Molluscan Nervous Systems - Wiley Online Library

Molluscan Nervous Systems

Advanced article Article Contents

Paul R Benjamin, Sussex Centre for Neuroscience, Brighton, UK . Introduction

Gyo¨rgy Kemenes, Sussex Centre for Neuroscience, Brighton, UK

. Anatomy of the Molluscan Nervous System: A Few Pairs of Ganglia

Kevin Staras, Sussex Centre for Neuroscience, Brighton, UK

. Large Identifiable Neurons in the Molluscan Nervous System

Neural circuits underlying reflex and rhythmic behaviours in molluscs can be understood at the level of single identified neurons. Using these systems, the most important advances have been made in the understanding of the cellular and molecular basis of learning and memory.

. Neural Circuits for Behaviour in Aplysia and Lymnaea . Learning and Memory in Aplysia and Lymnaea . The Complex and Highly Differentiated Cephalopod Nervous System . Summary

Introduction

doi: 10.1038/npg.els.0004073

The central nervous systems (CNSs) of one class of molluscs, the gastropod slugs and snails, have been of major importance in neurobiological studies aimed at understanding the cellular and molecular basis of behaviour. Individual neurons can be identified readily and their synaptic connectivity within a neural circuit can be determined by simultaneous electrophysiological recording from two or more neurons. This feature has allowed the neural circuitry underlying a number of simple behaviours to be elucidated. Significant progress has been made in understanding the neural control of reflexes such as those involved in defensive withdrawal (siphonal/gill/tail withdrawal and inking in Aplysia, whole body withdrawal in Lymnaea) and central programmes generating or modulating rhythmic motor behaviours (respiratory pumping in Aplysia; heart control in Aplysia and Lymnaea; locomotion in Tritonia and Clione; feeding in Aplysia, Limax, Helisoma and Lymnaea). Although initially these circuits were conceived as producing relatively stereotyped motor behaviours, it was later realized that several of them were capable of modification as a result of behavioural training, allowing the detailed neural and molecular mechanisms of learning and memory to be investigated. This work has been particularly successful in the neural network that underlies the gill/siphonal defensive withdrawal behaviour of the sea slug, Aplysia californica. A variety of different types of learning, both nonassociative (habituation and sensitization) and associative (classical conditioning), have been demonstrated in this system. It seems remarkable that such a simple reflex shows so many different types of plasticity. This work on learning and memory in Aplysia and the related work on other gastropod molluscan preparations (Hermissenda, Limax, Lymnaea and Pleurobranchaea) has produced results that are of general significance for the whole of neuroscience. See also: Mollusca (molluscs); Neural networks and behaviour Although the behaviour of the lowly slugs and snails is much more sophisticated than was previously thought, it is

still far simpler than that exhibited by the most ‘brainy’ group of molluscs, the cephalopods, exemplified by the squids and octopuses. Rather than the 10 000–20 000 neurons present in the CNSs of gastropod molluscs like Aplysia and Lymnaea, the octopus has many millions of cells, making it comparable with the brains of ‘lower’ vertebrates such as fish. Most of the gastropod molluscs of neurobiological interest are grazing herbivores (molluscan ‘cows’), whereas the octopuses and squids are voracious predatory carnivores with well-developed vision, fast muscles that allow jet propulsion and complex manipulative appendages (arms and suckers). With this large brain, it is not surprising that they have complex sexual and predator avoidance behaviour and that they possess elaborate visual and tactile learning capabilities. Information about the neural structures involved in these two types of memory is available in Octopus vulgaris, but there is little cellular and molecular work on this interesting animal to compare with the gastropod data on learning and memory.

Anatomy of the Molluscan Nervous System: A Few Pairs of Ganglia All the molluscs of neurobiological interest have a ring of ganglia (Figure 1a) that form a CNS or brain around the anterior gut (Figure 1b). They also have a considerable peripheral nervous system (PNS) in or near the skin, involved in sensory processing and local motor control. In the octopus there are substantially more cells in the PNS than in the CNS (Wells, 1978), reflecting the fact that most of the complex tasks that the arms and suckers perform can be organized at the arm nerve cord level. In opisthobranch sea slugs (Aplysia) and pulmonate snails (Lymnaea, Helix), it is possible to identify pairs (left and right) of homologous ganglia (Figure 1a) with similar general functions. These are the buccal, cerebral, pleural and parietal ganglia. The

ENCYCLOPEDIA OF LIFE SCIENCES & 2005, John Wiley & Sons, Ltd. www.els.net

1


Figure 1 Ganglia and neurons of the molluscan CNS. (a) Central ganglia of Aplysia, Lymnaea, Helix and Octopus. Green, buccal ganglia; yellow, cerebral ganglia; pink, pleural ganglia; blue, pedal ganglia; red, abdominal ganglion/viscero-parietal complex; purple, lobes of the Octopus brain that are involved in visual learning; light brown, lobes of the Octopus brain that are involved in tactile learning. (b) The central ganglia in snails are concentrated around the oesophagus (gut) in the anterior part (‘head’) of the animal’s body. (c) Some of the best-known identified neurons in the abdominal ganglion of Aplysia. L7, gill and siphon motor neuron; yellow, sensory neurons; green, R18 interneuron; purple, other neurons of particular interest. (d) Some of the best-known identified neurons in the buccal ganglia in the Lymnaea feeding system. The ganglia are shown with the right buccal ganglion twisted to expose the ventral side. B4, B7, B10, motor neurons; black, N1–3 CPG interneurons (d, dorsal; v, ventral; M, medial; L, lateral; t, tonic; p, phasic); green, SO modulatory interneuron. Insert shows an N2v type CPG interneuron filled with carboxyfluorescein.

parietal and visceral (abdominal) ganglia tend to be fused more or less completely into a single ganglion where the homologies are less clear. The buccal ganglia are exclusively concerned with eating and innervate the large muscular feeding organ called the buccal mass (Figure 1b). The cerebral ganglia are always the largest and most lobed ganglia in the CNS and have several motor and sensory functions. They are involved in the sensory and motor control of feeding, head waving and penis eversion and the sensory processing of information from the eyes, tentacles and lips. The centres involved in sensory processing are not obvious in gross anatomy except in terrestrial slugs (Limax), whose great sense of smell and olfactory learning capabilities are reflected in a large olfactory lobe (procerebrum). The pedal ganglia control locomotion and conjoint muscle movements involving the shell in snails. The pleural ganglia have no nerves and their function is generally unknown. The abdominal/parietal ganglia have been extensively investigated and are involved in gill and related respiratory movements, control of (myogenic) heartbeat and other visceral functions (nerves 2

innervate the gut, kidney and reproductive systems). Many more complex actions, such as locomotion and egg laying, require integration of movements of the whole body, involving many different ganglia, but little is known about how this coordination is achieved. The highly complex octopus brain is difficult to homologize with the gastropod CNS, but some ganglia like the paired inferior buccal ganglia that directly innervate buccal mass muscles may be homologous with their equivalents in the gastropods (Figure 1a).

Large Identifiable Neurons in the Molluscan Nervous System The best known large single neuronal structure in (molluscan) neuroscience is the squid giant axon. This axon was discovered in the squid, Loligo, by J. Z. Young in the early 1930s. Its large size (1 mm diameter) and retained excitability when isolated in seawater made it an ideal


preparation for electrophysiology. Its discovery was followed by the first detailed elucidation of the ionic mechanisms underlying the action potential in neural tissue. The giant axon is the largest of 10–12 giant thirdorder nerve fibres that mediate jet-propelled swimming in the squid (reviewed in Hanlon and Messenger, 1996). The technical advantage for electrophysiological (intracellular) microelectrode recording was also the main attraction for Angelique Arvanitaki, who was the first to use the giant neurons of the gastropod molluscs Aplysia and Helix (reviewed in Kandel, 1976). Arvanitaki and Tchou originally identified seven giant cells in the abdominal ganglion of Aplysia. However, later work by Kandel and his colleagues expanded this number to include more than 50 individually identifiable cells and 1000 other cells that cannot be identified as individuals. The latter are the elements of homogenous clusters with similar electrical properties. The total number of identified neurons represents more than two-thirds of the total cells in the abdominal ganglion (Kandel, 1976). Like the squid giant axon, individual giant cells have become famous in their own right; R2 because of its size ( 1 mm cell body diameter), R15 as a model bursting neuron and L10 as a multiaction cholinergic interneuron. One pair of giant neurons called the metacerebral cells (or cerebral giant cells) have been found in cerebral ganglia of all gastropod molluscs that have been examined, providing a remarkable example of evolutionary conservation at the level of a single cell. Most important for behavioural studies is the opportunity to directly record the chemical and electrotonic synaptic connections between molluscan neurons and thus determine the structure of circuits underlying specific behavioural acts. Examples of identified neurons that form behavioural circuits in Aplysia and Lymnaea are shown in Figure 1c and d. See also: Brain evolution and comparative neuroanatomy It is important to note that some of the most important neurons present in these examples of molluscan circuits are not, by any means, giants. However, they are still identifiable, because of their consistent position, firing properties and synaptic connectivity. The cell bodies of molluscan neurons lie conveniently on or just below the surface of the ganglia. Intracellular recording from these is feasible and it is believed that recording of synaptic potentials at the cell bodies can accurately monitor synaptic inputs to the whole cell. Most synapses are believed to be in the central neuropil, probably on the processes or neurites that originate from the axon in the deep part of the ganglion (the neuropil). These are particularly dense close to the cell body in most molluscan neurons (see dye-marked cell in Figure 1d). The longer, larger-diameter process(es) that projects to other ganglia or nerves is usually called the axon, but this can also branch so the distinction between axon and dendrite, either anatomically or functionally, is not very clear. In a few cases, the specific location of the synapse between two

identified cells has been localized in the neuropil by viewing dye-marked cells in the transmission electron microscope (Kandel, 1976). See also: Neurons; Synapses

Neural Circuits for Behaviour in Aplysia and Lymnaea Defensive withdrawal reflexes in Aplysia: gill withdrawal and inking Both gill withdrawal and inking are part of a local defensive response of the Aplysia body, elicited by a touch stimulus applied to the siphon or mantle skin. The two behaviours are similar in that they are both reflexes and can be elicited from the same region of the body surface. However, there are major differences in the behaviours that allow interesting questions to be asked about the neural circuitry involved and the biophysical properties of the constituent neurons. Defensive gill withdrawal is an example of a low-threshold graded behaviour, where the gill responds selectively to a short-duration weak tactile (pressure) or electrical stimulus applied to the siphon or mantle. The gill is a fleshy fan-shaped respiratory organ that is withdrawn for protection inside the mantle cavity of the animal’s back when a potentially noxious stimulus is applied to the siphon. The size of the contraction depends on the strength of the stimulus, indicating that the reflex is graded. For inking to occur, the stimulus must be stronger and longer in duration, lasting for at least 2 s. It is an all-ornothing reflex that elicits the release of purple ink from a special gland near the edge of the mantle cavity. When the strength of a noxious electrical stimulus applied to the siphon is gradually increased, there is a progressive increase in the size of gill withdrawal, with inking suddenly triggered as an additional response at high levels of stimulus intensity. Once a threshold stimulus is exceeded, the ink is released as a large purple cloud. The neurons responsible for gill withdrawal and inking are in the abdominal ganglion (Figure 1c). It is possible to make an isolated reflex preparation consisting of gill, siphon and part of the ink gland with abdominal ganglion and peripheral nerves intact that allows the circuitry underlying the two behaviours to be compared (Byrne, 1982). This preparation showed that the basic organization of the reflexes is similar. Both circuits consist of sensory neurons, motor neurons and interneurons, some of which are shared (Figure 2a). The cell bodies of more than 50 mechanosensory neurons that have overlapping touch sensory fields in the siphon and mantle skin areas are located in three clusters on the dorsal (RF, Figure 1c) and ventral surfaces (RE, LE). They mediate the sensory component of both reflexes. The same sensory neuron can have monosynaptic excitatory effects on the separate types of motor neurons that cause gill withdrawal (L7) or inking 3


Figure 2 Molluscan neural circuits underlying reflexes and rhythmic motor behaviour. (a) Simplified diagram of the overlapping neural networks involved in the defensive inking and gill-withdrawal reflexes of Aplysia. Yellow box, sensory neurons (SNs); green circles, modulatory interneurons; boxes with red or blue outline, motor neurons (MNs) for inking and gill-withdrawal, respectively. Solid lines, monosynaptic connections; dashed line, polysynaptic connection. Vertical bars, chemical excitatory synapses; horizontal bar, peripheral sensory endings. (b) The more complex neuronal network underlying rhythmic feeding behaviour in Lymnaea. The degree of simplification is the same as for the Aplysia gill withdrawal/inking circuitry. Yellow box, sensory neurons; green circles, modulatory interneurons; circles and boxes with red, blue or green outline, CPG interneurons and motor neurons active in the three different phases of feeding (protraction, rasp, swallow). Filled circles, chemical inhibitory synapses; vertical bars, chemical excitatory synapses; horizontal bar, peripheral sensory endings; zigzag lines, electrotonic connections.

(L14). This is then a direct excitatory pathway between the sensory periphery and motor response. There is also an indirect excitatory pathway to the motor neurons via the abdominal interneuron known as R18. R18 again excites motor neurons for both gill withdrawal and inking, unlike a second type of excitatory interneuron, L31, that only excites the ink motor neuron L14. A simplified summary of the two circuits involved in defensive withdrawal is summarized in Figure 2a. See also: Modulatory and command interneurons for behaviour; Neuronal firing pattern modulation What then are the differences in the operation of the two circuits that can explain the graded nature of gill withdrawal versus all-or-nothing inking? The gill-withdrawal 4

reflex is easier to understand. When the siphon is stimulated, a complex excitatory postsynaptic potential (EPSP) is produced in the gill motor neurons, like L7, causing them to fire and initiate contractions in the gill muscles. This compound EPSP is made up of direct contributions from the sensory neurons and further inputs from R18. This combination of activity drives L7 and the gill withdraws. There is a quantitative relationship between the strength of the stimulus and the final motor response that can be explained by relations between the different individual components of the circuit. The mechanoreceptor firing rate is proportional to the strength of the stimulus. A fairly linear relationship exists between this sensory neuron firing rate and the size of the motor neuron EPSP and motor neuron firing rate, and finally the muscular contraction is linearly related to the motor neuron discharge (Kandel, 1976). These quantitative relationships provide the neural mechanisms to account for the graded nature of the reflex. In inking, the same nominally excitatory input arrives on motor neurons like L14, but no immediate firing occurs even though the cell is depolarized. Only later in the stimulus, after a delay of about 2 s, does the cell fire and then only in an all-ornothing manner with no gradation of firing rate. The absence of the early firing in L14 is due to the presence of a special kind of K+ channel that shunts out the effects of the early depolarization and prevents L14 firing. This channel is closed at normal membrane potential but is activated by the initial sensory EPSP input. After about 2 s, the K+ channel becomes inactivated again, allowing L14 to finally respond by firing. This latter response is due to a late EPSP input resulting from activity in L31. L31 responds to sensory stimulation by producing a slow, long-lasting EPSP on L14, unlike R18, whose effects have largely disappeared by the time the K+ current has been inactivated. See also: Sensory processing in invertebrate motor systems The difference in the behaviour of the inking and gillwithdrawal circuits is due to (1) the presence of a particular type of ion channel in the L14 motor neuron that is not present in the different type of (L7) motor neuron in the gill-withdrawal circuit, and (2) the interneuron L31, which is again unique to the inking circuit. This difference illustrates an important principle of circuit design that generation of the appropriate behaviour depends both on synaptic connectivity and on the ion channel complement present in neurons. See also: Neural networks and behaviour

Central programmes for movement: the feeding system of Lymnaea The neural circuits underlying rhythmic movements like that of Lymnaea feeding involve a type of circuit design (Benjamin and Elliott, 1989) that differs from defensive


reflexes like gill withdrawal and inking. Sequences of movements are required that are repeated over and over again. These sequences rely on central neural networks called central pattern generators (CPGs). The isolated nervous system of Lymnaea has a CPG that can generate a basic feeding motor pattern, but regular repeated sequences of feeding movements seen in normal feeding behaviour require the presence of food on the lips. When observed over several minutes, feeding in Lymnaea is episodic with bouts of food-stimulated rhythmic feeding movements alternating with periods of quiescence. See also: Central pattern generators The neural network generating food ingestion movements is more complex (Figure 2b) than that for the gillwithdrawal reflex (Figure 2a). It consists of sensory neurons (chemosensory), CPG interneurons (3 types), modulatory interneurons (3 types) and motor neurons (10 types). These neurons are located mainly in the paired buccal ganglia (Figure 1d). Pond snails like Lymnaea are grazing herbivores that crawl across the surfaces of rocks, carrying out long sequences of rhythmic scraping or biting movements using a toothed tongue or radula that removes green algal material from the substrate. Alternatively, they float upside-down on the surface of the water, scraping food from floating pond weed, with the same repeated eating movements. Each feeding cycle in Lymnaea lasts 3–5 s and consists of a forward (protraction) and backward (retraction) movement of the buccal mass. Retraction has two phases: rasp when the food is scooped into the mouth, and swallow when the food is pushed into the anterior gut (oesophagus) by a backward rotational movement of an internal structure of the buccal mass called the odontophore. The buccal mass is highly complex (46 muscles), but analysis of its organization is simplified because the muscles and their innervating motor neurons are active during only one of three phases of the feeding cycle (protraction, rasp or swallow). This is also true of the CPG interneurons (the N1, N2 and N3 cells) that fire in sequence (N1, protraction; N2, rasp; N3, swallow) to control their respective types of motor neurons (Figure 2b). The synaptic connections of the N cells with the motor neurons are chemical and can be excitatory or inhibitory depending on type (Figure 2b), with recovery from inhibition (postinhibitory rebound) leading to spiking in some cells. Further sets of complex reciprocal inhibitory and excitatory synaptic connections exist between the N cells of the CPG and the N cells and the modulatory neurons called the slow oscillator (SO), cerebral ventral 1 (CV1) cells and the cerebral giant cells (CGCs), are also shown in Figure 2b. See also: Oscillatory neural networks How does this complex circuit generate rhythmic feeding movements? Chemical stimuli that drive rhythmic feeding behaviour in the intact snail excite all the cells in the network via chemosensory neurons in the lips. The N cells of the CPG produce the basic three-phase motor

programme. As in the gill-withdrawal network, no single factor is involved, but behaviours rely on combinations that include both synaptic connections and the electrical (intrinsic) properties of the neurons themselves. Work on isolated CPG neurons in cell culture (Straub et al., 2002) showed that only the N1 cells (plateauing) and the N3 cells (postinhibitory rebound) have important intrinsic properties within the CPG indicating that rhythmicity mainly depends on synaptic connectivity. Thus the Lymnaea feeding CPG can be considered as an example of a network oscillator. During bouts of food-activated feeding, the N1 (protraction phase neurons) fire first in each cycle because they are the cells with the lowest threshold for chemosensory excitation. Their plateaus are triggered by sugar applied to the lips. By a recurrent inhibitory synaptic loop between the N1 and N2 cells (Figure 2b), activity in the N1 cells in the protraction phase is followed by a plateauactivated burst in the N2, rasp phase of the cycle. The N3 (swallow phase) interneurons fire after the N2 s in a third consecutive phase of activity to complete the cycle. In fact, the N3 cells receive two consecutive phases of inhibitory input from the N1 and N2 cells (Figure 2b) that stop them firing during protraction and rasp, but then they recover and fire by postinhibitory rebound. These are the main chemically mediated synaptic interactions leading to patterned activity (Brierley et al., 1997) in the feeding network, but important results (Staras et al., 1998) have shown that a restricted population of motor neurons also play a role in motor pattern generation via electrotonic (electrical) synaptic connections they have with N cells firing in the same phase of the feeding cycle (Figure 2b). This suggests that pattern generation is a property of the whole neural network and is not restricted to the so-called CPG interneurons. In the absence of food, feeding behaviour is quiescent and actively suppressed. This is due to the inhibition of the feeding network by a subtype of the N3 CPG neurons, the N3 tonic (N3t). During quiescence, the N3t fires continuously to inhibit the N1 CPG cell and prevent rhythmic activity. During feeding the N3t switches into rhythmic mode and becomes part of the CPG (Staras et al., 2003). The cell thus alters function depending on the current behavioural requirements of the animal. By utilizing the same intrinsic member of the CPG network in both rhythm generation and suppression, the feeding system has developed a simple and efficient mechanism for generating episodic rhythmic behaviour. What is the role of the modulatory neurons not so far discussed? Modulation implies that they confer some type of flexibility on the circuit. The SO appears to be involved in controlling the frequency of the feeding oscillator via its excitatory synaptic connection with the N1 cells. The CV1 s control another aspect of the feeding pattern, the duration of the N2/rasp phase, by an independent pathway (Figure 2b). The CGCs and their homologues in other gastropods have long been known to modulate the feeding network of molluscs. In Lymnaea, their large size and convenient position has 5


enabled us to record their electrical activity in freely moving animals by supergluing fine wire electrodes on to their cell bodies. A precise level of firing, about 15 spikes min21, is required to allow food-activated feeding to occur. In other experiments using reduced preparations, it was shown that this activity could not simply activate feeding but promoted or facilitated it through widespread synaptic connections with almost every cell in the feeding network. These fascinating cells have a role in common with other cells in motor systems that is usually referred to as ‘gating’. See also: Chemosensory systems One important conclusion from this work is that the ensemble properties of the whole network are required to produce the behaviour. This, combined with redundancy of some elements, argues for a parallel/distributed type of organization, with neurons having overlapping and complementary functions.

Learning and Memory in Aplysia and Lymnaea Nonassociative learning The gill-withdrawal reflex of Aplysia shows both habituation and sensitization. When touch (usually by a calibrated water jet) is applied repeatedly to the siphon or adjacent mantle skin at intervals of between 30 s and 3 min, the gillwithdrawal response habituates (decrements) to about 30% of control values after 10–15 trials. If, prior to touch, a single strong noxious stimulus, such as an electric shock, is applied to the tail or neck, the subsequent touch-evoked response is enhanced or sensitized. Both of these nonassociative phenomena are central processes that involve changes in synaptic strength at the sensory to motor synapses that mediate the normal gill-withdrawal reflex (Figure 2a). Behavioural habituation was found to be paralleled by suppression of neurotransmitter release from the presynaptic terminals of the touch-sensitive mechanosensory neurons. A consequent reduction in the size of the motor neuron EPSP was followed by a reduction in the activation of the motor neurons, which consequently fired less. This homosynaptic depression is due to a progressive reduction in an inward Ca2+ current in the sensory neuron terminals that is required for transmitter release from secretory vesicles. Sensitizing stimuli have the opposite effect, causing an increase in transmitter release. This effect is due to the sensitizing stimulus activating three different classes of facilitatory neurons, one of which uses the transmitter serotonin. These facilitatory neurons terminate on the sensory neuron terminals, presynaptically, and facilitate synaptic transmission at the sensorimotor junction, causing the motor neurons to fire more after sensitization. This type of heterosynaptic facilitation causes an increase in transmitter release in two ways: (1) 6

by closing K+ channels, leading to a broadening of the sensory neuron action potential, thereby enhancing Ca2+ influx, and (2) by a separate mechanism that involves synaptic vesicle mobilization. Mobilization appears to be due to the serotonin-induced phosphorylation of synapsin, a phosphoprotein present in the synaptic vesicles of the sensory neuron terminals (Angers et al., 2002). See also: Heterosynaptic modulation

Associative learning Molluscs have been shown to exhibit both of the main types of associative learning, operant (instrumental) (Brembs, 2003) and classical conditioning (Benjamin et al., 2000). So far, only classical conditioning has been studied at both the cellular and molecular levels in molluscs and will be considered here. Classical conditioning involves the presentation of a neutral conditioning stimulus (CS) to the animal, followed by the presentation of an unconditioned stimulus (US) within a short interval. After a single or several pairings, the CS produces the conditioned response (CR), which is usually the same as the response evoked by the US. The aversive type of classical conditioning has been studied using the gill- and siphon-withdrawal reflex of Aplysia, phototactic behaviour of Hermissenda and feeding behaviour of Limax, Pleurobranchaea and Lymnaea. Reward or appetitive classical conditioning has been investigated in the feeding system of Aplysia, Limax, Pleurobranchaea and Lymnaea. We will consider the mechanisms of aversive classical conditioning in Aplysia and food-reward classical conditioning in Lymnaea as examples of the two main types of classical conditioning. See also: Learning and memory Carew and colleagues (see Carew and Sahley, 1986) showed that the gill- and siphon-withdrawal reflex is subject to aversive classical conditioning as well as habituation and sensitization. In their associative conditioning paradigm, they paired a weak tactile stimulus (the CS) to the siphon with a strong electric shock to the tail (the US). After 15 trials, the CS came to elicit a stronger gill-withdrawal than controls; the effect lasted for several days. Later they showed that differential classical conditioning also worked. A weak tactile stimulus applied to the siphon or mantle shelf was either paired with shock (CS+) or not paired (CS2) in the same animal. Differences between the effects on gill-withdrawal at the two sites were enhanced, even after a single trial. Cellular analysis of differential conditioning, using a reduced preparation, showed that the mechanism underlying classical conditioning of the gill-withdrawal response was the elaboration of presynaptic facilitation that was previously shown to underlie sensitization, except that the effect of the pairing of CS and US produced an even greater facilitation of the withdrawal response. This effect was called activitydependent amplification of facilitation. Evidence for this


mechanism was obtained in a reduced preparation (Hawkins et al., 1983), where the effects of touch were mimicked by stimulation of two different sensory neurons that made excitatory synaptic connections with siphonal motor neurons (Figure 3a). Differential conditioning produced a significantly greater enhancement of the sensorimotor synapse if weak sensory neuron spike activation (by current injection) was followed by the tail shock (CS+) than if the sensory neuron stimulation was unpaired (CS2) (Figure 3a) or if shock occurred alone (sensitization). This result supported the notion of activitydependent facilitation because of its dependence on temporal pairing. Like sensitization, this mechanism was shown to be presynaptic and could also be mimicked by the application of serotonin. A pairing-specific enhancement of Ca2+ influx into the presynaptic terminal of the sensory neurons was seen that was due to a reduction in the opening of the same serotonin-sensitive K+ channel that was shown to be involved in sensitization. See also: Serotonin Appetitive classical conditioning of the feeding response in Lymnaea has produced interesting neural correlates, although not yet at the level of single identified synapses. The neural network involved in feeding is more complex

than that of the gill-withdrawal network and cellular work initially concentrated on a system approach with the assumption that there will be a number of sites of plasticity in different parts of the network. This multisite hypothesis is current in many other systems where learning and memory are being studied. The most successful appetitive conditioning paradigm in Lymnaea involves the pairing of lip touch (CS) with sucrose reward (US). After 15 pairings spaced over 3 days, the touch CS produces a robust sequence of feeding movements (CR) in the intact snails, significantly greater than controls. This lasts for at least 10 days. The response can be differentially conditioned by applying touch to two different sites on the body (e.g. tentacles and tail) and comparing them as CS+ and CS2 sites, with sucrose as the US (Kemenes and Benjamin, 1989). More complex features of appetitive conditioning using the lip-touch paradigm include stimulus generalization, goal tracking and increased learning in a novel environment (context dependence). Neuronal analysis of appetitive conditioning in Lymnaea has involved two approaches. In the first, training is carried out in vitro in a reduced lip–brain preparation, just

Figure 3 Cellular models of classical conditioning in molluscs. (a) In vitro analogue of differential aversive classical conditioning in Aplysia. Conditioning consists of pairing spike activity in a sensory neuron (SN1) with a shock to the tail (aversive training) in reduced preparations. Spike activity in another SN (SN2) was explicitly unpaired in the same preparations. After training, the spikes evoked in the paired SNs are longer in duration and evoke larger EPSPs in motor neurons compared to spikes triggered in unpaired SNs (cellular correlate). The simplified cellular mechanism (model) is based on activity-dependent amplification of heterosynaptic facilitation of the sensory to motor neuron excitatory chemical synapse (large black triangle) by inputs from a facilitator modulatory neuron activated by tail shock. (b) Appetitive classical conditioning in intact Lymnaea followed by an electrophysiological analysis of the memory trace in a reduced preparation. Touch to the lip was paired with application of sucrose, a food stimulus (appetitive training). After training, touch to the lip evokes more fictive feeding cycles (indicated by dots at the beginning of each cycle) on the B3 motor neurons in preparations made from animals that received paired stimuli compared to those that received unpaired stimuli (cellular correlate). The circuit (model) is based on the observation that cellular traces of learning were recorded at several sites in the brain and indicate that there might be multiple sites of plasticity (large black triangle) in this more complex system. Small black triangles are nonfacilitating synaptic inputs.

7


as in the study of differential conditioning in Aplysia. The preparation is conditioned by pairing touch to the lip as the CS with depolarization of a modulatory neuron (the SO) as the US. Activation of the SO by current injection drives rhythmic activity in the feeding network (‘fictive feeding’). After fewer than 10 trials, the touch CS evokes a train of feeding bursts in the CPG interneurons and motor neurons (CR) (Kemenes et al., 1997). This in vitro preparation offers the opportunity to trace neuronal mechanisms that may be responsible for the acquisition of the learned response, but it is difficult to relate the neural data to learning in the intact animal. This requirement has led to the development of a second approach in which whole animals are trained and then examined to see whether neural traces of the memory survive into the reduced preparation. In this way, changes in neural activity can be more directly related to the behavioural plasticity (Figure 3b). This approach has been successful and neural correlates of behavioural conditioning have been obtained (Staras et al., 1999). Conditioned fictive feeding at the level of the feeding pattern and changes in specific components of the network were recorded electrophysiologically. Thus, conditioning-induced changes have been recorded at the level of the CS pathway, in the CPG network and in the motor neurons with several potential sites of synaptic plasticity involved. However, recent work suggests that changes in the cellular properties of individual neurons may also contribute to the memory trace within the feeding circuit. For instance, Jones et al. (2003) showed that CV1, a modulatory neuron in the Lymnaea feeding circuit (Figure 2b), undergoes a persistent change in membrane potential (depolarization) after tactile conditioning. This depolarization led to an enhanced responsiveness to the previously neutral CS (touch) stimulus in the CV1 and consequently in the whole feeding network. The change in the membrane potential in CV1 was shown to be both sufficient and necessary for the formation of the memory trace. This result emphasizes that changes in both synaptic strength and cellular properties such as excitability both play an important role in associative conditioning. Figure 3 compares the mechanisms of learning originally proposed for aversive learning in the Aplysia gill- and siphon-withdrawal system with those for appetitive conditioning in the Lymnaea feeding system. The early results for aversive conditioning using the reduced preparation (Figure 3a) show presynaptic facilitation by a modulatory interneuron that is activated by tail shock. Based on this finding, a single-site hypothesis involving enhancement at the synapse between sensory neurons and motor neurons was proposed to explain the facilitated response. Recent important work in Aplysia however has shown that persistent memory for aversive conditioning is likely to be mediated by a coordinated interaction of both presynaptic and postsynaptic processes (Antonov et al., 2003; Roberts and Glanzman, 2003). In the appetitive conditioning paradigm, the approach has been to relate the 8

electrical changes directly to the behaviour. The conditioned fictive feeding pattern produced by touch after training is a direct electrical correlate of the behavioural conditioned fictive feeding. The systems model for learning shown in Figure 3b is less detailed at the synaptic level than the Aplysia model but may be more realistic in terms of the likely complexity of the learned behaviour and underlying electrical changes.

Molecular mechanisms of memory formation in Aplysia and Lymnaea Detailed analysis of the molecular cascades involved in the regulation of ion channel activity that underlie both shortand long-term memory formation in Aplysia has been extremely successful. Early studies of sensitization both in reduced preparations and in cell culture concentrated on presynaptic mechanisms and have been particularly important in unravelling the mechanisms involved in the transition between short- and long-term memory (reviewed in Milner et al., 1998). It was realized early on that elevation of levels of the second messenger cyclic adenosine monophosphate (cAMP) in the mechanosensory neurons of the gill-withdrawal reflex were necessary for both sensitization and classical conditioning. Application of a single pulse of serotonin, which could mimic the effect of a sensitizing electrical shock, elevated cAMP, and if the sensory neurons received a brief application of serotonin (US) immediately after a train of spikes (CS), the cAMP levels were even more enhanced (4-fold). These experiments showed that both +ve short-term sensitization and classical conditioning utilized cAMP-dependent processes. It is now believed that serotonin acts via G-linked receptors located in the sensory cell terminals to activate adenylyl cyclase (AC) and increase the level of cAMP. The increase in cAMP activates the cAMPdependent protein kinase A (PKA), which then phosphorylates a receptor on the K+ channel, thereby closing it. Spike broadening of the sensory neuron action potential follows from this blockage, and results in an increased Ca2+ influx and potentiation of synaptic transmission, as discussed earlier. The fact that the cAMP levels are elevated more in classical conditioning than in sensitization can explain the activity-dependent effects of synaptic facilitation. Increased elevation of cAMP levels during classical conditioning is due to a dual activation of AC, both by Ca2+/calmodulin (due to CS-triggered spike activity in the presynaptic terminal) and serotonin (released from facilitating interneurons of the US pathway and interacting with G-protein-coupled receptors). The presynaptic mechanisms involved in short-term sensitization are summarized in Figure 4. See also: Calcium signalling and regulation of cell function; G proteincoupled receptors; Sensory transduction mechanisms; Serotonin receptors


Sensory neuron CREB-2

CREB-1

Nucleus CAAT Late

CRE Early CRE Early MAPK Adenylyl cyclase cAMP Facilitator neuron Tail

5-HT

K+ channel

C/EBP Persistent kinase Protein kinase A

Ca2+ channel Changes in morphology

Motor neuron

Figure 4 A comparison of the main presynaptic molecular events underlying short- and long-term memory formation during sensitization in Aplysia. Serotonin (5-hydroxytryptamines, 5-HT), released from facilitator interneurons activated by tail shock, activates AC, which in turn leads to the elevation of cAMP levels in the sensory neuron terminal presynaptic to the motor neuron. A cAMP-activated protein kinase enzyme (protein kinase A, PKA) phosphorylates a K+ channel, which closes, and this leads to action potential broadening and, consequently, to increased Ca2+ influx, resulting in increased transmitter release. Repeated tail shocks or repeated applications of 5-HT cause a greater elevation of cAMP levels and this results in the translocation of PKA and MAPK into the nucleus. Here, through the activation of the transcription factor CREB protein, a cascade of earlyimmediate and late genes are activated and this leads to the production of a persistently active PKA and long-term changes in neuronal morphology underlying long-term memory formation. Adapted and reproduced from Figure 7 of Milner et al. (1998) with permission from Cell Press.

All the different types of plasticity shown by the gillwithdrawal reflex (habituation, sensitization and classical conditioning) show both short- and long-term forms of memory. Short term proceeds into long term if the number of stimuli or pairings of CS and US is increased. In the case of sensitization, a single pulse of serotonin leads to a memory lasting a few minutes, whereas with five exposures the memory lasts for hours or even days. As long-term memory requires ribonucleic acid (RNA) and protein synthesis, it is clear that gene regulation must be involved, unlike the case of short-term memory, which relies on the activation of already existing proteins. Elegant work by Kandel and his colleagues, has elucidated many of the presynaptic mechanisms involved in long-term sensitization. Figure 4 shows that the same synapse that is involved in short-term memory can be modified for long-term memory. However, with long-term sensitization, PKA is additionally translocated into the nucleus of the sensory cell, along with another kinase enzyme, mitogen-activated protein kinase (MAPK), where they activate the transcriptional regulatory

protein, cAMP response element-binding protein 1 (CREB1). CREB1 leads to the further activation of a number of immediate-early genes. One of these encodes for an enzyme (ubiquitin hydrolase) that is part of a mechanism that eventually cleaves off the regulatory subunits of PKA and tags them for proteolytic degradation by the proteasome. This frees the catalytic subunits of PKA and establishes a persistently active PKA, crucial for both the recruitment of MAPK and subsequent CREB phosphorylation. As in short-term memory, this PKA can also phosphorylate proteins that form part of the K+ channel and both facilitate synaptic transmission and increase neuronal excitability. Another type of CREB (CREB2) is inhibited by PKA and MAPK. This is important because CREB2 normally has a repressive effect on the stimulatory activity of CREB1. The formation of long-term memory lasting for more than 10 h relies on a further nuclear mechanism that leads to the production of proteins that are important for the growth of new synaptic connections. Proliferation of synapses is known to occur during longterm sensitization and appears to be crucial for its maintenance. The genes producing these proteins for growth are stimulated by the transcriptional factor CCAAT enhancer-binding protein (C/EBP). This itself is the product of a second type of immediate-early gene, again activated by CREB1. Thus at least two types of immediateearly gene are activated by CREB1, one leading to the occurrence of persistent PKA activity, the other to de novo protein synthesis and sprouting of the sensory neuron terminal. This molecular model for Aplysia learning (summarized in Figure 4) is of general interest because CREB and PKA are known to be involved in memory formation in a number of other important invertebrate and vertebrate systems (Milner et al., 1998). See also: Long-term depression and depotentiation; Long-term potentiation; Memory in fruit flies and nematodes; Protein synthesis and long-term synaptic plasticity Recent work on the Aplysia learning system has raised the possibility that both pre- and postsynaptic mechanisms are involved in memory formation including classical conditioning (Roberts and Glanzman, 2003). It is suggested that purely presynaptic mechanisms can only account for the early formation of an association between CS and US and for a brief memory of this association, but not for more persistent associative memory, which requires a postsynaptic mechanism as well (Roberts and Glanzman, 2003). The important discovery that sensorimotor synapses of Aplysia possess the capacity for N-methylD-aspartate (NMDA) (glutamate)-receptor-dependent long-term potentiation (LTP) led to the hypothesis that classical conditioning might depend, in part, on a Hebbian LTP-type plasticity mechanism that depends on a facilitated activation of the presynaptic sensory neuron and a coincident postsynaptic depolarization of the motor neuron (Roberts and Glanzman, 2003). Support for this hypothesis initially came from studies that used the 9


so-called ‘cellular analogues’ of classical conditioning, which involve reduced preparations or synapses in dissociated cell culture. This work was extended by Antonov et al. (2003), who used a more intact preparation that permits simultaneous electrophysiological and behavioural investigations. The results of Antonov et al. (2003), together with those of the earlier studies provided strong evidence that Hebbian LTP actually mediates classical conditioning in Aplysia. Figure 5 outlines some of the most important interacting mechanisms that are proposed to underlie the late associative effects of classical conditioning in Aplysia (Roberts and Glanzman, 2003). It shows a hybrid pre/ postsynaptic model of associative enhancement of the sensorimotor synapse, based on two coincidence detection mechanisms, one resulting in a presynaptic change

Siphon tap

(activity-dependent presynaptic facilitation) and the other resulting in a postsynaptic change (NMDA-receptordependent LTP). The former leads to increased glutamate release from the presynaptic sensory neuron terminal while the latter leads to the insertion of new a-amino-3-hydroxy5-methyl-4-isoxalone propionic acid (AMPA) receptors into the postsynaptic motor neuron membrane. Importantly, this hybrid model, which is very similar to current models of NMDA-dependent mammalian LTP, also proposes that a rise in postsynaptic calcium levels during the LTP-induced depolarization of the motor neurons activates a (so far unidentified) retrograde signal, which appears to gate or amplify the presynaptic mechanisms, possibly through the persistent activation of PKA. The presynaptic molecular events (recruitment of MAPK, activation of CREB, etc.) set in motion by the action of

CS

Tail shock

US

INT

5HT

Presynaptic activation

G-protein

Postsynaptic depolarization PLC

Mg2+

NMDAR

IP3 Ca2+ +

Glu

Ca2+

Ca2+ store

Ca2+ MAPK, CREB?

AMPAR

Vesicle mobilization

PKA

PKC

K+ channel

Sensory neuron

Retrograde signal

Motor neuron

Figure 5 Interacting pre- and postsynaptic mechanisms contributing to classical conditioning in Aplysia. Paired CS-US stimulation activates postsynaptic NMDA receptors due to the coincidental presynaptic activation and resulting release of glutamate (Glu) and postsynaptic depolarization by US-activated excitatory interneurons (Int). The US, through the activation of the serotonergic (5-HT) facilitatory interneurons, also probably activates phospholipase C (PLC) in the motor neuron via a G protein-coupled 5-HT receptor, which in turn contributes to inositol-1,4,5-triphosphate (IP3) mediated Ca2+ release from internal Ca2+ stores. This, together with the Ca2+ influx through the activated NMDA receptors leads to protein kinase C (PKC) mediated upregulation of AMPA receptor function and stimulation of a retrograde signal. The former increases postsynaptic sensitivity to Glu, while the latter triggers the persistent presynaptic cellular changes that accompany classical conditioning, possibly through the transsynaptic activation of PKA in the sensory neuron. This model, which is based on work in the Glanzman laboratory, only indicates an indirect persistent effect of the 5-HT-ergic facilitatory interneurons on the sensory neurons, but this notion is still controversial (see Antonov et al., 2003). Adapted and reproduced from Figure 3 of Roberts and Glanzman (2003) with permission from Elsevier).

10


persistent PKA after classical conditioning in Aplysia have not been investigated but are assumed to be similar to those of long-term sensitization. More direct experimental evidence on the molecular mechanisms underlying behavioural classical conditioning has been obtained in Lymnaea. For this type of analysis, a chemical appetitive conditioning paradigm was used based on pairing of an amyl acetate CS with the sucrose US (Benjamin et al., 2000). Unlike tactile conditioning, chemical conditioning produces long-term memory after just one trial, allowing the establishment of sharply timed time windows for molecular events triggered by reward conditioning. Using this paradigm, it has been shown that the gaseous transmitter Nitric oxide (NO) plays a crucial role in an early (10 min) to intermediate ( 5 h) posttraining time-window in the consolidation of long-term memory (Kemenes et al., 2002). The highly conserved transcriptional activator CREB1 also has been identified in Lymnaea and shown to be phosphorylated both in response to PKA activation by cAMP in the isolated CNS and after single-trial reward conditioning in intact animals (Ribeiro et al., 2003). Importantly, phosphorylation of CREB1 after conditioning takes place selectively in the cerebral and buccal ganglia where the neurons that undergo learning-induced plastic changes are located (Ribeiro et al., 2003). The same ganglia also show significant rises in PKA catalytic subunit activity shortly after training and when this activity is inhibited, long-term memory is blocked. It seems therefore that many of the molecular mechanisms described using simple aversive learning paradigms in Aplysia and other molluscs are also shared by the more complex reward conditioning studied in Lymnaea.

The Complex and Highly Differentiated Cephalopod Nervous System The anatomy of the brain of octopus has been studied in considerable detail and functions have been localized and related to behaviour by lesion and gross electrical stimulation (reviewed by Wells, 1978). The structure of the brain can be broadly divided into the supraoesophageal and suboesophageal lobes linked by the perioesophageal magnocellar lobes. The suboesophageal lobes are mainly motor in function and control movements of arms, head, eyes and other body organs as well as the complex chromatophore skin colour system. There is a complex hierarchy of motor control, with at least three levels within the CNS and even more cells in the peripheral nervous system. Specific contractions of muscles can occur with suboesophageal electrical stimulation of motor neurons, but more complex movements, requiring integration of several peripheral muscular systems, only occur with stimulation of the higher centres, including those in the

supraoesophageal ganglia. The motor centres of the supraoesophageal lobes are more concerned with the decision-making processes involved in major sequences of behaviour, such as whether to attack or retreat from an object, rather than individual movements. The supraoesophageal lobes shown in Figure 1a are the highest centres of the brain and are mainly involved in perception and learning. Particularly obvious are the paired optic lobes, which are huge compared with the rest of the brain. The lobe on each side contains about 68 million cells and this complexity reflects that the octopus is a highly visual animal. Several of the supraoesophageal structures are involved with visual memory and there is also a separate highly developed lobe system underlying tactile learning, also important in the octopus. These two types of memory system have been a major focus of attention in octopus neurobiology.

Visual and tactile memory in Octopus Learning studies in the octopus were not generally aimed at understanding the mechanisms of learned behaviour that could be rigorously classified into categories such as associative or nonassociative, operant versus classical, etc., although undoubtedly octopus and other cephalopods show these distinct types of learned behaviours (reviewed in Hanlon and Messenger, 1996). Instead, convenient visual or tactile sensory discrimination procedures were developed that allowed the neural mechanisms of learning and memory to be investigated after lesioning of the brain. The same types of training procedures were also used for studying visual and tactile pattern recognition. Octopuses look at you with their large eyes when you approach them, and respond vigorously to any object that appears within their visual field. Their eyes are vertebratelike in their general optical design, although their retina is much simpler and the visual image is not inverted. They approach any small object that is moved up and down in front of them in the aquarium situation and rapidly attack anything that is known, by prior reinforcement, to be food (e.g. a crab). They withdraw from objects that are larger than themselves and their skin pales, indicating ‘fear’. The ability to evoke visually stimulated attack behaviour has been used extensively to study visual shape preference in naive animals and in discrimination experiments using classical conditioning. Usually, two different stimuli are presented either together or in rapid sequence. One is rewarded with food, the other punished with a mild electric shock. Positive reinforcement leads to further and more vigorous attacks, punishment to leaving the object alone. The repeated training gradually leads to a difference score between the two types of stimuli that is a measure of the ability to discriminate between the two shapes. The similarity between the two stimuli is measured quantitatively by the number of trials it takes to reach a certain level 11


of accuracy in the response or from the accuracy achieved after a certain number of training trials. Stimulus generalization tests are often carried out. Here, after discrimination training with reinforcement applied, further types of visual shapes are tested without reward or punishment to see whether they can be recognized. ‘Similarities’ are quantified by measuring the number of attacks or rejections compared with those evoked by the originally reinforced visual stimulus. See also: Eye anatomy Using these methods, it was discovered that octopuses can readily distinguish a variety of shapes. For instance, naive animals prefer vertical rectangles moved up and down to horizontally orientated rectangles moved up and down, and they can be trained to discriminate between them with a high degree of accuracy compared with bars held at a variety of intermediate angles, indicating orientation specificity. They can be trained to distinguish between shapes of more complex structure (form vision), differently sized patterns of the same shape, and the same shapes with different brightnesses and different planes of polarized light, but they appear to be colour blind (reviewed in Hanlon and Messenger, 1996). The optic lobes are essential for basic visual processes and if they are removed the animals are blind. Whether these lobes are involved in visual learning is more controversial. Partially dividing them, by transverse cuts, prevents intraocular transfer of learned responses, so they probably are involved. More certain are data on the vertical lobe and superior frontal lobe (the latter shown in Figure 1a). Removal impairs visual discrimination learning, which takes much more training. Animals with vertical lobes removed after training show little retention of discrimination learning, although retraining is followed by some saving compared with animals that were completely untrained prior to lesioning. These data are obviously complex and it would be incorrect to conclude that there is a single memory store in the brain or a complete separation of memory from other aspects of sensory processing of visual information. Presumably related to the role of the vertical lobes in visual memory is the recent interesting electrophysiological demonstration of changes in the strength of synaptic field potentials in vertical lobe ‘amacrine cells’ following tetanic stimulation of an afferent tract originating in the adjacent medial superior frontal lobe (Hochner et al., 2003). One type of synaptic plasticity revealed by this study appears to have similar features to the glutamate-dependent mechanism, LTP, in the vertebrate brain. Although this data on octopus were obtained in slices of brain tissue and could not be directly related to behaviour, it could be an important cellular mechanism for behavioural learning by analogy with similar work in vertebrates and another mollusc, Aplysia. A similar complexity is present in the tactile sensory system, on which experiments on basic sensory function and tactile discrimination learning have also been carried out. The mobile suckers located in two rows of 20 on each arm of the 12

octopus, each exhibit an extremely sensitive touch and chemical sensitivity on their rims that is probably used in the natural environment to detect the presence of small prey in the cracks and crevices of rocks that are explored with their arms. In the laboratory, rewarded objects will be grasped by the suckers and passed to the mouth, punished ones will be rejected on the basis of either taste or touch. Detailed experiments on the touch sense (Wells, 1978), show that octopuses can distinguish the texture of objects, for instance the different degrees of roughness shown by cylinders, but they cannot be taught to discriminate the direction of roughness, e.g. whether the grooves are in the vertical or horizontal orientation on the cylinder. The latter is probably due to a problem shared by all animals with no joints, lacking proprioceptors that could indicate the position of a limb in space. Neither can an octopus distinguish the shape, size and weight of objects on the basis of touch. They cannot manipulate objects very well with or without visual guidance. See also: Neural information processing; Touch Different supraoesophageal lobes are involved in tactile discrimination, compared with vision, giving rise to the stimulating idea of two memory systems in the octopus brain. As with vision, there is the problem of separating primary sensory processing from memory and again there is unlikely to be a single memory store. The inferior frontal lobe system (perhaps with the superior buccal lobe) is the main supraoesophageal structure involved. The subfrontal lobe forms the majority (95%) of the six and a half million cells of the inferior frontal system (Wells, 1978) and its removal can account for all of the effects of lesions in preventing the discrimination in learning of textural cues. This failure of discrimination appears to be due to the inability of the octopus to reject one object (the one associated with punishment) while continuing to take the other at the normal rate. Whether this is a failure of memory per se, or of some other decision-making process, is unclear. Neither is it established whether octopuses with subfrontal lobe lesions can actually detect objects on the basis of texture, so there may be deficits in sensory processing as well. Again it seems that learning and memory involve several different structures with difficult to define, complex functions. Unfortunately, no electrophysiological work has been done to examine the responses of neurons to different types of tactile stimuli. However, behavioural work using chemical blocking agents has given some clues to the molecular mechanisms involved in tactile memory formation in Octopus vulgaris. Tactile learning, involving the discrimination of smooth from rough balls, was blocked on one side of the animal if cytochalasin D was applied directly to the subfrontal lobe but not to the vertical (visual memory) lobe on the same trained side of the animal (Robertson, 1994). Cytochalasin D substance blocks growth cone extension in a number of systems, suggesting that sprouting of neurons in the subfrontal lobe is necessary for memory formation.


Robertson et al. (1994) also showed that the gaseous neurotransmitter NO was involved in tactile learning. The intramuscular injection of a substance (nitro-L-arginine methyl ester (L-NAME)) that blocks the synthesis of NO prevented the occurrence of this learning. See also: Axon growth; Nitric oxide as a neuronal messenger A final example, still controversial, illustrates the further complexity of octopus behaviour. Fiorito and Scotto (1992) first trained a group of ‘demonstrator’ octopuses to carry out a visual discrimination task. These learned to distinguish between red and white balls, on the basis of contrast, until they showed a 100% success rate, at test, in selecting the red ball. ‘Observer’ octopuses in an adjacent tank were allowed to watch the successful choices, without reward, and after only four observational ‘trials’ they selected the same red target with a frequency greater than chance, suggesting that they had learned by observation. Although this result has a variety of interpretations, it is still a remarkable example of learned behaviour in a mollusc.

Summary Detailed information is available on the neural mechanisms underlying a wide range of behaviours in molluscs. In gastropods, the ability to identify single neurons and determine their synaptic connectivity has led to comprehensive descriptions of the circuitry underlying simple reflex acts and rhythmic motor behaviour. All of these circuits are dynamic and change their properties following standard conditioning procedures. There are no unique locations for learning in gastropod ganglia; changes occur at the same sites that are involved in generating the basic unconditioned behaviour. Cascades of cellular and molecular events have been elucidated that can explain, in principle, how nonassociative (habituation, sensitization) and associative (classical conditioning) learning can occur, including short- and long-term memory. The same level of cellular analysis is not possible in cephalopods because their complex behaviour is generated by millions of neurons. However, the location of ‘centres’ in the brain that are involved in visual and tactile learning has been carefully analysed, offering the possibility of future, more detailed analyses. See also: Repetitive action potential firing

References Angers A, Fioravante D and Chin J et al. (2002) Serotonin stimulates phosphorylation of Aplysia synapsin and alters its subcellular distribution in sensory neurons. Journal of Neuroscience 22: 5412– 5422. Antonov I, Antonova I, Kandel ER and Hawkins RD (2003) Activitydependent presynaptic facilitation and Hebbian LTP are both required and interact during classical conditioning in Aplysia. Neuron 37: 135–147.

Benjamin PR and Elliott CJH (1989) Snail feeding oscillator: the central pattern generator and its control by modulatory interneurons. In: Jacklet J (ed.) Neuronal and Cellular Oscillators New York: Marcel Dekker. Benjamin PR, Staras K and Kemenes G (2000) A systems approach to the cellular analysis of associative learning in the pond snail Lymnaea. Learning and Memory 7: 124–131. Brembs B (2003) Operant conditioning in invertebrates. Current Opinion in Neurobiology 13: 710–717. Brierley MJ, Yeoman MS and Benjamin PR (1997) Glutamatergic N2v cells are central pattern generator interneurons of the Lymnaea feeding system: a new model for rhythm generation. Journal of Neurophysiology 78: 3396–3407. Byrne JH (1982) Cellular and biophysical mechanisms contributing to regulation of reflex excitability of inking behavior in Aplysia. Federation Proceedings 41: 2147–2152. Carew TJ and Sahley CL (1986) Invertebrate learning and memory: from behavior to molecules. Annual Review of Neuroscience 9: 435–487. Fiorito G and Scotto P (1992) Observational learning in Octopus vulgaris. Science 256: 545–547. Hanlon RT and Messenger JB (1996) Cephalopod Behaviour. Cambridge, UK: Cambridge University Press. Hawkins RD, Abrams TW, Carew TJ and Kandel ER (1983) A cellular mechanism of classical conditioning in Aplysia: activity-dependent amplification of presynaptic facilitation. Science 219: 400–405. Hochner B, Brown ER, Langella M, Shomrat T and Fiorito G (2003) A learning and memory area in the octopus brain manifests a vertebratelike long-term potentiation. Journal of Neurophysiology 90: 3547–3554. Jones NG, Kemenes I, Kemenes G and Benjamin PR (2003) A persistent cellular change in a single modulatory neuron contributes to associative long-term memory. Current Biology 13: 1064–1069. Kandel ER (1976) Cellular Basis of Behavior: An Introduction to Behavioral Neurobiology. San Francisco: WH Freeman. Kemenes G and Benjamin PR (1989) Appetitive learning in snails shows characteristics of conditioning in vertebrates. Brain Research 489: 163–166. Kemenes G, Staras K and Benjamin PR (1997) In vitro appetitive classical conditioning of the feeding response in the pond snail Lymnaea stagnalis. Journal of Neurophysiology 78: 2351–2362. Kemenes I, Kemenes G, Andrew RJ, Benjamin PR and O’Shea M (2002) Critical time-window for NO-cGMP dependent long-term memory formation after one-trial appetitive conditioning. Journal of Neuroscience 22: 1414–1425. Milner B, Squire LR and Kandel ER (1998) Cognitive neuroscience and the study of memory. Neuron 20: 445–468. Ribeiro MJ, Serfo˜zo˜ Z and Papp A et al. (2003) Cyclic AMP response element binding (CREB)-like proteins in a molluscan brain: cellular localisation and learning-induced phosphorylation. European Journal of Neuroscience 18: 1223–1234. Roberts AC and Glanzman DL (2003) Learning in Aplysia: looking at synaptic plasticity from both sides. Trends in Neuroscience 26(12): 662–670. Robertson JD (1994) Cytochalasin D blocks touch learning in Octopus vulgaris. Proceedings of the Royal Society of London B 258: 61–66. Robertson JD, Bonaventura J and Kohn AP (1994) Nitric oxide is required for tactile learning in Octopus vulgaris. Proceedings of the Royal Society of London B 256: 269–273. Staras K, Kemenes G and Benjamin PR (1998) Pattern-generating role for motoneurons in a rhythmically active neuronal network. Journal of Neuroscience 18: 3669–3688. Staras K, Kemenes G and Benjamin PR (1999) Cellular traces of behavioral classical conditioning can be recorded at several specific sites in a simple nervous system. Journal of Neuroscience 19: 347–357.

13


Staras K, Kemenes I, Benjamin PR and Kemenes G (2003) Loss of selfinhibition is a cellular mechanism for episodic rhythmic behaviour. Current Biology 13: 116–124. Straub VA, Staras K, Kemenes G and Benjamin PR (2002) Endogenous and network properties of Lymnaea feeding central pattern generator interneurons. Journal of Neurophysiology 88: 1569–1583. Wells MJ (1978) Octopus. Physiology and Behavior of an Advanced Invertebrate. London: Chapman and Hall.

Further Reading Bullock TH and Horridge GA (1965) Structure and Function in the Nervous Systems of Invertebrates. San Francisco: WH Freeman.

14

Byrne JH (1987) Cellular analysis of associative learning. Physiological Reviews 67: 329–439. Glanzman DL (1995) The cellular basis of classical conditioning in Aplysia californica – it’s less simple than you think. Trends in Neuroscience 18: 30–36. Kandel ER (1979) Behavioral Biology of Aplysia. San Francisco: WH Freeman. Krasne FB and Glanzman DL (1995) What we can learn from invertebrate learning. Annual Review of Psychology 46: 585–624. Squire LR and Kandel ER (1999) Memory. From Mind to Molecules. New York: Scientific American Library. Wilbur KM (ed. in chief) and Willows AOD (ed.) (1985, 1986) The Mollusca, vols 8 and 9, Neurobiology and Behavior, parts 1 and 2. Orlando, FL: Academic Press.