Fig. 5: Sweeps and internal direction signals persist during sleep.
From: Left–right-alternating theta sweeps in entorhinal–hippocampal maps of space
a, Sweeps and alternating internal direction pulses are preserved during REM sleep. Top, raster plot with spike times (black ticks) of internal direction cells sorted by preferred firing direction, and tracked head direction (blue line), during a 2.5 s extract from an epoch of REM sleep. Note the theta-paced, left–right alternating packets of direction-correlated activity. Bottom, decoded sweeps (filled circles, colour-coded by time) from grid cells of one module and decoded internal direction from direction-tuned cells (green arrows) during four successive theta cycles (1–4 in the top panel). The grey line shows the reconstructed 2.5 s trajectory after smoothing with a wide gaussian kernel (σ = 100 ms). For both position and internal direction, a Bayesian decoder was used, using tuning curves estimated by the LMT model during a preceding open-field foraging session. b, Internal direction-aligned, non-rhythmic trajectories during SWS. Top, raster plot (as in a) showing activity of internal direction cells during a 3 s extract from SWS. Note the sharp transitions between the up and down states. Bottom, decoded position from a single grid module (as in a) during each of three highlighted segments of an up state in the top panel. Note the sweep-like position trajectories aligned with the decoded internal direction (green arrows) in each segment. c, The decoded direction alternates from side to side during awake open-field running (Run) and REM, but not during SWS. Top, distribution of angles between decoded direction at successive peaks of activity (n = 9 rats, 1 session per rat) when the previous decoded direction was directed to the left (red) or right (blue). Bottom, autocorrelogram of decoded direction across brain states. Coloured dots, mean; whiskers, s.d. Note the rhythmic alternation during awake and REM. d, Top, decoded trajectories for 100 example sweeps from one example grid module (same decoding method as in a), coloured according to whether the previous decoded direction pointed to the right (blue) or left (red). Because spatial representations are decoupled from physical movement during sleep, sweep trajectories are referenced to the low-pass-filtered decoded trajectory (smoothed with a 100 ms gaussian kernel) and aligned to a ‘virtual head direction’ (low-pass-filtered decoded direction). Individual sweeps are plotted as separate lines. Bottom, averaged sweeps across brain states for all 18 grid modules of all 9 animals. Sweeps were referenced, rotated and sorted as in the top panel, normalized by the spacing of the grid module, and then averaged across all theta cycles. Each pair of red and blue lines corresponds to one module.
