Memory loss is one of the most obvious and devastating symptoms of Alzheimer disease (AD). But despite advances in researchers’ understanding of the molecular basis for memory and synaptic plasticity, they still do not understand to what extent a given memory is woven into the fabric of our neural circuitry. When Pavlov rang his bell, which neurons in his dog’s brain sprang to attention, and how many where they? This question gets at the heart of the memory engram, the sum of all morphological and molecular changes that represent a given memory. Are engrams traced through a few neurons, some specific subsets, or the majority of neurons? In yesterday’s Sciencexpress, Roberto Malinow and colleagues show that in the amygdala at least, there is precious little redundancy when it comes to storing memory, as blocking plasticity in as few as 10 percent of the neurons significantly impairs memory formation.
Working with colleagues at New York’s Cold Spring Harbor Laboratories and New York University, Malinow came to this conclusion after studying the role of AMPA glutamate receptor subunits in learning. Previous work from this group (Hayashi et al., 2000) suggested that incorporation of GluR1, one of the AMPA receptor subunits, into synapses might be important for long-term potentiation (LTP), a phenomenon that ensures a neuron can be “trained” by repeated stimulation to become more efficient in its own synaptic response; LTP is required for learning and memory.
To measure the contribution of GluR1 to associative learning, first author Simon Rumpel and colleagues turned to the modern-day rodent equivalent of Pavlov’s experiment—fear conditioning, in which a mouse or rat learns to associate a mild electric foot shock with an audible tone. To measure how this type of learning affected glutamate receptors, the authors used a “plasticity tag.” The tag, a green fluorescent protein-coupled GluR1, serves as a homomeric glutamate receptor that is a better conductor than the natural receptor, which is made up of the two endogenous subunits, GluR1 and GluR2. The additional current (or rectification because it only works inwardly, not outwardly) evoked by the tag allowed the researchers to measure GluR1-linked increases in synaptic activity.
The authors found that when rat amygdalas were infected with a vector expressing the plasticity tag, the animals responded normally in the fear conditioning response: After a period of training, the rats froze in expectation of a foot shock whenever they heard the tone. To measure the GluR1 contribution to this response, the authors tested the electrophysiology of brain slices taken after training. They found that slices expressing the tag showed significantly more rectification (almost 20 percent more) than wild-type slices, indicating that the GFP-GluR1 was indeed being incorporated into synapses in response to the learning paradigm.
Because the plasticity tag is labeled with GFP, the authors were able to estimate how many neurons were transfected in addition to how many exhibited increased rectification. The data indicates that about one-third of the neurons in the amygdala undergo synaptic remodeling in response to the fear conditioning training.
But it was when the authors used a “plasticity block” tag that they could try to answer the question of how many neurons are essential for the learning process. The blocking tag consisted of a GFP coupled to the C-terminal tail of the GluR1 receptor subunit. This chimera acts to prevent incorporation of GluR1 into synapses and so behaves in a dominant-negative fashion. When Rumpel and colleagues transfected rat amydgalas with this construct, they found that the animals froze only about half as often after fear conditioning as did wild-type animals.
What percentage of the amygdala neurons are required for fear conditioning? Rumpel and colleagues found that if less than 10 percent of the neurons were infected with the plasticity blocker, then the rats still learned. Infection of 10-20 percent of neurons was sufficient to impair learning, whereby after training, these animals only responded to the tone about half the time. The authors conclude that “Perturbing plasticity in a small fraction of lateral amygdala neurons appears to be sufficient to reduce memory function, suggesting little robustness or redundancy.” They also suggest that memory may require coordinated changes, and that perturbing a few key plastic units, e.g., synapses, may corrupt the whole process, much like a single out-of-tune violin in a symphony. Whether the same principle applies to retrograde memory or not remains to be determined. But if a similar lack of redundancy marks remembrance of things well-learned, then it may take only a few neuronal lesions to elicit symptoms of dementia.—Tom Fagan
- Hayashi Y, Shi SH, Esteban JA, Piccini A, Poncer JC, Malinow R. Driving AMPA receptors into synapses by LTP and CaMKII: requirement for GluR1 and PDZ domain interaction. Science. 2000 Mar 24;287(5461):2262-7. PubMed.
- Rumpel S, LeDoux J, Zador A, Malinow R. Postsynaptic receptor trafficking underlying a form of associative learning. Science. 2005 Apr 1;308(5718):83-8. PubMed.