Invertebrates

There is much evidence that learning and memory result from cellular changes occurring within individual neurons (Byrne, 1987), but the cellular mechanisms underlying these changes have resisted definitive elucidation. The mechanisms that produce a change in neuronal structure have received scant attention even though this class of mechanisms was first suggested more than a century ago (Ramón y Cajal, 1988). Researchers have observed morphological correlates of learning in several vertebrate and invertebrate model systems (Bailey and Kandel, 1993), but no one has proved that modifications of neuronal structure affect behavior. Nevertheless, the variety of model systems with morphological correlates suggests that this is one of the mechanisms that underlies learning.

An investigation of the morphological correlates of learning and memory should incorporate several features. First, there should be a readily observable behavior that can be modified by training. Second, the circuit mediating that behavior should be well known. Third, the morphology of the neurons in that circuit must be accessible to analysis. Although no real-world model system meets all of these criteria, a number of investigators have turned to invertebrates for model systems with few compromises.

The structural modifications vary among different systems. This is not surprising because morphological features may contribute to neuronal function in several ways. This article will focus on two broad classes of morphological changes: the large-scale remodeling of neuronal arborizations arising from outgrowth that seem to result in the formation of new synaptic connections and more localized structural changes that presumably affect preexisting connections.

Neurite Outgrowth

Cajal's original hypothesis suggested that learning is mediated by the outgrowth of new axons and the formation of new synaptic connections, a process that recapitulates brain development (Ramón y Cajal, 1988). This hypothesis has been tested in detail in the marine mollusk Aplysia. Two simple defensive reflexes in this animal have proved useful for studies of the cellular and molecular mechanisms of nonassociative learning: the siphon-gill withdrawal reflex and the tail-siphon withdrawal reflex.

Both withdrawal reflexes can be enhanced by sensitization, a form of nonassociative learning (Castellucci, Pinsker, Kupfermann, and Kandel, 1970; Walters, Byrne, Carew, and Kandel, 1983). Sensitization is the enhancement of a behavioral response resulting from the application of a novel or noxious stimulus to the animal. Sensitization in a variety of animals seems to be a short-lived phenomenon, recovering over the course of minutes to an hour. The long-term form of sensitization lasts from one day to roughly three weeks. Recent evidence suggests that the situation may be more complex. For example, there is a distinct form of sensitization with intermediate duration (Sutton and Carew, 2000). Investigators have identified several elements of the circuits mediating these reflexes. In both circuits, the sensory neurons are sites of short-and long-term plasticity. Neurite outgrowth followed long-term but not short-term sensitization.

Following sensitization of the siphon-gill withdrawal reflex, training increased the complexity of the axonal arbor of sensory neurons, as measured by total neurite length (Bailey and Chen, 1988a). In addition, the number of varicosities per neuron increased. These observations were consistent with enhanced convergence onto follower motor neurons. Tail sensory neurons yielded similar results (Wainwright, Zhang, Byrne, and Cleary, 2002). The effects of training, however, were limited to a region of the arborization in the pedal ganglion, which is where follower motor neurons are located. The total arborization length in this region was increased, as was the number of branch points and varicosities.

The honeybee is an interesting case because it has a powerful facility for learning the features and location of food. One of the most profound learning experiences for a foraging bee occurs on its first flight from the hive. Presumably, the animal learns navigational cues that the target reinforces, but the precise coding of the information remains unclear.

Given the complexity of learning, it likely draws on large portions of the brain. Some research has pointed to a correlation between learning and an increase in volume of two brain regions: the mushroom body and antennal lobe glomeruli (Farris, Robinson, and Fahrbach, 2001; Sigg, Thompson, and Mercer, 1997). In the mushroom body, the growth appears to be due at least in part to an increase in sprouting from Kenyon cell dendrites (Farris, Robinson, and Fahrbach, 2001). A form of olfactory learning in the fruit fly Drosophila yielded similar results (Balling et al., 1987).

A distinctive approach to morphological correlates has been taken with Drosophila. Rather than looking directly for correlates following a training session, investigators generate mutants that are deficient in learning and observe effects of the mutations on neuronal structure. In some cases these mutations have been found to affect the morphology of neurons in the central nervous system. Two commonly used mutations are dnc and rut. Both mutations impair a form of olfactory conditioning in which flies associate a particular odor with an electrical shock. Their critical effect is on the cAMP second-messenger pathway: dnc increases cAMP levels, and rut reduces them. Although they have opposite biochemical effects, both mutations result in larger numbers of spines and varicosities in a central process of the sensory neuron innervating the antero-notopleural thoracic bristle (Corfas and Dudai, 1991). More frequently, the effects of the mutations have been observed at the neuromuscular junctions of animals at the larval stage of development. In this preparation, the effects of the two mutations are different: dnc mutations increase motor neuron branching (Schuster et al., 1996), whereas rut mutations decrease branching (Cheung, Shayan, Boulianne, and Atwood, 1999).

The mollusk Hermissenda has been the subject of studies of associative conditioning between visual and rotational stimuli. A crucial site of plasticity is the photoreceptor. Following training, the arborization of the medial type B photoreceptor was reduced compared to controls (Alkon et al., 1990). This result is consistent with other biophysical and biochemical changes resulting in decreased synaptic strength to follower neurons.

In these model systems changes in the complexity of the arbor parallels the changes in synaptic efficacy. Neurite outgrowth presumably strengthens the behavioral response by forming new synapses to follower neurons. Neurite retraction presumably weakens the behavioral response by reducing the number of synapses. Because these changes appear to be associated only with long-lasting forms of learning, it seems that there is a high threshold for induction. Such changes would not seem to admit of easy reversal, but no testing has confirmed this hypothesis.

Modification of Synaptic Ultrastructure

Synaptic strength can be modulated by changes in neuronal structure that are more subtle than those outlined in the previous section. Usually such changes involve the structural modification of preexisting synapses. This possibility has been examined in great detail in Aplysia.

Following long-term sensitization training, the number of sensory neuron varicosities increases as does the size of the axonal arborization. This outgrowth results in an increase in the number of active zones, in the length of the active zone, and the number of vesicles in proximity to the release site (i.e., the vesicle complement; Bailey and Chen, 1983). These same features are smaller in animals subjected to long-term habituation. For example, sensory neurons from these animals had on average 35 percent fewer varicosities compared to controls. Quantitative analysis of the time course over which these anatomical changes occur during long-term sensitization suggests that alterations in the number of sensory neuron varicosities and active zones persist in parallel with the behavioral retention of the memory, whereas active zone length and vesicle complement do not (Bailey and Chen, 1989). Thus, there may be an overlapping cascade of mechanisms that sustain modifications in synaptic function.

In contrast with the extensive structural changes following long-term training, the morphological correlates of short-term memory in Aplysia (lasting minutes to hours rather than days to weeks) are far less pronounced and are primarily restricted to shifts in the proximity of synaptic vesicles adjacent to sensory neuron active zones, a phenomenon that may reflect altered levels of transmitter mobilization (Bailey and Chen, 1988b). These studies in Aplysia suggest a clear difference in the time course of structural events that underlie memories of differing durations. The transient duration of short-term memories probably involves the covalent modification of preexisting proteins and is accompanied by modest structural remodeling in the vicinity of the active zone, such as the translocation of synaptic vesicles to the release site.

Another model system in which synaptic structure seems correlated with function is the crayfish neuromuscular junction. This system does not permit an examination of the effects of a specific learning paradigm because the cellular plasticity underlying behavioral modifications does not occur at the neuromuscular junction. Instead, attention is focused on the effects of high-frequency stimulation, which produces both short-and long-term enhancements in synaptic strength.

In the crayfish preparation, a single axon makes numerous synaptic contacts with the muscle fiber. These synapses have different functional properties (Wojtowicz et al., 1994); some are affected by long-term facilitation, and some are not. Moreover, many synapses that release transmitter do not contribute to the EPSP, suggesting that changes in EPSP amplitude can be effected by modifying only a fraction of synapses.

Early ultrastructural studies supported the idea that high-frequency stimulation produced an increase in the percentage of synapses exhibiting presynaptic active zones. These observations suggest that alterations in neuronal activity can induce rapid structural transformations at the synapse, affecting primarily the active-zone. However, subsequent studies suggested that the effects of high-frequency stimulation were not persistent. Stimulation at 20 Hz for ten minutes produced an increase in the number of synapses per unit area of terminal membrane, a decrease in the synaptic area, and an increase in the number of dense bodies per unit of terminal membrane area or synapse area. Forty-five minutes later, however, the synapse was still facilitated, but these structural features were no longer different from control. Only one structural change was observed following long-term facilitation: In synapses that had more than one dense body, there was an increase in the number of synapses per unit area of terminal.

In Drosophila, mutants that interfere with learning also produce changes in the structure of the neuro-muscular junction (Renger et al., 2000). For example, dnc mutations, which increase cAMP levels, produced relatively modest structural changes: the ratio of docked to undocked vesicles increased, and pre-synaptic and postsynaptic specializations became less densely stained. Structural changes following the rut mutation, which decreases cAMP levels, were more profound. The number of synapses per unit area of terminal decreased, but the size of each synapse increased. The ratio of docked to undocked vesicles decreased, and staining intensity of presynaptic and postsynaptic specializations was unaffected. If the structural changes at these peripheral synapses reflect similar changes within the central nervous system, then they could contribute to the learning deficits.

Causal Relationship Between Structure and Function

Notwithstanding the structural changes that seem to correlate with learning, the relationship between structure and function is not necessarily a simple one. For example, at the crayfish neuromuscular junction, muscle fibers are innervated by both tonic and phasic motor neurons. Tonic motor neurons are typically highly active, and their terminals typically evince large synaptic varicosities with large numbers of vesicles. Phasic motor neurons with low levels of impulse activity have thinner terminals with fewer vesicles. Aside from the greater number of tonic terminals, there are no marked differences in synaptic structure in the two populations of neurons (Msghina, Govind, and Atwood, 1998). Therefore other mechanisms, presumably biophysical or biochemical, are more important determinants of synaptic strength between the two different types of motor neurons.

In an elegant series of experiments at the developing neuromuscular junction of larval Drosophila, moderate underexpression of a cell adhesion molecule resulted in axonal outgrowth and varicosity formation (Schuster et al., 1996). More severe underexpression resulted in axonal retraction and decreased varicosity formation. These morphological changes did not by themselves produce changes in synaptic strength, however. Unknown intrinsic mechanisms maintained synaptic strength at an appropriate level by adjusting the number and size of active zones (Stewart, Schuster, Goodman, and Atwood, 1996). Enhancing synaptic strength required an activation of the cAMP second-messenger pathway and phosphorylation of the nuclear regulatory protein CREB. Thus, in addition to the difficulties of demonstrating the structural mechanisms of learning, there are the further challenges of demonstrating that those changes contribute to the modulated function of the neural circuit underlying the behavior.

In recent studies in Aplysia discussed earlier (Wainwright, Zhang, Byrne, and Cleary, 2002), the structural changes induced by four days of long-term sensitization training did not occur after only one day of long-term sensitization training. Moreover, the neurite outgrowth that accompanied four days of training occurred on both sides of the animal, even though the behavioral modification was limited to the trained side. These dissociations do not rule out the role of structural changes in learning but caution against the overinterpretion of results in the absence of appropriate controls.

It is difficult indeed to nail down the possible causal relationship between structural changes and behavioral changes. Structural changes require a large number of cellular processes, such as protein synthesis, cytoskeletal reorganization, organelle translocation, and membrane-membrane interactions. Therefore, in most model systems there is no single drug or manipulation that can block selected structural changes without affecting other normal cellular processes. Researchers are always seeking systems and control experiments that might permit a more precise determination of the possible causal role of structural changes.

Bibliography

Alkon, D. L., Ikeno, H., Dworkin, J., McPhie, D. L., Olds, J. L., Lederhendler, I., Matzel, L., Schreurs, B. G., Kuzirian, A., Collin, I., and Yamoah, E. (1990). Contraction of neuronal branching volume: An anatomical correlate of Pavlovian conditioning. Proceedings of the National Academy of Sciences of the United States of America 87, 1,611-1,614.

Bailey, C. H., and Chen, M. (1983). Morphological basis of long-term habituation and sensitization in Aplysia. Science 220, 9,193.

—— (1988a). Long-term memory in Aplysia modulates the total number of varicosities of single identified sensory neurons. Proceedings of the National Academy of Sciences of the United States of America 85, 2,373-2,377.

—— (1988b). Morphological basis of short-term habituation inAplysia. Journal of Neuroscience 8, 2,452-2,459.

—— (1989). Time course of structural changes at identified sensory neuron synapses during long-term sensitization in Aplysia. Journal of Neuroscience 9, 1,774-1,780.

Bailey, C. H., and Kandel, E. R. (1993). Structural changes accompanying memory storage. Annual Review of Physiology 55, 397-426.

Balling, A., Technau, G. M., and Heisenberg, M. (1987). Are the structural changes in adult drosophila mushroom bodies memory traces? Studies on biochemical mutants. Journal of Neurogenetics 4, 65-73.

Byrne, J. H. (1987). Cellular analysis of associative learning. Physiological Review 67, 329-439.

Castellucci, V. F., Pinsker, H., Kupfermann, I., and Kandel, E. R.(1970). Neuronal mechanisms of habituation and dishabituation of the gill withdrawal reflex in Aplysia. Science 167, 1,745-1,748.

Cheung, U. S., Shayan, A. J., Boulianne, G. L., Atwood, H. L.(1999). Drosophila larval neuromuscular junction's responses to reduction of cAMP in the nervous system. Journal of Neurobiology 40, 1-13.

Corfas, G., and Dudai, Y. (1991). Morphology of a sensory neuron in Drosophila is abnormal in memory mutants and changes during aging. Proceedings of the National Academy of Sciences of the United States of America 88, 7,252-7,256.

Farris, S. M., Robinson, G. E., and Fahrbach, S. E. (2001). Experience-and age-related outgrowth of intrinsic neurons in the mushroom bodies of the adult worker honeybee. Journal of Neuroscience 21, 6,395-6,404.

Msghina, M., Govind, G. K., and Atwood, H. L. (1998). Synaptic structure and transmitter release in crustacean phasic and tonic motor neurons. Journal of Neuroscience 18, 1,374-1,382.

Ramón y Cajal, S. (1988). Anatomicophysiological considerations on the cerebrum. In J. DeFelipe and E. G. Gros, eds., Cajal on the cerebral cortex. New York: Oxford University Press.

Renger, J. J., Ueda, A., Atwood, H. L., Govind, C. K., and Wu, C.-F.(1996). Genetic dissection of structural and functional components of synaptic plasticity. II. Fasciclin II controls presynaptic structural plasticity. Neuron 17, 655-667.

—— (2000). Role of cAMP cascade in synaptic stability and plasticity: Ultrastructural and physiological analyses of individual synaptic boutons in drosophila memory mutants. Journal of Neuroscience 20, 3,980-3,992.

Schuster, C. M., Davis, G. W., Fetter, R. D., and Goodman, C. S.(1996). Genetic dissection of structural and functional components of synaptic plasticity. I. Fasciclin II controls synaptic stabilization and growth. Neuron 17, 641-654.

Sigg, D., Thompson, C. M., Mercer, A. R. (1997). Activity-dependent changes to the brain and behavior of the honey bee, Apis mellifera (L.). Journal of Neuroscience 17, 7,148-7,156.

Stewart, B. A., Schuster, C. M., Goodman, C. S. and Atwood, H. L.(1996). Homeostasis of synaptic transmission in Drosophila with genetically altered nerve terminal morphology. Journal of Neuroscience 16, 3,877-3,886.

Sutton, M. A. and Carew, T. J. (2000). Parallel molecular pathways mediate expression of distinct forms of intermediate-term facilitation at tail sensory-motor synapses in Aplysia. Neuron 26, 219-231.

Wainwright, M. L., Zhang, H., Byrne, J. H. and Cleary, L. J. (2002).

Localized neuronal outgrowth induced by long-term sensitization training in Aplysia. Journal of Neuroscience 20, 4,132-4,141.

Walters, E. T., Byrne, J. H., Carew, T. J., and Kandel, E. R. (1983). Mechanoafferent neurons innervating tail of Aplysia. II. Modulation by sensitizing stimuli. Journal of Neurophysiology 50, 1,543-1,559.

Wojtowicz, J. M., Marin, L., and Atwood, H. L. (1994). Activity-induced changes in synaptic release sites at the crayfish neuromuscular junction. Journal of Neuroscience 14, 3,688-3,703.

Craig H.Bailey

Revised byLenCleary