Circadian Rhythm Mechanisms Underlying Convergent Adaptation of Unihemispheric Slow-Wave Sleep in Marine Mammals

Authors: Daiqing Yin, Zhenpeng Yu, Haojia Jiang, Yujie Chong, Cuijuan Zhong, Shixia Xu, Guang Yang

Abstract

Marine mammals have evolved unihemispheric slow-wave sleep, a unique state during which one cerebral hemisphere sleeps while the other remains awake, to mitigate the fundamental conflict between sleep and wakefulness. However, the underlying mechanisms remain largely unclear. Here, we use a comparative phylogenetic approach to analyze genes associated with light-dependent circadian mechanisms, aiming to reconstruct the evolution of the circadian rhythm pathway in mammals and to identify adaptively changed components likely to have contributed to the development of unihemispheric slow-wave sleep. Specifically, among eight genes with shared signals of positive selection in two unihemispheric slow-wave sleep-specific lineages, seven genes showed direct evidence of affecting sleep and spontaneous movements. Both in vitro and in vivo experiments indicated that functional innovation in cetacean and non-phocid pinniped FBXL21, which was found to undergo positive selection, may be beneficial for decoupling sleep–wake patterns from daily rhythms to sustain continuous swimming. For cetaceans exhibiting only unihemispheric slow-wave sleep, we identified 73 genes as rapidly evolving and 92 genes containing unique amino acid substitutions. Functional assays showed that a cetacean-specific mutation (F411Y) in NFIL3 led to a decrease in repressor activity and protein stability. Furthermore, convergent amino acid replacements detected in genes related to Ca^2+^ signaling and CREB phosphorylation suggest their crucial role in unihemispheric slow-wave sleep adaptation. Overall, this study enhances our understanding of the evolutionary mechanisms underlying unihemispheric slow-wave sleep and provides a foundation for investigating how circadian rhythm changes contribute to variations in sleep and circadian behavior.

Introduction

Sleep, which serves multiple biological functions such as thermoregulation and energy management, is crucial for animal survival (Siegel 2005). Chronic sleep deprivation even leads to animal death more quickly than food deprivation (Rechtschaffen 1998). Sleep is known to be a highly conserved state throughout the animal kingdom (Joiner 2016). However, there is a great diversity of the sleep architecture (i.e. the amount, composition, continuity, and intensity of sleep), which reflects ecological and physiological demands (Ungurean et al. 2020). Unearthing the genetic basis of sleep diversity provides important clues to elucidate the molecular mechanisms of sleep regulation. Initial comparative studies in the fruit fly (Drosophila melanogaster), zebrafish (Danio rerio), and mouse (Mus musculus) identified conserved and novel regulators of sleep (Keene and Duboue 2018).

One of the most intriguing topics in sleep research arose from comparative studies of marine mammals (Aulsebrook et al. 2016). All terrestrial mammals studied to date have two different sleep stages involving the whole brain and bihemispheric slow-wave sleep (BSWS) and rapid eye movement (REM) sleep (Siegel 2005). However, marine mammals (except phocids) have evolved a distinct sleep behavior known as unihemispheric slow-wave sleep (USWS), in which one cerebral hemisphere sleeps while the other is awake to coordinate movement for surfacing and heat generation. Interestingly, fully aquatic cetaceans were found to have only USWS, with no signs of REM sleep recorded (Lyamin et al. 2008). Manatees, another fully aquatic mammal, exhibit not only USWS but typical BSWS and REM sleep found in terrestrial mammals, and they remain motionless underwater during all stages of sleep (Mukhametov et al. 1992). Regarding amphibious pinnipeds (i.e. walrus, otariids, and phocids), walrus and otariids exhibit BSWS and REM sleep, as well as typical USWS in water and on land (Mascetti 2016). In contrast, phocids have only BSWS and REM sleep, with a complete absence of USWS both in water and on land (Rattenborg et al. 2000). When sleeping under water, they hold their breath and periodically awaken to return to the surface to breathe (Karamanlidis et al. 2017). USWS allows marine mammals to experience the benefits of sleep, breathing, thermoregulation, and vigilance, which contributes to their adaptation to aquatic habitats. A greater understanding of USWS promises to provide novel insights into the evolution of sleep, yet knowledge on the molecular basis underlying USWS remains limited.

The circadian rhythm, a self-sustained biological oscillator system with a period close to 24 h, plays a crucial role in sleep regulation (Borbély et al. 2016). At the molecular level, the mammalian circadian system consists of two main the photic signal transduction pathway and the circadian clock pathway (Lowrey and Takahashi 2000). Light input from retinal photoreceptors reaches the master circadian pacemaker in the suprachiasmatic nuclei (SCN) via the photic signal transduction pathway, where it then interacts with the circadian clock to regulate rhythmic oscillations (Colwell 2011). Primary neurotransmitters, including glutamate and pituitary adenylate cyclase activating polypeptide (PACAP), act upon glutamate AMPA/kainate receptors, PACAP-specific receptor (PAC1R), and a number of protein kinases to alter the intracellular concentrations of Ca^2+^ and trigger the phosphorylation of cAMP response element-binding proteins (CREBs). Furthermore, an interacting positive and negative transcriptional–translational feedback loop (TTFL) composed of cycling circadian clock genes and their protein products is central to circadian clock function (Agostino et al. 2011). In brief, the positive regulators BMAL1 or BMAL2 and CLOCK or NPAS2, acting in the form of a heterodimer, activate the transcription of negative regulators, including three period genes (PER1, PER2, and PER3) and two cryptochrome genes (CRY1 and CRY2). Increases in PER and CRY proteins suppress CLOCK/BMAL1-dependent transactivation and thereby negatively regulate their expression (Ko and Takahashi 2006). Furthermore, additional stabilizing feedback loops composed of DBP/NFIL3 and ROR/REV-ERB further contribute to the timing and robustness of the cycle (Cox and Takahashi 2019). Other post-translational regulation (e.g. phosphorylation, acetylation, and ubiquitylation) of circadian clock proteins also has an important role in rhythm generation and regulation (Kojima et al. 2011). The loss or mutation of rhythm-related genes has been shown to cause sleep alterations or even sleep disorders (Wulff et al. 2009).

The molecular evolution of key candidate genes provides a feasible way to understand the genetic mechanisms underlying adaptive traits (Lynch et al. 2004; Xu et al. 2017). In present study, we investigated the molecular evolution of 147 circadian genes, evenly distributed across each major step in the circadian rhythm system, across mammalian phylogeny to reconstruct the evolution of the circadian rhythm pathway in mammals and to identify adaptively changed components likely to have contributed to the generation of USWS. Our evolutionary analyses, in combination with in vitro and in vivo functional experiments, indicate that many adaptive changes occurred specifically in USWS-specific marine mammals, suggesting that adaptive evolution of circadian system was necessary to meet the demands of sleep adaptation in marine mammals.

Results

Positive Selection Signals on Circadian Rhythm Genes in USWS-Specific Marine Mammals

We estimated the ancestral character states of the presence of USWS during mammalian evolution using the phylogenetic tree from TimeTree (Kumar et al. 2022). Reconstructing the ancestral state suggested that the emergence of USWS occurred independently three times in the respective ancestors of cetaceans, non-phocid pinnipeds (i.e. walrus and otariids), and the manatee. Selection analysis was further performed using the CODEML program implemented in phylogenetic analysis by maximum likelihood (PAML) to evaluate the selective pressure acting on circadian rhythm genes in mammals. First, the one-ratio model, a strict model that assumes the same ω (dN/dS) across all branches, showed that the ω values for all 147 rhythm genes were ranged from 0.001 to 0.402 (supplementary table S2, Supplementary Material online), providing good support for the expected presence of purifying selection acting on circadian rhythm genes. In the branch-site model, signals of positive selection were identified in 15 rhythm genes along the ancestral branch leading to cetaceans, and eight positively selected genes (PSGs) were significant for false discovery rate (FDR) (Fig. 1, supplementary table S3, Supplementary Material online). Furthermore, 28 rhythm genes were identified to be under positive selection along the terminal branches leading to cetaceans, and 15 PSGs were significant after FDR correction. Additionally, seven and six genes were found separately to be under positive selection along the ancestral and terminal branches leading to otariids and odobenids, with three and one PSGs remaining significant after FDR adjustment (Fig. 1, supplementary table S3, Supplementary Material online). For the manatee Trichechus manatus, two out of seven PSGs were significant after FDR correction (Fig. 1, supplementary table S3, Supplementary Material online). Among these, five PSGs were shared by cetaceans and non-phocid pinnipeds, and three PSGs were shared by cetaceans and the manatee (Fig. 1). For example, three key circadian elements (supplementary table S3, Supplementary Material online; CLOCK, BMAL2, and PER3) showed signals of positive selection in cetaceans, of which BMAL2 and PER3 were further identified to be positively selected in the manatee and non-phocid pinnipeds, respectively, although no significance was found after FDR correction. In contrast, after FDR correction, five PSGs (ADCY9, BTRC, CACNA1D, CRTC1, ITPR1) were found in the terrestrial counterpart branches (branches a, c, and e in Fig. 1). There was little overlap among the identified PSGs between USWS-specific lineages and their sister lineages (supplementary table S3, Supplementary Material online).

Convergent Amino Acid Substitutions Between Divergent USWS-Specific Lineages

To evaluate the molecular convergence, we reconstructed ancestral sequences for the internal nodes of the phylogenetic tree to identify convergent amino acid changes along three USWS-specific marine lineages. The algorithms PolyPhen-2, SIFT, and PROVEAN were then used examine the potential functional effects of amino acid replacements that were unique to USWS-specific lineages compared with the other species. No shared substitution was detected along all three USWS-specific lineages (branches b, d, and f in Fig. 1). However, 19 convergent amino acid changes were found in 12 rhythm genes in cetaceans and the manatee, as well as 7 convergent changes in 7 rhythm genes in the manatee and non-phocid pinnipeds, and 2 convergent changes in 2 rhythm genes in cetaceans and non-phocid pinnipeds (supplementary table S4, Supplementary Material online). Altogether, we identified 27 convergent mutations in 17 different genes among USWS-specific marine mammals. Only three genes (ADCY2, GRIN2C, SREBF1) show convergent mutations at different amino acid positions in manatee cetacean and manatee non-phocid pinniped. Three genes (CACNA1H, FBXL21, RYR1) were inferred to evolve under positive selection in USWS-specific lineages. Furthermore, 11 of these 27 convergent mutations harbored radical amino acid changes (Fig. 2a, supplementary table S4, Supplementary Material online), which are more likely to damage important biochemical properties of the original amino acid compared to conservative changes (Lyons and Lauring 2017), and were further predicted to impact function by PolyPhen-2, SIFT, and PROVEAN (supplementary table S5, Supplementary Material online). For example, SREBF1 has two shared substitutions (P505L, V726M) in the manatee and cetaceans and a A254V substitution in the manatee and non-phocid pinnipeds. These three convergent mutations were located in N-terminal transcription factor domain (residues 1 to 401) and C-terminal regulatory domain (residues 480 to 1,147), respectively, and were classified into radical and possibly damaging mutations (Fig. 2b, supplementary table S5, Supplementary Material online).

Evidence for Species-Specific Adaptations in Cetacean Rhythm Genes

Cetaceans have a unique sleep pattern, dominated by USWS and an almost complete absence of REM sleep. When comparing the evolutionary rates of the 147 rhythm genes with other mammals, the two-ratio model showed that 74 genes (50.3%, 74/147) have been rapidly evolving in cetaceans after FDR correction and the average evolutionary rate was more than four times greater than that in the control group (Fig. 3, supplementary table S3, Supplementary Material online). To test for a shift in selection pressures on cetacean rhythm genes relative to other USWS-specific lineages, we further analyzed selection patterns using clade model C (CmC), revealing that 31 rhythm genes (21.1%, 31/147) showed significant differences in selection between cetaceans and other USWS-specific lineages, with 17 out of them being found to evolve faster in cetaceans (supplementary table S7, Supplementary Material online). These 17 rhythm genes were also classified as rapidly evolving by the two-ratio model (supplementary table S3, Supplementary Material online), of which five genes (ADCY10, CREM, MEF2C, PLCB1, RYR1) showed signals of positive selection in cetaceans, providing strong support for the important roles of genetic changes in the circadian system for cetacean sleep evolution (Fig. 3).

Then, we reconstructed ancestral sequences and mapped species-specific amino acid changes along cetacean lineages. The results showed that 92 cetacean genes exhibited a total of 596 unique amino acid substitutions not found in any other mammals (supplementary table S8, Supplementary Material online). Furthermore, 66.3% (61/92) of these genes were identified as rapidly evolving and 59.1% (352/596) of cetacean-specific mutations generated radical amino acid changes. Notably, 13 non-frameshift deletions were identified in ten rhythm genes, which could have a greater functional effect (supplementary table S8, Supplementary Material online).

Detecting USWS-Associated Genes

Phylogenetic generalized least squares (PGLS) regression was used to assess the correlations between the evolutionary rates of rhythm genes (root-to-tip dN/dS) and USWS phenotype. Our data revealed 20 rhythm genes strongly correlated with the USWS phenotype in all marine mammals (supplementary table S9, Supplementary Material online), and 18 genes strongly correlated with cetacean sleep that is characterized by USWS and the near-complete absence of REM sleep (supplementary table S10, Supplementary Material online), as well as 21 genes strongly correlated with the sleep of non-cetacean marine mammals (supplementary table S11, Supplementary Material online), i.e. the combination of BSWS, USWS, and REM sleep. Of these, twelve genes (BMAL1, CHD9, CRY1, FABP7, KDM8, NOCT, RASD1, RBM4, RXRA, RYR3, SIRT, and ZFHX3) were found to have a significant correlation with two USWS-associated variables, and the evolutionary rates of ADCY2 and PER2 were associated with all three variables.

The Cetacean and Non-phocid Pinniped Positively Selected Gene FBXL21 Increased the Degradation of CRY1

To determine whether the functional convergence exists across independent lineages, we assessed one candidate, FBXL21 (F-box and leucine rich repeat protein 21), which was detected to be under positive selection in cetaceans and non-phocid pinnipeds (Fig. 1, supplementary table S3 and fig. S5, Supplementary Material online). FBXL21 has been shown to target the degradation of CRY within the cytoplasm and thus affects the period of the circadian rhythm (Yoo et al. 2013). Three-dimensional structure prediction indicated that dolphin (Tursiops truncatus) FBXL21 and fur seal (Callorhinus ursinus) FBXL21 lacked an α-helix in the F-box domain (39 to 85) as compared to mouse FBXL21 (Fig. 4a). The F-box domain is essential for ubiquitination activity by mediating the interaction between substrates and ubiquitin-conjugating enzymes (Kipreos and Pagano 2000). To test potential functional roles, we first examined the effects of mouse, dolphin, and fur seal FBXL21 on the stability of CRY1 following cycloheximide (CHX) treatment to prevent de novo protein synthesis in transfected cells. The expression of FBXL21 from all three species resulted in the degradation of CRY1 (Fig. 4b). However, both the addition of dolphin FBXL21 and the addition of fur seal FBXL21 caused a significantly faster degradation of CRY1 than the addition of mouse FBXL21; fur seal FBXL21 resulted in the fastest degradation rate (Fig. 4b). Additionally, mouse FBXL21 was more stable than dolphin FBXL21 and fur seal FBXL21 (Fig. 4b). Subcellular localization analysis subsequently revealed that all three FBXL21 proteins were predominantly distributed in the cytoplasm (Fig. 4c).

To test whether functional differences among these FBXL21 genes were sufficient to alter sleep behavior, we then monitored sleep and locomotor activity in zebrafish overexpressing mouse FBXL21, dolphin FBXL21, and fur seal FBXL21, respectively, under day–night cycles for 3 d between 5 and 8 dpf (Fig. 5a). We first found that larvae of these transgenic zebrafish were healthy and had normal morphological architecture at 4 dpf (supplementary fig. S1, Supplementary Material online). Sleep in zebrafish larvae is defined as an inactive bout of more than 1 min, based on the observations that larvae exhibit reduced responsiveness to stimuli after 1 min of inactivity (Prober et al. 2006). The analysis revealed that both the dolphin FBXL21 and fur seal FBXL21 overexpression zebrafish showed significantly less total sleep time than the mouse FBXL21 overexpression zebrafish (1,088 ± 132.1 min vs. 1,168 ± 101.5 min, P = 0.0035; 1,054 ± 179.0 min vs. 1,168 ± 101.5 min, P = 0.0002) and EGFP-negative control zebrafish (1,088 ± 132.1 min vs. 1,144 ± 118.4 min, P = 0.0400; 1,054 ± 179.0 min vs. 1,144 ± 118.4 min, P = 0.0044), whereas there were no differences between the mouse FBXL21 overexpression zebrafish and EGFP-negative control zebrafish (Fig. 5b and c). Both the dolphin FBXL21 and fur seal FBXL21 overexpression groups demonstrated less sleep time compared to the mouse FBXL21 overexpression group (603.4 ± 102.8 min vs. 654.3 ± 77.34 min, P = 0.0242; 583.8 ± 125.5 min vs. 638.3 ± 100.1 min, P = 0.0149) and EGFP-negative control group (603.4 ± 102.8 min vs. 638.3 ± 100.1 min, P = 0.0363; 583.8 ± 125.5 min vs. 654.3 ± 77.34 min, P = 0.0229) during the daytime (Fig. 5d), and no difference was observed among all groups at night (Fig. 5e). In parallel, locomotor activity analysis also showed that the dolphin FBXL21-overexpressing and the fur seal FBXL21-overexpressing larvae covered significantly greater total distances when compared to the mouse FBXL21-overexpressing larvae (1,361 ± 399.2 cm vs. 999.1 ± 261.8 cm, P = 0.0026; 1,761 ± 842.3 cm vs. 999.1 ± 261.8 cm, P < 0.0001) (Fig. 5f). Although the dolphin FBXL21-overexpressing and the fur seal FBXL21-overexpressing larvae showed no obviously increased mobile distance compared with the mouse FBXL21-overexpressing larvae during the daytime (Fig. 5g), both groups moved significantly greater distances than the EGFP-negative control group (796.7 ± 344.8 cm vs. 501.7 ± 243.8 cm, P = 0.0001; 718.7 ± 368.4 cm vs. 501.7 ± 243.8 cm, P = 0.0192) (Fig. 5g). Furthermore, the mobile distance recorded during the dark phase was significantly higher in the dolphin FBXL21-overexpressing and the fur seal FBXL21-overexpressing groups than in the mouse FBXL21-overexpressing group (540.1 ± 176.4 cm vs. 343.9 ± 102.7 cm, P = 0.0002; 1,042 ± 553.0 cm vs. 343.9 ± 102.7 cm, P < 0.0001) (Fig. 5h), while mobile distance was ∼443 cm shorter in the mouse FBXL21-overexpressing group versus the EGFP-negative control group (Fig. 5h). Further analysis of sleep architecture revealed that significantly decreased sleep time in the dolphin FBXL21-overexpressing and the fur seal FBXL21-overexpressing larvae was due to the reduction of the frequency of the specific sleep bout durations (i.e. >30 min) (Fig. 5i) rather than the number of sleep bouts and the sleep latency during the day–night switch (supplementary figs. S1 and S2, Supplementary Material online). Their data demonstrated that functional changes in dolphin FBXL21 and fur seal FBXL21 genes can lead to short sleep and increased locomotor activity.

Cetacean-Specific Mutation in NFIL3 Caused Reduced Repressor Activity and Lower Protein Stability

To investigate possible functional consequence of markedly adaptive signatures in cetaceans, we focused on NFIL3 (nuclear factor, interleukin 3 regulated, also called E4BP4), a key transcriptional repressor in circadian auxiliary feedback loop (Mitsui et al. 2001). NFIL3 is essential for determining the period length of the circadian rhythm (Yamajuku et al. 2011). In particular, PGLS analysis revealed that NFIL3 showed a significant association with cetacean sleep evolution; importantly, cetacean NFIL3 contains a unique nonsynonymous mutation (F411Y) in its key carboxyl terminal region (281 to 420) compared with other mammals (Fig. 6a and supplementary fig. S6, Supplementary Material online). This region has been confirmed to be necessary for transcriptional repression and sufficient for interaction with PER2 (Ohno et al. 2007). In vitro experiments were performed to determine the functional differences between mouse NFIL3, dolphin NFIL3, cetacean-specific mutated mouse NFIL3 (mouse_NFIL3_mut, F411Y), and mutated dolphin NFIL3 (dolphin_NFIL3_mut, Y411F). Dual-luciferase assays using the classic Per2 promoter construct showed that mouse NFIL3 caused nearly twice as much transcription repression activity compared with dolphin NFIL3 (P = 0.0043) (Fig. 6b). Furthermore, the repression activity of mouse_NFIL3_mut was significantly lower than that of mouse NFIL3 (P = 0.0137) (Fig. 6b), suggesting that the cetacean-specific mutation had a negative effect on the repression activity of NFIL3. In contrast, dolphin_NFIL3_mut had a significantly higher repression activity compared with dolphin NFIL3 (P = 0.0032) (Fig. 6b). Furthermore, protein stability assays showed that mouse NFIL3 was more stable than dolphin NFIL3 (Fig. 6c). Moreover, the mouse_NFIL3_mut protein was much less stable than mouse NFIL3, whereas the dolphin_NFIL3_mut protein was slightly more stable than dolphin NFIL3 (Fig. 6c).

Discussion

To mitigate the fundamental conflict between sleep and wakefulness, marine mammals (except phocids) have evolved a very special form of sleep termed USWS, a unique state during which one cerebral hemisphere sleeps while the other remains awake (Rattenborg et al. 2000). Although this unusual form of mammalian sleep has attracted much interest from researchers, the underlying genetic mechanisms remain elusive. As a general rule, the regulation of sleep is a complex process that involves dual modulation by both sleep homeostatic mechanisms and circadian rhythm (Borbély et al. 2016). The latter helps organisms to anticipate daily environmental changes and thus tailors sleep to the appropriate time of the day (Panda et al. 2002). Remarkably, the lateralized circadian driver has been reported as a major factor to determine the frequency of alternating USWS bouts based on the physiology-based quantitative model (Kedziora et al. 2012), which suggests that circadian system plays an important role in the evolution of USWS. In this study, we systematically examined the evolutionary characteristics of 147 circadian rhythm genes across representative mammals to uncover the evolutionary signals and molecular substitutions associated with USWS. Subsequently, we performed functional assays to evaluate functional consequences of the identified adaptive features in marine mammals. Together, our findings provide direct evidence for the molecular adaptations and corresponding functional changes in the circadian system that have facilitated adaptation to sleeping underwater.

Evolutionary Signatures of Rhythm Genes in Marine Mammals Reveal Potential Adaptations for Sleep Alterations

The evolution of USWS in marine mammals occurred to meet life-sustaining requirements, such as surfacing for respiration, more efficient monitoring environment, and thermogenesis (Lyamin et al. 2008). It is conceivable that an urgent demand for the rapid sleep–wake transition and flexible alternation of sleep between the two hemispheres may have been significant enough to drive genetic differences in sleep-related molecules. Concordantly, our analyses uncovered a much higher proportion of rhythm genes under positive selection in the USWS-specific lineages than the background lineages, for example, 1.36% (2/147), 2.73% (4/147), and 12.93% (19/147) rhythm genes were identified under positive selection in the manatee, non-phocid pinnipeds, and cetaceans after FDR correction, respectively. Although circadian regulation involves a complex signaling network, exogenous expression and the classical CHX chase assay have been widely applied to study the protein stability of rhythm genes and underlying molecular mechanisms (Sahar et al. 2010; Stojkovic et al. 2014). For example, CRY proteins are subject to degradation through FBXL21-mediated ubiquitination (Hirano et al. 2013). In our study, non-phocid pinniped and cetacean FBXL21 genes were found to undergo positive selection, and cell-based experiments also revealed the increased degradation of CRY1 and the decreased stability of FBXL21 itself. A growing body of evidence supports a key role of the stability of CRY proteins in the regulation of the mammalian circadian rhythm and sleep/wake timing (Busino et al. 2007; Hirano et al. 2017). A missense mutation in CRY2 has been demonstrated to destabilize CRY proteins in vivo and hence lead to perturbation of the circadian period (Hirano et al. 2016b). As expected, zebrafish behavioral assays further showed that functional changes in the FBXL21 gene from USWS-specific marine mammals caused significantly less sleep, and sleep alterations were more dramatic during the day, which is consistent with the fact that FBXL21 directs diurnal accumulation of CRY proteins in the cytoplasm and has strong rhythmic circadian expression with high mRNA levels during the mid to late day (Dardente et al. 2008; Hirano et al. 2013). Marine mammals exhibit a particular sleep rhythm as USWS, which allows them to maintain constant activity and alertness throughout the day (Yin et al. 2021). Therefore, the functional alterations of FBXL21 leading to destabilization of CRY proteins might help to generate the irregular circadian organization of activity and sleep during USWS in non-phocid pinnipeds and cetaceans by disturbing the circadian oscillator. These results strongly suggest that the circadian network has evolved adaptively and undergone functional modifications in USWS-specific lineages, which has allowed adaptation to the high regulatory demands for dramatic sleep changes during the transition from land to water.

We also found evidence of positive selection acting on rhythm genes in diverse marine mammals. For example, three PSGs (CACNA1C, CACNA1H, RYR3) associated with Ca^2+^ transmembrane transport were unique to the manatee, and three PSGs (BTRC, FBXL21, KDM8) relevant to ubiquitinated degradation occurred mainly along non-phocid pinniped lineages. In humans, altered sleep–wake cycles and sleep disturbance (e.g. difficulties falling or staying asleep, or excessive sleepiness) frequently result from the abnormal turnover or accumulation of circadian clock proteins (Hirano et al. 2016a). The Ca^2+^ concentration in SCN neurons is regarded as a major factor affecting the molecular oscillations of circadian transcription and light synchronization of daily rhythm (Ding et al. 1998; Colwell 2011). Consequently, a more flexible regulation of Ca^2+^ signaling may be important role for manatees in triggering the rapid onset of sleep in both two cerebral hemispheres or just in one hemisphere, as their sleep occurs during short apneas lasting several minutes when resting motionless (Mukhametov et al. 1992). In addition, ubiquitination, as a key posttranscriptional modulator, mediates the stability of the core circadian clock protein to maintain both the speed and the robustness of diurnal oscillations (Stojkovic et al. 2014). Semiaquatic pinnipeds may rely more on cumulative changes in circadian post-translational modifications to achieve sleep switching (from predominant BSWS when on land to USWS when in water), which is in part supported by the observation that fur seal FBXL21 significantly enhanced the degradation for CRY proteins compared with dolphin FBXL21 and mouse FBXL21. By contrast, cetacean PSGs were mainly found to be associated with cyclic AMP (cAMP) signaling and core clock elements in circadian TTFLs. Terrestrial ancestors of manatees, cetaceans, and pinnipeds have returned to the sea on multiple independent occasions and therefore under different genomic and environmental contexts. The imposition of similar constraints by the marine environment might dictate the occurrence of multiple evolutionary changes in different circadian components to attain similar sleep adaptations, as revealed from the evolution of mammalian hypoxia tolerance (Tian et al. 2017).

Unique Circadian System Adaptations Contribute to Cetacean Sleep

Cetacean sleep, characterized by USWS and the absence of REM sleep, is the most unusual form of mammalian sleep, even in marine mammals; sleep in other aquatic lineages usually comprises USWS intermingled with BSWS and REM sleep (Mascetti 2016). Our evolutionary analysis uncovered that half of the rhythm genes (50.3%) have been rapidly evolving and 34 genes exhibited a signature of positive selection across cetaceans, suggesting that the strong demand for sleep regulation has greater accumulation of nonsynonymous mutations in cetacean rhythm genes to allow adaptive changes in circadian regulation. Notably, cetacean PSGs are found widespread throughout the circadian rhythm pathway and have close functional interactions (supplementary fig. S4, Supplementary Material online). More importantly, a high proportion of PSGs was found to be associated with the core circadian TTFL loop and cAMP signaling, two critical areas closely related to sleep control (Hendricks et al. 2001; Franken 2013). Indeed, studies in mice with targeted disruption or mutation of core circadian clock genes have shown changes in circadian rhythmicity as well as changes in sleep architecture, amount, and quality (Franken and Dijk 2009). For example, the CLOCK gene was detected to be under positive selection in cetaceans. The deletion of the Clock gene in mice led to attenuation of the distribution of sleep and wakefulness rhythm, with a significant decline in sleep duration of 1 to 2 h, and a smaller increase in REM sleep after sleep deprivation (Naylor et al. 2000). Similarly, the DEC2 gene, which serves as an essential transcriptional repressor of the negative feedback loop in the circadian clock (Honma et al. 2002), was also found to have undergone positive selection in cetaceans. A missense mutation in the human DEC2 gene resulted in decreased suppression for orexin expression, thus resulting in increased vigilance time and shorter sleep time in transgenic mouse models (Hirano et al. 2018). Furthermore, the observation that the increased cAMP levels reduced sleep or sleep-like behavior in mice, flies, and nematodes (Hendricks et al. 2001; Graves et al. 2003; Raizen et al. 2008), provides strong evidence that cAMP signaling play a phylogenetically conserved role in promoting wakefulness. The identified evolutionary signals may be advantageous for improving the wake-promoting ability, which is important for cetaceans to overcome sleep-inducing effects in the whole brain to maintain unihemispheric waking during USWS. These findings further support our hypothesis that circadian rhythm genes were targets of natural selection in cetaceans and played a crucial role in the evolution of cetacean sleep by facilitating adaptation to stronger demands for sleep inhibition.

Interestingly, we observed that 62.58% (92/147) of rhythm genes exhibit cetacean-specific mutations in functional domains, further strengthening the conclusion that the cetacean circadian system has adaptively changed to create favorable conditions for adopting unique sleep. As an example, in the PER2 gene, which was significantly correlated with the evolution of USWS, 29 cetacean-specific mutations were identified. As it is a central element of the f circadian negative feedback loop, species-specific mutations in the PER2 gene have been associated with the extreme short sleep of Rothschild's giraffe (Giraffa rothschildi) and the circadian arrhythmicity in reindeer (Rangifer tarandus) and pygmy lorises (Xanthonycticebus pygmaeus) (Lin et al. 2019; Liu et al. 2021; Li et al. 2022). Unique amino acid substitutions and potential functional alterations in the cetacean PER2 gene were expected to help sleep escape from rhythmic constraints by weakening the ability of PER2 to bind with CRY proteins, which is required for sustained photic-induced circadian rhythms (Ukai-Tadenuma et al. 2011). In particular, the NFIL3 gene was found to be significantly associated with the evolution of cetacean sleep, and the identified cetacean-specific mutation (F411Y) was further shown to reduce repressor activity and the stability of NFIL3 inside the cell. NFIL3 plays an important role in determining the period length of the circadian oscillator by suppressing transcriptional activity to regulate the stabilization and fine-tuning of the core PER/CRY feedback loop (Mitsui et al. 2001). The knockdown of Nfil3 was shown to produce a long-period circadian rhythm in cells (Yamajuku et al. 2011). The observed instability of NFIL3 and weakened repression caused by the cetacean-specific mutation contribute to lengthened locomotor rhythm, which is consistent with the fact that cetaceans require more active time to surface regularly to breathe and to continuously produce heat in thermally challenging aquatic environments (Lyamin et al. 2008). In addition, two unique and radical amino acid changes (A76P and D111N) were found in the highly conserved binding region of the FABP7 gene, which is reported to be a clock-controlled gene involved in sleep regulation and cognitive function (Gerstner et al. 2023). The FABP7 T61M missense mutation was reported to lead to fragmented sleep in humans, mice, and flies (Gerstner et al. 2017). Cetaceans always maintained swimming, even during sleep (Castellini 2002); the increased sleep fragmentation might be beneficial to stimulate continuous motion without full awakening. Collectively, these results not only suggested that several key genes and proteins in the circadian network were changed in cetaceans but also that changes have occurred throughout the pathway, which may be important in facilitating the adaptation of cetaceans to a fully aquatic lifestyle.

Shared Shifts in the Circadian System Underpin the Convergent Evolution of USWS in Marine Mammals

In addition to lineage- or species-specific adaptations, signals of convergent evolution were also found in the circadian network across marine mammals. First, eight rhythm genes in total were shown to have convergent signatures of positive selection, including three PSGs (BMAL2, PLCB2, NCOA6) shared by cetaceans and the manatee and five PSGs (PER3, FBXL21, KDM2A, GRIA1, GNAO1) shared by cetaceans and non-phocid pinnipeds. Interestingly, these genes all had an established relationship with the regulation of sleep and movement, except the PLCB2 gene. For two major members of the circadian TTFL, the polymorphisms of the BMAL2 gene were significantly related to diurnal preference for sleep timing (Parsons et al. 2014), whereas mutations of the PER3 gene in mice caused phase shifts in the sleep–wake cycle and alterations in the homeostatic responses to sleep loss (Hasan et al. 2014; Zhang et al. 2016). Notably, two cetacean-specific mutations in BMAL2 have been found to enhance the PER2 expression specifically and the related wakefulness-promoting effect to help cetaceans to depress whole-brain sleep demand during USWS (Yin et al. 2024). KDM2A (also known as FBXL11), a histone demethylase, functions as a negative element in the mammalian circadian clock by repressing the CLOCK/BMAL1-induced transcription; the knockdown of the KDM2A gene leads to activity period shortening and its overexpression leads to period lengthening (Reischl and Kramer 2015). Furthermore, GRIA1 knockout mice revealed a striking reduction in sleep electroencephalogram spindle activity and longer REM sleep episodes (Ang et al. 2018). Accumulated changes in these sleep-related molecules may reflect the adaptive adjustment of sleep mechanisms by marine mammals to adopt unihemispheric sleep. Remarkably, both the NCOA6 and GNAO1 genes were reported to be associated with involuntary motor disturbances. The NCOA6 gene is one of the putative causal genes for a sleep-related sensorimotor disorder, restless legs syndrome, which is a strong and uncontrollable urge to move the legs during periods of rest, especially in the evening and night (Akçimen et al. 2020). Variations in the GNAO1 gene are responsible for a type of epileptic encephalopathy with severe hyperkinetic movements (Waak et al. 2018), and mouse models carrying the target mutation exhibited strong hyperactivity in behavioral assays (Silachev et al. 2022). Sleep is a dangerous state for marine mammals as they gradually lose muscular activity and awareness of their environment. The changes in NCOA6 and GNAO1 could therefore be conducive to promoting the excitation of motor impulses to achieve continuous movement during USWS to meet the needs of breathing and monitoring predators.

Next, we identified 27 convergent amino acid replacements in 17 rhythm genes. Notably, seven of the twelve convergent candidate genes shared by cetaceans and the manatee were related to Ca^2+^ transmembrane transport, including four genes encoding voltage-gated calcium channel subunits (CACNA1D, CACNA1I, CACNA1G, CACNA1H) and three genes involved in the release of intracellular calcium (ITPR3, RYR1, PLCB3) (Sorrentino et al. 2000; Perez-Reyes 2006). Besides the role of Ca^2+^ levels in circadian oscillations, Ca^2+^-dependent hyperpolarization also has a role in the regulation of sleep duration; specifically, the knockout of CACNA1G or CACNA1H decreases sleep duration in mice (Tatsuki et al. 2016). These convergent changes, in combination with specific positive selection, highlight the importance of Ca^2+^ signaling in the generation of USWS, especially in the manatee. Furthermore, three convergent amino acid substitutions were identified between the manatee and non-phocid pinnipeds in three genes (ADCY2, ADCY7, NOS1) relevant to CREB phosphorylation. Studies have documented that the increase in Ca^2+^ activates a number of signaling pathways such as ADCY-mediated signaling and NOS1-mediated signaling, to initiate the phosphorylation of CREB (Golombek and Rosenstein 2010). CREB phosphorylation only occurs when light induces behavioral phase shifts in the circadian cycle, and then phosphorylated CREB drives the expression of circadian clock genes (Colwell 2011). Convergent amino acid substitutions may affect light-regulated molecular responses, which could subsequently influence the sleep rhythms disrupted during USWS. It should also be noted that SREBF1 is among only three genes that have convergent amino acid substitutions identified in more than two USWS-specific lineages. In particular, three convergent mutations of SREBF1 generated radical and possibly damaging amino acid changes and were located in important functional domains. For example, two convergent mutations (P505L, V726M) were found in C-terminal transmembrane and regulatory domain. Mutations at the nearby amino acid residues, R527 and L530, were reported to affect nuclear translocation and decrease transcriptional activity of SREBP1 (Wang et al. 2020). Genome-wide association studies classified that nucleotide variants in the human SREBF1 gene were associated with excessive daytime sleepiness, a symptom of chronic insufficient sleep (Wang et al. 2019). The increased activity of SREBF1 was recently reported to induce sleep deficits and reduce sleep pressure at the night-time onset in the Cyfip heterozygous flies. Overall, although no convergent amino acid replacement was identified in all marine mammals, this observation suggests that alterations in different parts of the circadian rhythm system across USWS-specific lineages may converge to alter circadian aspects of sleep regulation. Our results provide evidence of the convergent selective pressures and genetic responses shared by marine mammals.

Furthermore, five rhythm genes were under positive selection in the terrestrial counterpart one (CRTC1) in Hawaiian monk seal (Neomonachus schauinslandi) and four (ADCY9, BTRC, CACNA1D, ITPR1) in the African elephant (Loxodonta africana). Phocid seals exhibit a lot of phenotypic differences from otariids other than the absence of USWS, for instance, phocids have greater ability to store energy than otariids (Costa and Williams 1999). The available data indicated that phocid mothers store significantly more fat than otariids, because they need to rear pups from body stores for the entire lactation period whereas otariids feed their pups by repeated foraging during lactation (Trillmich and Weissing 2006). CRTC1 engaged in affecting photic resetting of the circadian system (Jagannath et al. 2013) and was found to be required for appetite, fat accumulation and energy balance (Altarejos et al. 2008; Altarejos and Montminy 2011). The lack of CRTC1 triggers hyperphagic obesity, abnormal expansion of the white adipocytes, and reduced energy expenditure in mice (Breuillaud et al. 2009; Hu et al. 2021). The accumulated amino acid variations in phocid CRTC1 might be associated with their rapid fat storage for the high lactation energy consumption. Additionally, elephants are famous for their severely short sleep duration and excellent long-term chemical and social memory (Tobler 1992; Hart et al. 2008). Elephants are recorded to sleep for 2 h per day and are thus called the shortest sleeping mammals (Gravett et al. 2017). Concordantly, we found evidence of positive selection of two genes related to sleep duration. Rhythm gene BTRC is thought to act through glucocorticoid stress-related pathways to influence sleep duration (Dashti et al. 2019), which also showed signals of positive selection in non-phocid pinnipeds. CACNA1D (encoding CaV1.3 channel) is the main contributor to the pacemaking of substantia nigra dopaminergic neurons involved in maintaining arousal (Monti and Monti 2007; Surmeier et al. 2017). It should also be noted that a shared Ala-to-Thr mutation was observed at codon 865 of cetacean and non-phocid pinniped CACNA1D, which is close to the extracellular transmembrane helices S1–S2 linker of repeat III. A similar variation (Arg930His) was found to induce voltage-dependent alterations of Cav1.3 channel gating (Rinné et al. 2022). Convergent CACNA1D amino acid substitution may have an effect on the pacemaker activity of wake neurons. In addition, ITPR1 and ADCY9 are essential for synaptic plasticity (Mons et al. 2004; Tsuboi et al. 2015). ITPR1 dysfunction has been linked to learning and memory impairments (Tada et al. 2016). Adaptative changes in these two genes are possibly beneficial for memory consolidation of elephants.

Conclusion

Sleep is almost ubiquitous throughout the animal kingdom, although little is known about how ecological factors shape the diversity of sleep. Our work provides evidence for positive selection signals, convergent changes, and lineage-specific innovations in the circadian system of marine mammals, which may reflect molecular adaptations to the unique USWS in the face of environmental constraints of sleeping in water. In vitro and in vivo assays further demonstrate the functional modifications in the FBXL21 and NFIL3 genes, two important circadian regulators with adaptive signatures, which were likely beneficial for the formation of USWS in marine mammals by facilitating the escape from circadian control and lengthening locomotor rhythms. Furthermore, the observation that rapidly evolved genes and unique amino acid substitutions were frequently identified in the circadian network of cetaceans strengthens our conclusion that circadian rhythm genes were targets of natural selection that allow adaptive changes in circadian sleep regulation and thus played a crucial role in the development of unusual sleep pattern of cetaceans. In summary, this study revealed the influence of the circadian rhythm on the evolution of USWS and has shed novel insights into the genetic mechanisms underlying variations in mammalian sleep. In addition, the present investigation is limited by the experimental conditions; the USWS-related genes and sites, particularly adaptive modifications associated with ion channels and signaling, are still to be evaluated in future research, which possibly provides insights into understanding how genetic information influence and shape sleep.

Materials and Methods

Data Acquisition and Processing

We retrieved 147 genes that covered each major step in the circadian rhythm pathway as evenly as possible, based on the KEGG (https://www.genome.jp/kegg/) (Kanehisa and Goto 2000) and Reactome (https://reactome.org/) pathway databases (Croft et al. 2011). Protein-coding sequences for the human circadian rhythm genes were first extracted from the NCBI database as queries. BlastN and tBLASTn searches were then performed to identify one-to-one orthologs in 30 high-quality mammal genomes with the longest transcript retained. These 30 mammals included 14 USWS-specific marine mammals (nine cetacean lineages, four non-phocid pinniped lineages, and the manatee T. manatus), 1 seal species Hawaiian monk seal, and 15 terrestrial mammals (supplementary table S1, Supplementary Material online). The sequences were aligned using the OMM_MACSE pipeline (Douzery et al. 2014) by three alignment MACSE v2 (Ranwez et al. 2018), MAFFT v.7 (Katoh and Standley 2013), and HmmCleaner v. 0.180750 (Di Franco et al. 2019). To reduce the number of false-positive predictions, multiple sequence alignments were further refined with Gblocks v.0.91b (Talavera and Castresana 2007) by filtering for incorrect alignment. Each alignment was examined by hand and edited as necessary. The resulting data set that contained 147 rhythm genes from 30 major mammalian lineages was used in the subsequent analyses.

Ancestral State Reconstruction

The sleep phenotype data from multiple references (Lyamin et al. 2008; Mascetti 2016) and a well-supported mammal phylogeny from TimeTree (http://www.timetree.org/) (Kumar et al. 2017) were used to reconstruct the ancestral character states via the phytools and ape packages in R v4.0.5 (https://www.r-project.org/) (Paradis et al. 2004; Revell 2012). We first tested the fitness of the following Equal Rate discrete character evolution in which a single parameter governs all transition rates; All Rates Different all possible transitions can occur at different rates; and Symmetrical Rates forward and reverse transitions share the same parameter. We then used the model with the best Akaike information criterion score to reconstruct the ancestral state of mammalian sleep pattern.

Phylogenetic Comparative Approach

For each gene, we used phylogenetic analysis by maximum likelihood (PAML v4.9) to measure the rate of evolution across mammalian phylogeny and to search for molecular signatures of adaptive evolution (Yang 2007). A well-accepted species tree was obtained from the TimeTree (http://www.timetree.org/) and served as an input tree for all analyses (Fig. 1) (Kumar et al. 2022). The ratio of nonsynonymous (dN) to synonymous (dS) mutations (ω = dN/dS) was estimated using the CODEML program in PAML v4.9 to evaluate the selective pressure for each gene (Yang 2007). In brief, ω < 1, ω = 1, and ω > 1 indicate negative purifying selection, neutral evolution, and positive selection, respectively. First, we tested the one-ratio model (model = 0), which has a single ω ratio for the entire tree. Next, we used the free-ratio model (model = 1), which allows different ω values across branches to estimate linage-specific evolutionary rates for each branch. To identify candidates with adaptive signals in cetaceans as well as non-phocid pinnipeds and the manatee, respectively, we further analyzed the following two gene (i) rapidly evolving genes (REGs), which have a higher ω (regardless of whether it is greater than 1) than the background, were identified using the two-ratio model (model = 2); and (ii) PSGs, with specific codons in a particular branch under positive selection, were identified with an optimized branch-site model (model = 3) (Zhang et al. 2005). We also repeated the tests on closely related sister taxa (e.g. the cattle, Bos taurus as a sister species to cetaceans) (branches a, c, and e in Fig. 1). Nested models were compared using a likelihood ratio test (LRT) with a χ^2^ distribution. P-values of <0.05 indicates a significantly better fit by the alternative model. P-values were adjusted for multiple testing using the FDR procedure, and adjusted P-values of <0.05 were considered significant for branch-site model analysis (Anisimova and Yang 2007).

To assess the divergent evolution in rhythm genes along the cetacean clade, non-cetacean USWS clade, and non-USWS clade, we conducted the CmC test (Bielawski and Yang 2004) using the CODEML program in PAML v4.9. This model allows for tests of differences in selection at a subset of sites between two or more partitions of a tree. CmC has been shown to be extremely useful in testing for long-term shifts in selection pressure associated with changes in function and ecology (Schott et al. 2014; Schott et al. 2019). The CODEML setting chosen for CmC test was model = 3 with NS sites = 2 (selection). Using LRT, the divergent evolution of the selected clade (CmC test) was statistically compared to the null model M2a_rel (model: 0, NS 22), which does not allow for divergence across the phylogeny.

Detection of Convergent or Unique Amino Acid Substitutions

FasParser (version 2.3.0) (Sun 2018) was used to identify specific amino acids shared between each combination of two or three USWS-specific lineages. Only the specific amino acid changes presented in at least 80% of USWS lineages were considered as convergent candidates. Further, convergent amino acid substitutions in ancestral branches were detected via the method described in Zou and Zhang (2015). Briefly, the ancestral node sequences for each gene were reconstructed using the CODEML algorithm in PAML v4.9 package (Yang 2007). Then, a tree with branch lengths was extracted from the output file from the abovementioned calculation, and the relative substitution rates of all amino acid sites within one gene were calculated using the aaml model in CODEML. All observed and expected cases of convergent amino acid substitutions were calculated under the recommended JTT-fgene matrix. To filter out noise resulting from chance amino acid substitutions, the Poisson test was performed to assess the significance differences in the observed and expected numbers in the target gene. Additionally, an amino acid substitution was considered specific if the cetacean amino acid differed from that of all the other species.

To estimate the functional significance of identified changes, we further classified them as radical amino acid mutations if the ancestral and derived amino acid belong to a different physicochemical groups using an amino acid classification that is based on charge, polarity, and positive charge (R, K, H), negative charge (E, D), aliphatic and nonpolar (A, I, L, M, V, G), polar and uncharged (S, T, C, P, N, Q), and nonpolar and aromatic (F, W, Y) (Lee et al. 2018). Pfam (version 1.6, Pfam-A database with default parameters) was then used to determine whether mutations were located in the functional domain of the protein structure (Coggill et al. 2008). Three free online protein structure prediction programs, PolyPhen-2 (http://genetics.bwh.harvard.edu/pph2/index.shtml), SIFT (http://provean.jcvi.org/) and PROVEAN (http://provean.jcvi.org), were then used to predict the impact of these conserved mutations using the default cutoff values. Human protein sequences were used as queries for prediction programs. The predictions “possibly damaging” or “probably damaging” by PolyPhen-2, “damaging” by SIFT, and “deleterious” by PROVEAN were all regarded as function-altering substitutions. 3D protein structures were calculated with alphafold v2.3.1 using the “monomer_ptm” model (Jumper et al. 2021).

Gene–Phenotype Association Analyses

PGLS were used to detect different variable relationships while accounting for phylogenetic relationships. PGLS incorporates phylogenetic information into generalized linear models to provide a powerful method for analyzing continuous data, and this has been utilized to test for correlations among the evolutionary models and the relationships among life-history traits. In the PGLS regression analysis, the value of the lambda (λ) was estimated using ML to optimally adjust the degree of phylogenetic correlation among data sets. The λ values can range from 0 to 1; λ = 0 indicates no phylogenetic signal whereas λ = l indicates a strong phylogenetic signal (Xu et al. 2017). To explore the potential relationships between the evolutionary rate (ω) of rhythm genes and USWS traits, the root-to-tip ω that considers the evolutionary history of a locus was calculated for each gene using the free-ratio model from PAML v4.9. Genes with dS < 0.0001 were discarded in the present analysis. Finally, PGLS was performed in the caper package in R 3.5.1 (Orme et al. 2013) to evaluate the associations between Log10-transformed root-to-tip ω and USWS traits (Yu et al. 2021).

Plasmid Construction

The full-length cDNA sequences of mouse FBXL21, dolphin FBXL21, fur seal FBXL21, dolphin NFIL3, and mouse NFIL3 were synthesized by Beijing Tsingke Biotech Co., Ltd, which were cloned into pIRES2-EGFP vector with XhoI and EcoRI sites, respectively. Oligonucleotides encoding the FLAG epitope sequence were fused to the 3′-end of the three complete FBXL21 cDNA sequences and the 5′-end of the two complete NFIL3 cDNA sequences, respectively. For transient transgenic expression, the cDNA sequences of dolphin FBXL21, fur seal FBXL21, and mouse FBXL21 were PCR-amplified and subcloned in a pCS2-EGFP vector with BamHI and EcoRI sites to express fusion proteins with a C-terminal EGFP green fluorescent protein tag. All constructs were verified by DNA sequencing. Cetacean-specific change (F411Y) in the NFIL3 gene was confirmed by PCR amplification and direct sequencing. Dolphin tissue samples used in the present study were collected from dead individuals in the wild. The mouse_NFIL3_mut plasmid was created using QuickMutationPlus Site-Directed Mutagenesis Kit (Beyotime Biotechnology). After the mutagenesis, the entire coding region was sequenced to confirm the presence of the desired mutation and the absence of additional mutations.

In Vitro Functional Assays of FBXL21

The effects of FBXL21 proteins on CRY degradation rate were determined as described in Godinho et al. (2007). In brief, HEK293T cells were cultured in Dulbecco's modified Eagle medium supplemented with 10% fetal bovine serum. Cells were plated out into the six-well plates and transfected with 0.3 μg CRY1 and 0.7 μg FBXL21 (dolphin, fur seal, or mouse) using jetPRIME reagent (Polyplus) in accordance with the manufacturer's protocol. After 36 h, CHX (MedChemExpress) was added to a final concentration of 20 μg/mL and incubated for 0, 3, 6, or 9 h. Cells were lysed in 500 μL RIPA buffer and analyzed by Western blotting using anti-FLAG antibody (1:1,000, Sigma). For the subcellular localization studies, 2 × 10^5^ 293T cells were incubated in the 24-well plates with cell slide inserts on the day before transfection. At 48 h after transfection, cells were fixed at room temperature for 15 min in 4% paraformaldehyde in PBS (10 mM Na-phosphate buffer, 140 mM NaCl, 1 mM MgCl2, pH 7.4). Fixed cells were pretreated with 0.1% Triton X-100 in PBS at room temperature for 15 min and then incubated at room temperature for 20 min in a blocking PBS solution containing 1% (v/v) bovine serum albumin (Sigma). Then, the cells were incubated with anti-FLAG antibody (1:2,000, ABclonal) diluted in the blocking solution at 37 °C for 2 h. After rinsing with PBS, the cells were incubated with goat anti-mouse IgG antibody (1:200, Jackson ImmunoResearch) diluted in the blocking solution at 37 °C for 1 h, followed by nuclear staining with DAPI. Samples were observed under confocal laser scanning microscopy (Olympus FV3000).

Experimental Animals

All zebrafish lines were maintained on a 14 h/10 h light-dark dark cycle and fed three times daily at the zebrafish facility of Southern Marine Science and Engineering Guangdong Laboratory (Guangzhou). Embryos were obtained by natural crosses and were staged according to standard developmental conditions, and fertilized eggs were raised at 28.5 °C. All animal protocols were conducted in accordance with guidelines approved by the Animal Ethical and Welfare Committee of Southern Marine Science and Engineering Guangdong Laboratory (Guangzhou).

Transgenic zebrafish that overexpressed EGFP, mouse FBXL21, dolphin FBXL21, and fur seal FBXL21 transiently were generated as previously described (Yin et al. 2024). Briefly, the empty plasmid (pCS2-EGFP) and constructed plasmids were linearized with XbaI and then transcribed into mRNAs using mMESSAGE mMACHINE SP6 Transcription Kit (Thermo Fisher Scientific, Waltham, MA, USA), respectively. Embryos injected with EGFP mRNA were sampled at equivalent developmental stages and used as the negative control. mRNA products were diluted to 400 ng/μL with diethylpyrocarbonate-treated water for microinjection. The mRNA products were microinjected into the one-cell-stage wild-type embryos of the AB-type background using an Applied Scientific Instrumentation pressure injection system under a stereomicroscope (Leica).

In Vivo Behavioral Assays

Zebrafish larvae behavioral assays were performed as previously reported (Ricarte et al. 2024) using Noldus DanioVision tracking systems (Noldus Information Technology, Wageningen, The Netherlands). Free-swimming 4.5 dpf (days postfertilization) larvae were habituated in 96-well square plates, with one larva per well, in the observation chamber of the Danio Vision tracking system for one night. After 10 h of habituation, each larva was recorded under controlled lighting conditions (10 h of dark and 14 h of light cycles) from 5 dpf. The sleep architecture and locomotion activity of each larva were tracked for 3 consecutive days and analyzed by EthoVision XT 17.5 software (Noldus, Wageningen, The Netherlands).

Functional Assays of NFIL3

To examine the transcriptional repressor activity of NFIL3, dual-luciferase assays were performed in a 24-well plate format by transiently transfecting HEK293T cells. Cellular lysates were collected 36 h after transfection using passive lysis buffer. Luciferase activity was measured using the Dual-Luciferase Reporter Assay Kit (Vazyme Biotech Co., Ltd) on a Synergy H1 Multi-Mode microplate reader (BioTek, USA). To normalize the samples, light output from transcriptional activity was divided by the output from Renilla luciferase activity. Next, the protein stability of NFIL3 was evaluated by using CHX time-course assay in HEK293T cells overexpressing the mutants or the WT form of NFIL3 as previously described (Kostrzewski et al. 2018; Kurien et al. 2019).

Statistical Analysis

Statistical calculations were performed using the GraphPad Prism 9 software. The D’Agostino and Pearson normality and Shapiro–Wilk tests were used to check the data distribution. If data were normally distributed, N-way ANOVA (alpha = 0.05) was used with correction for multiple comparisons using Holm–Sidak multiple posthoc test. If non-parametric, the Kruskal–Wallis test was used with correction for multiple comparisons using Dunn–Sidak (alpha = 0.05). All experiments were independently repeated at least three times. All data were expressed as the mean ± SEM. The labels “ns,” “,” “,” “,” and “****” indicate not significant, P < 0.05, P < 0.01, P < 0.001, and P < 0.0001, respectively.

Supplementary Material