The SIK3-N783Y mutation is associated with the human natural short sleep trait

Significance

A mutation in salt-induced kinase 3 (hSIK3-N783Y) is identified in a human subject exhibiting the natural short sleep duration trait. A mouse model carrying this homologous mutation demonstrates reduced sleep duration, confirming the mutation’s causality to the sleep trait. This mutation leads to decreased SIK3 activity and altered global protein phosphorylation profiles, especially for synaptic proteins. Further data analyses reveal additional kinases that could participate in the modulating network for sleep duration. These findings advance our understanding of the genetic underpinnings of sleep, highlight the broader implications of kinase activity in sleep regulation across species, and provide further support for potential therapeutic strategies to enhance sleep efficiency.

Abstract

Sleep is an essential component of our daily life. A mutation in human salt induced kinase 3 (hSIK3), which is critical for regulating sleep duration and depth in rodents, is associated with natural short sleep (NSS), a condition characterized by reduced daily sleep duration in human subjects. This NSS hSIK3-N783Y mutation results in diminished kinase activity in vitro. In a mouse model, the presence of the NSS hSIK3-N783Y mutation leads to a decrease in sleep time and an increase in electroencephalogram delta power. At the phosphoproteomic level, the SIK3-N783Y mutation induces substantial changes predominantly at synaptic sites. Bioinformatic analysis has identified several sleep-related kinase alterations triggered by the SIK3-N783Y mutation, including changes in protein kinase A and mitogen-activated protein kinase. These findings underscore the conserved function of SIK3 as a critical gene in human sleep regulation and provide insights into the kinase regulatory network governing sleep.

Natural short sleeper (NSS) refers to people who have a lifelong requirement of only 4 to 6 h of daily sleep. Our familial NSS subjects show no adverse effects traditionally associated with chronic sleep loss (1). Furthermore, these people do not desire additional sleep and tend to feel “worse” if they do sleep more. Genes that carry the mutations or variations that are responsible for this familial natural short sleep trait were characterized as short-sleep genes. So far, four short-sleep genes have been identified in human families (2–6), which substantially improved our understanding of the “genetics of sleep health” and provided therapeutic targets to improve sleep efficiency and satisfaction (7). Studying human sleep genes not only expands our knowledge regarding the sleep regulatory network, but also may bring the basic research from mouse models to clinical relevance.

Dynamic phosphorylation of proteins, especially those located at synapses, were found to be coupled with sleep–wake cycles (8–10). Several kinases play important roles in sleep–wake regulation (10). Among them, Sik3 has been established as a critical sleep-regulating gene in mice (11). Studies have elucidated how different genetic manipulations/variations of Sik3 influence sleep in rodent models, particularly sleep duration and sleep need as measured by EEG delta power. The initial mouse model identified was termed “Sleepy,” characterized by a nucleotide substitution leading to exon skipping, which induced increased sleepiness (11). Following this, the Sik3 S551A mouse model was developed, where Serine 551 was replaced by Alanine, resulting in enhanced kinase activity and increased sleep need (12). Mutations equivalent to S551A in SIK1 (S577A) and SIK2 (S587A) also resulted in increased sleep need in mice (13). These mutations, together with those in the Sleepy and S551A models, are classified as gain-of-function. More recent models, such as T221A and the knockout first (kof) model, where Sik3 expression was significantly reduced, represent loss-of-function mutations (14). These models displayed a decrease in sleep need, EEG slow-wave activity, and sleep duration, albeit to varying degrees, supporting mouse Sik3 as a sleep-promoting gene. Yet, direct evidence for SIK3 as a human sleep gene remains undetermined.

In this study, we reported a genetic mutation, SIK3-N783Y, in a subject characterized with naturally short sleep durations. In vitro kinase assay demonstrated that this mutation resulted in decreased kinase activity. This observation aligns with previous findings indicating that SIK3 kinase activity is positively correlated with sleep duration, thereby underscoring the role of this kinase in sleep regulation across species.

Results

The SIK3-N783Y Mutation was Found in One FNSS Family.

In searching for more NSSs and the underlying human sleep genes, we characterized one human subject with a short sleep trait compared to the general population (8 to 8.5 h). At the time of initial participation, the subject was in her 70s, healthy, and had maintained a life-long active lifestyle. While she self-reported sleeping approximately 3 h per day, activity recordings indicated an average of 6.3 h of sleep per night (Fig. 1A). Whole exome sequencing of the subject’s DNA sample revealed more than 500 variants. After DNA variants data analyses, six variants remained including one in the SIK3 (SI Appendix, Table S1). Previously, a point mutation was found in Sik3 from a forward genetic screen for sleep mutants in mice (11). We therefore sought to validate this mutation’s role in sleep. Specifically, this mutation converts an asparagine (N) residue into a tyrosine (Y) at position 783 (SIK3^N783Y) (Fig. 1B and SI Appendix, Fig. S1A). This asparagine (N) residue is conserved among mammals and birds (Fig. 1B). SIK3^N783Y is a rare mutation with a frequency of 6.02 ×10⁻⁵ in the Genome Aggregation database.

Fig. 1.

SIK3-N783Y mutation was identified in a natural short sleep family. (A) Actogram of the subject carrying the SIK3 mutation. (B) SIK3-N783Y is localized to the Proline-rich domain and is conserved among mammals and birds.

The SIK3-N783Y Mutation Led to Dampened Catalytic Activity.

In exploring the functional consequences of the SIK3-N783Y mutation, we focused on its effects on Histone deacetylase 4 and 5 (HDAC4/5), known substrates of SIK3 involved in sleep–wake control (15, 16). Utilizing these substrates, we assessed the kinase activity of the SIK3 N783Y mutant. Full-length SIK3 and HDAC4/5 were expressed and purified from bacteria, and kinase reactions were reconstituted in the presence of ATP. Both wild-type (WT) and mutant SIK3 successfully phosphorylated HDAC4/5, as shown in Fig. 2 A and B. This phosphorylation process was inhibited by the pan-SIK inhibitor YKL-06-062 (17), as indicated in Fig. 2C. Notably, the mutant SIK3 exhibited significantly reduced kinase activity (a loss of function) for both HDAC4 and HDAC5.

Fig. 2.

SIK3-N783Y lead to decreased kinase activity. (A) In vitro kinase assay of recombinant SIK3-WT and SIK3-N783Y with recombinant HDAC4/5 as substrates and immunoblot with HDAC4/5 phosphorylation specific antibody. (B) Quantified results for (A). The phosphosites in HDAC4/5 are shown on the Top. (C) In vitro kinase assay with SIK inhibitor YKL-06-062 as control. (D) Schematic of different SIK3 mutations. (E) In vitro kinase assay similar to (A) but with different mutant forms of SIK3. (F) Quantified results for (E). (G) Representative immunoblots with an anti-SPP antibody for different forms of immunoprecipitated SIK3 in the transfected HEK293 cells. (H) Quantified results for (G). *P < 0.05, **P < 0.01, ****P < 0.0001, n.s. not significant. unpaired t test for (B). One-way ANOVA, post hoc Tukey’s multiple comparisons test (F and H). Data are mean ± SEM.

The Residues Surrounding SIK3 N783 Have Dramatic Effects On Kinase Activity.

The asparagine residue (N783) is situated within a proline-rich region and is surrounded by several residues, predominantly histidine and proline, which are characterized by their cyclic structures (Fig. 1B). According to predictive modeling conducted by AlphaFold, this region is identified as a disordered sequence, as indicated by low confidence values (SI Appendix, Fig. S1B). This characterization suggests that the sequence may not form classical structural domains, which implies a greater degree of flexibility and uncertainty in its spatial conformation. The introduction of the N783Y mutation, which adds an additional cyclic side chain, is likely to influence the conformation of this disordered sequence (SI Appendix, Fig. S1 C and D).

Notably, N783 is preceded by an SPP motif. Given that serine residues frequently undergo posttranslational modifications in vivo, additional mutations were introduced for further investigation (Fig. 2D). Our results demonstrated that the SIK3-S779A mutation significantly enhances kinase activity, whereas the double mutation (SIK3-S778A/S779A) abolishes this effect (Fig. 2 E and F). This indicates that the amino acid sequence surrounding N783 plays a crucial role in regulating its enzymatic activity. Immunoblot analysis using a phospho-specific antibody (anti-pSPP) confirmed that S779 can be phosphorylated in cultured cells (Fig. 2G). Although phosphorylation levels of S779 in the SIK3- N783Y were comparable to that of the WT transfected cells (Fig. 2 G and H), these findings suggest that N783 is located within a region that is particularly crucial for the enzymatic activity of SIK3, and that even minor structural changes in this region can have profound effects on its kinase activity.

Sik3-N783Y Mice Spend Less Time Sleeping.

To test whether the N783Y mutation contributes to the short sleep phenotype in the human subject, we generated the mouse model carrying the homologous N783Y mutation (Sik3^N783Yor Sik3^NY hereafter). Electroencephalogram and electromyography (EEG/EMG) recordings were conducted under a 12 to 12-h light–dark cycle to investigate their sleep patterns. Homozygous Sik3^NY/NY mice exhibited increased wakefulness, reduced non–rapid eye movement (NREM) sleep, and higher NREM sleep EEG delta power compared to WT littermates (Fig. 3 A–D and SI Appendix, Fig. S2 A–G). Reduced sleep was mainly due to the decreased NREM sleep during the dark phase (Fig. 3B). Daily sleep time in Sik3^NY/NY mice was 31.8 min shorter than WT littermates (Fig. 3 A–C). In contrast, heterozygous Sik3^N783Y/+ mice had comparable amounts of NREMS and rapid eye movement (REM) sleep to WT mice. Following sleep deprivation (ZT0-6), Sik3^NY/NY mice showed even less rebound sleep (~54 min less) than WT controls (Fig. 3 E–I and SI Appendix, Fig. S2 H–J). However, the recovery process is largely comparable among different genotypes (SI Appendix, Fig. S2 K–M). Together, these results suggested a potential causal link between the SIK3-N783Y mutation and the sleep phenotype in the mouse model and in the human subject.

Fig. 3.

Sik3-N783Y mice demonstrate reduced sleep time. (A–C) Total wake (A), NREM sleep (B), and REM sleep (C) time within 24 h, light phase, and dark phase were measured by EEG/EMG. (D) NREM sleep delta power was plotted hourly over 24 h. (E) Schematic of sleep deprivation (SD) workflow. (F–H) NREM (F), REM (G), and total sleep (H) time within 18 h (18 h), light phase (6 to 12 h), and dark phase (12 to 24 h) measured by EEG/EMG were calculated. (I) NREM sleep delta power was plotted hourly over 18 h following SD. Sleep measurements for Sik3^+/+ (WT) (n =16), Sik3^NY/+(n = 19), Sik3^NY/NY (n = 15) under basal conditions. Sleep times for Sik3^+/+ (WT) (n = 11), Sik3^NY/+ (n = 9) and Sik3^NY/NY (n = 13) after SD. *P < 0.05, **P < 0.01. One-way ANOVA (A to C, F to H); two-way ANOVA, post hoc Sidak’s multiple comparisons test (D and I). Data are mean ± SEM.

The Sik3-N783Y Mutation Resulted in Widespread Alterations At the Phosphoproteomic Level.

Previously, mutant mouse Sik3 was shown to have dramatic effects on the brain phosphoproteome (8). Next, we implemented quantitative proteomic and phosphoproteomic analyses using whole forebrain lysates from WT and Sik3^NY/NY (Mut) mice, analyzing a total of 12 samples via mass spectrometry with six samples from each group (Fig. 4A). Our proteomic analysis successfully quantified 8,265 proteins, of which 7,850 were detected in both genotypes (SI Appendix, Fig. S3A and Dataset S1). Furthermore, the phosphoproteomic analysis quantified a total of 17,992 phosphopeptides corresponding to 11,696 phosphosites and 4,013 phosphoproteins common to both genotypes (SI Appendix, Fig. S3 B and C and Dataset S2). The phosphopeptide intensities demonstrated high reproducibility across both biological replicates and genotypes (SI Appendix, Fig. S3D).

Fig. 4.

Brains of Sik3^N783Y/N783Y mutant (Mut) mice exhibited extensive hypophosphorylation. (A) Experiment workflow for proteomic and phosphoproteomic analysis. (B) Summary of quantified significant change in the phosphoproteome and proteome. (C and D) Volcano plots showing changes of all phosphopeptides in phosphoproteome (C) and proteins in proteome (D). (E and F) Pie chart indicating the fraction of hyperphosphorylated (E) and hypophosphorylated (F) peptides which also exhibited significant increase(red) or decrease (blue) in protein abundance. Light red and light blue indicate alterations exclusively in protein phosphorylation, without corresponding changes in total protein abundance.

We compared the phosphoproteomic profiles of the Mut samples to those of WT controls and found that a substantial fraction of phosphosites (782 out of 11,696; 6.7%) showed significant changes in abundance, with a fold change greater than 1.5 and a P-value less than 0.05 (Fig. 4 B–D). The majority of these changes were decreases (533 out of 782, 68.1%), indicating that the mutation primarily leads to hypophosphorylation (Fig. 4C). At the protein level, only a minor fraction of proteins (94 out of 7,850; 1.2%) exhibited significant changes, suggesting a general stability of the brain proteome in the mutant mice (Fig. 4B). Among these 94 proteins, 62 showed an increase and 32 a decrease in abundance in the mutant samples (Fig. 4D). Importantly, the majority of changes in phosphosite levels were not reflected at the protein level, with only a minimal number of altered phosphosites showing significant changes in corresponding protein abundance (1 out of 249, 0.4% in the increased group; 5 out of 533, 0.9% in the decreased group) (Fig. 4 E and F). Together, these results underscore the specificity of phosphoregulatory alterations induced by the Sik3-N783Y mutation.

Sik3-N783Y Induced Distinct Alterations Compared to the Sleepy Mutation.

The Sik3 ^Sleepy mutation was initially identified in Sleepy mice, characterized by a phenotype that contrasts with the N783Y mutants in terms of sleep duration (11) (SI Appendix, Fig. S4A). To test whether these two mutations caused opposite downstream effects, we aligned the differentially expressed phosphosites (DEPs) from the N783Y mutant with the dataset derived from Sleepy mice (8). Employing a less stringent cutoff (fold change > 1.2), we observed that 249 phosphosites showed significant alterations in both Sik3^NY and Sik3 ^Sleepy mice (SI Appendix, Fig. S4B). These DEPs were further categorized into four groups based on the directionality of their changes across different genotypes. Notably, 57.0% (142/249) of the DEPs were categorized in groups 3 and 4, which displayed opposing changes between N783Y and Sleepy mutations (SI Appendix, Fig. S4B and Dataset S3).

SIK3 belongs to the AMP-activated protein kinase (AMPK) family. We extracted the putative AMPK substrates from the DEPs in the Sik3^NY/NY mice and the overlapping dataset with the Sleepy model (8). In the Sik3^NY/NY (Mut) /WT comparison, a total of 66 phosphosites changed significantly; 52 were hypophosphorylated, which overshadowed the 14 in the hyperphosphorylation group (SI Appendix, Fig. S4C). Hypophosphorylated sites, but not those hyperphosphorylated, demonstrated significant enrichment compared to the global average (Fig. 4C and SI Appendix, Fig. S4C, Bottom panel). In the dataset overlapping with Sleepy, 25 putative AMPK substrates were identified, with Group 3 showing the most enrichment (SI Appendix, Fig. S4D and Dataset S3). This further supported that the Sik3^NY/NY and Sik3 ^Sleepy mutations exert opposite effects on their downstream targets.

A previous study revealed a list of proteins that physically interact with SIK3, predominantly binding with the Sleepy mutant form of SIK3 (8). We extracted these SIK3-binding proteins from the Sik3^NY/NY DEPs dataset. From a total of 1,283 phosphosites corresponding to these SIK3 interacting proteins, 28 and 45 exhibited hyper- and hypophosphorylation in the Mut mice, respectively (SI Appendix, Fig. S4E). However, there is no significant enrichment in either characterized hyperphosphorylated or hypophosphorylated group compared to the global average (SI Appendix, Fig. S4E). As control, 192 and 91 hyper- and hypophosphorylated peptides were identified from 2,879 phosphopeptides in Sleepy animals(8) (SI Appendix, Fig. S4F). Together, these findings aligned with the hypothesis that the N783Y mutation leads to a general reduction in kinase activity, contrasting with the gain-of-function property observed in Sleepy mice (8, 11).

Phosphorylation Changes at Synapses in Sik3^N783Y Mutant Mice.

Several studies have highlighted the critical roles of synaptic protein phosphorylation in regulating sleep–wake cycles, particularly emphasizing SIK3 as a key kinase(8–10). Gene ontology analysis revealed that DEPs in the Sik3^NY/NY (Mut)/WT group were primarily associated with synapse-related terms (Fig. 5 A and B). When aligning the DEPs from mutant samples with previously identified synaptic phosphoproteomes (10) (Dataset S4), 18.2% (142/782) of the DEPs in mutant mice precisely matched the phosphorylation sites identified at synapses. On the other hand, of the 5,748 phosphopeptides identified at synapses, 142 (2.5%) exhibited significant changes in Sik3^NY/NY mice (Upper Left panel, Fig. 5C). At the protein level, 294 of the 622 phosphoproteins (47.2%) with at least one significantly altered phosphorylation site in the N783Y mutant samples were detected at the synapse. Out of the total 1,657 phosphoproteins identified at the synapse, 17.7% showed differential phosphorylation in the mutant samples (Upper Right panel, Fig. 5C). The earlier study also indicated that over one-fourth of the phosphorylation sites in synaptic proteins undergo daily oscillations, correlating with sleep–wake cycles (10). However, there was no significant enrichment of cycling phosphosites over the total phosphopeptide dataset, suggesting that the N783Y mutation does not specifically affect the cycling phosphoproteome (Lower panel, Fig. 5C).

Fig. 5.

N783Y mutation led to phosphorylation changes at synapse. (A and B) Top20 Gene Ontology (GO) cellular component (CC) (A) and biological process (BP) (B) annotations for the DEPs in Mut/WT comparison. (C) Venn diagram of quantified DEPs (Left) and phosphoproteins (Right) overlapped with previously identified total (Upper) and cycling (Lower) synaptic phosphopeptides and phosphoproteins. (D) Heatmap profile of changes of SNIPPs in the phosphorylation state (Left panel) and total protein abundance (Right panel) in two comparisons. (E) Mean phosphorylation changes of SNIPPs in two animal models. (F) Representative immunoblots using indicated antibodies with the forebrain lysate collected from WT and Mut mice. (G) Quantified results for (F). Mut/WT: Sik3^NY/NY vs. WT; Slp/WT: Sik3^Sleepy/+ vs. WT. n.s. not significant, **P < 0.01. Chi square test (E), unpaired t test (G).

Previous research has pinpointed synaptic sleep-need-index phosphoproteins (SNIPPs), where phosphorylation changes are closely linked to variations in sleep need (8). We assessed the overall phosphorylation state changes by summarizing the log2 (fold change) (log2FC) value of all significant DEPs for individual SNIPPs (Fig. 5D and Dataset S5). Consistent with previous findings that SNIPPs exhibit hyperphosphorylation in Sleepy mice, the analysis here showed uniformly positive log2FC values for the Sleepy/WT group (Fig. 5E) (8). Notably, 45 out of 80 SNIPPs in the Mut/WT group had at least one significantly altered phosphorylated site. In contrast to the Sleepy/WT group, the majority of SNIPPs in the Mut/WT group exhibited decreased phosphorylation (68.9%, 31/45) (Fig. 5E). Decreased phosphorylation of NMDAR subunits NR2B at Ser 1303 was verified by western blotting (Fig. 5 F and G). Notably, protein changes in the mutant group were minimal (Right panel, Fig. 5D).

Kinase Network Changes Caused by Sik3^N783Y Mutation.

Kinases are pivotal regulators of nearly all cellular activities, including sleep. Our research uncovered a total of 47 DEPs from 37 kinases in the Mut/WT comparison as depicted in Fig. 6A. Notably, kinase PAK6 was the only kinase exhibiting increased protein abundance in the mutant mice, whereas reductions were observed in three others, namely EPHA6, NPR2, and EIF2AK4 (Fig. 6B).

Fig. 6.

N783Y mutation led to phosphorylation changes of kinases. (A and B) Volcano plot showing changes of all putative kinases in phosphoproteome (A) and proteome (B) in the Sik3^NY/NY (Mut)/WT group. (C) Enrichment results of kinase–substrate relationships were organized and visualized in a chord plot. The most enriched kinases were shown on the Right (red rings), while the substrates were displayed on the Left (green rings). Every colored chord represents the kinase–substrate relationship originated from one kinase. (D) KSEA enrichment in the Mut/WT group. (E) Representative immunoblots using antibody against phospho-PKA in the forebrain lysate from Sik3^WT and Sik3^NY/NY mice. (F) Quantified results for (E). (G) Same as (E) except antibody against phospho-MAPK substrate motif was used. (H) Quantified results for (G). *P < 0.05, unpaired t test.

Next, we annotated phosphorylated peptides using established kinase–substrate relationships (www.phosphosite.org). This revealed that CAMK2A, MAPK1, and PRKACA (PKA), along with their respective substrates, were among the top three enriched relationships (Fig. 6C). Given that site-specific phosphorylation does not directly correlate with changes in kinase activity, we also employed another computational approach, Kinase–substrate Enrichment Analysis (KSEA) (18). The KSEA results indicated a significant upregulation in MAPK1/3 activity and a downregulation in PRKACA activity in the Mut samples. Although not statistically significant, the KSEA algorithm also highlighted several potentially altered kinase activities, including those of MTOR, CAMK2, and AKT1, all of which are known to play roles in synaptic regulation during sleep/wake cycles (10, 19–22) (Fig. 6D).

To further investigate the kinase activity that participated in the altered signaling cascades caused by the Sik3^N783Y mutation, we analyzed the consensus motifs of the DEPs using MEME (23, 24). The analysis identified six consensus motifs significantly enriched among the down-regulated DEPs and three among the up-regulated DEPs (SI Appendix, Fig. S5 A and B). Notably, down-regulated Motif 3 aligns with the consensus for both SIKs (25, 26) and general AMPKs(27, 28), as well as matching the consensus for PKA (SI Appendix, Fig. S5 A and C). Additionally, down-regulated Motif 6 also matches PKA consensus (SI Appendix, Fig. S5 A and C). In contrast, up-regulated Motif 7 and 8 correspond with the consensus for MAPKs (SI Appendix, Fig. S5 B and C). These motif analyses are consistent with the KSEA findings, suggesting that PKA and MAPKs are the respective down-regulated and up-regulated kinases in the mutant samples. Immunoblotting with antibody against phospho-PKA substrate motif revealed a significant reduction in phosphorylated substrates in the mutant brain samples (Fig. 6 E and F). Conversely, when immunoblotting was performed with an antibody against the phospho-MAPK substrate motif, it was observed that some bands demonstrated an increase in phosphorylation (Fig. 6 G and H).

Employing SIGNORApp, which constructs regulatory networks from peer-reviewed publications (29), we aimed to develop a framework using our phosphoproteome profiles to elucidate the changes brought about by the Sik3^N783Y mutation. Several kinases emerged as nodes within this network, underscoring their central roles in the signaling cascade (SI Appendix, Fig. S6). Again, MAPK1 and PKA were identified as critical nodes, colored red and blue, respectively. The network also suggested roles for other signaling molecules in the Sik3 regulatory network, such as CSNK2A1, GRIN2B, and RAF1, whose functions might extend beyond sleep regulation (30, 31). It is important to mention that while the network was generated automatically by the algorithm, SIK3 and its two phosphorylated sites (T221 and S551) were manually added as they did not show significant changes in the mutant samples (SI Appendix, Fig. S7). This network is based on the DEPs enriched in the Sik3 mutant models, therefore, SIK3 should be an inherent start point being placed in the center. However, SIK3 is positioned on the periphery of the network. Therefore, SIK3 could potentially impact the entire network via putative, yet unidentified edges.

Discussion

Thus far, five natural short sleep mutations in four genes have been reported from the human population, including DEC2 (6), ADRB1 (5), NPSR1 (3), and GRM1 (4). Correspondingly, the sleep pattern of mouse models has been extensively studied. It is not uncommon that mouse models do not fully replicate the human phenotype regarding sleep duration. In this context, the sleep duration of heterozygotic mutant mice Sik3^NY/+ did not significantly differ from that of WT controls, although a slight increase in delta power was noted (Fig. 3D). Homozygous mice exhibited 31 min less sleep under basal conditions (Fig. 3A) and 54 min less sleep during the recovery period following sleep deprivation (Fig. 3F), a reduction nonetheless smaller than that observed in human subjects. This discrepancy may be attributed to the different sleep patterns between species, as rodents typically experience more fragmented sleep. Alternatively, it may result from genetic background effects of this inbred mouse strain (32). In models of natural short sleep, mutant animals usually display enhanced delta power (3, 5), which was also observed here in Sik3^NY/NY mutant mice (Fig. 3D). This suggests that the sleep pressure sensing system remains functional, and enhanced delta power is a consequence of reduced sleep. This observation aligns with clinical findings that short sleepers tend to awaken with higher sleep pressure and can tolerate greater homeostatic sleep pressure (1, 33).

Several seminal studies have identified Sik3 as a rodent sleep gene. Among the various animal models developed, the Sleepy mouse model is notably the foremost and extensively examined in sleep-related research (11). Phenotypically, the Sik3^N783Y model exhibits reduced sleep duration compared to the Sleepy mouse model, which demonstrates increased sleep. This observation aligns with biochemical functionalities where the N783Y mutation results in a loss of function, whereas the Sleepy mutation leads to a gain of function. Such findings corroborate the hypothesis that SIK3 acts as a sleep-promoting kinase. Utilizing the published phosphoproteomic datasets, a direct comparative analysis between our Sik3^N783Y and Sleepy models was conducted. The analysis revealed that the Sleepy mutation induces global hyperphosphorylation, whereas the Sik3^N783Y mutation largely causes hypophosphorylation of proteins, including SNIPPs (8). Nevertheless, the effects of the Sleepy and N783Y mutations do not uniformly oppose each other; for instance, a substantial proportion of DEPs exhibit similar alterations in both models (Group1 and Group2, SI Appendix, Fig. S4B), and an increase in delta power during NREM sleep was observed in both models.

Phosphorylation of SNIPPs were identified as associated with sleep need, with their phosphorylation levels presumed to positively correlate with NREM sleep delta power (8). However, in the Sik3^NY mouse model, most SNIPPs demonstrated decreased phosphorylation despite elevated delta power, suggesting a decoupling of SNIPPs phosphorylation from sleep need. One plausible explanation is that SIK3 is the principal mediator conveying sleep need to SNIPPs. Given that N783Y constitutes a loss-of-function mutation, the phosphorylation of SNIPPs may no longer reliably reflect sleep need. Similar results were observed for PKA. PKA activity decreased in the Sik3^NY/NY samples (Fig. 6). Accordingly, phospho-PKA substrate was downregulated in Sik3^NY/NY (Fig. 6E) and upregulated in Sik3^Sleepy mice (8). This is inconsistent with the recent declaration that PKA is presumably a wake-promoting kinase (34). Previous research indicated that PKA directly inhibits SIK3 by phosphorylating S551 (12). Therefore, SIK3 may be epistatic to PKA. In the context of the Sik3^NY mutation, the downregulation of PKA, which would typically promote sleep, could be obstructed by the reduced activity of SIK3. The decrease in PKA activity could serve as a feedback mechanism in response to the attenuated function of SIK3. Yet, the precise feedback mechanism needs further investigation.

Recent research has highlighted that the LKB1–SIK3–HDAC4/5 signaling cascade plays a pivotal role in modulating sleep and wakefulness (15, 16). However, our studies did not reveal any observable changes in the phosphorylation levels of HDAC4/5 in the whole forebrain lysate from Sik3^N783Y mice, either through immunoblotting (SI Appendix, Fig. S7 A–D) or quantitative mass spectrometry. This lack of detectable difference may be attributed to the regional specificity of SIK3 expression and its role in sleep regulation, which may not be discernible in the brain lysate. Additionally, no significant changes in phosphorylation at the T221 site were detected, suggesting that the N783Y mutation does not impact phosphorylation at this position (SI Appendix, Fig. S7 E–H). Our studies also identified S779 as a potential phosphorylation site on SIK3, with mass spectrometry results confirming its phosphorylation in vivo (Dataset S2). However, the N783Y mutation did not affect phosphorylation at the S779 site. Given the influence of mutations at the S779 site on enzyme activity (Fig. 2E), it is speculated that this site might regulate sleep under certain physiological conditions. This serine residue and its neighboring sequences correspond to the phosphorylation consensus for MAPK1/3, which exhibited increased activity in our KSEA (Fig. 6D). Interestingly, the SPP motif could also be a target for other kinases such as CDK5, which showed reduced activity in the Sik3^N783Y mutants, indicating a possible competitive interaction among different kinases at this site. Further research is necessary to clarify the integration of SIK3 activity within the kinase network and its precise role in sleep regulation.

The phosphorylation of synaptic proteins is pivotal in regulating the sleep–wake cycle in mammals (8–10). SIK3, inherently a synaptic kinase (10), plays a significant role in this process(15, 16). Our data revealed that approximately one fifth of the phosphorylated synaptic proteins in our study exhibited at least one phosphosite significantly altered by the N783Y mutation (Fig. 5C). Notably, among the genes associated with short sleep duration in humans, GRM1 is noteworthy as it encodes the group I metabotropic glutamate receptor (mGluR1). The signaling pathways of mGluR1, in conjunction with PKA, are instrumental in mediating the downscaling of excitatory synapses during sleep (9, 35). Additionally, noradrenergic signaling and the potential involvement of the beta receptor (ADRB1) are likely to modulate this process (9). Therefore, the finding of human short sleep genes strongly supports that sleep quantity/quality is directly associated with synapse homeostasis.

In sum, our findings underscore the evolutionary conservation of SIK3 as a sleep gene. Similar to various previously identified receptors, SIK3 emerges as a promising therapeutic target to treat sleep-related disorders. On another note, our study also identifies specific conserved regions within SIK3 that could bidirectionally regulate its enzymatic activity, providing additional insights for future drug design.

Materials and Methods

Short Sleeper Characterization and Identification of Candidate Gene.

Human subject for this study was a healthy voluntary participant (no health issue physically and/or mentally). All studies conducted were approved and human participant signed consent forms approved by the Institutional Review Board at the University of California, San Francisco (IRB# 10-03952). Self-reported habitual sleep–wake schedules were obtained during structured interviews. Activity recording was obtained by using wrist actigraphy (Ambulatory Monitoring Inc). Blood sample collection and DNA preparation were performed as previously described (6). Whole exome sequencing was performed by Novogene (Novogene Corporation Inc, Sacramento CA). Variants were screened using the following criteria: 1) it was heterozygous in affected individual; 2) the allele frequency in both the 1,000 Genomes (36) and ExAC (37) datasets was less than 10⁻⁴; 3) was potentially deleterious because it had a “high” predicted impact in SnpEff(38), or was called as “damaging” by SIFT (39), or was categorized as either “probably damaging” or “possibly damaging” by HumDiv-trained Polyphen-2 (40), 4) it did not belong to a gene with a high load of rare deleterious mutations, and 5) it did not fall into the list of “in-house” exclusion gene list. All variants passing these criteria were then examined by researchers to prioritize its candidacy based on literature investigation.

Animals.

Sik3-N783Y mice were customer-made by GemPharmatech company (Nanjing,China). Male mice were used for all behavioral experiments. Mice from both sexes were used for molecular analysis. All experimental animals were housed under a 12:12 LD cycle and given ad libitum access to food and water. All experimental procedures were conducted as approved by Animal Care and Use Committees at Shanghai Institute of Materia Medica, Chinese Academy of Sciences.

EEG/EMG Recording and Scoring.

EEG/EMG surgeries, recordings, and scoring were carried out as described previously (5, 41). For spectral analysis, artifacts and state transition epochs were excluded. The relative NREM delta power density was calculated by the ratio of delta power (1–4 Hz) to total power (1–25 Hz) of EEG signals during NREM sleep. EEG power spectra analysis was conducted by calculating the ratio of each 1 Hz bin to total 1–25 Hz of EEG signals during NREMS, which was averaged over ZT1 to ZT24 (8). Sleep deprivation was carried out from ZT0 to ZT6 by gentle handling (3).

Structural Prediction.

In the protein tertiary structure prediction section, we employed the AlphaFold3 prediction platform (https://alphafoldserver.com/) to predict the structures of both WT and mutant proteins. Specifically, the amino acid sequences of the WT and mutant proteins were submitted to the server for global structure prediction. During the prediction process, the random seed was set to the default state. Based on the quality assessment metrics of the predicted models, the structures with the highest average predicted Local Distance Difference Test (pLDDT) scores and minimal prediction errors were selected as the final predicted structures for the WT and mutant proteins. Notably, the pLDDT scores for both WT and mutant models were below 50, suggesting that the prediction results should be interpreted cautiously and validated using additional structural analysis methods. Experimental data revealed that the interface predicted template modeling scores for the WT and mutant proteins were 0.35 and 0.30, respectively, while the global predicted template modeling scores were 0.36 and 0.32.

Mass Spectrometry Sample Preparation.

Forebrains from 12 male mice (6 for WT and 6 for Sik3 ^NY/NY) were quickly dissected at ZT1, rinsed with PBS and flash frozen in liquid nitrogen. During experiment, the mouse brain was individually homogenized by MP FastPrep-24 homogenizer (24×2, 6.0 M/S, 60 s, twice), and then SDT buffer (4%SDS, 100 mM Tris-HCl, pH7.6) was added. The lysates were further sonicated and boiled for 15 min. After being centrifuged at 14,000×g for 40 min, the supernatant was quantified with the BCA Protein Assay Kit. 15 μg of protein for each sample were mixed with 5X loading buffer respectively and boiled for 5 min. The proteins were separated on 4 to 20% SDS-PAGE gel (constant voltage 180 V, 45 min). Protein bands were visualized by Coomassie Blue R-250 staining. An equal aliquot from each sample in this experiment was pooled into one sample for quality control. DTT (with the final concentration of 20 mM) was added to each sample respectively, mixed at 600 rpm for 1 min and incubated at 37 °C for 1.5 h. After the samples cooled to room temperature, IAA (with the final concentration of 20 mM) was added into the mixture to block reduced cysteine residues and incubated for 30 min in darkness. Then, 8 M UA was diluted with 450 μL 100 mM NH4HCO3 buffer to a final concentration below 1.5 M. Finally, trypsin was added to the samples (the trypsin: protein (wt/wt) ratio was 3:200) and incubated at 37 °C for 15 to 18 h (overnight). 50 μL 0.01% TFA was added to the peptides and the 10%TFA was then added to adjust pH to 3 for stopping trypsin digestion. The peptides of each sample were desalted on C18 Cartridges (Empore™ SPE Cartridges MCX, 30UM, waters) and lyophilized.

The enrichment of phosphopeptides was carried out using the High-SelectTM Fe-NTA Phosphopeptides Enrichment Kit according to the manufacturer’s instructions (Thermo Scientific). After lyophilization, the phosphopeptides were resuspended in 20 μL loading buffer (0.1% formic acid). iRT standard peptides were added to 2 μg phosphopeptides of each sample before analysis.

Mass Spectrometry Assay for Data Independent Acquisition (DIA).

The peptides from each sample were analyzed by OrbitrapTM AstralTM mass spectrometer (Thermo Scientific) connected to a Vanquish Neo system liquid chromatography (Thermo Scientific) in the DIA mode. Precursor Ions were scanned at a mass range of 380 to 980 m/z, MS1 resolution was 240,000 at 200 m/z, Normalized AGC Target: 500%,Maximum IT: 5 ms. 300 windows were set for DIA mode in MS2 scanning, Isolation Window: 2 m/z, HCD Collision Energy: 25ev, Normalized AGC Target: 500% for proteome and 300% for phosphoproteome, Maximum IT: 3 ms.

Mass Spectrometry Data Analysis.

DIA data for proteome were analyzed with DIA-NN 1.8.1. Main software parameters were set as follows: enzyme is trypsin, max missed cleavages is 1, fixed modification is carbamidomethyl (C), dynamic modification is oxidation (M) and acetyl (Protein N-term). DIA data for phosphoproteome were analyzed with Spectronaut. Main software parameters were set as follows: enzyme is trypsin, max missed cleavages is 2, fixed modification is carbamidomethyl (C), dynamic modification is oxidation (M) and acetyl (Protein N-term). All protein or peptide identification was determined by false discovery rate (FDR) ≤ 1%. For site-specific quantification, the abundance of a phosphorylation site was calculated by summing up all the abundances of precursors/peptides containing this site (42).

Motif Analysis.

The motifs were analyzed by MeMe (http://meme-suite.org/index.htm). We extracted the amino acid sequences containing the modified site and six upstream/downstream amino acids from the modified site (13 amino acid sites in total). These sequences were used to predict motifs in this study (parameters: width: 13, occurrences: 20, background: species).

GO Annotation.

Genes with a P value <0.05 and |log2fold-change| >0.585 were considered differentially expressed. GO enrichment analysis was performed using the clusterProfiler R package (v4.14.4, RRID: SCR_016884), focusing on BP and CC. The results were imported into the ggplot2 R package for visualization.

KSEA.

KSEA were performed in R (4.4.1) implementing the package KSEAapp (https://github.com/casecpb/KSEA) using phosphosite data according to its manual with the cutoff of P < 0.05 and substrate count > 3(18, 43, 44).

NETWORK Analysis.

Network analysis was performed in Cytoscape (45) (v3.10.3) using SIGNOR App(29) (v1.2). The phosphosite (with significant change) information in our dataset was imported to the application to build the subnetwork according to its manual. SIK3 T221 and S551 were manually added to the dataset as these two phosphosites did not exhibit significant changes in the mutant mice.

In Vitro Kinase Assay.

GST-tagged SIK3(WT), SIK3(NY), SIK3(SA), SIK3(2SA), and HDAC4/5 recombinant proteins were expressed and purified from BL21 using Glutathione Sepharose 4 Fast Flow (17513201, cytiva). Kinases and substrates were incubated with 1 × Kinase buffer (20 mM Tris-HCl PH8.0, 10 mM MgCl2, 25 mM β-glycerophosphate, 250 μM ATP) and reacted in thermo shaker in 37 °C 1000 rpm for 30 min. Then the activity of SIK3 was analyzed by immunoblot.

Immunoblot and Antibodies.

Western blotting was performed according to standard procedures. Primary antibody used in this study were as follow: phospho-SIKs (SIK1-T182, SIK2-T175, SIK3-T163) (ab199474,Abcam), phospho-HDAC4(S246)HDAC5(S259)/HDAC7(S155) (#3443,CST), phospho-PKA Substrate (RRXS*/T*) (#9624,CST), phospho-Ser-Pro-Pro motif [p-SPP] (#14390,CST), HDAC4(D15C3) (#7628,CST), HDAC5(B-11) (sc133106,Santa Cruz), QSK(C-2) (sc515408,Santa Cruz), Flag M2 (F1804,sigma), Phospho-MAPK Substrates Motif (PXpTP) (#14378,CST), Phospho- GluN2B (Ser1303)(#20274,CST), GluN2B(D15B3) (#4212,CST),β-actin (LF201,Epizyme). Band intensities were determined using ImageJ software (NIH).

Quantification and Statistical Analysis.

Statistical parameters including sample sizes (n = number of animals per group), the statistical test used, and statistical significance are reported in the Figures and Figure legends. No outliers were excluded from analysis unless specified. Data are judged to be statistically significant when P < 0.05. In figures, asterisks denote statistical significance *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. All statistical analysis was performed using GraphPad PRISM 9 software.

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (XLSX)

Dataset S02 (XLSX)

Dataset S03 (XLSX)

Dataset S04 (XLSX)

Dataset S05 (XLSX)

Data, Materials, and Software Availability

All study data are included in the article and/or supporting information. Previously published data were used for this work (Datasets in refs. 8 and 10 were used).