-161 .

Dubois C ., Manuguerra J ., Hanttecoeur C . B ., and Mazy J . (1990)Monoclonal A2B5 which detects cell surface antigens, binds toGT3 (113(NeuAc)3 Lac Cer) and to its 9-O acetylated deriva-tives. J. Biol. Chem. 265, 2797-2803 .

Durbec P ., Gennarini G ., Goridis C ., and Rougon G. ( 1992) Asoluble form of the F3 neuronal cell adhesion molecule pro-motes neurite outgrowth . J. Cell Biol. 117, 877-887 .

Durbec P ., Gennarini G ., Buttiglione M ., Gomez S ., and Rougon G .(1994) Different domains of the F3 neuronal adhesion moleculeare involved in adhesion and neurite outgrowth promotion . Eur.J. Neurosci . 6, 461-472.

Dustin M . and Springer T. (1991) Role of lymphocyte adhesionreceptors in transient interactions and cell locomotion . Anna.Rev. Immunol . 9, 27-66 .

Edidin M . (1990) Molecular associations and membrane domains .Curr. Top . Membr. Trans. 36, 81-96 .

Faissner A., Teplow B ., Kubler D ., Keilhauer G ., Kinzel V ., andSchachner M . (1985) Biosynthesis and membrane topographyof the neural cell adhesion molecule L1 . EMBO J . 4, 3105-3113 .


S. OLIVE ET AL.

Faivre-Sarrailh C ., Gennarini G ., Goridis C ., and Rougon G . (1992)F3/F11 cell surface molecule expression in the developingmouse cerebellum is polarized at synaptic sites and within gran-ule cells . J. Neurosci. 12, 257-267 .

Ferguson M . ( 1992) Glycosylphosphatidyl inositol anchors : the taleof a tail . Biochem. Soc. Trans. 20, 243-256 .

Furley A ., Morton J . S ., Manalo B . D ., Karagogeos D ., Dodd J .,and Jessell T . ( 1990) The axonal glycoprotein TAGI is animmunoglobulin superfamily member with neurite outgrowthpromoting activity . Cell 61, 157-170 .

Gennarini G ., Cibelli G ., Rougon G ., Mattei M ., and Goridis C .(1989a) The mouse neuronal cell surface protein F3 : a phospha-tidylinositol-anchored member of the immunoglobulin superfamily related to the chicken contactin . J. Cell Biol. 109, 775-788 .

Gennarini G ., Rougon G ., Vitiello F ., Corsi P ., Di Beneditta C ., andGoridis C . (1989b) Identification and cDNA cloning of a newmember of the L2/HNK-1 family of neural surface glycopro-teins . J. Neurosci. Res . 22, 1-12 .

Gennarini G ., Durbec P., Boned A ., Rougon G ., and Goridis C .(1991) Transfected F3/F11 neuronal cell surface protein medi-ates intercellular adhesion and promotes neurite outgrowth .Neuron 6, 595-606 .

Glenney J . and Soppet D . (1992) Sequence and expression ofcaveo-lin, a protein component of caveolae plasma membrane domainsphosphorylated on tyrosine in RSV transformed fibroblasts .Proc. Nad. Acad. Sci. USA 89, 10517-10521 .

Goodman C . (1994) The likeness of being : phylogenetically con-served molecular mechanisms of growth cone guidance . Cell78, 353-356 .

Grant S ., O'Dell T ., Karl K ., Stein P., Soriano P., and Kandel E .(1992) Impaired long term potentiation, spatial learning, andhippocampal development in fyn mutant mice . Science 258,1903-1910 .

Hsi E ., Siegel D . J ., Minami Y ., Luong E ., Klausner R ., and Same]-son L . (1989) T cell activation induces rapid tyrosine phosphor-ylation ofa limited number of cellular substrates . .I. Biol . Chern.264, 10836-10842 .

Hynes R . and Lander A . ( 1992) Contact and adhesive specificitiesin the associations, migrations, and targeting of cells and axons .Cell 68, 303-322.

Ignelzi M ., Miller D ., Soriano P ., and Maness P . (1994) Impairedneurite outgrowth of src-minus cerebellar neurons on the celladhesion molecule Lt . Neuron 12, 873-884 .

Irwin C. C . and Irwin L . (1979) A simple rapid method for ganglio-side isolation from small amounts of tissue . Anal . Biochem. 94,335-339 .

Janeway C . ( 1992) The T cell receptor as a multicomponent signal-ing machine : CD4/CD8 coreceptor and CD45 in T cell activa-tion . Anna . Rev . Immunol. 10, 645-674 .

Kuhn T ., Soeckli E ., Condrau M ., Rathjen F., and Sonderegger P .(1991) Neurite outgrowth on immobilized axonin-1 is mediatedby a heterophilic interaction with Lt . J. Cell Biol . 115, 1113-1126 .

Le Bivic A ., Quaroni A ., Nichols B ., and Rodriguez-Boulan E .(1990) Biogenic pathways of plasma membrane proteins inCaco2, a human intestinal cell line . .I. Cell Biol. 111, 1351-1361 .

Lemmon U ., Farr K ., and Lagenaur T . (1989) L1 mediated axonoutgrowth occurs via a hemophilic binding mechanism . Neuron2, 1597-1603 .

Lisanti M ., Caras I ., Davitz M ., and Rodriguez-Boulan E . (1989)A glycolipid membrane anchor acts as an apical targetting signalin polarized epithelial cells . J. Cell Biol. 109, 2145-2156.

Lisanti M ., Scherer P ., Vidugiriene J ., Tang Z., Hermanowski-Vo-satka A ., Tu Y ., Cook F ., and Sargiacomo M . ( 1994) Character-ization of caveolin-rich membrane domains isolated from anendothelial-rich source : implication for human disease . J . CellBiol . 126, 111-126 .

Low M. G . (1989) The glycosyl-phosphatidylinositol anchor ofmembrane proteins . Biochem . Biophys. Acta 988, 427-454 .


Maness P. F. (1992) Non-receptor protein tyrosine kinases associ-ated with neuronal development. Dev. Neurosci . 14, 257-270.

Mayor S., Rothberg K., and Maxfield F. (1994) Sequestration ofGPI anchored proteins in caveolae triggered by cross-linking .Science 264, 1948-1951 .

Nédelec J ., Pierres M., Moreau H., Barbet J., Naquet P., Faivre-Sarrailh C., and Rougon G. (1992) Isolation and characteriza-tion of a novel glycosyl-phosphatidylinositol-anchored glycoconjugate expressed by developing neurons. Eur. J. Biochem.203, 433-442.

Persohn E. and Schachner M. (1987) Immunoelectron microscopiclocalization of the neural cell adhesion molecules L1 andNCAM during postnatal development of the mouse cerebellum.J. Cell Biol. 105, 569-576.

Pesheva P., Gennarini G., Goridis C., and SchachnerM. (1993) TheF3/11 cell adhesion molecule mediates the repulsion of neuronsby the extracellular matrix glycoprotein JI-160/180 . Neuron10, 69-82.

Rathjen F. (1991) Neural cell contact and axonal growth . Carr.Opin. Cell. Biol. 3, 992-1000 .

Rathjen F. and Schachner M. (1984) Immunocytological and bio-chemical characterization ofa new neuronal cell surface compo-nent (LI antigen) which is involved in cell adhesion . EMBOJ. 3, I-10 .

Rothberg K., Ying Y., Kamen B., and Anderson R. G. (1990) Cho-lesterol controls the clustering of the glycosyl-phosphatidylino-sitol membrane receptor for the 5-methyltetrahydrofolate. J.Cell Biol . 111, 2931-2938.

Rothberg K., HenserJ., DonzellW., Ying Y., Glenney J., and Ander-son R. G. (1992) Caveolin, a protein component of caveolaemembrane coats. Cell 68, 673-682.

Rutishauser U. and Jessell T. (1988) Cell adhesion molecules invertebrate neural development. Physiol. Rev. 68, 819-857.

Sadoul K., Sadoul R., Faissner A., and Schachner M. (1988) Bio-chemical characterization of different molecular forms of theneural cell adhesion molecule LI . J. Neurochem. 50, 510-521 .

Sadoul R., Kirchhoff F., and SchachnerM. (1989) A protein kinaseactivity is associated with and specifically phosphorylates the

231 7

neural cell adhesion molecule L1 . J. Neurochem. 53, 1471-1478 .

Sargiacomo M., Sudol M., Tang Z.-L., and Lisanti M. (1993) Signaltransducing molecules and glycosyl-phosphatidylinositol-linkedproteins form a caveolin-rich insoluble complex in MDCK cells.J. Cell Biol. 122, 789-807.

Schachner M. (1991) Neural recognition molecules and their influ-ence on cellular functions, in The Nerve Growth Cone (Letour-neau P. C., Kater S. B., and Macagno E. R., eds), pp . 237-254. Raven Press, New York .

Schachner M. (1994) Neural recognition molecules in disease andregeneration . Curr. Opin . Neurobiol. 4, 726-734.

Sonnino S., Ghidoni R� Masserini M., and Tettamanti G. (1983)Recognition by two dimensional thin layer chromatography anddensitometric quantification of alkalilabile gangliosides fromthe brain of different animals. Anal. Biochem. 128, 104-114.

Stefanova L, Horejsi V., Ansotegui L, Knapp J. W., and StockingerH. (1991) GPI-anchored cell-surface molecules complexed toprotein tyrosine kinases. Science 254, 1016-1019.

Th6veniau M., Malapert P., and Rougon G. (1990) Antibody againstB. thurengiensis phosphatidylinositol phospholipase C. Someexamples of its potential uses . Immunol. Cell. Biol . 68, 87-93.

Thomas J., Dwek R., and Rademacher T. (1990) Structure, biosyn-thesis and function of glycosylphosphatidylinositols. Biochem-istry 29, 5413-5422.

Thomas P. and Samelson L. (1992) The glycosylphosphatidyl-inosi-tol anchored Thy-1 molecule interacts with the p60 fyn proteintyrosine kinase in T cells. J. Biol . Chem . 267, 12317-12322.

Yoshihara Y., Kawasaki M., Tani A., Tamada A., Nagata S., Kaga-miyama H., and Mori K. (1994) BIG-1 : a new TAG-1/173related member of the Ig superfamily with neurite outgrowthpromoting activity . Neuron 13, 415-426.

Zuellig R., Rader C., Schroeder A., Kalousek M., Von Bohlen F.,Osterwalder T., Inock C., Stoeckli E., Affolter U., Fritz A.,Hafen E., and Sonderegger P. (1992) The axonally secretedcell adhesion molecule, axoninl : primary structure, immuno-globulin-like and FNIII-like domains and glycosylphosphatidyl-inositol anchorage. Ear. J. Biochem. 204, 453-463 .


the f3 neuronal glycosylphosphatidylinositol-linked molecule is localized to glycolipid-enriched...

Documents