Eppendorf and Science Prize

An Uninstall Function for Fear Memory

Roger Clem

Memory is the essence of human identity. For some, however, a painful past is like a debilitating emotional storm lurking behind the most insignificant of reminders. Coping in the aftermath of trauma can be aided by pharmacology and psychological therapy, but these approaches leave underlying emotion subject to resurface when least expected. The question of how to permanently silence traumatic memories, however, has long eluded clinicians. In my postdoctoral work under Rick Huganir, I made a critical advance toward this goal.

Neuroscientists have localized fearful memories of trauma to an almond-shaped structure, the amygdala, in the brain’s limbic system. Sensory information relayed to the amygdala plays an instructive role in memory formation (1, 2). By way of neural plasticity, terror can be permanently linked with the sights, sounds, and smells of past events (3). Although traumatic memories can be quite elaborate, their essential features can be modeled in rodents by classical fear conditioning. In this form of associative learning, a neutral cue, such as nonthreatening tone, is presented to the animal together with an aversive stimulus, usually a mild shock. Rodents quickly learn that the tone predicts the shock, such that the tone later elicits a quantifiable and stereotyped “freezing” response.

Previous experiments suggest that this fear conditioning results from enhanced communication between auditory neurons and amygdala neurons (4, 5), a process referred to as synaptic strengthening. Thus, to build fear memories the brain follows one basic rule: amplify the amygdala reaction to predictors of harm. By obtaining electrical recordings from amygdala neurons, I found that this synaptic strengthening was mediated by increased function of AMPA receptors (AMPARs), which bind glutamate released by auditory neurons. By dialing up AMPAR activity, fear conditioning allows the amygdala to signal imminent threat. But how can we dial this activity back down, if at all? Luckily, this was not the end of the story.

While enhancing AMPAR activity, fear conditioning also changed the composition of these receptors, which are assembled from combinations of the related proteins, or subunits, GluA1-4. While in naive animals GluA2-subunit-containing AMPARs predominated, these were transiently replaced after fear conditioning with GluA2-lacking receptors, before a return to the original subunit configuration after 2 to 3 days. Experience-dependent trafficking of GluA2-lacking receptors has been observed elsewhere (69), but how these receptors modify neural function has been unclear. Because AMPAR trafficking exhibits subunit-dependence, one possibility was that GluA2-lacking receptors are more efficiently inserted or eliminated from synapses.

Indeed, I found that by delivering repetitive electrode stimulation, amygdala neurons could be induced to selectively and completely remove GluA2-lacking receptors, leaving synaptic strength permanently reduced. This removal required activation of metabotropic glutamate receptor 1 and N-methyl-D-aspartate (NMDA)—type glutamate receptors. Because this protocol did not affect GluA2-containing receptors, it was much less effective at dampening transmission beyond a few days after fear conditioning. Therefore, during a brief window after fear conditioning, stimulation can remove GluA2-lacking receptors to effectively negate synaptic strengthening that underlies memory. If this mechanism could be harnessed therapeutically, I thought, traumatic fear memories could be permanently alleviated.

While electrodes are too invasive for human subjects, amygdala inputs can instead be activated behaviorally. In fact, a common strategy for attenuating fear conditioning, as well as posttraumatic stress disorder, entails repeated exposure to the fear-eliciting stimulus (10). Exposure-based therapy, as well as its rodent equivalent, extinction training, merely suppresses fear in most cases, allowing it to resurface later. In contrast, a more powerful approach to inhibiting fear takes advantage of the momentary instability of memories after they are retrieved or accessed. Administration of amnesic drugs at this time can permanently disrupt memory (11). While the toxicity of these drugs precludes their therapeutic use, intriguing new work suggests that memories can also be effectively erased when retrieval is shortly followed by extinction training (12, 13). If a modified extinction protocol can erase fear memories, what molecular mechanisms need to be engaged? I hypothesized that GluA2-lacking AMPARs were part of the answer.

Armed with a tailored extinction protocol, I examined whether emotional memory could be eliminated from fear-conditioned mice. To confirm that fear was not simply masked by an inhibitory process, I challenged extinguished mice with a multitude of triggers, all of which failed to elicit fear. In mice subjected to erasure, GluA2-lacking AMPARs were removed from synapses, and synaptic strength was reduced. These data confirmed that trafficking of GluA2-lacking receptors from synapses contributed to memory erasure.

While erasure was quite easily generated during the time when GluA2-lacking receptors were abundant (1 day after fear conditioning), a gradual return to GluA2-containing receptors within 1 week after training suggested that fear memories may grow less prone to erasure over time. To examine this possibility, I repeated extinction experiments at 1 week after fear conditioning. Under this condition, fear memories strongly resisted erasure. Thus, changes in AMPAR subunit composition mediate the maturation of fear memory from malleable to stable following its acquisition.

Why do GluA2-lacking receptors accumulate during an early phase of memory, and do they alone determine its malleability? Through the use of genetic mutants, I determined that GluA2-lacking-receptor accumulation depended on phosphorylation of a second AMPAR subunit, GluA1, at the protein kinase A target residue serine-845. When this event was blocked, GluA2-lacking receptors failed to accumulate at synapses, and as a result, neither artificial stimulation nor extinction training could reduce amygdala transmission. These mice were therefore recalcitrant to our fear-erasure protocol and relapsed when reminded of their previous trauma.

Regrettably, conventional therapies for fear and anxiety disorders succeed at hiding or perhaps muffling emotion, only to find it has later returned with a vengeance. In harnessing the brain’s own self-editing capabilities, I have explained how troubled minds can be forever cleansed of fearful memories. As my results show, a provisional window of vulnerability defined by GluA2-lacking AMPARs allows these memories to be modified prior to their permanent storage and could be a useful point for intervention following emotional trauma. Tweaked by pharmacology, these mechanisms can potentially ensure that patients never face the pain of trauma again.

References and Notes

  • 1. J. E. LeDoux, A. Sakaguchi, D. J. Reis, J. Neurosci. 4, 683 (1984).
  • 2. S. Campeau, M. Davis, J. Neurosci. 15, 2312 (1995).
  • 3. H.-C. Pape, D. Pare, Physiol. Rev. 90, 419 (2010).
  • 4. M. T. Rogan, U. V. Staubli, J. E. LeDoux, Nature  390, 604 (1997).
  • 5. M. T. Rogan, J. E. LeDoux, Neuron 15, 127 (1995).
  • 6. R. L. Clem, A. Barth, Neuron 49, 663 (2006).
  • 7. K. L. Conrad et al., Nature 454, 118 (2008).
  • 8. C. Bellone, C. Luscher, Nat. Neurosci. 9, 636 (2006).
  • 9. R. L. Clem, V. Anggono, R. L. Huganir, J. Neurosci.  30, 6360 (2010).
  • 10. R. J. McNally, Clin. Psychol. Rev. 27, 750 (2007).
  • 11. K. Nader, O. Hardt, Nat. Rev. Neurosci. 10, 224 (2009).
  • 12. M.-H. Monfils, K. K. Cowansage, E. Klann, J. E. LeDoux, Science 324, 951 (2009).
  • 13. D. Schiller et al., Nature  463, 49 (2010).