Cell, Vol. 85, 501–512, May 17, 1996, Copyright 1996 by Cell Press

The Receptor Tyrosine Kinase MuSK Is Requiredfor Neuromuscular Junction Formation In Vivo

Thomas M. DeChiara,* David C. Bowen,* been shown to be essential for development of normalblood vessels (Fong et al., 1995; Sato et al., 1995; Sha-David M. Valenzuela,* Mary V. Simmons,*laby et al., 1995). We have recently described the ratWilliam T. Poueymirou,* Susan Thomas,† Erika Kinetz,†and human versions of a receptor-like tyrosine kinaseDebra L. Compton,* Eduardo Rojas,* John S. Park,*that appears to be restricted to skeletal muscle in itsCynthia Smith,† Peter S. DiStefano,* David J. Glass,*expression, and termed it MuSK for muscle-specific ki-Steven J. Burden,† and George D. Yancopoulos*nase (Valenzuela et al.,1995); a very close relative, which*Regeneron Pharmaceuticals, Inc.may represent the mouse MuSK ortholog, has been777 Old Saw Mill River Roadtermed Nsk2 (Ganju et al., 1995).Tarrytown, New York 10591

MuSK is first detectable in proliferating myoblasts but†Molecular Neurobiology Programis dramatically up-regulated as myoblasts exit the cellSkirball Institutecycle and fuse to form multinucleated myotubes (Valen-New York University Medical Centerzuela et al., 1995). In vivo, innervation by motor axonsNew York, New York 10016shortly follows myoblast fusion (Hall and Sanes, 1993).Functional innervation depends on the formation of ahighly specialized synaptic connection, termed the neu-Summaryromuscular junction (NMJ), between the incoming motornerve and the muscle fiber (Hall and Sanes, 1993). AsFormation of neuromuscular synapses requires a se-muscle matures, MuSK becomes highly localized to aries of inductive interactions between growing motorsmall region of the myofiber surface that comprises theaxons and differentiating muscle cells, culminating inpostsynaptic side of the neuromuscular junction (Valen-the precise juxtaposition of a highly specialized nervezuela et al., 1995). Thus, MuSK appears to be strategi-terminal with a complex molecular structure on thecally positioned to recognize and to be activated by apostsynaptic muscle surface. The receptors and sig-nerve-derived ligand and, in turn, to relay a key signalnaling pathways mediating these inductive interac-to the muscle. Indeed, a potential ortholog of MuSK hastions are not known. We have generated mice withbeen isolated from the electric organ of the electric raya targeted disruption of the gene encoding MuSK, aTorpedo californica, a tissue homologous to skeletalreceptor tyrosine kinase selectively localized to themuscle and highly enriched in synapses, using a strat-postsynaptic muscle surface. Neuromuscular syn-egy designed to identify tyrosine kinases that play im-apses do not form in these mice, suggesting a failureportant roles at the NMJ (Jennings et al., 1993). Thus,

in the induction of synapse formation. Together withit appears as if MuSK and its distant orthologs have

the results of an accompanying manuscript, our find-maintained crucial and particular roles at the NMJs ofings indicate that MuSK responds to a critical nerve-evolutionarily distant species.derived signal (agrin), and in turn activates signaling

It has long been known that important signals arecascades responsible for all aspects of synapse for-

exchanged across the NMJ (Nitkin et al., 1987; Hall andmation, including organization of the postsynaptic

Sanes, 1993; Bowe and Fallon, 1995; Burden et al., 1995;membrane, synapse-specific transcription, and pre-

Sanes, 1995). These signals include the chemical trans-synaptic differentiation. mitter acetylcholine, which is released from vesicles in

the nerve terminal, recognized by acetylcholine recep-Introduction tors (AChRs) on the muscle, and ultimately results in

electrical activation and contraction of the muscle. Mus-Intercellular communication is often mediated by recep- cle also provides neurotrophic factors that support sur-tors, on the surface of one cell, that recognize and are vival of motor neurons (DeChiara et al., 1995), and theactivated by specific protein ligands released by other nerve may in turn provide myotrophic factors that main-cells. Members of one class of cell surface receptors, tain muscle mass (Helgren et al., 1994). Reciprocal sig-known as receptor tyrosine kinases (RTKs), are charac- naling interactions are also critical both for the formationterized by having a cytoplasmic domain containing in- and maintenance of the neuromuscular junction itself.trinsic tyrosine kinase activity (Schlessinger and Ullrich, Such signals regulate recognition of nerve-to-muscle1992). This kinase activity is regulated by the binding contact, arrest the growth of the incoming nerve ending,of a cognate ligand to the extracellular portion of the and induce formationof a highly specialized nerve termi-receptor. The few RTKs known to be expressed in cell nal marked by a polarized arrangement of synaptic vesi-type–specific fashions have been shown to play roles cles and active zones. Simultaneously, precisely juxta-critical for the growth and differentiation of those cell posed with respect to the nerve terminal, a complextypes. For example, members of the neural-specific Trk molecular apparatus forms on the muscle membrane.family of RTKs—which recognize nerve growth factor This specialized postsynaptic structure, termed the mo-and its relatives (Barbacid, 1993; Glass and Yanco- tor endplate, comprises a tiny patch onthe muscle mem-poulos, 1993)—are absolutely required for the survival brane that is characterized by a dense clustering ofand development of discrete neuronal subpopulations particular proteins; some of these may receive nerve-(Snider, 1994). Similarly, several RTKs that are specifi- derived signals, as AChRs are known to do, while others

may be involved in creating the molecular scaffold forcally expressed in vascular endothelial cells have all

Cell502

this postsynaptic specialization. Signals produced by AChRs, and inhibitors of tyrosine phosphorylation blockthe nerve induce postsynaptic clusters by at least two agrin-mediated clustering (Wallace et al., 1991; Qu andmechanisms. First, these signals can induce redistribu- Huganir, 1994; Wallace, 1994, 1995; Ferns et al., 1996).tion of pre-existing molecules that are initially expressed Intriguing recent findings have revealed that agrin canthroughout the myofiber, and second, they can induce directly bind to a-dystroglycan, an extrinsic peripherallocalized transcription of specific genes only by subsy- membrane protein that is attached to the cell surfacenaptic nuclei underlying the NMJ. Between the nerve by covalent linkage to b-dystroglycan, which in turnterminal and the motor endplate is a narrow synaptic couples to the intracellular cytoskeletal scaffold via ancleft containing a complex basal lamina. This basal lam- associated protein complex (Bowe et al., 1994; Campa-ina is distinguished from the adjacent extracellular ma- nelli et al., 1994; Gee et al., 1994; Sealock and Froehner,trix by the accumulation of a number of proteins, such 1994; Sugiyama et al., 1994). Extrasynaptically, the dys-as acetylcholinesterase and s-laminin; as noted below, troglycan complex binds laminin on its extracellular facethe synaptic basal lamina also serves as a reservoir and couples to the actin scaffold via a spectrin-like mol-for signaling molecules exchanged between nerve and ecule known as dystrophin. At the synapse however,muscle. agrin (via its own laminin-like domains) may be able to

While the reciprocal interactions between nerve and substitute for laminin, whereas utrophin (a dystrophin-muscle have been intensively explored for decades, related protein) replaces dystrophin as the link to actinmany questions still remain concerning the precise na- (reviewed in Bowe and Fallon, 1995). The dystroglycanture of the signals involved in formation of the NMJ. complex coclusters with AChRs in response to agrin inThe realization that empty sheaths of the synaptic basal vitro, and components of this complex are concentratedlamina could induce formation of both nerve terminal at the endplate in vivo (reviewed in Bowe and Fallon,specializations and motor endplates suggested that key 1995). Recent evidence suggests that a 43 kDa cyto-signaling molecules might be embedded in the extracel- plasmic protein, known as rapsyn, anchors AChRs to alular matrix (Sanes et al., 1978; Burden et al., 1979; subsynaptic cytoskeleton complex, probably via inter-McMahan and Slater, 1984; Kuffler, 1986). Indeed, re- actions with the dystroglycan complex (Cartaud andcent findings indicate that a protein discovered for its Changeux, 1993; Apel et al.,1995). Gene disruption stud-AChR-inducing activity and thus termed ARIA (Jessell ies reveal that rapsyn is absolutely necessary for cluster-et al., 1979; Falls et al., 1990, 1993), which can increase ing of AChRs, as well as of the dystroglycan complexthe expression of several of the AChR subunit genes (Gautam et al., 1995). However, other aspects of NMJ(Harris et al., 1989; Martinou et al., 1991; Chu et al., formation, involving presynaptic differentiation and syn-1995; Jo et al., 1995), is localized to the synaptic basal apse-specific transcription, are seen in mice lackinglamina (Goodearl et al., 1995; Jo et al., 1995). Molecular rapsyn (Gautam et al., 1995).cloning has revealed that ARIA corresponds to a factor Despite the findings that agrin can bind directly toalternatively referred to as neuregulin, NDF, heregulin a-dystroglycan, and that AChRs and the dystroglycanor glia growth factor, and binds to the ErbB family of complex are linked and cocluster in response to agrin,RTKs (Carraway and Burden, 1995). Interestingly, neu- the role of dystroglycan as an agrin receptor remainsregulin production has been demonstrated in motor neu- unclear (Sealock and Froehner, 1994; Ferns et al., 1996).rons, and neuregulin receptors, ErbB3 and ErbB4, have Dystroglycan could be directly involved in activatingrecently been localized to the motor endplate, support- signaling pathways that appear to be required for clus-ing the idea that nerve-derived neuregulin provides an tering, such as those involving tyrosine phosphorylation,important signal to muscle that regulates transcription by an unknown mechanism (for example, via associationfrom subsynaptic nuclei (Altiok et al., 1995; Moscoso et

with a cytoplasmic tyrosine kinase). Alternatively, dys-al., 1995; Zhu et al., 1995).

troglycan could be involved in couplings of agrin notAnother protein, known as agrin, was isolated from

only to AChRs but to a novel signaling receptor. On thethe synaptic basal lamina based on its ability to cause

other hand, it remains possible that dystroglycan doesredistribution of pre-exisiting AChRs into clusters on thenot play an active or required role in initiating clusteringsurface of cultured myotubes (Godfrey et al.,1984; Ruppand is merely among an assortment of postsynapticet al., 1991; Tsim et al., 1992). In contrast to neuregulin,molecules that undergo clustering.agrin does not appear to regulate AChR expression.

To determine the role of MuSK during skeletal muscleHowever, agrin causes the clustering of a number ofdevelopment, we generated mice homozygous for asynaptic components, along with AChRs, in culturedMuSK gene disruption. While it appears that early skele-myotubes (Wallace, 1989). A variety of data are consis-tal muscle development proceeds in these mice, we findtent with the notion that agrin also acts in vivo to induceno evidence of NMJs in these mice, suggesting a failureand maintain the postsynaptic membrane specializa-in the induction of synapse formation. In an accompa-tion. Most important among these are the findings thatnying paper (Glass et al., 1996 [this issue of Cell]), wethe most active forms of agrin are exclusively made bydemonstrate that MuSK serves as a signaling compo-neurons and are deposited in the synaptic basal laminanent of a complex receptor for neural forms of agrin.(Ruegg et al., 1992; Ferns et al., 1993; Hoch et al., 1993),Thus agrin, acting via the MuSK receptor, apparentlyand that antibodies to agrin block nerve-induced clus-activates signaling cascades critical for all aspects oftering of AChRs oncultured myotubes (Reist etal., 1992).NMJ formation in vivo, including organization of theThe precise mechanism of action of agrin remains apostsynaptic membrane, synapse-specific transcrip-mystery (Sealock and Froehner, 1994; Ferns et al., 1996).

Agrin is known to induce tyrosine phosphorylation of tion, and presynaptic differentiation.

MuSK Is Required for NMJ Formation503

vector was designed that would delete most of the thirdkinase exon upon homologous recombination into theendogenous mouse MuSK locus (Figure 1A); this tar-geting vector contained a total of 3.8 kb of homologywith the mouse MuSK gene. Successful gene targetingusing this construct was predicted to result in the gener-ation of a novel 3.8 kb EcoRI fragment from the targetedallele as detected by a 59 probe, as well as loss of twoNcoI fragments hybridizing to a kinase probe (Figure1A). Southern blot screening for these fragments re-vealed that successful targeting of the mouse MuSKgene was achieved in 4 of z400 embryonic stem (ES)cell clones obtained using a double selection schemeintended to enhance for selection of targeted clones;the ES cells were derived from the 129 strain of mice.Male chimeras derived from all four of these targetedclones were bred with C57BL/6 females. Chimeras fromtwo of the targeted clones transmitted the mutant alleleto the F1 generation. The resulting F1 progeny heterozy-gous for the MuSK mutation were viable and appearednormal and fertile.

MuSK Gene Disruption Results in Perinatal LethalityThe heterozygous F1 progeny were interbred to gener-ate mice homozygous for the MuSK gene disruption(designated MuSK2/2 mice). Among the F2 litters de-rived from these crosses were newborn mice that diedperinatally. Genotype analysis of tail DNA revealed thatthe dead pups were homozygous for the mutant MuSKallele (Figure1B); significantly,not a single mouse homo-zygous for the mutation survived the perinatal period(37 homozygotes were noted among the first 138 pupsthat were genotyped, corresponding to a 26.8% fre-quency of homozygotes).

To determine the phenotype of the MuSK2/2 new-Figure 1. Targeting of the MuSK Gene borns immediately at birth, we were careful to observe(A) Schematic representation of genomic DNA encompassing the the births of several litters derived from heterozygotethree kinase domain exons of the mouse MuSK gene (D. M. V. et

crosses. Though normal in their gross anatomy andal., unpublished data), of the targeting vector constructed, and ofbody weight, the MuSK2/2 pups differed in several strik-a mutant locus following successful targeting. The three exons ofing ways from their littermate controls. First, theythe MuSK kinase domain are indicated as black boxes, containing

the indicated kinase subdomains (SD). The PGK-neo and MC1-tk showed no spontaneous movement and didnot respondcassettes are indicated as open boxes. The novel EcoRI (R) and to a mild tail or leg pinch (Figure 2, top left). Only aNcoI (N) fragments generated following successful targeting are strong tail pinch was able to elicit a weak uncoordinatedlabeled. The 59 EcoRI/HpaI probe used to detect the endogenous

movement. By contrast, littermate controls showed ex-and mutant EcoRI fragments was derived from genomic DNA nottensive movement and responded vigorously to a mildincluded in the targeting construct. B, BamHI;Hp, HpaI; S, SpeI (sitestail pinch (Figure 2, top right). Second, the MuSK2/2included within parentheses are destroyed in the cloning process).

(B) A representative Southern blot of tailDNA from wild-type, hetero- pups were cyanotic at birth (Figure 2, top left) and ap-zygous, and homozygous F2 progeny showing the endogenous and peared not to breathe, although their hearts continuedmutant EcoRI fragments detected by the 59 RI/HpaI probe, as well to beat for a short time after birth. To determine whetheras the endogenous NcoI fragments detected by the kinase region

the MuSK2/2 pupshad ever taken a breath, we examinedprobe, which are absent in the homozygous mutant.the lungs histologically. Lung alveoli are collapsed inutero and expand with the first breath of life; even if

Results respiration is then terminated, the alveoli remain ex-panded. Histological examination revealed that the alve-

Homologous Recombination to Disrupt oli of MuSK2/2 pups were not expanded (Figure 2A),the MuSK Gene indicating that the pups had never taken a breath. InThe tyrosine kinase domain of MuSK is comprised of contrast, the lungs of the littermate controls displayed11 subdomains that are divided among three coding expanded alveoli (Figure 2B).exons (D. M. V. et al., unpublished data). Subdomain Iencoding the ATP-binding domain is located on the first Normal Skeletal Muscle but Aberrantkinase exon, while subdomains 5–11 encoding the cata- Neuromuscular Junctions in MuSK2/2 Micelytic region are located on the third kinase exon (Figure Because MuSK is localized to synaptic sites in skel-

etal muscle (Valenzuela et al., 1995), and because1A). To disrupt MuSK tyrosine kinase activity, a targeting

Cell504

Figure 2. Newborn Mice Carrying a MuSKGene Disruption Are Immobile and Do NotTake a Single Breath, but They Do ContainGrossly Normal Skeletal Muscle

(Top) Images depict stop-gap photographyof newborn F2 littermates showing thatMuSK2/2 pups, unlike control pups, com-pletely lack spontaneous movements and arenotably cyanotic.(A and B) Although the hearts of MuSK2/2

newborn pups beat for a few minutes afterbirth, postmortem histological analysis of thelung demonstrates that the alveoli air sacs inthe MuSK2/2 newborn (A) are not expandedas they are in the lung of the control littermate(B), indicating that mutant pups do not takea single breath.(C and D) Postmortem histological analysisof hindlimb musculature reveals that theMuSK2/2 mice (C) possess grossly normalmuscle architecture similar to that of controlmice (D).

MuSK2/2 mutant mice are immobile at birth and die stained with antibodies to synaptophysin to label synap-tic vesicles, which are normally concentrated in nerveshortly thereafter, we reasoned that neuromuscular syn-

apse formation might be aberrant in MuSK2/2 mutants. terminals (Buckley and Kelly, 1985), and with a-bungaro-toxin to label AChRs, which are concentrated in theWe first examined the diaphragm muscle because its

simple organization and thin structure allows synaptic postsynaptic membrane of normal skeletal muscle (Halland Sanes, 1993). Since nerve terminals normally arisesites to be visualized in whole-mount preparations. The

diaphragm muscle is innervated by the phrenic nerve, from short branches of the main intramuscular nerve,synaptophysin staining in control animals reveals a cuffwhich normally enters near the center of the diaphragm

muscle. The main intramuscular nerve is oriented per- of nerve terminals bordering the main nerve in the cen-tral region of the diaphragm muscle (Figure 3A). Inpendicular to the long axis of the muscle fibers and

extends through the central region of the muscle. The MuSK2/2 mutant mice, synaptophysin-stained nerveprocesses are not restricted to the central region ofarrangement and gross structure of the muscle fibers

(compare Figures 2C and 2D), as well as of the main diaphragm muscle and instead are found throughoutthe muscle (Figures 3B and 4 [top image]). Indeed, syn-intramuscular nerve (Figure 3), appears to be unaltered

in MuSK2/2 mutant mice. Thus, although the onset of aptophysin-stained nerve processes are nearly as abun-dant at the ends of the muscle as in the central regionMuSK expression occurs at about embryonic day 11

in developing mouse somites (within the presumptive (Figure 4, halftone images). Moreover, the shape of syn-aptophysin-stained processes in MuSK2/2 mutant micemyotome), MuSK does not appear to be essential for

the generation, proliferation and fusion of myoblasts, or resembles that of preterminal axons rather than that ofnerve terminals, and highly arborized nerve terminalsfor the growth of motor axons from spinal cord to mus-

cle. Because a detailedanalysis of myofiber sizewas not are absent in MuSK2/2 mice, indicating that MuSK isrequired for at least some aspects of presynaptic differ-performed, we cannot exclude the possiblity, however,

that some myofiber atrophy may be present. entiation. Since MuSK is expressed in skeletal muscleand not in motor neurons (Valenzuela et al., 1995), it isTo determine whether synapse formation is aberrant

in MuSK2/2 mutant mice, we stained whole-mounts of likely that the exuberant growth of motor axons and theabsence of nerve terminal arborization is due to a lackdiaphgram muscle with probesthat allowed us toassess

presynaptic and postsynaptic differentiation. Whole- of MuSK function in muscle and a consequent loss ofappropriate retrograde signals that normally preventmounts of diaphragm muscle from newborn mice were

MuSK Is Required for NMJ Formation505

Figure 3. Pre- and Postsynaptic Differentia-tion Are Aberrant in Muscle from MuSK2/2

Mutant Mice

A whole-mount of a diaphragm muscle froma wild-type (A and C) or a MuSK2/2 mutant(B and D) mouse was simultaneously stainedwith antibodies against synaptophysin (SYN)to label vesicles in presynaptic nerve termi-nals and with rhodamine-labeled a-bungaro-toxin (a-BGT) to label AChRs on the postsyn-aptic muscle membrane, and the wholemounts were viewed with optics selective foreither fluorescein (A and B) or rhodamine (Cand D). Synaptic sites in normal muscle (Aand C) are located adjacent to the main intra-muscularnerve (arrows in[A]) and are charac-terized by short arborized nerve terminals (A)and an accumulation of clustered AChRs un-derlying the terminals (C). MuSK2/2 mutantmuscle (B and D) lacks arborized nerve termi-nals (B) and clustered AChRs (D). Althoughsynaptophysin-stained nerve branches arepresent in MuSK2/2 mutant muscle (B), theyare aberrantly long and extend throughoutthe muscle and do not terminate near themain intramuscular nerve (arrows in [B]).Whole-mounts of diaphragm muscle fromwild-type (E) or MuSK2/2 mutant (F) mousewere also stained for AChE activity (Karnov-sky, 1964; Letinsky et al., 1976). AChE is local-ized to the centrally located endplate band indiaphragm muscle from wild-type mice (E),whereas accumulations of AChE are absentin muscle from MuSK2/2 mutant mice AChE(F). Bar, 100 mm.

continued axon growth and also induce presynaptic dif- membrane of neuromuscular synapses, or locally de-posited in the overlying synaptic basal lamina, and serveferentiation (see Discussion).as markers of appropriate postsynaptic differentiationAChR clusters, which are found in postsynaptic mem-(Hall and Sanes, 1993; Bowe and Fallon, 1995). We ex-brane patches underlying synaptophysin-stained nerveamined the distributions of several of these markers toterminals in normal muscle (compare Figure 3C withfurther characterize the extent of postsynaptic deficitsFigure 3A), are absent from MuSK2/2 mutant musclein MuSK2/2 mice. To examine the distribution of a pro-(Figure 3D). The absence of AChR clusters in MuSK2/2

tein that normally becomes localized to the synapticmutant muscle is not due to a lack of AChR expression,basal lamina in MuSK2/2 mutant muscle, we stainedsince AChR transcripts are expressed in MuSK2/2

whole-mounts of diaphragm muscle from control andmutant muscle (see below) and quantitation usingMuSK2/2 mutant pups for acetylcholinesterase (AChE)[125I]a-bungarotoxin reveals more—albeit diffusely dis-activity. Whereas AChE activity is present in a band intributed—AChRs in MuSK2/2 mutant muscle as com-the central region of normal diaphragm muscle, no suchpared with normal muscle (data not shown), and similarlocalization is present in diaphragm muscle fromnumbers of AChRs in cultured MuSK2/2 and controlMuSK2/2 mutant mice (compare Figures 3E and 3F); themyotubes (Glass et al., 1996). Thus, MuSK is requireddifference between normal and mutant mice once againto cluster but not to express AChRs.appears to reflect a failure in localization as opposed toThe absence of differentiated nerve terminals andexpression, since wild-type and mutant muscle containAChR clusters in MuSK2/2 mutant mice is consistentsimilar levels of AChE mRNA (data not shown).with the notion that these mice die at birth owing to

To further determine the extent of postsynaptic differ-severe abnormalities in both presynaptic and postsyn-

entiation in MuSK2/2 mutant mice, we stained frozenaptic aspectsof NMJ formation, and that MuSK isessen-

sections of skeletal muscle from newborn MuSK2/2 micetial for normal formation of neuromuscular synapses with antibodies to synaptic vesicle proteins (either syn-in vivo. aptophysin or SV2; Buckley and Kelly, 1985), which are

normally concentrated in nerve terminals, and with anti-Postsynaptic Differentiation Is Absent bodies against proteins normally found in the synapticin MuSK2/2 Mutant Mice basal lamina or postsynaptic membrane. While clustersIn addition to the AChRs, approximately one dozen pro- of AChE, ErbB4, utrophin, and rapsyn are all precisely

juxtaposed with nerve terminals in sections of muscleteins are known to be concentrated in the postsynaptic

Cell506

Figure 4. Neurites from MuSK2/2 Mice Do Not Undergo Growth Arrest

Montage views of synaptophysin-stained diaphragm muscles from control and MuSK2/2 mice, showing that synaptophysin-stained neuritesterminate proximally to the main intramuscular nerve (arrowheads) in the diaphragm from control mice (middle, white arrows depict nervebranch lacking nerve sprouts), but extend to the ends of the diaphragm from MuSK2/2 mice (top). Quantitation of this observation is depictedin the bar graph; the muscle was divided into bins at increasing distance from the main nerve, and the total length of neurites within eachbin is presented; solid arrows mark the ends of the diaphragms in all the panels. Bar, 100 mm.

from control and MuSK1/- pups, these proteins were of the muscle (Gautam et al., 1995). We do not observean enrichment of AChRs in the central region of theneither concentrated at sites stained with antibodies to

synaptic vesicle proteins nor anywhere else in muscle muscle by using rhodamine-labeled a-bungarotoxin(see Figure 3D), and [125I]a-bungarotoxin binding indi-from MuSK2/2 mutant pups (Figure 5A–5H). These re-

sults indicate that MuSK is required for many, if not all, cates that AChRs are uniformly distributed in MuSK2/2

muscle (data not shown). These data suggest that MuSKaspects of postsynaptic differentiation.mutant mice may lack synapse-specific transcription ofAChR genes. To learn whether MuSK is indeed requiredMuSK2/2 Mutant Mice Lack Synapse-Specific

AChR Gene Expression for synapse-specific expression, we used in-situ hybrid-ization to determine the expression pattern of tran-Two different pathways are thought to be responsible

for clustering AChRs at synaptic sites. One pathway scripts encoding the a and d subunits of the AChR. Incontrol mice, these transcripts are concentrated nearcauses a redistribution of pre-exisiting AChRs to the

newly formed synapse (McMahan, 1990), while a second myofiber nuclei that are situated near synaptic sites,which appear as a band in the central region of thepathway, thought to involve neuregulin, increases AChR

gene expression in muscle fiber nuclei situated near muscle (e.g. Figure 6, left images). In contrast, tran-scripts encoding both the a (Figure 6, right images) andthe synaptic site, resulting in an accumulation of AChR

mRNA near synaptic nuclei (Burden et al., 1995; Chu et d (data not shown) subunits are distributed uniformlythroughout the myofibers of MuSK2/2 mice. Thus, MuSKal., 1995; Duclert and Changeux, 1995). In rapsyn2/2

mice, which lack clustered AChRs, synapse-specific is required for synapse-specific expression of AChRgenes, and these results suggest that an inability togene expression appears normal, and as a consequence

AChRs are still notably enriched within the central region selectively express AChR genes in synaptic nuclei is

MuSK Is Required for NMJ Formation507

Figure 5. Immunocytochemical Analysis Re-veals That Specializations of the Postsynap-tic Membrane and Synaptic Basal Lamina AreLacking in Muscle from MuSK2/2 MutantMice

Frozen sections of muscle from newbornmice were stained with antibodies as de-scribed in Experimental Procedures, and sin-gle sections, which were stained both withantibodies against a synaptic vesicle protein(SV2 or synaptophysin) and with antibodiesagainst a postsynaptic protein, were viewedwith optics selective for rhodamine or fluores-cein. AChRs, AChE, ErbB4, utrophin, andrapsyn are clustered at synaptic sites,marked by SV2 or synaptophysin (syn), inmuscle from MuSK1/2 (A–J) but not fromMuSK2/2 mutant (K–T) mice. a-BGT, a-bun-garotoxin; syn, synaptophysin; AChE, acetyl-cholinesterase. Bar, 5 mm.

responsible, at least in part, for the lack of AChR cluster- to test the possibility that agrin acts via the MuSK recep-tor. In an accompanying paper (Glass et al., 1996), weing in muscle from MuSK2/2 mice.demonstrate that MuSK indeed serves as a requiredsignaling component of a complex receptor that specifi-Discussioncally responds to the neural forms of agrin. Altogether,our findings indicate that agrin, acting via the MuSKWe describe mice homozygous for a MuSK gene disrup-receptor, is required for all aspects of NMJ formation.tion and report that, while early skeletal muscle develop-

The defects in NMJ formation in mice lacking MuSKment apparently proceeds rather normally in MuSK2/2

are more profound than might have been predicted formice, every aspect of NMJ formation that we examinedmice lacking a functional agrin receptor. The previouslyis absent in these mice. Branches of the main intramus-described in vitro actions of agrin are largely limitedcular nerve do not establish normal contacts with theto its organizing effects on postsynaptic differentiationmuscle, do not form correctly positioned or specialized(McMahan, 1990; Hall and Sanes, 1993; Bowe and Fal-nerve terminals, and are apparently not given appro-lon, 1995). The aberrant behavior of the preterminal ax-priate signals to stop their wandering aimlessly acrossons and the lack of synapse-specific transcription repre-

the muscle. Furthermore, we find no evidence of post-sent deficits that would not necessarily have been

synaptic differentiation, since muscle-derived proteinsexpected to occur in mice in which agrin-mediated sig-

that are normally localized to the synaptic basal lamina naling pathways had been disrupted. However, recentor the postsynaptic membrane are instead uniformly generation of mice lacking agrin reveal NMJ defects thatdistributed in MuSK2/2 myofibers. The defects in NMJ are indeed as profound as those described here for miceformation are sufficient to account for the perinatal le- lacking MuSK (Gautam et al., 1996 [this issue of Cell ]),thality of the MuSK2/2 mice resulting from their inability consistent with the idea that MuSK is a required compo-to breathe, and for their immobility. nent of the functional receptor for agrin. Since MuSK

Agrin is a nerve-derived molecule that has been sug- is located only on the postsynaptic side of the NMJgested to have a critical organizing role at the motor (Valenzuela et al., 1995), it seems likely that the aberrantendplate based on its in vitro clustering activities (Mc- behavior of the presynaptic terminals in MuSK2/2 mutantMahan, 1990; Hall and Sanes, 1993; Bowe and Fallon, mice is due to indirect actions of the agrin/MuSK signal-1995). Although the mechanism of action of agrin is not ing system (Figure 7). Thus, we favor the idea that agrinunderstood, evidence suggests that tyrosine phosphor- released from the nerve terminal causes the postsynap-ylation may be involved (Wallace et al., 1991; Qu and tic muscle cell, via MuSK activation, to release recipro-Huganir, 1994; Ferns et al., 1996; Wallace, 1994, 1995). cally a recognition signal back to the nerve, or toBased onthe absence of NMJs in mice lacking theMuSK Schwann cells (Son and Thompson, 1995), to indicate

that a functional contact has occurred. In response toRTK, as reported in this manuscript, we have proceeded

Cell508

Figure 6. AChR Genes Are Expressed Selectively in Synaptic Nuclei of Normal Muscle but Uniformly in Muscle from MuSK2/2 Mutant Mice

Sections of intercostal muscles from wild-type and MuSK2/2 mice, hybridized to an AChR a probe, are depicted in bright-field (top images)and dark-field (bottom images) views; the intercostal muscle fibers are oriented obliquely to the ribs to which they are attached (M, muscle;R, ribs). AChR a transcripts are concentrated in the centrally located endplate band from wild-type mice (arrowheads, bottom left image), butdistributed uniformly in muscle from MuSK2/2 mutant mice (bottom right image). Binding of sense probes to muscle from wild-type andMuSK2/2 mutant mice is uniform and not above background.

this muscle-derived recognition signal, the nerve stops we do not favor the idea that MuSK is the receptor forthe extracellular signal that activates synapse-specificgrowing and undergoes presynaptic differentiation; the

ability of basal lamina sheaths to induce presynaptic transcription. At present, the best candidate for such asignal is neuregulin, acting via the ErbB3 and/or ErbB4differentiation in the absence of the underlying muscle

suggests that muscle-derived signals may be embed- receptors localized in the postsynaptic membrane (Al-tiok et al., 1995; Moscoso et al., 1995; Zhu et al., 1995).ded in synaptic basal lamina, like agrin and neuregu-

lin (Sanes et al., 1978). Alternatively, we cannot elimi- Thus, it seems likely that the loss of synapse-specifictranscription is another indirect by-product of the factnate the possibility that excessive neuronal growth in

MuSK2/2 mutant mice is caused by persistent produc- that disruption of agrin/MuSK signaling blocks normalsynaptic differentiation, resulting in an inability of thetion of a muscle-derived sprouting factor that is normally

down-regulated following functional innervation (Brown muscle cell to cluster ErbB receptors and possibly aninability of mutant nerve terminals to provide neuregulin.et al., 1981), or from the absence of appropriate synaptic

activity (Balice-Gordon and Lichtman, 1994). It should Consistent with the former possibility, recent studiesdemonstrate that ectopic agrin induces ectopic cluster-be noted that despite the defects displayed by nerve

branches, the main intramuscular nerves appear to be ing of ErbBs (M. Rimer, I. Cohen, T. Lomo, C. R. Slater,S. J. B., U. J. McMahan, unpublished data).properly positioned within the muscle in MuSK2/2 mice,

indicating that guidance mechanisms that bring motor The deficits in mice lacking either agrin or MuSK aremore severe than in mice lacking rapsyn. Thus, althoughaxons to muscle remain intact and operate indepen-

dently of the agrin/MuSK signaling system. presynaptic differentiation is disturbed in mice lackingrapsyn, as evidenced by excessive growth of mutantAs with the deficits in presynaptic differentiation, it

seems likely that the lack of synapse-specific transcrip- axons that end without arborizing, the nerve branchesin rapsyn2/2 mice terminate quite close to the main intra-tion in MuSK2/2 mice also results from an indirect mech-

anism (Figure 7). That is, although AChR genes are ex- muscular nerve (Gautam et al., 1995). Various aspectsof postsynaptic differentiation are also maintained inpressed uniformly in muscle fibers from MuSK2/2 mice,

MuSK Is Required for NMJ Formation509

Figure 7. Schematic Representation of RolePlayed by the Agrin/MuSK Signaling System

MuSK appears to activate signaling cascadesthat are responsible for all aspects of NMJformation, including postsynaptic organiza-tions, synapse-specific transcription, as wellas presynaptic growth and differentiation(perhaps by regulating the elaboration of ret-rograde signals, or simply by promoting syn-aptic activity that down-regulates productionof sprouting factors). MuSK activates at leasttwo independent pathways, one that is rap-syn dependent and leads to AChR and dys-troglycan clustering, and one that appears tobe rapsyn independent and results in syn-apse-specific transcription. In the modeldepicted, synapse-specific transcription ispresumed to involve clustering of ErbB re-ceptors. However, it has recently been shownthat synapse-specific transcription occurs inrapsyn-deficient mice although they lackclustered ErbB3 receptors (Moscoso et al.,1995), indicating that synapse-specific ex-pression does not require clustering of ErbB3in particular; it may still involve clustering ofother ErbB receptors (e.g. ErbB4), or alterna-tively it may depend on ErbB-independentpathways.

rapsyn2/2 mice. For example, although normal cluster- MuSK signaling toa subset of its downstream pathways,in much the same way rapsyn appears to be required.ing of a number of synaptic proteins—including AChRs

as well as the dystroglycans and their associated pro- Analysis of rapsyn2/2 mutant mice (Gautam et al., 1995)provides support for the latter possibility, suggestingteins—is lacking in rapsyn2/2 mice, synapse-specific

transcription remains intact, which apparently results in that dystroglycan is not absolutely required for at leastcertain aspects of agrin-mediated signaling throughthe enrichment of AChRs within the central region of

the muscle. In addition, synaptic basal lamina proteins, MuSK. Thus, in rapsyn2/2 mutant mice, dystroglycanand its complex do not localize to the postsynapticsuch as AChE, s-laminin, and agrin, are localized to the

synaptic basal lamina in rapsyn2/2 mice. Thus, while it membrane, yet presynaptic differentiation and synapse-specific transcription do occur (Gautam et al., 1995); inseems as if there is only a single nerve-derived signaling

pathway—involving agrin acting via MuSK—that initi- addition, recent data suggests that MuSK transcriptscluster postsynaptically in rapsyn2/2 mutant mice (J. R.ates all aspects of NMJ formation, rapsyn is required

for only a subset of the events downstream from this Sanes, L. M. Moscoso, D. J. G, and G. D. Y., unpublisheddata). Thus, at least some aspects of agrin/MuSK-medi-initiating signal (Figure 7). The enrichment of AChRs

within the central region of muscle from rapsyn2/2 mice, ated differentiations seem to occur independently ofcoclustering of the dystroglycan complex with MuSK atwhere the mutant nerve endings are also localized, may

allow for some neurotransmission and account for the the motor endplate.The essential role of MuSK at the NMJ raises thelimited neonatal movement observed in rapsyn2/2 mice,

which contrasts with the immobility of MuSK2/2 pups. possibility that MuSK or related RTKs are playing similarroles in other synapses, particularly since low levels ofIt seems plausible that a stop signal for nerve terminals

is, like the AChR proteins, enriched within the central MuSK mRNA are detectable in central nervous system(Valenzuela et al., 1995; and data not shown). Agrin, asregion of muscle from rapsyn2/2 mutant mice but uni-

formly distributed in muscle from MuSK2/2 mutant mice, well as binding sites for agrin, have been detected inthe central nervous system (Bowe and Fallon, 1995),explaining the difference in nerve branching noted be-

tween rapsyn2/2 and MuSK2/2 mice. We would expect suggesting that analysis of synapses in the central ner-vous systems of MuSK2/2 mice is warranted. In thisthat the gene or genes encoding this stop signal, like

those encoding the different AChR subunits, are tran- regard, a pair of receptors known as the RORs (Masia-kowski and Carroll, 1992), which are expressed duringscribed preferentially in synaptic nuclei of normal and

rapsyn2/2 muscle (Figure 7). early development of the central nervous system andwhose closest relative is MuSK, should probably alsoDespite our findings, the role of dystroglycan as an

agrin receptor remains unclear. The comparison of be evaluated for roles in synapse formation. In any case,the complex types of spatial organizing events initiatedMuSK2/2 and agrin2/2 mutant mice, together with the

results of our accompanying manuscript (Glass et al., by MuSK have not previously been noted in responseto other RTKs, which have instead been largely charac-1996), indicates that MuSK is the critical receptor com-

ponent that initiates agrin-mediated signaling pathways. terized for their ability to mediate responses to growthfactors that elicit mitogenic, survival, or differentiativeHowever, dystroglycan could still be an essential acces-

sory component of a MuSK receptor complex (see Glass actions (Schlessinger and Ullrich, 1992). It remains pos-sible that, in addition to its organizing role at the NMJ,et al., 1996), or it could instead be necessary to couple

Cell510

used at 1/40; antibodies to ErbB4 (616) were kindly provided bythe agrin/MuSK system is also involved in more classicalDr. C. Lai (Scripps Institute) and were used at 1/2000. Followinggrowth factor–like responses. Further analysis will beincubation with primary antibodies, the sections were washed threerequired to clarify initial impressions that the muscle intimes for 30 min in PBS and incubated with appropriate fluoro-

MuSK2/2 pups may be atrophic; while such atrophy may chrome-conjugated secondary antibodies (Cappel, Inc.) in PBS withsimply be due to lack of appropriate electrical activity 0.1% BSA for 2 hr at room temperature, washed three times in PBS

for 15 min, and mounted and viewed as described for whole-mountsin MuSK2/2 muscle, loss of an agrin/MuSK-mediatedof muscle.growth factor–like signal might also be contributory.

Along these lines it is worth noting that, in the adult,In Situ HybridizationsMuSK is dramatically up-regulated in situations—suchNewborn mice were fixed in 4% paraformaldehyde in PBS at 48Cas those involving denervation or forced immobiliza-overnight andrinsed briefly inPBS. Ribs and the attached intercostal

tion—where the muscle is prone to atrophy (Valenzuela muscles were dissected from newborn pups, frozen in liquid nitro-et al., 1995). Providing exogenous agrin in these situa- gen-chilled isopentane, and embedded in OCT (Tissue Tek). Frozen,

longitudinal sections (10 mm) were collected and subjected to intions, or analyzing conditional mutants in which MuSKsitu hybridizations as previously described (Valenzuela et al., 1995),expression is specifically ablated in adult muscle whichusing either an AChR d probe (Simon et al., 1992), or an AChR ais then subjected to treatments that can result inprobe derived by amplifying a fragment of the mouse AChR a cDNA

atrophy, may tell us whether the agrin/MuSK pathway (corresponding to nucleotides 795 to 1382, GenBank Accessioncan provide myotrophic signals as well as organiza- number X03986) using primers tailed with either the T7 (sense) or thetional cues. T3 (antisense) polymerase promoter sites. 35S-radiolabeled sense

or antisense strand probes were transcribed off the gel-purifiedlinearized AChR d plasmid or the AChR a fragment using an RNAExperimental ProceduresTranscription Kit from Stratagene.

Targeting Vectors and Embryonic Stem Cell TechnologyThe MuSK gene targeting vector was constructed from mouse geno- Acknowledgmentsmic DNA fragments isolated from a lambda FIX II phage libraryprepared with 129 strain mouse genomic DNA (Stratagene). The We thank Drs. L. S. Schleifer and P. Roy Vagelos, along with the rest1.7 kb SpeI fragment depicted in Figure 1A was ligated into the of the Regeneron community, for enthusiastic support. We thank Dr.compatible ends of a unique XbaI site upstream of the PGK-neo J. Sanes for useful discussions and communicating the descriptioncassette (destroying the SpeIand XbaI sites), while the 2.1 kb BamHI of his mice lacking agrin prior to their publication; Drs. Fredrick AltDNA fragment depicted in Figure 1A was blunt-end ligated into and Kirk Thomas for the PGK-neo and MC1-tk plasmids, respec-the unique HindIII site between the PGK-neo cassette and MC1-tk tively; and Dr. Richard Murray for the 129/Ola ES cells (E14.1). Weexpression cassettes (destroying the BamHI and HindIII sites). The thank Albert Fliss, Anthony Lucarelli, and Rosemary Rodriguez fortargeting vector was linearized by digestion with NotI and then elec- expert technical assistance, and Monique Gisser and Jennifer Grif-troporated into E14.1 embryonic stem cells, which were subjected fiths for DNA sequencing and oligonucleotide synthesis. This workto a double selection protocol (gancyclovir addition resulted in a was supported in part by grants from the National Institutes of Health5- to 10-fold enrichment compared with selection in G418 alone) to S. J. B. (NS21579 and NS27963). We are also indebted to Claudiaand then used to generate chimeric mice as previously described Murphy, Eric Hubel, and Jon Weider for excellent graphics work.(Conover et al., 1995; DeChiara et al., 1995).

Received February 16, 1996; revised March 22, 1996.ImmunofluorescenceFor whole-mount diaphragm preparations, newborn mice were fixed Referencesin 1% paraformaldehyde in phosphate-buffered saline (PBS) at 48C

Altiok, N., Bessereau, J.L., and Changeux, J.P. (1995). ErbB3 andfor several hours and then rinsed briefly in PBS. Diaphragm musclesErbB2/neu mediate the effect of heregulin on acetylcholine receptorwere dissected out, washed twice for 10 min in PBS, incubated ingene expression in muscle: differential expression at the endplate.0.1 M glycine in PBS for 15 min, rinsed for 5 min in PBS, andEMBO J. 14, 4258–4266.permeabilized with 0.5% Triton X-100 in PBS (PBT) for 5 min. The

muscles were then incubated with rabbit antibodies to synaptophy- Apel, E.D., Roberds, S.L., Campbell, K.P., and Merlie, J.P. (1995).sin (provided by Dr. R. Jahn, Yale University Medical School), which Rapsyn may function as a link between the acetylcholine receptorwere diluted 1/1000 in PBT with 2% BSA, overnight at 48C, subse- and the agrin-binding dystrophin-associated glycoprotein complex.quently rinsed for 5 min in PBT, washed three times for 1 hr in PBT, Neuron 15, 115–126.and then incubated simultaneously with flourescein-conjugated Balice-Gordon, R.J., and Lichtman, J.W. (1994). Long-term synapsesheep anti-rabbit IgG (Boehringer Mannheim) and tetramethlyrho- loss induced by focal blockade of postsynaptic receptors. Naturedamine-conjugated a-bungarotoxin (Molecular Probes, Oregon) 372, 519–524.overnight at 48C. The tissues were then washed three times for 1

Barbacid, M. (1993). Nerve growth factor: a tale of two receptors.hr with PBT, rinsed once with PBS for 5 min, fixed in 100% methanolOncogene 8, 2033–2042.at 2208C and mounted in 90% glycerol, 0.1 M Tris (pH 7.5) with 1Bowe, M.A., and Fallon, J.R. (1995). The role of agrin in synapsemg/ml p-phenylenediamine. The whole-mounts were viewed withformation. Annu. Rev. Neurosci. 18, 443–462.epiflourescence and filters that were selective for rhodamine or

flourescein, and images were recorded either on film or on a CCD Bowe, M.A., Deyst, K.A., Leszyk, J.D., and Fallon, J.R. (1994). Identi-camera (Princeton Instruments). fication and purification of an agrin receptor from Torpedo postsyn-

Frozen sections (10 mm) of leg muscles from newborn mice were aptic membranes: a heteromeric complex related to dystroglycans.prepared as described previously (Zhu et al., 1995) and incubated Neuron 12, 1173–1180.overnight at 48C with the following primary antibodies diluted in PBS Brown, M.C., Holland, R.L., and Hopkins, W.G. (1981). Motor nervecontaining 1% BSA: rabbit antibodies to acetylcholinesterase were sprouting. Annu. Rev. Neurosci. 4, 17–42.provided by Dr. T. Rosenberry (Case Western University Medical

Buckley, K.,and Kelly, R.B. (1985). Identification of a transmembraneSchool) and were used at 1/1000; hybridoma supernatant containingglycoprotein specific for secretory vesicles of neural and endocrinea monoclonal antibody to SV2 was provided by Dr. K. Buckley (Har-cells. J. Cell Biol. 100, 1284–1294.vard Medical School) andwas used at 1/100; hybridoma supernatant

containing a monoclonal antibody to utrophin (Novocastra Labora- Burden, S.J., Sargent,P.B., and McMahan, U.J. (1979). Acetylcholinereceptors in regenerating muscle accumulate at original synaptictories) was used at 1/10; affinity-purified rabbit antibodies to recom-

binant mouse rapsyn (S. Dyer and S. Burden, unpublished) were sites in the absence of the nerve. J. Cell Biol. 82, 412–425.

MuSK Is Required for NMJ Formation511

Burden, S.J., Jo, S.A., Tang, J., Zhu, X., Yeadon, J.E., and Simon, that cause aggregation of acetylcholine receptors on cultured mus-cle cells. J. Cell Biol. 99, 615–627.A.M. (1995). Polarity in skeletal muscle cells is induced by innerva-

tion. Sem. Dev. Biol. 6, 59–65. Goodearl, A.D., Yee, A.G., Sandrock, A.W., Corfas, G., and Fisch-bach, G.D. (1995). ARIA is concentrated in the synaptic basal laminaCampanelli, J.T., Roberds, S.L., Campbell, K.P., and Scheller, R.H.of the developing chick neuromuscular junction. J. Cell Biol. 130,(1994). A role for dystrophin-associated glycoproteins and utrophin1423–1434.in agrin-induced AChR clustering. Cell 77, 663–674.

Hall, Z.W., and Sanes, J.R. (1993). Synaptic structure and develop-Carraway, K.L., and Burden, S.J. (1995). Neuregulins and their recep-ment: the neuromuscular junction. Cell 72/Neuron 10 (Suppl.),tors. Curr. Opin. Neurobiol. 5, 606–612.99–121.Cartaud, J., and Changeux, J.P. (1993). Post-transcriptional com-Harris, D.A., Falls, D.L., and Fischbach, G.D. (1989). Differential acti-partmentalization of acetylcholine receptor biosynthesis in the sub-vation of myotube nuclei following exposure to an acetylcholineneural domain of muscle and electrocyte junctions. Eur. J. Neurosci.receptor-inducing factor. Nature 337, 173–176.5, 191–202.

Helgren, M.E., Squinto, S.P., Davis, H.L., Parry, D.J., Boulton, T.G.,Chu, G.C., Moscoso, L.M., Sliwkowski, M.X., and Merlie, J.P. (1995).Heck, C.S., Zhu, Y., Yancopoulos, G.D., Lindsay, R.M., and DiStef-Regulation of the acetylcholine receptor e subunit gene by recombi-ano, P.S. (1994). Trophic effects of ciliary neurotrophic factor onnant ARIA: an in vitro model for transynaptic gene regulation. Neurondenervated skeletal muscle. Cell 76, 493–504.14, 329–339.Hoch, W., Ferns, M., Campanelli, J.T., Hall, Z.W., and Scheller, R.H.Conover, J.C., Erickson, J.T., Katz, D.M., Bianchi, L.M., Poueymirou,(1993). Developmental regulation of highly active alternativelyW.T., McClain, J.,Pan, L., Helgren, M., Ip, N.Y., Boland, P., Friedman,spliced forms of agrin. Neuron 11, 479–490.B., Weigand, S., Vejsada, R., Kato, A., DeChiara, T.M., and Yanco-

poulos, G.D. (1995). Mice lacking NT4 thrive and retain motoneu- Jennings, C.G., Dyer, S.M., and Burden, S.J. (1993). Muscle-specificrones, but exhibit sensory neuron deficits distinct from mice lacking trk-related receptor with a kringle domain defines a distinct class ofBDNF. Nature 375, 235–238. receptor tyrosine kinases. Proc.Natl. Acad. Sci. USA 90, 2895–2899.

DeChiara, T.M., Vejsada, R., Poueymirou, W.T., Acheson, A., Suri, Jessell, T.M., Siegel, R.E., and Fischbach, G.D. (1979). InductionC., Conover, J.C., Friedman, B., McClain, J., Pan, L., Stahl, N., Ip, of acetylcholine receptors on cultured skeletal muscle by a factorN.Y., Kato, A., and Yancopoulos, G.D. (1995). Mice lacking the CNTF extracted from brain and spinal cord. Proc. Natl. Acad. Sci. USA 76,receptor, unlike mice lacking CNTF, exhibit profound motor neuron 5397–5401.deficits at birth. Cell 83, 313–322. Jo, S.A., Zhu, X., Marchionni, M.A., andBurden, S.J. (1995). Neuregu-Duclert, A., and Changeux, J.P. (1995). Acetylcholine receptor gene lins are concentrated at nerve-muscle synapses and activate ACh-expression at the developing neuromuscular junction. Physiol. Rev. receptor gene expression. Nature 373, 158–161.75, 339–368. Karnovsky, M.J. (1964). J. Cell Biol. 23, 217–232.Falls, D.L., Harris, D.A., Johnson, F.A., Morgan, M.M., Corfas, G., Kuffler, D.P. (1986). Accurate reinnervation of motor end plates afterand Fischbach, G.D. (1990). Mr 42,000 ARIA: a protein that may disruption of sheath cells and muscle fibers. J. Comp. Neurol. 250,regulate the accumulation of acetylcholine receptors at developing 228–235.chick neuromuscular junctions. Cold Spring Harbor Symp. Quant.

Letinsky, M.S., Fischbeck, K.H., and McMahan, U.J. (1976). Preci-Biol. 55, 397–406.sion of reinnervation of original postsynaptic sites in frog muscle

Falls, D.L., Rosen, K.M., Corfas, G., Lane, W.S., and Fischbach, after a nerve crush. J. Neurocytol. 5, 691–718.G.D. (1993). ARIA, a protein that stimulates acetylcholine receptor

Martinou, J.-C., Falls, D.L., Fischbach, G.D., and Merlie, J.P. (1991).synthesis, is a member of the neu ligand family. Cell 72, 801–815.Acetylcholine receptor-inducing activity stimulates expression of

Ferns, M., Camapnelli, J.T., Hoch, W., Scheller, R.H., and Hall, Z.W. the e-subunit gene of the muscle acetylchoine receptor. Proc. Natl.(1993). The ability of agrin to cluster AChRs depends on alternative Acad. Sci. USA 88, 7669–7673.splicing and on cell surface proteoglycans. Neuron 11, 491–502.

Masiakowski, P., and Carroll, R.D. (1992). A novel family of cellFerns, M., Deiner, M., and Hall, Z. (1996). Agrin-induced acetylcho- surface receptors with tyrosine kinase–like domain. J. Biol. Chem.line receptor clustering in mammalian muscle requires tyrosine 267, 26181–26190.phosphorylation. J. Cell Biol. 132, 937–944.

McMahan, U.J. (1990). The agrin hypothesis. Cold Spring HarborFong, G.H., Rossant, J., Gertenstein, M., and Breitman, M.L. (1995). Symp. Quant. Biol. 55, 407–418.Role of the Flt-1 receptor tyrosine kinase in regulating assembly of McMahan, U.J., and Slater, C.R. (1984). The influence of basal laminavascular endothelium. Nature 376, 66–70. on the accumulation of acetylcholine receptors at synaptic sites inGanju, P., Walls, E., Brennan, J., and Reith, A.D. (1995). Cloning and regenerating muscles. J. Cell Biol. 98, 1453–1473.developmental expression of NsK2, a novel receptor tyrosine kinase Moscoso, L.M., Chu, G.C., Gautam, M., Noakes, P.G., Merlie, J.P.,implicated in skeletal myogenesis. Oncogene 11, 281–290. and Sanes, J.R. (1995). Synapse-associated expression of an acetyl-Gautam, M., Noakes, P.G., Mudd, J., Nichol, M., Chu, G.C., Sanes, choline receptor-inducing protein, ARIA/heregulin, and its putativeJ.R., and Merlie, J.P. (1995). Failure of postsynaptic specialization receptors, ErbB2 and ErbB3, in developing mammalian muscle. Dev.to develop at neuromuscular junctions of rapsyn-deficient mice. Biol. 172, 158–169.Nature 377, 232–236. Nitkin, R.M., Smith, M.A., Magill, C., Fallon, J.R., Yao, Y.M.-Y., Wal-Gautam, M., Noakes, P.G., Moscoso, L., Rupp, F., Scheller, R.H., lace, B.G., and McMahan, U.J. (1987). Identification of agrin, a syn-Merlie, J.P., and Sanes, J.R. (1996). Defective neuromuscular synap- aptic organizing protein from Torpedo electric organ. J. Cell Biol.togenesis in agrin-deficient mice. Cell, this issue. 105, 2471–2478.

Gee, S.H., Montanaro, F., Lindenbaum, M.H., and Carbonetto, S. Qu, Z., and Huganir, R.L. (1994). Comparison of innervation and(1994). Dystroglycan-a, a dystrophin-associated glycoprotein, is a agrin-induced tyrosine phosphorylation of the nicotinic acetylcho-functional agrin receptor. Cell 77, 675–686. line receptor. J. Neurosci. 14, 6834–6841.

Glass, D.J., and Yancopoulos, G.D. (1993). The neurotrophins and Reist, N.E., Werle, M.J., and McMahan, U.J. (1992). Agrin released bytheir receptors. Trends Cell Biol. 3, 262–268. motor neurons induces the aggregationof acetylcholine receptors at

neuromuscular junctions. Neuron 8, 865–868.Glass, D.J., Bowen, D.C., Stitt, T.N., Radziejewski, C., Bruno, J.,Ryan, T.E., Gies, D.R., Shah, S., Burden, S.J., DiStefano, P.S., Valen- Ruegg, M.A., Tsim, K.W.K., Horton, S.E., Kroger, S., Escher, G.,zuela, D.M., DeChiara, T.M., and Yancopoulos, G.D. (1996). Agrin Gensch, E.M., and McMahan, U.J. (1992). The agrin gene encodesacts via a MuSK receptor complex. Cell, this issue. for a family of basal lamina proteins that differ in function and distri-

bution. Neuron 8, 691–699.Godfrey, E.W., Nitkin, R.M., Wallace, B.G., Rubin, L.L., and McMa-han, U.J. (1984). Components of Torpedo electric organ and muscle Rupp, F., Payan, D.G., Magill-Solc, C., Cowan, D.M., and Scheller,

Cell512

R.H. (1991). Structure and expression of a rat agrin. Neuron 6,811–823.

Sanes, J.R. (1995). The synaptic cleft of the neuromuscular junction.Dev. Biol. 6, 163–173.

Sanes, J.R., Marshall, L.M., and McMahan, U.J. (1978). Reinnerva-tion of muscle fiber basal lamina afterremoval of myofibers. Differen-tiation of regenerating axons at original synaptic sites. J. Cell Biol.78, 176–198.

Sato, T.N., Tozawa, Y., Deutsch, U., Wolburg-Buchholz, K., Fujiwara,Y., Gendron-Maguire, M., Gridley, T., Wolburg, H., Risau, W., andQin, Y. (1995). Distinct roles of the receptor tyrosine kinase Tie1 andTie-2 in blood vessel formation. Nature 376, 70–74.

Schlessinger, J., and Ullrich, A. (1992). Growth factor signaling byreceptor tyrosine kinases. Neuron 9, 383–391.

Sealock, R., and Froehner, S.C. (1994). Dystrophin-associated pro-teins and synapse formation: is a-dystroglycan the agrin receptor?Cell 77, 617–619.

Shalaby, F., Rossant, J., Yamaguchi, T.P., Gertsenstein, M., Wu,X.F., Breitman, M.L., and Schuh, A.C. (1995). Failure of blood-islandformation and vasculogenesis in Flk-1-deficient mice. Nature 376,62–66.

Simon, A.M., Hoppe, P., and Burden, S.J. (1992). Spatial restrictionof AChR gene expression to subsynaptic nuclei. Development 114,545–553.

Snider, W.D. (1994). Functions of the neurotrophins during nervoussystem development: what the knockouts are teaching us. Cell 77,627–638.

Son, Y.J., and Thompson, W.J. (1995). Nerve sprouting in muscleis induced and guided by processes extended by Schwann cells.Neuron 14, 133–141.

Sugiyama, J., Bowen, D.C., andHall, Z.W. (1994).Dystroglycan bindsnerve and muscle agrin. Neuron 13, 103–115.

Tsim, K.W., Ruegg, M.A., Escher, G., Kroger, S., and McMahan, U.J.(1992). cDNA that encodes active agrin. Neuron 8, 677–689.

Valenzuela, D.M., Stitt, T.N., DiStefano, P.S., Rojas, E., Mattsson,K., Compton, D.L., Nunez, L., Park, J.S., Stark, J.L., Gies, D.R.,Thomas, S., Le Beau, M.M., Fernald, A.A., Copeland, N.G., Jenkins,N.A., Burden, S.J., Glass, D.J., and Yancopoulos, G.D. (1995). Re-ceptor tyrosine kinase specific for the skeletal muscle lineage: ex-pression in embryonic muscle, at the neuromuscular junction, andafter injury. Neuron 15, 573–584.

Wallace, B.G. (1989). Agrin-induced specializations contain cyto-plasmic, membrane, and extracellular matrix-associated compo-nents of the postsynaptic apparatus. J. Neurosci. 9, 1294–1302.

Wallace, B.G. (1994). Staurosporine inhibits agrin-induced acetyl-choline receptor phosphorylation and aggregation. J. Cell Biol. 125,661–668.

Wallace, B.G. (1995). Regulation of the interaction of nicotinic acetyl-choline receptors with the cytoskeleton by agrin-activated proteintyrosine kinase. J. Cell Biol. 128, 1121–1129.

Wallace, B.G., Qu, Z., and Huganir, R.L. (1991). Agrin induces phos-phorylation of the nicotinic acetylcholine receptor. Neuron 6,869–878.

Zhu, X., Lai, C., Thomas, S., and Burden, S.J. (1995). Neuregulinreceptors, erbB3 and erbB4, are localized at neuromuscular syn-apses. EMBO J. 14, 5842–5848.