Fire together, wire together: building a memory with long-term potentiation

When an axon of cell A is near enough to excite a cell B and repeatedly or persistently takes part in firing it, some growth process or metabolic change takes place in one or both cells such that A's efficiency, as one of the cells firing B, is increased. – Donald Hebb (Canadian!), Organization of Behaviour, 1949. Image credit: theratchet.c

When an axon of cell A is near enough to excite a cell B and repeatedly or persistently takes part in firing it, some growth process or metabolic change takes place in one or both cells such that A's efficiency, as one of the cells firing B, is increased. – Donald Hebb (Canadian!), Organization of Behaviour, 1949. Image credit: theratchet.c

Scientists have long assumed that our memories are stored in the chatter between connected neurons. Novel learning experiences tweak the strength of select synapses, and these changes in transmission efficiency allow us to encode and remember new memories. These days “synaptic plasticity” is almost synonymous with “learning and memory”, yet the exact biological nature of the phenomenon remains elusive.

One dominant candidate is long-term potentiation (LTP), a type of synaptic plasticity induced experimentally in brain slices or animals. In the mid-1960s, a Norwegian neuroscientist named Terje Lomo discovered LTP by chance while studying synaptic transmission in rabbits. Armed with a pair of electrodes, he used one to deliver a burst of high-frequency electrical pulses to neuronal fibres leading to the hippocampus, and the other to simultaneously record changes in the activity of downstream hippocampal neurons. Together with his collaborator Tim Bliss, the team noticed that after the train of stimulation, a small burst of activity in one neuron would subsequently trigger a much larger synaptic response in its partner. In other words, the signaling between them became more efficient. This “potentiation” of synaptic strength lasted for hours, thus earning the phenomenon its name.

The field rapidly accepted LTP as a potential molecular substrate of memory. They had plenty of cause: LTP is prominent in the hippocampus and other brain circuits crucial for memory; one stimulation protocol that induces LTP mimics a type of brain wave called theta rhythms that naturally occur in the hippocampus; drugs or genetic manipulations that block LTP usually (but not always) generate some sort of memory deficit. Yet a key piece of evidence was missing: can artificially induced LTP in animals produce a fake memory?

A team of neuroscientists set out to do just that.

They implanted light-responsive proteins into parts of the amygdala of rats that receive auditory information. The amygdala is an almond-shaped structure involved in forming memories of high emotional value. By shining light on these proteins through an optic fibre, the scientists can rapidly switch the neurons on and off.

The researchers trained the rats on an auditory fear conditioning task, but with a modern twist. Instead of repeatedly pairing a tone with a shock (the classical version, a below), they used light to artificially stimulate amygdala neurons in place of the tone (b). When researchers re-exposed the animals to the optical stimuli after conditioning, the rats froze in fear ("paired condition" in graph below). Without ever consciously experiencing the light pulses, the animals had learned to associate them with pain! Since researchers specifically activated neurons that process auditory information, in a sense the animals had learned to fear a “tone” that was completely in their heads. When researchers examined the rats’ brain later, they found telltale signs of LTP within the amygdala.

Left (a): classical tone-shock conditioning; Right (b) light-shock conditioning. When the tone and light was NOT simultaneously delivered (unpaired), the animals did not freeze (measured as a decrease in lever presses). When they came together (paired), the rats froze.

Left (a): classical tone-shock conditioning; Right (b) light-shock conditioning. When the tone and light was NOT simultaneously delivered (unpaired), the animals did not freeze (measured as a decrease in lever presses). When they came together (paired), the rats froze.

Crucially, the same light pulses did not trigger LTP when delivered without the foot shocks nor when their deliveries were staggered in time ("unpaired" in graph above). It is the association between the two events that correlates with the presence of LTP in the brain.

If LTP is the biological basis of the fear memory, then reversing it should “erase” the memory. Using a different pattern of light stimulation, the researchers weakened the synaptic connections back to baseline – and the rats were no longer afraid of the light pulses (b below). (The authors called this process long-term depression, or LTD, but it should be depotentiation. Although both processes weaken the connection between synapses and share many signaling molecules, they do differ in some respects.) Memory “deletion” was reversible: when researchers gave the animals a set of LTP-promoting light pulses, the fear returned (c below).

Playing with memory like a yo-yo.

Playing with memory like a yo-yo.

What were those LTP-promoting light pulses doing in the brain? Were they strengthening transmission between synapses involved in the original memory trace? Or were they simply amping up random synaptic connections in the amygdala and triggering fear by chance? A quick control experiment showed that the LTP-promoting light pulses could only reactivate the fear association after it’s been learned. This strongly suggests that the pulses were re-potentiating the same cohort of synapses that stored the initial memory.

Light-induced LTP or LTD is obviously not natural, but do they at least resemble their biological counterparts? To answer this question, researchers tried to artificially inhibit a naturally learned fear memory. They trained a new group of rats on the standard tone-shock conditioning task (a below). Once the memory was established, they then depotentiated the amygdala with light – and the fear disappeared (b)! Considering that only a handful of neurons are required for storing a memory, this suggests that the depotentiating light pulses managed to selectively reverse LTP at synapses involved in the initial fear memory trace. (But really, were the rats so bored that only those synapses were potentiated? What other things did the light wipe away?) Finally, LTP-promoting light pulses could not bring back a fear memory that had undergone extinction (e,f). This makes sense, since extinction training generally produces a new memory rather than erase the original (but see here).

The lower the red line, the stronger the fear memory. a,b: Light-induced depotentiation ("LTD") sort of inactivated the fear memory. e,f: Light-induced LTP did not bring back a memory after extinction.

This is quite the ambitious study and a great first step. There are many questions that still need to be answered before crowning LTP as a cellular mechanism of memory. For example, memories lasts a lifetime, but LTP generally decays within hours, at most a week (although memories do get moved out of the hippocampus for storage). Memories are encoded at select synapses, yet LTP at one site may trigger LTP at others. How much overlap is there in the cellular pathways that produce synaptic changes during memory encoding and during light-induced LTP? And finally, blocking LTP does not always affect learning and memory, so what other forms of synaptic plasticity are at play?

Donald Hebb's legacy continues half a century after the Canadian psychologist first postulated the synaptic plasticity model of memory. Hats off to Hebb, and HAPPY CANADA DAY!


Nabavi S, Fox R, Proulx CD, Lin JY, Tsien RY, & Malinow R (2014). Engineering a memory with LTD and LTP. Nature PMID: 24896183