Aplysia provides a cellular model of learning and memory

Kandel and co-workers [13] pioneered the use of the sea snail Aplysia californica as a model system for the study neuronal events associated with learning. Aplysia has a relatively simple nervous system, consisting of approximately 10,000 neurons, many of which are sufficiently large to allow them to be manipulated directly within defined neuronal circuits.

Three behaviors, each of which has adaptive value, have been studied in Aplysia: habituation, sensitization and classical conditioning. Habituation describes a learned response in which a decrease is observed in a specific behavioral response after repeated stimulation. In Aplysia, habituation serves as a rudimentary model that shares some aspects of regulation relevant to learning. A snail ordinarily withdraws its gill after gentle tactile stimulation of the siphon. Upon repeated stimulation, the gill-withdrawal reflex diminishes in both magnitude and duration. If the habituation experience consists of one training session of relatively few stimulations, say, less than ten, over a short period of time, less than 1 hr, then the habituation lasts for only a few hours after the training. If, however, four or more individual training sessions are given, the habituation response can last for several weeks. These two forms of habituation have been interpreted as models of short- and long-term memory.

There is a difference in how short-term and long-term memory are measured in Aplysia experiments and in behavior in vertebrates such as the goldfish. In Aplysia, short-term memory refers to a fleeting memory resulting from few training trials. With additional trials, memory appears to be more robust and to last longer. In experiments with goldfish and rats, the same training session is thought to give rise to both a short- and long-term form of memory. Short-term memory is formed during the training session, while long-term memory proceeds by a consolidation process that takes place after the training session. In both fish and Aplysia, it appears that the short-term memory formation does not require ongoing protein synthesis, while the long-term form does.

The relatively simple neuronal circuitry involved in habituation of the gill-withdrawal reflex is shown in Figure 50-3. Stimulation of a sensory neuron innervating the siphon causes stimulation of motor neurons that innervate the gill muscle. As habituation proceeds, the number of postsynaptic potentials produced in the motor neuron decreases. When completely habituated, the motor neuron does not depolarize and there is no gill withdrawal, while depolarization of the sensory neuron in response to siphon stimulation is unaffected. Kandel and co-workers [13] have shown that habituation is the result of a decrease in synaptic efficacy between the sensory neuron and the motor neuron due to altered permeability of the Ca²⁺ required for neurotransmitter release at the presynaptic terminal. The Ca²⁺ enters the presynaptic terminal through voltage-sensitive Ca²⁺ channels, and during habituation, Ca²⁺ channels in the presyn- aptic terminal become inactivated. Although specific mechanisms for this inactivation have been postulated, its nature in the synaptic terminal of the sensory neurons is unknown. Furthermore, since the long-term habituation appears to involve protein synthesis, the known mechanisms for inactivation of Ca²⁺ channels do not suffice to explain it.

Figure 50-3

A: A simplified diagram of the circuitry involved in sensitization, habituation and classical conditioning in Aplysia. Sensory neurons, motor neurons and facilitating interneurons are indicated. Sensory neurons from the mantle shelf (B) and siphon skin (more...)

A second form of learning in Aplysia, known as sensitization, involves a more complex learning paradigm and depends on a more complicated cellular regulatory mechanism. In one type of sensitization experiment, a mild tail shock is given to the animal shortly preceding tactile stimulation of the siphon. The prior tail shock sensitizes the animal so that the normal gill-withdrawal reflex associated with siphon stimulation is increased in magnitude and duration. Like habituation, sensitization can be either short-term or long-term in nature, depending on the duration and number of training sessions involved.

The neuronal pathway involved in the tail-shock-sensitization experiment is schematized in Figure 50-3A. Stimulation of sensory neurons in the tail causes generation of an action potential in the specific interneurons that facilitate sensitization. These facilitating interneurons form a specific synaptic connection with axons of the sensory neuron that innervate the siphon. The synapses of the facilitating interneurons are positioned so that release of neurotransmitter from their axons is targeted to the axons of the sensory neuron, forming an axo-axonic synapse. In tail shock sensitization, the facilitating interneurons release serotonin onto the axonal terminals of the sensory neurons. Specific serotonin receptors on the sensory axons respond to the serotonin and increase axonal cAMP concentrations (Fig. 50-3B). The elevated cAMP then activates cAMP-dependent protein kinase (PKA), which has multiple substrates in the sensory neuron. Initially, PKA phosphorylates K⁺ channels in the sensory axon. The resulting decrease in K⁺ influx prolongs the action potential and increases the duration of Ca²⁺ influx through voltage-sensitive Ca²⁺ channels. The net effect of this phosphorylation event is that more Ca²⁺ flows into the axon, and since Ca²⁺ is required for synaptic vesicle fusion with the membrane, greater neurotransmitter release is observed when the neuron is depolarized following activation of the serotonin receptor. The greater release of neurotransmitter from sensory axons onto the motor neurons results in increased contraction of muscles involved in gill withdrawal. This mechanism, involving only phosphorylation of existing proteins, is sufficient to account for most of the changes seen during short-term sensitization.

Long-term sensitization can last for days to weeks and has been shown to result in morphological changes which suggest that the conversion from short-term to long-term memory involves the translation of synaptic transmission efficiency into morphological changes of the synapse. Long-term sensitization requires RNA and protein synthesis in order to allow the growth of new synaptic contacts between pre- and postsynaptic cells. The basic mechanism for long-term sensitization relies on the same serotonergic activation of PKA; however, during the repeated training sessions required for long-term sensitization, the persistent activation of PKA causes nuclear translocation of the catalytic subunit of the kinase (Fig. 50-3B). In the nucleus, the catalytic subunit phosphorylates transcription factors, which regulate transcription from specific genes. One class of transcription factor binds to enhancer sequences which are responsible for increasing transcription after activation of PKA. These cAMP response element (CRE) enhancer sequences generally conform to the consensus sequence TGACGTCA and have been characterized in mammals as well as in Aplysia (see Chap. 26). The CRE-binding (CREB) protein specifically recognizes the CRE sequences near the promoter of responsive genes and mediates the increase in transcription only when it is phosphorylated by PKA. Microinjection of a synthetic CRE oligonucleotide into the nucleus of the sensory neuron has been shown to abolish long-term sensitization, presumably by diverting CREB protein away from the target genes required for long-term changes in gene expression. Two forms of CREB have been described in Aplysia: ApCREB1, which acts as a positive mediator of cAMP regulation of transcription, and ApCREB2, which shows homology to ApCREB1 but lacks the PKA phosphorylation site. Experiments suggest that ApCREB2 acts as a repressor of long-term sensitization and that phosphorylation by other kinases may be required to relieve this repression [14,15].

Some of the events following activation of ApCREB1 have been defined [13]. One of the genes which is downstream of the initial phosphorylation of CREB has been identified as the Aplysia CCAAT enhancer-binding protein (ApC/EBP). This protein is a homologue of a mammalian transcription factor known to regulate the immediate early gene c-fos. It is likely that the induced ApC/EBP protein is responsible for a secondary induction of a number of genes responsible for the formation of new synaptic connections. A second gene induced directly by ApCREB1 is a ubiquitin hydrolase. This enzyme has been implicated in the proteolytic degradation of the regulatory subunit of PKA, resulting in the prolonged activation of the catalytic subunit seen during long-term sensitization. The ubiquitin hydrolase may also play a role in the degradation of Aplysia cell adhesion molecule (ApCAM). Degradation of the ApCAM has been postulated to allow the activation of endocytosis required for growth of new synapses [13].

The most complex form of learning which can be studied conveniently in Aplysia is classical conditioning. In this paradigm, tail shock results in withdrawal of the gill, and if the siphon is stimulated shortly before (<1 sec) the tail shock, the animal eventually learns to associate siphon stimulation with the tail shock. In contrast, if the mantle is stimulated shortly after the tail response, there is no learned response. The cellular mechanisms underlying classical conditioning in Aplysia are similar to those involved in sensitization, but the crucial difference lies in the strict temporal dependence found in classical conditioning. At the same axo-axonal synapse where serotonin release from the sensory neuron stimulates cAMP production during sensitization, prior depolarization of the conditioned stimulus pathway (siphon stimulation in Fig. 50-3 above) results in depolarization of the presynaptic terminal. This depolarization results in elevated concentrations of Ca²⁺ in the presynaptic terminal as a result of influx through voltage-sensitive Ca²⁺ channels. The increase is transient, and Ca²⁺ eventually returns to resting intracellular concentrations as it is pumped out through plasma membrane Ca²⁺ pumps. If, however, the presynaptic terminal is stimulated with serotonin while the Ca²⁺ concentration is elevated within the presynaptic terminal, the effects of serotonin on generation of cAMP are potentiated by the elevated Ca²⁺. This effect appears to be due to a Ca²⁺-sensitive form of adenylyl cyclase, which shows greater stimulation of cyclase activity by serotonin in the presence of high Ca²⁺ concentration. Thus, if serotonin release triggered by the US closely follows activation of the CS pathway, the potentiation of the cyclase activity results in higher concentrations of cAMP, greater activation of PKA, greater phosphorylation of the same K⁺ channels involved in sensitization and increased release of neurotransmitter from the presynaptic terminal. This then constitutes a synaptic model of conditioning, regulated at the molecular level.

Drosophila are useful as a genetic system for studying learning and memory

Seymour Benzer and his associates pioneered the use of genetics to study learning and memory in Drosophila [16,17]. These researchers have identified a number of mutant strains that appear normal except for the inability to learn or to store memory of a specific training task. Some forms of sensitization, habituation and classical conditioning paradigms have been demonstrated in Drosophila. Furthermore, behavioral screens have been devised for the detection of strains defective in these behaviors. One of the commonly used screening strategies for characterization of Drosophila mutants involves operant conditioning and olfactory cues (Fig. 50-4). Flies are exposed to an odorant, such as 4-methylcyclohexanol, spread onto an electrified wire grid. Eventually, normal flies learn to associate the electric shock with the odorant and avoid it. After this type of training, learning can be quantitated by exposing the flies to two chambers that have different odorants and determining the percentage of flies that have learned to avoid the conditioning odorant. This percentage is known as the avoidance index. Fly stocks can be mutagenized, and mutant flies that are altered in their ability to associate the electric shock with the odorant selected. Avoidance indices of 0.9 are common for wild-type Drosophila, while mutant strains may have avoidance indices of 0.3 or less.

Figure 50-4

A: The training and testing paradigm for Drosophila learning and memory. Flies are first trained to avoid an odorant, such as cyclohexanol, on an electrified grid. After training, the flies are allowed to choose between two chambers, one (a) containing (more...)

One of the best-documented behavioral mutants is dunce (dnc), in which a deficiency in the structural gene for a cAMP phosphodiesterase (PDE) has been established (Fig. 50-4B) [18]. These PDEs specifically hydrolyze cAMP to 5′-AMP and are responsible for returning cellular concentrations of cAMP to resting values following stimulation of adenylyl cyclase. Many isoforms of PDE have been characterized that are differentially regulated and expressed in a tissue-specific manner (Chap. 22). The dnc gene organization is extremely complex and is distributed over at least 150 kb of the Drosophila genome. Multiple RNA transcripts from this gene encode several different PDE protein isoforms. Antibodies generated to a conserved region of these isoforms have been used to show that the dnc PDE is concentrated within a region of the Drosophila nervous system known as the mushroom body. The mushroom body has been implicated as the anatomical site of olfactory learning and memory on the basis of many experiments. The relatively restricted expression of the dnc PDE to this region suggests that the dnc protein plays a direct role in mediating memory formation. How the PDE gene defect is translated into a behavioral deficit is not known. The behavioral deficit does not seem to be due to a generalized developmental defect in the nervous system since the brains of dnc flies appear morphologically normal and the flies are able to form memory, albeit extremely short-term.

A related Drosophila behavioral mutant is rutabaga (rut). Like dnc mutants, memory forms in the rut mutant but decays rapidly. Adenylyl cyclase from rut flies has an increased K_m for ATP. Two forms of adenylyl cyclase activity can be distinguished in normal Drosophila, based on sensitivity to calmodulin. The rut mutants appear to be deficient in calmodulin-stimulated adenylyl cyclase. Mammalian cDNA clones for adenylyl cyclase have been used to isolate the Drosophila homologues. The rut gene product encodes a calmodulin-stimulated adenylyl cyclase [19]. The rut mutants contain a single base change, which results in an amino acid substitution of arginine for the glycine normally found at residue 1026 of the adenylyl cyclase. This single amino acid substitution is sufficient to completely abolish cyclase activity. These observations suggest that, in Drosophila, complex behaviors such as learning can be drastically altered by relatively simple alterations to the genome.

Mutant flies producing defective enzymes required for neurotransmitter metabolism, such as DOPA decarboxylase, have been reported to be deficient in olfactory and behavioral learning paradigms. DOPA decarboxylase is necessary for the synthesis of dopamine, serotonin and octopamine. Because this enzyme is essential not only for neurotransmitter synthesis but also for synthesis of the cuticle, these mutants are impaired in their exoskeletal development and the phenotype is complex [17]. This problem has been overcome by use of temperature-sensitive alleles of the mutation. These alleles function normally and permit normal development at 25°C. Subsequent induction of the neural phenotype can be brought about by relatively brief exposure of the adult flies to an elevated temperature which inhibits the mutant enzyme.

The amnesiac (amn) mutant has been characterized at a molecular level, and the gene responsible has been identified as a neuropeptide precursor which contains two peptides related to the mammalian pituitary adenylyl cyclase-activating peptide (PACAP) [20]. As the name implies, PACAP elevates cAMP levels by activating adenylyl cyclase. However, it has not yet been demonstrated that any of the neuropeptides derived from the amn precursor elevate cAMP in Drosophila neurons.

In addition to these behavioral mutant fly strains obtained by forward-genetic screening, a number of reverse-genetic experiments have demonstrated that all of the following play essential roles in learning and memory: the α subunit of G_s, both the regulatory and catalytic subunits of PKA and a fly homologue of CREB designated dCREB2. Interestingly, different transcripts of the same dCREB2 gene can be formed by alternative splicing to produce both an activator and a repressor of gene transcription, much like the mammalian CRE modulator gene (see Chap. 26).

Other invertebrates are being developed as model systems for experimental studies of memory

In the marine invertebrate Hermissenda crassicornis, the CS is a positive phototaxis that is paired with high-speed rotation, the US, which leads to suppression of the unconditioned response [21]. Daily sessions of 50 to 100 pairings for a few days result in retention of the learned response for over 2 weeks. Of a number of cellular changes found following conditioning, a reduced K⁺ current in type B photoreceptors has been pursued experimentally. From experiments involving various drugs and microinjections of enzymes, it has been inferred that Ca²⁺/calmodulin-activated protein kinases and PKA mediate associative learning in Hermissenda. However, in Hermissenda, unlike Aplysia, Ca²⁺ influx is the result of depolarization, as is proposed in hippocampal LTP, whereas in Aplysia, it is neurotransmitter-mediated. Inhibitors of protein synthesis prolong the altered biophysical correlates of conditioning in Hermissenda. It has been speculated that this seemingly paradoxical result is related to short-term, rather than long-term, memory formation. Microinjection experiments and the application of phorbol esters indicate that protein kinase C (PKC) and the phosphoinositide pathway may also play an important role in associative learning, as is further discussed below, in relation to LTP.

Caenorhabditis elegans represents another species which has been subjected to genetic analysis in order to identify the molecular components of neural development and, ultimately, of learning and memory [22]. Mutant animals which show defects in both short-term and long-term memory have been isolated, but identification of the genes affected has not been completed. Although C. elegans is less impressive than the fruit fly in terms of its behavioral repertoire, it has the anatomical advantage of possessing only a few hundred neurons. Thus, if altered phenotypic expression in the morphology of the nervous system occurs, it can be more readily correlated with altered behavior.