Disruption of ripple-associated hippocampal activity during rest impairs spatial learning in the rat

Abstract

The hippocampus plays a key role in the acquisition of new memories for places and events. Evidence suggests that the consolidation of these memories is enhanced during sleep. At the neuronal level, reactivation of awake experience in the hippocampus during sharp-wave ripple events, characteristic of slow-wave sleep, has been proposed as a neural mechanism for sleep-dependent memory consolidation. However, a causal relation between sleep reactivation and memory consolidation has not been established. Here we show that disrupting neuronal activity during ripple events impairs spatial learning. We trained rats daily in two identical spatial navigation tasks followed each by a 1-hour rest period. After one of the tasks, stimulation of hippocampal afferents selectively disrupted neuronal activity associated with ripple events without changing the sleep-wake structure. Rats learned the control task significantly faster than the task followed by rest stimulation, indicating that interfering with hippocampal processing during sleep led to decreased learning.

Figure 1. Learning is slowed down by ripple disruption during post-rest

A1. Spatial trajectory of Rat 3 in the two mazes T (test) and C (control) during the sixth day of exploration. The two mazes are created from a unique wagon wheel structure by placing walls and doors on the path; they are separated here for clarity. The trajectory is derived from diodes located on the headstage ~10 cm above the skull. Except to prevent crossing between maze arms, there were no walls on the sides, so that the rat could extend its head well outside of the maze floor (gray areas). In each maze, the rat has to navigate back and forth from the Start point to the Finish point; reward is supplied at both. Exploration of wrong arms is indicated by the black portions of the trajectory. The test maze is chosen randomly for each rat (right/left maze). A2. On the first day, the rat initially rests for 1 hour. Exploration of the first maze (C or T, randomized across rats) for 15 minutes is followed by a 1-hour rest period; this is repeated for the second maze. Microstimulation is applied during the rest period following maze T. The same sequence is repeated each day for 8 to 10 days, with alternation of the order of the mazes. B. Behavioral deficit in the test maze compared to the control maze. A trial was defined as a complete trajectory from Start to Finish. Left; the number of trials completed per day (± SEM) shows a delay of learning for the test maze compared to the control maze (ANOVA, Day × Maze interaction, P < 0.009, 5 rats). Right; the mean distance per trial (± SEM) was longer for the test maze (ANOVA, Maze term, P < 0.025, and Day × Maze interaction, P < 1.10⁻⁵). C. Summary of the days in which the mean distance per trial was similar (gray), longer (black) or shorter (light gray) for the the test maze compared to the control maze (two-tailed unpaired Student's t test, P < 0.05) for all the rats that learned the task. Rat 5 was terminated on Day 4 due to technical problems. For each rat, the first significant difference observed was a longer mean distance in the test maze, and this difference disappeared (and was reversed for Rat 2) at the end of testing. D. Increase in trajectory errors accompanying the lengthening of trial distances. Only data from days in which the mean distance per trial was longer in the test maze are plotted (black squares of C). Left; the number of errors per trial (exploration of wrong arms; see black segments in A1) was systematically higher in the test maze compared to the control maze (two-tailed paired Student's t test, P < 0.03, n = 12; gray lines indicate individual days, black symbols indicate mean ± SD). Right; concomitantly, the fraction of the trajectory traversing the wrong arms was larger for the test maze (two-tailed paired Student's t test, P < 0.005, n = 12).

Figure 2. The sleep/wake architecture is not modified by stimulation during rest

A. Classification into sleep/wake states for a window of fifteen minutes of rest with stimulation applied. A normal pattern is observed, consisting of mostly non-REM sleep and, in this case, one episode of REM sleep followed by a Wake period. Bin, 2s. B. Fraction of time spent in Wake, non-REM and REM sleep for the two post-rest periods, averaged over all rats and days. Error bars on the inside of each box indicate ± SEM; everything is relative to the full length of the bar, equal to 1. Below, average rate of ripple events (including the ones interrupted, when applicable) and of stimulation events in each state. The overall structure of the sleep/wake cycle was not affected by the stimulation (two-tailed paired Student's t test, P > 0.05, n = 33, for each state fraction) and this was true throughout the experiment (ANOVA, Day, Stim and Day × Stim interaction terms, P > 0.05 for each state fraction). C. Number and mean duration of long (> 20s) REM episodes averaged for the two rest periods. There was no significant difference in these distributions (two-tailed paired Student's t test, P > 0.8 and P > 0.3, n = 33). D. Rate of ripple events (including the ones interrupted) as a function of days for the non-REM and Wake states. REM episodes were omitted as they represent a minor fraction of time and have very few ripples (less than one percent of the total). Ripple occurrence in non-REM sleep was quite constant throughout the experiment, whereas in the Wake state, there was a significant drop in the rate of ripple events in the first days, before stabilization at about half the initial value (ANOVA, Days term, P < 0.0001 for all data or data restricted to the condition without stimulation). Differences between the conditions with and without stimulation in the non-REM periods are due to suppression of a fraction of ripples by the stimulation (see Figure 3 and Text).

Figure 3. Ripple events are interrupted and further ripples suppressed by stimulation

A. Top; raw LFP, LFP filtered online in the ripple band (100-400 Hz), power in the ripple band and multi-unit activity on five different tetrodes during rest, around the detection of a ripple that would have been stimulated if in a stimulation rest period. The boxes indicate ripples as detected offline by a classic algorithm (see Methods). The inside points correspond to the peaks in ripple power, which had to exceed a threshold (dotted line). Note that there is slight delay (7 ms) introduced by the online-filtering. Bottom; same signals around a stimulation event for the other rest period (same rat and day). The artifact in the raw LFP yields a stereotyped oscillation in the filtered LFP, which contaminates the measurement of the ripple power during a 60-ms period after the stimulation; likewise, the multi-unit activity was contaminated during a 30-ms time window (dashed lines). Signal in the ripple band is suppressed for several hundreds of milliseconds, concomitant with a suppression of firing. B. Ripple power averaged around stimulations (solid line) or around ripple detections that would have been stimulated (dashed line). The shaded areas indicate ± SD across sessions (n = 33). Electrical artifacts precluded measurement of ripple power during a 60-ms period after the stimulation, and traces were offset by 7 ms to compensate for the online-filtering delay. Top, box plots showing the duration of ripples detected by the threshold method in the two conditions. The left of the rectangle box is the median value of the ripple start relative to triggers, and error bars indicate the 10^th and 90^th percentiles; the right side of the box gives the same values for the ripple end. C. Top; Ripple power (± SEM) after triggers on a longer time scale. During rest with stimulation, power in the ripple band reaches on average the control value after one second. Bottom; number of further ripple events (± SEM) developing after triggers. Again, the ripple activity reaches the control value after one second. In the non-stimulation condition, there is an excess of new ripples at 200-300 ms above the baseline, illustrating that ripples often come as trains of events.

Figure 4. The multi-unit activity is suppressed by stimulation

A. Raster plots and peri-stimulus time histograms of the multi-unit activity recorded by a single tetrode, around ripple detections (Top) and stimulations (Bottom) for the two rest periods of one experimental day. Bin size, 1 ms. The points and error bars on the right indicate the mean spontaneous activity (± SD) calculated on the window [1.8 s - 2 s]. Only 50 trials, picked at random and sorted by time, are illustrated in each raster plot, out of 736 (Top) and 853 (Bottom). B. Average multi-unit activity around stimulations (solid line) or around ripple detections that would have been stimulated (dashed line). Bin size, 5 ms. The shaded areas indicate ± SD across sessions and tetrodes (n = 175). Unsorted electrical artifacts contaminated the activity during a 30-ms period after the stimulation. Multi-unit activity was suppressed for several hundred milliseconds by the stimulation.