Can sleep protect memories from catastrophic forgetting?
- Department of Medicine, University of California, San Diego, United States
- Department of Computer Science and Engineering, University of California, San Diego, United States
Figures
Figure 1
Network architecture and baseline dynamics.
(A) Basic network architecture (PY: excitatory pyramidal neurons; IN: inhibitory interneurons; TC: excitatory thalamocortical neurons; RE: inhibitory thalamic reticular neurons). Excitatory synapses …
Figure 2 with 1 supplement
Two spatially separated memory sequences show no interference during training and both are strengthened by subsequent sleep.
(A) Network activity during periods of testing (T), training of two spatially separated memory sequences (S1/S2), and sleep (N3). Cortical PY neurons are shown. Color indicates voltage of neurons at …
Figure 2—figure supplement 1
Sleep replay improves performance for complex non-linear sequences.
(A) Example of the training protocol used for training a long non-linear sequence - BACEDFHGIJ. (B) Average group activations during baseline testing (left), after sequence training (middle), and …
Figure 3
Sleep replay strengthens synapses to improve memory recall.
(A) Change in synaptic weight over entire sleep period as a function of the number of Up states where a given synapse was replayed. Each star represents a synapse in the direction of S1. Dashed line …
Figure 4 with 1 supplement
Training of overlapping memory sequences results in catastrophic interference.
(A) Network activity (PY neurons) during training and testing periods for three memory sequences in awake-like state. Note, sequence 1 (S1) and sequence 1* (S1*) are trained over the same population …
Figure 4—figure supplement 1
Interleaved training of the old and new memory sequences prevents the old sequence from forgetting and improves performance for both memories.
(A) Network activity during sequential training of memory sequences S1→ S2→ S1* (blue bars) followed by interleaved training of S1/S1* (green bar). (B) Example of stimulation protocol used for …
Figure 5 with 1 supplement
Sleep prevents the old memory sequence from forgetting and improves performance for all memories.
(A) Network activity (PY neurons) during sequential training of sequencies S1/S2/S1* (blue bars) followed by N3 sleep (red bar). No stimulation was applied during sleep. (B) Examples of testing for …
Figure 5—figure supplement 1
Training of a new memory that interferes with previously consolidated old memory leads to forgetting that can be reversed by subsequent sleep.
(A) Network activity (PY neurons) during training of S1 (150 s), S1* (350 s) (blue bars) and N3 sleep (red bars). No stimulation was applied during sleep. (B) Examples of testing periods for each …
Figure 6
Sleep promotes replay of both overlapping memory sequences during each Up state.
(A) Change in synaptic weight over entire sleep period as a function of the number of Up states where a given synapse was preferentially replayed. Each star represents a synapse in the direction of …
Figure 7 with 2 supplements
Sleep promotes unidirectional synaptic connectivity with different subsets of synapses becoming specific to the old or new memory sequences.
(A) Dynamics of synaptic weight distributions from the trained region. Top row shows strength of synapses in direction of S1. Bottom row shows strength of synapses in direction of S1*. Blue shows …
Figure 7—figure supplement 1
Interleaved training revealed synaptic weight dynamics that are similar to sleep but result in less segregation of synaptic weights.
(A) Dynamics of synaptic weight distributions from the trained region. Top row shows strength of synapses in direction of S1. Bottom row shows strength of synapses in direction of S1*. Blue shows …
Figure 7—figure supplement 2
Synaptic plasticity that is biased towards LTP or LTD also results in memory orthogonalization during sleep .
Synaptic dynamics for LTP/LTD ratio biased towards LTD (A+/A-=0.0019/0.002) (A) or LTP (A+/A-=0.0021/0.002) (B). Top, Scatter plots showing synaptic weights for all reciprocally connected pairs of …
Figure 8
Population of neurons participating in reliable replay during sleep overlaps with the early responders during memory recall.
(A) Characteristic examples of the network activity showing spiking events during stimulations of each individual ‘letter’ of a memory sequence in awake. (B) Distributions of the differences in …
Author response image 1
Total number of partial sequence replays for sequence S1 (blue) and S1* (red) during the first half (left) and last half (right) of up states during sleep.
Author response image 2
Sequence replays during sleep.
(A) Representative up state showing full sequence replays within the trained region of the network. (B) Histogram of the total in-degree of the neurons in our model. The x-axis is the in-degree or …
