In the brain of a fruit fly, sleep control hinges on a connection that should not work. Two groups of clock neurons demonstrably communicate: the small ventrolateral neurons (s-LNv) release the neuropeptide sNPF, and their larger neighbors, the l-LNv neurons, carry the receptors that translate that signal into consolidated nighttime sleep. Yet the most detailed wiring diagrams of the fly brain show no chemical synapse between the two groups. The message that controls sleep exists; the standard route for sending it does not. A preprint posted July 30 on bioRxiv resolves the contradiction: the sending neuron broadcasts three ways, each driving a different behavior, from its axon terminals, from its cell body, and through thin filopodia-like protrusions called cytonemes that touch neighboring neurons without forming a synapse.
The Terminal Message That Goes Nowhere
The default assumption in neurobiology is that a neuron releases neuropeptides from its axon terminals, the same sites that handle fast neurotransmitters, and that the peptide then diffuses to nearby targets. In a fly brain less than a millimeter wide, diffusion should be easy. To watch the process in real time, the study relied on a fluorescent sensor called GRABsNPF1.0, engineered to light up when sNPF docks on its receptor, expressed in the clock neurons of living brain explants under a spinning-disk confocal microscope. Control experiments showed the sensor responds to applied sNPF in a dose-dependent way, while a mutated, nonfunctional version does not respond at all.
The molecular clock drives a burst of sNPF release from s-LNv terminals in midmorning, at a time the field calls ZT 3. At that peak moment and at every other time point tested, the sensor registered nothing at the l-LNv neurons and nothing at the s-LNv cell bodies either. Even inside an insect’s tiny brain, the peptide released from terminals did not spread far enough to reach the cells that need it.
The Soma Broadcast
The signal that matters appears at the far end of the night. At ZT 23, about an hour before lights-on, extracellular sNPF rises at the s-LNv cell bodies, and the l-LNv somata respond in turn. The hour is meaningful: earlier work had shown that dense-core vesicles at the s-LNv soma fuse with the membrane at ZT 23 in an IP3-dependent process tied to sleep consolidation, and the new experiments tie the two observations together. Expressing an IP3 sponge that buffers this signaling, or knocking down sNPF production in the sending cells, abolished the late-night signal at both the s-LNv and the l-LNv somata. The l-LNv neurons make no sNPF of their own, so the peptide they detect at their cell bodies can only have arrived from outside.
The conclusion is that the sleep-relevant conversation runs soma to soma. The cell body of one neuron releases sNPF, the cell body of the receiving neuron picks it up, and the terminals are left out of the exchange entirely.
The Cytoneme Bridge
Diffusion between cell bodies is not the whole story, because the two neuron groups also touch. Reconstructions from three electron-microscopy connectome datasets, the Hemibrain, the optic lobe connectome, and the male central nervous system connectome, all built from FIB-SEM volumes, show s-LNv and l-LNv somata sprouting slender outgrowths up to about 20 micrometers long and only 150 to 400 nanometers wide. In live imaging these protrusions proved highly dynamic, extending and retracting over seconds. They carry dense-core vesicles, perform exocytosis inside the outgrowths themselves, and are labeled by an actin-binding protein, all hallmarks of signaling filopodia. They qualify as cytonemes, the same class of protrusion first described in developing fly tissues, and RNA interference against diaphanous, a formin required for cytoneme growth, shortens them as expected.
The contacts between cytonemes are what conventional connectomics misses. A trans-Tango assay, which labels any neuron touching a cell of interest, found contacts between s-LNv and l-LNv neurons in all 12 brain hemispheres examined. In the electron microscopy volumes, those contacts are direct membrane-to-membrane appositions wrapped in glia, with dense-core vesicles inside and omega-shaped profiles marking vesicle fusion. What is absent is just as telling: no T-bars, no postsynaptic densities, no clusters of small clear vesicles, and no microtubules. By every criterion used to annotate a synapse, these contacts do not register. They are real, direct, and behaviorally relevant, yet invisible to standard connectome analysis.
Three Compartments, Three Behaviors
The division of labor across the neuron is clean. Knocking down diaphanous in LNv neurons, which shortens their cytonemes, changed none of the circadian hallmarks: total activity, period, rhythmicity, and morning anticipation all stayed normal, consistent with the known role of terminal-released PDF peptide in driving circadian behavior. Nighttime sleep, the behavior governed by somatic sNPF, was also unaffected. What changed was sleep latency, the time it took the flies to fall asleep. The effect held across two independent RNAi lines and their insertion-site controls, with and without the sensor expressed.
The cytoneme effect is also independent of sNPF signaling. Cell-specific knockout of the sNPF receptor reduced nighttime sleep, reproducing the known role of the peptide, but left sleep latency untouched. Terminals, the soma, and cytonemes each own a distinct behavior: PDF at terminals steers circadian timing, somatic sNPF consolidates nighttime sleep, and the thin non-synaptic bridges tune how quickly sleep begins.
Why It Matters
The results are a pointed reminder that a wiring diagram is not the whole brain. Connectomics aims to explain behavior from synaptic connectivity, but the fly’s clock neurons do a large share of their work through channels the diagrams omit: peptide diffusion between cell bodies and cytoneme contacts that carry no conventional synaptic markers. The findings imply that connectome analysis may need to incorporate the proximity of peptidergic somata and cytoneme contacts to account for behavior. They also enlarge the picture of what a single neuron is: not a device that sends one signal from one place, but a cell that can broadcast different messages from different addresses, soma for sleep, terminals for timing, thin bridges for the speed of falling asleep.
Limits
Three caveats apply. The work is a preprint and has not been through peer review. It was done in Drosophila, and whether mammalian neurons use cytoneme-like contacts for similar signaling remains untested. And the diaphanous knockdown used to probe cytoneme function is a broad perturbation: it shortens these protrusions, but the behavioral assignment to cytonemes, though consistent across lines, rests on a single tool.
Bottom Line
In the fly clock circuit, sleep is not a single signal but a portfolio: peptide broadcast from the cell body consolidates nighttime sleep, peptide released at terminals steers circadian timing, and non-synaptic bridges of membrane decide how quickly sleep begins. The standard view that neurons communicate mainly at synapses is not wrong, just incomplete. The next generation of brain maps may need to draw the connections that do not look like connections.
Source
Klose MK, Rivlin PK, Bulgari D, Kim J, Gregg SN, Xia X, Li Y, Schmidt BF, Deitcher DL, Levitan ES. Adult Clock Neuron Somatic Neuropeptide Release and Cytonemes Regulate Sleep. bioRxiv preprint, posted July 30, 2026. DOI: 10.64898/2026.07.27.741016. Not yet peer reviewed. Supported in part by NINDS R01NS032385 and NIMH R24MH114785. No competing interests declared.