Attentional Modulation of Hippocampal Place Field Stability: What Can it Tell Us About the Brain?

Sean M. Montgomery
Spring, 2002

Dr. Kentros’ finding was one of serendipity. He was originally investigating the effect of a calcineurin knockout in place field stability, recording from hippocampal pyramidal cells in the mouse. Early in his investigation he found that the mice with which he was working had a robust endogenous drive to run around the testing arena and he therefore didn’t have to throw food randomly into the arena to ensure the animals would visit all locations in a session. After running a cohort of knockout and wild-type animals, Dr. Kentros found that not only was the place field stability of knockout mice very low, but the place field stability of wild-type mice was equally low. This finding was in stark contrast to a whole body of work with rats showing very stable place fields over many days. From this finding, Dr. Kentros decided to further explore the underpinnings of this discrepancy.

In order to formally explore this issue, Dr. Kentros designed an experiment in which animals were run in the arena in one of three conditions. In the first, ‘no task’ condition, animals were put in the environment and allowed to run for 30 minutes. In the second, ‘foraging’ condition, similar to the typical rat experiments, food was randomly thrown into the arena at periodic intervals for 30 minutes. In the final, ‘spatial’ condition, animals ran around the open arena, but every so often bright lights would turn on and loud noises would sound until the animal sat in an arbitrarily specified, but unmarked location in the arena. This task, similar to the water maze, requires that the animal use spatial cues to determine the location of the ‘safe spot’. The 30 minute (within session) and 24 hour (between session) stability of the place fields was determined over many days. The floor of the arena was changed after each session to reduce the effect of changing local cues. In order to quantify the results, Dr. Kentros calculated the average firing rate of each neuron for each pixel area of the arena (rate map) and correlated the rate maps between and across sessions. Because of slight variability of hippocampal firing a correlation of 0.5 using this method turns out to generally represent what would generally be judged by eye to be perfect stability.

After running animals in these three conditions, Dr. Kentros found that while most animals in the place condition learned the location of the safe spot over six days of training, other animals didn’t learn the location of the safe spot. Observing the animals’ behavior, there wasn’t a marked difference between performers and non-performers on day one, but on day six performers would run to the safe spot soon after the light and sound began, while the non-performers would exhibit frantic thigmotaxic running around the arena. Dr. Kentros thus divided the animals in the place condition into groups based on their performance. Animals were designated to the performer group if the mean escape time on day 6 was less than half the mean escape time on day 1.

The average path length and running speed were similar across all the groups. All groups also exhibited similar high stability within a session. However, the 24-hour stability of place fields differed markedly across groups. The no-task group had an average rate map correlation of 0.15 for consecutive days. The foraging group had an average rate map correlation of 0.28. In the spatial condition, the performers had an average rate map correlation of 0.5 (in fact, 100% stability), while the non-performers had only a 0.14 correlation. In order to show that animals were using the stable reference cues in the arena rather than floor cues that change from day to day, Dr. Kentros rotated the cylinder for a subset of the no-task animals and found that the place fields rotated with the cylinder. Dr. Kentros also noticed something interesting when he looked at the successive daily rate maps of no-task animals for which stable recordings could be maintained for long periods (up to three weeks). Specifically, he found that while the no-task group had a very low rate map correlation for consecutive days, cells seemed to have a few different place representations that would be randomly represented in any given session.

Dr. Kentros argued that these three conditions differed in the amount of attention that the animal used to perform the task. Accordingly he argued that the difference in stability of place fields can be explained by differences in attention. With this line of reasoning he set out to find an alternative way to manipulate attention. Looking up pharmacological agents thought to affect attention, he ended up with an extensive list from which he chose dopamine. By giving systemic injections of D1/D5 agonists and D1/D5 antagonists to a new set of animals, Dr. Kentros hoped to manipulate attention and observe the effects on place field stability.

Giving the dopamine agonist to animals in the no-task condition, Dr. Kentros found an increase in the 24-hour place field stability from a rate map correlation of 0.11 to 0.25. Giving the dopamine antagonist to animals in the foraging condition, he found a decrease in the 24-hour stability from a rate map correlation of 0.28 to 0.07. He did note, however, that the animals given dopamine antagonist were rather ‘messed up’.

I found the experiments described by Dr. Kentros quite intriguing. However, I think that to claim that attention is the root of the change in place field stability is a rather nebulous blanket statement. In general, attention is a poorly defined, overused term, but in this case, it is blatantly enigmatic. Because the term is so poorly defined, I think several people at the talk found themselves unable to argue with the speaker because any argument could be turned back around to say that it was attention and that was the point of the talk. At the end of the talk the speaker made a hand-waving gesture to the notion of following up this work with future work trying to work out whether this was ‘selective’ or ‘general’ attention. While these terms seem to cut the nebulum in half, it isn’t clear that these two half nebulas are any better to work with.

Regardless of what words you want to use, however, I think it’s a remarkable finding that different tasks can result in different levels of place field stability. Importantly, Dr. Kentros showed that the average running speed and path length didn’t differ between the animals to show that it wasn’t a non-specific effect of activity level. I think that because Dr. Kentros showed that rotating the cylinder leads to place field rotation, one can also rule out the uninteresting possibility that animals weren’t using the stable cues on the cylinder walls to generate place maps. This uninteresting possibility is also made less likely by the fact that animals with unstable place fields seemed to have place fields that seemed to vacsilate between several maps. With both of these controls, I think one can make the claim from these data that the internal state of the animal can affect the information represented in the hippocampus. While this isn’t a new idea, I think it is quite interesting.

In order to take this idea to the next level, though, I think that we need to learn something about how this ‘internal state’ is affecting the stability of representations in the hippocampus. As an approachable first step to understanding how, I think it might be useful to know where in the brain this change in processing is occurring. Specifically, is it occurring in the hippocampus itself or is it occurring in regions upstream from the hippocampus? It’s possible that the information reaching the hippocampus changes from day to day. This could be imagined to occur because of different ‘filtering’ of information as it is processed through the cortex. In this case, different internal states could be said to make the filtering process more or less regular over different sessions. It is also possible that very similar information reaches the hippocampus every day, but the hippocampus transforms this information in a slightly different manner on each day, leading to different maps. This type of behavior could be generated from a simple attractor mechanism in which on different days, slightly different starting parameters of the network lead to divergent outcomes. The different internal states in this case could be changing the stringency of the attractor network so that slight changes in starting parameters don’t change the outcome. Of course it’s possible that the internal state could be affecting both the processing in the cortex and the hippocampus.

So, how can we test these possibilities? One way would be to record from entorhinal cortex in the three conditions that Dr. Kentros used. If the representations in the entorhinal cortex change from day to day, it could be said that processing prior to the hippocampus is causing different information to reach the hippocampus on different days. Alternatively, if the representations in the entorhinal cortex were stable while representations in the hippocampus changed from day to day, it could be inferred that processing in the hippocampus is responsible for the differential stability.

With another experiment one could determine whether changes in hippocampal region processing were sufficient to change stability of hippocampal representations. At the 2001 SFN meeting, Brazhnik & Fox found that injecting picrotoxin into the medial septum selectively caused putative cholinergic cells (based on wave-shape characteristics) in the medial septum to increase firing. Recording simultaneously from neurons in CA1, the authors found that there was a general ‘sharpening’ of place fields. Specifically, the in-field and the center-of-field rate of firing increased, while out-of-field firing decreased. Using systemic scopolamine injections in addition to picrotoxin injections, the authors found no change in in-field or center-of-field firing, while out-of-field firing increased. Put in terms of an attractor network, it could be said that increasing ACh in the hippocampal region increases stringency of the outcome attractor state. By using picrotoxin injections into the medial septum in Kentros experiment, one could investigate whether increasing ACh into the hippocampal region was sufficient to stabilize place fields in the hippocampus even in the face of potentially somewhat variable inputs.

These relatively simple experiments could provide very interesting results. If cortical processing leads to changing inputs to the hippocampus on different days, it would be interesting, but it might be a dead end. However, if representation in the entorhinal cortex were stable while hippocampal representations were unstable, it would be an incredibly interesting finding that could be the launch point for a whole host of investigations. Similarly, if picrotoxin injection into the medial septum was found to stabilize otherwise unstable place fields, this could point to a functional role of ACh in the hippocampus and help to understand the dynamic computational properties of the hippocampus. Using these experiments as a starting point, one could look at the timing of the sources and sinks in the different regions of the hippocampus and relate these to the timing of unit firing in hippocampus and entorhinal cortex. Using these measurements in combination with behavioral and pharmacological manipulations, one may begin to better understand how the computational properties of the hippocampus contribute to the formation of stable representations and how these properties may be dynamically regulated.