Agrin Inhibits Neurite Outgrowth but Promotes Attachment of Embryonic Motor and Sensory Neurons

DEVELOPMENTAL BIOLOGY 181, 21–35 (1997)ARTICLE NO. DB968435

Agrin Inhibits Neurite Outgrowth but PromotesAttachment of Embryonic Motorand Sensory Neurons

David Chang,* J. Susie Woo,* James Campanelli,†Richard H. Scheller,‡ and Michael J. Ignatius**Department of Molecular and Cell Biology, University of California, Berkeley, Berkeley,California 94720-3200; †Department of Biochemistry, B4, University of Illinois, Urbana,Urbana, Illinois 61801; and ‡Department of Molecular and Cellular Physiology, HowardHughes Medical Institute, Stanford University Medical Center, Stanford, California 94305-5426

Agrin is a secreted glycoprotein with the ability to cluster cell surface molecules, including the nicotinic acetylcholinereceptor (AchR) on muscle cells. Alternate splicing of agrin mRNA results in a family of agrin proteins which differ intheir clustering potency. Neuronal-specific isoforms with the highest clustering activity play a role in clustering postsynap-tic proteins at the neuromuscular junction. However, the function of agrin isoforms expressed in many nonneuronal tissues,and only weakly active in clustering assays, remains obscure. Monolayer cultures of Chinese hamster ovary (CHO) cellsexpressing a neuronal (agrin-19) or a nonneuronal (agrin-0) form of agrin were used to assay the effect of agrin on neuriteoutgrowth and cell attachment. These results were compared to outgrowth on control CHO cells expressing only drugresistance and on regions of CHO–agrin monolayers not expressing detectable levels of agrin. Neurite extension on confluentmonolayers of agrin-0- or -19-expressing CHO cells was reduced substantially below that of controls. In one experimentneurite lengths were compared at 2 and 3 days after plating and suggested that neurite outgrowth may be stopped and notsimply retarded. Attachment of sensory or motoneurons was nearly twofold higher to agrin monolayers than to controlcells, showing that the inhibition is not a result of a nonpermissive environment. An agrin construct missing the C-terminal half, removing the major site of variability and clustering activity, was also tested. This construct did not reduceoutgrowth, suggesting that the C-terminal half of the protein may be important in stopping growth as well as inducingclustering. These results expand the role of agrin in synaptogenesis as it may provide a stop signal at the myofiber surfaceand may anchor the presynaptic fibers to the eventual motor endplate. q 1997 Academic Press

INTRODUCTION However, tissue distribution studies suggest an expandedrole for agrin in synaptogenesis or development in general

Agrin is a large heparin-sulfate-containing glycoprotein (Ma et al., 1994; Patthy and Nikolics 1993; Biroc et al.,made by many neurons and their targets during synaptogen- 1993; Thomas et al., 1993; Hoch et al., 1993). The mostesis. In the best-studied example, motoneurons make agrin, compelling observation is that a function for the predomi-as do their target myotubes in advance of innervation. The nant isoform of agrin expressed in the body has not beenagrin hypothesis states that nerve cells release agrin, which established. Four alternative transcripts of agrin with inser-induces the clustering of Ach receptors underneath the tions at the Y and Z sites (rat agrin terminology) have beeneventual presynaptic element (McMahan and Wallace, seen, and generate multiple isoforms of the protein (Rupp1989; McMahan, 1990). Agrin released by nerves (Cohen et al., 1991; Hock et al., 1993; Stone and Nikolics, 1995;et al., 1994), added exogenously (Nitkin et al., 1987), or Ma et al., 1995; Ruegg et al., 1992; Tsim et al., 1992). Whileexpressed by CHO cells (Campanelli et al., 1991), can clus- as many as eight agrin forms are possible, only five areter Ach receptors. Soluble agrin can also induce clustering expressed. Agrins Y0Z0 and Y4Z0 (agrin-0 in this paper)of additional junctional proteins (Nitkin and Rothschild, contain either zero or one insert and have no clustering1990) and associated cytoskeletal proteins (Shadiack and activity. Agrins with 8, 11, and 19 (agrin-19 in this paper)

amino acid inserts at the Z site are neural specific (SmithNitkin, 1991).

21

0012-1606/97 $25.00Copyright q 1997 by Academic PressAll rights of reproduction in any form reserved.

AID DB 8435 / 6X15$$$$21 12-06-96 09:37:00 dba

22 Chang et al.

et al., 1994; Hoch et al., 1993). These forms show the high- ited to motor systems and acts on additional populationsof cells, our results suggest that agrin performs an expandedest level of clustering activity, with the order of activity

Y4Z8 Å Y4Z19 (agrin-19) ú Y4Z11 @ Y4Z0 (agrin-0) (Rupp role in stabilizing synapses throughout the nervous system.et al., 1991; Hoch et al., 1993; Ferns et al., 1992; 1993). Thisleaves unexplained a role for the most abundant forms ofagrin, agrin Y0Z0 or agrin Y4Z0, which are synthesized by MATERIALS AND METHODSnonneuronal cells, including muscle (Ferns et al., 1993).Significantly, agrin-0 is abundantly expressed prior to in-

Materials. F12 culture medium was from BioWhittaker (Gaith-nervation throughout the muscle fiber (Hoch et al., 1993)ersburg, MD); F12/DME high glucose in a 1:1 mixture and fetaland in aneural muscle (Fallon et al., 1989), arguing for abovine serum (FBS) were obtained from Irvine Scientific (Santa Ana,role independent of receptor clustering.CA). Geneticin (G418) was from GibcoBRL (Gaithersburg, MD);Agrin may stabilize or anchor growing axons at their syn-glass coverslips were purchased from Carolina Biologicals, (Burling-

aptic sites. During peripheral nerve regeneration, adult ax- ton, NC) and all tissue culture plastic was from Corning (Corning,ons find their synaptic sites, cease growing, and differentiate NY). A mouse monoclonal antibody to rat agrin, mAb 131, wasa presynaptic element, all in the absence of the muscle itself used as conditioned medium at 1:50 dilution. This mAb has no(Sanes and Chiu, 1983). Extracellular matrix (ECM) proteins effect on agrin-induced clustering but does label agrin on fixed

and live tissue. Fluoromount-G was from Southern Biotechspecific to the original synaptic fold, including agrin and S-(Birmingham, AL); 2*,7*-bis(2-carboxyethyl)-5-carboxyfluorescein,laminin, survive myofiber ablation (Sanes et al., 1978).acetoxymethyl ester (BCECF-AM) was from Molecular Probes (Eu-These components may act in concert or separately to or-gene, OR). Laminin and fibronectin were from Collaborative Re-chestrate the reorientation and anchoring of the nerve ter-search, Inc. (Lexington, MA). Rabbit antiserum to neurofilament-minal to the original synaptic site. For example, S-laminin200, all secondary antibodies, BSA, poly-d-lysine (pdl) type V, andis inhibitory to motoneuron outgrowth, an effect abrogatedany unspecified reagents were from Sigma Immunochemicals (St.

by LRE peptides (Porter et al., 1995). LRE is a motif con- Louis, MO). Phosphate-buffered saline (PBS) is 200 mg/liter KCl,served in rat and mouse S-laminin, but not chick, agrin. 200 mg/liter K2SO4 , 8 g/liter NaCl, 2.16 g/liter Na2HPO4r7H2O, 1The picture during development is less clear. S-laminin, mM CaCl2 , 0.5 mM MgCl2 adjusted to pH 7.4.unlike agrin, is absent from myoblasts prior to innervation. Cell culture. CHO cells were transfected with either agrin-0 or

agrin-19 constructs, and stable G418-resistant colonies were se-This leaves unresolved what signals arrest growth at synap-lected according to Campanelli et al. (1991). Pairs of sublines weretic sites during development.used in the study, hereafter termed agrin-0-1, agrin-0-2, agrin-19-Experiments are presented that show that both nonneu-1, and agrin-19-2. The agrin constructs alone are simply referredronal and neuronal forms of agrin promote attachmentto as agrin-19 and agrin-0.while inhibiting neurite outgrowth of neurons from three

The R4HA transfectant line was the generous gift of G. G. Gayer.sources. These studies are in agreement with the recentIt spans nucleotide positions 1–3863, and includes two LRE se-

finding that mutant mice missing all agrin isoforms by ho- quences. Two drug-resistant control CHO lines, DR-1 and DR-2,mologous recombination do not form mature motor end- were derived from colonies of cells transfected with G418-resistantplates. Instead, axons grow past normal synaptic sites (Gau- plasmid only and maintained throughout in G418. CHO cells weretam et al., 1996). The present study, while in general agree- propagated in 1:1 F12/DME high glucose, 10% heat-inactivated

FBS, 100 mg/ml streptomycin, 100 U/ml penicillin G, 292 mg/ml L-ment with another study examining agrin’s effect in vitroglutamine, and 500 mg/ml G418 in 7% CO2/93% air at 377C. CHO(Campagna et al., 1995), differs considerably in the approachcells were passaged at 70% confluence and passaged at 1/20 dilutionused and certain conclusions. The earlier study assays theto allow the cells to continue logarithmic growth. All CHO cellsresponse of ciliary ganglion neurons when first growing onwere maintained for less than 1 month. The agrin-19 (equivalenta fibronectin substrate, that then confront an island of agrin-to Y4Z19) isoform was maintained for less than 14 days to maintainexpressing CHO cells (Campagna et al., 1995). Two con-adequate expression levels. All data presented were for subclones

cerns of their assay are avoided in our present study. First, designated agrin-0-1 and agrin-19-1. Two additional subclones orneurites may be deflected by the physical properties of the agrin-expressing clones were used to rule out subclone variability,CHO cells present in the island—all of the CHO sublines designated agrin-0-2 and agrin-19-2 (data not shown).used show marked variations in their degree of spreading The degree of cell spreading was variable for different sublines of

agrin-expressing and drug-resistant control cells. Conditions wereversus rounding. Second, neurons growing on laminin havesought that minimized these differences since round or nonspreadbeen shown to avoid substrates, like collagen IV, that incells could create a more severe physical barrier to neurons at-isolation are excellent substrates for growth (Gunderson,tempting to grow onto the CHO cells. Laminin gave the best re-1987). Assays of this sort may merely determine the hierar-sponse but appeared to be inducing neurite outgrowth-promotingchy of substrate preferences rather than measuring a truemolecules in the CHO cells, effectively shielding any inhibitoryavoidance response. By assaying growth of cells on conflu-influence that agrin might be having. Eventually, conditions were

ent monolayers of equally spread sublines, we have been found that allowed CHO cells to be grown on pdl alone with consis-able to abrogate this concern. In addition, our study shows tent attachment and spreading response. To further minimize thisthat the agrin’s effects do not appear limited to motor neu- concern the additional subclones were assayed, and found to haverons, but can alter the growth and promote the adhesion of consistently similar results.

Fertile White Leghorn eggs were obtained from Feather Hill Farmsensory neurons as well. Since agrin expression is not lim-

Copyright q 1997 by Academic Press. All rights of reproduction in any form reserved.

AID DB 8435 / 6X15$$$$21 12-06-96 09:37:00 dba

23Agrin Inhibits Neurite Outgrowth

(Petaluma, CA) and maintained at 387C and 95% humidity until cies-specific, fluorescently labeled secondary antibodies were addedfor 30 min in the dark. After three more washes in PBS/glycerol,use. Ciliary ganglion neurons were dissected from E8 embryos andcoverslips were mounted onto slides using Fluoromount-G.cultured according to Nishi and Berg (1977), sensory neurons from

Data analysis. Agrin and neurofilament immunostaining weredorsal root ganglion neurons (DRG) were from E5–6.5 embryossimultaneously imaged using a dual wavelength filter set (Techni-according to Banker and Goslin (1991) and motoneuron culturescal Instruments, San Francisco, CA). Images were recorded with awere from ventral neural tube of E5–6 embryos (Bloch-Gallego etPhotometrics (Tuscon, AZ) series 200 cooled CCD (charged coupledal., 1991). Greater than 50% of the cells in this region of E5–6device) camera and the sum of all neurite lengths for an individualneural tube are immunoreactive with the motoneuron specificneuron on agrin-positive regions was measured using Image 1 soft-mAb, DM-1 (Burns et al., 1991; P. B. Quebada and M.J.I., unpub-ware (Universal Imaging, West Chester, PA). All of our images andlished observation.)studies were performed on neurons growing on top of cells, a pointNeurite outgrowth assay. Agrin-expressing or drug-resistantreadily established by rolling the tissues through successive focalcontrol CHO cells were plated onto 200 mg/ml pdl-coated glassplanes. Slides were analyzed by a double-blind procedure, with onecoverslips taking care to evenly distribute them to establish a uni-investigator identifying neurons completely contained within aform density. CHO cells were grown on these coverslips until 70%uniform region of agrin or control expression, and another measur-confluence for dissociated cultures of ventral spinal cord and ciliarying lengths captured on a black and white video screen. Neuriteganglion neuron cocultures or until 85 –90% confluence for DRG-lengths for 20 to 40 neurites were binned and plotted to show thederived sensory neuron cocultures. This variation in plating timesdistribution of neurite lengths as a percentage of the total popula-was used because the two types of neurons had different rates oftion. Lengths were normalized to the longest neurite in a set ofneurite outgrowth. Neurons (approximately 10,000/well) wereexperiments to allow comparison of multiple data sets. The dataadded such that the final medium of the cocultures was 75% F12/in Table 1 were compiled by determining the length of neurites25% DME high glucose, in addition to previously stated amounts offrom each substrate (shown in Figs. 5–10) that were 50% of theFBS, streptomycin, penicillin G, and L-glutamine, although withoutnormalized total. Responses to control substrates were divided byG418. Appropriate growth factors were added: 3% E16 chick eyethose responses on agrin-expressing cells to establish the net reduc-extract for ciliary ganglion-derived motor neurons (Nishi and Berg,tion in neurite outgrowth. These ratios were subjected to the Stu-1981), 50 mg/ml chick muscle extract for ventral spinal cord (Bloch-dent t test to establish confidence limits.Gallego et al., 1991), or 50 mg/ml NGF for sensory neurons. Unless

otherwise specified, sensory and ciliary neurons were allowed togrow for 24 hr, and ventral neural tube cultures were grown for 48hr, due to their slower rate of axon extension. Cells were then RESULTSfixed, stained, mounted, and photographed with a Nikon MICRO-PHOT-SA fluorescent microscope, or measured directly as detailed

A mAb to rat agrin was used to establish whether agrinbelow for data analysis.expression in the transfected cells was uniform in all theCell attachment assay. To test cell attachment, freshly dis-sublines. A variable expression pattern was seen for agrin-sected ciliary or sensory neurons were added to confluent mono-19 (Fig. 1A), while agrin-0 was expressed at high and nearlylayer cultures of agrin-0-transfected or drug-resistant control CHOuniform levels (Figs. 1D and 3A). Expression levels werecells. To visualize the neurons, they were first labeled with the vitalhigher in unpermeabilized cells (Fig. 4). Most of the agrinfluorescent dye, BCECF-AM, and then added at a concentration of

50,000 cells/well. Cells were also assayed for their attachment to was on the surface of cells and surface expression levelslaminin, pdl, or fibronectin-coated coverslips, that were first were more uniform than that seen on permeabilized cellsblocked with PBS and 1% BSA according to Ignatius et al. (1988). (compare Figs. 3A and 4). In agrin-0-expressing lines, agrinCells were allowed to attach for 2 hr with constant agitation pro- immunoreactivity was detected in unusual tracks ex-vided by a rotating platform. Nonadherent cells were washed off tending beyond the limits of the cell (Figs. 1C, 1D, and 4). Inby several streams of PBS and then fixed with 2% paraformalde- some instances where cells have been removed by washinghyde/2% sucrose for 15 min. The total number of attached cells

steps, a uniform coating of agrin was left behind (notin duplicate wells was determined by counting the equivalent areasshown), suggesting that agrin can deposit an extracellularunder epi-fluorescent illumination, and each trial repeated once.matrix-like coating. This acellular form of deposited agrinImmunofluorescence. For immunodetection of agrin and neu-does not appear to inhibit neurite extension, although cellsrofilament, coverslips were fixed in 2% paraformaldehyde/2% su-expressing agrin are avoided, as can be seen by the neuritecrose for 15 min. After four washes in PBS with 10% glycerol,

coverslips were incubated for 15 min in 10 mM ammonium chlo- skirting the cell boundary but not the region of acellularride with 2% sucrose and then 100 mM glycine with 2% sucrose to agrin deposition (Figs. 1C and 1D).quench any reactive formaldehyde groups. To image neurofilament, In Fig. 2, a dorsal root ganglion neuron is shown grow-cells were then treated with 0.25% Triton X-100 in PBS/glycerol ing on pdl as it confronts a boundary of CHO cells ex-for 10 min to permeabilize them. To detect surface expression of pressing agrin-0. In agreement with earlier studies whereagrin, cells were not permeabilized. Following two more washes, fibronectin was used as a substrate (Campagna et al.,cells were blocked with PBS, 7% normal goat serum, 1.5% BSA,

1995), neurite growth is affected when agrin expressing10% glycerol, and 0.1% Triton X-100 for 30 min. Cells were thencells are encountered (Figs. 2A and 2B). Non-agrin-ex-washed twice in the blocking solution with 0.3% Triton X-100.pressing cells did not slow neurite outgrowth (Figs. 2CPrimary mouse mAb to agrin and rabbit anti-neurofilament 200and 2D). Similar results were seen when laminin wasantibodies were added for 2 hr, separately or in combination forused as a substrate or when agrin-19-1 cells were useddouble labeling. Coverslips were then washed twice in large vol-

umes of PBS/glycerol and placed in blocking solution for 1 hr. Spe- (data not shown).


AID DB 8435 / 6X15$$$$21 12-06-96 09:37:00 dba

24 Chang et al.

TABLE 1Response of Sensory and Ventral Neural Tube Derived Neurons to Agrina

DR-1/0 DR-1/19 DR-2/0 DR-2/19 19(0)/DR-2 DR-2/R4HA

Ventral tube neurons 2.3 (0.6) 2.3 (0.2) 2.7 (1.2) 3.0 (0.5) 1.03 (.36) naSensory neurons / NGF na na 2.5 (0.6) 2.3 (0.3) 1.3 (0.1) 0.4 (1.1)

Note. All data are expressed as ratios of neurite lengths on nonexpressing controls versus various agrin isoforms.a Neurite lengths were derived from plots, similar to those in Figs. 5–9, by determining the estimated length of a neurite that was 50%

of the normalized total. Numbers represent the ratio of these 50% values for neurite outgrowth on three isoforms of agrin-expressingcells (agrin-0, agrin-19, and agrin R4HA) versus three nonexpressing controls: DR-1, DR-2, and 19(0). Values in parentheses are the rangeof ratios derived from at least duplicate trials. A value of one represents no change, while greater than one reflects a reduction in neuritelengths. Values for all but 19(0)/DR-2 and DR-2/R4HA are statistically different (P § 0.001).

A concern of this approach and that used by others ence prior to addition of neurons directly on top of themonolayers. Only those neurites growing completely on(Campagna et al., 1995) is the presence of a cell boundary,

in particular those created by partially rounded cells agrin immunopositive or -negative cells were measured.Agrin expression was determined with agrin mAb inwhich occur in these cultures. Such boundaries may de-

flect growth as neurites attempt to leave the coverslip combination with neurofilament antibody. These dou-ble-labeled cultures were used for the remaining neuritesurface to grow on the surface of cells. To avoid this

physical barrier, CHO cells were grown to 70% conflu- outgrowth assays. Population distributions of all the ax-

FIG. 1. Immunohistochemical staining for agrin reveals high levels of expression of agrin-0 and more heterogeneous expression of agrin-19 in transfected, stable lines of CHO cells. CHO cells were grown on glass coverslips coated with pdl, then fixed and immunostained.In A are agrin-19-expressing CHO cells, double stained for neurofilament in B, and in C agrin-0-expressing cells are shown with thecorresponding neurofilamant stain imaged in D. In C the deposition of agrin-0 beyond the cell boundary can be seen (arrow) when comparedto the fluorescent image of neurofilament staining shown in D where the CHO cells show some faint background staining which definesthe actual cell boundary. This faint staining seen in fluorescence aligned with cell boundaries established by phase micrography. In bothB and D, an axon can be seen avoiding agrin-expressing cells, while it is able to grow over the deposited agrin material to some degree.Scale bar in D, 20 mm.


AID DB 8435 / 6X15$$$$21 12-06-96 09:37:00 dba


FIG

.2.

Agr

in-e

xpre

ssin

gC

HO

cell

sca

nde

flec

tth

egr

owth

ofse

nso

ryn

euro

ns

exte

ndi

ng

anax

onon

pdl.

InA

and

Ba

hig

hly

bran

ched

neu

ron

isse

enw

hos

egr

owth

has

arre

sted

atth

ebo

un

dary

ofag

rin

-0-e

xpre

ssin

gC

HO

cell

s.In

con

tras

t,n

euro

ns

grow

exu

bera

ntl

yon

todr

ug-

resi

stan

tco

ntr

olm

onol

ayer

s(C

,D

).A

grin

expr

essi

onis

show

nin

Aan

dC

,w

hil

en

euro

ns

are

reve

aled

wit

hn

euro

fila

men

tim

mu

nor

eact

ivit

yin

Ban

dD

.Sc

ale

bar

inD

,20

mm

.


AID DB 8435 / 6X15$$8435 12-06-96 09:37:00 dba

26 Chang et al.

ons elaborated by one neuron, growing within a confluent Indeed, agrin is disrupting neurite outgrowth to a markeddegree.monolayer of cells uniformly positive or negative for

agrin, were measured. The effects seen could be caused by the upregulation ofa CHO-specific diffusible molecule isolated during clonalCocultures, like those shown (Figs. 3A–3E), were mea-

sured while imaging both agrin and neurofilament immuno- selection, unrelated to agrin. To address this concern, thevariable expression in agrin-19-1 cultures was exploited toreactivity with dual wavelength filter sets. As can be seen,

the effect on total neurite lengths is even more marked compare outgrowth on agrin-expressing islands adjacent tononexpressing areas. Agrin-19 regions reduce axon out-when assayed in this manner (compare Figs. 3B and 3E).

Non-agrin-expressing cells support exuberant growth, while growth (Fig. 6A) and yet neurite outgrowth on nonexpress-ing regions was indistinguishable from DR-controls (Fig.the agrin-expressing monolayer supports little growth. This

assay also circumvented the concern that static analysis 6B). These data suggest that the only difference in the re-sponse generated by the CHO cells is the presence or ab-of growth cones near an agrin-expressing cell can give the

impression of arresting growth. Without time lapse infor- sence of agrin expression. This control also argues againstthe appearance of undetected, inhibitory molecules thatmation, such results cannot distinguish between an ap-

proaching axon, an arrested axon, or a retracting one. In our might arise during the selection of the agrin positive linesfor study, since their downregulation is unlikely to occurassays on continuous monolayers we are, however, able to

measure the net response to a maintained interaction with in parallel with agrin expression.Agrin, originally found to be effective on acetylcholineagrin. However, we cannot rule out that reductions in net

lengths were not due to increased cycles of growth and re- receptor clustering, is expressed by motoneurons, and wastherefore thought to be specific to motor systems. However,traction on agrin.

Dissociated cells from ventral spinal cord were used to distribution studies show that agrin is expressed in manyregions of the nervous system, including sensory neuronsisolate motoneuron-enriched cultures, with over 50% of the

cells positive for a motor neuron-specific marker (see Mate- and their eventual targets (Hoch et al., 1993; Stone andNikolics, 1995). It was therefore of interest to see if therials and Methods). Figure 5 summarizes the result for out-

growth on two agrin sublines, agrin-0-1 and agrin-19-1, com- effect on sensory cell neurite outgrowth was similar to theeffect on motoneurons. Figure 7 shows that agrin had thepared to drug-resistant controls. Control lines, termed

DR-1 and DR-2, differed in division rates and overall appear- same potent effect on neurite outgrowth from sensory cellswhen compared to the same two drug-resistant lines (Figs.ance, thus providing a good control for clonal variations in

CHO lines which might affect neuronal interactions. Neu- 7A and 7B) or when comparing growth on agrin-positiveregions to agrin-negative regions (Figs. 7C and 7D). Becauserite outgrowth was reduced to a similar degree on both alter-

natively spliced forms of agrin compared to either DR-1 the sensory cells were grown in the continued presence ofNGF, the neuronal subtypes would be primarily nociceptive(Figs. 5A and 5B) or DR-2 (Figs. 5C and 5D). Presence of an

insert at the Z sites (agrin-19), important for AchR cluster- and touch, without proprioceptive neurons that normallyinnervate motor fibers. Similar reductions in neuriteing activity, showed no detectable differences in this assay.

Moreover, comparable differences were seen when mea- lengths were also obtained using peripheral motoneuronsfrom ciliary ganglia (S. Combes and M.J.I., unpublished ob-sured against two drug-resistant lines and two subclones

of agrin-expressing cells, agrin-19-2 and agrin-0-2 (data not servation), in agreement with previous studies (Campagnaet al., 1995). Table 1 summarizes the net effect on neuriteshown). These results ensured that clonal variations in se-

lected lines, differing in their spreading capability or other outgrowth, for both motor and sensory neurons on all CHOcell lines.undefined molecular variations, were not altering the obser-

vations. For comparison, these changes in neurite distribu- DRG outgrowth was allowed to continue for longer timesto see if axonal growth was being stopped or merely slowed.tions are equivalent to those seen in other systems where

a nearly complete blockade of neurite outgrowth has been Neurons were prepared on the same day and added to theidentical preparations of agrin-0-1–CHO monolayers. Oneachieved. For example, coaddition of antibodies to N-cad-

herin and integrin or NgCAM and integrin to ciliary neu- set was stopped after 48 hr and the other was stopped after72 hr, and they were compared to each other. The curvesrons growing on Schwann cell monolayers achieves a simi-

lar shift in neurite length distribution (Bixby et al., 1988). representing the neurite distribution overlapped, (Fig 8A).

FIG. 3. Confluent islands of agrin-expressing CHO cells inhibit neurite outgrowth, while nonexpressing cells do not. Multiple imagesare shown in which agrin expression (A, D) neurofilament (B, E) and cell spreading under phase optics (C) are visualized. Ventral spinalcord cultures were grown on agrin-0-1 (A, B) or DR-1 (C–E) monolayers for 48 hr. Levels of agrin immunoreactivity are high and uniformon agrin-0-1 cells (A) while barely detectable on DR-1, nonexpressing cells (D). Neurofilament staining shows restricted neurite outgrowthon agrin-0-expressing cells (B), while long neurites can be seen on control CHO, non-agrin-expressing drug-resistant control cells (E). Cshows a phase image to reveal the underlying cell monolayer, since the lack of agrin immunoreactivity obscures this result (D). Filledarrows in A and B and open arrows in C–E point to identical regions of the culture. Scale bar in E, 20 mm.


AID DB 8435 / 6X15$$$$21 12-06-96 09:37:00 dba


12-06-96 09:37:00 dba

28 Chang et al.

FIG. 4. Unpermeabilized agrin-0-1 cells were stained with agrin mAb and then fixed and photographed. Agrin expression was even higherin unpermeabilized cells, strongly concentrated on the cell surface and extended past the detectable cell boundary. Scale bar, 25 mm.

This indicates a trend toward cessation of growth, rather for 2 hr with continuous agitation on a rotating platform.After removal of loosely attached cells with brief washing,than a mere slowing. The effect of stopping growth at these

longer times was more dramatic with agrin-0-expressing the coverslips were examined and the cells were countedto determine the number of neurons that remained attachedCHO cells than that with agrin-19-expressing cells (not

shown). The diminished response with agrin-19-1 cells in to control vs agrin-expressing cells. Attachment to agrin-0-1 CHO cells was nearly twofold higher than to DR-2 con-this one experiment may reflect a real difference, but we

cannot rule out that the more heterogeneous agrin expres- trols for all neuronal types tested (Fig. 10A). Since attach-ment to DR-2 controls is suitable for extensive neurite out-sion in the agrin-19-1 cells may have produced the differ-

ence in these longer-term cultures. It is difficult to ensure growth, the increased cell attachment with agrin expressionis all the more significant. For comparison, attachment ofcontinuous interactions of growing neurites on monolayers

with variations in expression. ciliary ganglion neurons under these conditions to laminin,fibronectin and pdl was only 1.2- to 1.4-fold higher (Fig.A candidate stop signal specific for motoneurons has been

identified as the tripeptide, LRE in S-laminin (Porter et al., 10B). This suggests that agrin does not slow neurite exten-sion by creating a nonpermissive substrate for general adhe-1995). Two LRE sequences appear in rat and mouse agrin

near the N-terminal half. A construct was used, R4HA, that sion. Instead, agrin can promote attachment to levels be-tween that seen for ECM proteins and control CHO cells,includes this domain, but is missing the acetylcholine re-

ceptor clustering activity in the C-terminal portion of agrin. both of which can promote robust neurite outgrowth. Thisresult also shows that agrin is not anchoring cells so tightlySensory neurons were analyzed for their response to CHO

cells expressing this construct (Fig. 9). Unlike the other that neurite outgrowth is completely prevented by toomuch adhesion.agrin constructs, sensory neuron outgrowth on R4HA-ex-

pressing CHO cells was indistinguishable from that on con-trol cells.

The attachment of all three nerve cell types to agrin-0-1 DISCUSSIONmonolayers was measured to determine if the reduced neu-rite outgrowth might be due to decreased adhesiveness. Results presented here show that in addition to agrin’s

known capacity to aggregate acetylcholine receptors, thisBCECF-AM, a vital fluorescent dye, was used to label neu-ronal cultures. Labeled neurons were added to live, nearly ECM protein can slow the extent of process outgrowth in

vitro from two classes of motor neurons and from DRG-confluent cultures of CHO cultures, and allowed to attach


AID DB 8435 / 6X15$$$$21 12-06-96 09:37:00 dba


FIG. 5. Neurite length distribution for ventral neural tube neurons growing on agrin and non-agrin-expressing CHO substrates showthat agrin slows the rate of neurite outgrowth. Neurite lengths from multiple experiments were measured and normalized to each otheras described under Materials and Methods to allow comparison of experiments performed on different days. Open symbols are used forone pair of results and filled symbols for another set of experiments performed in a separate trial. (Distributions that shift the data set tothe left indicate a reduction in the average length of the population of neurites.) Neurites extending on agrin-0-1-expressing (A and C,circles) and agrin-19-1-expressing (B and D, triangles) CHO cells were compared to two drug-resistant subclones, DR-1 (A and B, squares)and DR-2 (C and D, diamonds). Neurites were measured after 48 hr growth and only those neurites growing on agrin-positive cells weremeasured. Agrin-19-1 and agrin-0-1 slowed growth to a comparable degree (for summary, see Table 1).

derived sensory neurons. Agrin also promotes attachment terminal laminin-G and EGF domains, but containing thefollistatin domains and the LRE sequences. Agrin appearedof all these cells, and eventually may completely stop

growth. It is apparent that agrin is a cell surface protein to prevent growth optimally when expressed on the surfaceof cells.that is eventually deposited in a matrix-like form on the

underlying and surrounding substrate. This suggests as well Our conclusions are in general agreement with thosedrawn from another recent study that has shown that ciliarythat it could operate in vivo on myofibers as an acellular

environment, arresting axon growth. Unlike their differen- ganglion neurons growing on fibronectin are deflected whenthey confront the boundary of a CHO cell expressing agrin-tial effects on acetylcholine receptor clustering, most of

these effects were similar for the two full-length isoforms, 19 or -O (Campagna et al., 1995). Yet the present studyextends these earlier findings in a number of importantagrin-0 or -19, although agrin-0 was more effective in com-

pletely stopping outgrowth. No effect on neurite outgrowth ways. First, we were able to show that agrin inhibits notonly motoneuron outgrowth but also sensory outgrowth.was seen with a truncated agrin isoform, missing the C-


AID DB 8435 / 6X15$$$$21 12-06-96 09:37:00 dba

30 Chang et al.

FIG. 6. Agrin alters the outgrowth of only those neurites growing directly on agrin-positive cells, and not on adjacent agrin-negativecells. Neurites from ventral neural tube motoneurons were measured as described in the legend to Fig. 5. Only neurons with neuritesfully contained on agrin-19-negative (A and B, open, inverted triangles) or agrin-19-positive (A, filled triangles) were measured. Neuritelengths on agrin-19-negative cells were longer than those on agrin-19-positive cells, and indistinguishable from that of a drug-resistantcontrol, DR-1 (B, open squares).

Sensory neurons were tested only for their relative adhesion each agrin subtype, we were able to rule out any additionaleffects produced by heterogeneity in cell spreading capacityto agrin in the earlier study. Agrin seems to affect multiple

classes of neurons, suggesting a role outside of the motor or clonal variations selected for during their isolation. An-other novel finding in the present study is that by 3 daysendplate. More significantly, the present study demon-

strates that this inhibition is evidenced when axons are neurite outgrowth is essentially arrested. While this conclu-sion is based on one experiment, it suggests that agrin doesgrown on a continuous monolayer of agrin-expressing cells

without the potential confounding influence of a physical not merely slow growth, but may in fact stop growth. Fi-nally, our studies do not support their finding that ciliarybarrier created by a cell boundary. Such a physical barrier

was evidenced in our studies if the test cells, both agrin and ganglion neurons are more adherent than sensory cells toagrin monolayers. Instead, we found that attachment of allnon-agrin-expressing, were not well spread prior to addition

of neurons. This lack of spreading was impossible to avoid classes of neurons to agrin was roughly equivalent, and sim-ilar to the attachment seen on several ECM substrates. Thein some sublines used. Since both reports used the same

agrin-expressing CHO cells, it is of significant concern. We source of this difference in our two studies is not readilyapparent.also measured considerably more variability in agrin expres-

sion, especially for agrin-19 cells, motivating our double Target recognition and synapse formation by motoneu-rons requires the temporal orchestration of pre- and post-staining of all assays, to establish concretely that neurons

were interacting with agrin-expressing cells. synaptic signals and events (Hall and Sanes, 1989; Bixby,1985; Nastuk and Fallon, 1993; Connor and Smith, 1994).Of additional concern in the Campagna et al. (1995) assay

is that others have shown that neurons will remain on a The motoneuron needs to find its target, cease growing atthat site, and organize a synaptic release site. The musclelaminin substrate when confronted with the choice be-

tween laminin and surrounding areas of collagen IV (Gund- cell may create the signals that direct this innervation andthen in turn assemble the necessary postsynaptic machin-erson, 1987). Yet collagen IV alone promotes excellent neu-

rite outgrowth, although not as robustly as laminin, sug- ery to stabilize this specialization in response to presynapticactivity. The results presented here suggest that agrin couldgesting that neurons remain on laminin, unless provided

with an even better option. It is hard to conclude from the serve a fundamental role in synaptogenesis by providing astop signal that allows the eventual anchoring of the grow-assay design of Campagna et al. (1995) whether agrin is

inhibitory or only not as favorable as fibronectin, the sub- ing axon to its target tissue. Stabilization of the contact sitewould then allow neural forms of secreted agrin to clusterstrate chosen in their report. The monolayer experiments

performed in the present study, while more difficult to per- receptors underneath the release site. Consistent with thismodel, motoneuron axons in mice deficient for both theform, do eliminate this concern since cell/substrate bound-

aries are absent; therefore, neurons do not need to make a neural and aneural forms of agrin [agrin-0 and 19 (Gautamet al., 1996)] or for MuSK, a tyrosine kinase receptor thatchoice.

By assaying additional control lines and separate lines for agrin may act through (DeChiara et al., 1996; Glass et al.,


AID DB 8435 / 6X15$$$$21 12-06-96 09:37:00 dba


FIG. 7. Sensory neuron outgrowth was also reduced on agrin-0 and agrin-19 CHO monolayers. Neurites extending on agrin-0-expressing(A, circles) and agrin-19-positive-expressing (B, triangles) CHO cells were compared to DR-2 control cells (A and B, diamonds). Agrin-19and agrin-0 slowed growth to a comparable degree. Outgrowth on agrin-19-negative regions (C and D, inverted triangles) was similar tooutgrowth on DR-2 cells (C, diamonds), while agrin-19-positive and agrin-19-negative regions showed very different distributions (D).Neurites were measured after 24 hr growth in the presence of NGF. Open symbols are used for one pair of results and filled symbols foranother set of experiments performed in a separate trial except for panel D.

1996), show exuberant growth, extending considerable dis- Agrin levels are indeed low in growing neurons, during bothdevelopment and regeneration, and therefore would not betances beyond their normal synaptic target region. Along

with other generalized defects in synapse differentiation, expected to slow their own growth. It is also possible thatyet to be described agrin receptors involved in this responsethese mice demonstrate that agrin in vivo may provide a

stop signal for at least developing motoneurons. are not expressed by growing axons until they reach themotor endplate, thereby avoiding any inhibitory influence.Agrin-Y0Z0 and Y4Z0 are secreted by myofibers and other

nonneuronal cells early in development (Rupp et al., 1991; All the test neurons used in the present study were derivedfrom embryos in which these populations would have al-Hoch et al., 1993). They are therefore in a position to arrest

neurite outgrowth prior to innervation. Both of these forms ready reached their targets. In this context it is worth notingthat younger neurons failed to show as significant a re-have little or no aggregating activity. Why the neural form

of agrin that can slow growth is synthesized (Magill-Solc sponse in our assays (D.C. and M.J.I., unpublished observa-tion.) It is also possible that additional interactions in theand McMahan, 1990) and released by motor neurons in cul-

ture (Cohen et al., 1995), potentially arresting their own growing environment of the nerve override this negativeinfluence and favor growth until the targets are reached.outgrowth, is unresolved. However, the agrin message has

not been seen in motor neurons before they reach their Both agrin isoforms reduced outgrowth when expressedon the surface of cells in our studies, despite their differ-targets (Ma et al., 1995). Postganglionic axotomy, which

induces nerve sprouting, reduces agrin-19 levels in regener- ences in AchR-clustering activity. It is possible that thislack of differentiated functions for the two agrin isoformsating chick ciliary ganglion neurons (Thomas et al., 1995).


AID DB 8435 / 6X15$$$$21 12-06-96 09:37:00 dba

32 Chang et al.

FIG. 8. Neurite growth was arrested on agrin between 48 and 72 hr. Separate cultures of sensory neurons were prepared under identicalconditions, with one set fixed and analyzed at 48 hr (open squares) and the other at 72 hr (filled squares). Neurites continued to grow oncontrol cells (B), but stopped extending on agrin-0-expressing cells during the same time period (A).

on neurite outgrowth is due to concentrations of agrin in expressed proteins. It is possible that soluble agrins, atknown concentrations, are needed to distinguish their func-our assays that might exceed levels needed to separate their

functions. Transfected cells allow a more physiological as- tions. However, our observation that neurite lengths onagrin-19-negative regions of the substrate were longer thansessment of agrin’s function as a target-derived signal, since

they mimic the in vivo situation in which agrin is presented on immediately adjacent agrin-19-expressing regions sug-gests that agrin must act directly on cells, possibly anchoredon the surface of a cell. However, these assays do suffer the

drawback that it is difficult to establish concentrations of

FIG. 10. (A) Attachment of motor, ciliary ganglion, and DRG neu-rons to agrin-0-expressing CHO cells is higher than that to controlCHO cells. Neurons were prepared and labeled with a fluorescentFIG. 9. Outgrowth of sensory neurons on R4HA (diamond sym-

bols) was identical to that of a control line (/’s and /’s in squares). marker as described under Materials and Methods, and then addedto confluent monolayers of agrin-0-1 CHO cells (open bars) andDuplicate sets of trials on CHO cells expressing R4HA and DR-2

revealed no effect produced by this truncated construct. R4HA is DR-1 control cells (filled bars). (B) Attachment of ciliary ganglionderived neurons to coverslips coated with pdl, laminin, or fibronec-missing the C-terminal half of agrin that contains a putative aggre-

gation domain. Cells were assayed as in Fig 5. Open diamond sym- tin was compared as well and shows roughly equal levels to thatof agrin-0 with diminished binding to DR controls. Error bars reflectbols are used for one pair of results and filled symbols for another

set of experiments performed in a separate trial. the range of values from duplicate sets of experiments.


AID DB 8435 / 6X15$$$$21 12-06-96 09:37:00 dba


to an ECM. The lack of agrin inhibition evidenced on depos- that includes both the osteonectin and LRE domains, noeffect on normal neurite outgrowth was seen, indicatingited material argues that agrin may need to remain mobile,

possibly through linkage to plasma membrane receptors. that these domains are not sufficient for the inhibitory func-tion of agrin. Further experiments are designed to try toThis hypothesis suggests that purified agrin isoforms would

show little activity when anchored as a substrate in vitro. identify the region slowing axon growth, either in concertwith LRE domains or in isolation and to determine whetherAt present we can conclude that both agrins are equally

potent at slowing growth when expressed on the surface of the affect bears similarity to that observed for osteonectin.In conclusion, agrin may act as a stop signal at bothcells.

S-laminin, another ECM protein concentrated at synap- developing and regenerating motor endplates. Agrin ispresent at the right times and in the appropriate regionstic sites, can also alter neurite outgrowth and has been

suggested to provide the stop signal at developing and to effect such a response. Interestingly, partial denervationof nerve terminals or paralysis with botulinum toxin isregenerating synapses. However, S-laminin’s inhibitory

effects are only seen when used to compete for the neurite known to induce sprouting of intact fibers that grow to-ward the denervated synapses (Thompson and Jansen,outgrowth-promoting activity of laminin (Porter et al.,

1995); S-laminin by itself can support growth. Equally 1977). Recent studies have shown that this treatment firstinduces sprouting of glial processes and glial cell migra-problematic for a role in synaptogenesis is the absence of

S-laminin from developing synapses. Its abundance and tion from the synaptic terminal Schwann cells. Presum-ably this growth provides a permissive substrate for thepersistence at denervated sites argues for a role more lim-

ited to regeneration. Agrin, present during both these secondary response of neurite growth out of the synapticregion (Son and Thompson, 1995). Such a response by thetime frames, is therefore a good candidate to serve as a

stop signal at both developing and regenerating synapses. glial cells would be necessary if axons were anchored andinhibited from growing by agrin. Masking of agrin or S-In addition, our studies indicate that agrin can fully stop

growth, while S-laminin has been shown to only delay laminin by the glial cells at these mature synapses wouldallow growth, an hypothesis we are currently addressinggrowth on laminin substrates for 1 or 2 hr.

The structural characterization of agrin reveals a variety using the cultures described here.of motifs, consistent with an expanded role for agrin (re-viewed in Rupp et al., 1992; Patthy and Nikolics, 1993). TheC-terminal half contains three laminin-G domains between ACKNOWLEDGMENTSEGF motifs 2 and 3. Another region bracketed by the EGFrepeats is homologous in both agrin and slit, a Drosophila

The authors acknowledge the help of Teresa Huang and Staceyneuronal guidance molecule (Rothberg and Artavanis-Tsa-Combes for initial efforts in developing the neurite outgrowth

konas, 1992). Finally, a synaptic cell surface protein, neu- assays used, David Bentley for use of his microscope facility forrexin, shares homology to the laminin-G domain found in measuring neurite lengths, and Karen Zachow for comments onboth agrin and slit (Ushkaryov et al., 1992). No function for the manuscript. This work was funded by NIH RO1NS31366-01A1any of these domains in agrin, other than heparin binding, and the ALS Association to M.J.I., R.H.S. is an investigator in the

HHMI.has been established. However, homology to those proteinsinvolved in pathfinding suggests a role for agrin in axoninteractions, consistent with the results found in the pres-ent study. REFERENCESAll agrin isoforms contain nine follistatin domains in theN-terminal half that may allow the binding of certain

Banker, G., and Goslin, K. (1991). In ‘‘Culturing Nerve Cells’’ (G.growth factors such as TGF-b or PDGF. These N-terminalBanker and K. Goslin, Eds.), pp. 119–136. MIT Press, Cambridge,

repeats are also related to Kazal-type protease inhibitors MA.(Patthy and Nikolics, 1993; review in Rupp et al., 1991) that Biroc, S. L., Payan, D. G., and Fisher, J. M. (1993). Isoforms of agrinhave been shown in agrin to inhibit trypsin and plasmin are widely expressed in the developing rat and may function as(Biroc et al., 1993). The nine follistatin repeats in the N- protease inhibitors. Dev. Brain Res. 75, 119–129.

Bixby, J. L., Lilien, J., and Reichardt, L. F. (1988). Identification ofterminus also show homology to osteonectin (or SPARC)the major proteins that promote neuronal process outgrowth onprotein, an inhibitor of cell attachment when added to cer-Schwann cells in vitro. J. Cell Biol. 107, 353–361.tain nonneuronal cells on ECM substrates. Osteonectin dis-

Bixby, J. L., and Reichardt, L. F. (1985). The expression and localiza-rupts attachment by dissolving or dissociating integrin-de-tion of synaptic vesicle antigens at neuromuscular junctions inpendent focal contact sites (Murphy-Ullrich et al., 1995).vitro. J. Neurosci. 5, 3070–3080.However, unlike osteonectin, agrin can promote cell attach-

Bloch-Gallego, E., Huchet, M., el M’Hamdi, H., Xie, F. K., Tanaka,ment and no role for integrins in agrin interactions has been H., and Henderson, C. E. (1991). Survival in vitro of motoneuronsreported. Two LRE tripeptides, competitive inhibitors of S- identified or purified by novel antibody-based methods is selec-laminin-dependent neurite extension, reside in the N-ter- tively enhanced by muscle-derived factors. Development 111,minal domain of rat agrin (although chick agrin does not 221–232.

Burns, F. R., von Kanner, S., Guy, L., Raper, J. A., Kamholz, J., andhave a LRE). In our assay of the rat agrin construct R4HA


AID DB 8435 / 6X15$$$$21 12-06-96 09:37:00 dba

34 Chang et al.

Chang, S. (1991). DM-GRASP, a novel immunoglobulin super- ants are differentially regulated in developing chick embryo spi-family axonal surface protein that supports neurite extension. nal cord and sensory ganglia. J. Neurobiol. 26, 585–597.Neuron 7, 209–220. Magill-Solc, C., and McMahan, U. J. (1990). Synthesis and transport

Campagna, J. A., Ruegg, M. A., and Bixby, J. L. (1995). Agrin is a of agrin-like molecules in motor neurons. J. Exp Biol. 153, 1–10.differentiation-inducing ‘‘stop signal’’ for motoneurons in vitro. McMahan, U. J. (1990). The agrin hypothesis. Cold Spring HarborNeuron 15, 1365–1374. Symp. Quant. Biol. 55, 407–418.

Campanelli, J. T., Hoch, W., Rupp, F., Kreiner, T., and Scheller, McMahan, U. J., and Wallace, B. G. (1989). Molecules in basal lam-R. H. (1991). Agrin mediates cell contact-induced acetylcholine ina that direct the formation of synaptic specialization at neuro-receptor clustering. Cell 67, 909–916. muscular junctions. Dev. Neurosci. 11, 227–247.

Cohen, M. W., Moody-Corbett, F., and Godfrey, E. W. (1995). For- Murphy-Ullrich, J. E., Lane, T. F., Pallero, M. A., and Sage, E. H.mer neuritic pathways containing endogenous neural agrin have (1995). SPARC mediates focal adhesion disassembly in endothe-high synaptogenic activity. Dev. Biol. 167, 458–468. lial cells through a follistatin-like region and the Ca(2/)-binding

Connor, E. A., and Smith, M. A. (1994). Retrograde signaling in the EF-hand. J. Cell. Biochem. 57, 341–350.formation and maintenance of the neuromuscular junction. J. Nastuk, M. A., and Fallon, J. R. (1993). Agrin and the molecularNeurobiol. 25, 722–739. choreography of synapse formation. Trends Neurosci. 16, 72 –76.

deChiara, T. M., Bowen, D. C., Valenzuela, D. M., Simmons, M. V., Nishi, R., and Berg, D. K. (1977). Dissociated ciliary ganglion neu-Poueymirou, W. T., Thomas, S., Kinetz, E., Compton, D. L., Ro- rons in vitro: Survival and synapse formation. Proc. Natl. Acad.jas, E., Park, J. S., Smith, C., DiStefano, P. S., Glass, D. J., Burden, Sci. USA 74, 5171–5175.S. J., and Yancopoulos, G. D. (1996). The receptor tyrosine kinase Nishi, R., and Berg, D. K. (1981). Two components from eye tissueMuSK is required for neuromuscular junction formation in vivo. that differentilly stimulte the growth and development of ciliaryCell 85, 501–512. ganglion nurons in cell culture. J. Neurosci. 1, 505–513.

Doherty, P., Williams, E., and Walsh, F. S. (1995). A soluble chime- Nitkin, R. M., and Rothschild, T. C. (1990). Agrin-induced reorgani-ric form of the L1 glycoprotein stimulates neurite outgrowth. zation of extracellular matrix components on cultured myotubes:Neuron 14, 57–66. Relationship to AChR aggregation. J. Cell Biol. 111, 1161–1170.

Fallon J. R., and Gelfman, C. E. (1989). Agrin related molecules arePatthy, L., and Nikolics, K. (1994). Agrin-like proteins of the neuro-

concentrated at acetylcholine receptor clusters in normal andmuscular junction. Neurochem. Int. 24, 301–316.aneural developing muscle. J. Cell Biol. 108, 1527–1535.

Porter, B. E., Weis, J., and Sanes, J. R. (1995). A motoneuron-selec-Ferns, M., Hoch, W., Campanelli, J. T., Rupp, F., Hall, Z. W., andtive stop signal in the synaptic protein S-laminin. Neuron 14,Scheller, R. H. (1992). RNA splicing regulates agrin-mediated549–559.acetylcholine receptor clustering activity on cultured myotubes.

Rothberg, J. M., and Artavanis-Tsakonas, S. (1992). Modularity ofNeuron 8, 1079–1086.the slit protein: Characterization of a conserved carboxy-terminalFerns, M. J., Campanelli, J. T., Hoch, W., Scheller, R. H., and Hall,sequence in secreted proteins and a motif implicated in extracell-Z. (1993). The ability of agrin to cluster AChRs depends on alter-ular protein interactions. J. Mol .Biol. 227, 367–370.native splicing and on cell surface proteoglycans. Neuron 11,

Ruegg, M. A., Tsim, K. W., Horton, S. E., Kroger, S., Escher, G.,491–502.Gensch, E. M., and McMahan, U. J. (1992). The agrin gene codesGautam, M., Noakes, P. G., Moscoso, L., Rupp, F., Scheller, R. H.,for a family of basal lamina proteins that differ in function andMerlie, J. P., and Sanes, J. R. (1996). Defective neuromusculardistribution. Neuron 8, 691–699.synaptogenesis in agrin-deficient mutant mice. Cell 85, 525–535.

Rupp, F., Payan, D. G., Magill-Solc, C., Cowan, D. M., and Scheller,Glass, D. J., Bowen, D. C., Stitt, T. N., Radziejewski, C., Bruno,R. H. (1991). Structure and expression of a rat agrin. Neuron 6,J., Ryan, T. E., Gies, D. R., Shah, S., Mattsson, K., Burden, S. J.,811–823.diStefano, P. S., Valenzuela, D. M., DeChiara, T. M., and Yanco-

Sanes, J. R., and Chiu, A. Y. (1983). The basal lamina of the neuro-poulos, G. D. (1996). Agrin acts via a MuSK receptor complex.muscular junction. Cold Spring Harbor Symp. Quant. Biol. 48,Cell 85, 513–523.667–678.Gunderson, R. W. (1987). Response of sensory neurites and growth

Sanes, J. R., Marshall, L. M., and McMahan, U. J. (1978). Reinnerva-cones to patterned substrata of laminin and fibronectin in vitro.tion of muscle fiber basal lamina after removal of myofibers. J.Dev. Biol. 121, 423–431.Cell Biol. 78, 176 –198.Hall, Z. W., and Sanes, J. R (1993). Synaptic structure and develop-

Shadiack, A. M., and Nitkin, R. M. (1991). Agrin induces alpha-ment: The neuromuscular junction. Cell 72, 99–121.actinin, filamin, and vinculin to co-localize with AChR clustersHoch, W., Campanelli, J. T., Harrison, S., and Scheller, R. H. (1994).on cultured chick myotubes. J. Neurobiol. 22, 617–628.Structural domains of agrin required for clustering of nicotinic

Smith, M. A., and O’Dowd, D. K. (1994). Cell-specific regulation ofacetylcholine receptors. EMBO J. 13, 2814–2821.agrin RNA splicing in the chick ciliary ganglion. Neuron 12,Hoch, W., Ferns, M., Campanelli, J. T., Hall, Z. W., and Scheller,795–804.R. H. (1993). Developmental regulation of highly active alterna-

Son, Y. J., and Thompson, W. J. (1995). Nerve sprouting in muscletively spliced forms of agrin. Neuron 11, 479–490.is induced and guided by processes extended by Schwann cells.Ignatius, M. J., and Reichardt, L. F. (l988). Identification of a neu-Neuron 14, 133–141.ronal laminin receptor: An Mr 200K/120K integrin heterodimer

Stone, D. M., and Nikolics, K. (1995). Tissue and age-specific ex-that binds laminin in a divalent cation-dependent manner. Neu-pression patterns of alternatively spliced agrin mRNA transcriptsron 1, 713–725.in embryonic rat suggest novel developmental roles. J. Neurosci.Ma, E., Morgan, R., and Godfrey, E. W. (1994). Distribution of agrin15, 6767–6778.mRNAs in the chick embryo nervous system. J. Neurosci. 14,

2943–2952. Thomas, W. S., Jacob, M. H., O’Dowd, D. K., and Smith, M. A.(1995). Agrin gene expression in ciliary ganglion neurons follow-Ma, E., Morgan, R., and Godfrey, E. W. (1995). Agrin mRNA vari-


AID DB 8435 / 6X15$$$$21 12-06-96 09:37:00 dba


ing preganglionic denervation and postganglionic axotomy. Dev. Tsim, K. W., Ruegg, M. A., Escher, G., Kroger, S., and McMahan,Biol. 168, 662–669. U. J. (1992). cDNA that encodes active agrin. Neuron 8, 677–689.

Thomas, W. S., O’Dowd, D. K., and Smith, M. A. (1993). Develop- Ushkaryov, Y. A., Petrenko, A. G., Geppert, M., and Sudhof, T. C.mental expression and alternative splicing of chick agrin RNA. (1992). Neurexins: Synaptic cell surface proteins related to theDev. Biol. 158, 523–535. alpha-latrotoxin receptor and laminin. Science 257, 50–56.

Thompson, W., and Jansen, J. K. S. (1977). The extent of sproutingReceived for publication June 6, 1996of remaining motor units in partly denervated immature and

adult rat soleus muscle. Neuroscience 2, 523–535. Accepted October 3, 1996


AID DB 8435 / 6X15$$$$21 12-06-96 09:37:00 dba

Documents

Agrin Inhibits Neurite Outgrowth but Promotes Attachment of Embryonic Motor and Sensory Neurons