I want to discuss a very interesting new paper from Jim Knierim’s group at Johns Hopkins, which has the potential to be a major advance in our understanding of episodic memory. It also perhaps is the beginning of an answer to a question that has been in my mind for 20 years.
The paper is: Joseph D Monaco, Geeta Rao, Eric D Roth, James J Knierim. Attentive scanning behavior drives one-trial potentiation of hippocampal place fields. Nature Neuroscience, 2014.
Before I go into its content, let me explain why it caught my attention.
Twenty years ago I was a graduate student at the University of Arizona, with Bruce McNaughton as my supervisor. Matt Wilson had recently joined the lab as a postdoc, and had worked with Bruce and Casey Stengel to develop a system capable of recording neural activity using 12 tetrodes at once. With tetrodes implanted in the rat hippocampus, the system allowed populations as large as 100 neurons to be recorded from simultaneously—occasionally even larger—in rats that were awake and freely moving. Nobody had ever been able to record data of that sort before, and it was full of wonderful things, some of which were reported in two Science papers on which Matt was first author. At that time my main skills were in the area of data analysis, and Matt and Bruce very kindly gave me access to his data, to see if I could get anything out of it that would be useful to me (my main focus was the hippocampal theta rhythm).
For his first project, Matt wanted to examine the reaction of rat hippocampal cells to exposure to a novel environment. The apparatus he used was a rectangular box divided into two equal-sized parts (north and south) by a high wall running down the middle. The rats were trained to forage for randomly scattered food pellets. During training the wall was always in place, and the rats were exposed only to the north box. They were never even given a sight of the south box, or any clues at all about what was behind the south wall. After the rats were thoroughly familiarized with this task, a “hyperdrive” containing 12 tetrodes was implanted; the rats were allowed to recover; then they were allowed to familiarize themselves with foraging while the hyperdrive was in place and recording cables were attached. Meanwhile the electrodes were gradually moved downward through the brain until as many as possible of the tips were located in the CA1 sector of the hippocampus.
When the day finally arrived on which everything was set up as well as possible, Matt carried out his experiment. The recording session was divided into three parts: first for a period of 10 minutes the rat was placed in the north box and allowed to forage exactly as during training; then the center wall was lifted out, allowing the rat to see and enter the south area for the first time in its life; for the next 20 minutes it was allowed to explore and forage freely, moving back and forth between the two regions at will; finally the wall was put back into position, and the rat was allowed to spend another 10 minutes foraging in the north region.
The data from the experiment showed a number of striking things, including a notable difference in hippocampal activity levels between the north and south regions during the middle part of the session. I won’t try to recapitulate them; they are described in Matt’s 1993 Science paper. I will, however, relate the point that is relevant to the present story, which is that during all this the activity in the north region was largely unchanged. For cells that had place fields in the north box during the first 10 minutes, they retained those place fields for the middle 20 minutes and the final 10 minutes, with no noticeable changes.
There was, however, one cell that did something remarkable. Looking over the data, I saw that during the first 10 minutes, while the rat was doing the familiar task in the north box, this cell showed no activity whatsoever. Not one single action potential during the whole time. But during the few seconds in which the wall was lifted out, while the rat was frozen in place watching this happen, the cell fired a vigorous burst of action potentials. From that moment on, the cell showed a place field at that location. Each time the rat passed through the place from which the rat had observed the wall being removed, the cell fired a burst of action potentials. This continued not only during the middle 20 minutes (while the rat could move back and forth between north and south areas), but also during the final 10 minutes, in which the environment looked exactly the same as during the first 10 minutes.
This was extraordinarily interesting, because it could immediately be interpreted in terms of a very simple theory about the function of the hippocampus in episodic memory. The theory postulates that when the brain needs to store a memory trace, the hippocampus allocates a small group of neurons, chosen essentially at random, and links them (by strengthening synapses) to each other and to neurons in the cerebral cortex that represent features of the event. Every aspect of the cell I spotted seemed to fit perfectly with that mechanism.
Why had nobody ever reported anything like this before? Because hippocampal activity is very sparse. During active wakefulness, only on the order of 1% of hippocampal pyramidal cells are active at any given time. With single-neuron recording, the chances of having the good luck to spot a cell that participates in a memory trace are minimal. With populations of 100 neurons, though, the chances of finding one or two become reasonably decent.
I looked around in Matt’s data for more cells doing similar things, and found a couple more that were suggestive, but none that were nearly as striking as the first one.
Ever since then I have had my eyes open for further examples of that phenomenon. I didn’t think it would be easy to find relevant data, because it seemed necessary for the rat to experience something unusual enough to be worth memorizing, and for technical reasons the great majority of experiments involve animals repeating tasks on which they have been very thoroughly trained.
In my laboratory at the University of Pittsburgh I tried some preliminary experiments that I called “Surprise the Rat”, in which we exposed rats to various stimuli that seemed likely to surprise them, in the hope that we would see a fraction of cells developing place fields at the place where the rat was standing when the event happened. We did have some successes, most commonly when the event involved reconfiguring the environment spatially in some way—but we also saw numerous instances of cells developing new place fields at various locations for no obvious reason. It became obvious that in order to convincingly relate the new place fields to memory we would need to redo the experiment with stronger controls.
As it turns out, the new paper by Monaco et al. shows that we were working too hard. Apparently it isn’t necessary for the rat to experience something especially memorable—all that is necessary is for the rat to take a few moments to attentively scan the environment.
Joe Monaco’s experiment involved a type of task very commonly used in hippocampal studies—rats walking along a circular track foraging for scattered pellets. While the rats did this, they frequently paused to look around. Monaco and his colleagues were able to observe a number of instances in which this act of scanning the environment was associated with formation of a new place field by a hippocampal pyramidal cell. It looks like exactly the same sort of thing I saw in that cell of Matt Wilson’s so many years ago.
“We found that many cells that were previously silent would suddenly start firing during a specific head-scanning event,” Jim Knierim said (I am quoting from a Johns Hopkins press release). “On the very next lap around the track, many of these cells had a brand new place field at that exact same location and this place field remained usually for the rest of the laps. We believe that this new place field marks the site of the head scan and allows the brain to form a memory of what it was that the rat experienced during the head scan.”
Of course more work needs to be done, but I believe this is a tremendously important discovery. The nature of the “engram”—the trace left in the brain by an episodic memory—has been one of the central problems in neuroscience for almost a century. There is a reasonable chance that this discovery will turn out to be the key that unlocks the puzzle.