How does the amygdala mediate memory storage in fear conditioning?As described above, there is considerable evidence suggestingthat the CS-US association occurs in the basolateral complex.Inputs to the basolateral complex use glutamate as the transmitterand activate synapses expressing both AMPA and NMDA receptors.Results from several laboratories have shown that infusionof the NMDA receptor antagonist D,L-2-amino-5-phosphonovalericacid (AP5) into the basolateral complex blocks the acquisitionof amygdala-dependent conditioning (24, 68, 100, 176). Withinthe basolateral complex, selective lesions of the lateral amygdaladisrupt fear conditioning (9, 72, 117, 184, 281). Recordingsof single units with the LA in vivo have shown that these neuronsrespond to both auditory (tone; CS) and somatosensory (shock;US) stimulation (199, 232). Auditory fear conditioning enhancesshort latency CS firing in LA neurons (216). The enhanced neuronalfiring to the CS observed in the LA precedes behavioral expressionof the conditioned response (225). Conditioning-associatedincreases in firing to the CS are also observed in the auditorycortex, but this change is subsequent to the firing rate changesin the LA (215). Furthermore, animals can undergo auditoryfear conditioning following lesions to either the thalamoamygdalaor cortico-amygdala pathways, but not both (24, 233). Finally,the auditory evoked potentials recorded in the LA are enhancedin conditioned rats, but not in pseudoconditioned controlsthat receive unpaired CS and US presentations (229).
Most recently, a study by Rosenkrantz and Grace (238) providescompelling evidence suggesting that the LA is indeed a siteof plasticity during fear conditioning. Using intracellularrecordings in vivo, these authors show that pairing an odor(CS) with a foot-shock (US) enhanced the amplitude of the responseto the paired odor. Simultaneously they also showed that theresponse to an unpaired odor is unaffected. The potentiationof the paired response could be blocked by hyperpolarizingthe postsynaptic cell during pairing, indicating that the plasticityrequired the postsynaptic neuron.
Together, these studies have led to the suggestion that NMDAreceptor-mediated synaptic plasticity (long-term potentiation,LTP) within the basolateral complex underlies the acquisitionand storage of memory related to fear conditioning (140). Moreover, it has been suggested that these changes occur inprojection neurons in the LA (116, 117). In agreement withthis suggestion, in vivo recordings have shown that tetanicstimulation of thalamic inputs to the LA enhance both electricallyevoked as well as auditory evoked responses in the LA (228).Unfortunately, whether the changes in synaptically evoked responseswere sensitive to NMDA receptor blockade was not tested. Ina different set of experiments, McKernan et al. (171) haveshown that the amplitude of AMPA evoked currents at thalamicinputs to neurons in the LA were larger in animals that hadundergone auditory fear conditioning. With the use of pairedpulse facilitation as an index of release probability (301),it has been suggested that the site of plasticity during fearconditioning is presynaptic (171). Finally, if fear conditioningresults from LTP at inputs to neurons in the basolateral complex,one might expect that the induction of LTP at synapses thathave participated in fear conditioning would be occluded. Thishas recently been demonstrated at inputs to pyramidal neuronsin the LA (273). In summary, there is convincing evidencethat the response of neurons in the LA to CS stimulation isenhanced after fear conditioning.
The basic properties of LTP, input specificity, cooperativity,and associativity, have made it an attractive cellular modelof associative learning (18). While LTP has been studied atmany other synapses, the most extensively studied form occursat excitatory synapses between Schaffer collaterals and CA1pyramidal neurons in the hippocampus (133, 137). These studieshave shown that the induction of LTP requires a rise in postsynapticcalcium (138). In the classical model of LTP, depolarizationof the postsynaptic neuron relieves the Mg2+ blockade of NMDAreceptors (191), allowing Ca2+ to enter via NMDA receptorchannels to trigger signal transduction cascades that initiatethe molecular changes that underlie LTP (18, 139). The depolarizationrequired to activate NMDA receptors is provided either by AMPAreceptors (during tetanically induced LTP) or by backpropagatingaction potentials (134). While the postsynaptic inductionof this LTP is certain, the final locus of the change that underliesLTP, whether it is presynaptic or postsynaptic, has been muchdebated (139). NMDA receptor-independent forms of LTP havealso been described. In NMDA-independent induction of LTP,postsynaptic calcium comes from voltage-gated calcium channels(VGCCs) (296) or calcium-permeable AMPA receptors (135). Finally, at some synapses, a presynaptic form of LTP, whichdoes not require NMDA receptors, has also been described (187).
Both cortical and thalamic afferents to the LA are capable oflong-term plasticity after tetanic stimulation. Tetanic stimulationof afferents to the basolateral complex has been shown to resultin LTP both in vivo (36, 143, 228) and in vitro (10, 32,33, 84, 289). All studies are in agreement that a rise inpostsynaptic calcium is required for the induction of LTP atboth cortical and thalamic inputs to basolateral neurons. Thesite of plasticity has been proposed to be presynaptic withan increase in the probability of transmitter release (84,143, 273). However, whether induction of LTP at synapsesto LA neurons requires activation of NMDA receptors is notclear. Although some groups have shown NMDA-dependent LTP invitro (84), others have found that tetanically induced LTPdoes not require activation of NMDA receptors (32). Furthermore,a recent study found that tetanic stimulation alone was ineffectivein inducing LTP, but pairing action potentials with thalamicexcitatory postsynaptic potetials did lead to a form of LTPthat required activation of L-type VGCCs for induction (273,289). As mentioned above, NMDA receptors have also been suggestedto contribute to basal transmission at thalamic inputs to thelateral amygdala (127, 291). In agreement with this finding,behavioral evidence has suggested that AP5 may block the acquisitionof fear conditioning by disrupting normal synaptic transmissionin the amygdala (62, 123). These results challenge the notionthat NMDA receptor-mediated plasticity within the amygdala isthe mechanism underlying the acquisition and storage of fear-relatedmemories (23).
One potential problem has been that all the in vivo studies(and some in vitro studies) have relied on field potentialmeasurements of inputs to the amygdaloid complex. However,unlike in clearly layered structures like the hippocampus orcerebellum, there is little organization to the architectureof the neurons and where they receive their synaptic inputs.Thus, because there are no clear sources and sinks when inputsare stimulated, it is difficult to separate field potentialsassociated with synaptic currents and those associated withaction potentials (143). Thus changes in field potentialsare difficult to interpret with regard to the locus of changethat is being measured. Neurons in the lateral amygdala haveextensive local connections (153). Following fear conditioning,an increase in the correlated firing of neurons in the lateralamygdala have been reported (199, 215), suggesting that theremight also be changes in the local connections or cell propertiesfollowing conditioning. Thus it is possible that the changesin field potential measurements following fear conditioningmay result from changes in the strength of connections betweenneurons rather than of excitatory inputs to these cells. Onlyin a few studies have changes in field potentials been correlatedwith changes in synaptic potentials (84).
Recent in vitro studies have also demonstrated several otherforms of plasticity in the basolateral amygdala. First, glutamatergicinputs to interneurons in the basolateral complex activatesynapses that do not express NMDA receptors. LTP at these inputsis triggered by a rise in calcium by influx via AMPA receptors(135). Because interneurons are the only source of inhibitorypotentials in the basolateral complex, potentiation of excitatoryafferents to interneurons leads to an increase in the amplitudeof the disynaptic inhibitory potentials (135). Second, as inthe hippocampus, low-frequency stimulation of inputs to thebasal nucleus leads to long-term depression (LTD) (75, 282).This LTD is input specific and requires activation of metabotropicreceptors and a rise in postsynaptic calcium levels (75, 128, 223). Lastly, low-frequency stimulation of inputs tothe basal nucleus evokes a slow onset of facilitation of theseinputs (125). Its induction was shown to result from activationof kainate receptors present in the basal nucleus (125). Thisis a novel form of synaptic plasticity that has not been previouslydescribed, and its relationship to LTD described by others,is not clear. The possible roles of these phenomena in amygdala-mediatedlearning are not known at present. In summary, although therole of the basolateral complex in the induction of fear conditioningis clear, there is considerable debate as to the nature ofthe changes that occur during the conditioning.
If the underlying mechanism for the acquisition and storageof fear-related memories is LTP within the amygdala, it isuseful to consider how this might operate mechanistically.As originally formulated, LTP results from the coincident activationof two different inputs to single cells. One of these is consideredthe "weak" input and represents the CS. The other input isa "strong" input and represents the US. The "strong" inputis capable of activating postsynaptic cells causing a strongdepolarization and/or evoking repetitive action potential firing.The weak input, however, is not as effective as the stronginput in driving the postsynaptic cell. During conjunctivestimulation, these two inputs undergo a pairing in which theweak input is potentiated, that is, undergoes LTP and becomescapable of driving the postsynaptic cell. This is the principleof associativity. Neurons in the basolateral complex have beenshown to respond to both the CS and US. While the responseto the CS is perhaps not as strong as to that of the US (215),it is currently unclear how CS-US association during fear conditioningwould lead to LTP of the CS inputs. Interestingly, in a recentstudy, intracellular recordings maintained during the CS-USpairing revealed that plasticity of the CS occurs despite thelack of significant postsynaptic depolarization during theconjunctive stimulation (238). This result suggests that alternatemechanisms to classical LTP may be operating in the LA duringfear conditioning. Conditioning to different modalities hasbeen shown to require activation of basolateral neurons. Fearconditioning to one modality preserves the response to a differentmodality showing that the associated plasticity is restrictedto the input that has undergone conditioning (24). In vitrorecordings in acute brain slices prepared from fear-conditionedanimals show enhanced responses when inputs to LA neurons arestimulated (171). These results suggest that different modalitiesmust converge on all neurons and does not resolve how inputspecificity of conditioning is generated. These considerationsshow that a simple model involving plasticity of synapses madeby a given input in one structure cannot entirely account forthe complexity of the phenomenon of learning.
Answers to all your Biology Questions
