New research from Fudan University reveals that cholinergic neurons in the brain’s oculomotor nucleus are organized into two parallel circuits: one that moves the eyes and a separate one that helps terminate REM sleep. The finding solves a longstanding puzzle about how the brain selectively spares motor outputs during rapid eye movement sleep.
During REM sleep, the brain paralyzes nearly all skeletal muscles. The sleeper is effectively locked in place, unable to act out their dreams. Yet the eyes move freely, darting back and forth behind closed lids. This selective motor sparing has long been a riddle for sleep neuroscience: how does the brain suppress body movement while preserving the precise, coordinated eye movements that define the REM stage?
A team led by researchers at Fudan University in Shanghai has now identified the neural circuitry that makes this possible. In a study published June 24 in Nature Communications, the group used optogenetics and cell-type-specific projection mapping in mice to show that cholinergic neurons in the oculomotor nucleus (nIII) are not a uniform population. They are split into two genetically distinct subtypes, each wired to a different downstream target and each serving a completely different function.
What they found
The oculomotor nucleus is best known for its role in eye movement. It sits in the midbrain and sends motor commands to four of the six extraocular muscles that rotate the eyeball. The Fudan team confirmed that a subset of cholinergic neurons in nIII (which they label nIII^ChAT) projects directly to these muscles and reliably triggers eye movements when activated. This subset has no effect on sleep or wake states. It does its job and nothing else.
The surprise came from a second subset of nIII^ChAT neurons. These cells project to the ventrolateral periaqueductal gray (vlPAG), a midbrain region already known to regulate sleep-wake transitions. Unlike their eye-moving neighbors, these neurons show no activity correlated with eye movements at all. Instead, their firing ramps up progressively just before REM sleep ends, as though they are preparing the brain to exit the REM state.
When the researchers used optogenetics to artificially activate this vlPAG-projecting subset, REM sleep was promptly suppressed. Conversely, silencing the same cells prolonged REM episodes. Crucially, activating these neurons changed nothing about eye movements or body motion. The two circuits coexist in the same tiny nucleus but operate independently, one dedicated to moving the eyes and the other dedicated to ending REM sleep.
The upstream control of this system also turned out to be organized along parallel lines. The nucleus papilio, a smaller structure that sends input to nIII, contains two genetically distinct populations of its own: Vglut2-expressing (glutamatergic, or excitatory) neurons and Vgat-expressing (GABAergic, or inhibitory) neurons. These two upstream channels appear to coordinate the two downstream nIII^ChAT circuits, providing a push-pull mechanism that helps the brain decide when to stay in REM and when to leave it.
Why it matters
This dual organization resolves a conceptual tension in sleep physiology. REM sleep is defined by rapid eye movements, yet the same neuronal population that produces those movements has also been implicated in sleep-wake regulation. How could the same cells do both things without conflict? The answer, the Fudan team shows, is that they are not the same cells. The cholinergic oculomotor nucleus harbors two parallel circuits that share a neurotransmitter and a home nucleus but diverge in every other respect: projection target, activity pattern, behavioral function, and effect on brain state.
The finding also identifies a specific microcircuit for REM sleep termination. What pulls the brain out of REM and into wakefulness or non-REM sleep has remained unclear. The vlPAG-projecting nIII^ChAT neurons appear to be part of that exit pathway. Their progressive ramp-up in activity before REM ends suggests they function as a timer or a trigger, gradually building pressure until the brain switches states. This could have clinical relevance for conditions in which REM sleep is dysregulated, including narcolepsy, REM sleep behavior disorder, and depression, where REM timing and duration are often abnormal.
Limits
This is a mouse model study. While the basic organization of the oculomotor nucleus is conserved across mammals, the specific circuit details and functional roles identified here need to be confirmed in humans. The study used optogenetics, a technique that requires genetic modification and is not transferable to human research. The behavioral readouts in mice (electromyography, electroencephalography, video monitoring) capture indirect correlates of REM sleep but cannot access the subjective experience of dreaming. Additionally, the study focused on cholinergic neurons; other neurotransmitter systems in the oculomotor nucleus may also contribute to the effects observed.
Bottom line
The oculomotor nucleus contains two parallel cholinergic circuits that serve different masters. One moves the eyes. The other helps the brain exit REM sleep. Their separation explains how the sleeping brain can move the eyes while keeping everything else still.
Source
Jiang C, Luo Y, Tan X, et al. Parallel cholinergic circuit in oculomotor nucleus to control eye movements and REM sleep. Nature Communications. 2026 Jun 24. doi: 10.1038/s41467-026-74768-5. PMID: 42342690.