Brain replays motor skills during sleep without help from the memory center

When we sleep, our brains replay the events of the day to solidify new skills, a process scientists previously thought was always controlled by a region called the hippocampus. A new study in mice reveals that the brain network responsible for habits and motor skills can generate these sleep replays entirely on its own. The findings, published in Nature Neuroscience, suggest that different memory systems in the brain can function and reinforce themselves independently.

Memory is not a single phenomenon. The brain handles facts and events, known as declarative memory, differently than it handles procedural memory, which covers motor skills and habits like riding a bicycle. For decades, researchers debated how the brain cements these different types of memories during sleep.

A prevailing theory suggested that the hippocampus acts as a central conductor for the sleeping brain. Under this model, the hippocampus replays the spatial and temporal context of the day’s events, sending signals that trigger the rest of the brain to lock in all types of memories. However, some past research hinted that motor skills might not need the hippocampus to become permanent.

To resolve this tension, a research team led by Emmett Thompson and Lars Rollik at University College London looked at exactly how the brain processes procedural memories offline. Marcus Stephenson-Jones supervised the work. The researchers wanted to know if the brain circuits that store motor skills possess an autonomous ability to replay memories, or if they rely entirely on the hippocampus.

The researchers first designed a physical task for a small study of mice. The animals learned to poke their noses into a specific sequence of five ports to receive a water reward. As the mice practiced, their movements became highly stereotyped, resembling the acquisition of a physical habit.

Free daily newsletter

To confirm which brain area drove this learning, the researchers disabled a region called the dorsolateral striatum, part of the dorsal striatum. This structure sits deep in the brain and is heavily involved in coordinating movement and processing rewards. Mice with a disabled dorsolateral striatum failed to learn the sequence from memory, and they showed highly variable, uncoordinated movements during the physical task.

Next, the team blocked chemical receptors required for neural plasticity in the dorsal striatum while the mice slept. The researchers waited until the animals had finished a day of training before delivering the blocking agent. The following day, the animals performed worse on the sequence task, dropping back to earlier levels of training.

The researchers then implanted high-density electrodes to monitor individual brain cells in the dorsal striatum. They recorded the mice both while they performed the task and while they slept afterward. To analyze the data, they used a machine learning tool that identifies hidden patterns in neural firing without relying on human assumptions.

During the awake physical task, the algorithm identified distinct sequences of brain cell activity that aligned with specific phases of the behavioral sequence. Different groups of neurons fired in precise orders depending on whether the mouse was moving between ports, consuming a reward, or performing other behaviors such as grooming.

When the researchers applied the same algorithm to the animals’ sleeping brain activity, they found that these exact neural sequences reactivated. The brain cells fired in the same order they did during the physical task, a phenomenon known as replay. This offline replay occurred at normal speeds, but it also often progressed much faster, compressing the time it took to complete the neural sequence into a fraction of a second.

The content of the sleep replay was highly structured. The sleeping brain prioritized sequences related to the physical task over random behaviors. It also heavily favored sequences associated with rewards. When the brain replayed sequences in a forward direction, and when it used neurons that fired consistently during the daytime task, the mice showed better physical performance the next day.

Finally, the team tested whether the hippocampus was orchestrating this striatal replay. They used a viral injection to perform large bilateral lesions, completely disabling the hippocampus in a group of mice. They then monitored the animals’ learning and their brain activity during sleep.

The mice without a hippocampus learned the five-step physical sequence just as well as control animals. They completed the task with the exact same speed and accuracy. This confirmed that the hippocampus was not required to learn or perform this specific motor skill.

When the researchers looked at the sleep data, they found that the dorsal striatum still generated neural replay. Every characteristic of the replay remained intact despite the missing hippocampus. The sequences played in forward and reverse orders, sped up and slowed down, and occurred across multiple stages of sleep. The brain also continued to prioritize task-related sequences just as it did in healthy mice.

While these results show that procedural memory replay can happen without the hippocampus, the findings do not mean the hippocampus is universally uninvolved in all forms of motor learning. In physical tasks that heavily rely on spatial navigation or learning an environment, the hippocampus might still play a role in shaping memory.

Because this research was conducted in mice, the exact mechanisms of memory replay will need to be confirmed in humans. Additionally, the researchers point out that future studies will need to identify the exact trigger for this independent replay in the striatum. They plan to investigate how other chemicals, like dopamine, might coordinate with replay to cement habits and skills during rest.

The study, “Replay of procedural memory is independent of the hippocampus,” was authored by Emmett J. Thompson, Lars B. Rollik, Benjamin Waked, Georgina Mills, Sthitapranjya Pati, Jasvin Kaur, Ben Geva, Haoyu Li, Rodrigo Carrasco-Davis, Tom George, Clementine Domine, William Dorrell, and Marcus Stephenson-Jones.