NREM Oscillations Mediate Synaptic Proteome Remodelling to Support Synapse Stabilisation

Synapses are known to remodel their proteome during sleep . However, the exact mechanisms driving this remodelling and its impact on synaptic function or cognition are not well understood. We combine 24 -hour EEG recordings with time -resolved synaptic proteomics in a model of SYNGAP1-Related Disorders to reveal a mechanism by which slow-waves and spindles, two NREM sleep oscillations, mediate the remodelling of the synaptic proteome . Moreover, we uncover that this remodelling promotes synaptic stabilization , which could support sleep-dependent memory consolidation. In contrast, the increase of slow-waves and decrease of spindles found in Syngap1+/- mice would activate molecular pathways invol ved in synaptic weakening instead of stabilization. This is consistent with the proposed roles of slow -waves and spindles in synaptic downscaling and potentiation, respectively. Here, w e provide evidence on how NREM oscillations regulate the synaptic proteome and reveal a pathological mechanism that could be of relevance to all neurodevelopmental disorders coursing with sleep disturbances.

Syngap1, sleep /wake cycles , NREM sleep, synapse, proteomics, brain oscillations, electroencephalography, spindles, slow-waves, mouse model, neurodevelopmental disorder .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 3

Sleep is an evolutionarily conserved process1,2 that serves essential biological functions. These i nclude cognitive processes, such as learning or memory consolidation3–6, or cellular restoration events, such as metabolic waste clearance 7,8. Consistently , sleep alterations lead to impairments in neural processing and behavior 9–11. Furthermore, sleep disturbances are highly prevalent across neurological and psychiatric conditions , being particularly prominent in genetic neurodevelopmental disorders (NDDs) 12–15. These observations highlight sleep as a critical component of brain health and a potential contributor to disease-related neural dysfunction. Several studies have sho wn that sleep influences the proteomic composition of synapses16–20. Indeed, the synaptic proteome undergoes daily remodelling that are linked to sleep and wake cycles18. Importantly, these daily proteome dynamics are not caused by tissue-wide changes in gene expression or protein turn-over, instead, they are restricted to synapses18, and thus should be driven by synapse-specific mechanisms. This suggest that the marked differences in brain activity occurring between sleep and wakefulness lead to distinct patterns of synaptic stimulation, resulting in the remodeling of the synaptic proteome. Therefore, it is plausible that specific brain oscillations within a given state, such as slow-waves (SWs) and spindles in NREM sleep, are also able to influence synaptic composition in different manners. Nevertheless, the exact influence of different NREM oscillations onto the synaptic proteome remains unexplored. It is well -known that both SWs and spindles play an important role in cogn ition, since their interaction facilitates the transfer of information from short- to long-term storage, contributing to memory consolidation during sleep3,4,21. SWs are oscillations in the 0.5- 4 Hz range that emerge from alternating periods of firing and silence in cortical neurons22, while spindles are waxing and waning oscillations in the 9–16 Hz range that originate in thalamocortical circuits 21. Both oscillations have been proposed to participate in synaptic plasticity. SWs would promote synaptic downscaling during sleep23, indeed, different patterns of stimulation delivered onto pyramidal neurons during up states produce synaptic weakening 24–26. On th e other hand, spindles closely correlate with calcium transients in cortical dendrites, suggesting they may facilitate synaptic potentiation27–29. Accordingly, a theoretical framework of sleep function has proposed that spindle-rich sleep supports synaptic and circuit potentiation, while SWs- rich sleep promotes homeostatic synaptic downscaling30. Pathogenic variants in the SYNGAP1 gene cause an autosomal dominant NDD involving intellectual disability and epilepsy 31–34. Beyond these hallmark features, sleep disturbances are highly frequent in SYNGAP1-Related Disorders (SYNGAP1 -RD), affecting about two -thirds of these individuals 32,33. Importantly, among the slee p disturbances identified, we have recently shown that individuals with this condition present a marked decrease in spindle activity33. Alterations in this oscillatory activity .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 4 have been reported in related NDD s, including Rett35, Angelman36, and rodent models of Fragile X 37 or Dravet 38. At the molecular level, the SYNGAP1 gene encodes for a Ras/Rap GTPase-activating protein (SynGAP) highly enriched in synapses 39,40, where it plays important functions in both Hebbian and homeostatic synaptic plasticity41–45. Given the high prevalence of sleep disturbances in SYNGAP1-RD, the central role of the SynGAP protein in synaptic physiology, and the emerging evidence linking sleep to the regulation of the synaptic proteome, we asked whether Syngap1 heterozygosity disturbs daily synaptic proteome dynamics. Here, we integrate 24-hour electroencephalographic ( EEG) recordings with time -resolved synaptic proteomics in Syngap1+/- mice to provide evidence linking slow-waves and spindles to synaptic proteome remodelling . We reveal a potential pathological mechanism in which disrupted brain oscillations would lead to dysregulated synaptic proteome dynamics. This could have important cognitive implications for NDDs and other neurological conditions with aberrant brain oscillations. .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 5

Sleep Architecture and Circadian Behaviour are Preserved in Syngap1+/- Mice We first performed 24h video recordings in wild-type (WT) and Syngap1+/- (HET) mice to track their locomotor activity . These started at Zeitgeber Time (ZT) 0, that is , at the beginning of the light /rest period. As expected for nocturnal animals, both genotypes presented clearly enhanced locomotion during the dark /active period (Figure 1a,b and S1a), suggesting a well-preserved circadian behaviour. However, as previously reported46,47, Syngap1+/- animals had a larger light to dark increase in activity (Figure 1a,c). To directly assess sleep -wake dynamics we conducted 24 h EEG recordings ( Figure S1b,c). First, we quantified the percentage of daytime spent awaken, in non -rapid eye movement (NREM) or REM sleep and found no differences between genotypes for any state (Figure 1d,e). Furthermore, both genotypes showed higher Wake amount during the dark/active period and higher NREM and REM amounts during the light/rest period (Figure 1d,f). Finally, we also studied sleep consolidation by looking into in the number and length of sleep and wake episodes (see Methods) . We found no differences between genotypes in the number of episodes from each state ( Figure 1g) or in the distribution of their lengths (Figure 1h). Overall, these findings indicate that our model of SYNGAP1-RD does not present substantial alterations in sleep architecture and circadian behaviour. Syngap1 Heterozygosity Principally Disrupts NREM Sleep Brain Activity Given the absence of gross phenotypes in sleep and circadian behaviour, we next assessed brain activity by applying a Fast Fourier Transform (FFT) to the EEG data. However, Syngap1+/- mice are known to present interictal spikes (IIS) 48,49 that could interfere with this analysis, thus, we applied a procedure to identify them (Figure S2a- c). Indeed, IIS were increased in Syngap1+/- mice (Figure S2d), particularly during NREM sleep (Figure S2e,f). Therefore, detected IIS were removed prior to applying FFT to the data. Analysis of the power spectrums obtained by FFT identified a slowing of the theta peak during wakefulness in Syngap1+/- mice (Figure S3a). However, this was specific of the electrodes placed above hippocampi (hippocampal derivations) , since no differences were observed in the parietal electrode. Regarding sleep, no significant differences were found in the power spectrums of REM sleep between genotypes (Figure S3b). Contrarily, we observed robust EEG power alterations in NREM sleep, as these affected all derivations (Figure 2a). More precisely, NREM sleep alterations in Syngap1+/- animals included an increase in delta (δ, 1-4 Hz) and a decrease in sigma (σ, 9–16 Hz) power, two hallmark rhythms of NREM sleep. .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 6 Increased Slow -Waves and Reduced Spindles Underpin NREM Sleep Brain Activity Alterations During NREM sleep, delta and sigma powers reflect the occurrence of slow-waves (SW) and spindles, respectively . In addition, delta power follows a temporal pattern that closely mirrors homeostatic sleep pressure50–53, whereas sigma power is regulated in a circadian manner, presenting a peak in the light-dark transition (ZT12) 54,55. Therefore, the observed power alterations in Syngap1+/- mice could arise from differences in SW/spindles, from an impairment of the δ /σ temporal regulation or both. The analysis of SWs ( Figure 2b ) revealed an increased density of these events in Syngap1+/- mice ( Figure 2c ). Furthermore, SWs also presented morphological differences, having higher amplitude and duration ( Figure 2 d,e), and reduced slope (Figure 2f ). To study the homeostatic regulation of delta power, we fitted a mathematical model of sleep pressure to the temporal pattern of delta power (see Methods). The fitted trajectories closely tracked empirical δ in both genotypes ( Figure S4a-c). Importantly, the estimated time constants for build -up (Ti; Figure S4d ) and dissipation (Td; Figure S4e ) of sleep pressure were similar between genotypes . Together, these data indicate that the increased delta power found in Syngap1+/- mice stems from a higher amount and amplitude of SWs , not from differences in the homeostatic regulation of the δ power. Spindle detection (Figure 2g) revealed a reduced density of these events in heterozygous mice (Figure 2h). This was accompanied by a decrease in their amplitude, duration and the number of cycles within individual spindles (Figure 2i-k). In addition, we also found decreased coherence of spindles between EEG channels , meaning that they are less coordinated across electrodes (Figure 2l). Regarding the temporal regulation of σ, both genotypes showed a clear time modulation with similar amplitude and a peak occurring near ZT12 (Figure S4f-i). Together, these results show that the sigma power decrease in Syngap1+/- mice is driven by a reduction of spindle density, while its circadian modulation is preserved. Finally, we have previously shown that in Syngap1+/- mice IIS mostly appear during NREM sleep. Thus, we asked whether alterations in SWs or spindles were related to this abnormal activity. Interestingly, while no correlation was found between SWs and IIS (Figure S3c ), a significant negative correlation was found for spindles ( Figure S3d ), suggesting that the epileptic activity in Syngap1+/- mice is linked to the reduction of spindle density. .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 7 Syngap1+/- Mice Present Attenuated Infra-Slow Oscillations and Fewer Microarousals Spindles are organized into clusters that appear approximately every 50s (~0.02 Hz) . Spindle clustering drive a rhythm of the sigma power that is known as the Infra-Slow Oscillation (ISO)56–58. Importantly, this rhythm is imposed by fluctuations in noradrenaline (NA) levels57–59, with the ISO peak coinciding with NA troughs. Since we found reduced spindle density in heterozygous mice, we investigated whether their ISO was also disrupted. We first confirmed that the sigma power of WT animals presented a period ic modulation that coincides with spindle clustering (Figure 3a ). In contrast, these clusters were not well organised in Syngap1+/- mice, resulting in a poor definition of the ISO in these animals . Indeed, FFT of sigma power temporal profile revealed a robust peak between 0.01 –0.04 Hz for WTs that was significantly attenuated in our model of SYNGAP1-RD (Figure 3b). Beyond spindle clustering, NA fluctuations critically shape NREM sleep microarchitecture, being linked to the onset of microarousals (MA), which occur in NA peaks and, thus, in ISO troughs56,60,61. Therefore, we next quantified the phase relationship between the temporal profile of sigma power and the appearance of MAs (Figure 3c–d). MA timing was tightly phase-locked to the ISO in both genotypes near the σ trough. Despite preserved phase -locking, Syngap1+/- mice showed a significant reduction in MA density ( Figure 3e ), which is consistent with their attenuated ISO. Overall, the ISO impairment combined with the MA reduction suggest a dysfunction in noradrenergic signalling during NREM sleep in Syngap1+/- mice. Taken together, these findings show that not only spindle density is reduced but their temporal organization is also disrupted. Combined with the alterations in SWs, our

point to an important reorganization of NREM brain activity in Syngap1+/- mice, with no major changes in sleep-wake organization and circadian behaviour. Disorganization of Daily Synaptic Proteome Dynamics in Syngap1+/- Mice Since the synaptic proteome undergoes a daily remodelling during sleep/wake cycles18 and Syngap1+/- mice present alterations in NREM sleep , we aimed to investigate if proteome dynamics were affected in these mice . To achieve this, we conducted a proteomics analysis of cortical synaptic fractions collected every 4h along the day (ZT0, ZT4, ZT8, ZT12, ZT16 and ZT20; Figure 4a), from WT and Syngap1+/- mice, identifying a total of 5,112 proteins (Table S1 ). As previously reported 62, these synaptic fractions contain proteins spanning all major sub-synaptic compartments (Figure 4b), offering a comprehensive view of the synaptic proteome. In parallel, homogenates used to generate these synaptic fractions were processed using the same proteomics workflow (Table S1 ). By c omparing these datasets, we aimed to determine whether temporal .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 8 changes in synaptic protein abundance are synapse-specific or if, instead, they reflect broader, tissue-wide dynamics. To identify proteins with daily abundance changes (i.e., cycling proteins) we applied a model selection framework 63 (see methods , Table S2 ). This includes 7 cosine-based models encompassing all possible expression profiles between genotypes : a shared- cycling model (identifying proteins with the same rhythmic expression pattern in both genotypes; Synapses Figure 4c; Homogenates Figure S5a;), two genotype -specific models (identifying proteins with rhythmic expression pattern only in one genotype, Synapses Figure 4d,e; Homogenates Figure S5b,c), three models in which a protein cycles in both genotypes but with shifted amplitude and/or phase (phase being the time of maximal abundance; Synapses Figures 4f,g; Homogenates Figure S5d) and, finally, a null model, which captures proteins that do not cycle (Figure 4h). In WT mice , approximately 10% (533 proteins) of the synaptic proteome exhibited a cycling behaviour, whereas this proportion increased to 15% in Syngap1+/- mice (Figure 4i,j). More importantly, only 193 synaptic proteins presented a similar regulation in both genotypes (Table S2). Indeed, two-thirds of synaptic proteins cycling in WT, either lost or changed their temporal expression in HET mice (Figure 4i,j). Moreover, Syngap1+/- mice had 526 newly cycling proteins in synaptic fractions (Table S2). Noticeably, in WT samples, only a small subset of proteins cycled in both homogenates and synapses (Figure S5e) and, of those, few showed similar phases (Figure S5f). The same scenario was found for HET samples (Figure S5g,h). These differences manifest that homogenate and synaptic fractions largely capture different functional pools of the same proteins and that expression patterns found in synaptic fractions are highly specific to this compartment. To clarify if genotype -specific proteins reflect robust biological differences, we ranked all proteins by their goodness-of-fit (R²) to a cosine model and partitioned them into deciles. Notably, proteins identified as WT-specific cyclers were significantly enriched in the lowest R² deciles of the Syngap1+/- rank (Figure S6a,b). Similarly, cycling proteins specific to heterozygous mice were significantly enriched in the lowest R² deciles of the WT rank (Figure S6c,d). In contrast, proteins assigned to the shared-cycling, phase-shift and amplitude-shift models exhibit a significant enrichment within the highest R² deciles in both genotypes (Figure S6e-h). These results indicate that proteins classified by the genotype-specific models reflect robust loss or gain of cycling proteins in Syngap1+/- mice. Synaptic Functions in Wild-Type Mice Are Organized Sequentially Along the Day We first wanted to investigate how regulation of synaptic protein abundance shape s synaptic function along the day in wild-type mice (Figure 5a). To do this, we performed .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 9 an enrichment analysis of signalling pathways and Gene Ontology (GO) terms using all proteins with a temporal regulation . However, we argued that for a given term to be considered as time -regulated most of its constituent proteins should cycle in a synchronized manner. Thus, we introduced the notion of temporal coherence in this analysis (see methods). In addition, temporally coherent terms (Table S3) were grouped into the following 9 broader categories, termed Synaptic Functional Modules (SFM) : Signal Transduction, Translation, Metabolism, Protein Stability, Intracellular Trafficking, Endocytosis, Synaptic Vesicle, Synaptic Assembly and Synaptic Stabilization. The dark/active period (ZT12–ZT24) was characterised by three major SFMs : Signal Transduction, Translation and Protein Stability (Figure 5 a2). Signal Transduction pathways, with terms related to regulation of MAPK, calmodulin -dependent kinase activity, or PKC biology (Table S3). Notably, of the 202 kinases present in our synaptic proteome 30 cycled in a daily fashion, with 26 of them peaking in the dark (Table S2). Examples include Camk2a, Camk2g, P kcg or Sik3. Noteworthy, Sik3, which has been consistently implicated in sleep regulation 64–68, cycled with a maximum at ZT19 in WT synapses. Protein Translation control was also activated early in the dark phase, with multiple initiation factors (EIF1AX, EIF3E, EIF3G, EIF4A1, EIF4A2) peaking at ~ZT16. Subsequently, in the second half of this period, several processes related to Protein Stability emerged. Proteins relevant to autophagy (ATG5, ULK2, WDR45 ) peaked at ~ZT18, while terms related to the unfol ded protein response (ER/UPR) , known to be upregulated by extended wakefulness69–71, emerged near the dark-to-light transition (Figure 5a1, Table S3). Interestingly, terms from the Synaptic Assembly module appeared at the end of the dark phase and spanned into the early-light period. These include “Regulation of cell junction assembly” at ZT23, the term with the largest number of proteins (23), as well as others like “Neuron recognition” or “Regulation of axon guidance” . These terms contain key proteins for synaptic formation, such as neuroligins (NLGN2 , NLGN3), Bai1 (ADGRB1), NRP1, NEGR1 or FLRT3; as well as neurotrophic receptors EPHA7, SLITRK2, SLITRK4, TrkC (NTRK3) and their ligands, SLIT1 and EFNB3 (Table S3). Remarkably, during the early- light period terms related to the Synaptic Stabilization module were prominent, including functions important for postsynaptic organization or for the localization and activity of neurotransmitter receptors. These terms included core PSD scaffolds (DLG3, DLGAP3, HOMER2, SHANK2), glutamate receptors (GRIA4, GRIN2B, GRM5), their auxiliary subunits (CACNG2/7; NETO1) and actin regulators (BAIAP2, RHOA). The first half of the light period was also characterized by the emergence of pa thways related to metabolism and mitochondrial function (e.g., β -oxidation or TCA cycle). Strikingly, proteins involved in mitochondrial translation, including many ribosome components and two translation initiation factors , showed maximal expression at ZT3, .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 10 suggesting intense mitochondrial protein synthesis at this time. Finally, starting at the second half of the light period and spanning into the dark one , synaptic function was dominated by the Intracellular Trafficking mo dule (e.g. Golgi vesicle biogenesis, microtubule motor activity, axonal transport). Overall, in WT animals, SFMs were clearly segregated along the day, with little overlap between them. Remarkably, our results suggest that synaptic functions are dynamically adjusted to support information processing during the active period, resulting in synaptic remodeling or formation towards the dark -to-light transition . These events would culminate, during the first hours of the rest phase, in a process of synaptic stabilization (Figure 5a2). Marked Alterations in Synaptic Functions During the Light/Rest Period in Syngap1+/- Mice In Syngap1+/- mice synaptic function organization in the dark/active period was overall well preserved (Figure 5 b and Table S3 ). As in WT mice , this period was mostly characterized by terms related to S ignal Transduction (e.g., CaMKIV–CREB, p38 MAPK, kinase activator activity), Protein Translation and Protein Stability (e.g., folding/stress- response terms; chaperone -mediated autophagy) . For instance, 16 of the 26 kinases peaking in the dark in WT animals also did so in HET synapses (e.g. CAMK2A, PRKCG or ROCK2), although proteins with gained regulations also contributed to Signal Transduction functions ( e.g. CAMK2B, PLCD3, DAP2IP or TIAM1). Regarding Protein Translation, it is noteworthy that Syngap1+/- synapses not only presented increased levels of translation initiation factors , as observed in WTs, but also many ribosomal subunits peaked at ZT17. Finally, as in WTs, terms from the Synaptic Assembly module (e.g., neuron remodelling, focal adhesion assembly, positive regulation of cell -junction assembly) rose towards the end of the dark cycle. These contained shared cycling proteins such as NRP1, NLGN3, FLRT3 but also gained ones as EFNB2, PTPRF or RELN. Nevertheless, some differences were also observed during the dark/active period. For instance, terms from the Synaptic Stabilization module, which in WT mice appear in the early light/rest period, were found in the dark. These included terms such as ‘ionotropic glutamate receptor activity ’, ‘neurotransmitter receptor complex ’, or ‘ regulation of neurotransmitter receptor activity ’, containing proteins like GRIK2, GRIN2D, GRID1, DLGAP4 or SHISA6. Similarly, a cluster of metabolism and mitochondrial -related terms, which in WTs are characteristic of the light period, also appeared during darkness. These involved cellular respiration and mitochondrial protein synthesis , meaning that the prominent peak of mitochondrial ribosome components appearing around ZT3 in WT samples was shifted close to 12 hours in heterozygous mice. .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 11 In contrast to the better conservation of SFMs found in the dark/active period, m ore pronounced divergences emerged in the light/rest one. Most noticeably, terms related to Synaptic Stabilization in the early light phase were absent from Syngap1+/- synapses. Instead, we observed multiple terms from two related SFMs unique to HET samples : Endocytosis and Synaptic Vesicles. These SFMs were related to the internalization of neurotransmitter receptors and to clathrin-mediated endocytosis, which were driven by the cycling of clathrin subunits (CLTA and CLTC), clathrin adaptors (PICALM and SNAP91), cargo selection proteins (e.g. ARRB2 or DAB2) and many synaptic vesicle constituents (e.g., SYN1, SYP, SVOP, SCNA). In addition, while in WTs Signal Transduction were mostly confined to the dark period, in HET synapses this module was also abundant during the light. However, these Signal Transduction terms were clearly different from those in the dark, including p hosphatases and kinases related to synaptic depression (PPP3CA, PPP3CB, GSK3B and CDK5)72–75. Together with the appearance of endocytic processes, these results suggest a process of synaptic weakening in Syngap1+/- synapses during the light/rest period. Further differences in synaptic functions from the light/rest period include a strong presence of Protein Stability terms, which in WTs appeared mainly confined to the dark. Indeed, a very large number of components from the proteasome (19/33) and the Chaperonin Containing TCP -1 (CCT) complex (8/9) cycled coherently in Syngap1+/- synapses with maximal abundances around ZT6. On the other hand, two SFMs characteristic of the light were preserved in HET mice : Metabolism/Mitochondria (e.g., TCA, glycolysis/gluconeogenesis) and Intracellular Trafficking (e.g., axonal transport, dynein heavy-chain binding). Overall, the temporal organization of SFMs in Syngap1+/- mice was importantly altered, showing increased temporal overlap and clustering either in the mid -light or mid-dark period (Figure 5b2). However, we found the identity of SFMs t o be better preserved during the dark period than in the light one. Our results suggest that, as in WTs, synapses are primed for information processing in the active phase, leading to synaptic formation. However, the synaptic stabilization occurring during early-rest would be switched for a process of synaptic downscaling. NREM Sleep Is the Main Driver of Synaptic Proteome Dysregulation in Syngap1+/- Mice Since the disorganization of synaptic functions in HET mice was most pronounced during the light period, we examined whether genotype-specific cycling proteins preferentially peaked during this circadian window. We therefore analysed the distribution of phases among shared and genotype-specific cycling proteins. Interestingly, phases within each group were not randomly distrib uted across the day but, instead , clustered at specific times (Figure 6a-c). Shared cycling proteins tend to peak at ZT19 ( Figure 6a), proteins with lost regulation in HET mice (i.e WT-specific) do so around ZT4 (Figure 6b) and HET- .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 12 gained cycling proteins exhibit a biphasic distribution, with peaks at the middle of the light (ZT6.5) and dark (ZT18.5) periods (Figure 6c). Thus, shared cycling proteins tend to peak at the dark/active period (Figure 6a ), where synaptic functions are better preserved (Figure 5b), while proteins from the genotype-specific models tend to do so in the light/rest period, which presented marked functional differences (Figure 5b). The non-uniform distribution of protein phases within each group indicates that they might be regulated by a common driver. Two plausible candidates would be circadian rhythms and sleep/wake cycles18. To explore if a circadian dysregulation existed at the molecular level in heterozygous animals , despite their preserved circadian behaviour (Figure 1, S1 and S3), we analysed the daily gene expression pattern of six circadian clock components in the same cortical samples used for proteomics. In line with our behavioural data, we found cycling patterns of these genes to be well preserved in HET mice (Figure 6d-i), further arguing against circadian rhythms acting as the main driver of synaptic proteome dysregulation. We thus explored the contribution of sleep/wake cycles to synaptic proteome dynamics by looking at the temporal distribution of wakefulness, NREM and REM in our EEG data. Wakefulness distribution is centred around ZT16 in WTs and ZT18 in HET mice (Figure 6j), close to the mean phase of shared-cycling proteins (Figure 6a), suggesting they could be upregulated by wakefulness . Regarding sleep, REM shows a distribution centred around ZT11 in both genotypes (Figure 6l), yet this does not coincide with the mean phase of any protein cycling group. Indicating that REM sleep is not a major driver for any of these protein groups. On the other hand, NREM sleep in WTs is centred around ZT4 (Figure 6k), matching the mean phase of WT-specific cycling proteins (Figure 6b). In HETs, this centre is at ZT6 ( Figure 6k), also matching the main peak of HET -specific cycling proteins (Figure 6c). This would support that genotype-specific cycling proteins are influenced by NREM sleep. To further investigate the connection between synaptic proteome dynamics and sleep/wake cycles, we took advantage of a recently published synaptic proteomics study, which uses a sleep deprivation (SD) paradigm, allowing to unambiguously determine proteins that are regulated by sleep/wake cycles 16. Importantly, we found a highly significant overlap between our genotype-shared cycling proteins, mostly peaking during the active period, and those upregulated by wakefulness in that study (Figure 6m). These results strongly suggest that most proteins with preserved rhythms in HET mice are upregulated by wakefulness. On the contrary, there was little overlap between the sleep regulated proteins and the cycling proteins from any group in our study. Thus, to explore the association between genotype-specific cycling proteins and NREM sleep, we collected cortical tissue fr om mice at ZT4 either after 4h of SD or normal rest. We immunoblotted synaptic fractions from these mice for GRIN2B, DLG3 and BAIAP2 (Figure 6n -p), three WT-specific cycling proteins related to Synaptic Stabilization. .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 13 GRIN2B and DLG3 were significantly reduced after SD in WT and BAIAP2 showed a similar non-significant trend (p=0.078). These results indicate that these proteins are upregulated by sleep in WT animals. Of note, SD did not affect the level of these proteins in Syngap1+/- mice (Figure 6n-p), consistent with their loss of cycling behaviour. Together, our findings suggest that the greater disruption of synaptic functions occurring during the light period in Syngap1+/- mice is linked to the preferential peak of genotype-specific cycling proteins during this period. Importantly, while shared cycling proteins are related to wakefulness , genotype-specific ones would be upregulated by NREM sleep . These findings are in line with NREM sleep being the most affected behavioural state in Syngap1+/- mice as observed in the EEG analysis. .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 14

While the importance of NREM oscillations on sleep-dependent memory consolidation is well -stablished3,4,21, their impact onto the proteomic composition of synapses is ignored. Here, through the integrative analysis of EEG recordings and time -resolved synaptic proteomics in a NDD model, we propose that slow -waves and spindles , two brain oscillations characteristic of NREM sleep, would mediate synaptic proteome remodeling during sleep. Importantly, our results indicate that this sleep-dependent remodeling results in a synaptic stabilization, which could serve as the molecular substrate of memory consolidation occurring during sleep. In WT animals 10% of the cortical synaptic proteome is subjected to daily expression changes. Consistent with previous findings18, these dynamics are strictly synaptic, since changes at the whole cortical level are essentially different. Importantly, this reveals that daily proteome dynamics would be driven by synapse -specific mechanisms. As previously shown 18, we found these proteomic dynamics to be linked to sleep/wake cycles. Indeed, proteins peaking in the active period tend to do so when wake density is maximal. Furthermore, these proteins increase their levels with sleep deprivation (SD), which is indicative of a wake-driven upregulation. Likewise, proteins peaking in the rest period do so near the time of highest NREM sleep density and decrease upon SD, pointing to a NREM-driven upregulation. Together, these findings suggest that different brain states impose distinct patterns of synaptic stimulation, leading to specific synaptic proteome changes independently of tissue-wide changes. Remarkably, daily synaptic proteome dynamics result in the temporal segregation of key synaptic functions. Throughout the active period, WT synapses are strongly engaged in signal transduction, in agreement with theoretical frameworks proposing that synaptic signalling increases during wakefulness as a result of experience -dependent neuronal activity23,76,77. Simultaneously, the peak of translation initiation factors would support activity-dependent translation. Later in the active period, protein degradation and folding pathways appeared, suggesting that sustained synaptic activity during this period leads to molecular stress. This is in line with prolonged wakefulness activating the unfolded-protein response (UPR)69–71. Towards the dark–light transition, the upregulation of proteins involved in cell-adhesion and synaptic assembly suggests a period of enhanced synaptic remodelling and formation, which extends into the early light phase. This is accompanied by the upregulation of core PSD scaffolds, glutamate receptors and their regulators, which we interpret as a period of synaptic stabilization. The early light/rest period is also characterised by mitochondrial and metabolic events. Especially notable was the coordinated o verexpression of mitochondrial ribosome proteins at ZT3, which is indicative of a window for synaptic mitochondria regeneration. Consistently, SD leads to mitochondrial deficits 78–81, and mitochondrial metabolism c ontributes to sleep -wake .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 15 regulation79,82. Finally, the second half of the light/rest period presented enrichment of intracellular transport, in line with transcriptomics studies showing that sleep upregulates these processes83–85. Overall, our findings in WT animals support the notion that wakefulness -related signalling prime synapses for later stabilization during sleep77. Indeed, we observed that synaptic functions are nicely segregated along th e day, supporting information processing during the active period and leading to synaptic remodeling or formation towards the dark -to-light transition. These events would culminate, during the first hours of the rest period, with a process of synaptic stabilization. Importantly, this orchestrated organization of synaptic functions across the day was severely disrupted in Syngap1+/- synapses, with all SFMs showing great temporal overlap. This functional disorganization was more pronounced during the light period , when animals are mostly asleep . Accordingly, WT cycling proteins upregulated by wakefulness were better preserved in Syngap1+/- mice than those regulated by sleep , pointing to a sleep -dependent proteomic dysregulation in Syngap1+/- mice. However, the amount of wake, NREM and REM sleep was equivalent between genotypes, thus sleep architecture per se is unlikely to explain the proteomic differences observed. Sleep alterations in Syngap1+/- mice were essentially restricted to the two major oscillations characteristic of NREM sleep, slow-waves, which were enhanced, and spindles, which were reduced and temporally disorganised , suggesting that these oscillations mediate the remodelling of the synaptic proteome. Indeed, the alterations in NREM oscillations are well suited to explain the proteomic dysregulation found during the rest period, since both have been proposed to play important, yet opposed, roles in synaptic plasticity during sleep30. While SWs facilitate synaptic downscaling 24–26, spindles promote synaptic potentiation trough Ca 2+ influx into dendrites27–30. Therefore, increased SWs combined with reduced density and infra- slow organization of spindles, may underlie a shift from synaptic stabilization toward s synaptic weakening. This shift is reflected by the enrichment in Syngap1+/- synapses of processes related to endocytosis, neurotransmitter receptor internalization, and kinase/phosphatase signaling associated with LTD and homeostatic downscaling 72–75. Thus, our findings are in line with the proposed roles for NREM sleep oscillations in synaptic plasticity, suggesting that spindle -rich NREM sleep promotes synaptic potentiation while SW-rich NREM sleep promotes downscaling30. In addition, the loss of synaptic stabilization in Syngap1+/- mice might affect their capacity to consolidate memories during sleep, impacting their cognitive performance. Disturbances in sleep are more common in children with NDDs than in their neurotypical peers12–14. Furthermore, alterations in NREM oscillat ions have been reported in monogenic NDDs, including Rett 35,86, Angelman 36,87,88, Fragile X 37,89–91, Phelan – McDermid92,93 or Dravet38,94,95. Among these alterations, the decrease in spindle density .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 16 seems to be a common trait, both in human patients and rodent models of this conditions35,36,92–95,37,38,86–91. Noticeably, we recently reported a decrease in spindle activity in SYNGAP1-RD patients33, what underscores the relevance of our findings in the mouse model. This raise s the possibility that dysregulation of synaptic proteome dynamics is a common feature in NDDs coursing with sleep disorders. Further studies directly manipulating specific features of NREM sleep, particularly SWs and spindles, would contribute strengthening the observations linking these oscillations with proteome remodelling. Furthermore, targeting these oscillations may hold therapeutical potential for SYNGAP1 -RD and other NDDs , since the recovery of normal synaptic proteome dynamics might result in improved cognitive performance. Overall, our studies on a NDD-causing gene allows us to propose that NREM oscillations mediate the synaptic proteome remodelling occurring during sleep, which would be linked to the stabil ization of synapses and , potentially, to memory consolidation. Furthermore, we reveal a putative pathological mechanism that could be common to neurological conditions involving alterations in NREM oscillations. .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 17

Animal handling Animal research was done with equal numbers of male and female inbred mice from the conditional Syngap1 rescue line (Syngap1 +/lx-st, Jackson Laboratories, strain code JAX: #029304) in accordance with national and European legislation (Decret 214/1997 and RD 53/2013). Research procedures were approved by the Ethics Committee on Animal Research from the: i) Institut de Recerca Sant Pau (IR -SANT PAU) ii) Centro Andaluz de Biologia Molecular y Medicina Regenerativa (CABIMER) iii) The Herbert Wertheim UF Scripps Institute for Biomedical Innovation & Technology . These procedures were also approved by the Departament de Territori i Sostenibilitat from the Generalitat de Catalunya (approval reference numbers 9,655 and 164.16). Maintenance and experimental procedures were conducted at the animal facilities of the IR -SANT PAU and CABIMER. Mice were housed at a 12h light/dark cycle, at a temperature of 22 ± 1 °C and a relative humidity of 55% ±10%. Mice had access to fresh water and food ad libitum. We used animals of both sexes and 9-14 weeks of age. Mouse Brain Dissection Mice were culled by cervical dislocation minimizing animal suffering. Mouse head was severed with dissecting scissors and placed onto a glass petri dish with a filter paper soaked with chilled 1x PBS (10 mM Na2HPO4, 1.8 mM KH2PO4, 137 mM NaCl, 2.7 mM KCl, pH 7.4). The skin was removed by a cut through the long axis of the head, using 14.5 cm dissecting scissors, whereas the skull and meninges were removed using 11.5 cm Iris scissors and tissue forceps 1:2 (Thermo Scientific). The brain was then extracted using curved Gerald thumb forceps (VWR International, Darmstadt, Germany), rinsed with chilled PBS, and dissected with two spatulas following Spijker (2011)96. Tissue weighted, snap-frozen in liquid nitrogen and stored in a -80ºC freezer (Thermo Fisher Scientific). Mouse Video Recording and Animal Tracking Video recordings were conducted to assess circadian behaviour in HET mice. Four cages were recorded simultaneously using an infrared camera (Basler acA640-750um camera, ref. 106748) positioned above the cages, at a sampling rate of 30Hz. Recordings started at Zeitgeber Time 0 (ZT0; 7:00 a.m.) and lasted for 48 hours. Mice were individually placed in clean cages at the start and remained undisturbed throughout. Lighting conditions matched the 12 h light/dark cycle of the animal facility. Videos were downsampled 60 -fold using ffmpeg (https://www.ffmp eg.org/). Animal tracking was performed with DeepLabCut v3.0 97–99 using the pretrained SuperAnimal - .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 18 TopViewMouse model100. The following body parts were tracked: nose, eyes, ears, head, neck, shoulders, back, hips, and tail. The animal’s centroid was computed from major body parts (excluding the tail), omitting coordinates with a detection likelihood below 0.95. Additionally, to avoid misidentification of body parts, the area covered by the mouse's body was calculated per time point and whenever it was >100 pixels the centroid position was discarded and filled with the next valid position. Distance travelled was calculated as the pixel difference between consecutive centroid coordinates. Activity during light and dark periods was quantified as total distance travelled per periods and compared using a mixed ANOVA, with Genotype as the between -subject factor and Circadian Phase as the within-subject factor. Electrode Implantation and Electroencephalogram (EEG) Recordings Stereotaxic surgeries and EEG recordings were conducted at CABIMER (Seville, Spain). Mice were anesthetized with isoflurane (2–3% induction, 1–2% maintenance; 0.8 L/min O₂ flow), and eight Teflon-coated silver-nichrome wire electrodes (125 µm diameter; A- M Systems, USA) were bilaterally implanted in the frontal cortex (±1 mm lateral, +1 mm anterior to bregma), parietal cortex (±2 mm lateral, 0 mm anterior), and above the dorsal hippocampus (±2.5 mm lateral, −2.5 mm posterior). Two additional electrodes served as ground (−2.0 mm lateral, −4.8 mm posterior) and reference for both hippocampal electrodes (+2.0 mm lateral, −3.5 mm posterior). Front al and parietal electrodes were referenced to the contralateral hemisphere. Electrodes were connected to a pin header linked to a signal amplifier (8406-SE4, Pinnacle Technologies) and secured with dental cement, with the entire assembly encased in acrylic resin. After surgery, mice were allowed to recover for at least 7 days. Recordings were performed on 14 -16 weeks-old freely moving animals using a synchronized video -EEG monitoring system (8400 -K1-SE4, Pinnacle Technologies) equipped with four channels and managed by Sirenia® Acquisition software. Sampling rate was 250 Hz and preamplification gain was set at 100x. Mice were housed individually in cages and monitored for 24-hour sessions. Sleep Architecture Analysis Raw EEG data were visually inspected usi ng SleepSign (Kissei Comtec, Japan) and classified into Wake, NREM, and REM stages in 4-second epochs. Artifacts were marked and excluded to ensure data quality. The proportion of time spent in each vigilance state (Wake, NREM, REM) was calculated for the full recording period and normalized by the total duration of the recording. To assess circadian variation, data were binned according to the light/dark phases. For .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 19 comparisons of sleep patterns over the 24 -hour cycle, data were binned by Zeitgeber Time (ZT). Statistical analyses were performed using mixed ANOVA with Genotype as the between-subject factor and Vigilance State, Circadian Period, or ZT as within-subject factors. Episodes were defined as continuous periods of a given state allowing brief interruptions by other states, with minimum durations of ≥52 s (13 epochs) for Wake/NREM and ≥32 s (8 epochs) for REM. Interruptions ≤4 consecutive epochs (16 s) and <30% of total duration were permitted 27,101. Episodes were detected using a custom -made Python function. The number of episodes were normalized by total recording time and compared between genotypes using independent t-test. Episode stability was assessed with the lifelines Python package 102 using Cox regression models accounting for repeated measures and circadian phase and Kaplan–Meier curves for visualization. Interictal Spike (IIS) Detection and Quantification Raw EEG signals were processed using MNE -Python103. IIS detection was conducted by adapting previously described methods 104. Specifically, the find_peaks function from SciPy was applied to both raw and inverted EEG signals to detect positive and negative peaks, respectively. Final detection criteria were as follows: Prominence ≥ 20 0 µV, Height ≥ 2 × baseline, Width < 200 ms, Inter -spike interval ≥ 100 ms. Baseline was dynamically calculated for each subject and vigilance state as the 99th percentile of the absolute signal amplitude. IIS frequency was quantified as spike density (spi kes/hour) for each animal and compared using mixed -design ANOVA with Genotype as the between-subject factor and Stage as the within -subject factor. Post hoc tests were conducted using pairwise comparisons with Holm correction. To visualize the shape of IIS, segments centred on the peak of each spike (± 300 ms) were extracted and averaged to generate mean waveforms per animal, channel, and polarity (positive, negative). Spectral Power Analysis EEG signals were pre-filtered in SleepSign using a 0.5–30 Hz FIR band-pass filter, and Fast Fourier Transforms (FFT; 1024 points, 4.1 s epochs, Hanning window) were applied, yielding a frequency resolution of ~0.25 Hz. Only power values within 0.5 –30 Hz we re analysed. For each 4-second epoch, relative power was computed by normalizing each frequency bin to total epoch power, then averaged across all epochs of the same state (Wake, NREM, REM). Epochs containing previously detected IIS events were excluded. Data from left and right hippocampi, referenced to the same electrode, were averaged. Frontal electrode was excluded from spectral analyses due to noisy data. Spectral power was compared between genotypes for each vigilance state and channel using .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 20 independent t-tests per frequency bin, with multiple comparisons corrected via False Discovery Rate (FDR). Homeostatic Regulation of Delta Power (Process S Simulation) To investigate the homeostatic regulation of delta power (1 -4 Hz), Process S dynamics were simu lated using a well -established model 55,63,105,106. In this model, Process S increases during Wake and REM according to a saturating exponential function and decays during NREM following an exponential decay function. During Wake/REM: (1) 𝑆!"# = (𝑆! − 𝐿𝐴)𝑒$%!/'% + 𝐿𝐴 During NREM: (2) 𝑆!"# = 𝑈𝐴 − (𝑈𝐴 − 𝑆!)𝑒$%!/'( St is the simulated Process S at time t, dt the time step between consecutive measures (4 seconds), Ti and Td are time constants for increase and decay and UA and LA are the upper and lower asymptotes. The initial value St=0 was also treated as a free parameter. Model parameters (Ti, Td, S t=0) were optimized per animal by minimizing the mean squared error between simulated and smoothed empirical delta power time series. To minimize inter-individual variability, UA and LA were fixed to 250 and 50. Empirical delta power was calculated by collapsing power from 1 –4 Hz frequency bins during NREM epochs. Epochs with power beyond ±2.5 z -score were excluded. Values were normalized within each animal as a percentage of mean NREM delta power across the 24h recording. The temporal profile was then smoothed using a centred rolling average with 1h window, requiring at least 50% valid data points per window. Circadian Modulation of Sigma Frequency Band EEG power was summed across the sigma band (9–16 Hz) for each channel and vigilance state using FFT data, including only NREM epochs. Data were resampled into 1-hour bins and normalized within each animal by dividing each time point by the mean sigma power over 24 hours, centring values around 1. Sigma power was modelled for each animal and channel using a cosine function with a fixed 24-hour period: (3) y(t) = 𝐴 · cos 2 )*! ' − 𝜑4 + 𝐶 + 𝜖 .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 21 where 𝐴 and 𝜑 represent the amplitude and phase (i.e. time of maximal sigma power), 𝐶 the baseline level and 𝜖 residual noise. Model fitting was performed using the curve_fit function from scipy.optimize, with non -linear least squares optimization. Amplitudes were constrained to non -negative values and phases to the interval 0 -2π. Goodness of fit was assessed using the coefficient of determination (R²). Amplitude and R² values were averaged across channels per animal and compared using independent t-test. Phases were also averaged per animal and compared using circular statistics. For each genotype, the circular m ean (μ), resultant vector length (r), and concentration parameter (κ) were computed to quantify the mean phase, phase clustering, and dispersion, respectively. Homogeneity of concentration parameters (κ) between genotypes was first tested using the equal k appa test. Differences in mean phase angles were then evaluated using the Watson–Williams test for circular means or angular randomization test (1,000 permutations) depending on the results of the equal kappa test. Slow-Waves Detection Slow-waves (SWs) du ring NREM sleep were automatically detected using the sw_detect() function from the YASA Python package107. Prior to SWs detection interictal spikes IIS were removed. SWs were def ined by the following criteria: negative peak duration 125 ms –1 s, amplitude 1 –3 z -units; positive peak duration 125 ms –1 s, amplitude 0.5–3 z-units; and peak-to-peak amplitude 2.5–6 z-units. For each animal and channel, SW counts were normalized to total NREM time to compute SW density. Peak- to-peak amplitude, duration, and slope were extracted for each animal and compared between genotypes using t -tests, Welch’s t -tests, or Mann –Whitney tests as appropriate. Spindle Detection Spindle detection during NREM sleep was performed using the spindles_detect() function from the YASA Python package, based on the A7 algorithm 108. The detection combined three criteria: 1) the ratio of sigma (9–16 Hz) to broadband (1–30 Hz) power, thresholded at the 75th percentile; 2) the Pearson correlation (r ≥ 0.65) between sigma- and broadband-filtered signals and 3) the moving root mean square of the sigma-filtered signal, thresholded at 2.5 SD above the mean. Only events lasting 0.4–3 s were retained109, and spindles occurring within 500 ms were merged. For o ne WT and two HET animals no spindles were detected, these were excluded from all subsequent analyses. For each animal, the density (spindles per minute), mean amplitude, mean duration and mean number of oscillations were .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 22 calculated and compared between genotypes using t -tests, Welch’s t -tests, or Mann – Whitney tests as appropriate. Spindle coheren ce between channels was estimated as the proportion of overlapping events between channels (≤0.5 s window) using compare_channels(). Infra-Slow Oscillation Analysis of Sigma Power Infra-slow fluctuations in sigma power during uninterrupted long NREM episodes (≥96s) were analysed as described previously 58,110. For each channel, sigma power was estimated using a Morlet wavelet transform (tfr_array_morlet, MNE) with 0.2Hz frequency resolution and 4-cycle wavelets. Power was collapsed across frequencies per timepoint, mean-centred, and resampled at 10Hz. For visualization, the 10Hz resampled sigma signal was low-pass filtered (FIR, order 100, cutoff 0.025 Hz). The resulting sigma signal was transformed via FFT to obtain the infra-slow power spectrum (0–0.12 Hz). Each spectrum was normalized by (1) episode length and (2) mean power across 0 –0.12 Hz frequencies. Because different episode lengths result in different frequency resolution, the normalized spectra were binned into 0.005 Hz -wide bins and averaged ac ross episodes. Spectra were further normalized by subtracting mean power in the baseline plateau (0.08–0.12 Hz). The relative power within the 0.01– 0.04 Hz band was collapsed and compared with an independent t-test. Microarousal Analysis and Phase Coupling to Infra-Slow Oscillations Long NREM episodes (≥96s) containing brief interruptions (≤16 s and <30% of total episode duration) were detected, and infra-slow fluctuations in sigma power were analysed as described above. The instantaneous phase of t he low-pass filtered sigma signal (FIR, order 100, cutoff 0.025 Hz) was extracted using the Hilbert transform 61. Phases were circularly aligned such that sigma peaks corresponded to 0° and troughs to 180°. The phase of each microarousal (MA) onset within NREM episodes was interpolated from the sigma phase time series. For each animal, mean preferred phase (μ) and phase concentration (r) were computed using circular statistics, and phase distributions were tested for non -uniformity with the Rayleigh test. MA density was calculated as the number of MAs per hour of NREM sleep and c ompared between genotypes using independent t-tests. Tissue Fractionation and Purification of Synaptic Fractions Preparation of synaptic fractions followed a previously described protocol62. Tissue was homogenized (9:1 v/w) using 1 mL glass -glass tissue grinders (Borosilicate Dounce .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 23 homogenizer, Wheaton) with 20 strokes on ice in homogenization buffer (HB: 50 mM Tris, 0.32 M Sucrose, 5 mM EDTA, 1 mM EGTA, 20 uM ZnCl2, 50 mM NaF, 1 mM Na - Orthovanadate, 2.5 mM Na -Pyrophosphate, 1 ug/ml Aprotinin, 1ug/ml Leupeptine, 1/2500 PMSF, pH 7.4). The homogenate was centrifuged at 1,400×g for 10 minutes at 4ºC using an Eppendorf 5417R centrifuge. The supernatant was retained, and the pellet was re-homogenized twice under the same conditions. 150 μL of the first supernatant were aliquoted to use in MS experiments as a nuclei -depleted homogenate. The combined supernatants from all three rounds of homogenization were centrifuged at 700×g for 10 minutes. The re sulting supernatant was centrifuged at 21,000×g for 30 minutes at 4ºC and the pellet resuspended in HB and layered onto a 0.85/1/1.2 M sucrose gradient in a 11×60 mm tube (Beckman Coulter, CA, USA). The gradient was centrifuged at 82,500×g for 2 hours at 4ºC using a Sw60 Ti rotor (Beckman Coulter) in an ultracentrifuge (Optima L -90K Ultracentrifuge, Beckman Coulter). Synaptosomes were collected from 1-1.2M interface, diluted to 10% sucrose with 50 mM Tris buffer (pH 7.4) and centrifuged at 21,000×g for 30 minutes at 4ºC using an Eppendorf 5117R rotor. The resulting pellet was resuspended in 50 mM Tris buffer (pH 7.4) containing 0.5% Triton X-100, incubated on ice for 15 minutes and centrifuged again (21,000×g, 30 min, 4ºC). The Triton X-100 insoluble fraction was used for MS experiments. Homogenate and synaptic fractions were extracted in 5% SDS and 50 mM Tris -HCl (pH 8) for 10 minutes at 37°C with gentle shaking, then centrifuged at 10,000×g for 1 minute, and supernatants was collected. Protein was quantifie d using a micro -BCA assay (Thermo Fisher Scientific, Waltham, MA, USA) following the standard microplate procedure. Absorbance measurements were taken at 560 nm using the xMark™ Microplate Absorbance Spectrophotometer (Bio-Rad). Sample Preparation for MS Mass spectrometry (MS) sample preparation was carried out using a modified version of the DNA micro spin column suspension trapping protocol described by Thanou et al., 2023111. 20 µg of protein was combined with S -Trap lysis buffer (5% SDS, 50 mM Tris - HCl, pH 8.5) supplemented with 5 mM tris(2-carboxyethyl)phosphine (TCEP) and 20 mM chloroacetamide (CAA). Reduction and alkylation were performed sequentially, with samples incubated at 95°C for 15 minutes followed by 55°C for 15 minutes in a thermomixer set at 1500 RPM. After treatment, samples were acidified to a final phosphoric acid concentration of 1.1% (from a 12% stock solution) and mixed with six volumes of binding/washing buffer (90% methanol, 100 mM Tris -HCl, pH 8). The mixture was then loaded onto a plasmid DNA micro column (HiPure, Magen Biotechnology). Proteins were retained on the column by centrifugation at 1400×g for 1 minute, followed by four washes with binding/washing buffer. .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 24 Columns were subsequently transferred to new LoBind tubes (Eppendorf), where digestion was carried o ut by adding 0.8 µg of Trypsin/Lys -C (Promega) in 50 mM NH₄HCO₃. Samples were incubated overnight at 37°C in a humidified incubator. Tryptic peptides were sequentially eluted using 50 mM NH₄HCO₃, 0.1% formic acid, and 0.1% formic acid in acetonitrile. The collected peptides were dried using a SpeedVac and stored at −80°C until further analysis. LC-MS Analysis Each tryptic digested sample was reconstituted in 0.1% formic acid, and peptide concentration was measured using a tryptophan fluorescence assay112. A total of 150 ng of peptides was loaded onto an Evotip Pure (Evosep) before chromatographic separation. Peptide separation was performed using the Evosep One liquid chromatography system following the standardized 30 samples per day method. The separation was carried out on a 15 cm × 150 μm reverse-phase column packed with 1.5 µm C18 beads (EV1137, Evosep), which was connected to a 20 µm ID ZDV emitter (Bruker Daltonics). Peptides were introduced into the timsTOF HT mass spectrometer (Bruker Daltonics) via electrospray ionization using a CaptiveSpray source. The mass spectrometer was operated with the following settings: scan range of 100–1700 m/z, ion mobility range of 0.65–1.5 Vs/cm², ramp time of 100 ms, and accumulation time of 100 ms. Collision energy was dynamically adjusted, decreasing linearly with inverse ion mobility from 59 eV at 1.6 Vs/cm² to 20 eV at 0.6 Vs/cm². Data were acquired in dia-PASEF mode, with each acquisition cycle lasting 1.38 seconds and consisting of one MS1 full scan followed by 12 dia-PASEF scans. Each dia-PASEF scan included two isolation windows, collectively covering an m/z range of 300–1200 and an ion mobility range of 0.65–1.50 Vs/cm². The placement of dia-PASEF isolation windows was optimized using the py -diAID tool 113. Ion mobility calibration was automatically performed at the start of each sample using the following calibrant ions (m/z, 1/K₀): 622.029, 0.992 Vs/cm²; 922.010, 1.199 Vs/cm²; and 1221.991, 1.393 Vs/cm². Peptide Identification and Protein Mapping DIA-PASEF raw data were analysed using DIA-NN 1.8.1114,115. A predicted spectral library was generated in silico from the UniProt mouse reference proteome (SwissProt a nd TrEMBL, including canonical and additional isoforms, release 2023-5). Peptide digestion was simulated using Trypsin/P, allowing up to one missed cleavage. Carbamidomethylation was set as a fixed modification, while oxidation and N -terminal methionine excision (with a maximum of one modification per peptide) were specified .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 25 as variable modifications. Peptide length was restricted to 7 –30 amino acids, precursor charge state was set between 2 –4, and precursor m/z was limited to 280 –1220. Both MS1 and MS2 mass accuracy were set at 15 ppm, with a scan window of 9. The double- pass mode and match-between-runs feature were enabled, with protein isoforms used for protein inference. All other parameters remained at default settings. For downstream analysis, MS -DAP 1.0.8116 was used to process the DIA -NN output. Peptide-level filtering was set to retain only peptides that were confidently dete cted in at least 75% of samples per group. Peptide abundance values were normalized using the VSN (Variance Stabilizing Normalization) algorithm, followed by mode -between normalization at the protein level. One outlier sample was identified in the quality control analysis generated by the MS -DAP report in the synaptic fractions dataset and two in the Homogenate dataset. These outliers were excluded from further analyses. Additionally, for each protein, values within a specific ZT group and genotype were cleaned using the Interquartile Range (IQR) method. The IQR was calculated as the difference between the 1st (Q1) and 3rd (Q3) quartile of the data and any values falling below Q1 – 1.5·IQR or above Q3 + 1.5·IQR were removed. Model Selection Analysis for Temporal Regulation Identification To investigate rhythmic protein expression across genotypes over time we implemented a model selection framework63 with 7 different models: ● Null Model: Assumes constant expression over time (no rhythmicity). (4) y(t) = 𝐶 + 𝜖 ● Shared Cosine Model: Captures oscillatory patterns with shared amplitude and phase between genotypes. Follows eq. 3. ● WT-Specific Model: Assumes oscillatory pattern in WT but constant expression in HET. (5) 𝑦(𝑡, 𝐺) = ;𝐴 · cos 2 )*! ' − 𝜑4 + 𝐶 + 𝜖, 𝑖𝑓 𝐺 = 0 𝐶, 𝑖𝑓 𝐺 = 1 ● HET-Specific Model: Assumes constant expression in WT but oscillatory pattern in HET. (6) 𝑦(𝑡, 𝐺) = ; 𝐶, 𝑖𝑓 𝐺 = 0 𝐴+,' · cos 2 )*! ' − 𝜑+,' 4 + 𝐶 + 𝜖, 𝑖𝑓 𝐺 = 1 .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 26 ● Genotype-Dependent Amplitude Model: Fits different amplitudes for WT and HET while sharing a common phase. (7) 𝑦(𝑡, 𝐺) = @ 𝐴 · cos 2 )*! ' − 𝜑4 + 𝐶 + 𝜖, 𝑖𝑓 𝐺 = 0 𝐴+,' · cos 2 )*! ' − 𝜑4 + 𝐶 + 𝜖, 𝑖𝑓 𝐺 = 1 ● Genotype-Dependent Phase Model: Fits different phases for WT and HET while keeping amplitude constant. (8) 𝑦(𝑡, 𝐺) = @ 𝐴 · cos 2 )*! ' − 𝜑4 + 𝐶 + 𝜖, 𝑖𝑓 𝐺 = 0 𝐴 · cos 2 )*! ' − 𝜑+,'4 + 𝐶 + 𝜖, 𝑖𝑓 𝐺 = 1 ● Fully Genotype-Dependent Model: Allows both amplitude and phase to vary by genotype. (9) 𝑦(𝑡, 𝐺) = @ 𝐴 · 𝑐𝑜𝑠 2 )*! ' − 𝜑4 + 𝐶 + 𝜖, 𝑖𝑓 𝐺 = 0 𝐴+,' · 𝑐𝑜𝑠 2 )*! ' − 𝜑+,' 4 + 𝐶 + 𝜖, 𝑖𝑓 𝐺 = 1 In all cases, the period (T) was fixed to 24 h, 𝐶 represents the baseline expression level, and ϵ residual noise. 𝐴 and 𝜑 represent the amplitude and phase (i.e. time of maximal abundance) derived from WT fitting. In some equations, these parameters are shared with HET, while in others the parameters 𝐴+,' and 𝜑+,' were introduced to account for genotype differences. In those cases, the genotype variable (G) was encoded as G=0 for WT and G=1 for HET mice. Prior to model fitting, protein expression values (log₂-transformed) were normalized by subtracting the genotype specific mean to remove baseline differences. Model fitting was implemented using non-linear least squares optimization (curve_fit, scipy.optimize). Amplitudes were constrained to non-negative values and phase parameters restricted to 0-2π. Model fits assumed independent Gaussian errors with constant variance. For each protein, the sum of squared residuals was used to compute the Bayesian Information Criterion (BIC; Astropy function bayesian_info_criterion_lsq). Schwarz weights were derived from BIC values to quantify model support, with the highest-weight model defining the best fit63. To refine detection of genotype-specific rhythms, proteins initially classified as non- rhythmic (Null model) but showing a WT- or HET-specific cosine model as the second- best fit were reanalysed using only data from the corresponding genotype. BIC-based Schwarz weights were recalculated, and proteins were classified as rhythmic in WT or HET when the genotype-specific cosine outperformed the Null model. .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 27 Decile-Based Enrichment Analysis of Subthreshold Trends To determine whether genotype-specific proteins reflect real biological changes instead of merely subthreshold rhythms in the other genotype, cosine models were independently fitted to expression profiles from WT and HET animals using constrained non-linear least squares ( eq. 3). For each genotype, the R² was computed as a quantitative measure of goodness-of-fit and used to rank proteins (excluding genotype- specific rhythmic proteins) and divide into deciles. The distribution of WT - and HET - specific rhythmic proteins across R² deciles of the opposite genotype was then assessed using a hypergeometric test to evaluate enrichment in particular deciles. This approach allowed us to test whether proteins rhythmic in one genotype tend to exhibit some degree of oscillatory behavio ur in the other genotype but fail to reach statistical significance due to increased noise or variability. As a positive control, proteins rhythmic in both genotypes (shared, phase -shifted, or amplitude -differential) were analysed similarly. Pathways and GO Terms Enrichment Analysis To explore the biological relevance of the rhythmic protein subsets identified by model selection, Pathway and Gene Ontology (GO) enrichment analyses were performed using the pathfindR package in R117, following a previously published approach62. PathfindR first builds a Protein Interacting Network (PIN) from all differentially expressed (DE) molecules (genes/proteins), or oscillatory proteins in this case, using a protein–protein interaction (PPI). A high -confidence PPI network was c onstructed by integrating data from two complementary sources: STRING v12.0 and BioGRID v4.4.237. STRING interactions for Mus musculus and Homo sapiens with combined confidence ≥0.9 were retrieved, while BioGRID physical interactions were filtered to retai n only same-species pairs. Protein identifiers were mapped to gene symbols using biomaRt118,119, redundant and low -quality entries (self -interactions, reciprocal duplicates) were removed, and homologous mappings betwee n species were standardized. The final integrated PPI network contained 972,209 unique interactions. Active subnetworks were then identified using a greedy search algorithm constrained to a maximum depth of 1, thereby including only direct neighbours of oscillatory proteins. Subnetworks with ≥10 molecules were retained and overlapping ones (≥50% shared genes) were filtered to keep the highest-scoring network. Pathway enrichment was then performed for each retained subnetwork using a one -sided hypergeometric test, with the full PIN used as background. Pathways investigated with pathfindR were taken from MSigDB ( https://www.gsea- msigdb.org). From the C2 collection, only REACTOME, WikiPathways, and KEGG term s were used (3,224 gene sets); from the C5 collection, all GO Biological Process (BP), .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 28 Cellular Component (CC), and Molecular Function (MF) terms were included (10,454 sets). As the greedy search is stochastic, the analysis was repeated 50 times, retaining only terms enriched in ≥13 iterations (>25% occurrence). Additionally, terms containing <3 proteins were discarded. To avoid biologically irrelevant enrichments in our context, terms related to viral infections (HIV, SARS-CoV-2) were also excluded. Finally, to reduce complexity and facilitate interpretation, enriched terms were clustered using pathfindR’s cluster_enriched_terms() function, and only the most representative term from each cluster was reported. This was done separately for Pathways and for each GO sub-ontologies (BP, CC, MF). Temporal Coherence Filtering of Pathway and GO Terms To incorporate temporal synchronization of proteins within enriched Pathways and GO Terms, the circular standard deviation (Std) of protein peaks within each term wer e analysed. A threshold of 3.13 circular Std was applied to identify pathways in which ≥80% of proteins oscillated within an 8 -hour window centred around the mean ( ±4 h), assuming a normal distribution. This cutoff corresponds to twice the 4 -hour sampling interval, ensuring detection of pathways with tightly synchronized oscillations while accounting for biological and technical variability. Sleep Deprivation. Sleep deprivation (SD) was achieved by novel environmen t and novel object exposure. Animals were transfer to a novel cage containing unfamiliar objects at ZT0 . The procedure was continuously monitored by an experimenter to ensure the maintenance of wakefulness throughout the deprivation period. If behavioural signs of sleep were observed, additional novel objects were introduced to re-engage exploratory activity. If animals continued to display sleep-like behaviour, mild sensory stimulation was applied, including gentle tapping on the cage or disturbance of the bedding material . At ZT4 animals were sacrificed and cortices were collected for further analysis. Immunoblot analysis After isolation of synaptic fraction s from cortical tissue , protein concentrations were quantified using a micro bicinchoninic acid (BCA)-based protein assay kit (Thermo Fisher Scientific, Waltham, MA, USA), mixed with 10xLSB (500 mM Tris, 20% (w/v) SDS, 10% (v/v) beta-mercaptoethanol, 0.1% (w/v) bromophenol blue, pH 7.4) and boiled at 95ºC for 5 minutes. The prepare d samples were then boiled at 95ºC for 5 minutes. Aliquots were subsequently generated and stored at –80ºC until further use. Proteins were separated by electrophoresis using Fast -Casting TGX Stain -Free gels (Bio -Rad) in a .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 29 vertical Mini -PROTEAN system kit (Bio-Rad). Following electrophoresis, the polyacrylamide gels were activated under UV light for 2 minutes using a ChemiDoc Imaging System (Bio -Rad), Next, transfer was conducted in a vertical Mini -PROTEAN system kit (Bio-Rad), using a custom-made transfer buffer (48 mM Tris, 39 mM Glycine, 0.04% (w/v) SDS, 20% (v/v) Methanol, pH 9 ), at a constant voltage of 100 V for 90 minutes. PVDF membranes were then fixed following in pure acetone at 4ºC for 30 minutes with continuous agitation and dried for 30 minutes at 60ºC120. Following fixation, membra nes were reactivated in methanol and washed in TBS -T buffer ( 50 mM Tris, 150 mM NaCl, 1% (v/v) Tween -20, pH 7.4 ) for 5 minutes. The membranes were then blocked for 15 minutes using a commercial solution (EveryBlot Blocking Buffer, Bio -Rad, #12010020) and i ncubated with the primary antibody (1:500 goat anti-Baiap2 (ab15697 Abcam); 1:1000 mouse anti-Grin2b (610417 BD Bioscience); 1:1000 rabbit anti-Dlg3 18036-1-AP ProteinTech)) diluted in TBS-T for either 2 hours at room temperature (RT) or overnight at 4ºC. Following incubation, membranes were washed three times for 5 minutes each with TBS -T and then incubated for 1 hour at RT with a secondary antibody conjugated to an infrared fluorophore (1:7500 IRDye® 680RD anti-Rabbit (9226-68073 Licor-Bio), IRDye® 8 00CW anti-Mouse (9226-32212 Licor-Bio) or IRDye® 800CW anti -Goat (9226 -32214 Licor -Bio)). Membranes were next washed three more times for 5 minutes each with TBS -T, followed by a final wash in TBS. Membranes were imaged using an LI-COR 9120 Odyssey or Odyssey M Infrared Imaging System for infrared fluorescence detection. Band intensity was analysed using Image Studio Lite v5.2 (Li -Cor), and total protein quantification previously obtained was used for normalization. Real Time – quantitative Polymerase Chain Reaction (RT-qPCR) For RNA experiments, nuclear fractions obtained during the brain fractionation protocol were used. RNA was extracted using the RNeasy Lipid Tissue Mini Kit (Qiagen) following the manufacturer’s recommended protocol. The RNA integrity n umber (RIN) was measured using an Agilent 2100 Bioanalyzer, quantified using a Nanodrop 2000 (ND - 200, Thermo Scientific) and transcribed into cDNA. To quantify gene expression, a total of 100 ng of cDNA was used for each reaction on a 384 -well plate (Therm o Fisher, AB1384). The reaction mixture included 2 μL of 50 ng/ μL of cDNA, 5 μL of PowerUp SYBR Green Master Mix (Applied Biosystems™), 300 nM of specific primers (see Table 2), and DNA- and RNA-free H₂O, making up a final volume of 10 μL. The plate was then analyzed using a 7900HT Fast Real -Time PCR System with 384 -Well Block Module (Applied Biosystems™) in combination with the software Sequence Detection System v2.4 Enterprise Edition (SDS 2.4; Applied Biosystems™). All samples were analyzed in triplicate for the gene of interest and the reference gene Gapdh, which was used to normalize values gene expression following the equations: .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 30 (10) ∆𝐶𝑡 = 𝐶𝑡!-./0! /020 − 𝐶𝑡3456+ (11) 𝑅𝐸 = 2∆8! For each sample, the Mean and Standard Deviation (Std) of the Ct values within technical replicates were calculated. Whenever the Std exceeded 0.4, Z -scores were calculated, and the replicate with the highest value was removed to eliminate outliers resulting from technical errors. .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 31 Data Availability All the data generated by the bioinformatics analysis performed in this manuscript can be found in the supplementary tables. Mass spectrometry proteomics data has been deposited to the ProteomeXchange Consortium via the PRIDE partner repository 121 with the dataset identifier PXD073485. Code Availability All custom-made code is available from GitHub: https://github.com/Molecular-Physiology-of-the- Synapse/synapse_sleep_2026/releases/tag/v1.0 .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 32

1. Campbell SS, Tobler I. Animal sleep: A review of sleep duration across phylogeny. Neurosci Biobehav Rev. 1984;8(3):269-300. doi:10.1016/0149-7634(84)90054-X 2. Zimmerman JE, Naidoo N, Raizen DM, Pack AI. Conservation of sleep: insights from non -mammalian model systems. Trends Neurosci . 2008;31(7):371 -376. doi:10.1016/j.tins.2008.05.001 3. Klinzing JG, Niethard N, Born J. Mechanisms of systems memory consolidatio n during sleep. Nat Neurosci. 2019. doi:10.1038/S41593-019-0467-3 4. Brodt S, Inostroza M, Niethard N, Born J. Sleep —A brain-state serving systems memory consolidation. Neuron. 2023;111(7):1050 -1075. doi:10.1016/J.NEURON.2023.03.005 5. Walker MP, Stickgo ld R. Sleep, memory, and plasticity. Annu Rev Psychol . 2006;57:139-166. doi:10.1146/ANNUREV.PSYCH.56.091103.070307, 6. Lutz ND, Harkotte M, Born J. Sleep’s contribution to memory formation. https://doi.org/101152/physrev000542024. 2025;106(1):363-483. doi:10.1152/physrev.00054.2024 7. Lüthi A, Nedergaard M. Anything but small: Microarousals stand at the crossroad between noradrenaline signaling and key sleep functions. Neuron. 2025;113(4):509-523. doi:10.1016/J.NEURON.2024.12.009 8. Hauglund NL, Pavan C, Nedergaard M. Cleaning the sleeping brain – the potential restorative function of the glymphatic system. Curr Opin Physiol . 2020;15:1 -6. doi:10.1016/J.COPHYS.2019.10.020 9. Krause AJ, Simon E Ben, Mander BA, et al. The sleep -deprived human brain. Nat Rev Neurosci 2017 187. 2017;18(7):404-418. doi:10.1038/nrn.2017.55 10. Lim J, Dinges DF. A Meta-Analysis of the Impact of Short-Term Sleep Deprivation on Cognitive Variables. Psychol Bull. 2010;136(3):375. doi:10.1037/a0018883 11. Lowe CJ, Safati A, Hall PA. The neurocognitive consequences of sleep restriction: A meta -analytic review. Neurosci Biobehav Rev . 2017;80:586 -604. doi:10.1016/j.neubiorev.2017.07.010 12. Didden R, Sigafoos J. A review of the nature and treatment of sleep disorders in individuals with developmental disabilities. Res Dev Disabil. 2001;22(4):255-272. doi:10.1016/S0891-4222(01)00071-3 13. Ballester P, Richdale AL, Baker EK, Peiró AM. Sleep in autism: A biomolecular approach to aetiology and treatment. Sleep Med Rev . 2020;54:101357. doi:10.1016/J.SMRV.2020.101357 14. Tamir S, Dye TJ, Witt RM. Sleep and Circadian Disturbances in Children With Neurodevelopmental Disorders. Semin Pediatr Neurol . 2023;48:101090. doi:10.1016/J.SPEN.2023.101090 .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 33 15. Freeman D, Sheaves B, Waite F, Harvey AG, Harrison PJ. Sleep disturbance and psychiatric disorders. The Lancet Psychiatry . 2020;7(7):628 -637. doi:10.1016/S2215-0366(20)30136-X 16. Gay SM, Chartampila E, Lord JS, et al. Developing forebrain synapses are uniquely vulnerable to sleep loss. Proc Natl Acad Sci U S A . 2024;121(44):e2407533121. doi:10.1073/PNAS.2407533121 17. Diering GH, Nirujogi RS, Roth RH, Worley PF, Pandey A, Huganir RL. Homer1a drives homeostatic scaling-down of excitatory synapses during sleep. Science (80- ). 2017;355(6324):511-515. doi:10.1126/science.aai8355 18. Noya SB, Colameo D, Brüning F, et al. The forebrain synaptic transcriptome is organized by clocks but its proteome is driven by sleep. Science (80 - ). 2019;366(6462). doi:10.1126/science.aav2642 19. Brüning F, Noya SB, Bange T, et al. Sleep -wake cycles drive daily dynamics of synaptic phosphorylation. Science (80 - ). 2019;366(6462). doi:10.1126/science.aav3617 20. Vyazovskiy V V, Cirelli C, Pfister-Genskow M, Faraguna U, Tononi G. Molecular and electrophysiological evidence for net synaptic potentiation in wake and depression in sleep. 2008;11. doi:10.1038/nn2035 21. Fernandez LMJ, Lüthi A. Sleep spindles: Mechanisms and functions. Physiol Rev. 2020;100(2):805-868. doi:10.1152/PHYSREV.00042.2018, 22. Vyazovskiy V V., Cirelli C, Tononi G. Electrophysiological correlates of sleep homeostasis in freely behaving rats. Prog Brain Res . 2011;193:17 -38. doi:10.1016/B978-0-444-53839-0.00002-8 23. Tononi G, Cirelli C. Sleep and the Pric e of Plasticity: From Synaptic and Cellular Homeostasis to Memory Consolidation and Integration. Neuron. 2014;81(1):12- 34. doi:10.1016/j.neuron.2013.12.025 24. Bartram J, Kahn MC, Tuohy S, Paulsen O, Wilson T, Mann EO. Cortical Up states induce the select ive weakening of subthreshold synaptic inputs. Nat Commun. 2017;8(1):1-11. doi:10.1038/s41467-017-00748-5 25. González-Rueda A, Pedrosa V, Feord RC, Clopath C, Paulsen O. Activity-dependent downscaling of subthreshold synaptic inputs during slow-wave-sleep-like activity in vivo. Neuron. 2018;97(6):1244-1252.e5. doi:10.1016/j.neuron.2018.01.047 26. Kim J, Gulati T, Ganguly K. Competing roles of slow oscillations and delta waves in memory consolidation versus forgetting. Cell. 2019;179(2):514 -526.e13. doi:10.1016/j.cell.2019.08.040 27. Seibt J, Richard CJ, Sigl-Glöckner J, et al. Cortical dendritic activity correlates with spindle-rich oscillations during sleep in rodents. Nat Commun . 2017;8(1):684. doi:10.1038/s41467-017-00735-w 28. Niethard N, Ngo HV V., Ehrlich I, Born J. Cortical circuit activity underlying sleep slow oscillations and spindles. Proc Natl Acad Sci USA . 2018;115(39):E9220 - .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 34 E9229. doi:10.1073/pnas.1805517115 29. Contreras D, Destexhe A, Steriade M. Intracellular and Computational Characterization of the Intracortical Inhibitory Control of Synchronized Thalamic Inputs In Vivo. https://doi.org/101152/jn1997781335. 1997;78(1):335 -350. doi:10.1152/JN.1997.78.1.335 30. Genzel L, Kroes MCW, Dresler M, Battaglia FP. Light sleep versus slow wave sleep in memory consolidation: a question of global versus local processes? Trends Neurosci. 2014;37(1):10-19. doi:10.1016/J.TINS.2013.10.002 31. Berryer MH, Hamdan FF, Klitten LL, et al. Mutations in SYNGAP1 Cause Intellectual Disability, Autism, and a Specific Form of Epilepsy by Inducing Haploinsufficiency. Hum Mutat. 2013;34(2):385-394. doi:10.1002/humu.22248 32. Vlaskamp DRM, Shaw BJ, Burgess R, et al. SYNGAP1 encephalopathy: A distinctive generalized developmental and epileptic encephalopathy. Neurology. 2019;92(2):E96-E107. doi:10.1212/WNL.0000000000006729/VIDEO-2 33. Ribeiro-Constante J, Tristán-Noguero A, Martínez Calvo FF, et al. Developmental outcome of electroencephalographic findings in SYNGAP1 encephalopathy. Front Cell Dev Biol. 2024;12:1321282. doi:10.3389/FCELL.2024.1321282/BIBTEX 34. Hamdan FF, Gauthier J, Spiegelman D, et al. Mutations in SYNGAP1 in Autosomal Nonsyndromic Mental Retardation . N Engl J Med . 200 9;360(6):599-605. doi:10.1056/NEJMOA0805392, 35. Cataldi M, Cordani R, Chiarella L, et al. N°164 – Reduced sleep spindle activity in Rett syndrome: A polysomnographic case -control study. Clin Neurophysiol . 2023;150:e124. doi:10.1016/J.CLINPH.2023.03.162 36. Den Bakker H, Sidorov MS, Fan Z, et al. Abnormal coherence and sleep composition in children with Angelman syndrome: a retrospective EEG study. Mol Autism 2018 91. 2018;9(1):32-. doi:10.1186/S13229-018-0214-8 37. Martinez JD, Wilson LG, Brancaleone WP , et al. Hypnotic treatment improves sleep architecture and EEG disruptions and rescues memory deficits in a mouse model of fragile X syndrome. Cell Rep . 2024;43(6):114266. doi:10.1016/J.CELREP.2024.114266 38. Kalume F, Oakley JC, Westenbroek RE, et al. S leep impairment and reduced interneuron excitability in a mouse model of Dravet Syndrome. Neurobiol Dis . 2015;77:141-154. doi:10.1016/J.NBD.2015.02.016 39. Chen HJ, Rojas-Soto M, Oguni A, Kennedy MB. A synaptic Ras -GTPase activating protein (p135 SynGAP) inhibited by CaM kinase II. Neuron. 1998;20(5):895-904. doi:10.1016/S0896-6273(00)80471-7 40. Kim JH, Liao D, Lau LF, Huganir RL. SynGAP: A synaptic RasGAP that associates with the PSD -95/SAP90 protein family. Neuron. 1998;20(4):683 -691. doi:10.1016/S0896-6273(00)81008-9 41. Komiyama NH, Watabe AM, Carlisle HJ, et al. SynGAP regulates ERK/MAPK .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 35 signaling, synaptic plasticity, and learning in the complex with postsynaptic density 95 and NMDA receptor. J Neurosci . 2002;22(22):9721 -9732. doi:10.1523/jneurosci.22-22-09721.2002 42. Araki Y, Zeng M, Zhang M, Huganir RL. Rapid Dispersion of SynGAP from Synaptic Spines Triggers AMPA Receptor Insertion and Spine Enlargement during LTP. Neuron. 2015;85(1):173-189. doi:10.1016/j.neuron.2014.12.023 43. Araki Y, Rajko vich KE, Gerber EE, et al. SynGAP regulates synaptic plasticity and cognition independently of its catalytic activity. Science (80- ). 2024;383(6686). doi:10.1126/SCIENCE.ADK1291 44. Wang C-C, Held RG, Hall BJ. SynGAP Regulates Protein Synthesis and Homeostatic Synaptic Plasticity in Developing Cortical Networks. PLoS One. 2013;8(12):83941. doi:10.1371/journal.pone.0083941 45. Lee KJ, Lee Y, Rozeboom A, et al. Requirement for Plk2 in Orchestrated Ras and Rap Signaling, Homeostatic Structural Plasticity, and Memory. Neuron. 2011;69(5):957-973. doi:10.1016/j.neuron.2011.02.004 46. Aten S, Kalidindi A, Yoon H, Rumbaugh G, Hoyt KR, Obrietan K. SynGAP is expressed in the murine suprachiasm atic nucleus and regulates circadian -gated locomotor activity and light -entrainment capacity. Eur J Neurosci . 2021;53(3):732-749. doi:10.1111/EJN.15043 47. Nakajima R, Takao K, Hattori S, et al. Comprehensive behavioral analysis of heterozygous Syngap1 kn ockout mice. Neuropsychopharmacol Reports . 2019;39(3):223-237. doi:10.1002/npr2.12073 48. Creson TK, Rojas C, Hwaun E, et al. Re-expression of SynGAP protein in adulthood improves translatable measures of brain function and behavior. Elife. 2019;8. doi:10.7554/ELIFE.46752 49. Sullivan BJ, Ammanuel S, Kipnis PA, Araki Y, Huganir RL, Kadam SD. Low -Dose Perampanel Rescues Cortical Gamma Dysregulation Associated With Parvalbumin Interneuron GluA2 Upregulation in Epileptic Syngap1+/− Mice. Biol Psychiatry . 2020;87(9):829-842. doi:10.1016/J.BIOPSYCH.2019.12.025 50. Tobler I, Borbély AA. Sleep EEG in the rat as a function of prior waking. Electroencephalogr Clin Neurophysiol . 1986;64(1):74 -76. doi:10.1016/0013 - 4694(86)90044-1 51. Huber R, Deboer T, Tobler I. Topography of EEG dynamics after sleep deprivation in mice. J Neurophysiol. 2000;84(4):1888-1893. doi:10.1152/jn.2000.84.4.1888 52. Achermann P, Borbély AA. Mathematical models of sleep regulation. Front Biosci. 2003;8:s683-93. doi:10.2741/1064 53. Huber R , Felice Ghilardi M, Massimini M, Tononi G. Local sleep and learning. Nature. 2004;430(6995):78-81. doi:10.1038/nature02663 54. Vyazovskiy V V, Achermann P, Borbély AA, Tobler I. The dynamics of spindles and EEG slow-wave activity in NREM sleep in mice. Arch Ital Biol . 2004;142(4):511- .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 36 523. 55. Franken P, Dudley CA, Estill SJ, et al. NPAS2 as a transcriptional regulator of non - rapid eye movement sleep: Genotype and sex interactions. Proc Natl Acad Sci U S A. 2006;103(18):7118-7123. doi:10.1073/PNAS.0602006103 56. Lecci S, Fernandez LMJ, Weber FD, et al. Coordinated infraslow neural and cardiac oscillations mark fragility and offlin e periods in mammalian sleep. Sci Adv . 2017;3(2). doi:10.1126/SCIADV.1602026 57. Osorio-Forero A, Foustoukos G, Cardis R, et al. Infraslow noradrenergic locus coeruleus activity fluctuations are gatekeepers of the NREM–REM sleep cycle. Nat Neurosci 2024 281. 2024;28(1):84-96. doi:10.1038/s41593-024-01822-0 58. Osorio-Forero A, Cardis R, Vantomme G, et al. Noradrenergic circuit control of non-REM sleep substates. Curr Biol . 2021;31(22):5009 -5023.e7. doi:10.1016/J.CUB.2021.09.041 59. Kjaerby C, Andersen M, Hauglund N, et al. Memory-enhancing properties of sleep depend on the oscillatory amplitude of norepinephrine. Nat Neurosci 2022 258 . 2022;25(8):1059-1070. doi:10.1038/s41593-022-01102-9 60. Yüzgeç Ö, Prsa M, Zimmermann R, Huber D. Pupil Size Coupling to Cortical States Protects the Stability of Deep Sleep via Parasympathetic Modulation. Curr Biol. 2018;28(3):392-400.e3. doi:10.1016/j.cub.2017.12.049 61. Cardis R, Lecci S, Fernandez LMJ, et al. Cortico-autonomic local arousals and heightened somatosensory arousability during nrems of mice in neuropathic pain. Elife. 2021;10. doi:10.7554/ELIFE.65835 62. Reig-Viader R, del Castillo -Berges D, Burgas -Pau A, et al. Synaptic proteome diversity is shaped by the levels of glutamate receptors and their regulatory proteins. Nat Commun 2025 161 . 2025;16(1):10487 -. doi:10. 1038/s41467-025- 65490-9 63. Hor CN, Yeung J, Jan M, et al. Sleep –wake-driven and circadian contributions to daily rhythms in gene expression and chromatin accessibility in the murine cortex. Proc Natl Acad Sci U S A . 2019;116(51):25773 -25783. doi:10.1073/PNAS.1910590116 64. Honda T, Fujiyama T, Miyoshi C, et al. A single phosphorylation site of SIK3 regulates daily sleep amounts and sleep need in mice. Proc Natl Acad Sci U S A . 2018;115(41):10458-10463. doi:10.1073/PNAS.1810823115 65. Kim SJ, Hotta-Hirashima N, Asano F, et al. Kinase signalling in excitatory neurons regulates sleep quantity and depth. Nat 2022 6127940. 2022;612(7940):512-518. doi:10.1038/S41586-022-05450-1 66. Zhou R, Wang G, Li Q, et al. A signalling pathway for transcriptional regulati on of sleep amount in mice. Nat 2022 6127940 . 2022;612(7940):519 -527. doi:10.1038/S41586-022-05510-6 67. Xu J, Zhou R, Wang G, et al. Regulation of sleep quantity and intensity by long and .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 37 short isoforms of SLEEPY kinase. Sleep. 2022;45(11). doi:10.1093/SLEEP/ZSAC198 68. Li Y, Li C, Liu Y, et al. Sleep need, the key regulator of sleep homeostasis, is indicated and controlled by phosphorylation of threonine 221 in salt -inducible kinase 3. Genetics. 2023;225(1). doi:10.1093/GENETICS/IYAD136 69. Tononi G, C irelli C. Modulation of Brain Gene Expression during Sleep and Wakefulness: A Review of Recent Findings. Neuropsychopharmacol 2001 251 . 2001;25(1):S28-S35. doi:10.1016/s0893-133x(01)00322-0 70. Naidoo N, Giang W, Galante RJ, Pack AI. Sleep deprivation ind uces the unfolded protein response in mouse cerebral cortex. J Neurochem. 2005;92(5):1150-1157. doi:10.1111/J.1471-4159.2004.02952.X, 71. Naidoo N. Roles of endoplasmic reticulum and energetic stress in disturbed sleep. NeuroMolecular Med . 2012;14(3):213 -219. doi:10.1007/S12017 -012-8179- 9/METRICS 72. Bradley CA, Peineau S, Taghibiglou C, et al. A pivotal role of GSK -3 in synaptic plasticity. Front Mol Neurosci . 2012;5(JANUARY 2012):13. doi:10.3389/FNMOL.2012.00013 73. Peineau S, Bradley C, Taghibiglou C, et al. The role of GSK-3 in synaptic plasticity. Br J Pharmacol. 2008;153(Suppl 1):S428. doi:10.1038/BJP.2008.2 74. Seeburg DP, Feliu-Mojer M, Gaiottino J, Pak DTS, Sheng M. Critical Role of CDK5 and Polo-like Kinase 2 in Homeostatic Synaptic Plasticity during Elevated Activity. Neuron. 2008;58(4):571-583. doi:10.1016/J.NEURON.2008.03.021 75. Mansuy IM. Calcineurin in memory and bidirectional plasticity. Biochem Biophys Res Commun. 2003;311(4):1195-1208. doi:10.1016/J.BBRC.2003.10.046 76. Ode KL, Ueda HR. Phosphorylation Hypothesis of Sleep. Front Psychol . 2020;11:2700. doi:10.3389/fpsyg.2020.575328 77. Seibt J, Frank MG. Primed to sleep: The dynamics of synaptic plasticity across brain states. Front Syst Neurosci . 20 19;13:430280. doi:10.3389/FNSYS.2019.00002 78. Zhao H, Wu H, He J, et al. Frontal cortical mitochondrial dysfunction and mitochondria-related β -amyloid accumulation by chronic sleep restriction in mice. Neuroreport. 2016;27(12):916-922. doi:10.1097/WNR.0000000000000631, 79. Hill VM, O’Connor RM, Sissoko GB, et al. A bidirectional relationship between sleep and oxidative stress in Drosophila. PLOS Biol . 2018;16(7):e2005206. doi:10.1371/JOURNAL.PBIO.2005206 80. Rodrigues NR, Macedo GE, Martins IK, et al. Short -term sleep deprivation with exposure to nocturnal light alters mitochondrial bioenergetics in Drosophila. Free Radic Biol Med. 2018;120:395-406. doi:10.1016/J.FREERADBIOMED.2018.04.549 81. Haynes PR, Pyfrom ES, Li Y, et al. A neuron–glia lipid metabolic cycle couples daily sleep to mitochondrial homeostasis. Nat Neurosci 2024 274. 2024;27(4):666-678. doi:10.1038/s41593-023-01568-1 .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 38 82. Sarnataro R, Velasco CD, Monaco N, Kempf A, Miesenböck G. Mitochondrial origins of the pressure to sleep. Nat 2025 6458081 . 2025;645(8081):722 -728. doi:10.1038/s41586-025-09261-y 83. Cirelli C, Gutierrez CM, Tononi G. Extensive and Divergent Effects of Sleep and Wakefulness on Brain Gene Expression. Neuron. 2004;41(1):35 -43. doi:10.1016/S0896-6273(03)00814-6 84. Basheer R, Brown R, Ramesh V, Begum S, McCarley RW. Sleep deprivation - induced protein changes in basal forebrain: Implications for synaptic plasticity. J Neurosci Res. 2005;82(5):650-658. doi:10.1002/JNR.20675, 85. Mackiewicz M, Shockley KR, Romer MA, et al. Macromolecule biosynthesis: A key function of sleep. Physiol Genomics . 2007;31(3):441 -457. doi:10.1152/PHYSIOLGENOMICS.00275.2006, 86. Ammanuel S, Chan WC, Adler DA, et al. Heightened Delta Power during Slow - Wave-Sleep in Patients with Rett Syndrome Associated with Poor Sleep Efficiency. PLoS One. 2015;10(10):e0138113. doi:10.1371/JOURNAL.PONE.0138113 87. Sidorov MS, Deck GM, Dolatshahi M, et al. Delta rhythmicity is a reliable EEG biomarker in Angelman syndrome: a parallel mouse and human analysis. J Neurodev Disord 2017 91. 2017;9(1):17-. doi:10.1186/S11689-017-9195-8 88. Copping NA, Silverman JL. Abnormal electrophysiological phenotypes and sleep deficits in a mouse model of Angelman Syndrome. Mol Autism 2021 121. 2021;12(1):9-. doi:10.1186/S13229-021-00416-Y 89. Hooper AWM, Wong H, Niibori Y, et al. Gene therapy using an ortholog of human fragile X mental retardation protein partially rescues behavioral abnormalities and EEG activity. Mol Ther Methods Clin De v. 2021;22:196 -209. doi:10.1016/j.omtm.2021.06.013 90. Wong H, Hooper AWM, Kang HR, et al. CNS -dominant human FMRP isoform rescues seizures, fear, and sleep abnormalities in Fmr1 -KO mice. JCI Insight . 2023;8(11). doi:10.1172/JCI.INSIGHT.169650 91. Wong H , Hooper AWM, Niibori Y, et al. Sexually dimorphic patterns in electroencephalography power spectrum and autism -related behaviors in a rat model of fragile X syndrome. Neurobiol Dis . 2020;146:105118. doi:10.1016/J.NBD.2020.105118 92. Ingiosi AM, Schoch H, Wintler T, et al. Shank3 modulates sleep and expression of circadian transcription factors. Elife. 2019;8. doi:10.7554/ELIFE.42819 93. Medina E, Schoch H, Ford K, Wintler T, Singletary KG, Peixoto L. Shank3 influences mammalian sleep development. J Neurosci Res . 2022;100(12):2174 -2186. doi:10.1002/JNR.25119;WGROUP:STRING:PUBLICATION 94. Hall JC, Bashir S, Tsuboyama M, et al. EEG band pow er and phase -amplitude coupling in patients with Dravet syndrome. Ann Child Neurol Soc . 2024;2(2):92- 105. doi:10.1002/CNS3.20061 .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 39 95. Riga MS, Pérez-Fernández M, Miquel-Rio L, et al. Scn1a haploinsufficiency in the prefrontal cortex leads to cognitive impairment and depressive phenotype. Brain. 2024;147. doi:10.1093/brain/awae167 96. Spijker S. Dissection of Rodent Brain Regions. Neuromethods. 2011;57:13 -26. doi:10.1007/978-1-61779-111-6_2 97. Mathis A, Mamidanna P, Cury KM, et al. DeepLabCut: markerless pose estimation of user-defined body parts with deep learning. Nat Neurosci. 2018;21(9):1281- 1289. doi:10.1038/S41593-018-0209-Y 98. Nath T, Mathis A, Chen AC, Patel A, Bethge M, Mathis MW. Using DeepLabCut for 3D markerless pose estimation across species and behaviors. Nat Protoc . 2019;14(7):2152-2176. doi:10.1038/S41596-019-0176-0 99. Lauer J, Zhou M, Ye S, et al. Multi -animal pose estimation, identification and tracking with DeepLabCut. Nat Methods 2022 194 . 2022;19(4):496 -504. doi:10.1038/s41592-022-01443-0 100. Ye S, Filippova A, Lauer J, et al. SuperAnimal pretrained pose estimation models for behavioral analysis. Nat Commun. 2024;15(1):1-19. doi:10.1038/S41467-024- 48792-2 101. Grosmark AD, Mizuseki K, Pastalkova E, Diba K, Buzsáki G. REM Sleep Reorganizes Hippocampal Excitability. Neuron. 2012;75(6):1001 -1007. doi:10.1016/j.neuron.2012.08.015 102. Davidson-Pilon C. lifelines: survival analysis in Python. J Open Source Softw . 2019;4(40):1317. doi:10.21105/JOSS.01317 103. Larson E, Gramfort A, Engemann DA, et al. MNE-Python. Zenodo. December 2024. doi:10.5281/ZENODO.14519545 104. Kyle JJ, Sharma S, Tiarks G, Rodriguez S, Bassuk AG. Fast Detection and Quantification of Interictal Spikes and Seizures in a Rodent Model of Epilepsy Using an Automated A lgorithm. Bio-protocol. 2023;13(6):e4632. doi:10.21769/BIOPROTOC.4632 105. Franken P, Tobler I, Borbély AA. Sleep homeostasis in the rat: Simulation of the time course of EEG slow -wave activity. Neurosci Lett . 1991;130(2):141 -144. doi:10.1016/0304-3940(91)90382-4 106. Franken P, Chollet D, Tafti M. The Homeostatic Regulation of Sleep Need Is under Genetic Control. 2001. 107. Vallat R, Walker MP. An open-source, high-performance tool for automated sleep staging. Elife. 2021;10. doi:10.7554/ELIFE.70092 108. Lacourse K, Delfrate J, Beaudry J, Peppard P, Warby SC. A sleep spindle detection algorithm that emulates human expert spi ndle scoring. J Neurosci Methods . 2018;316:3. doi:10.1016/J.JNEUMETH.2018.08.014 109. Kim D, Hwang E, Lee M, Sung H, Choi JH. Characterization of Topographically Specific Sleep Spindles in Mice. Sleep. 2015;38(1):85. doi:10.5665/SLEEP.4330 .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 40 110. Dimitriades ME, Osorio-Forero A, Fattinger S, et al. The Infraslow Fluctuation of Sigma Power During Sleep: Links to Markers of Arousal and Memory Reactivation Across Development. bioRxiv. November 2024:2024.11.06.620875. doi:10.1101/2024.11.06.620875 111. Thanou E, Koopmans F, Pita -Illobre D, et al. Suspension TRAPping Filter (sTRAP) Sample Preparation for Quantitative Proteomics in the Low µg Input Range Using a Plasmid DNA Micro-Spin Column: Analysis of the Hippocampus from the 5xFAD Alzheimer’s Disease Mouse Mo del. Cells. 2023;12(9). doi:10.3390/CELLS12091242, 112. Wis̈niewski JR, Gaugaz FZ. Fast and sensitive total protein and peptide assays for proteomic analysis. Anal Chem. 2015;87(8):4110-4116. doi:10.1021/AC504689Z, 113. Skowronek P, Thielert M, Voytik E, et al. Rapid and In -Depth Coverage of the (Phospho-) Proteome With Deep Libraries and Optimal Window Design for dia - PASEF. Mol Cell Proteomics. 2022;21(9). doi:10.1016/j.mcpro.2022.100279 114. Demichev V, Szyrwiel L, Yu F, et al. dia -PASEF data analysis using FragPipe and DIA-NN for deep proteomics of low sample amounts. Nat Commun. 2022;13(1). doi:10.1038/S41467-022-31492-0, 115. Demichev V, Messner CB, Vernardis SI, Lilley KS, Ralser M. DIA -NN: neural networks and interference correction enable deep pr oteome coverage in high throughput. Nat Methods. 2020;17(1):41-44. doi:10.1038/S41592-019-0638-X, 116. Koopmans F, Li KW, Klaassen R V., Smit AB. MS -DAP Platform for Downstream Data Analysis of Label -Free Proteomics Uncovers Optimal Workflows in Benchmark Data Sets and Increased Sensitivity in Analysis of Alzheimer’s Biomarker Data. J Proteome Res . 2023;22(2):374 -386. doi:10.1021/ACS.JPROTEOME.2C00513 117. Ulgen E, Ozisik O, Sezerman OU. PathfindR: An R package for comprehensive identification of enriched pathways in omics data through active subnetworks. Front Genet. 2019;10(SEP):425394. doi:10.3389/FGENE.2019.00858/XML/NLM 118. Durinck S, Moreau Y, Kasprzyk A, et al. BioMart and Bioconductor: A powerful link between biological databases and microarray d ata analysis. Bioinformatics. 2005;21(16):3439-3440. doi:10.1093/BIOINFORMATICS/BTI525, 119. Durinck S, Spellman PT, Birney E, Huber W. Mapping identifiers for the integration of genomic datasets with the R/ Bioconductor package biomaRt. Nat Protoc . 2009;4(8):1184-1191. doi:10.1038/NPROT.2009.97, 120. Xu J, Sun H, Huang G, et al. A fixation method for the optimisation of western blotting. Sci Reports 2019 91. 2019;9(1):1-10. doi:10.1038/s41598-019-43039-3 121. Perez-Riverol Y, Bai J, Bandla C, et al. The PRIDE database resources in 2022: a hub for mass spectrometry -based proteomics evidences. Nucleic Acids Res . 2022;50(D1):D543-D552. doi:10.1093/nar/gkab1038 122. Koopmans F, van Nierop P, Andres -Alonso M, et al. SynGO: An Evidence -Based, Expert-Curated Knowledge Base for the Synapse. Neuron. 2019;103(2):217 - .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 41 234.e4. doi:10.1016/J.NEURON.2019.05.002 123. Kirkcaldie MTK. Neocortex. Mouse Nerv Syst . January 2012:52 -111. doi:10.1016/B978-0-12-369497-3.10004-4 .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 42 Author Contributions DdCB, OZR, AP, CVR, ACG, DAA, GE, DZ, TV, CR, FK, RVK and BDV performed experiments. DdCB, MAD, JS, GR, ABS and AB designed and supervised all studies. MAD, JS, GR, ABS and AB secured funding. DdCB and AB wrote the manuscript. All authors reviewed and approved the manuscript. Acknowledgments DdCB, OZR, AP, ACG, DAA, GE and AB financial support was provided by: PID2024 - 160538OB-I00 and PID2021-124411OB-I00 ( MICIU/AEI/10.13039/501100011033/ and FEDER-EU), Award AC17/00005 by ISCIII through AES2017 and within the NEURON framework, Ramón y Cajal Fellowship (RYC-2011-08391p), IEDI-2017-00822, Universitat Autònoma de Barcelona and AGAUR (201 7 SGR 1776 and 2021 SGR 01005). BDV and MAD for the financial support provided by: MCIU/AEI/FEDER Grants PID2024 - 162613OB-I00; PID2021 -127044OB-I00. DdCB and ACG thank AGAUR/Generalitat de Catalunya/FEDER, EU for ‘Ajuts per a la contractació de personal investigador novel (FI)’ Ref. 2020FI_B00130 and 2022 FI_B 00996 . DdCB thanks Fundació Universitària Agustí Pedro i Pons for ‘Ajuts per a estudis o projectes fora de Catalunya’ . OZR thanks the Universitat Autonoma de Barcelona for Investigador Predoctoral en formación (PIF-UAB) Ref. B21P0024. DdCB, OZR, AP, ACG, DAA, GE and AB thank the CERCA Programme/Generalitat de Catalunya for institutional support. The study was supported by the generous funding provided by the association of Spanish SYNGAP1-DEE families, SYNGAP1 España. Declaration of Interests The authors declare no competing interests. .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 43 Figure 1. Syngap1+/- mice show preserved circadian regulation of locomotor activity and normal sleep architecture. (a) Mean distance (m) travelled per hour over a 24 -hour period, shown for wild -type (WT n=8, blue) and heterozygous (HET n=8, orange) mice. Shaded areas surrounding the means correspond to the standard error of the mean (SEM). Gray shaded area indicates dark period. Time expressed as Zeitgeber Time (ZT). (b) Mean distance travelled during light and dark periods for WT (blue) and HET (orange) animals. Data presented as mean ± SEM. A Mixed-ANOVA with between-subject variable h e f d g a b c .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 44 ‘Genotype’ and within-subject ‘Circadian Period’ variable revealed a significant effect of the 'Circadian Period' (Light vs Dark; p = 2. 5e-5) and a significant Genotype x Circadian Period interaction (p = 0.027) . Post-hoc analysis was performed using paired t -tests to compare light and dark period within genotypes (WT adj -p=0.047; HET adj -p =0.0044) and independent t-tests to compare between genotypes in each circadian period (Light adj.-p=0.21, Dark adj-p=0.62). Post-hoc contrasts were corrected using Holm correction. (c) Change in activity levels (Dark -Light) for WT and HET animals . Data presented as mean ± SEM. Statistics: independent t-test, p=0.027. (d) Hourly distribution of vigilance states across the 24 -hour cycle for WT ( n=11, blue) and HET ( n=13, orange) mice: Wake (left), NREM (middle), and REM (right). Lines represent the mean and shaded areas indicate the SEM. Gray shaded area indicates dark period. (e) Percentage of total time spent in Wake (W), NREM (NR), and REM (R) sleep across 24 hours for WT (n=11, blue) and HET (n=13, orange) mice. No significantly differences found. Statistics: independent t-tests to compare Wake (p=0.59) and NREM (0.47) and Mann-Whitney to compare REM (p=0.91) (f) Percentage of time spent in each vigilance state (Wake, NREM, REM) during the light or dark periods for each genotype. Paired individual values are shown as dots connected by lines. Boxplots represent group distribution. Statistics: Mixed -ANOVA with ‘Genotype’ as between -subjects variable and ‘Circadian Period’ as within -subject variable for each state. ‘Circadian Period’ was significant in all state (W p= 5.9e-11, NR p=1.13e-11, R p= 1.2e-5). Post -hoc analysis within genotypes was performed using a paired t-tests with Holm correction (W: WT adj -p=7.93e-6, HET adj-p=7.93e-6; NR: WT adj-p=3.2e-6, HET adj-p=2.53e-6; R: WT adj-p=0.001, HET adj-p=0.008). (g) Number of episodes per hour and state: Wake (W), NREM (NR) and REM (R) . Bars represent mean ± SEM across animals. Statistics, unpaired t-tests or Mann-Whitney for non-normal data. ns, non-significant. (h) Kaplan–Meier survival curves for Wake (left), NREM (middle), and REM (right) episodes, reflecting the proportion of bouts that remain active as a function of duration. Shaded areas represent 95% confidence intervals. No significant differences were found between genotypes: Cox-Regression Analysis, Wake: HR= 1.10, p=0.16; NREM: HR= 1.14, p=0.09; REM: HR= 1.15, p=0.09. Significance levels, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 45 a b c d e f g h i jk l .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 46 Figure 2. Altered delta and sigma power reflect alterations in slow-waves and spindles in Syngap1+/- mice. (a) Normalized power spectra during NREM sleep after excluding IIS epochs for both WT (n=11, blue) and HET ( n=13, orange) animals in hippocampal channels (left) and the parietal channel (right). Red lines below the traces indicate frequencies with significant differences (multiple independent t-test with Benjamini-Hochberg (BH) FDR correction, q < 0.05). The shaded areas around the line plots represent the SEM. (b) Average waveform of detected slow-waves across the three EEG channels (EEG1–3) in representative WT (left , blue) and HET (right , orange) animals. Traces represent the mean ± SEM. Traces are aligned to the main negative peak (time 0). (c) Slow wave density (SW/min) during NREM sleep for WT (n=11) and HET (n=13) animals. Dots represent the total number of temporally non -overlapping events across all channels for each animal, normalized to the total artifact-free NREM time. Statistics: Mann-Whitney, p=0.018. (d) Average Z-score normalized peak -to-peak amplitude of detected slow-waves per animal. Statistics: Mann-Whitney, p=0.049. (e) Mean duration of slow-waves in WT (blue) and HET (orange) animals . Statistics: Mann-Whitney, p=0.002. (f) Mean Z-score normalized slope (SD/ms) of slow-waves. Slope was calculated as the peak-to-trough amplitude divided by time between the negative peak and the next 0 crossing. Statistics: Mann-Whitney, p=0.049. (g) Average waveform of detected spindles in a representative WT mouse (left) and HET mouse (right) across three EEG channels. Each trace represents the average signal aligned to the spindle peak (time = 0 s). (h) Spindle density (sp/minute) during NREM sleep for WT (n=10) and HET (n=11) mice. Dots re present the total number of temporally non -overlapping events across all channels for each animal, normalized to the total artifact -free NREM time. Statistics: independent t-test, p=0.015. (i) Mean Z-score normalized amplitude of spindles per animal. Statistics: independent t- test, p=0.02. (j) Mean spindle duration (in seconds) calculated across spindles per animal. Statistics: independent t-test, p=0.021. (k) Mean number of oscillatory cycles per spindle calculated across spindles per animal. Statistics: independent t-test, p=0.023. (l) Mean spindle coherence across EEG channels. Coherence was measured as the mean recall between each pair of EEG channels. Statistics: Mann-Whitney, p=0.018. Boxplots show median, interquartile range (IQR), and 1.5×IQR whiskers; dots represent individual animals. .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 47 Figure 3. Attenuation of the Infra-Slow Oscillation of Sigma Power in Syngap1+/- mice. (a) Representative example of a NREM episode from WT mouse (left) and a HET mouse (right). The upper panels show time –frequency spectrograms (Short -Time Fourier Transform; 1–30 Hz), with power z-scored across time and frequency. The middle panels display the slow fluctuations of sigma power (9 –16 Hz) obtained with a Morlet wavelet and low-pass filtered at 0.025 Hz to isolate infra-slow dynamics. The bottom panels show spindles detected during the represented episode. (b) Group-averaged normalized power spectra of the sigma signal for WT and HET mice after baseline subtraction. Baseline power was estimated from the 0.08–0.12 Hz plateau and subtracted from all frequency bins 58. The spectra were computed using the FFT of a b c d e .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 48 sigma power traces within NREM episodes l onger than 96 seconds. Shaded areas represent SEM. Dashed vertical lines indicate the 0.01 –0.04 Hz frequency range of interest. For statistical comparison the mean power within the 0.01 –0.04 Hz band was calculated for each animal and compared between genot ypes with an independent t - test, p = 0.005. (c) Examples of a NREM episodes from a WT (top) and a HET mouse (bottom). The slow sigma power trace is overlaid with vertical red bars marking microarousal (MA) onset. (d) Circular histograms (20° bins) of the MA phase relative to the σ infra -slow cycle for WT (left, n = 4590 events in 11 animals) and HET (right, n = 4254 events in 13 animals). Zero degrees indicates the σ peak and 180° the trough. Both distributions deviate strongly from uniformity (Rayleigh test: WT p < 10⁻15, HET p < 10⁻15) and cluster around the descending phase / trough of the oscillation. (e) Mean density of MAs (events per minute of NREM) per mouse. Bars represent mean ± SEM; dots are individual animals. Statistical comparison: independent t-test, p=0.016. An outlier was detected through the Z -score method (>2.5 SD) and excluded from the analysis, WT n=11, HET n=12. .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 49 Figure 4. Extensive dysregulation of Homogenate and Synaptic proteome dynamics in Syngap1+/- mice. (a) Schematic representation of the mass spectrometry -based proteomic s workflow used in this study. Cortical samples from WT and Syngap1+/- (HET) mice were collected a Synaptic Purification LC-MS/MS 0 12 16 204 8 24 Tissue Collection every 4h (4 WT & HET mice/time point) Cortical Dissection b c Sorted by WT ΦSorted by HET Φ f d g i e h j postsynapsepostsynapse presynapse presynapse membrane membrane synapticcleft synaptic cleftextrasynaptic extrasynaptic gene count 0 1 5 10 15 20 25 30 35 40 ≥45 .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 50 at 6 different Zeitgeber Times (ZT). Synaptic fractions were isolated from each individual animal. These fractions were finally profiled by mass spectrometry. (b) Distribution of synaptic proteins across sub -synaptic compartments as defined by SynGO CCs122. (c) Top, expression abundance along the day of an example protein (AK5) classified in the shared-regulation group . Dots represent individual data points, while triangles connected with a solid blue (WT) or orange (HET) line represent the mean values for each time point. The shaded coloured region surrounding the mean lines represents the SEM. The horizontal dashed repr esents the baseline (C) parameter. The solid red line represents the fitted cosine function. Shaded grey areas indicate the dark period of the light/dark cycle. The amplitude (A) and phase (Φ) of each protein are shown in the inset. Bottom, h eatmap displa ying normalized expression of all proteins from the shared - regulation group in both genotypes. Z-scored protein abundances are color-coded, with red indicating higher expression and blue indicating lower expression. Proteins (rows) were sorted by their phase. (d) Top, expression abundance along the day of an example protein (CAMKK1) classified in the WT -specific model . Bottom, h eatmap displaying normalized expression of all proteins from the WT-specific group in both genotypes. (e) Top, expression abundance along the day of an example protein (PRICKLE2) classified in the HET -specific group. Bottom, h eatmap displaying normalized expression of all proteins from the HET-specific group in both genotypes. (f) Top, expression abundance along the day of an example protein (EIF4A2) classified in the phase-shifted group. Bottom, h eatmap displaying normalized expression of all proteins from the phase-specific group in both genotypes. (g) Expression abundance along the day of an example protein (MICAL2) classified in the amplitude-shifted group. (h) Expression abundance along the day of an example protein (VPS41) falling in the null model. (i) Pie chart s presenting the proportion of proteins that best fit each model in WT synapses. No proteins were found in the model of amplitude and phase shifting. (j) Pie charts presenting the proportion of proteins that best fit each model in and HET synapses. No proteins were found in the model of amplitude and phase shifting. .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 51 Figure 5. Temporal coherence analysis of enriched pathways reveals altered distribution of synaptic functions along the day. (a) Synaptic function organization along the day in WT animals. (a1) Scatter plot displaying the top 20 time-coherent enriched pathways and GOBP for the dark/active period (left semicircle) and the light/rest period (right semicircle). Each dot represents a term, with the angular position representing the mean phase of its constituent proteins. The distance from the centre (radial depth) reflects fold enrichment (log2-transformed), Signal Transduction Synaptic Assembly Intracellular Trafficking Translation Synaptic Stabilization Endocytosis Protein Stability Metabolism / MitochondriaSynaptic Vesicle a b a.1 a.2 b.1 b.2 Dark/Active Period Light/Rest Period Dark/Active Period Light/Rest Period .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 52 dot size reflects the number of proteins in the term and the colour of dots indicates the adjusted p-value according to the presented scale. Terms were classified in different Synaptic Functional Modules (SFM) and their numeric label and term description (right and left boxes) were colour-coded as stated in the upper legend. (a2) Summary representation of SFMs along the day. Each colour represents a SFM (see legend in the top) and its position in the graph represents the mean (dot) +/ - circular std (line) calculated using the phase of all terms included in such category. (b) Synaptic function organization along the day in HET animals. (b1) Scatter plots displaying the top 20 t ime-coherent enriched pathways and GOBP for the dark/active period (left semicircle) and the light/rest period (right semicircle). (b2) Summary representation of SFMs along the day . Light and dark periods were treated separately to better show the biphasic distribution of their temporal organization. .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 53 Figure 6. Synaptic proteome dynamics are driven by sleep-wake cycles. (a-c) Circular histogram showing the phase distribution of common rhythmic proteins (a), WT-specific oscillatory proteins (b) or HET-specific oscillatory proteins (c). (d-i) Expression abundance along the day of circadian clock genes: (d) BMAL1, (e) CLOCK, (f) CRY2, (g) PER1, (h) CRY1 and (i) PER2. Dots represent individual data points, while triangles connected with a solid blue (WT) or orange (HET) line represent the mean a c ed g h j k l m n f i b op .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 54 values for each time point. The shaded region surrounding the mean lines represents the SEM. The horizontal dashed represents the baseline (C) parameter. The solid red line represents the fitted cosine function. Shaded grey areas indicate the dark period of the light/dark cycle. The amplitude (A) and phase (Φ) of each protein are shown in the inset. (j-l) Circular distribution of Wake (j) , NREM (k) and REM sleep (l) states. Blue (WT) and orange (HET) lines represent the mean proportion of time in the state per hour of recording. Shaded regions surrounding the mean represent SEM. Dots represent the circular mean (μ) of the distribution of individual animals . Arrows represent the mean resultant vector for each genotype , where the length of the vector (r) represents the data aggregation. Wake: WT μ=16.6 and r=0. 89, Rayleigh test p= 1.69e-5; HET μ=18.4 and r=0.95, Rayleigh test p=1.05e-7. NREM: WT μ=4.4 and r=0.69, Rayleigh test p=0.003; HET μ=6.3 and r=0.72, Rayleigh test p=0.005. REM: WT μ=10.7 and r=0.44, Rayleigh test p=0.12; HET μ=11.1 and r=0.64, Rayleigh test p=0.003. (m) Comparison of cycling proteins in our study with those previously published sleep/wake regulated proteins. Left panel: Enrichment of the proteins in each of our models in proteins found upregulated by Wake (red) or Sleep (green) in a previous publication16. Right panel: Phase distribution of overlappin g proteins between each of our models and Wake- or Sleep-upregulated proteins. Each dot represents an individual protein. (n-p) Immunoblot quantification of (n) GRIN2B, (o) DLG3 and (p) BAIAP2 in WT and HET undisturbed (Green, ZT4) or sleep -deprived (Red, ZT4SD) animals collected at ZT4 . Boxplots show median, interquartile range (IQR), and 1.5×IQR whiskers; empty dots represent individual animals outside the IQR. Statistics: independent t-test. GRIN2B: WT p= 0.037, ZT4 n=12, ZT4SD n=12; HET p=0.45, ZT4 n=12, ZT4SD n=12. DLG3: WT p= 0.037, ZT4 n=9, ZT4SD n=9; HET p=0. 86, ZT4 n=9, ZT4SD n=9. BAIAP2: WT p= 0.0 6, ZT4 n=9, ZT4SD n=9; HET p=0.75, ZT4 n=9, ZT4SD n=9. .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 55 Figure S1. Video-tracking of locomotor activity, EEG experimental set-up and temporal distribution of sleep/wake states. (a) Activity patterns of WT (blue; top two rows) and HET (orange; bottom two rows) mice over a 24 -hour period as measured by the distance travel led (m) per hour. Grey shaded areas indicate the dark period of the light/dark cycle. (b) Electrode implantation scheme showing the stereotaxic placement of 8 electrodes resulting in 4 EEG channels: frontal cortex (black), parietal cortex (orange), and bilateral hippocampus (green and yellow). Adapted from: Kirkcaldie, 2012123. (c) Representative hypnogram from a wild -type (WT) mouse displaying manual scoring of vigilance states (Wake, NREM, REM) across the light (top) and dark period (bottom). Ch1 (hipp. R) Ch2 (hipp. L) Ch3 (parietal) Ground

Ch3-4- Ch4 (frontal) b c a .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 56 Figure S2. Detection and analysis of Interictal Spikes (IIS). (a) Example of IIS detection performance in a 10-minute period with high IIS activity on a HET mouse. The blue line represents the 99th percentile of the absolute amplitude of the signal within each behavioural state (baseline), while the red line represents the threshold for spike detection (2·baseline). Detected spikes are marked by red dots. (b, c) Average waveforms of positive (left) and negative (right) spikes detected in a WT (b) and HET (c) animal for a given channel. Spikes are aligned to their peak (time 0). Shaded areas represent the standard deviation. The total number of detected events for each category is indicated above each plot. (d) Comparison of IIS frequency between genotypes. Data are presented as boxplots with individual data points overlaid. Boxplots show median, interquartile range (IQR), and 1.5×IQR whiskers. Statistics: independent t-test, WT n=11, HET n=13, p=0.001. (e, f) Comparison of IIS frequency between behavioural states in WT (e) and HET (f). Statistics: paired t -test or Wilcoxon signed -rank test depending on data normality , followed by Holm correction. WT n=11, HET n=13. WT: non-significant. HET: W vs NR adj. p-value = 0.003, NR vs R adj. p-value = 0.003, W vs R adj. p-value = 0.006. b a c e fd .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 57 Significant differences are indicated as follows: *p < 0.05, **p < 0.01, ***p < 0.001. .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 58 Figure S3. Power spectra analysis of wake and REM sleep and correlations between NREM oscillations and IIS. (a, b) Normalized power spectra during Wake (a) and REM (b) after excluding IIS epochs for both WT (blue) and HET (orange) animals in hippocampal c hannels (left panel) and parietal channel (right panel). The shaded areas around the line plots represent the SEM. Red lines below the traces indicate frequencies with significant differences (multiple independent t-test with Benjamini-Hochberg (BH) FDR correction, q < 0.05). (c) Correlation between SW density and IIS density in HET mice. Pearson r = 0.43, p = 0.14. (d) Correlation between spindle density and IIS density in HET mice. Pearson r = - 0.74, p = 0.006. a b c d .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 59 Figure S4. Homeostatic regulation of delta and circadian regulation of sigma bands. (a) Representative examples of Process S simulation and comparison with empirical delta power in a WT (left , blue) and a HET (right , orange) animal. Process S (black line) was simulated using established methods with fixed asymptotes (UA = 250, LA = 50). Dark blue (WT) and orange (HET) traces represent the smoothed delta power using a 1h rolling window. Lighter traces correspond to delta power calculated with a 2 min rolling window. The R2 and Pearson's r for each animal are indicated. a eb cd f gh i .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 60 (b, c) Comparison of model fit across genotypes using R2 (b) and Pearson r (c) calculated for each animal. Statistics: independent t-tests: WT n=11, HET n=13, R2 p=0.49, r p=0.68. (d, e) Comparison of optimized time constants Ti (d) and Td ( e) between WT and HET mice. Values are shown as boxplots with in dividual data points overlaid. Statistics: Mann-Whitney U tests: WT n=11, HET n=13, Ti p=0.73, Td p=0.86. (f) Average 24-hour dynamics of sigma power during NREM sleep for WT (blue) and HET (orange) mice in the hippocampal (left) and parietal (right) channels. Data are plotted as mean ± SEM for each genotype. Grey shaded area represents the dark period of the light/dark cycle. (g) Comparison of amplitude of sigma oscillations between genotypes. Statistics: independent t-test, WT n=11, HET n=13, p=0.27. (h) Comparison of the time of peak of the 24 -h sigma power rhythm. Dots outside the circle represent the phase of individual anima ls obtained (blue = WT, orange = HET). Arrows indicate the genotype mean phase, with arrow length proportional to the mean resultant length (r, i.e. phase concentration). Statistics: Watson-Williams Test, WT n=11, HET n=13, p=0.41. (i) Comparison between genotypes of R² values obtained from the cosine fitting . Statistics: independent t-test, WT n=11, HET n=13, p=0.39. Boxplots show median, interquartile range (IQR), and 1.5×IQR whiskers. .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 61 Figure S5. Daily expression dynamics of homogenate proteins. (a-d). Heatmaps displaying normalized expression of cycling proteins from the following expression models: shared -regulation (a), WT-specific (b), Syngap1+/--specific (c) and 0 12 16 2048 0 12 16 2048 WT HET 0 12 16 2048 0 12 16 2048 WT HET 0 12 16 2048 0 12 16 2048 WT HET 0 12 16 2048 0 12 16 2048 WT HET Sorted by WT ΦSorted by HET Φ a b c d ie h f g j .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 62 phase-shift (d). Z-scored protein abundances are color-coded, with red indicating higher expression and blue indicating lower expression. Proteins (rows) were sorted by their phase. (e) Venn diagram showing the overlap between WT cycling proteins in homogenates and synaptic fractions. (f) Scatter plots comparing the phase of WT cycling proteins in both compartments. Dashed lines indicate a ±4-hour window around the identity line. (g) Venn diagram showing the overlap between WT cycling proteins in homogenates and synaptic fractions. (h) Scatter plots comparing the phase of WT cycling proteins in both compartments. Dashed lines indicate a ±4-hour window around the identity line. (i, j) Pie chart showing the proportion of proteins best fit by each of the tested models in homogenates for WTs ( i) and HETs (j). No proteins were classified in the model with simultaneous phase and amplitude shift. .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint 63 Figure S6. Enrichment analysis of genotype-specific rhythmic proteins across R² deciles in the opposing genotype. (a, b) Enrichment analysis of WT -specific cycling proteins in homogenates (a) and synapses (b) across R² deciles in HET-ranked proteins of their respective compartment. (c, d) Enrichment analysis of HET -specific cycling proteins in homogenates (c) and synapses (d) across R² deciles in WT-ranked proteins of their respective compartment. (e-h) Enrichment analysis of cycling proteins in both genotypes (Shared oscillations + Phase-shifted + Amp. Shifted) in R² deciles in WT homogenates (e), WT synapses (f), HET homogenates (g) and HET synapses (h). The red dashed line represents the significance threshold (p = 0.05). The y-axis displays the p-value of the hypergeometric test for each decile, plotted on a logarithmic scale. ab cd ef gh .CC-BY-NC 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 8, 2026. ; https://doi.org/10.64898/2026.04.07.716883doi: bioRxiv preprint