Development 106, 483-491 (1989)Printed in Great Britain © The Company of Biologists Limited 1989

483

Influence of extracellular matrix proteins on the expression of neuronal

polarity

B. CHAMAK and A. PROCHIANTZ

1NSERM V114, Chaire de Neuropharmacologie, College de France, 11, place Marcelin Berthelot 75231 Pans Cedex 05 France

Summary

The influence of laminin (LN) and fibronectin (FN) onthe differentiation of individual neurones from theembryonic rat central nervous system was studied invitro. In control cultures or in the presence of solubleFN, most neurones had several dendrite-like and oneaxon-like processes. On substratum-bound LN, multi-polar and unipolar cells were present. Soluble LN andbound FN induced a very simple neuronal morphology,most neurones having only one axon-like neurite asdefined by morphological and immunocytochemicalcharacteristics. The significant reduction of neuronal

adhesion and spreading in conditions leading to thegrowth inhibition of dendrite-like processes suggeststhat, contrary to that of axons, dendrite growth stronglydepends on neuronal adhesion. We propose a model inwhich the different dependency of axonal and dendriticoutgrowth towards adhesion and spreading is explainedby the respective physical properties of the two types ofneurites.

Key words: matrix, adhesion, neurone, axon, dendrite,polarity.

Introduction

Neuronal process outgrowth in vitro is regulated bynumerous chemical and physical factors. Among thechemical factors, NGF and basic FGF are known tostimulate the elongation of neurites in specific periph-eral and central neuronal populations (Levi-Montalcini,1982; Korshing et al. 1985; Morrisson et al. 1986;Walicke et al. 1986; Unsicker et al. 1987). The import-ance of physical parameters, such as tension andadhesion, has also been emphasized (Letourneau, 1975;Fraser, 1980; Bray, 1984). In addition, the cellularsubstrate and the extracellular matrix (ECM) proteinsinfluence neuronal process outgrowth. The role of theseproteins has been studied mainly in the peripheralnervous system (PNS) where they have been shown topromote the migration of neural crest derivatives, theoutgrowth and guidance of neurites and the inductionof proteins involved in neurotransmitter metabolism(Baron-van Evercooren et al. 1982; Thiery et al. 1982;Edgar et al. 1984; Smalheiser et al. 1984; Davis et al.1985; Lander et al. 1985; Hammarback et al. 1985;Acheson et al. 1986; Millaruelo et al. 1988). In thecentral nervous system (CNS), ECM proteins and inparticular laminin (LN) and fibronectin (FN) are likelyto play a role in neuronal differentiation and regener-ation and in the process of neurite elongation (Man-thorpe et al. 1983; Smalheiser et al. 1984; Faivre-Bauman et al. 1984; Carri et al. 1988). LN and FN aresynthesized in the CNS during development and after

brain injury as well as by astrocytes in culture (Liesi etal. 1983, 1984; Price & Hynes, 1985). Further, LN andFN receptors are present on neurones in culture(Bozyczko & Horwitz, 1986).

In a previous series of studies, we examined theregulation of neuronal process outgrowth by CNSneurones in vitro from a different point of view. Wedemonstrated that dendritic and axonal growth areseparate events under distinct regulatory controls.Using short-term in vitro cocultures of neurones andastrocytes from two regions of the brain (mesencepha-lon and striatum), we observed vigorous dendritegrowth in neuroastroglial cocultures of neurones andastrocytes derived from the same region (homotopiccocultures). In comparison, neurones grown on hetero-topic astrocytes had a poorly developed dendritic treeand exhibited a long and thin axon-like process (Denis-Doniniefa/. 1984; Chamak etal. 1987; Autillo-Touati etal. 1988). This finding that regulation of dendritic andaxonal growth is differentiable has been confirmed byexperiments with other neuronal populations (Bruck-enstein & Higgins, 1988a). Furthermore, Bruckenstein& Higgins (1988b) have demonstrated that a purifiedfraction of serum stimulated dendritic outgrowth bysympathetic neurones in culture.

In a recent study, we have started to characterize thefactors regulating dendrite outgrowth in rat mesen-cephalic neurones. We found that molecules able tostimulate or inhibit the growth of dendrites werepresent in astrocyte-conditioned media (Rousselet et al.

484 B. Chamak and A. Prochiantz

1988). Preliminary analysis of these media revealed thatthey contain large amounts of different isoforms of LNand FN (Rousselet et al. in preparation).

Therefore, in order to set up a model system we havestarted to analyse the effects of LN and FN on theexpression of neuronal polarity in CNS neurone cul-tures. In this report, we demonstrate that LN and FNdo regulate dendritic and axonal extension but that thephysical form (soluble or sustrate bound) of thesemolecules is a critical factor in the differential growth ofthe two neuronal compartments.

Materials and methods

Cell culturesDissociated mesencephalic neurones were prepared from 14-day-old rat embryos (IFFA CREDO, France), plated at lowdensity (104 cells cm" ) on polyornithine (1-5/igmF1, Sigma)-coated culture dishes (16 mm diameter wells) and culturedfor 2 days in serum-free medium (SFM) consisting ofDMEM/F12 (1:1, Gibco), 33mM-D-glucose, 2mM-L-gluta-mine, 3mM-NaHCO3, 5-mM-Hepes pH7-4, 5i.u.ml~' peni-cillin and 5|<gml~ streptomycin. This SFM was sup-plemented with 25/igml insulin, lOO^gml"1 transferrin,2xlO"8M-progesterone, 6xlO~5M-putrescine, 3X10"8M-selenium and 0-1% ovalbumin (chemically defined medium,CDM). In some cases, dissociated brain stem and striatalneurones were prepared from 14- and 16-day-old rat embryos,respectively.

Laminin (Bethesda Research Laboratories or provided byDr N. Brunet de Carvalho) was isolated from an Engelbreth-Holm-Swarm (EHS) sarcoma, a transplantable mouse tu-mour. Fibronectin (Collaborative Research Incorporated)was obtained from the fibrinogen fraction of human plasma.In order to obtain the concentrations of bound LN (bLN) andbound FN (bFN) reported in Table 1, culture dishes wereincubated overnight at 37°C with 50 jig ml"1 (bFN) or1 jigml"1 (bLN) of the matrix molecules. They were washedthree times with serum-free medium before neuronal plating.Iodination of the matrix molecules was achieved with Iodogen(Pierce), according to the manufacturers' instructions. Sol-uble fibronectin or laminin diluted in SFM at the indicatedconcentrations were added to the medium 1 h after neuronalplating on polyornithine-coated (1-5/igmr1) culture dishes.48 h later (or more when indicated), cultures were fixed for 1 hat 4°C with 2-5% glutaraldehyde in phosphate-bufferedsaline (PBS), washed three times with PBS and incubated for5 to lOmin with 0-5% toluidine blue in 2-5% Na2CO3.Stained cells were washed with distilled water and air dried.For reversion experiments, the cells were detached after 24 h

Table 1. Amounts of soluble and bound matrixmolecules in the different culture conditions

sFN bFN sLN bLN

Soluble (jig/well)Bound (/ig/well)

Bound/soluble

3-50-8

0-23

0081-2

15

400-6

0-15

0020-5

25

Laminin and fibronectin were mixed with tracer amounts of 125I-iodinated molecules. After 48 h, the amounts of radioactivitypresent in the supematants and attached to the substratum weredetermined. Similar results were obtained in the presence or in theabsence of neurones.

by a brief (5 min) trypsin treatment (0-25 % in SFM) at roomtemperature and replated as indicated in the Results section.

ImmunocytochemistryFor immunocytochemistry cells were grown on 16 mm diam-eter polyornithine-coated glass coverslips, fixed for 1 h at 4°Cwith freshly prepared 4% paraformaldehyde in PBS, andwashed three times with Tris-buffered saline (TBS) contain-ing 10 % fetal calf serum and 01 % Triton X 100 (buffer A).Incubations with the first antibodies were for 1 h at 37 °C, thecells were washed three times with buffer A, further incu-bated for lh at 37°C with fluorescein- or rhodamine-linkedsecond antibodies (Byosis) diluted 100-fold in buffer A,washed three times and mounted in PBS:glycerol (5:95) formicroscopic observation. The polyclonal anti-MAP2 antibodywas provided by Dr A. Fellous (Chamak et al. 1987) and thepolyclonal anti Glial Fibrillary Acidic Protein (GFAP) werepurchased from Dakopatts. The axon-specific monoclonalantibodies recognizing phosphorylated isoforms of the 160and 200 neurofilament proteins (Foster et al. 1987) werecharacterized and generously provided by Dr P. Levitt or byDr R. Liem.

Adhesion assaysCells were seeded on plastic culture dishes, with or withoutpolyornithine. After 2 days in culture, the plates were shakenat 300 revs min"1 for 1 h. Attached cells were fixed with 2-5 %glutaraldehyde in PBS, washed, stained with toluidine blueand counted in 10 fields chosen at random.

Analysis of neuronal morphologyThe number of cells was determined by counting 50 indepen-dent fields. Cell survival was calculated as a fraction of thenumber of cells extruding trypan blue 2h after plating. Inorder to analyse neuronal morphology, 50 to 100 neuronestaken at random were drawn with the help of a drawing tube.In order not to introduce a bias in the choice of the neurones,we proceeded to a double blind analysis, starting at a givenpoint of the culture dish and drawing all isolated neuronesencountered, field by field, until the desired number of cellswas obtained. In the culture conditions used for this study,most neurones were clearly isolated and only a few aggregateshad to be eliminated from the statistics. For each neurone, thenumber of primary neurites and branching points wascounted. From the drawings, neurite length and cell bodysurfaces were calculated using the Jandel Scientific SigmaScancomputer program. For each parameter studied (neuriticlength, number of primary neurites, number of branchpoints), the neurones were separated into several classes andthe percentage distributions were calculated. In some cases,extreme classes had to be pooled in order to obtain theminimum number of neurones per class necessary for thevalidity of the statistical analysis. Differences between distri-butions were analysed with a test of homogeneity (Chi-square).

Results

Cellular survival and characterizationMesencephalic cells were plated at low density and 2 hlater non-attached cells were washed away. A 40%plating efficiency (number of cells extruding trypanblue) was calculated. After 2 days, the number of liveattached cells had decreased by 40% in the controlmedium or with soluble extracellular matrix proteins

Separate regulation ofaxonal and dendritic outgrowth 485

Table 2. Percentage of neurite-bearing cells in differentculture conditions

Control sFN bFN sLN bLN

Cells with neurites% (±S.E.D.)

35 ±7 31 + 8 71 + 8 36±3 75 ±4

Each value represents the mean of 3 independent experiments.

and by 25 % when bound LN or FN were used.Immunostaining of the cells after 2 days in vitrodemonstrated that more than 98% of them could berecognized by the neurone-specific anti-neurofilamenttriplet antiserum, the remaining 2 % being stained bythe anti-GFAP antibody. This clearly demonstrated theneuronal nature of the majority of the cells present atthe time of morphological analysis.

The amounts (in jig/well) of soluble fibronectin(sFN), bound fibronectin (bFN), soluble laminin (sLN)and bound laminin (bLN), were calculated with thehelp of 125I-iodinated matrix molecules. They are pre-sented in Table 1. This table also gives the amounts ofmatrix molecules which attach to, or detach from, thesubstratum during the time of culture, as well as theratios between bound and soluble forms in the differentculture conditions.

Effects of matrix molecules on neuronal morphologyThe presence of matrix proteins had a strong influenceon the number of cells bearing neurites longer than twocell diameters. As shown in Table 2, compared tocontrol medium, bFN and bLN increased the number ofcells with processes whereas sFN and sLN had no effect.Fig. 1 illustrates the typical morphologies of the mesen-cephalic neurones in the different culture conditions. Incontrol medium or in sFN, the cell bodies had severalcytoplasmic extensions (Fig. 1A). On bFN or with sLNmost of the neurones were unipolar (Fig. IB and 1C)whereas on bLN both multipolar and unipolar neuroneswere observed (Fig. ID). The presence of these twomorphologies on bLN was probably due to the solubil-ization of 4 % of bound laminin (Table 1). In factadding sLN to bLN increased dramatically the percent-age of unipolar neurones (not shown). Similar resultswere obtained with neuronal populations from thestriatum and the hindbrain (not shown).

Quantitative morphological analysis was done on allneurite-bearing cells. A typical experiment in which theeffects of soluble and bound ECM molecules werecompared is illustrated in Fig. 2. The large majority ofneurones in control medium and in the presence of sFNwere multipolar. In contrast with sFN, which permittedthe extension of several neurites (Fig. 2A), bFNstrongly reduced the number of multipolar neurones(Fig. 2B). The same unipolar morphology was alsoobserved in the presence of sLN (Fig. 2C). In thepresence of bLN, both unipolar and multipolarneurones were present (Fig. 2D).

In order to verify that the unipolar neurones obtainedwith sLN did not reflect the selective loss of multipolarcells, reversion experiments were performed. The re-

* 1• /

1A 20Mm

'%<• i.

fFig. 1. Typical morphologies of mesencephalic neuronescultured for 48h in control medium (A), on bound FN (B),with soluble LN (C) and on bound LN (D). Themorphology of neurones in presence of soluble FN wassimilar to that of the control culture (not shown).

suit of such an experiment is shown in Fig. 3. Mesen-cephalic neurones were cultured 48 h either in controlconditions (Fig. 3A) or in the presence of sLN(Fig. 3B). Neurones were cultured in sister wells for24 h in sLN where they adopted the simple unipolarphenotype. They were then trypsinized, replated onpolyornithine and cultured for another 24 h. The multi-polar morphology of the cells shown in Fig. 3C demon-strates the complete phenotypic reversal of the unipolarneurones.

In another experiment, mesencephalic neuroneswere cultured for 24 h in control conditions and sLNwas then added for another 24 h. In these conditions,we observed a small population of unipolar and a largepopulation of multipolar neurones (Fig. 3D). Theunipolar neurones were present in control conditionsafter 24h (not shown), and the effect of adding LN wasonly to increase the length of their single neurite.Multipolar neurones differed from those seen in controlmedium alone (after 48 h). They now had an asym-

486 B. Chamak and A. Prochianti

metrical shape, one neurite elongating much faster thanthe others (arrowheads).

From these experiments, we could first conclude thatthe unipolar population seen in sLN (and also in bFN)did not reflect the selective loss of a multipolar popu-lation, since they could be induced to adopt themultipolar phenotype. Second it appeared that amongthe several neurites present in the control condition oneof them started to elongate at a faster rate when sLNwas added to the culture medium. These conclusionsare partially summarized in the scheme presented inFig. 4.

Better quantification and characterization of neur-onal morphologies were obtained from a statisticalanalysis in which we included the following parameter:number of primary neurites, number of branch points,total neuritic length and length of the longest neurite.Fig. 5 shows that sFN had no effect on the number ofprimary neurites and branch points. On the contrary,sLN and bFN reduced the mean number of primaryneurites from 5 to 1 and 1-5, respectively. The numberof branch points was also significantly reduced in thesetwo culture conditions. Laminin adsorbed to the sub-stratum had an intermediate effect easily explained bythe presence of the two neuronal populations present-ing different morphological traits. As indicated in

fc%^£̂

Fig. 2. Drawings of mesencephalic neurones grown for 48 hin different culture conditions. Neurones were chosen atrandom and drawn with the help of a light chamber.(A) Soluble FN; (B) bound FN; (C) soluble LN; (D) boundLN.

X

J1\

Fig. 3. Drawings of mesencephalic neurones cultured 48 hin control medium (A) or in presence of soluble LN (B).(C) Neurones cultured for 24 h in soluble LN, trypsinized,replated on polyornithine in control medium and left foranother 24 h. (D) Neurones cultured for 24 h in controlmedium and supplemented for another 24 h with solubleLN, note the presence of longer neurites (compare with A)indicated by arrows.

Oh 24 h 48h

(20%)

ControlMedium ^ '

sLN

Control

^Mediumi

Control,^,

(100%), Medium

sLN "* "-̂

^

(100%)

(20%)

(80%)

(100%)

< (100%)

Fig. 4. Schematic representation of the results of theexperiments presented in Figs 2 and 3 and of the treatmentsleading to the different morphologies.

Separate regulation of axonal and dendritic outgrowth 487

Effect of matrix molecules on neuronal morphology

Number of primary neurites Number of branch points

U n £ in

Fig. 5. Number of primary neurites and branch points ofneurones cultured in different conditions. 50 neuroneschosen at random were studied in each condition. Thesignificance of the differences in neuronal distributionsbetween the control (no matrix proteins added) and theexperimental conditions were calculated using the Chi-square analysis and are indicated as follows: ••*, P< 0-001;n.s., not significant; bFN, bound fibronectin; sFN, solublefibronectin; bLN, bound laminin; sLN, soluble laminin.

Table 3. Influence of matrix molecules on neuritelength

ControlsFNbFNsLNbLN

SH4L (±S.E.M.)(jmi)

342 ±15342 ±14233 ±12212 ±23301 ±22

L (±S.E.M.)(/*m)

99±6127 ± 27200 ±13210 ±21156 ±15

In each condition, 50 neurones chosen at random were drawnand the total neurite length (ELi + L) of the length of the longestneurite (L) were determined using Jandel Sigma Scan measurementcomputer program.

Table 3, total neuritic length per neurone was identicalin control medium and in the presence of sFN butdecreased in the presence of bLN, bFN and sLN.Decrease in total length was particularly striking withbFN and sLN, conditions favouring the unipolar pheno-type. However, the length of the longest neurite wasgreatly enhanced in these latter conditions (bFN andsLN), some of the neurites reaching 900^m after 48h.The intermediate values found on bLN are explained bythe presence of two populations with different mor-phologies (Fig. 2).

Identification of 'dendrite-like' and 'axon-like' neuritesThe single long neurite observed in bFN and sLN hadthe morphological aspect of an axon suggesting that, in

Fig. 6. Anti-MAP2 staining of mesencephalic neuronesafter 2 days in culture. (A,B) Control medium;(C,D) soluble laminin; (E,F) bound laminin. In B, D and Fthe negatives were overexposed in order to visualize theaxons. Strong staining was limited to cell bodies anddendrites.

these conditions, axonal growth only was permitted. Onthe contrary, several neurites present in the controlcultures and with sFN or bLN had a dendritic mor-phology. In order to confirm these morphologicalcharacterizations, we used immunocytochemicalmarkers preferentially staining the dendritic or axonalcompartments of the neurones. Fig. 6 illustrates theanti-MAP2 staining of the dendrites and cell bodiesafter 2 days in culture. In this figure, the same negativeswere exposed either normally (Fig. 6A, C, E) or wereoverexposed (Fig. 6B, D, F). This later conditionpermitted visualization of some neurites which wereonly faintly immunostained. In control medium alone(Fig. 6A, B) or with sFN (not shown), neurones ex-tended several short neurites stained by the antiserum.In sLN, the single 'axon-like' neurite was stronglystained only in the region proximal to the cell body(Fig. 6C, D). The staining decreased in the most distalpart and its visualization was made possible only byoverexposure. The same patterns of staining wereobtained in sLN and on bFN (not shown). The anti-MAP2 staining of neurones grown on bLN is illustratedin Fig. 6E and F. Several dendrites are strongly labelledand the long putative axon is faintly stained, as clearlydemonstrated in the overexposed condition (Fig. 6F).

488 B. Chamak and A. Prochiantz

Fig. 7. Staining by the axon-speciiic antibody ofmesencephalic neurones cultured for 5 days in differentconditions. (A) Control medium; (B) bound FN;(C) soluble LN; (D) bound LN. Note that in all conditions,only the longest neurite is stained by the antibody.

In order to better characterize the 'axon-like'neurites, the cells were cultured for 2 or 5 days andstained with a monoclonal antibody directed against a200xlO3 MT phosphorylated neurofilament isoformspecifically expressed in the immature axons (Liem,personal communication). Identical results were ob-tained with a monoclonal antibody directed againstaxon-specific phosphorylated 160 and 200 x l(r Mrproteins (Levitt etal. in preparation). After 2 days, onlyone neurite in multipolar cells and the single neurite inmonopolar cells were stained by the axon-specificantibodies. This staining pattern was even more strikingafter 5 days (Fig. 7). From this figure, it is also clear thatthe stained neurites bear axonal morphological charac-ters: they are long, unbranched in their proximalsegment and their thin diameter does not taper as thedistance from the cell body increases. In addition, thepearly pattern of the staining indicated that, at thisstage, the phosphorylated isoforms were not uniformlydistributed along the 'axon-like' shaft. It can thereforebe proposed that axon-like and dendrite-like neurites

Table 4. Influence of matrix molecules on the size ofthe cell body surface

Control sFN bFN sLN bLN

Cell body surface(/an2) (±S.E.M.)

274 ±7 172 ± 5 155 ±5 141 ± 4 182 ±5

In each condition, 100 neurones chosen at random were drawnand the surface of each soma was determined using Jandel SigmaScan measurement computer program.

are present in the control conditions and in sFN but thatin sLN and bFN the neuritic arborization is almostentirely restricted to the axon-like compartment.

Cell body surfaces and adhesionThe morphology of the cell bodies in the differentconditions suggested a decreased spreading of the somain conditions in which only axon-like neurites werepresent. In order to verify this point, 100 cell bodies ofprocess-bearing neurones were chosen at random,drawn and their surfaces were calculated using theSigmaScan digitizing program. Results of one of theseexperiments are shown in Table 4. Compared to polyor-nithine (PORN) alone, the presence of bound orsoluble matrix proteins decreased the apparent surfacesof the neurones. However, these surfaces were signifi-cantly reduced in the two conditions leading to thegrowth of a unique 'axon-like' process.

We hypothesized that the apparent surface of the cellbody was an index of cellular adhesion and spreading.Thus, in order to verify if adhesion was decreased in thetwo conditions inhibiting dendritic growth and favour-ing axonal elongation (sLN and bFN), adhesion assayswere performed. After 2 days in culture, the plates wereshaken for lh at 300 revs min"1, and the number ofattached cells was determined. As shown in Table 5, thenumber of attached cells was significantly decreased insLN and bFN conditions, even in the presence of theadhesive substratum of polyornithine. In the absence ofPORN, the effect of sLN and bFN on cellular adhesionwas even more dramatic, and the few attached cells, ifleft to develop for 48 h, exhibited the expected unipolarphenotype. From these observations, it was concludedthat, compared to axon-like neurites, the growth of thedendrite-like arborization was strongly dependent uponadhesion.

Table 5. Adhesion assayWith PORN Without PORN

Cells cm"Cells/field

±S.E.M. Cells cm"Cells/field

±S.E.M.

ControlsFNbFNsLNbLN

33003200260022003300

27 ±0-726 ±0-921 ±118 ±0-526 ±0-4

31002900

10050

2900

25 ±0-924 ± 0-5

1 ±0-10-4 ±0-124+1

Cells are seeded on plastic dishes with or without polyornithine(PORN). After 2 days in culture, the cells are shaken for 1 h at300revs min"1. They are then fixed, washed, stained by toluidineblue and counted. In each condition, 10 fields were examined.

Separate regulation of axonal and dendritic outgrowth 489

Discussion

In this article, we report the effects of LN and FN onneurite outgrowth and on the establishment of neuronalpolarity. The results obtained with mesencephalicneurones were reproduced with striatal and brain stemneurones (not shown). They therefore present a certaindegree of generality.

LN and FN had different effects depending on theirphysical (soluble or attached) form. These differencesprincipally concerned two parameters: the number ofprocess-bearing neurones and neuronal morphology.Compared to control medium, bFN and bLN increasedthe number of process-bearing neurones. This con-firmed the observation of several authors on the influ-ence of matrix substrata on neuritic extension (Man-thorpe et al. 1983; Smalheiser et al. 1984; Faivre-Bauman et al. 1984; Cam et al. 1988).

However, the precise examination of neurite-bearingcells allowed by the sparse culture conditions led us todemonstrate that, in control medium and sFN, mostneurones were multipolar whereas in sLN and bFN theyadopted a unipolar morphology. On bLN the twomorphological types were present, a result which can beexplained by the partial solubilization of the boundmolecule, as calculated from experiments with U5l-iodinated LN. This was validated by the fact that addingsLN to bLN increased the number of unipolar neurones(not shown). The latter experiment also confirmed thetheory that, in sLN alone, the simple morphologicaltype was indeed due to the presence of the molecule inits soluble state.

In the present study, most analysis was done after 2days in culture and with neurones plated at low densityin a chemically defined medium. These culture con-ditions allowed a good morphological description ofindividual neurones and prevented the conditioning ofthe medium by the cells. Indeed neurone-conditionedmedium has been shown to modify neuronal mor-phology (Bartlett & Banker, 1984; Chamak etal. 1987).The choice of a short-term culture was dictated by ourinterest in the initial steps in neurite formation. How-ever, experiments performed after 5 days in culture ledto the same conclusions regarding the developmentalpatterns created by the different forms of matrixmolecules.

A special difficulty was encountered because of theneuronal death observed in the conditions of lowneuronal density. Indeed the unipolar or multipolarmorphologies observed in the different conditions couldhave resulted from the differential loss of specificneuronal subpopulations. This possibility was ruled outby reversion experiments in which unipolar neurones,trypsinized and replated in control medium, were stillable to adopt a multipolar morphology. Conversely, theaddition of sLN to multipolar neurones promoted theelongation of only one of the several neurites.

Classification of the neurites was based on morpho-logical and immunocytochemical criteria. The neuritesof multipolar neurones (on average 4 to 5) were thickand branched in the region proximal to the cell body, as

expected for dendrites. The neurite of unipolarneurones conformed to axonal criteria, they wererather long, did not taper and showed a limited numberof branch points after two days in culture. After 5 days,some branching was observed in regions distal to thecell body.

Immunocytochemical methods were used to confirmthe morphological characterization of axon-like anddendrite-like neurites. We first used an antibody di-rected against MAP2, a protein highly enriched in thedendritic compartment of most neurones (Matus et al.1981; Caceresetal. 1984; De Camilli etal. 1984; Cacereset al. 1986). The cell bodies and the short neuritesexhibiting dendrite-like morphology were stronglystained with the anti-MAP2 antiserum. However, wealso observed a strong MAP2 staining in the proximalportion and in the varicosities of some axon-likeneurites, a staining pattern already observed in imma-ture axons by Caceres et al. (1986).

Neurite characterization was confirmed with the helpof antibodies directed against phosphorylated forms of160 and 200 xlO3 Mr neurofilament proteins which arespecifically enriched in the axonal compartment (Stern-berger & Sternberger, 1983; Shaw et al. 1985; Dahl &Bignami, 1985; Peng et al. 1986; Foster et al. 1987;Levitt, personal communication; Liem, personal com-munication). In all conditions, only the neurites withaxonal morphological traits were stained by the twodifferent axon-specific antibodies and the dendrite-likeneurites never showed any immunoreactivity.

Therefore we proposed that bFN and sLN inducedaxonal outgrowth and inhibited dendritic development,whereas bLN, sFN and medium without ECM proteinspermitted both axonal and dendritic growth.

This interpretation of an inhibiting effect of sLN andbFN on dendritic growth was further substantiated bythe comparative analysis of neuritic lengths in thedifferent culture conditions. Clearly the acceleratedoutgrowth of the axon in bFN and sLN was associatedwith an almost complete absence of dendritic growth. Apossible interpretation of this result is that, in a definedmetabolic state, neurones synthesize a given amount ofmembrane and cytoskeleton but that external con-ditions dictate in which compartments (axonal oraxonal and dendritic) this neurite-building material isdispatched. One must underline the possible physio-logical importance of a mechanism (dendrite growthinhibition) by which all neurite-building elements aredispatched into the axonal compartments, reducingthus the time necessary for the axonal growth cone toreach its target.

In the two culture conditions favouring axonal out-growth (sLN and bFN), quantitative measurements ofthe apparent cell body surfaces demonstrated a de-creased spreading of the neuronal soma. This obser-vation is similar to that of Bruckenstein & Higgins(1988a) on sympathetic neurones. The decrease inneuronal spreading suggests a simple physical expla-nation for the separate regulation of dendrite and axonoutgrowth. This hypothesis draws on D'Arcy Thomp-son's and Plateau's calculations for fluids mechanic

490 B. Chamak and A. Prochiantz

(1917). In high-surface-tension conditions (low ad-hesion), process outgrowth strongly depends on neuri-tic viscosity. In fact, in our experiments (Table 5),dendritic outgrowth was observed in all conditionspermitting strong neuronal adhesion (possibly implicat-ing the low viscosity of the dendroplasm) whereas axonswere able to grow in weak adhesion conditions, prob-ably because of their high viscosity, easily explained bytheir large amount of fasciculated microtubules(Autillo-Touati et al. 1988; Chamak et al. in prep-aration).

In order to understand the direct effect of LN and FNon neuronal morphology and polarity, we must assumethe existence of neuronal receptors for these molecules.Bozycko & Horwitz (1986) have already demonstratedthe participation of a putative cell surface receptor forLN and FN in peripheral neurite extension. We arepresently studying the presence of ECM protein recep-tors on CNS neurones in culture. Specific staining ofsuch neurones with an antibody directed against theintegrin /3 subunit has already been observed (notshown).

We have not yet investigated the mechanism impliedin the decreased adhesion observed in conditions lead-ing to the unipolar morphology. We speculate that sLNoccupies LN-binding sites on the cell surface, decreas-ing the number of receptors available for the formationof adhesion plaques between cells and substrate. Con-versely, it is possible that bFN occupies importantcellular sites for adhesion without leading to the intra-cellular changes required for adhesion and spreading.In fact, the numerous matrix molecules, matrix recep-tors and adhesion molecules involved in cell-cell orcell-substratum adhesion indicate that several mechan-isms could be involved (Bozyczko & Horwitz, 1986;Cohen et al. 1987; Bixby et al. 1988; Tomaselli et al.1988; Werz & Schachner, 1988). Among these mechan-isms, a direct trophic activity of the matrix moleculescannot be excluded (Panayotou et al. 1989).

In our previous studies, we showed that mesen-cephalic and striatal neurones develop their dendriticarborization much faster on homotopic than on hetero-topic astrocytes (Chamak et al. 1987). This finding cannow be interpreted in term of preferential adhesion ofthe neuronal cells on astrocytes derived from the sameregion of the brain. Such a region-specific preferentialadhesion of neurones to astrocytes could be explainedin several ways. First, one can speculate on the exist-ence of region-specific cell adhesion molecule isoformscreated through post-translational modifications andalternative splicing (Rutishauser & Goridis, 1986).Second, homotopic situations could induce a fasterneuronal maturation, due to the existence of region-specific astrocyte-derived factors which, in turn, accel-erate the transition from the embryonic sialylated N-CAM to the more adhesive adult form (Edelman,1986).

Finally, the study presented here raises the possibilitythat the equilibrium between substrate-attached andsoluble LN or FN plays an important role in theseparate regulation of axonal and dendritic growth. The

region specificity of astrocyte-induced dendrite growththerefore suggests the presence of different isoforms ofmatrix molecules and of their receptors during develop-ment in the different regions of the brain. We arepresently investigating this hypothesis of a role ofmatrix receptors as positional markers in the nervoustissue. It is, however, noteworthy that such a positionalmarker role for the high-affinity LN receptor and forthe a subunit of the integrin heterodimer has beendemonstrated in completely different systems (Leptin etal. 1987; Rabacchi et al. 1988).

We thank Dr J. Glowinski for his useful advices andconstant support. We are also grateful to Drs N. Brunet deCarvalho, A. Rousselet, F. Lafont and M. Vigny. This workwas supported by grants from INSERM, Rh6ne-PoulencSante and DRET.

References

ACHESON, A., EDGAR. D., TIMPL, R. & THOENEN, H. (1986).Laminin increases levels and activity of tyrosine hydroxylase incalf adrenal medullary cells. J. Cell Biol. 102, 151-159.

AUTILLO-TOUATI, A., CHAMAK, B., ARAUD, D., VUILLET, J., SEITE,R. & PROCHIANTZ, A. (1988). Region-specific neuro-astroglialinteractions:ultrastructural study of the in vitro expression ofneuronal polarity. / . Neurosci. Res. 19, 326-342.

BARON-VAN EVERCOOREN, A., KLEINMAN, H. K., OHNO, S.,MARANGOS, P., SCHWARTZ, J. P. & DUBOIS-DALQ, M. (1982).Nerve growth factor, and fibronectin promote neurite growth inhuman fetal sensory ganglion cultures. / . Neurosci. Res. 8,179-193.

BARTLETT, W. P. & BANKER, G. A. (1984). An electron microscopicstudy of the development of axons and dendrites by hippocampalneurons in culture. I. Cells which develop without intercellularcontacts. / . Neurosci. 4, 1944-1953.

BIXBY, J. L., LILUEN, J. & REICHAKDT, L. F. (1988). Identificationof the major proteins that promote neuronal process outgrowthon Schwann cells in vitro. J. Cell Biol. 107, 353-361.

BOZYCZKO, D. & HORWITZ, A. F. (1986). The participation of aputative cell surface receptor for laminin and fibronectin inperipheral neurite extension. /. Neurosci. 6, 1241-1251.

BRAY, D. (1984). Axonal growth in response to experimentallyapplied mechanical tension. Devi Biol. 102, 379-389.

BRUCKENSTEIN, D. A. & HIGGINS, D. (1988a). Morphologicaldifferentiation of embryonic rat sympathetic neurons in tissueculture. I. Conditions under which neurons form axons but notdendrites. Devi Biol. 128, 324-336.

BRUCKENSTEIN, D. A. & HIGGINS, D. (1988ft). Morphologicaldifferentiation of embryonic rat sympathetic neurons in tissueculture. II. Serum promotes dendritic growth. Devi Biol. 128,337-348.

CACERES, A., BANKER, G. A. & BINDER, L. (1986).Immunocytochemical localization of tubulin and microtubule-associated protein 2 during the development of hippocampalneurons in culture. J. Neurosci. 6, 714-722.

CACERES, A., BANKER, G. A., STEWART, O., BINDER, L. & PAYNE,M. (1984). MAP2 is localised to the dendrites of hippocampalneurons which develop in culture. Dev. Brain Res. 13, 314-318.

CARRI, N. G., PERRIS, R., JOHANSSON, S. & EBENDAL, T. (1988).Differential outgrowth of retinal neurites on purified extracellularmatrix molecules. J. Neurosci. Res 19, 428—439.

CHAMAK, B., FELLOUS, A., GLOWINSKI, J. & PROCHIANTZ, A.(1987). MAP2 expression and neuritic outgrowth and branchingare co-regulated through region specific neuroastroglialinteractions. / . Neurosa. 7, 3163-3170.

COHEN, J., BURNE, J. F., MCKINLAY, C. & WINTEJI, J. (1987). Therole of laminin and the laminin/fibronectin receptor complex inthe outgrowth of retinal ganglion cell axons. Devi Biol. 122,407-418.

Separate regulation of axonal and dendritic outgrowth 491

D'ARCY THOMPSON. (1917). On Growth and Form. J. T. Bonner,editor. Cambridge University Press.

DAHL, D. & BIGNAMI, A. (1985). Neurofilament phosphorylation indevelopment. A sign of axonal maturation? Expl Cell Res. 162,220-230.

DAVIS, G. E., MANTHORPE, M., ENGVAIX, E. & VARON, S. (1985).Isolation and characterization of rat Schwannoma neuritepromoting factor: evidence that the factor contains laminin. J.Neurosci. 5, 2662-2671.

D E CAMILLI, P., MILLER, P. E., NAVONE, F., THEURKAUF, W. E. &VALLEE, R. B. (1984). Distribution of microtubule-associatedprotein 2 in the nervous system of the rat studied byimmunofluorescence. Neuroscience 11, 819-846.

DENIS-DONINI, S., GLOWINSKI, J. & PROCHIANTZ, A. (1984). Glialheterogeneity may define the three-dimensional shape of mousemesencephalic dopaminergic neurones. Nature, Lond. 307,641-643.

EDELMAN, G. M. (1986). Cell adhesion molecules in the regulationof animal form and tissue pattern. A. Rev. Cell Biol. 2, 81-116.

EDGAR, D., TIMPL, R. & THOENEN, H. (1984). The heparin bindingdomain of laminin is responsive for its effect on neuriteoutgrowth and neuronal survival. EMBO J. 3, 1463-1468.

FAIVRE-BAUMAN, A., PUYMIRAT, J., LOUDES, C , BARRET, A. &TIXIER-VIDAL, A. (1984). Laminin promotes attachment andneurite elongation of fetal hypothalamic neurons growth inserum-free medium. Neurosci. Lett. 44, 83-89.

FOSTER, G. A., DAHL, D. & LEE, V. M.-Y. (1987). Temporal andtopographic relationships between the phosphorylated andnonphosphorylated epitopes of the 200 kD neurofilament proteinduring development in vitro. J. Neurosci. 7, 2651-2663.

FRASER, S. E. (1980). A differential adhesion approach to thepatterning of nerve connections. Devi Biol. 79, 453-464.

HAMMARBACK, J. A., PALM, S. L., FURCHT, L. T. & LETOURNEAU,P. C. (1985). Guidance of neurite outgrowth by pathways ofsubstratum-adsorbed laminin. J. Neurosci. Res. 13, 213-220.

KORSHING, S., AUBURGER, G . , HEUMANN, R . , SCOTT, J. &THOENEN, H. (1985). Levels of nerve growth factor and itsmRNA in the central nervous system of the rat correlate withcholinergic innervation. EMBO J. 4, 1389-1393.

LANDER, A. D., FUJII, D. K. & REICHARDT, L. F. (1985). Laminin

is associated with the "neuritic outgrowth-promoting factors"found in conditioned media. Proc. natn. Acad. Sci. U.S.A. 82,2183-2187.

LEPTIN, M., AEBERSOLD, R. & WILCOX, M. (1987). Drosophila

position-specific antigens resemble the vertebrate fibronectin-receptor family. EMBO J. 6, 1037-1043.

LETOURNEAU, P. C. (1975). Cell-to-substratum adhesion andguidance of axonal elongation. Devi Biol. 44, 92-101.

LEVI-MONTALCINI, R. (1982). Developmental neurobiology and thenatural history of nerve growth factor. A. Rev. Neurosci. 5,341-362.

LIESI, P., DAHL, D. & VAHERI, A. (1983). Laminin is produced byearly rat astrocytes in primary cultures. J. Cell Biol. 96, 920-924.

LIESI, P., KAAKKOLA, S., DAHL, D. & VAHERI, A. (1984). Laminin

is induced in astrocytes of adult brain by injury. EMBO J. 3,683-686.

MANTHORPE, M., ENGVALL, E., RUOSLAHTI, E., LONGO, F. M.,

DAVIS, G. E. & VARON, S. (1983). Laminin promotes neuriticregeneration from cultured peripheral and central neurons. J.Cell Biol. 97, 1882-1890.

MATUS, A., BERNHARDT, R. & HUGH-JONES, T. (1981). High

molecular weight microtubule-associated proteins are

preferentially associated with dendritic microtubules in brain.Proc. natn. Acad. Sci. U.S.A. 78, 3010-3014.

MILLARUELO, A. I., NIETO-SAMPEDRO, M. & COTMAN, C. W. (1988).

Cooperation between nerve growth factor and laminin orfibronectin in promoting sensory neuron survival and neuriteoutgrowth. Dev. Brain. Res. 38, 219-228.

MORRISON, R. S., SHARMA, A., DE VELLIS, J. & BRADSHAW, R. A.

(1986). Basic fibroblast growth factor supports the survival ofcerebral cortical neurons in primary culture. Proc. natn. Acad.Sci. U.S.A. 83, 7537-7541.

PANAYOTOU, G., END, P., AUMAILLEY, M., TIMPL, R. & ENGEL, J.

(1989). Domains of laminin with growth-factor activity. Cell 56,93-101.

PENG, I., BINDER, L. I. & BLACK, M. M. (1986). Biochemical and

immunological analysis of cytoskeletal domains of neurons. J.Cell Biol. 102, 252-262.

PRICE, J. & HYNES, R. O. (1985). Astrocytes in culture synthesizeand secrete a variant form of fibronectin. J. Neurosci. 5,2205-2527.

RABACCHI, S. A., NEVE, R. L. & DRAGER, U. C. (1988). Molecular

cloning of the "dorsal eye antigen": homology to the highaffinity laminin receptor. Soc. for Neurosci. Abst. 14, 311.4.

ROUSSELET, A., FETLER, L., CHAMAK, B. & PROCHIANTZ, A. (1988).Rat mesencephalic neurons in culture exhibit differentmorphological traits in the presence of media conditioned onmesencephalic or striatal astroglia. Devi Biol. 129, 495-504.

RUTISHAUSER, U. & GORIDIS, C. (1986). NCAM: the molecule andits genetics. TIG. 2, 72-75.

SHAW, G., BANKER, G. A. & WEBER, K. (1985). An

immunofluorescence study of neurofilament protein expressionby developing hippocampal neurons in tissue culture. Eur. J. CellBiol. 39, 205-216.

SMALHEISER, N. R., CRAIN, S. M. & REID, L. M. (1984). Lamininas a substrate for retinal axons in vitro. Dev. Brain Res. 12,136-140.

STERNBERGER, L. A. & STERNBERGER, N. H. (1983). Monoclonalantibodies distinguish phosphorylated and nonphosphorylatedforms of neurofilaments in situ. Proc. natn. Acad. Sci. U.S.A. 80,6126-6130.

THIERY, J. P., DUBAND, J.-L. & DELOUV£E, A. (1982). Pathwaysand mechanisms of avian trunk neural crest cell migration andlocalization. Devi Biol. 93, 324-343.

TOMASELLI, K., NEUGEBAUER, K. M., BIXBY, J. L., LILIEN, J. &

REICHARDT, L. F. (1988). N-cadherin and integrins: two receptorsystems that mediate neuronal process outgrowth on astrocytesurfaces. Neuron. 1, 33-43.

UNSICKER, K., REICHERT-PREIBSCH, H., SCHMIDT, R., PETTMAN, B.,

LABOURDETTE, G. & SENSENBRENNER, M. (1987). Astroglial andfibroblast growth factors have neuronotrophic functions forcultured peripheral and central nervous system neurons. Proc.natn. Acad. Sci. U.S.A. 84, 5459-5463.

WALICKE, P., COWAN, W. M., VENO, N., BAIRD, A. & GUILLEMIN,

R. (1986). Fibroblast growth factor promotes survival ofdissociated hippocampal neurons and enhances neurite extension.Proc. natn. Acad. Sci. U.S.A. 83, 3012-3016.

WERZ, W. & SCHACHNER, M. (1988). Adhesion of neural cells toextracellular matrix constituents. Involvement ofglycosaminoglycans and cell adhesion molecules. Dev. Brain Res.43, 225-234.

(Accepted 11 March 1989)