Sleep maintains excitatory synapse diversity in the cortex and hippocampus

Abstract

Insufficient sleep is a global problem with serious consequences for cognition and mental health.¹ Synapses play a central role in many aspects of cognition, including the crucial function of memory consolidation during sleep.² Interference with the normal expression or function of synapse proteins is a cause of cognitive, mood, and other behavioral problems in over 130 brain disorders.³ Sleep deprivation (SD) has also been reported to alter synapse protein composition and synapse number, although with conflicting results.^4,5,6,7 In our study, we conducted synaptome mapping of excitatory synapses in 125 regions of the mouse brain and found that sleep deprivation selectively reduces synapse diversity in the cortex and in the CA1 region of the hippocampus. Sleep deprivation targeted specific types and subtypes of excitatory synapses while maintaining total synapse density (synapse number/area). Synapse subtypes with longer protein lifetimes exhibited resilience to sleep deprivation, similar to observations in aging and genetic perturbations. Moreover, the altered synaptome architecture affected the responses to neural oscillations, suggesting that sleep plays a vital role in preserving cognitive function by maintaining the brain's synaptome architecture.

Keywords: PSD-95; SAP102; protein lifetime; protein turnover; proteostasis; sleep; sleep deprivation; synapse; synapse types; synaptome; synaptome architecture.

Graphical abstract

Figure 1

Workflow for synaptome mapping Brain tissue from mice expressing the fluorescent proteins GFP and mKO2 fused to PSD95 and SAP102, respectively, was collected from the mice at different stages of the circadian cycle or after sleep deprivation. Synapse labeling and imaging shows genetic modification of PSD95 with eGFP and SAP102 with mKO2, which labels the proteins and their respective multiprotein complexes, which are distributed into synapse types/subtypes that can be visualized in brain sections using confocal spinning disc microscopy. The image analysis pipeline detects, segments, classifies, and quantifies synapse puncta, which are categorized into types and subtypes by machine learning and the spatial distribution plotted in synaptome maps.

Figure 2

SD alters brain synaptome architecture within the cortex and HPF SD reduces the size of PSD95-expressing synapses in cortical (A) and HPF (A and B) subregions. SD reduces synapse subtype diversity in cortical (C and G) and HPF (C, D, and G) subregions. Blue regions (A–D), Cohen’s d effect size for regions with significant changes shown (p < 0.05, Bayesian test with Benjamini-Hochberg correction); gray regions (A–D), not significantly altered. (G) X axis shows change (Cohen’s d) in synaptome diversity induced by SD. Asterisks indicate significant changes (p < 0.05, Bayesian test with Benjamini-Hochberg correction) in the diversity. Key for brain regions (E and F). Regions: CTX, isocortex; OLF, olfactory areas; HPF, hippocampal formation; CTXsp, cortical subplate; STR, striatum; TH, thalamus; HY, hypothalamus. Related to Figures S1–S4.

Figure 3

SD differentially impacts synapse subtypes (A) Heatmap of SD-induced changes in the density (Cohen’s d) of synapse subtypes in cortex (CTX) and HPF ranked from longest to shortest PSD95 lifetime. For a heatmap showing significant changes corrected for multiple testing, see Figure S4. (B) SD-induced synapse subtype density changes in the cortex (average of all subregions) plotted against PSD95 lifetime (normalized percentage). (C) Comparison of the density changes of six LPL (2, 3, 5, 20, 34, 35) with six SPL (6, 8, 11, 28, 29, 31) subtypes after SD. Red signifies subregions with greater change (Cohen’s d) in LPL than SPL synapses, whereas blue signifies subregions with greater change in SPL than LPL synapses; white subregions show no significant differences. Related to Figures S3 and S4.

Figure 4

Computational modeling of physiological responses in the CA1sr to patterns of activity after SD (A) The model simulates a 2D (11 × 11) array of synapses (boxes) expressing PSD95 and SAP102 measured along the radial and tangential axes of the CA1sr.^, (B) Five patterns of neuronal activity were used for CA1sr stimulation in the computational model. The summed excitatory postsynaptic potential (EPSP) response amplitudes in the ZT6 and ZT6SD groups were quantified (color bar, arbitrary units). (C) Comparison of the extent of disruption caused by SD for each of the five patterns of activity. ED, Euclidian distance.