The discovery hinges on a classic piece of neural circuitry: the half-center oscillator. Engineers and neuroscientists have long known that two populations of neurons wired to inhibit each other can generate rhythm from nothing more than mutual antagonism. When one side fires, it suppresses the other; the suppressed side then escapes from inhibition, fires in its turn, and pushes its opponent back down. This arrangement powers the central pattern generators that control swimming, walking and breathing throughout the animal kingdom. The Oxford team has now shown that the same architecture, usually associated with movement, has been repurposed to regulate something as profound as the need for sleep.

To watch this circuit in action, the researchers implanted tiny translucent windows into the heads of living flies and used two-photon microscopy to image the calcium signals of dFBN dendrites over many minutes. What they saw was strikingly regular: fluorescence transients repeating rhythmically at 0.2 to 1 hertz, the frequency band that defines slow-wave activity. Simultaneous whole-cell patch-clamp recordings revealed the electrical events beneath these optical signals. The bistable neurons oscillated between depolarized UP states, during which they fired high-frequency bursts of action potentials, and hyperpolarized DOWN states, during which they fell silent. The alternation was metronome-like, mirroring the UP and DOWN membrane states that cortical neurons occupy during mammalian non-REM sleep.

Crucially, the slow rhythm did not arise from every cell in the ensemble, nor from cell-intrinsic pacemakers acting alone. When the team imaged individual dFBN somata within one hemisphere, their fluorescence traces were, on average, uncorrelated with one another. Instead, a handful of pacemaker-like neurons accounted for most of the variance in the compound dendritic signal. The real source of the rhythm emerged when the researchers imaged both brain hemispheres at once: the left and right dFBN populations oscillated in antiphase, each firing exactly when the other fell silent. Connectomic analysis of two independent electron-microscopy datasets, the hemibrain and FlyWire connectomes, confirmed that dFBNs are linked by dense, reciprocal, and almost perfectly symmetric connections, precisely the wiring diagram a half-center oscillator requires.

The neurotransmitter carrying this cross-hemispheric inhibition turned out to be glutamate, which is unusual because glutamate is best known as the brain’s main excitatory signal. In the fly central nervous system, however, glutamate frequently acts as an inhibitor. The researchers verified this assignment anatomically, using genetic intersection strategies that isolated dFBNs from other glutamatergic neurons, and functionally, with genetically encoded glutamate and acetylcholine sensors. When they optogenetically activated one random half of the dFBN mosaic, the other half hyperpolarized and stopped spiking, and this inhibition persisted in the presence of tetrodotoxin, proving that the connection is monosynaptic. dFBNs, in short, inhibit each other directly through inhibitory glutamatergic synapses.

Was this rhythm actually sufficient to put an animal to sleep? The team answered this with an elegant optogenetic replay experiment. Because stimulating both hemispheres in synchrony can drive the mutually inhibitory neurons into a depolarization block or a stalemate, earlier attempts to activate dFBNs had produced confusing results. Pulsed illumination, which mimics the natural burst structure, worked beautifully. When the researchers drove dFBNs with light patterns that reproduced naturalistic slow-wave activity, 500-millisecond bursts recurring at roughly 0.5 to 1 hertz, the flies went to sleep. The effect scaled with intraburst frequency up to a saturation point of about 40 hertz, matching the peak firing rates observed in spontaneous recordings. Tonic stimulation at the same average rate was far less effective, showing that the temporal structure of the signal, not just its intensity, carries the sleep instruction.

The necessity side of the argument was just as convincing. Depleting the vesicular glutamate transporter from dFBNs, or silencing them with potassium channels, reduced both baseline sleep and the rebound sleep that normally follows deprivation, all without hyperactivity or anatomical defects that might confound the interpretation. Sleep became fragmented, bouts shortened, and the slow oscillations themselves weakened. Interference with glutamatergic transmission thus dissolved both the rhythm and the behavior it was proposed to drive, tying the half-center oscillator directly to the homeostatic regulation of sleep.

Perhaps the most striking finding is that dFBN oscillations do not pause when the fly is awake. In mammals, slow waves are largely confined to deep sleep. In the fly, the team imaged head-fixed animals walking on a spherical treadmill and roused them with brief pulses of infrared heat aimed at the abdomen. The heat triggered dopamine release onto dFBN dendrites, locomotion, and a brief collapse of the calcium oscillations, but the rhythm quickly resumed through a powerful postinhibitory rebound. Slow-wave power in dFBNs rose with the duration of sleep deprivation, grew roughly linearly through the first half of the night, decayed over 12 to 24 hours of recovery, and followed a natural daily rhythm peaking in the evening. The oscillation therefore behaves as a continuous, amplitude-modulated readout of sleep pressure, streaming information about sleep need to downstream circuits even during waking, in a fly-mammalian inversion of convention.

How does the circuit tune its own gain? The answer is a surprising form of synaptic depression. Sleep deprivation is usually assumed to potentiate synapses, and indeed active zone markers like bruchpilot increase in most regions of the sleep-deprived fly brain. dFBNs do the opposite. Using genetic tagging of endogenous bruchpilot, the team showed that dFBN active zones shrink after sleep loss, and direct measurement of optically evoked glutamate release confirmed that transmission from dFBN synapses was attenuated. In a computational model of two reciprocally inhibitory dFBNs embedded in recurrent excitatory circuitry, this weakening was essential. Increased intrinsic excitability and stronger external drive raise slow-wave power only if the output synapses depress in parallel; without that homeostatic brake, the oscillating neurons would smother their own excitatory inputs. Two opposing forces, heightened excitability and weakened output, thus conspire to encode sleep need in the amplitude of the rhythm.

The broader picture is one of elegant engineering. dFBNs inherit their sleep pressure from a biochemical memory described in previous work, in which lipid peroxidation products oxidize the cofactor of the Shaker potassium channel’s Hyperkinetic subunit, one bit at a time. The half-center oscillator now provides the output stage: it converts that slowly accumulating chemical record into a periodic electrical signal whose amplitude tracks sleep need from moment to moment, resists the antagonism of arousing dopaminergic neurons in a flip-flop arrangement reminiscent of the mammalian hypothalamic sleep switch, and, when strong enough, tips the animal into sleep. The remaining puzzle is analog-to-digital conversion: sleep-wake states are binary, while the signal is graded. The authors suggest that strongly facilitating synapses or the winner-take-all architecture of the fan-shaped body could perform the transformation. Whatever the mechanism, the lesson is clear: the same rhythmic circuits that make legs move and hearts beat have been conscripted, over evolutionary time, to make brains rest.

Subject of Research: A half-center oscillator in Drosophila sleep-control neurons that generates slow-wave activity encoding and regulating sleep pressure

Article Title: A half-center oscillator regulates sleep pressure

Article References: A half-center oscillator regulates sleep pressure. (n.d.). https://doi.org/10.1038/s41593-026-02431-9

Image Credits: AI Generated

DOI: 10.1038/s41593-026-02431-9

Keywords: sleep pressure, half-center oscillator, slow-wave activity, Drosophila, dorsal fan-shaped body, glutamatergic transmission, optogenetics, sleep homeostasis, synaptic depression, two-photon imaging, connectomics, neural oscillations