176muscle differentiation such as contractile filaments, theirproper attachment to the epidermis, and their ability tofuse into multinucleate fibres. Micrographs of late-stage17 mef2 mutant embryos never revealed proper cell junctionsbetween motoneurons and improperly differentiatedmuscle fibres, but instead their cell surfaces are coveredby basement membrane (Fig. 2 K; <strong>Prokop</strong> et al. 1996).This phenotype could be explained by loss of mef2-dependentlate synaptic CAMs, directly maintaining theneuromuscular cell junction. Alternatively, mef2-functionmight be required to exclude basement membrane receptorsfrom the NMJ (e.g. via synapse-specific cytoskeletalproperties), thus preventing separation of the pre- andpostsynaptic membranes by competitive invasion of basementmembrane material (white arrowheads in Fig. 4B).So far, basement membrane receptors have not yetbeen identified. Loss of cell surface adhesion to basementmembrane in laminin A mutant embryos suggeststhat basement membrane receptors bind to Laminin orLaminin-dependent extracellular matrix components(<strong>Prokop</strong> et al. 1998a). In addition, lack of Laminin Aleads to a partial detachment of neuromuscular boutons 4(Fig. 2J), even though basement membrane is normallyabsent from the narrow neuromuscular cleft in Drosophila(Fig. 3A; <strong>Prokop</strong> et al. 1998a). This suggests that athin layer of Laminin in the synaptic cleft might existand mediate cell adhesion via appropriate receptors atthe NMJ. Alternatively, basement membrane, whichspans over the neuronal terminal and adheres closely tosurrounding muscle surfaces, might press the terminalagainst the muscle (Fig. 4B,C). If this were the case,weak adhesion at the neuromuscular cleft during the processof NMJ formation would suffice (provided mef2-dependentfactors kept basement membrane receptors awayfrom the NMJ; white arrowheads in Fig. 4B). Low neuromuscularadhesion would facilitate the extensive reorganisationof the terminal’s shape during NMJ formation.This would be consistent with the fact that plasticreshaping of neuronal terminals in Drosophila larvae orin the sea slug Aplysia is at least partly dependent on adecrease in synaptic CAMs, i.e. a reduction in cell adhesion(Schuster et al. 1996a; Thompson et al. 1996; Baileyet al. 1997).Differentiation of nerve terminal shape and sizeDifferentiation and shape changes of growth cones intobouton-forming terminals take place during 4–5 h afterinitial neuromuscular contact at early stage 13 (Broadieand Bate 1993c; Yoshihara et al. 1997). These changesappear to be accelerated or facilitated by late bloomergene function, which encodes a member of the tetraspaninfamily of receptor-complex-associated proteins.4 Reduction of neuromuscular contact was also found in s-lamininmutant mice; s-laminin is part of the basement membrane withinthe ca. 50-nm-wide neuromuscular cleft and might well mediateadhesion to synaptic receptors (Noakes et al. 1995).The Drosophila late bloomer protein is localised in motoneuronalaxons (Kopczynski et al. 1996), suggesting thatmotoneurons receive signals. Such signals might comefrom the target muscles, inducing the differentiation processof the presynaptic terminal (not shown in Fig. 4). Inagreement with this interpretation, retrograde signallingat the onset of NMJ differentiation has been described inother systems. For example, at the vertebrate NMJ muscle-releasedAgrin seems to serve such a function (Rüeggand Bixby 1998). Also certain Heliosoma motoneurons inculture show presynaptic calcium influx upon contactwith appropriate muscle fibres (Funte and Haydon 1993;Zoran et al. 1993). However, in the case of Drosophila,loss of late bloomer function only delays NMJ differentiationbut does not block it (Kopczynski et al. 1996), suggestingthat a potential muscle-derived signal might onlybe of minor importance. This is in accordance with otherfindings that differentiation of bouton-like motoneuronalstructures can take place even in the complete absence ofmuscles (<strong>Prokop</strong> et al. 1996).Also the size of the motoneuronal terminal appearsrelatively independent of the target muscle. This is suggestedby observations in myoblast city mutant embryos,where myoblast fusion is blocked due to lack of myoblastcity function within the muscle (Erickson et al.1997). As a result myoblast city mutant muscles aremononucleated and much smaller, and also NMJ size isseverely reduced. However, the motoneuronal terminalsdo not adjust to this fact, but instead grow ectopicbranches which contain displaced presynaptic structuresFig. 4A–C Summary model of embryonic NMJ formation in Drosophila.The scheme shows the situation before contact (A), in theimmature (B) and the mature junction (C). Stages are indicated topright, symbols in the box below. A At about stage 15 the growthcone (GC) approaches the muscle, both forming filopodial extensionsand expressing cell-specific cell adhesion molecules (specificCAMs) and non-specific Fas2 (Fas2). Repellents (on muscle2) preventestablishment of false contact and transient extensions are retracted.Early CAMs (on growth cone and muscle1) might attracteach other (stippled lines) during the process of target recognition.Glutamate (little dots) is detectable in the terminal around the timeof contact formation and vesicles (stippled circle) must form soon,as first transmission occurs within 30 min after neuromuscular contact.B The early cell junction might be formed by the specificCAMs involved in target recognition (connected rectangles). Basementmembrane (BM) forms and adheres to receptors on muscleand neuron (BM-recept.), perhaps holding the presynaptic terminalagainst the muscle surface. Basement membrane appears to be excludedfrom the synaptic cleft (white arrowheads) by a functiondownstream of the transcription factor mef2 (arrow from nucleus).Glutamate receptors (Glu-recept.) are initially evenly distributedbut, upon neuromuscular contact, cluster in response to presynapticelectrical activity (zigzag arrow). kakapo function mediatesbranching of the presynaptic terminal, and active zones (AZ) startassembling. C Specific CAMs either fade from the NMJ or theybecome restricted to the NMJ (stippled rectangles). Fas2 is restrictedto the NMJ pre- and postsynaptically and appears to manifestthe status of the synapse at late embryogenesis. On the postsynapticside, Fas2 is clustered, together with Shaker channels, by Discslarge (Dlg). Dlg might localise to the NMJ via binding to synapsespecificcytoskeletal elements (three-pronged stars; so far purelyhypothetical) or to synaptic transmembrane proteins (not shown).The amount of Glu-Rs is upregulated as a function of presynapticelectrical activity (arrow from nucleus)
177(Fig. 2D,Q; <strong>Prokop</strong> et al. 1996). A comparable intrinsicdetermination of neurons to elaborate a typically sizedterminal has been demonstrated in vivo for sensory neuronsof crickets, cockroaches and Drosophila (Murpheyand Lemere 1984; Bacon and Blagburn 1992; Canal etal. 1998) and in vitro for motoneurons of crayfish (Arcaroand Lnenicka 1995; Zoran et al. 1996).Like in myoblast city, kakapo mutant embryos showseverely reduced NMJs (and reduced dendritic trees inthe central nervous system; see Fig. 2C for NMJ phenotype).In contrast to myoblast city, embryos carryingstrong alleles of kakapo have normal-sized muscles andmotoneuronal terminals form no ectopic branches. Theseobservations suggest a presynaptic requirement for kakapofunction and, accordingly, anti-Kakapo antisera detectthe protein at motoneuronal terminals (<strong>Prokop</strong> et al.1998b). Cloning data suggest that kakapo encodes a cytoskeletalelement with actin-binding properties (Gregoryand Brown 1998; Strumpf and Volk 1998). Furthermolecular components involved in the growth of the em-