 Described initially at the beginning of the 20th century
by Ivan Pavlov, classical conditioning involves the pairing of neutral
stimuli with aversive
or appetitive cues. This results in learning. Formally neutral stimuli
now predict salient events. One form of Pavlovian conditioning is fear
conditioning, the primary associative learning mechanism involved in aversive
emotional learning. It is through this process that we learn to be fearful
of people, animals, objects, and places. From a psychological perspective,
it is one of best understood types of learning in mammals because the stimulus
and response properties can be very carefully controlled. For these reasons,
Pavlovian fear conditioning has served as a powerful animal model of fear
and anxiety disorders including phobia, panic disorder, posttraumatic stress
disorder, and possibly many of the “neurotic” associations
of everyday life. If we can understand the underlying molecular basis for
how a previously innocuous stimulus leads to intense fear in an animal
model, the hope is that such an understanding will eventually lead to better
treatments in humans disabled by these crippling disorders.
Besides the obvious usefulness of this model system for understanding fear
and anxiety disorders, fear conditioning has received considerably more
attention in recent years for two reasons. Investigators are able to carefully
control the stimulus-response properties of fear conditioning in rodents.
This has led to very precise behavioral paradigms for studying learning
and memory. In addition, the ability to combine behavioral, anatomical,
and electrophysiological techniques, both in vitro and in vivo, has contributed
to an advanced understanding of the neural circuitry underlying this form
of learning. This combination of approaches has made fear conditioning
and its neural substrates among the best models for the understanding of
learning and memory across the cellular, neural systems, and organismal
levels of organization. In this brief review, we will first describe the
experimental paradigm involved in generating and measuring fear-conditioned
learning. We will then discuss the neural circuits and their cellular substrates
that are thought to mediate the learning and memory process.
In general, fear conditioning involves the pairing of a previously neutral
cue, which then becomes the conditioned stimulus (CS), with an aversive
cue, the unconditioned stimulus (US). In rodents, the US is typically given
in the form of a mild foot-shock (Fig.
1A). Although any form of sensory
cue may be used, the CS is typically a discrete tone, light, or odor. In
fact, rodents can easily discriminate among numerous discrete cues in this
form of conditioning. After several pairings of the CS and US, the animal
will display a conditioned response to the CS alone which is indicative
of a central state of fear. Any time after this training, from hours to
months, a fear response can be elicited by exposure to the CS alone.
The most commonly used measures of conditioned fear are the freezing response
and the fear-potentiated acoustic startle reflex. With both of these measures,
the animal is placed in a neutral environment and the CS is presented in
the absence of the US. Freezing is measured as immobility during the presentation
of the CS. The percentage of time spent freezing is found to be significantly
greater in the presence than in the absence of a fear-conditioned CS. Alternatively,
the acoustic startle reflex is elicited by presenting the animal with a
short noise burst which serves to elicit a reproducible startle response
(Fig. 1A). Measurement of the startle reflex can be entirely automated
such that a computer may be used to record the magnitude of the startle
response as well as to control the presentations of the CS and US. In the
presence of the CS (e.g., light, odor), the magnitude of the startle elicited
by the noise burst is reproducibly greater than it is in the absence of
the CS, and this difference is referred to as fear-potentiated startle.
The amygdala is the primary brain region involved in fear-conditioned learning.
However, “the amygdala” in fact refers to a number of functionally
separate nuclei that are anatomically grouped together. The learned state
of fear involves activation of the basolateral nucleus of the amygdala
which in turn activates the central nucleus of the amygdala (Fig.
1B).
As its name implies, the central nucleus is a “hub” that serves
to initiate the full fear response whether it is activated physiologically
via the basolateral nucleus or experimentally through electrical stimulation.
The “fear response” is generated through the hardwired neural
connections that exist between the central nucleus and a number of other
neural pathways. For example, activation of various midbrain nuclei by
the central amygdala results in freezing, potentiation of reflexes such
as the acoustic startle reflex, and increased respiration. Projections
to the lateral hypothalamus activate the sympathetic nervous system leading
to cardiovascular effects, pupil dilation, and increased sweating. Activation
of the paraventricular nucleus of the hypothalamus activates the glucocorticoid
response. Lesions of these individual brain regions that are downstream
of the central nucleus serve to block specific aspects of the fear response,
whereas ablation of the central nucleus itself blocks the entire fear response.
In addition to the well-understood circuitry underlying the outputs of
fear response, the sensory inputs representing the CS and US pathways have
also been studied. Auditory, visual, and somatosensory projections arrive
at the basolateral amygdala (BLA) through both thalamic and cortical pathways.
Olfactory pathways, on the other hand, take a more direct route from the
olfactory bulb through the piriform cortex and directly into the amygdala.
In all cases, the pathways representing the aversive US and the previously
neutral CS converge in the BLA. It is within the BLA that the most critical
cellular processes underlying associative learning are thought to occur.
There is a large amount of data supporting a role for the BLA in initiating
associative fear learning. Chemical and electrical lesions of the BLA have
been shown to block new fear learning. Furthermore, the principal electrophysiological
measure of learning, long-term potentiation (LTP), appears to occur between
the neurons projecting to and the target neurons within the BLA in brain
slices. Electrical stimulation of neurons within the BLA before and after
fear conditioning in freely moving rats has shown that there is potentiation
of the pathway mediating many of the sensory connections to neurons within
the BLA. Taken together, these data suggest that fear conditioning may
result in an alteration in the functional connectivity between sensory
pathways and the BLA, such that future presentation of the CS alone is
now sufficient to activate the central nucleus.
One model for the circuitry changes underlying fear conditioning proposes
that when the associative CS-US pairing occurs, strong firing of neuronal
projections that mediate the aversive stimuli are paired with weaker firing
representing the neutral stimuli. This coincident firing leads to a number
of events that result in the strengthening of the BLA connections that
represent the neutral stimuli. The strengthening of BLA connections could
occur through at least two mechanisms: either increasing the efficiency
of presynaptic neurotransmitter release, or increasing the sensitivity
of postsynaptic terminals through structural modifications at the synapse.
An active area of research is aimed at determining whether the alteration
in neuronal connectivity is limited to changes in synaptic efficacy, to
actual structural change, or to both mechanisms.
The cellular events that mediate the associative learning between axons
representing the CS and US pathways are also beginning to be understood
(Fig. 1C). The glutamatergic N-methyl-D-aspartate (NMDA) receptors are
thought to perform the primary function of mediating associative learning
at the level of the synapse. Blockade of these receptors has been shown
to block the acquisition of new fear associations, but not to block the
expression of previously learned fear responses. As with other memories,
fear conditioning appears to have short-term and long-term phases of consolidation.
Short-term memory formation appears to be dependent on a number of receptors
including NMDA receptors, voltage-gated calcium channels, and norepinephrine
receptors. In contrast, the transition from short-term to long-term representation
appears to require the addition of new mRNAs and protein synthesis.
The initial cellular events that occur following the coincident association
of the CS and US involve an influx of calcium into the postsynaptic neuron.
This occurs when the NMDA receptor is activated through glutamatergic release
as well as when the voltage-gated calcium channel is activated by membrane
depolarization. In addition, other neurotransmitters such as norepinephrine
and growth factors such as brain-derived neurotrophic factor (BDNF) have
been implicated in the activation of the neuronal events mediating memory
formation. A number of intracellular factors are induced by receptor activation
during CS-US pairing including kinases, phosphatases, and transcription
factors. Some of the better understood pathways are illustrated in Figure
1C.
A central point is that the signaling cascades that are activated lead
ultimately to the phosphorylation of transcription factors within the nucleus.
It is thought that these early events initiate long-term memory formation
by facilitating a later wave of mRNA and protein synthesis, involving the
production of other transcription factors, synaptic clustering molecules,
and molecules that mediate axonal or dendritic structural plasticity. Through
this cascade of events—occurring over time courses ranging from milliseconds
to days—the associative pairing of the CS and US leads to alterations
in synaptic efficacy. This results in a stronger activation of the basolateral
and central amygdala following presentation of a seemingly neutral stimulus
(the CS) than occurs before conditioning.
We have outlined the role of the BLA as a site for convergence of CS and
US representations and as a likely site for mediating the cellular and
molecular processes underlying fear learning. Much progress has been made
as numerous investigators have focused on this brain area using fear conditioning
as a behavioral tool. Undoubtedly, this approach will continue to provide
an excellent model of learning as the amygdala appears to serve as an integrator
between input associations and output behavior.
The amygdala is not the only region in the brain undergoing changes when
fear conditioning takes place. Accumulating evidence suggests that salient
learning alters other brain areas such as limbic, cortical, and subcortical
regions. Furthermore, there is a large body of literature devoted to the
role of the amygdala in activating the norepinephrine system when a fearful
event occurs, and it is clear that this and other neurotransmitters participate
in consolidating memories in multiple brain regions.
In summary, Pavlovian fear conditioning is a simple model of associative
learning. It serves as a powerful experimental tool because of its position
at the interface between behavior and neurobiology. Behaviorally, precise
experimental control is possible through subtle manipulation of sensory
cues. Neuro-biologically, the amygdala is an important mediator of convergent
sensory associations and a source of divergent outputs related to learned
fear. Numerous investigators have used this circuitry to show that fear
conditioning may be dependent on LTP and is mediated by both short-term
and long-term consolidation processes. Finally, rapid progress in understanding
the requisite cellular and molecular mechanisms has made fear conditioning
one of the most useful models for understanding learning from major perspectives:
molecular, neural systems, and behavioral.
Such progress in understanding the basis of a rudimentary form of learning
and memory may prove to be of great benefit to psychiatry. Other forms
of learning are also fascinating, but there may be no other learning process
more relevant to psychiatric disorders than amygdala-dependent fear learning.
A full understanding of these mechanisms will surely have direct implications
for clarifying the pathophysiology of and developing treatments for anxiety,
phobia, panic, and posttraumatic stress disorders.
top of page
|
