662 Opinion TRENDS in Neurosciences Vol.26 No.12 December 2003 Learning in Aplysia:lookingatsynaptic plasticity from both sides Adam C. Roberts1 and David L. Glanzman2 1Interdepartmental PhD Program in Molecular, Cellular and Integrative Physiology, UCLA, Los Angeles, CA 90095-1606, USA 2Departments of Physiological Science and Neurobiology, and the Brain Research Institute, David Geffen School of Medicine at UCLA, Los Angeles, CA 90095-1761, USA Until recently, learning and memory in invertebrate have convinced significant numbers of neuroscientists organisms was believed to be mediated by relatively that Hebbian LTP does indeed underlie certain forms of simple presynaptic mechanisms. By contrast, learning learning and memory in vertebrates [13]. Meanwhile, and memory in vertebrate organisms is generally knowledge of the cellular mechanisms of learning in thought to be mediated, at least in part, by postsyn- invertebrates has also progressed. Postsynaptic mechan- aptic mechanisms. But new experimental evidence isms, including NMDA-receptor-dependent plasticity, from research using a model invertebrate organism, the have been shown to play crucial roles in learning in marine snail Aplysia, indicates that this apparent dis- Aplysia. Here, we review recent data that indicate the tinction between invertebrate and vertebrate synaptic importance of postsynaptic mechanisms in both classical mechanisms of learning is invalid: learning in Aplysia conditioning and sensitization in Aplysia, and discuss cannot be explained in terms of exclusively presynaptic their implications for a general understanding of the mechanisms. NMDA-receptor-dependent LTP appears synaptic mechanisms of learning and memory. to be necessary for classical conditioning in Aplysia. Furthermore, modulation of trafficking of postsynaptic Contributions of presynaptic and postsynaptic ionotropic glutamate receptors underlies behavioral mechanisms to classical conditioning in Aplysia sensitization in this snail. Exclusively presynaptic pro- Classical conditioning, a form of learning originally cesses appear to support only relatively brief memory described in dogs by the Russian physiologist and in Aplysia. More persistent memory is likely to be psychologist Ivan Pavlov [14], occurs when a more-or- mediated by postsynaptic processes, or by presynaptic less behaviorally neutral stimulus (the conditioned stimu- processes whose expression depends upon retrograde lus or CS) is presented to an animal together with a signals. reinforcing stimulus (the unconditioned stimulus or US). Before training, presentation of the US evokes a reflexive Approximately a decade ago it was widely accepted that response in the animal (the unconditioned response or simple forms of invertebrate learning, such as habituation, UCR), whereas the CS does not. However, as a result of sensitization and classical conditioning, could be explained paired stimulation with the CS and US, the CS alone in terms of comparatively simple forms of synaptic becomes able to evoke a response in the animal (the plasticity. As epitomized by the marine snail Aplysia, conditioned response or CR) that resembles the reflexive learning-related synaptic plasticity in invertebrates was UCR [15]. Most experimental investigations of classical thought to involve exclusively presynaptic cellular conditioning have used mammals but such conditioning changes, including presynaptic facilitation and presyn- has also been demonstrated in a variety of invertebrate aptic depression [1–3]. By contrast, vertebrate learning is organisms, including mollusks and arthropods [16]. Clas- believed to depend on more complex forms of synaptic sical conditioning of a simple defensive withdrawal reflex plasticity that involve postsynaptic changes. Most promin- in the marine snail Aplysia was first described in 1981 by ent among the candidate cellular mechanisms of learning Eric Kandel and colleagues [17]. This form of invertebrate in vertebrates is NMDA-receptor-dependent long-term learning was originally hypothesized to be due to an potentiation (LTP) [4,5] – also known as Hebbian LTP exclusively presynaptic mechanism, known as activity- [6,7]. NMDA-receptor-dependent synaptic plasticity was dependent presynaptic facilitation (ADPF; Box 1). How- not believed to exist in the nervous systems of inverte- ever, the discovery that sensorimotor synapses of Aplysia brates [8], which seemed to provide a partial explanation possess the capacity for NMDA-receptor-dependent LTP for their more primitive learning capabilities. [18–20] led to an alternative hypothesis: that classical Although many remain skeptical of the solidity of the conditioning might depend, in part, on Hebbian LTP [21]. experimental link between Hebbian LTP and vertebrate Support for this hypothesis has come mostly from studies learning [9–11], convergent results from a broad array of that used so-called ‘cellular analogs’ of classical condition- studies, many of which have used transgenic mice [12], ing, which involve reduced preparations or synapses in dissociated cell culture [22–25]. But a new study by Corresponding author: David L. Glanzman ([email protected]). Antonov et al. [26], who used a reduced preparation of http://tins.trends.com 0166-2236/$ - see front matter q 2003 Elsevier Ltd. All rights reserved. doi:10.1016/j.tins.2003.09.014 Opinion TRENDS in Neurosciences Vol.26 No.12 December 2003 663 Box 1. LTP and classical conditioning in Aplysia: a brief history Three influential papers published together in 1983 in Science [98–100] Adecadelater,whetherAplysia sensorimotor synapses can express demonstrated differential classical conditioning of the withdrawal Hebbian LTP was re-investigated. Lin and Glanzman [19,20] demon- reflex of Aplysia, and proposed a simple synaptic mechanism to strated unambiguously Hebbian LTP at synapses in dissociated cell account for this form of invertebrate learning. The mechanism, activity- culture. Like LTP of CA1 hippocampal synapses [5], this invertebrate LTP dependent presynaptic facilitation (ADPF), involves a presynaptic depends on postsynaptic depolarization, activation of NMDA receptors 2 2 interaction between elevated intracellular Ca þ levels and serotonin [47] and elevated postsynaptic Ca þ concentration. The demonstration 2 (5-HT). The elevated presynaptic Ca þ concentration results from the of Hebbian LTP of sensorimotor synapses raised the question of conditioned stimulus (CS) – weak tactile stimulation of the siphon – whether LTP mediates learning in Aplysia.MurphyandGlanzmanused whereas 5-HT is released onto the sensory neurons in response to the a cellular analog of classical conditioning to re-examine whether 2 unconditioned stimulus (US) – strong electrical shock of the tail [40]. In Hebbian LTP was involved. They found that injecting the Ca þ chelator ADPF there is greater activation of presynaptic adenylyl cyclase, and 1,2-bis(2-aminophenoxy)ethane-N,N-N0,N0-tetraacetic acid (BAPTA) into consequently greater synthesis of cAMP [37,38], than in standard the motor neuron before training blocked synaptic enhancement [23]. presynaptic facilitation in Aplysia. The enhanced synthesis of presyn- Murphy and Glanzman further showed that the NMDA-receptor- aptic cAMP results in enhanced downstream modulatory actions [41] antagonist Dl-2-amino-5-phosphonovalerate (APV) blocked associative (Figure 3a of main text), and greater transmitter release from the synaptic enhancement due to paired CS–US stimulation. By contrast, presynaptic terminals. APV did not affect non-associative enhancement of the synapse due to A1984study[101]comparedADPFwithaHebbian[6]mechanismasa unpaired delivery of the CS and US [22]. A study by Bao et al. [102] synaptic explanation for classical conditioning in Aplysia.Hebbian found that postsynaptic BAPTA blocked associative enhancement of plasticity depends on the conjunctive occurrence of presynaptic activity sensorimotor synapses in vitro.Finally,newevidencefromastudyby and strong postsynaptic depolarization [5,7]. A cellular analog of Antonov et al. [26] supports a role for NMDA-receptor-dependent LTP classical conditioning was used to test for the involvement of Hebbian in behavioral classical conditioning. plasticity. Here, direct activation of a sensory neuron via a microelec- How was the contribution of Hebbian plasticity to classical condition- trode serves as the CS, and electrical shock of the pedal (tail) nerves ing originally missed [101]? It seems probable that the hyperpolariz- serves as the US. Successful conditioning is represented by associative ation and depolarization of the motor neuron were insufficient. The enhancement of the sensorimotor excitatory postsynaptic potential postsynaptic hyperpolarizing current used might have declined at distal (EPSP). Two tests were performed. In the first, firing of the postsynaptic postsynaptic sites such that the postsynaptic membrane was only motor neuron with a microelectrode was used as the US rather than weakly hyperpolarized, if at all [103]. (Interestingly, a similar mistake nerve shock. Paired stimulation using postsynaptic stimulation for the appears to have led to an early, erroneous conclusion that postsynaptic US did not enhance the sensorimotor EPSP. In the second test, the depolarization is not required for the induction of LTP at the Shaffer- motor neuron soma was hyperpolarized to prevent it from firing during collateral-to-CA1-neuron synapse [104,105].) Furthermore, action the US (tail-nerve shock). Postsynaptic hyperpolarization during the