Using Place Cells in Goal-Directed Behaviour: Rodent Electrophysiology

Francois Windels

Theme 2A project focuses on how contextual information from the hippocampus is used in learnt and newly acquired behaviour. The underlying hypothesis is that the hippocampus provides both contextual and positional information to other brain regions like the amygdala, where it can be associated with other relevant information to form new memories. We use deep brain moveable multi-wire electrodes to record neuronal assembly activity over days or weeks (figure 1). Electrical activity is screened over several days and electrode placement adjusted accordingly to ensure that the largest number of neurons can be recorded with a high signal to noise ratio.

The neuronal mechanisms underlying associative learning in the amygdala were first studied using non-spatial tasks to establish a reference framework that can be compared with published results. An auditory stimuli (CS+) was repeatedly paired with an aversive shock (US), while a non paired auditory tone (CS-) was also played to provide a baseline response to auditory stimuli throughout the experiment (figure 2A). Single neuron responses to the acquisition of this new association are shown on Figure 2B and 2C.

The dynamic response of individual cells to each pairing was also studied using a statistical model (state-space general linear model, SS-GLM, Czanner et al. 2004), which accounts for the pre-stimuli spiking history and pairing to pairing changes (Figure 3).

Figure 1

Figure 1. Individual neurons in the basolateral amygdala were identified from tetrode recordings (left). A. Spike clustering of a single tetrode. Representative 2D projections comparing peak-to-trough (amplitude; A) and voltage at a specified time (V) between tetrode channels are shown. Each colour represents a different BLA neuron. B, C, and D are spike waveforms of clustered spikes in A. Waveforms recorded from three channels from a single tetrode are shown. The smoothed lines on the right are the average waveforms. (Right) Electron microscopy picture of a 4 twisted wire electrode (scale bar 10 microns).

Figure 2

Figure 2. Percentage of freezing time (A) in the 3 phases of an auditory fear conditioning protocol (CS white noise or 7.5kHz tone, US: 0.5mA; 0.5s; habituation 12 presentations, acquisition 7 presentations, test 10 presentations one day after). Raster plot (B) of all acquisition trial for CS- (upper panel) and CS+ (lower panel) of one experiment, peristimulus time histogram shown under raster plots show the delayed response to the US. CS duration are represented by the 2 blue lines. Peristimulus time histogram (C) z-scored, aligned to the onset of the CS for the acquisition phase of the conditioning (CS-, upper trace; CS+, lower trace).

Figure 3. Spike history effect; SS-GLM estimates of the effect of recent spike history on current spike rate for A. CS+ trials, and B. CS- trials. The black line is the SS-GLM estimate. Grey bars are 95% CI. Positive modulation of the current spike rate is seen with history effect > 1; negative modulation of spike rate is seen with history effect < 1. The spike history in the preceding 11-15 ms positively modulated the current spike rate (95% CI: 1.004-1.483). Learning Dynamic. Change in firing rate difference between the 500-900 ms post-stimulus interval and the 100-500 ms post-stimulus interval across (A). CS+ trials, and (B). CS- trials. The difference is the 500-900 ms rate minus the 100-500 ms rate. The grey lines show the differences in firing rate for each trial. The solid green line shows the SS-GLM estimate of the difference in firing rate. The dotted green lines are the 95% CI for the SS-GLM estimate. There was a significant increase in the difference for the CS+ trials (p < 0.004).

We are also studying changes in neuronal activity at the network level (within or between brain regions) by measuring the correlation of firing between pairs of neurons (Figure 4) to then relate these changes against behavior.

In separate experiments we studied position related neuronal activity in the hippocampus (Figure 5), in accordance with previously published findings we observed place fields, which are zones of the foraging area where certain neurons are active. This activity is specific to the context in which the recording is performed, hence when the context is altered over a certain level the place cells show a different location specific firing.

Figure 4. Cross-correlogram of neuronal activity recorded simultaneously within the same brain region during a learning task. Left: activity of one cell decreases in the 20 ms preceding spikes from the reference cell. Right: no specific pattern seems to link the activity of the 2 cells analyzed.

Navigation related memory formation is studied in rats using multi-unit recording implanted into both the amygdala and the hippocampus during acquisition and recall of a spatial task, active place avoidance (figure 6). An animal is placed in a cylindrical rotating arena and receives a mild footshock every time it enters an unmarked quadrant (60 degrees). The orientation cues are kept to a minimum to force the animal to hold a representation of space to avoid shocks, giving us the opportunity to study short term memory representation by neuronal activity. The same experiments are conducted during the ‘recall session’ (no shock) to study long term memory. For each part of this experiment we are also using place field mapping to investigate how space representation changes throughout training and how this neural activity is related to these changes.

Figure 5. The same single neuron’s activity was recorded in CA1 hippocampus region and mapped onto its foraging space from two different environment (0.72 sqm circular arena, left; 1.0 sqm square platform, right; colour coded activity: blue, place visited no activity; red 3.3 Hz left, 5Hz right).

Figure 6: Colour coded place preference recorded during the habituation (left), training (middle) and retention test (no shock, right) during an active avoidance task (blue to red gradient indicates an increase in presence time). Shock intensity 0.2 mA, rotation 1 rpm counter clock-wise. A preference for the same “safe zone” is observed during the acquisition and test trials. This indicates that the rat held a representation of its position in space relative to the shock zone (an unmarked 60 deg quadrant at the top of the arena).

Ongoing Collaborative Research

  • Action potential waveform is not a reliable tool for spike sorting in collaboration with Thinking Systems’ colleagues Peter Stratton and Allen Cheung.
  • Dorso ventral representation of place fields in CA3 in collaboration with Thinking Systems’ colleague Chris Nolan.
  • Idiothetic navigation for early Alzheimer disease diagnostic in collaboration with Adam Hamelin (Coulson laboratory, QBI).


  • Crane, J.W.*, Windels, F.,* Sah, P. (2009) Oscillations in the basolateral amygdala: aversive stimulation is state dependent and resets the oscillatory phase. Journal of Neurophysiology 102: 1379-1387.   (PDF File 1,751 KB)
  • Windels, F.*, Crane, J.W.,* Sah P. (2010) Inhibition dominates the early phase of up-states in the basolateral amygdala. Journal of Neurophysiology. doi:10.1152/jn.00531.    (PDF File 1,341 KB)
    * Authors contributed equally to the work.

Papers in Preparation

  • Martin, T., Windels, F.  RatTrack, software based video analysis for learning and memory experiment.
  • Stratton, P., Cheung, A., Kiyatkin, E.A., Wiles, J., Sah, P., Windels, F. Action potential waveform consistency between neurons limits spike sorting accuracy.

Conference Poster

  • Brain Plasticity Symposium, (2010) Queensland Brain institute.
  • Australian Neuroscience Society meeting, (2010) Sydney.

Laboratory Visits

  • June 2009, Visit to Matthew Wilson laboratory, Picower institute, MIT.
  • July 2009, Gordon conference, Neural circuit and plasticity, Newport.

Where to Next?

Following the work I did with the Thinking System Project, I am now working on a project investigating memory formation at the network level. I am using a conditioning paradigm that involves navigation and simultaneous recording of neural activity in multiple brain regions. We are more precisely investigating the changes in activity during the acquisition phase of the learning process. Based on published studies we are testing the hypothesis that a glutamate dependent replay of activity sequences are required for the early phase of memory consolidation. We received a 5 years NHMRC project grant for this project.