Two fundamental things happen to cortical synapses during sleep. One has been suspected for decades: the brain prunes back dendritic spines, paring down the connections strengthened during waking hours. The other, revealed in a study published July 29 in iScience, is the opposite of what most sleep researchers would have predicted.
At the same time the postsynaptic tree is being thinned, the presynaptic terminal is quietly restocking its arsenal of neurotransmitter vesicles. Sleep does not simply renormalize the synapse by cutting excess connections. It actively replenishes the machinery of transmission on the sending side while reducing receiving structures on the other.
The finding comes from a team led by David Elmenhorst at Forschungszentrum Juelich in Germany, who combined two complementary methods to watch both sides of the synapse across sleep, wake, and sleep deprivation in mice. The result reshapes how scientists understand the synaptic homeostasis hypothesis (SHY), the leading theory of why sleep exists at the molecular level.
The Two Faces of Sleep at the Synapse
The synaptic homeostasis hypothesis, proposed by Giulio Tononi and Chiara Cirelli in the early 2000s, holds that waking experience strengthens synapses across the brain, and that sleep is necessary to scale them back down. Without this renormalization, synapses would saturate, preventing new learning and wasting energy. Most evidence for SHY has come from postsynaptic markers: spine density, AMPA receptor levels, phosphorylation state.
The Juelich team wanted to test both sides of the synapse simultaneously. They used two techniques on the same animals: Golgi impregnation to visualize and count dendritic spines (the postsynaptic receiving ends), and ex vivo [3H]UCB-J autoradiography to measure synaptic vesicle glycoprotein 2A (SV2A), a protein expressed uniformly on presynaptic vesicles and a reliable proxy for synaptic vesicle density.
Mice were sampled at three time points: ZT0 (light onset, when mice begin their rest phase), ZT12 (light offset, the start of the active dark phase), and after six hours of sleep deprivation (also sampled at ZT12). The team focused on basal dendrites of layer 2/3 and layer 5 pyramidal neurons in primary somatosensory and visual cortex.
The spine data aligned with SHY. Spine density decreased during sleep, falling from ZT0 to ZT12, and increased after sleep deprivation. That is the classic renormalization signature: sleep pares back the postsynaptic structure built up during wake.
But SV2A told a completely different story.
A Surprising Presynaptic Reversal
SV2A levels moved in the opposite direction from spine density across every condition. During sleep, when spines were being pruned, SV2A levels actually increased. After sleep deprivation, when spines were dense, SV2A dropped.
This was not random. The two measures were tightly coordinated: in adult mice, the correlation between spine density change and SV2A change was r = -0.93 (p = 0.0008). In adolescents, the correlation was r = -0.83 (p = 0.0111). The relationship was so consistent that knowing one value effectively predicted the other.
The authors interpret this as a coordinated bidirectional regulatory strategy. During sleep, reduced postsynaptic demand frees resources for the presynaptic side to accumulate vesicle reserves. Sleep deprivation, by contrast, forces synapses to maintain transmission without the restorative period, gradually depleting the vesicle pool. The synapse is not simply turned down during sleep. It is selectively reconfigured: receiving ends shrink while sending ends restock.
The layer specificity reinforced the pattern. Layer 5 pyramidal neurons, the main output neurons of the cortex, showed stronger sleep-wake modulation in adult mice than layer 2/3 neurons. The system adjusts most at the level of cortical output, where the energy cost of sustained transmission is highest.
Adolescents versus Adults: Different Synaptic Strategies
The study also included adolescent mice (four to six weeks old, roughly equivalent to teenage humans) alongside adults (nine to twenty-four weeks). The differences were telling.
Adolescent mice had higher baseline levels of both spine density and SV2A across the board. Their synapses were more numerous and more heavily equipped with vesicles at every time point. This fits with what is known about developmental synaptic proliferation: the adolescent brain overproduces connections, then prunes selectively during maturation.
But the responsiveness to sleep and deprivation differed by age. Adolescent SV2A was more sensitive to sleep deprivation: the drop in vesicle protein after sleep deprivation was larger in young animals than in adults. Adult mice, on the other hand, showed stronger physiological sleep-related modulation in the normal sleep-wake cycle.
The authors also ran longitudinal PET imaging with [18F]SynVesT-1, a radiotracer that binds SV2A in living animals. After sleep deprivation, SV2A binding potential dropped by roughly 20 to 22 percent across every brain region measured: striatum (down 21.6 percent), cortex (22.7 percent), hippocampus (21.8 percent), and thalamus (20.8 percent). The effect was widespread, not confined to sensory cortex.
Crucially, recovery was incomplete even after 24 hours of ad libitum sleep. The vesicles did not bounce back fully, suggesting that a single recovery sleep cycle is not enough to undo the presynaptic cost of extended wake.
Why the Presynaptic Side Matters
Most sleep research has treated the synapse as if its two sides are a single unit. If SHY says synapses are strengthened during wake and weakened during sleep, the implicit assumption has been that both sides follow the same trajectory. This study shows they do not. The postsynaptic side follows the classic SHY prediction. The presynaptic side inverts it.
That matters because SV2A is a clinical target. It is the binding site for the antiepileptic drug levetiracetam, and [18F]SynVesT-1 PET is increasingly used in humans to study synaptic density in neurodegenerative disease, epilepsy, and psychiatric disorders. If SV2A levels fluctuate with sleep-wake state by 20 percent or more, clinical PET scans will need to control for time of day and prior sleep history to avoid confounding synaptic pathology with normal circadian dynamics.
The study also opens a new question: if sleep restocks vesicles while pruning spines, what molecular signals coordinate the two sides? The tight inverse correlation suggests a shared regulatory mechanism, but the authors note that the signaling pathways linking postsynaptic pruning to presynaptic vesicle accumulation remain unknown.
Limits of the Study
The study has several limitations. The researchers did not use EEG or EMG to confirm sleep stages, relying instead on behavioral monitoring of the mouse light-dark cycle. Spine subtypes (mushroom, stubby, thin) were not classified separately, and Golgi staining in two dimensions may underestimate absolute spine counts compared with three-dimensional reconstruction methods. The analysis was limited to primary somatosensory and visual cortex, leaving open whether the same pattern holds in association cortex, hippocampus, or subcortical structures.
The Juelich team plans to extend the work with longitudinal two-photon imaging in living mice to track the same synapses across sleep-wake cycles, and to investigate the molecular pathways that couple presynaptic vesicle turnover to postsynaptic spine dynamics.
For now, the takeaway is clear. Sleep does not just turn down the volume on cortical synapses. It reconfigures them, trimming postsynaptic receivers while loading presynaptic transmitters for the next day. The synapse goes into sleep with a different balance than it wakes up with, and that balance shifts in opposite directions on either side of the gap.
Source: Ma, J., Braun, A., Oskamp, A., Kricsfalussy-Hrabar, L., Hermes, N., et al. (2026). Bidirectional pre- and postsynaptic regulation of synaptic homeostasis in the mouse cortex across sleep-wake states and development. iScience, 29, 116676. DOI: 10.1016/j.isci.2026.116676. Open access (CC BY).