Thursday, May 11, 2006

Now where was I?   posted by Coffee Mug @ 5/11/2006 11:23:00 PM

I put my thang down, flip it, and reverse it. – Missy Elliot

I have been invited on as an author at Gene Expression, so this post will go up at The Genius and at GNXP. To GNXPers: Hello. I'm a grad student interested broadly in neuroscience and genetics. Hopefully I'll be adding signal and not noise. This post is about an electrophysiological phenomenon in the hippocampus that may play a role in locking in lasting memories of event sequences.

Sharp-wave ripples in memory consolidation are sooo hot right now. I have been following the leads in this Wendy Suzuki mini-review to some really interesting multi-cellular recording studies in the hippocampus. She focuses on three recent papers, one of which I discussed here but hope to revisit. The most straightforward paper of the three is this report from Foster and Wilson. They show that cells in the hippocampus replay firing patterns that represent the places a rat has just been, but they do it in reverse. Another good avenue into the paper is this commentary by Colgin and Moser.

To get the paper you have to get a grip on sharp-wave ripples and place cells. To get the theoretical issues you probably need to know more about dopamine and its role in reward learning. I’ve discussed place cells at some length here, here, and here, but for the uninitiated here are the basics.

Place cells are found in the hippocampus and closely related areas (such as the part of the cortex that feeds directly into the hippocampus). They have place fields, just like cells in the visual processing system have receptive fields. They fire more as the rat moves toward the spatial location that is at the center of the place field. There are enough place cells with different place fields to make up a full-coverage map of a rat’s immediate environment. When the rat is moved to a new environment a new configuration of place cells comes online and represents a map of that environment. Observations in this vein led to a theory of the hippocampus as a cognitive mapping system espoused by John O’Keefe and Lynn Nadel in a book called The Hippocampus as a Cognitive Map in 1978, which I was amazed to find that you can download the full-text of for free here and it’s only 3.3 MB.

Sharp-wave ripples (SWRs), I am only beginning to become friends with, but they are super-interesting. So when you are recording in the extracellular space in most layers of the hippocampus there are network-level potential sweeps that occur in different behavioral states. An example that I’ve discussed before is theta oscillation, in which the extracellular potential fluctuates at a rate of ~6-12 Hz or an oscillation about every 100 milliseconds. Theta oscillations usually happen while a rat is walking around and exploring. SWRs are measured the same way, but they are much higher frequency and usually happen at times when theta is not happening, for instance during consummatory behaviors (chowing down), waking immobility, and during slow-wave sleep. Some of the most compelling research with regard to SWRs has shown that patterns of place cell activity that occured during the day are recapitulated in miniature during SWRs in slow-wave sleep, and that these SWRs are coordinated with another electrophysiological phenomenon called spindles in the neocortex. This latter finding is very interesting in terms of the standard systems consolidation view of memory, wherein memories intially stored in the hippocampus are moved over time to permanent cortical storage sites.

Foster and Wilson are not content with replay during sleep. They recorded from pyramidal cells in the CA1 region of the hippocampus using electrode arrays of 17-18 tetrodes (I'm not sure if you need one tetrode for each cell of if they get 72 cells out of this deal) as rats ran back and forth on a linear track to retrieve goodies at either end. First they show a set of place cells that represent the whole journey (i.e. one fires in Los Angeles, the next in Las Vegas, the next in Flagstaff (yes, we're going to Amarillo)). But once the rat gets to Amarillo and chills out, the all the cells spike together...well not exactly together. They fire at a greatly compressed timescale in reverse order with regard to the recent journey. It's like the rat is thinking back over it's trip starting with the most recent event.

Several of these reverse replay events can happen per resting period, and they happen during SWRs. So somehow this fast wave of activity sweeping across the hippocampus must be working its way backwards from the most recently spiking cell assembly to the most remote. Now that I'm thinking about it, I wonder if the spikes are setting up some sort of slow-decaying process that makes a neuron more reactive, like a release of intracellular calcium stores or maybe a signal transduction event leads to increase of current flow through ion channels or something. Then the cells most recently activated would be closest to threshold when the SWR activity shoots through.

The authors actually propose something quite close to the above now that I look back at their model. Although they don't bother with a fancy mechanism, they just assume that recently activated cells are at a higher level of excitation. The reason why I skipped over that simpler idea is because it's my impression that cells that have just fired are hyper-polarized (refractory period) and less likely to fire, so I needed a process that didn't just rely on action potential mechanics. It's probably best to assume that they took this into account since they are smarty-pantses.

We get some intriguing theoretical stuff when the authors attempt to explain why this reverse replay would be useful. It goes something like this:
  1. Dopamine is thought to reflect an error-signal when reward-prediction is off. This can work both ways, but in this case the rat is supposed to be under-expecting reward.
  2. When the rat gets a tasty reward at the end of its trip (that is something more wonderful that the rat had been expecting), there is a fast-spiking, slow-decaying dopamine input into the hippocampus.
  3. The apex of the dopamine spike coincides with the place cell firing that is closest to the reward location and then gradually falls off as place cells further away fire.
  4. This creates a reward gradient that can motivate and guide goal-seeking behavior in the future as the rat only has to move up the gradient to the final glorious dopamine peak spot where there is a delicious chocolate sprinkle.

There are some good reasons to believe in this model. The authors note that other (non-dopaminergic) parts of reward circuitry can fire in synch with SWRs, and we know that in some cases the presence of dopamine determines whether or not activity-induced changes in the neural circuitry actually stick. The only problem I have is that I have a hunch that these reverse replay events might happen even if there wasn't a reward. The authors report that several of these SWRs can happen per stopping period, and the rat isn't getting a new treat for each one. Sooo why are they happening in the absence of reward? Maybe you can just explain the phenomenon in terms of cellular excitability in reverse order, and you don't need to tack on the reward issue to explain the phenomenon but it does add an intriguing twist.

Also, I was thinking about the idea that these are in reverse and the ones during sleep are forward. I haven't read the papers about the sleepy-time ones, but I wonder if its worth looking for a slow reverse replay prior to the SWR-forward play. Seems like you might find something like that if the reason SWRs activate cells in a particular order is because of their recent spiking history.


Suzuki WA. 2006. Encoding new episodes and making them stick. Neuron 50:19-21.

Foster DJ. Wilson MA. 2006. Reverse replay of behavioural sequences in hippocampal place cells during the awake state. Nature 440:680-683.

O'Keefe J. Nadel L. 1978. The hippocampus as a cognitive map. Clarendon, London.