One Midbrain Hub, Two Jobs: The Neurons That Run REM Sleep and Wakefulness
REM sleep has always been the odd one out. The dreaming brain hums almost as loudly as a waking one, the eyes sweep beneath closed lids, and the body switches off its own muscles so completely that researchers once called it paradoxical. It is the closest a sleeping brain gets to being awake, yet the two are usually studied apart. A study in Cell Reports now suggests they share an overlord: a small population of excitatory neurons tucked in the lateral periaqueductal gray (LPAG), a midbrain structure better known for escape, attack, and freezing.
A hub for two waking states
The periaqueductal gray is an ancient column of gray matter wrapped around the fluid-filled channel at the midbrain’s core. Its lateral division helps run flight-or-fight responses, pain suppression, cardiovascular control, breathing, maternal behavior, and reward. Sleep did not appear on that resume. The first hint came from whole-brain c-Fos mapping, a stain for recently active neurons: after REM sleep deprivation, REM rebound, and total sleep deprivation, LPAG clusters lit up.
The cells’ identity was a clue. In the neighboring ventrolateral periaqueductal gray, sleep-related cells are GABAergic, releasing the inhibitory transmitter GABA. The active LPAG cells were not, pointing to glutamatergic neurons, which release the excitatory transmitter glutamate.
Active in wake, active in REM
The team, from Fudan University in Shanghai with partners in Lyon and Tsukuba, watched the cells in real time. Fiber photometry tracks calcium signals in genetically targeted neurons, and in freely moving mice the LPAG glutamatergic cells ranked consistently: most active during wakefulness, moderately active during REM sleep, quietest during NREM sleep, and quick to fire as animals woke. Single-neuron recordings with optic tetrodes confirmed it: light-tagged LPAG glutamatergic neurons fired more than their neighbors during wake and REM but not during NREM, ramping up in the ten seconds before a mouse woke. Staining put a number on it: roughly half of the LPAG neurons activated by REM deprivation and rebound in rats were glutamatergic, almost 59 percent during total sleep deprivation in mice.
The brake experiment
To learn what these neurons do, the researchers silenced them. Chemogenetics drugs a specific cell type: the neurons carry a receptor, hM4Di, that responds only to an otherwise inert compound called CNO, dampening their firing. In mice whose LPAG glutamatergic neurons carried the receptor, CNO cut REM sleep by 31 to 47 percent in the hours after injection, boosted NREM sleep by roughly 14 to 26 percent, and trimmed wakefulness by about 11 percent in the late afternoon. Locomotion and muscle tone were untouched, so the effect was specific to sleep architecture. A second detail pointed beyond sleep: in a foot-shock test, silenced mice froze about 48 percent less than controls, a sign that these neurons help hold an animal still under threat.
The accelerator experiment
Then the researchers flipped the switch the other way. Activating the cells with a second chemogenetic receptor, hM3Dq, kept mice awake for roughly eight hours: wakefulness rose 195 percent, NREM fell 87 percent, and REM nearly vanished, down 97 percent. The wakefulness looked unusual: the EEG was dominated by high theta power, the signature of alert attention, while the mice sat immobile with reduced muscle tone. They were not paralyzed, still responding to tones and foot shocks; at home, only resting time increased.
Optogenetics, which drives neurons with light-sensitive proteins at millisecond precision, delivered the same message with better timing. A 30-hertz train of blue light delivered 20 seconds into NREM sleep snapped mice awake almost instantly; 20 to 50 hertz worked faster than 10 hertz, and an hour of intermittent stimulation kept mice awake throughout, REM falling to zero.
The REM pathway: to the SLD
Where do these commands go? Anterograde tracing showed LPAG glutamatergic axons reaching three brainstem targets: the sublaterodorsal tegmental nucleus (SLD), the locus coeruleus (LC), and the ventral gigantocellular reticular nucleus (GiV). Retrograde tracing with three different tracers showed individual LPAG neurons project almost exclusively to one of the three targets, like commuters who each ride a single line, so each could be tested alone.
The SLD is the brain’s REM engine, and the LPAG-to-SLD line proved to be the REM line. Silencing only the SLD-projecting LPAG cells cut REM sleep by roughly 125 percent relative to vehicle controls in the 09:00 to 11:00 window while leaving wake and NREM untouched. Activating those terminals at 30 hertz pushed sleeping mice from NREM into REM. And after REM rebound, LPAG neurons traced from the SLD were packed with c-Fos, while the same tracer placed in the GiV labeled almost none. The REM-active cells talk to the SLD, not the GiV.
The wake pathways: to the LC and GiV
The wake side of the ledger has two routes. The locus coeruleus is the brain’s noradrenergic alarm bell, and it answered: 65 percent of recorded LC neurons responded to LPAG stimulation, and about half of the LPAG inputs targeted noradrenergic cells. The GiV, a medullary nucleus involved in motor control and passive defense, responded too, with 77 percent of its neurons firing to LPAG input. Activating either set of terminals roused sleeping mice, boosting theta, damping delta, lowering muscle tone, and promoting the same still-but-alert immobility. The two routes differ in stamina: an hour of steady excitation of the LC line sustained wakefulness, while the GiV line produced arousal that did not persist.
Why it matters
At first glance, a sleep switch seems an odd second job for the brain’s threat center. The authors argue the roles do not conflict: arousal and vigilance are the same currency, and a structure built to keep an animal alert to danger is a natural place to run states that share the waking brain’s fast rhythms. The theta-rich immobility they observed looks like orienting, an animal scanning a scene while holding perfectly still.
There may also be a clinical payoff: the REM-wake boundary goes wrong in REM sleep behavior disorder and in narcolepsy, and the paper proposes LPAG glutamatergic circuits as candidate therapy targets.
Limits
The authors flag the usual caveats. c-Fos staining after REM sleep deprivation may partly reflect the stress of the deprivation procedure itself, not sleep pressure alone. The wake-active and REM-active cells appear to be two intermingled subpopulations with no molecular marker yet available to separate them, so the groups cannot yet be studied in isolation. In the activation experiments, the dominant wake-promoting effect may have masked the REM-promoting influence of the subset that projects to the SLD. And the immobility state, though it resembles freezing, was not fully characterized as such: heart rate and other autonomic signs were not measured.
Bottom line
A set of glutamatergic cells in the lateral periaqueductal gray switches on during both wakefulness and REM sleep, promotes REM through its projection to the SLD, and promotes wakefulness through projections to the locus coeruleus and the ventral gigantocellular reticular nucleus. The finding makes the LPAG one of the few regions known to coordinate both sides of the waking brain’s ledger, and a plausible target for disorders that blur the sleep-wake line.
Source
Wang Y-Q, Li L, Ma W-X, Chen L, Jiang J-B, Liu W-Y, Shi K, Arthaud S, Kong L-X, Hayashi Y, Qu W-M, Luppi P-H, Huang Z-L. Control of REM sleep and wakefulness by lateral periaqueductal gray glutamatergic neurons. Cell Reports. 2026;45(8):117821. DOI: 10.1016/j.celrep.2026.117821. PMID: 42585020.