Activity-dependent and age-dependent recruitment and regulation of synapses in identified Crustacean Neurones¹

Substantial advances in understanding mechanisms of synapse formation have been made with tissue culture techniques, which permit the environment of the neurone to be controlled, and with genetic mutations, which provide a means of controlling its phenotypic capabilities. These approaches are providing information about the roles of growth factors and other materials as regulators of synapse formation. However, neural activity is thought to play an important role in regulating synapse formation and the physiological properties of previously established synapses in many systems, including vertebrate visual pathways (Debski et al. 1990) and mammalian autonomic and somatic peripheral nervous systems (Betz, 1987; Betz et al. 1990). The role of activity varies with age in many instances, but it is likely that most neurones respond adaptively to changes in activity throughout their existence.

This hypothesis can be advantageously explored in selected systems in which identified neurones can be monitored throughout their lifetimes. Crustacean motor systems provide an opportunity for this approach. The motor neurones have peripheral synapses which are much more similar to those typically found in the central nervous system than are the peripheral neuromuscular synapses of most vertebrates. In addition, the large size of the individual neurones allows techniques of presynaptic recording and injection to be employed. The long lifetimes of some crustaceans opens the possibility of ‘cradle to grave’ observation of individual identified neurones over a period of many years, and the remarkable powers of regeneration in crustaceans provide an opportunity to study reformation of synapses and physiological respecification by neurones at different stages of development.

At first glance, crustacean (and other arthropod) nervous systems seem to be unpromising places to study neuronal plasticity, since their development appears to be tightly controlled by genetically specified programmes. But in both insects (Murphey and Matsumoto, 1976; Bloom and Atwood, 1980) and crustaceans (Govind et al. 1987), modification of developmental outcome by early experience or manipulation of the environment has been demonstrated. In crustaceans, alteration of neuronal and synaptic phenotype occurs with altered activity of selected neural pathways (Atwood and Lnenicka, 1987). Such observations lend support to the hypothesis that even in relatively ‘hard-wired’ nervous systems, adaptation to altered activity and environmental change is a continuous process. The advantages of size, ease of experimental approach, neuronal lifetime and neuronal identity make it worthwhile to explore mechanisms of neuronal and synaptic adaptations to activity in crustacean motor systems.

We review here the present status of work on crustacean motor systems as it relates to synapse recruitment and regulation through activity-dependent mechanisms and growth.

Motor neurones of crustacean limb muscles were the first uniquely identifiable neurones ever described. They were originally designated as ‘fast’ or ‘slow’ types on the basis of the responses they evoked in their target muscles. However, this type of classification suffers from the awkward circumstance that slow-acting and fast-acting muscle fibres (responsible, respectively, for the characteristic ‘slow’ and ‘fast’ contractions of limb muscles) are both innervated by the so-called ‘slow’ and ‘fast’ motor axons, though to a varying extent (Atwood, 1976). The clean separation of motor units found in mammalian muscles is not seen in crustaceans. Thus, at present, the characteristic patterns of impulse activity found in different motor neurones provide a more reasonable starting point for defining them. We can refer broadly to ‘phasic’ and ‘tonic’ neurones (Kennedy and Takeda, 1965) and to intermediates.

The physiological and structural properties of several phasic and tonic neurones are well studied (Atwood and Wojtowicz, 1986). These neurones differ considerably in phenotype. Phasic neurones, exemplified by the neurones supplying phasic abdominal extensor and flexor muscles in crayfish and lobsters, are typically silent most of the time, and are recruited to fire brief bursts of closely spaced impulses when vigorous activity is required. In this, they resemble the ‘fast fatiguable’ motor units of mammalian muscles (Burke, 1980). The phasic neurones have large, rapidly conducting axons that taper down into numerous thin, filamentous terminals which profusely innervate the target muscle fibres. Synaptic transmission is characterized by production of large excitatory postsynaptic potentials (EPSPs) in the target muscle fibres. The EPSPs show rapid diminution in amplitude (synaptic depression) when neuronal activity is sustained. Correspondingly, muscle action potentials and accompanying contractions soon cease with maintained neural activity.

Axons of tonic neurones are of smaller diameter but typically supply enlarged varicose endings to the target muscle fibres. The innervation patterns range from diffuse to localized, but are always multiterminal. Synaptic transmission typically leads to EPSPs of variable amplitude, which often show considerable facilitation (both short-term and long-term) with repetition. Transmission is extremely resistant to fatigue, as is the case for ‘slow’ motor units of mammalian muscles.

Although the mechanisms responsible for the differences in neuronal morphology and synaptic transmission are not fully worked out (Atwood and Wojtowicz, 1986), it is clear that the functionally distinct phasic and tonic neurones fulfil different roles in normal activity and behaviour. The extent to which their properties are rigidly determined and their synapses ‘stable’ can be investigated by modifying their activity patterns.

Inspection of nerve terminals at the ultrastructural level has shown that each of them possesses substantial numbers of relatively small specialized contact points with regional postsynaptic elaborations of the muscle fibres (Figs 1 and 2). The contacts (synapses) are seen.in transverse sections as electron-dense membrane appositions between pre- and postsynaptic elements. The presynaptic side of the contact region often has a small ‘dense body’ or ‘dense bar’, similar in appearance, and possible function, to the ‘active zones’ of vertebrate neuromuscular junctions (Couteaux and Pecot-Dechavassine, 1970). Synaptic vesicles cluster selectively at this location (Fig. 2). Not all synapses of a nerve terminal possess these structures, and this circumstance has led to the hypothesis that the synapses vary in their physiological potency (Jahromi and Atwood, 1974).

Evidence for a physiological role for the dense bar or ‘active zone’ region in transmitter release has come from freeze-fracture studies (Pearce et al. 1986; Fig. 2). Images of vesicular release appear around the putative ‘active zones’, where vesicular release can also be seen with transmission electron microscopy (Fig. 2). The ‘active zone’ has an associated series of large membrane particles facing the synaptic cleft; in other systems, these are postulated to be calcium channels (Pumplin et al. 1981; Walrond and Reese, 1985).

The dense bar component of the ‘active zone’ seems to be labile, since it is rapidly removed by toxins which admit calcium to the terminal (Fritz et al. 1980) and also apparently by repetitive nerve activity (Wojtowicz et al. 1989), which is known to increase the calcium concentration in the terminal (Delaney et al. 1989). Thus, paradoxically, during tetanic stimulation, when transmitter release from the terminal is increasing, the number of dense bars is apparently decreasing. It is possible that the dense bar indicates a physiologically potent synapse, but may not be essential for transmitter release during tetanic activation. The associated calcium channels may operate even without the dense bar. As yet, the molecular composition of the dense bar is not known, so its precise function remains obscure.

It is not clear that all synapses with dense bars can normally release transmitter. At the very least, the probability of release of quantal units of transmitter is not likely to be uniform among the synapses on a nerve terminal. There are several indications of this. First, direct stimulation of a terminal with a macro-patch electrode (Dudel, 1981) never releases more than 6–8 quanta at a time, even though the number of available synapses with dense bars is typically much higher (Figs 1,2). Second, statistical methods applied to quantal release provide better fits to data sets with non-uniform probability of release for responding units (B. R. Smith, J. M. Wojtowicz and H. L. Atwood, in preparation). At present, the best statistical definitions of release are based upon small numbers (typically 1–6) of Responding units (active synapses?) with non-uniform release probabilities. Quite possibly, many of the numerous morphologically defined synapses on a terminal are not acutely available for transmitter release, but can be recruited at high frequencies of activity, during long-term facilitation (Wojtowicz and Atwood, 1986), or by neuromodulators. This provides a possible substrate for plasticity and adaptation.

The larger decapod species undergo an enormous increase in size during development, maturation and adulthood; also, they live for many years. It is possible to follow structural and physiological changes in single identifiable neurones over a long period. Preliminary studies have been made in the American lobster Homarus americanus (Govind and Pearce, 1981, 1982; De Rosa and Govind, 1978) and freshwater crayfish Procambarus clarkii (Lnenicka and Mellon, 1983a, b).

Following hatching and embryonic development, the American lobster undergoes several rapid moults which transform the initial pelagic larval form into a small juvenile lobster. Detailed structural information is available for motor axon terminals supplying the distal accessory flexor muscle of the walking leg, which is innervated by one excitatory and one inhibitory axon. The excitatory axon innervates only the central region of the larval muscle, and has only a few branches (Fig. 3). As the muscle grows during development, the axon terminals also grow and develop additional branches to keep pace with muscle enlargement. Synapses disappear from their initial locations on the axon terminals, and appear in greater numbers along the developing secondary and tertiary branches. Synapses may be the starting points for growth of new branches. The individual synapses do not differ greatly in structure or size; the major change is in their number and location. These results indicate a progressive turnover of synapses during development and growth; each synapse probably has a lifetime lasting approximately from one moult cycle to the next. As the animal matures and the time between moults Increases, individual synapses may persist for progressively longer times.

Once early development has been accomplished, growth continues, and neuromuscular transmission must be adjusted to match the larger size and reduced input resistance of the muscle fibres. That this actually occurs is shown by the relatively constant size of the EPSPs in fibres from different-sized animals. The most exacting study of this type was performed by Lnenicka and Mellon (1983a), who measured EPSPs produced by one motor neurone in a single identifiable muscle fibre in the superficial flexor muscle of crayfish, and demonstrated size constancy of EPSP with growth.

The general physiological picture in crayfish and American lobsters is that several mechanisms continually adjust synaptic transmission to target size. First, there is an increase in quantal content at individual nerve terminals (De Rosa and Govind, 1978; Lnenicka and Mellon, 1983a). This accounts for much of the increase in transmitter release necessary to maintain EPSP amplitude. Part of the increase in quantal content may arise from addition of active synapses, while part may be attributed to increased size of active zones at individual synapses. The latter change implies a higher probability of transmitter release for at least some of the synapses. A second significant mechanism of adjustment is a change in the duration and amplitude of the quantal current (Lnenicka and Mellon, 1983a; Fig. 4). In larger animals, the quantal currents increase considerably in both amplitude and duration; this provides for a greater postsynaptic quantal effectiveness than would occur if quantal current amplitude and duration remained constant (Fig. 4).

Interactions between the neurone and its target apparently occur in order to regulate the overall effect of transmission. Both pre- and postsynaptic mechanisms are involved, with the former being physiologically more significant. The nature of the regulatory signal(s) is not currently known. However, growth-dependent Increases in quantal content can be halted or reversed when target size is decreased experimentally (Lnenicka and Mellon, 1983b). Thus, synaptic changes are not strictly age-dependent, but are adaptive to alterations in the target muscles.

Crustacean motor axons are well known for their ability to sustain transmission for months after they have been disconnected from the central nervous system (Velez et al. 1981). The longevity of these severed distal processes may be attributed in part to importation of proteins (for a review see Bittner, 1988) or nuclei (Atwood et al. 1989) from adjacent glial cells. In some muscles, the amplitude of the EPSP decreases (Velez et al. 1981); a concurrent loss of presynaptic dense bars is observed at this time (Govind and Chiang, 1979). Remarkably enough, the EPSP persists, even though diminished in amplitude, for long periods after transection of the axon. Synapses can persist in the absence of the normal regulatory influence of the neuronal nucleus.

Quantal currents recorded at terminals of transected axons exhibit greatly prolonged time course and amplitude (I. Parnas, J. Dudel and H. L. Atwood, unpublished observations). To some extent, these changes counteract the decrease in quantal content that occurs after axonal transection. Although the change in quantal currents could be construed as adaptive if the synapses could be used for movement, disconnection from the central nervous system renders the latter impossible, so the functional implications of this synaptic change are not clear. In the crayfish opener muscle, transmission can be restored by ephaptic connection between outgrowing processes of the regenerating axon and the surviving distal processes (see review by Bittner, 1988), but in abdominal muscles this does not occur: the regenerating axons pass over the residual distal processes to innervate the muscle directly.

The signal for the prolongation of quantal currents, and its mechanism, are not known at present.

Neurones are sensitive to altered activity and chemical stimuli. Two general classes of effect can be discerned at the presynaptic level: (a) effects mediated locally in the neuronal membrane and neuroplasm which do not require additional protein synthesis; and (b) effects mediated by the neuronal nucleus which involve protein synthesis or its regulation. The first class of effect includes short-term and long-term facilitation, synaptic depression and modulation by neurotransmitters and neurohormones which act through presynaptic receptors. The second class of effect includes neuronal adaptation to altered activity or to particular patterns of input.

Locally mediated effects - short- and long-term facilitation, depression and neuromodulation - have been extensively studied in crustacean motor neurones, which exhibit these processes to a striking degree (see Fig. 5). Brief bursts of activity produce very large increases in transmission for tonic neurones such as the excitatory neurone of the crayfish opener muscle. The tetanic increase is linked to intraterminal accumulation of sodium and calcium (Atwood et al. 1975; Delaney et al. 1989). Supranormal calcium concentration is probably the main factor in enhanced release of quantal units of transmitter (Delaney et al. 1989). Available evidence from statistical analyses of quantal release suggests that the average probability of transmitter release increases at transmitting synapses (Wernig, 1972; Zucker, 1973; Hatt and Smith, 1976).

Following tetanic stimulation, transmission often remains at higher than normal levels for many hours (late phase of long-term facilitation: Wojtowicz and Atwood, 1985, 1986). The late enhancement may be produced with little or no calcium entry during stimulation (Wojtowicz and Atwood, 1988) but it is uncertain whether calcium level is, or is not, elevated in the terminal following stimulation (Delaney et al. 1989). Recent evidence implicates the adenylate cyclase system as an essential factor for induction of the late phase of long-term facilitation (Dixon and Atwood, 1989).

Statistical analyses of quantal release during the long-lasting residual phase of long-term facilitation have indicated that the number of quantal units available for release appears to increase (Wojtowicz and Atwood, 1986; Wojtowicz et al. 1988). An interpretation of this result is that additional synaptic release sites become available for transmitter release following tetanic stimulation. To some extent, this interpretation is supported by morphological evidence: the number of dense bars increases for synapses sampled with electron microscopy (Wojtowicz et al. 1989). Indeed, repeated bouts of tetanic stimulation can lead to a long-lasting increase in the incidence of synaptic dense bars (Mearow and Govind, 1989). Although the link between dense bars and transmitter release (active synapses) remains tenuous, these morphological changes could be an indicator of an increase in the number of active synapses.

Phasic motor neurones typically show depression during repetitive stimulation (Fig. 5). Short-term facilitation is often superimposed on the depression. Following tetanic stimulation, post-tetanic enhancement of transmission emerges from the depression (Fig. 5; Pahapill et al. 1987). This potentiation is ion-dependent and may lead to an after-effect resembling the late phase of long-term facilitation (Lnenicka and Atwood, 1985b). To date, statistical analyses of quantal events have not been carried out to see whether there is evidence for addition or subtraction of active synapses.

The general problem of synaptic stability and modification in the face of altered activity or experience can be addressed in crustacean motor neurones, where it is feasible to manipulate the electrical activity of selected neurones experimentally or to perturb sensory pathways that impinge upon them.

Studies on development of muscles and their neuromuscular connections in the asymmetric claws of the American lobster have shown that sensory input from claw mechanoreceptors at a critical stage has a strong influence on claw phenotype (Govind et al. 1987). If one claw is more active than the other, or receives more sensory stimulation, it is likely to develop into a tonic ‘crusher’ claw, while the less active or less stimulated claw becomes a phasic ‘cutter’ claw. It is likely that central comparison of sensory information from the two sides of the animal determines the outcome. Neuromuscular synaptic properties as well as muscle fibre properties are involved in estabfishing the appropriate phenotype. During development, therefore, properties of motor axons and synapses in the claws are influenced by neural activity in sensory nerves and within the central nervous system.

In adult crustaceans, shifts in phenotype of phasic motor neurones can be produced by direct stimulation of motor axons (Lnenicka and Atwood, 1985a) or through selective activation (Lnenicka and Atwood, 1988) or inactivation (Pahapill et al. 1985) of sensory pathways. The most easily observed effects (Fig. 6), which appear with as little as 1 day of extra impulse activity in a phasic neurone, are: (a) lowered amplitude of the EPSP during low-frequency stimulation; and (be increased resistance to synaptic depression during maintained stimulation. Both changes are adaptive in the sense that the neuromuscular synapses become more able to function with tonic, rather than phasic, activity patterns. Decreasing the activity of sensory structures through joint immobilization changes neuromuscular properties in the opposite direction (Pahapill et al. 1985).

The signal for these adaptive synaptic changes appears to be depolarization of the motor neurone, particularly its central processes. Experimental evidence supporting this interpretation was obtained by reversibly blocking impulse propagation at the mid-point of the motor axon (Lnenicka and Atwood, 1989). Impulse activity induced in the central part of the neurone led to the usual adaptational changes in the distal synapses, even though they did not experience any direct effects of the conditioning stimulation. Similarly, augmented sensory input to the motor neurone pool led to synaptic adaptation of the neuromuscular synapses, even though few, if any, extra impulses appeared in the motor neurone (Lnenicka and Atwood, 1988). The general conclusion from these lines of experimentation is that the central part of the motor neurone regulates the distal synapses in an adaptive manner, possibly by adjusting synthesis of materials that affect synaptic performance.

Adaptive changes in nerve terminal morphology occur in parallel with the physiological modification. The most obvious change is in the shape of the nerve terminals: they are transformed from a filiform type to a more varicose configuration characteristic of tonic motor neurones (Lnenicka et al. 1986) (Fig. 6B). There is evidence also that the terminals become more compact, implying that they retract from some of their original locations (Lnenicka and Le Page, 1988). Substantial increases in mitochondrial volume are found in the terminals and also in the axons. The latter change could be one basis for enhanced resistance to synaptic depression, since ATP supply would be better sustained during long periods of moderate activity, and prolonged transmission in these terminals requires an adequate energy supply (Lang and Atwood, 1973).

The two major synaptic physiological effects of adaptation, reduced initial EPSP amplitude and increased fatigue resistance, do not appear to be obligatorily linked. In experiments in which low frequencies of stimulation were used over extended periods, lowering of EPSP amplitude occurred without substantial fatigue resistance (A. J. Mercier, H. Bradacs and H. L. Atwood, unpublished observations). Also, interruption of axonal transport after 1 or 2 days of stimulation prevents development of fatigue resistance in the terminals, but not the lowering of EPSP amplitude (Nguyen and Atwood, 1990). Thus, it is likely that adaptation is a multi-faceted process.

Quantal analysis of a few examples has shown that quantal content of representative nerve terminals is lower after adaptation has taken place (Fig. 7). This change accounts for the smaller EPSP (Fig. 6), and confirms the predominantly presynaptic locus of the adaptive changes. There is no sign of a change in size or shape of the unitary quantal currents. Binomial parameters yield reduced values for probability of quantal release. The number of active synapses many remain about the same at locations sampled by focal recordings. However, since the distribution of synapses becomes more non-uniform with adaptation (Lnenicka et al. 1986), it is possible that a smaller percentage of the total synapse population is used at low frequencies, while more facilitation occurs at higher frequencies.

Direct evidence for involvement of protein synthesis in the adaptive changes has been provided by Nguyen and Atwood (1990). A protein synthesis inhibitor, eycloheximide, applied at a concentration that has no acute or direct effects on synaptic transmission but is sufficient to inhibit most of the protein synthesis in the central nervous system for about 2 h, suppresses neuronal adaptation, but only if is applied before the conditioning stimulation (Fig. 8). The implication of this experiment is that short-lived proteins or precursors must be available at the time of stimulation to get adaptive transformation of synaptic transmission. The nature of these materials is not known at present.

The phasic motor axon of the crayfish claw shows little adaptation in older animals (Lnenicka and Atwood, 1985a), but phasic axons of deep abdominal extensor muscles adapt readily in older animals (Mercier and Atwood, 1989) Preliminary observations have indicated that adaptation in regenerated claws of older animals can be more readily induced than in unregenerated claws. Hence, regeneration may involve a partial restoration of plasticity, even in ‘older’ neurones.

Polyneuronal innervation is the rule in crustacean muscles. There is usually at least one excitatory and one inhibitory input. When two or more excitatory axons supply the same target muscle, the strength of synaptic input provided by the different axons to individual muscle fibres follows a definable pattern, often correlated with muscle fibre type (Atwood, 1973,1976). The factors that set up and consolidate the pattern of innervation are not well understood, but several hypotheses have been advanced. Broadly speaking, they fall into the following categories, (a) Muscle fibres are specified genetically and, in turn, dictate the type of synapse that can be formed by the innervating axons (Frank, 1973). (b) The type of synapse that can be formed is time-dependent during early development or muscle regeneration, and muscle fibre type is influenced by the innervation (Atwood, 1973; Govind et al. 1973). (c) A trans-muscle gradient is established, possibly by the growing nerve, and synapse formation is a probabilistic event within the gradient (Velez and Wyman, 1978).

Competitive interactions amongst motor neurones during growth may regulate the final pattern of innervation to the polyneuronally innervated deep abdominal extensor muscles in the American lobster (Stephens and Govind, 1981). Input from an excitor and inhibitor neurone goes initially to this muscle in its own segment as well as in the immediately adjacent ones, but subsequently becomes restricted to its own segment.

Once the pattern of synapse distribution in crustacean muscles is well specified, it can nevertheless be modified when the innervation or the muscle is altered. In the tonic abdominal flexor muscles of the crayfish abdomen, elimination of half the muscle causes the axons that preferentially supply it to hyperinnervate the remaining muscle fibres and to weaken or suppress the pre-existing innervation from other axons (Worden et al. 1988). If the nerve bundle is simply transected part way across the muscle, leaving all the target muscle fibres in place, the axons grow back to their normal locations and competitive suppression of synergistic inputs does not occur (Rhee and Velez, 1989). This experiment demonstrates a suppressive effect of ‘new’ innervation on pre-existing innervation when target size is altered.

Experiments on the deep abdominal extensor muscles of rock lobsters also point to competitive interaction among synergistic inputs. Selectively killing one of the excitatory axons to a muscle bundle by injecting it with pronase causes a synergistic axon innervating the same bundle to increase its quantal content (see review by Parnas, 1987). The effect is specific for those terminals innervating the region vacated by the killed axon: terminals of the synergistic axon innervating nearby muscle fibres not originally supplied by the killed axon do not increase their quantal content.

If, in the same preparation, a segmental nerve is cut, the distal processes survive for some time, though they remain electrically silent. Under these conditions, overlapping synergistic inputs do not markedly increase their quantal content. Thus, there appears to be a local interaction between synergistic inputs to the same muscle, with resultant mutual suppression of quantal release.

In the above experiments, little attention was paid to the possible role of impulse activity as a factor influencing the interactions among synergistic inputs. Recent experiments on the deep extensor muscles of the American lobster suggest that altered impulse activity in one neurone can influence synaptic transmission in nearby terminals of other neurones (Bradacs et al. 1990). Elevating the impulse production in one of the segmental nerves of the deep extensor muscles by chronic stimulation leads to long-term adaptation, as described above (Figs 6 and 7). The EPSPs associated with the stimulated axons become smaller at low frequencies of stimulation. Surprisingly, the EPSPs of unstimulated axons supplying the same muscle bundles increase in amplitude (Fig. 9).

One possible explanation is suggested by the recent work of Lnenicka and LePage (1988). In the crayfish closer muscle, phasic nerve terminals undergoing long-term adaptation gradually retract while at the same time becoming more varicose. If a similar effect occurs in terminals of the phasic axons supplying the abdominal extensor muscles, retraction of stimulated terminals could remove the local suppression of some of the synergistic terminals. This could lead to an increase in EPSP size for the unstimulated axons.

In crustacean motor systems, it is feasible to conduct long-term studies on the development and fate of synapses formed by identifiable neurones, often on identifiable target cells. Furthermore, the role of controlled changes in neuronal activity can be assessed through measurements of synaptic physiology and morphology, and both pre- and postsynaptic structures can be recorded from or injected with pharmacological agents.

One of the basic features of crustacean motor terminals that makes them more similar to central neurones than to vertebrate neuromuscular junctions is their possession of large numbers of discrete synapses. These synapses may function as ‘all-or-nothing’ units, each capable of generating a single quantal unit of transmitter action for each successful transmission (Atwood and Wojtowicz, 1986). This idea is similar to that proposed for terminal boutons of neurones in the central nervous system (Korn et al. 1982; Redman, 1990). There is as yet no firm proof for this hypothesis. A fundamental question that has not been answered by direct experiment is whether the transmitter substance released by one (or a few) synaptic vesicle is sufficient to saturate the postsynaptic receptors at a single synapse. If so, the quantal unit size would be defined by the number of active receptors and their associated ion channels at a synapse; variations in quantal unit size would arise from variations in synapse size. Although normal changes in quantal size with growth (Fig. 4) could be accommodated in this hypothesis, it is more difficult to account for the extreme examples of altered quantal amplitude and time course seen in decentralized terminals of the rock lobster deep extensor muscles.

The morphological evidence suggests that there may be numbers of inactive synapses at many crustacean nerve terminals (Fig. 1). However, if this is the case, it is possible that some of them can be kicked into an active mode by increased neural activity, as suggested by studies on long-term facilitation (Fig. 6). In extreme cases, synapses may fail to respond at all until they have been ‘warmed up’ by repetitive activity (Acosta-Urquidi, 1978). Evidence for silent synapses has appeared in various studies of the mammalian central nervous system (e.g. Jack et al. 1981).

How stable are the synapses of crustacean nerve terminals? During development and regeneration they are formed rapidly (Govind et al. 1973), but disappear from their original locations as the terminal grows. They appear to enlarge or become perforated with elevated neural activity (Pearce et al. 1985; Wojtowicz et al. 1989). On morphological grounds, we imagine that synapse turnover coincides with periods of terminal growth that are probably related to the moult cycle. The physiological potency of a synapse in relation to its time of formation is not known. However, synapses of decentralized terminals acquire lower quantal content with time, and become more fatiguable. This may partly reflect the lack of generation of new synapses, as well as gradual loss of metabolic competence in the axon. Chronic changes in activity, especially in phasic neurones, have been shown to be capable of altering terminal morphology and physiology. On the morphological level, the terminals appear to become more consolidated and the synapses shift mainly to varicosities (Lnenicka et al. 1986). Enlargement of mitochondria accompanies, and may partly explain, increased resistance to synaptic depression. Down-regulation of initial transmitter release seems to be a separate effect, but its basis is not yet known. Similar alterations in initial transmitter output were described at the frog neuromuscular junction (Hinz and Wernig, 1988).

As in many well-studied vertebrate systems, competition between synaptic inputs to a common target is evident in crustacean muscles. Although the role of activity in determining the outcome of such competitive interactions among converging inputs has been explored only in preliminary investigations during early development in crustaceans, it appears to be significant in adult systems during reorganization or changes in activity (e.g. Fig. 9). Long-term adaptation of terminal morphology and metabolism in chronically active neurones could be one contributing factor in this interaction.

The totality of observations indicates both growth-dependent and activity-dependent adjustment of synapse number and potency, providing for considerable plasticity in relatively parsimonious ‘hard-wired’ motor systems. It seems likely that adjustment to altered activity goes on all the time in these neurones, as in many others.