Review. Wiring by Fly: The Neuromuscular System of the Drosophila Embryo. Neuron, Vol. 15, , September, 1995, Copyright 1995 by Ceil Press

Transcription

1 Neuron, Vol. 15, , September, 1995, Copyright 1995 by Ceil Press Wiring by Fly: The Neuromuscular System of the Drosophila Embryo Michael Bate and Kendal Broadie Department of Zoology University of Cambridge Downing Street Cambridge CB2 3EJ England The formation of neuromuscular systems encapsulates many of the issues that lie at the heart of our exploration of the way nervous systems develop: the diversification of neurons and muscles, the oriented growth of axons, the selection of targets, the formation of connections, and the maturation and plasticity of synapses. We would like to be able to dissect the underlying machinery of neural development into its component parts. Ideally, we would use a genetic approach to identify each of the elements required and assess their operation by altering or removing each of them systematically. The tiny Drosophila embryo in its impermeable egg shell must be one of the more obviously unattractive candidates for neurobiologists interested in the formation of synapses. However, the possibility for a rigorous genetic analysis makes Drosophila an important system. The genetic approach requires that we work with embryos because genes encoding functisns essential to the construction of neural networks will certainly mutate to give phenotypes that render the embryo nonviable and unable to hatch. Perhaps the major surprise to date is that, far from being an intractable system, the Drosophila embryo is amenable not only to experimental embryology (Sink and Whitington, 1991c; Broadie et al., 1992) but also to electrophysiology (Broadie and Bate, 1993a, 1993b). This means that genetic methods can be combined with an in vivo analysis of lethal loss of function phenotypes at the embryonic neuromuscular junction (NMJ) (Broadie, 1994). The Neuromuscular System The nerves and muscles of the larval fly are laid out in a straightforward way: about 34 motoneurons (Sink and Whitington, 1991a) innervate 30 muscles in each half segment of the abdominal body wall (Figure 1). All the muscles and many of the motoneurons are known and have been described in detail (reviewed in Bate, 1993; Keshishian et al., 1993). The majority of the motor nerves leave the CNS through a common nerve route containing two principle nerve trunks that separate at an exit glial plexus (Kl&mbt and Goodman, 1991; Van Vactor et al., 1993). The intersegmental nerve (ISN) projects dorsally along the anterior margin of the segment, and the segmental nerve (SN) diverges ventrally to form a consistent pattern of branches (SNa, SNb, SNc, and SNd) innervating ventral and lateral muscle groups (Bate, 1982; Johansen et al., 1989b). The organization of skeletal muscle and its innervation in Drosophila differs significantly from its well-studied vertebrate counterpart. Instead of a muscle formed from a bundle of fibers and innervated by a pool of motoneurons, Review each muscle in the fly is a single syncytial fiber that receives a unique and consistent pattern of innervation. The details of this innervation vary for different muscles: some are innervated by a single motoneuron, whereas others are multiply innervated by the overlapping terminals of up to three different classes of motoneurons (Jan and Jan, 1976; Johansen et al., 1989a; Atwood et al., 1993; Jia et al., 1993). This contrasts strikingly with the organization of the mammalian endplate, which is exclusively occupied by the terminals of one motoneuron. In Drosophila, the innervation is initially focused to a single terminal, which expands, particularly during postembryonic life, to provide a more distributed coverage of the muscle surface. It is during this postembryonic phase, as the terminals expand over the surfaces of the muscles, that the presynaptic endings first become distinguishable, according to the size and morphology of the boutons, as either type 1 (shorter branches, larger boutons) or type 2 (more dispersed branches, numerous smaller boutons) (Johansen et al., 1989a). Recent ultrastructural work suggests that both type 1 and type 2 boutons can be further subdivided on the basis of size and synaptic vesicle types (Atwood et al., 1993; Jia et al., 1993). All terminals contain the excitatory transmitter glutamate, and some contain additional peptides and amines, including insulin, proctolin, and octopamine (Keshishian et al., 1993). At the core of experimental and genetic analyses lies a small group of ventral muscles, innervated by identified motoneurons. The development of the normal pattern of this innervation has been described in detail (Figure 1). Evidence for Specificity As motoneuron growth cones leave the CNS and grow toward their muscle targets, their behavior is strikingly similar to that of their vertebrate counterparts. They defasciculate initially at the exit glial plexus and later at additional, more peripheral choice points. These choice points lie en route to particular parts of the muscle field and subdivide the growing nerves into specific branches (Sink and Whitington, 1991b; Broadie et al., 1993; Van Vactor et al., 1993). A series of consistent, motoneuron-specific choices leads each growth cone into the vicinity of its appropriate muscle, although as the growth cone extends, it puts out filopodia within a region of at least 15 p.m radius (Johansen et al., 1989b; Sink and Whitington, 1991b; Keshishian et al., 1993), which is sufficient to sample many alternative muscle surfaces. Once contact with the target is achieved, branches to other muscles are withdrawn, and a nerve terminal forms at a specific location on the muscle surface. If the target muscle is removed, surgically (Sink and Whitington, 1991c), with a laser (Cash et al., 1992), or genetically (Chiba et al., 1993), the innervating motoneuron grows normally into its proper domain and ramifies over neighboring fibers (Figure 2). These branches are not necessarily withdrawn, and ectopic, functional junctions may form on nearby muscles (Sink and Whitington, 1991c; Cash et al., 1992). The formation of proper connections

2 Neuron 514 Figure 1. Diagram Illustratingthe Organizationof Nerves and Muscles in the Embryonic Body Wall of Drosophila (Left) Two representativeabdominalsegmentsas viewed in an embryoopened dorsally and dissectedflat. The ventral CNS and nervetracts are shown shaded,flanked by the muscles and their innervation,viewed from inside (top) and outside (bottom). (Center) The area outlined by the rectangle(left) is enlargedto show the ventral muscles and their innervation,focusing on muscles 12, 13, 6, and 7 and their innervatingmotoneuronsrp5, RP1 and RP4, and RP3, respectively. (Right) The ventral musclesare viewed from the side, to show the paths taken by innervating motoneuronsas they navigatethrough the muscle field (basedon Van Vactor et al., 1993). ISN, intersegmentalnerve; SNa-SNd, segmental nerve branches. For further details and references,see text. by other nerves, whose targets are still available, is unaffected by the loss of a neighboring muscle (Chiba et al., 1993). On the other hand, there is some evidence that in the absence of its appropriate motoneuron a muscle will elicit the formation of ectopic branches by other neurons (Halfon et al., 1995). In addition, muscles can be duplicated before the arrival of the motor axon that innervates them, and in this case, the appropriate neuron selectively connects with the two fibers and expands its terminal field to form normally sized junctions on both (Chiba et al., 1993). Thus, a great deal seems to depend on a system that matches motoneurons and muscles. Motoneurons apparently select their targets through a specific recognition process, and when they reach this point, they stop growing and withdraw branches to other muscles. In the absence of a target, exploration continues and an ectopic junction may form at another location on a muscle that is already innervated appropriately by another neuron. This ectopic innervation persists, in contrast to the mammalian endplate, where polyneuronal innervation is eliminated in the normal course of development (Redfern, 1970) by competitive interactions between neighboring terminals (reviewed by Hall and Sanes, 1993). Guidance and Targeting Because of the reliability and precision of the connections and the ease with which the pattern of innervation can be revealed by staining embryos with antibodies, the neuromuscular system is ideal for studying the machinery of growth cone guidance and selective connection in combination with genetic screens (Van Vactor et al., 1993). The consistency of motoneuron pathways from embryo to embryo shows that in normal development growth cones are operated on by a robust guidance system that leads each nerve to execute a stereotyped series of movements as it grows from the CNS toward its target. While we have an increasingly detailed description of the cellular landscape through which growth cones navigate (Van Vactor et al., 1993), the molecular landscape is far tess well defined, although some of the principle features are becoming clear. In particular, growth cones and muscles express a series of cell surface and secreted molecules in characteristic patterns that suggest roles in guidance and/or targeting (see Table 1). These include cell adhesion molecules (CAMs) of the immunoglobulin- and leucine-rich families. It might be expected that a genetic screen would reveal more putative guidance molecules, but the pheno-

3 Review: Embryonic Drosophila Neuromuscular System ol 0 o o 0 BO c 0 Figure 2. Nerve Branching and Growth Cone Targeting in the Neuromuscular System (A-C) Misexpression of fasciclin II prevents defasciculation of axons at specific choice points to form subbranches of the motor nerve projection (shown here for SNb [blue]; the ISN is shown as red). In wild-type embryos (A), SNb branches from the common nerve root and innervates a subset of the ventral muscles as shown (not all muscles are included in this diagram). Overexpression of fasciclin II in motoneurons (B and C) causes failures of defasciculation at the branch point, with SNb continuing to grow dorsally along the ISN (B) or stalled branches forming either at the normal branch point or after a detour (C). Similar stalls at specific choice points occur in pathfinding mutants (Van Vactor et al., 1993). Later in development, stalled and misrouted growth cones may connect with target muscles by growing along abnormal paths to reach them (after Lin and Goodman, 1994). (D) In wako and clu embryos, growth cones in SNb fail to discriminate target from nontarget muscles and arborize over ventral muscles without forming specific connections. SNb motoneurons experimentally deprived of their targets branch over the same set of muscles. There appears to be a limited domain of ventral muscles (shown here by blue background; not all muscles are shown) available to this subset of the motoneuron projection for sampling as potential targets (for further details, see text and Van Vactor et al., 1993). types of mutations in genes coding for known CAMs are not encouraging: embryos carrying null mutations in fasciclin III and connectin, both putatively homophilic CAMs expressed on the surfaces of a subset of muscles and their innervating motoneurons, have no detectable phenotype in their neuromuscular systems (Nose et al., 1994; Chiba et al., 1995). It appears that the CAMs described so far function as part of a more complex network of context-dependent cues that provides a well-buffered system for directing growth cones to their destinations. The redundancy (implied by the benign phenotypes of mutations that delete single CAMs) is probably in the system of cues itself, so that the advantages of single gene copies in Drosophila for producing clear-cut loss of function phenotypes no longer apply. So far, the results of genetic screens seem to confirm this view. The sort of systematic errors in the pathways taken by individual axons that might have been predicted to result from the loss of single guidance cues have not yet been reported. As would be expected, some of the genes recovered in these screens code for essential elements of the machinery of growth cone extension that produce failures of axon growth along appropriate nerve pathways; others mutate to give interesting phenotypes that suggest a general failure of axon targeting within a subsection of the muscle field (Van Vactor et al., 1993). Although the loss of individual CAMs does not produce a conspicuous phenotype, and this makes it difficult to ascribe functions to those that have been identified, it is relatively straightforward to misexpress putative guidance molecules using heterologous promoters or the GAL4 system (Brand and Perrimon, 1993). These misexpression experiments produce striking phenotypes that indicate specific functions for the three CAMs so far tested in this way. One, perhaps naive, way of thinking about the different outcomes of these experiments is to compare the loss of function phenotype for the letter a in the word pathfinding (pthfinding) with the misexpression phenotype (paaatahafaianadaianaga). The system has a redundancy in it, which ensures that the loss of a single cue is insufficient to derange the appropriate response, whereas overexpression of the same cue throughout the system has catastrophic consequences. In the case of fasciclin II, which is normally expressed on a subset of CNS axons and on all motoneurons, overexpression in motoneurons produces failures of defasciculation at normal nerve branching points (Figure 2) (Lin and Goodman, 1994). These failures suggest that, in the normal development of the efferent projection, fasciclin tl could serve as a context-dependent CAM. Lower levels of axon-axon adhesion at prospective branch points would allow growth cones to respond specifically to additional cues by defasciculating and forming a subbranch. Thus, for example, when fasciclin II is overexpressed in the ISN, axons normally forming the SNb branch fail to leave the ISN and continue to grow dorsally (Lin and Goodman, Table 1. Cell Surface and Secreted Proteins in the Embryonic Drosophila Neuromuscular System Protein Type Expression Fasciolin II a Ig superfamily Motoneuron growth cones Muscle subset Fasciclin lip Ig superfamily Growth cone subset Muscle subset Neuromusculin c Ig superfamily Most muscles Connectin d Leucine-rich repeat Growth cone subset Muscle subset Toll ~ Leucine-rich repeat Muscle subset Semaphorin IIf Secreted semaphorin Single muscle Grenningloh et al., 1991 ; Van Vactor et al, 1993; Lin and Goodman, b Patel et al., 1987; Snow et al., 1989; Halpern et al., 1991; Chiba et al., Kania et al., Nose et al., 1992, 1994; Gould and White, e Nose et al., 1992; Halfon et al., f Kolodkin etal., 1993; Matthes et al., 1995.

4 Neuron RP1 RP3 C 0 Figure 3. Differential Role of Fasciclin III in Targeting Motoneurons RP1 and RP3 to Specific Muscles within the Ventral Domain (A) Growth cones of both neurons express fasciclin III, but only RP3 is targeted to fasciclin Ill-expressing muscles (6 and 7; green outline). (B) Absence of fasciclin III from growth cones and muscles in fasciclin III null mutant embryos does not affect the specificity of target recognition for either neuron. (C) Ectopic expression of fasciclin Ill in all muscles causes RP3, but not RP1, to make targeting errors. RP3 forms aberrant connections with muscles as shown (thickness of line indicates relative probability of connection). RP1, which normally innervates the fasciclin III negative muscle 13, is unaffected (after Chiba et al. 1995; see text for further details). 1994). Fasciclin II is related to vertebrate neural cell adhesion molecule (N-CAM), and this failure of defasciculation upon overexpression is reminiscent of the aberrant sorting of motor axons that can be induced by enzymatically removing polysialic acid (PSA) in the chick embryo (Tang et al., 1992). N-CAM-mediated axon-axon adhesion falls as the levels of associated PSA increase (Rutishauser et al., 1988), and endogenous levels of PSA are strikingly higher as motoneuron growth cones enter the plexuses in which, like the nerves in the exit glial plexus of Drosophila, they defasciculate and sort out into their specific pathways (Lance-Jones and Landmesser, 1981; Tosney and Landmesser, 1985a, 1985b). When PSA is removed, the motoneurons that normally sort out in the plexus make numerous projection errors. In both the fly and the chick, motor axons fasciculate as a bundle as they grow out into the periphery, and in normal development reductions in this CAM-mediated adhesion could provide the context for the operation of cues that cause the defasciculation and divergence of specific growth cones to form a particular subbranch of the projection (Lin and Goodman, 1994). The nature of these cues is so far undefined, although there is some evidence that the muscle field is subdivided into domains within which growth cones explore and select targets from a subset of the muscles. One reason for supposing that there are such domains is that there appears to be some restriction to the branching of targetless motoneurons: RP3, for example, when deprived of its normal targets, muscles 6 and 7, ramifies within a fairly restricted domain of ventral muscles that would normally include its proper targets (Sink and Whitington, 1991 c). At the same time, genetic screens produce a very interesting targeting phenotype, in addition to the arrested growth phenotypes described earlier. Mutations in two genes, walkabout (wako) and clueless (clu), produce a characteristic aberration in the motoneuron projection in which growth cones innervating ventral muscles via SNb fail to terminate on their proper targets (Figure 2). Instead, they ramify over a restricted domain of ventral muscles that corresponds to the region explored by RP3 in the absence of its targets (Van Vactor et al., 1993). Thus, while the lesions in both clu and wako prevent growth cones from recognizing their specific targets, they leave unaffected a capacity to recog- nize a subsect ion of the muscle field within which the target is located. The implication of these findings is that there is an underlying molecular organization to the muscle field that is critical to the behavior of growing axons. It is important to remember that the muscles, like axons in the CNS, not only provide targets for neuronal connection but also represent a significant substrate for nerve growth. Thus, molecules expressed on muscle surfaces may be at least as significant in routing axons through the muscles as in guiding individual axons to their particular targets. The expression of connectin (Nose et al., 1992; Gould and White, 1992) is particularly interesting in this context. Connectin is normally expressed on a subset of lateral muscles and their innervating growth cones, and in tissue culture assays it seems to promote homophilic adhesion (Nose et al., 1992). However, when connectin is expressed ectopically on ventral muscles lying along the path of axons that would normally enter this domain, these axons stall, although they may subsequently reach their targets by a circuitous route (Nose et al., 1994). The misbehavior of these growth cones could be caused by a nonspecific blocking action of ectopic connectin, but it certainly suggests that, in wild-type embryos, ventrally projecting axons would be barred from the immediately neighboring lateral domain of muscles by the presence of connectin on the muscle surfaces. Thus, while connectin is likely to promote homophilic adhesion between lateral growth cones and their target muscles, it may also serve as an inhibitory guidance cue excluding other axons from this domain (Nose et al., 1994). The conditional nature of growth cone responses to the available cues that this implies is also neatly demonstrated by the results of overexpressing fasciclin III (Figure 3). Like connectin, fasciclin III promotes homophilic adhesion in culture assays (Snow et al., 1989) and is expressed on a subset of muscles and innervating growth cones (Patel et al., 1987; Halpern et al., 1991). Fasciclin III is conspicuously expressed at the muscle 6/7 interface, where motoneuron RP3 forms its terminal, and on the RP3 growth cone as it enters the muscle field. Fasciclin I11 is also expressed by the growth cone of RP1, but RP1 bypasses muscles 6 and 7 and innervates the neighboring fasciclin

5 Review: Embryonic Drosophila Neuromuscular System 517 Ill-negative muscle, number 13. When fasciclin III is ectopically expressed by all muscles under the control of a muscle myosin promoter, the responses of the RP3 and RP1 growth cones are strikingly different (Figure 3): RP3 now wrongly innervates other muscles within the ventral domain, with or without forming a connection with muscle 6/7, while the RP1 growth cone is unaffected and continues accurately to target its appropriate muscle, number 13 (Chiba et al., 1995). This experiment shows that there is a significant difference between the two growth cones that contributes to the specificity of normal targeting. RP3 is unaffected by the absence of fasciclin Ill in mutant embryos (Chiba et al., 1995), showing that there is at least one other cue involved in its targeting to 6/7. Nonetheless, as shown by the overexpression experiments, RP3 is influenced by levels of fasciclin III on neighboring cells that it samples within its target domain. On the other hand, RP1, which is fasciclin Ill-positive but innervates a fasciclin Ill-negative muscle, is indifferent to fasciclin tll expression on neighboring cells or ectopic fasciclin III on its target muscle. By focusing on the different responses of two growth cones to the same experimentally manipulated cue, the experiment reveals a small part of the likely structure of choices and responses that guides growth cones to their targets. RP3 is diverted by fasciclin III expression into establishing connections with muscles on which it might normally form transitory branches. RP1 is indifferent to fasciclin III both as a cue and as a possible interference in forming connections with its normal target. The behavior of each growth cone indicates that the cell, or more likely its growth cone surface, contains an integrative mechanism that responds in specific ways to the set of environmental cues presented to it and feeds onto the machinery of growth cone extension. Termination The transition from growth and target selection to termination and synaptogenesis involves a radical change of state, including the withdrawal of branches to neighboring muscles and the formation of differentiated presynaptic endings. Presumably, this switch involves the activation of a signaling pathway in innervating motoneurons by an interaction between their growth cones and the surfaces of muscles. This is a general pathway activated in all motoneurons upon reaching their targets. The evidence that the specific muscle normally innervated by a neuron is not essential for its axon to terminate and form a functional synapse (Whitington, 1985; Cash et al., 1992) shows that the transition can be activated by contact with neighboring muscles as well as with the target itself. Thus, the particular combination of cues that identifies the target is not essential for triggering termination, but is a condition that favors it. Ectopic expression of a molecule such as connectin, which blocks growth cone extension for some axons, also blocks terminal formation on those muscles that ectopically express it (Nose et al., 1994). Growth cones that stall in response to ectopic connectin but reach their target by a circuitous route will terminate normally only if the target itself is not expressing connectin. Ectopic expression of fasciclin III, a homophilic CAM, diverts some fasciclin Ill-expressing growth cones into forming terminals on neighboring muscles that now express fasciclin Ill (Chiba et al., 1995). Thus, adhesion or barriers to adhesion may be important determinants of termination or continued growth and exploration. In this view, a domain of muscles (such as that defined by the clu and wako phenotypes; Van Vactor et al., 1993) would provide a permissive landscape for growth cone extension and exploration by an appropriate subset of the motoneurons. We might speculate that termination would be possible within this domain but would be triggered first by the particular combination of cues that normally occurs only at the target muscle and brings each nerve into close contact with a muscle surface. However, if the duration of contact were important as well as its strength, then termination would be triggered rapidly by close contact with the target but more slowly on other muscles. Under these conditions functional synapses would eventually develop on neighboring muscles if the target muscle were removed. If branches are prevented from forming on potential targets, as occurs specifically for RP3 in the presence of ectopically expressed semaphorin II in the ventral muscle domain, then terminal formation is blocked until the inhibitory cue disappears (Matthes et al., 1995). Termination itself might be a process with several steps, beginning with an arrest to growth. For example, in mice, S-laminin, a component of the subsynaptic basal lamina at the NMJ, contains a site that is specifically adhesive for motoneurons (Hunter et al., 1991) and blocks further growth of their axons (Porter et al., 1995; Sanes, 1995). The Neuromuscular Synapse as a Model System Since the formation of neuromuscular connections is essentially a problem of cellular morphogenesis, it can be effectively analyzed using antibody probes or dyes that reveal the extent to which the normal pattern of connections forms in wild-type, mutant, and experimental embryos (Van Vactor et al., 1993). However, while it is patterns of connectivity like this that give synaptic transmission its behavioral significance, there are additional issues associated with the function and development of synapses that are central to any understanding of neural circuits (for review, see Jessell and Kandel, 1993). First, it is important to understand in detail the specialized form of secretion that allows for rapid Ca2+-controlled vesicle release from the presynaptic terminal in response to neural activity (reviewed by SOdhof, 1995). Second, because an essential feature of transmission is that it is focused at specialized sites on pre- and postsynaptic cells, we need to know how the developmental process organizes the formation and localization of the pre- and postynaptic terminals (reviewed by Hall and Sanes, 1993). Third, while vesicle release is a rapid response to neural activity, the actual characteristics of transmission at the synapse depend on local conditions, in particular on the long-term effects of interactions between pre- and postsynaptic terminals. Clearly, therefore, we need to understand how the possibility for adjustment at the synapse is extended beyond the phase of its construction and how this plasticity is regulated to allow for long-term change, learning, and memory.

6 Neuron 518 Ideally, to look at any of these essential attributes, we should work with a synapse with accessible pre- and postsynaptic terminals, where we can record transmission, manipulate or delete synaptic proteins, observe and experiment with development, and analyze the long-term consequences of changes such as alterations in the level of neural activity. No synapse fulfills all these conditions. However, it is clear that there are considerable advantages in being able to work with mutated synaptic proteins, particularly if the effects of these alterations can be monitored directly by assaying synaptic transmission. In the mouse, for example, it has been possible to use slice preparations or cultured embryonic neurons to make extracellular or whole-cell patch recordings of synaptic activity from animals with targeted knockouts for several synaptic proteins, including synaptotagmin (Geppert et al., 1994b), synapsins (Rosahl et al., 1993, 1995), and the small GTPbinding protein Rab3A (Geppert et al., 1994a). At the same time, genetic screens provide a way of identifying novel synaptic proteins. For example, in the nematode, although it is not possible to monitor synaptic activity directly, behavioral phenotypes such as those of the uncoordinated (Unc) class can be used to identify mutations in genes that are required for neural or muscular activity. As would be expected, some of the genes identified in this way have now been shown to encode proteins that are essential for synaptic transmission (reviewed by Jorgensen and Nonet, 1995). However, it is so far only at the Drosophila NMJ that the two approaches of assaying for the effects of mutated proteins and of screening for new proteins can be combined with a synapse that is accessible for electrophysiological recording of synaptic transmission. In addition, the NMJ can be studied as it develops and during an extended phase of plastic modification. It is this unique combination of features that makes the Drosophila NMJ such a useful system for investigating synapses and synaptic transmission. Physiological Characteristics of the Embryonic NMJ An essential feature of the Drosophila work is that it is possible to carry out physiological experiments on the NMJ in the embryo (Broadie and Bate, 1993a, 1993b; Kidikoro and Nishikawa, 1994). This means that it is possible to assay the phenotypic consequences of mutating or deleting synaptic proteins by making a detailed analysis of the effects on synaptic transmission, even if, as is usually the case, the overall effect of such mutations is lethal. Physiological analysis at the embryonic NMJ depends almost entirely on using whole-cell patch-clamp techniques to record responses in the postsynaptic muscle. Muscles are voltage clamped, and excitatory junctional currents (EJCs) are recorded in response to evoked or spontaneous transmitter release from the presynaptic nerve ending. At the same time, the characteristics of the postsynaptic receptor field can be tested directly by recording the response of the muscle to iontophoretic release of the transmitter, glutamate. The significance of the recorded currents depends on the extent to which they truly reflect events in the muscle that is clamped. At early stages, there is extensive electrical and dye coupling between neigh- boring muscles (Johansen et al., 1989b; Broadie and Bate, 1993a), which is mediated by gap junctions (Gho, 1994). At the time of first contact between growth cones and muscles, this dye coupling is lost (Johansen et al., 1989b; Broadie and Bate, 1993a), although there is persistent electrical coupling between muscles that are attached to each other at their ends (Kidikoro and Nishikawa, 1994). Despite this coupling, muscles can be successfully clamped throughout embryogenesis and during early larval stages. However, because neighboring muscles are electrically coupled, two classes of endplate currents are detectable: one with a fast rise and decay, representing the true endplate current of the clamped muscle, and a lower amplitude current with a much slower time course that reflects the spread of current from neighboring cells (Kidikoro and Nishikawa, 1994). There are no direct methods for recording from the presynaptic terminal, although it has proved possible to use a macropatch technique to record the contribution of single presynaptic boutons to the EJC in larval muscles (Kurdyak et al., 1994). There are also elegant optical methods for monitoring vesicle traffic in the presynaptic terminal at the late larval NMJ (Ramaswami et al., 1994). Nonetheless, knowledge of the presynaptic terminal and its constituent proteins is far more advanced than our understanding of the postsynaptic side of the junction. The reasons for this imbalance lie in very rapid advances in the understanding of the constitutive and regulated pathways of secretion, and in the discovery that the machinery of secretion is highly conserved in eukaryotic cells (reviewed by Ferro- Novick and Jahn, 1994). This, together with methods for the purification of abundant synaptic constituents, has led to the identification of a conserved set of interacting proteins associated with synaptic vesicles and their release sites (reviewed by S~dhof, 1995). Many of these proteins have now been detected or identified in Drosophila (Figure 4), and there has been a flurryof activity directed at revealing their function in synaptic vesicle release by an analysis of their mutant phenotypes at the NMJ in the Drosophila embryo. Presynaptic Proteins and the Machinery of Transmitter Release Synaptic vesicle release involves a complex of proteins: some are elements of the general secretory pathway, and others are unique to the synapse, and therefore likely to be specialized elements of the Ca2 -controlled release mechanism. In Drosophila, identified components of the constitutive secretory pathway include the presynaptic membrane protein syntaxin (syx; Schulze et al., 1995), the integral vesicle protein synaptobrevin (syb; DiAntonio et al., 1993a), and the cytoplasmic proteins rop (Salzberg et al., 1993; Harrison et al., 1994; Schulze et al., 1994), N-ethylmaleimide-sensitive fusion protein (dnsf), and soluble NSF attachment protein (d~snap; Ordway et al., 1994). There are two proteins so far identified that are unique to the synapse: synaptotagmin (syt; Perin et al., 1991), an integral synaptic vesicle protein, and cysteine string protein (CSP), which is vesicle associated (Zinsmaier et al., 1990, 1994; Van de Goor et al., 1995).

7 Review: EmbryonicDrosophilaNeuromuscularSystem 519 S'~ ~VT ROP l Figure 4. RepresentativeNMJs from Third Instar LarvaeRevealedby Stainingwith Antibodiesagainst Synaptic Proteins Syntaxin (SYX) (Schulzeet al., 1995); synaptotagmin(syt) (Littletonet al., 1993b); rop (ROP) (Harrisonet al., 1994); and synaptobrevin(syb) (Shone et al., 1993).Terminals(largearrows)are type 1. Syntaxinand rop proteinsare presentin axons as well as terminals;synaptotagminand synaptobrevinappearto be confinedto terminals. For further details,see text. N, motor nerves. The similarity of the components associated with vesicle release at the synapse in animals as diverse as Drosophila, the nematode, and vertebrates is remarkable and reflects the general conservation of the secretory pathway from yeast to higher eukaryotes. All of the components so far identified are coded for by single genes in Drosophila, with the exception of synaptobrevin, which (as in vertebrates) is present as a neural-specific form, n-syb, and a ubiquitous form, syb (Sfidhof et al., 1989; Chin et al., 1993; DiAntonio et al., 1993a). The absence of multiple forms of synaptic proteins facilitates the analysis of mutant phenotypes but may present real difficulties when the developmental effects of loss of function in the general pathway of secretion are sufficient to mask specific effects on exocytosis at the NMJ. For example, rop encodes the fly homolog of yeast Seclp (Salzberg et al., 1993; Harrison et al., 1994), a protein essential for late stages of secretion (Novick and Schekman, 1979). rop is expressed in many tissues in the fly embryo, and null mutations produce pleiotropic effects that effectively prevent the embryonic NMJ being used to assay the role of rop in synaptic transmission (Harrison et al., 1994). However, the fly provides an answer to difficulties of this kind in the form of temperature-sensitive mutations and simple assays for synaptic transmission in the adult eye. While these assays, which depend on summed extracellular recordings from the retina, give only a relatively crude estimate of the presence or absence of synaptic transmission, they provide a backup technique that may be essential for detecting or analyzing the synaptic phenotypes of mutations in genes required early in development. In the case of rop, the work with the eye shows that temperature-sensitive mutations cause a reversible block in synaptic transmission at the restrictive temperature, possibly associated with a depletion in the pool of vesicles available for fusion (Harrison et al., 1994). The likelihood that some genes will be required generally for embryogenesis is an important consideration for any genetic screen designed to identify novel synaptic proteins. It may well be that at least some of the genes involved will be identified first as viable hypomorphs with effects on adult behaviour, dnsf, for example, which has recently been cloned (Ordway et al., 1994), has now been shown to be the product of comatose (Pallanck et al., 1995), a gene originally isolated on the basis of mutations that cause temperature-sensitive paralysis in the adult (Siddiqi and Benzer, 1976). Given that NSF is required generally for secretion, null mutations in comatose might be expected to disrupt embryonic development and, like

8 Neuron 520 mutations in rop, prevent detailed analysis of the synaptic phenotype at the NMJ. Despite this caveat, for most of the synaptic proteins so far identified, it has been possible to use the embryonic NMJ as an assay for their function in the presynaptic secretory pathway. Since the function of synaptic proteins for the most part has been deduced from biochemical experiments or by manipulating the proteins in artificial systems, and not by detailed analysis of transmission at a synapse, the results of this work are particularly timely. For example, monitoring the effects on synaptic transmission of removing synaptobrevin and syntaxin provides a stringent test of the hypothesis (SSllner et al., 1993) that vesicle docking requires a specific interaction between an integral vesicle protein (v-snar E: n-syb) and an integral target membrane protein (t-snare: syx). There are two general ways of manipulating the secretory machinery at the synapse: using bacterial toxins that inactivate specific components or working with null mutations that remove individual proteins, n-syb is cleaved specifically by tetanus toxin (Niemann et al., 1994). Previous experiments at synapses with these toxins have relied on injection to introduce the toxin into the presynaptic terminal (e.g., Mochida et al., 1990; Hunt et al., 1994). One enormous advantage in Drosophila is that the tetanus gene itself can be expressed in the presynaptic cell under the control of a neural-specific promoter (Sweeney et al., 1995). This not only removes the need for experimental manipulations such as injection but also ensures that the toxin is present at high levels throughout the life history of the presynaptic cell, so that there is likely to be complete inactivation of the target protein, n-syb. The effect of removing n-syb in this way is to block all evoked synaptic transmission, although spontaneous miniature EJCs are still detectable (Sweeney et al., 1995; Broadie et al., 1995). In the case of syntaxin, in null mutant embryos there is no transmission at the NMJ, either in the form of evoked EJCs or spontaneous miniatures (Schulze et al., 1995). However, the ultrastructure of the NMJ is remarkably normal in the absence of syb or syntaxin, with vesicles clustered and docked at release sites (Broadie et al., 1995). In both cases, the release of synaptic vesicles can be provoked by the nonspecific stimulus of exposing the junction to hyperosmotic saline. The implication seems to be that, despite the predictions of the SNARE hypothesis, neither protein is required for vesicle docking, at least as detectable ultrastructurally (Broadie et al., 1995). n-syb is clearly not required for fusion (spontaneous fusions continue in its absence) but seems to be an essential element in the evoked fusion pathway. On the other hand, syntaxin appears to be essential for fusion itself, and this probably reflects a central role for syntaxin in the mechanism of exocytosis in neural and nonneural cells (Schulze et al., 1995). In contrast, n-syb is specific to the synapse and is likely to be a specialized component of the evoked release pathway. There is, of course, likely to be a family of proteins required for the coupling of Ca 2+ and rapid vesicle release at the synapse, and these proteins will be neural specific. CSP appears to be of this class (Zinsmaier et al., 1994). The phenotype of syt mutants at the embryonic NMJ suggests that it too is part of this synapse-specific pathway. In null syt mutants, evoked vesicle release is reduced, but spontaneous vesicle fusion is enhanced (DiAntonio et al., 1993b; Littleton et al., 1993a). This would support the view that syt acts as a clamp that blocks vesicle fusion in the absence of Ca 2+ and enhances the efficiency of excitation secretion coupling (Broadie et al., 1994; Littleton et al., 1994). The Development of the Neuromuscular Synapse Synaptogenesis is extremely rapid at the embryonic NMJ, taking about 8 hr, from the first contacts between motoneuron growth cones and muscles to the mature synapse just before hatching. All stages of the process can be studied in living embryos that have been dissected or cultured during the late stages of embryogenesis (Broadie et al., 1992). The earliest signs of transmission between nerve and muscle occur as growth cones explore the muscle field, and small amplitude EJCs can be recorded either endogeneously or in response to motor nerve stimulation (Broadie and Bate, 1993a; Kidikoro and Nishikawa, 1994). Since the first detectable transmission between nerve and muscle occurs before the motoneuron makes a restricted terminal and withdraws its other branches, it is not excluded that these early interactions are part of the process that leads to terminal formation. Clearly, the synthesis of at least some of the components of the functional synapse begins before contacts between nerve and muscle are established. Transmitter expression, detected by antibodies, begins as motoneurons enter the m uscle field, although synaptic bouton-like structures first appear several hours later at the newly formed endings of motoneurons, when alternate branches have been withdrawn (Broadie and Bate, 1993a). Expression of glutamate receptor (GluR) mrna (Schuster et al., 1991) in muscles also occurs early, several hours before motor axons enter the muscle field, and may even occur in single myoblasts prior to fusion with forming muscles (Currie et al., 1995), However, the first functional GluRs are not seen (with iontophoresis) until after initial nerve muscle contacts (Broadie and Bate, 1993a), although the extensive electrical coupling between adjacent muscles may hinder their detection up to this point. The NMJ between motoneuron RP3 and muscle 6 (see Figure 1) has been the focus for most of the studies on synaptic development and synaptic transmission in the embryo so far. It matures from a rapidly fatiguing junction that generates low amplitude EJCs during early stages to a robust synapse with a larger amplitude EJC that continues to increase until the end of embryogenesis (Broadie and Bate, 1993a; Kidikoro and Nishikawa, 1994), as a consequence of an enrichment of functional GluRs at the postsynaptic site. There is also a steady increase in the consistency and reliability of transmission that seems to represent a true maturation of the junction, because, in contrast to the steadily changing characteristics of transmission at the synapse, there is a consistent response (of increasing amplitude) to iontophoretically applied glutamate from the earliest stages (Broadie and Bate, 1993a).

9 Review: Embryonic Drosophila Neuromuscular System 521 Work is now in progress to provide a detailed description of the ultrastructural development of the NMJ to complement the electrophysiological evidence of progressively maturing synaptic transmission. While the electrophysiological time line of junctional development provides a basis for the analysis of phenotypes in mutations that affect synaptogenesis, it also poses obvious questions about mechanisms. These questions are elementary and partially solved at the vertebrate endplate (Hall and Sanes, 1993), but they have to be reinvestigated at the embryonic NMJ in the fly to find out the extent to which the developmental mechanisms of synaptogenesis have been conserved. For example, one of the fundamental questions associated with the development of any neural circuit is the extent to which the development of preand postynaptic elements is autonomous or interdependent. The prospects at the embryonic NMJ for a thorough analysis of this question and of the nature of the underlying mechanisms are good, particularly because it is possible to study the development of the junction in embryos with genetic lesions that specifically affect pre- or postsynaptic cells (Broadie and Bate, 1993c). So far, the work in the fly has focused on the postsynaptic effects of preventing innervation during the normal period of synaptogenesis, or of reducing or eliminating activity in the presynaptic neuron. In mutant embryos with delayed or absent motor innervation, the initial phases of postsynaptic differentiation (i.e., the localized expression of adhesion molecules such as fasciclin Itl and the synthesis of functional GluRs) occur in the absence of the presynaptic endings. Although the electrical and contractile properties of the muscle as a whole can also develop without innervation (Broadie and Bate, 1993d), the further differentiation of the postsynaptic site requires the presence of a presynaptic ending (Broadie and Bate, 1993c). For example, as in the case of the vertebrate NMJ, transmitter receptors cluster postsynaptically beneath the presynaptic terminals of the developing N MJ in the embryonic fly. However, in the absence of innervation, the receptors remain diffusely distributed over the muscle surface. In addition, when innervation is delayed or absent, later phases of increased functional GluR expression at the terminal also fail to occur (Broadie and Bate, 1993c). By contrast, ectopic innervation redirects the localization of the receptor field to a novel site beneath the ectopic presynaptic endings. Thus, as at the vertebrate endplate (Hall and Sanes, 1993), both receptor clustering and the full expression of functional receptors by the postsynaptic cell are dictated by presynaptic input. Nonetheless, some of the mechanisms involved in the assembly of the postsynaptic receptor field in Drosophila may be different from those in the vertebrates, and this may reflect the specialized nature of the mammalian twitch fiber, which is the basis for most of the vertebrate studies. For example, evidence from in situ hybridization (Currie et al., 1995) shows that there is no localization of GluR transcripts to the subsynaptic region, as there is at the vertebrate endplate (Merlie and Sanes, 1985; Fontaine et al., 1988). Furthermore, attempts to discover Drosophila homologs of the agrins (McMahan, 1990), which have a central role in receptor clustering in vertebrates, have so far failed. It is not yet clear whether there are Drosophila homotogs of ARIA (Falls et al., 1990) or CGRP-like molecules (New and Mudge, 1986; Fontaine et al., 1986), which, in vertebrates, are involved in mediating the enhanced expression of acetylcholine receptors during synaptogenesis. Another apparent difference is highlighted by a requirement for electrical activity in the innervating motoneuron. Using temperature-sensitive mutations that block nerve transmission, it is possible to make a precise assay of the requirement for presynaptic activity during the formation of the NMJ. The surprising finding from these studies is that a low level of neuronal activity is essential both for receptor clustering and for the subsequent enrichment of receptors at the maturing NMJ (Broadie and Bate, 1993e). However, the exact role of this presynaptic activity is unclear, since vesicle-mediated synaptic transmission is not required for receptor clustering or enrichment, and evoked activity in the muscle is therefore apparently not involved (Broadie et al., 1994). In vertebrates, by contrast, receptor clustering seems to be activity independent, and instead, electrical activity has the effect of down-regulating receptor expression in nuclei outside the endplate (reviewed by Hall and Sanes, 1993). In this case, electrical activity is operating through a postsynaptic pathway that depends on muscle activity. However, the vertebrate twitch fiber is a specialized muscle, and it is likely that at least some of the regulatory mechanisms that lead to a localized synthesis of receptors beneath the endplate reflect the fact that only a small proportion of the muscle nuclei lie close to the synaptic site. In Drosophila, where the fibers are smaller and all nuclei are relatively close to synaptic terminals, there is an enrichment of GluRs beneath the presynaptic endings, but no requirement for a special class of transcriptionally active subsynaptic nuclei. Synaptic Plasticity In Drosophila, growth and modification of the NMJs continues postembryonically. In particular, the branched patterns of endings formed by the motoneurons, especially those of the type II class, extend over the muscle surfaces as the muscles increase in size during larval life. This process of growth and elaboration of the presynaptic endings is modulated by electrical activity (Budnik et al., 1990), an important discovery which suggests that it will be possible to use the Drosophila NMJ to dissect machinery involved in the long-term modification and adjustment of synaptic structure and function. Hyperexcitability induced by mutations in genes encoding K channels (e.g., ether-ago-go [eag], Shaker [Sh] double mutants) causes an expansion of the larval NMJ and an increase in the number of presynaptic branches and boutons (Budnik et al., 1990). An important clue to the mechanism involved in this plasticity at the NMJ is the finding that there is a similar expansion of the presynaptic endings in flies mutant for the gene dunce (dnc). dnc encodes phosphodiesterase II, an enzyme that hydrolyzes camp (Byers et al., 1981), and in dnc mutants intracellutar camp levels are raised (Byers et al., 1981). Mutations in dnc interact with mutations in eag or Sh to enhance the expansion of the presynaptic

10 Neuron 522 terminal, and this suggests that, at least in part, the effect of hyperexcitabilityon the presynaptic endings is mediated through the camp pathway. Strong confirmation for this view comes from the finding that mutations in another gene, rutabaga (rut), interact with mutations in dnc, supressing both the expansion of the terminal in dnc mutants and the enhanced expansion in dnc; Sh double mutants (Zhong et al., 1992). rut encodes a subunit of the Ca2+/ calmodulin-dependent adenylate cyclase, and mutations in rut depress the levels of camp (Dudai and Zvi, 1984; Livingstone et al., 1984), Thus, it seems highly likely that expansion of the terminal is mediated by an activitydependent accumulation of Ca 2, which in turn operates on the camp second messenger cascade. Interestingly, mutations in dnc and rut also affect the characteristics of transmission at the NMJ, disrupting both facilitation and potentiation (Zhong and Wu, 1991). Both dnc and rutwere originally isolated as learning and memory mutants (Dudai et al., 1976; Aceves-Pina et al., 1983). Technically, it would be extremely difficult to work with the central circuitry involved in the learning tasks that are disrupted in these mutants, to show directly how the camp pathway is involved in long-term synaptic change. Although the biological significance of modifying the NMJ is unclear, the structural and functional alterations that occur in these mutant backgrounds are important because the NMJ seems to provide a way of working directly with at least some of components that affect the plasticity of the synapse. The strength of this approach is exemplified by the recent finding that the gene amnesiac encodes a neuropeptide homologous to mammalian pituitary adenylyl cyclase activating peptide (PACAP), which plays an essential part in learning retention (Feany and Quinn, 1995). Zhong and PeSa (1995) have shown that a second PACAP-like peptide is expressed in the terminals of larval motoneurons at the NMJ. This peptide is released by high frequency stimulation of the motoneurons and causes a slow depolarization of the muscle membrane and a delayed activation of K channels, which enhances the outward K + current about 100-fold. An elegant analysis of the peptide's mode of action at the NMJ shows that it operates through the rut-encoded adenylate cyclase and the Ras/Raf second messenger pathway and requires the cooperative activation of both (Zhong, 1995). The precision of this analysis and the identification of the Ca2+/calmodulin-dependent adenylate cyclase as a key element in plastic modification of the NMJ confirm the view that the NMJ is an exceptionally valuable tool for the analysis of genetically encoded synaptic mechanisms, coinciding as it does with extraordinarily rapid advances in our understanding of the molecular mechanisms of learning and memory in the fly operating through the camp pathway (Yin et al., 1994, 1995; Kandel and Abel, 1995). Conclusion The neuromuscular system of the Drosophila embryo offers a way of making a genetic and cellular analysis of the mechanisms underlying neural connectivity and synaptogenesis, Because analysis of embryos is feasible, it is pos- sible to screen directly for the effects of lethal mutations and use these to identify genes likely to be essential for the normal development of the nervous system and behavior. At the same time, the NMJ can be used to make a rigorous analysis in vivo of the effects of loss of function on synaptic transmission. This combination of genetics and electrophysiology in the embryo means that it is possible to move directly from the identification of genes and molecules to an interpretation of their functional significance in the wild-type nervous system. It is clear that developmental mechanisms and the molecules that run them are conserved across wide stretches of evolutionary time. The genetic analysis of the developing nervous system in Drosophila has the potential to lead directly to the essential machinery of neural development in other organisms. As Seymour Benzer, who began the enterprise, wrote in 1971, "Experience thus far with the fly as a model system for unravelling the path from the gene to behavior is encouraging. In any case, it is fun." Acknowledgments The authors' work is supported by grants from the Wellcome Trust. We are very grateful to our colleagues for their comments, and in particular to Mary Baylies, Andrea Brand, and Nick Brown for their constructive criticism and advice. August 18, 1995 References Aceves-Pina, E. O., Booker, R., Duerr, J. S., Livingstone, M. S., Quinn, W. G., Smith, R. F., Sziber, P. P., Tempel, B. L., and Tully, T. P. (1983). Learning and memory in Drosophila, studies with mutants. Cold Spring Harbor Syrup. Quant. Biol. 48, Atwood, H. L., Govind, C. K., and Wu, C.-F. (1993). Differential ultrastructure of synaptic terminals on ventral longitudinal abdominal muscles in Drosophila larvae. J. Neurobiol. 24, Bate, C. M. (1982). Pioneer neurons and axonal pathways. Neurosci. Res. Prog. Bull. 20, Bate, M. (1993). The mesoderm and its derivatives. In The Development of Drosophila melanogaster, M. Bate and A. Martinez-Arias, eds. (Cold Spring Harbor, New York: Cold Spring Harbor Press). Benzer, S. (1971). From the gene to behavior. J. Am. Med. Assoc. 218, Brand, A. H., and Perrimon, N. (1993). Targeted gene expression as a means of altering cell fates and generating dominant phenotypes. Development 118, Broadie, K. (1994). Synaptogenesis in Drosophila: coupling genetics and electrophysiology. J. Physiol. (Paris) 88, Broadie, K. S., and Bate, M. (1993a). Development of the embryonic neuromuscular synapse of Drosophila melanogaster. J. Neurosci. 13, t Broadie, K. S., and Bate, M (1993b). Development of larval muscle properties in the embryonic myotubes of Drosophila melanogaster. J. Neurosci. 13, Broadie, K. S., and Bate, M. (1993c). Innervation directs receptor synthesis and localization in Drosophila embryo synaptogenesie. Nature 361, Broadie, K. S., and Bate, M. (1993d). Muscle development is independent of innervation during Drosophila embryogenesis. Development 119, Broadie, K, S., and Bate, M. (1993e). Activity-dependent development of the neuromuscular synapse during Drosophila embryogenesis. Neuron 11, Broadie, K. S., Skaer, H., and Bate, M. (1992). Whole embryo culture

11 Review: Embryonic Drosophila Neuromuscular System 523 of Drosophila: development of embryonic tissues in vitro. Roux's Arch. Dev. Biol. 201, Broadie, K. S., Sink, H,, Van Vactor, D., Fambrough, D., Whitington, P. M., Bate, M., and Goodman, C. S. (1993). From growth cone to synapse: the life history of the RP3 motor neuron. Development (Suppl.), Broadie, K. S., Bellen, H. J., DiAntonio, A., Littleton, J. T., and Schwartz, T. L. (1994). The absence of synaptotagmin disrupts excitation-secretion coupling during synaptic transmission. Proc. Natl. Acad. Sci. USA 91, Broadie, K. S., Prokop, A., Bellen, H. J., O'Kane, C. J., Schulze, K. L., and Sweeney, S. T. (1995). Syntaxin and synaptobrevin function downstream of vesicle docking in Drosophila. Neuron, this issue. Budnik, V., Zhong, Y., and Wu, C.-F. (1990). Morphological plasticity of motor axons in Drosophila mutants with altered excitability. J. Neurosci. 10, Byers, D., Davis, R. L., and Kiger, J. A. (19'81). Defect in cyclic AMP phosphodiesterase due to the dunce mutation of learning in Drosophila melanogaster. Nature 289, Cash, S., Chiba, A., and Keshishian, H. (1992). Alternate neuromuscular target selection following the loss of single muscle fibers in Drosophila. J. Neurosci. 12, Chiba, A., Hing, H., Cash, S., and Keshishian, H. (1993). The growth cone choices of Drosophila motorneurons in response to muscle fibre mismatch. J. Neurosci. 13, Chiba, A., Snow, P., Keshishian, H., and Hotta, Y. (1995). Fasciclin III as a synaptic target recognition molecule in Drosophila. Nature 374, Chin, A. C., Burgess, R. W., Wong, B. R., Schwartz, T. L., and Scheller, R. H. (1993). Differential expression of transcripts from syb a Drosophila melanogaster gene encoding VAMP (synaptobrevin) in nonneuronal cells. Gene 131, Currie, D. A., Truman, J. W., and Burden, S. J. (1995). Drosophila glutamate receptor RNA expression in embryonic and larval muscle fibres. Dev. Dyn. 203, DiAntonio, A., Burgess, R. W., Chin, A. C., Deitcher, D. L., Scheller, R. H., and Schwartz, T. L. (1993a). Identification and characterization of Drosophila genes for synaptic vesicle proteins. J. Neurosci. 13, DiAntonio, A., Parfitt, K. D., and Schwartz, T. L. (1993b). Synaptic transmission persists in synaptotagmin mutants of Drosophila. Cell 73, Dudai, Y., and Zvi, S. (1984). Adenylate cyclase in the Drosophila memory mutant rutabaga is defective in its responsivenes to Ca 2+. NeuroscL Lett. 47, Dudai, Y., Jan, Y. N., Byers, D., Quinn, W. G., and Benzer, S. (1976). dunce, a mutant of Drosophila deficient in learning. Proc. Natl. Acad. Sci. USA 73, Falls, D. L., Harris, D. A., Johnson, F. A., Morgan, M. M., Corfas, G., and Fischbach, G. D. (1990). Mr 42,000 ARIA: a protein that may regulate the accumulation of acetyicholine receptors at developing chick neuromuscular junctions. Cold Spring Harbor Symp. Quant. Biol. 50, Feany, M. B., and Quinn, W. G. (1995). A neuropeptide defined by the Drosophila memory mutant amnesiac. Science 268, Ferro-Novick, S., and Jahn, R. (1994). Vesicle fusion from yeast to man. Nature 370, Fontaine, B., Klarsfeld, A., H6kfelt, T., and Changeux, J.-P. (1986). Calcitonin gene-related peptide, a peptide present in spinal cord motoneurons, increases the number of acetylcholine receptors in primary culture of chick embryo myotubes. Neurosci. Lett. 71, Fontaine, B., Sassoon, D., Buckingham, M., and Changeux, J.-P. (1988). Detection of the nicotinic acetylcholine receptor o~-subunit mrna byin situ hybridization at neuromuscular junctions of 15-day-old chick striated muscles. EMBO J. 7, Geppert, M., Bolshakov, V. Y., Siegelbaum, S. A., Takei, K., De Camilli, P., Hammer, R. E., and SQdhof, T. C. (1994a). The role of Rab3A in neurotransmitter release. Nature 369, Geppert, M., Goda, Y., Hammer, R. E., Li, C., Rosahl, T. W., Stevens, C. F., and SQdhof, T. C. (1994b). Synaptotagmin I: a major Ca 2+ sensor for transmitter release at a central synapse. Cell 79, Gho, M. (1994). Voltage clamp analysis of gap junctions between embryonic muscles in Drosophila. J. Physiol. 481, Gould, A. P., and White, R. A, (1992). Connectin, a target of homeotic gene control in Drosophila. Development 116, Grenningloh, G., Rehm, E. J., and Goodman, C. S. (1991). Genetic analysis of growth cone guidance in Drosophila: Fasciclin II functions as a neuronal recognition molecule. Cell 67, Hatfon, M. S., Hashimoto, C., and Keshishian, H. (1995). The Drosophila Toll gene functions zygotically and is necessary for proper motoneuron and muscle development. Dev. Biol. 169, Hall, Z. W., and Sanes, J. R. (1993). Synaptic structure and development: the neuromuscular junction. Cell 721Neuron 10 (Suppl.), Halpern, M. E., Chiba, A., Johansen, J., and Keshishian, H. (1991). Growth cone behavior underlying the development of stereotypic synaptic connections in Drosophila embryos. J. Neurosci. 11, Harrison, S. D., Broadie, K., Van de Goor, J., and Rubin, G. M. (1994). Mutations in the Drosophila Rop gene suggest a function in general secretion and synaptic transmission, Neuron 13, Hunt, J. M., Bommert, K., Charlton, M. P., Kistner, A., Habermann, E., Augustine, G. J., and Betz, H. (1994). A post-docking role for synaptobrevin in synaptic vesicle fusion. Neuron 12, Hunter, D. D., Cashman, N., Morris-Valero, R., Bullock, J. W., Adams, S. P, and Sanes, J. R. (1991). An LRE (leucine-arginine-glutamate) dependent mechanism for adhesion of neurons to S-laminin. J. Neurosci. 11, Jan, L. Y., and Jan, Y. N. (1976). Properties of the larval neuromuscular junction in Drosophila melanogaster. J. Physiol. 262, Jessell, T. M., and Kandel, E. R. (1993). Synaptic transmission: a bidirectional and self-modifiable form of cell-cell communication. Cell 72/Neuron 10 (Suppl.), Jia, X.-X., Gorczyca, M., and Budnik, V. (1993). Ultrastructure of neuromuscular junctions in Drosophila: comparison of wild type and mutants with increased excitability. J. Neurobiol. 24, Johansen, J., Halpern, M. E., Johansen, K. M., and Keshishian, H. (1989a). Stereotypic morphology of glutamatergic synapses on identified muscles cells of Drosophila larvae. J. Neurosci. 9, Johansen, J., Halpern, M. E., Johansen, K. M., and Keshishian, H. (1989b). Axonal guidance and the development of muscle fiberspecific innervation in Drosophila embryos. J. Neurosci. 9, Jorgensen, E. M., and Nonet, M. L. (1995). Neuromuscular junctions in the nematode C. elegans. Sem. Dev. Biol. 6, Kandel, E., and Abel, T. (t995). Neuropeptides, adenylyl cyclase and memory storage. Science 268, Kania, A., Han, P.-L., Kim, Y.-T., and BeUen, H. (1993). neuromusculin a Drosophila gene expressed in peripheral neuronal precursors and muscles, encodes a cell adhesion molecule. Neuron 11, Keshishian, H., Chiba, A., Chang, T. N., Halfon, M. S., Harkins, E. W., Jarecki, J., Wang, L., Anderson, M., Cash, S., Halpern, M. E., and Johansen, J. (1993). Cellular mechanisms governing synaptic development in Drosophila melanogaster. J. Neurobiol. 24, Kidikoro, Y., and Nishikawa, K. (1994). Miniature endplate currents at the newly formed neuromuscular junction in Drosophila embryos and larvae. Neurosci, Res. 19, Kl~.mbt, C., and Goodman, C. S. (1991). The diversity and pattern of gila during axon pathway formation in the Drosophila embryo. Gila 4, Kolodkin, A. L., Matthes, D. J., and Goodman, C. S. (1993). The semaphorin genes encode a family of transmembrane and secreted growth cone guidance molecules. Cell 75, Kurdyak, P., Atwood, H. L., Stewart, B. A., and Wu, C.-F. (1994). Differential physiology and morphology of motor axons to ventral longitudinal muscles in larval Drosophila. J. Comp. Neurol. 350, Lance-Jones, C., and Landmesser, L. (1981). Pathway selection by

12 Neuron 524 chick lumbrosacral motorneurons during normal development. Proc. R. Soc. Lond. (B) 214, Lin, D. M., and Goodman, C. S. (1994). Ectopic and increased expression of Fasciclin II alters motoneuron growth cone guidance. Neuron 13, Littleton, J. T., Stern, M., Schulze, K., Perin, M, and Bellen, H. J. (1993a). Mutational analysis of Drosophila synaptotagmin demonstrates its essential role in Ca~+-activated neurotransmitter release. Cell 74, Littleton, J. T., Bellen, H. J., and Perin, M. S. (1993b). Expression of synaptotagmin in Drosophila reveals transport and localization of synaptic vesicles in the synapse. Development 118, Littleton, J. T., Stern, M., Perin, M. S., and Bellen, H. J. (1994). Calcium dependence of neurotransmitter release and rate of spontaneous vesicle fusions are altered in Drosophila synaptotagmin mutants. Proc. Natl. Acad. Sci. USA 91, Livingstone, M. S., Sziober, P. P., and Quinn, W. G. (1984). Loss of calcium/calmodulin responsiveness in adenylate cyclase of rutabaga, a Drosophila learning mutant. Cell 37, Matthes, D. J., Sink, H., Kolodkin, A. L., and Goodman, C. S. (1995). Semaphorin II can function as a selective inhibitor of specific synaptic arborizations. Cell 81, McMahan, U. J. (1990). The agrin hypothesis. Cold Spring Harbor Symp. Quant. Biol. 50, Medie, J. P., and Sanes, J. R. (1985). Concentration of acetylcholine receptor mrna in synaptic regions of adult muscle fibres. Nature 317, Mochida, S., Poulain, B., Eisel, U., Binz, T., Kurazano, H., Niemann, H., and Tauc, L. (1990). Exogeneous mrna encoding tetanus or botulinum neurotoxins expressed in Aplysia neurons. Proc. Natl. Acad. Sci. USA 87, New, H. V., and Mudge, A. W. (1986). Calcitonin gene-related peptide regulates muscle acetylcholine receptor synthesis. Nature 323, Niemann, H., Blasi, J., and Jahn, R. (1994). Clostridial neurotoxins: new toots for dissecting exocytosis. Trends Cell Biol. 4, Nose, A., Mahajan, V. B., and Goodman, C. S. (1992). Connectin: a homophilic cell adhesion molecule on a subset of muscles and motorneurons that innervate them in Drosophila. Cell 70, Nose, A., Takeichi, M., and Goodman, C. S. (1994). Ectopic expression of connectin reveals a repulsive function during growth cone guidance and synapse formation. Neuron 13, Novick, P. J., and Schekman, R. (1979). Secretion and cell surface growth are blocked in a temperature sensitive mutant of Saccharomyces cerevisiae. Proc. Natl. Acad. Sci. USA 76, Ordway, R. W., Pallanck, L, and Ganetzky, B. (1994). Neurally expressed Drosophila genes encoding homologs of the NSF and SNAP secretory proteins. Proc. Natl. Acad. Sci. USA 91, Pallanck, L., Ordway, R. W., and Ganetzky, B. (1995). A Drosophila NSF mutant. Nature 376, 25. Patel, N. H., Snow, P. M., and Goodman, C. S. (1987). Characterization and cloning of Fasciclin Ill, a glycoprotein expressed in a subset of neurons and axon pathways in Drosophila. Cell 48, Perin, M. S., Johnston, P. A., Ozcelik, T., Jahn, R., Franke, U., and SQdhof, T. C. (1991 ). Structural and functional conservation of synaptotagmin (p65) in Drosophila and humans. J. Biol. Chem. 266, Porter, B. E., Weis, J., and Sanes, J. R. (1995). A motoneuron-selective stop signal in the synaptic protein S-laminin. Neuron 14, Ramaswami, M., Krishnan, K. S., and Kelly, R. B. (1994). Intermediates in synaptic vesicle recycling revealed by optical imaging of Drosophila neuromuscular junctions. Neuron 13, Redfern, P. A. (1970). Neuromuscular transmission in newborn rats. J. Physiol. 209, Rosahl, T. W., Geppert, M., Spillane, D., Herz, J., Hammer, R. E., Malenka, R. C., and SL~dhof, T. C. (1993). Short-term synaptic plasticity is altered in mice lacking Synapsin I. Cell 75, Rosahl, T. W., Spillane, D., Missler, M., Herz, J., Selig, D. K., Wolff, J. R., Hammer, R. E., Malenka, R. C., and SL~dhof, T. C. (1995). Essen- tial functions of synapsins I and II in synaptic vesicle regulation. Nature 375, Rutishauser, U., Acheson, A., Hall, A. K., Mann, D. M., and Sunshine, J. (1988). The neural cell adhesion molecule (N-CAM) as a regulator of cell-cell interactions. Science 240, Salzberg, A., Cohen, N,, Halachmi, N., Kimchie, Z., and Lev, Z. (1993). The Drosophila Ras 2 and Rop gene pair: a dual homology with a yeast Ras-like gen e and a suppressor of its loss of function phenotype. Development 117, Sanes, J. R. (1995). The synaptic cleft of the neuromuscular junction. Sem. Dev. Biol. 6, Schulze, K. L., Littleton, J. T., Salzberg, A., Halachmi, N., Stern, M., Lev, Z., and Bellen, H. J. (1994). rop, a Drosophila homolog of yeast secl and vertebrate n-sec1/munc-18 proteins is a negative regulator of neurotransmitter release in vivo. Neuron 13, Schulze, K. L., Broadie, K., Perin, M. S., and Bellen, H. J. (1995). Genetic and electrophysiological studies of Drosophila syntaxin-la demonstrate its role in non-neuronal secretion and its essential role in neurotransmitter release. Cell 80, , Schuster, C. M., Ultsch, A., Schloss, P., Cox, J. A., Schmitt, B., and Betz, H. (1991). Molecular cloning of an invertebrate glutamate receptor subunit expressed in Drosophila muscle. Science 254, Shone, C. C., Quinn, C. P., Wait, R., Hallis, B., Fooks, S. G., and Hambleton, P. (1993). Proteolytic cleavage of synthetic fragments of vesicle associated protein isoform-2 by botulinum type B neurotoxin. Eur. J. Biochem. 217, Siddiqi, O., and Benzer, S. (1976). Neurophysiologioal defects in temperature sensitive paralytic mutants of Drosophila melanogaster. Proc. Natl. Acad. Sci. USA 73, Sink, H., and Whitington, P. M. (1991a). Location and connectivity of abdominal motoneurons in the embryo and larva of Drosophila melanogaster. J. Neurobiol. 22, Sink, H., and Whitington, P. M. (1991b). Pathfinding in the central nervous system and periphery by identified embryonic Drosophila motor axons. Development 112, , Sink, H., and Whitington, P. M. (1991c). Early ablation of target muscles modulates the arborisation pattern of an identified embryonic Drosophila motor axon. Development 113, Snow, P. M., Bieber, A. J., and Goodman, C. S. (1989). Fasciclin II1: a novel homophilic cell adhesion molecule in Drosophila. Cell 59, SSIIner, T., Whitheart, S. W., Brunner, M., Erdjument-Bromage, H., Geromanos, S., Tempst, P., and Rothman, J. E. (1993). SNAP receptors implicated in vesicle targeting and fusion. Nature 362, SQdhof, T. C. (1995). The synaptic vesicle cycle: a cascade of proteinprotein interactions. Nature 375, S(Jdhof, T. C., Baumert, M., Perin, M. S., and Jahn, R. (1989). A synaptic vesicle membrane protein is conserved from mammals to Drosophila. Neuron 2, Sweeney, S. T., Broadie, K., Keane, J., Niemann, H., and O'Kane, C. J. (1995). Targeted expression of tetanus toxin light chain in Drosophila specifically eliminates synaptic transmission and causes behavioral defects. Neuron 14, Tang, J., Landmesser, L., and Rutishauser, U. (1992). Polysialic acid influences specific pathfinding by avian motoneurons. Neuron 8, Tosney, K. W., and Landmesser, L. (1985a). Specificity of early motorneuron growth cone outgrowth in the chick embryo. J. Neurosci. 5, Tosney, K. W., and Landmesser, L. (1985b). Growth cone morphology and trajectory in the lumbrosacral region of the chick embryo. J. Neurosci. 5, Van de Goor, J., Ramaswami, M., and Kelly, R. (1995). Redistribution of synaptic vesicles and the proteins in temperature sensitive shibire t,~ mutant Drosophila. Proc. Natl. Acad. Sci. USA 92, Van Vactor, D., Sink, H., Fambrough, D., Tsoo, R., and Goodman, C. S. (1993). Genes that control neuromuscular specificity in Drosophila. Cell 73,

13 Review: Embryonic Drosophila Neuromuscular System 525 Whitington, P. M. (1985). Functional connections with foreign muscles made by a target-deprived insect motorneuron. Dev. Biol. 107, Yin, J. C. P., Wallach, J. S., Del Vecchio, M., Wilder, E. L., Zhou, H., Quinn, W. G., and Tully, T. (1994). Induction of a dominant negative CREB transgene specifically blocks long-term memory in Drosophila. Cell 79, Yin, J. C. P., Del Vecchio, M., Zhou, H., and Tully, T. (1995). CREB as a memory modulator: induced expression of a dcreb2 activator isoform enhances long-term memory in Drosophila. Cell 81, Zhong, Y. (1995). Mediation of PACAP-like neuropeptide by coactivation of Ras/Raf and camp transduction pathways in Drosophila. Nature 375, Zhong, Y., and Pe 5a, L. A. (1995). A novel synaptic transmission mediated by a PACAP-like neuropeptide in Drosophila. Neuron 14, Zhong, Y., and Wu, C.-F. (1991). Altered synaptic plasticity in Drosophila memory mutants with a defective cyclic AMP cascade. Science 251, Zhong, Y., Budnik, V., and Wu, C.-F. (1992). Synaptic plasticity in Drosophila memory and hyperexcitable mutants: role of camp cascade. J. Neurosci. 12, Zinsmaier, K. E., Hofbauer, A., Heimbeck, G., Pflugfelder, G. O., Buchner, S., and Buchner, E. (1990). A cysteine string protein is expressed in retina and brain of Drosophila. J. Neurogenet. 7, Zinsmaier, K. E., Eberle, K. K., Buchner, E., Walter, N., and Benzer, S. (1994). Paralysis and early death in cysteine string protein mutants of Drosophila. Science 263,