If long-term memory involves transcription and therefore the nucleus, an organelle [that is] in contact with every synapse a neuron makes, and a neuron can make up to a 1,000 synapses, does that mean every time you turn on the long-term process, and therefore transcription, you change the synaptic strength of every single synapse? That long-term memory is cell-wide? With short-term memory we're able to show it's synapse-specific. So clearly this is something we needed to explore.

Now fortunately, a number of years earlier, Sam Schacher, when he was in my lab, took advantage of a wonderful discovery that had been made by Arnie Kriegstein, another spectacular MD-PhD student.

I was very much influenced by Stanley Crain on the one hand, and by Gerry Fischbach, a relative of mine, who was getting beautiful results with the dissociated cell culture of motor neurons and muscle. And I wondered whether we could get the gill-withdrawal reflex, a sensory neuron and a motor neuron and a serotonergic cell, in culture. So for that, we needed young animals, and we didn't know how to get young animals because Aplysia has a step in metamorphosis, and no one had been able to trap that in a laboratory. So Arnie Kriegstein went to Woods Hole, and he realized that metamorphosis is triggered by a particular species of seaweed called Laurencia pacifica. And he found it, and he saw the metamorphosis in front of his eyes. He was able to get baby Aplysia in large numbers and that revolutionized the study of learning and memory in Aplysia, because now Eric Proshansky working with Sam Schacher was able to dissect out individual cells, put them in culture, and they survived perfectly well.

Actually we were very much influenced at the time by John Nicholls who had developed dissociated cell culture for leech cells, and we went out to visit him at Cold Spring Harbor, and he showed us how he dissected it. Turned out in Aplysia the cells are much larger, and you don't need such fancy techniques, but anyway, we were stimulated to do this. And we could show that you put a sensory neuron together with a motor neuron, and you put a serotonergic cell in, the serotonergic cell produces very nice facilitation. Moreover, you don't need the serotonergic cell, you can puff on serotonin. One puff of serotonin, short-term facilitation. Five puffs, long-term facilitation. Growth of new synaptic connections. Blocked by inhibitors of CREB. We could repeat the whole thing. Beautiful.

So Kelsey Martin comes into the lab, and she takes this culture system and she revolutionizes it. She takes a single sensory neuron and she cultures it so it has a bifurcating branch that ends on two separate motor neurons. So you have a single sensory neuron bifurcating axons ending on two motor neurons. So she puffs on one puff of serotonin here, nothing there. You get short-term facilitation here, nothing there. Five puffs of serotonin here, nothing there. Long-term facilitation here, nothing there. That's interesting. So despite the fact it involves CREB, apparently, this can be synapse-specific. So she asked, are we sure CREB is involved? She does again the same thing, five puffs here. She blocks CREB, she blocks the long-term facilitation. So clearly a signal has gone back to the nucleus, activating CREB, and is going back to the initial site. But is that the only place it's going to or is it going to all sites and all we lack is a marking signal? So she said, well let's see whether a mark will do it. She gave five pulses of serotonin here, and she gave a single pulse here. Single pulse produced facilitation lasting minutes. But if you gave that single pulse together with the five pulses here, she captured long term-facilitation, in reduced form, only 50%—but nonetheless, persistent, at the other branch.

So clearly gene products were shipped all over, you needed a marking signal. So that raised the question, what is the nature of the marking signal? Well, she began to explore it. She gave five pulses of serotonin here, together with an inhibitor of PKA. She blocked everything. She repeated the experiment, and she realized that one thing that you have at the synapses is the machinery for local protein synthesis. So maybe that has something to do with it. So she applied five pulses of serotonin here, with an inhibitor of protein synthesis. And she found she produced a long-term facilitation. It persisted for 24 hours. She had the growth of new synaptic connections, but it collapsed.

So this was the first evidence that there were wheels within wheels. That long-term memory has sub-steps. There's an initiation of long-term memory, and there's a maintenance process. The maintenance process is separate, it requires local protein synthesis. Wow! Just fabulous.

So this raised the question: what's the nature of that maintenance process?