The zebrafish mutant dreammist implicates sodium homeostasis in sleep regulation

Sleep is a nearly universal feature of animal behaviour, yet many of the molecular, genetic, and neuronal substrates that orchestrate sleep/wake transitions lie undiscovered. Employing a viral insertion sleep screen in larval zebrafish, we identified a novel gene, dreammist (dmist), whose loss results in behavioural hyperactivity and reduced sleep at night. The neuronally expressed dmist gene is conserved across vertebrates and encodes a small single-pass transmembrane protein that is structurally similar to the Na+,K+-ATPase regulator, FXYD1/Phospholemman. Disruption of either fxyd1 or atp1a3a, a Na+,K+-ATPase alpha-3 subunit associated with several heritable movement disorders in humans, led to decreased night-time sleep. Since atpa1a3a and dmist mutants have elevated intracellular Na+ levels and non-additive effects on sleep amount at night, we propose that Dmist-dependent enhancement of Na+ pump function modulates neuronal excitability to maintain normal sleep behaviour. Significance statement Sleep is an essential behavioral state, but the genes that regulate sleep and wake states are still being uncovered. A viral insertion screen in zebrafish identified a novel sleep mutant called dreammist, in which a small, highly-conserved transmembrane protein is disrupted. The discovery of dreammist highlights the importance of a class of small transmembrane-protein modulators of the sodium pump in setting appropriate sleep duration.


INTRODUCTION 37
The ability of animals to switch between behaviourally alert and quiescent states is of screening strategies in revealing novel sleep-wake regulatory genes suggests that more 65 sleep signals likely remain to be discovered. 66 One of the lessons from these genetic screens is that many of the uncovered genes play 67 conserved roles across species. For example, Shaker also regulates mammalian sleep 68 identified orthologs in most vertebrate clades, including other species of teleost fish, birds, 162 dmist expression was reduced by 70% in the viral insertion line, suggesting that dmist vir is a 213 hypomorphic allele. To confirm that the sleep phenotypes observed in dmist vir/vir animals are 214 due to the loss of Dmist function, we used CRISPR/Cas9 to create an independent dmist loss 215 of function allele. We generated a zebrafish line in which the dmist gene contains an 8 bp 216 insertion that causes a frameshift and early stop codon (dmist i8 , Figure 3A). The dmist i8 allele 217 is predicted to encode a truncated protein lacking the complete signal peptide sequence and 218 transmembrane domain ( Figure 3B), indicating this is likely a null allele. RT-qPCR showed that 219 dmist transcript levels were 60% lower in dmist i8/i8 fish compared to wild type siblings, 220 consistent with nonsense-mediated decay ( Figure S4A We next assessed the sleep and activity patterns of dmist i8/i8 fish. As seen in exemplar 222 individual tracking experiments, dmist i8/i8 larvae sleep less at night due to fewer sleep bouts 223 and also show an increase in waking activity relative to wild type and heterozygous mutant 224 siblings ( Figure 3C-H). This significant night-time reduction in sleep and increase in 225 hyperactivity is also apparent when combining 5 independent experiments with a linear mixed 226 effects (LME) model to normalize behaviour across datasets ( Figure 3I). Although dmist vir/vir 227 larvae also sleep less at night ( Figure 1G), the large day-time reduction in sleep observed in 228 dmist vir/vir larvae is absent in dmist i8/i8 animals, perhaps due to differences in genetic 229 background that affect behaviour. Because the dmist vir is likely a hypomorphic allele, we 230 focused subsequent experiments on the CRISPR-generated dmist i8/i8 larvae. 231 To test whether the increased night-time activity of dmist i8/i8 mutants persists in older 232 animals, we raised dmist i8/i8 mutants with their heterozygous and wild type siblings to adulthood 233 in the same tank and tracked individual behaviour for several days on a 14:10 light:dark cycle. 234 As in larval stages, dmist i8/i8 adults were hyperactive relative to both dmist i8/+ and dmist +/+ 235 siblings, maintaining a higher mean speed at night ( Figure 3J-L). This suggests that either 236 Dmist affects a sleep/wake regulatory circuit during development that is permanently altered in 237 dmist mutants, or that Dmist is continuously required to maintain normal levels of night-time 238 locomotor activity. 239 240 Dmist is distantly related to the Na + /K + pump regulator Fxyd1 (Phospholemman) 241 Because Dmist is a small, single pass transmembrane domain protein without any clear 242 functional motifs and has not been functionally characterized in any species, we searched for 243 similar peptides that might provide clues for how Dmist regulates behaviour. Using the multiple 244 sequence alignment tool MAFFT to align the zebrafish, mouse, and human Dmist peptides 245 Fxyd1 share 27-34% amino acid homology, including an RRR motif at the C-terminal end, 250 although Dmist lacks a canonical FXYD sequence ( Figure 4A). In addition, computational 251 predictions using the AlphaFold protein structure database revealed structural similarities 252 between Dmist and Fxyd1 (Jumper et al., 2021), suggesting that Dmist may belong to a class 253 of small, single pass transmembrane ion pump regulators. 254 Using In situ hybridisation, we found that fxyd1 is expressed in cells along the brain ventricle 255 and choroid plexus ( Figure 4C) in contrast to the neuronal expression of dmist ( Figure 2D). 256 Despite these different expression patterns, based on their sequence similarity we reasoned 257 that Fxyd1 and Dmist may regulate the same molecular processes that are involved in sleep. 258 To test this hypothesis, we used CRISPR/Cas9 to generate a 28 bp deletion in the third exon 259 of the zebrafish fxyd1 gene, causing a frameshift that is predicted to encode a truncated protein 260 that lacks the FXYD, transmembrane, and C-terminal domains ( Figure 4B). Contrary to a 261 previous report based on morpholino knockdown (Chang et al., 2012), fxyd1 28/28 larvae were 262 viable with no detectable defect in inflation of the brain ventricles. We therefore tested fxyd1 263 mutant larvae for sleep phenotypes. Like dmist mutants, fxyd1 28/28 larvae slept less at night 264 ( Figure 4D-F). Interestingly, this sleep loss is mainly due to shorter sleep bouts ( Figure 4F), 265 indicating that fxyd1 mutants initiate sleep normally but do not properly maintain it, unlike dmist 266 mutants, which initiate fewer night-time sleep bouts, although in both cases there is 267 consolidation of the wake state at night ( Figure 3I, 4F). Thus, despite the non-neuronal 268 expression of fxyd1 in the brain, mutation of the gene most closely related to dmist results in a 269 similar sleep phenotype. 270 271 The brain-wide Na + /K + pump alpha subunit Atp1a3a regulates sleep at night 272 Given the similarity between Dmist and Fxyd1 and their effects on night-time sleep, we 273 hypothesized that mutations in Na + /K + pump subunits known to interact with Fxyd1 might also 274 affect sleep. Consistent with this hypothesis, a low dose of the Na + /K + pump inhibitor, ouabain, 275 reduced night-time sleep in dose-response studies ( Figure S5A). When applied in the late 276 afternoon of 6 dpf, 1 µM ouabain decreased subsequent night-time sleep by 16.5% relative to 277 controls, an effect size consistent with those observed in dmist mutants ( Figure 5A, C). Night-278 time waking activity was also significantly increased after low-dose ouabain exposure ( Figure  279 5B, D). Ouabain binds to specific sites within the first extracellular domain of Na + /K + pump 280 alpha subunits (Price and Lingrel, 1988), and species-specific changes to these sites confers 281 species-specific ouabain resistance, as in the case of two naturally occurring amino acid 282 substitutions present in the Atp1a1 subunit of mice (Dostanic et al., 2004). Alignment of the 283 ouabain sensitive region of zebrafish and mouse Na + /K + pump alpha subunits revealed that 284 zebrafish Atp1a1a lacks the conserved Glutamine at position 121 ( Figure 5E), suggesting that 285 one of the other subunits with conserved ouabain-binding sites is responsible for the low dose 286 ouabain sleep effects. We focused on the Na + /K + pump alpha-3 subunit (Atp1a3), as this has 287 13 been shown to directly interact with Fxyd1 in mammalian brain tissue (Feschenko et al., 2003). 288 Murine Dmist expression also correlates well with the Atp1a3 distribution across 5 brain cell 289 types in mouse (Pearson correlation coefficient = 0.63), which has the strongest correlation 290 score with neuronal markers ( Figure S5B compared to Figure S3F). In contrast, zebrafish 291 atp1a2a is reportedly expressed in muscle at larval stages, while atp1a1b is confined to cells  Zebrafish have two Atp1a3 paralogs, atp1a3a and atp1a3b. Similar to dmist, atp1a3a is 294 widely expressed in the larval zebrafish brain ( Figure 5F, compare to Figure 2D). While atp1a3b 295 is also expressed in the zebrafish brain, its expression is more limited to regions of the midbrain 296 and hindbrain ( Figure S5C). To test whether these genes are involved in regulating zebrafish 297 sleep, we used CRISPR/Cas9 to isolate an allele of atp1a3a containing a 19 bp deletion and 298 an allele of atp1a3b containing a 14 bp deletion. Both mutations are predicted to generate null 299 alleles due to deletion of the start codon ( Figure 5G The similar night-time reduction in sleep in dmist and atp1a3a mutants, combined with the 319 similarities between Dmist and Fxyd1, suggested that Dmist may regulate the Na + /K + pump. 320 We therefore exposed wild type and mutant larvae to pentylenetetrazol (PTZ), a GABA-321 receptor antagonist that leads to globally heightened neuronal activity and elevated intracellular 322 sodium levels that must be renormalized by Na + /K + pump activity. Consistent with the 323 hypothesis that Dmist and Atp1a3a subunits are important for a fully functional Na + /K + pump, 324 brains from both dmist i8/i8 and atp1a3a 19/19 larvae had elevated intracellular sodium levels 325 after exposure to PTZ ( Figure 6A). Thus, neither dmist nor atp1a3a mutants were able to 326 restore intracellular sodium balance after sustained neuronal activity as quickly as wild type 327 siblings. Consistent with the night-specific alterations in sleep behaviour, we also found that 328 baseline brain Na + levels in dmist mutants were significantly elevated at night but not during 329 the day ( Figure 6B). Collectively, these data are consistent with the hypothesis that night-time 330 sleep duration is affected by changes in Na + /K + pump function and that Dmist is required to 331 maintain this function both at night and after sustained high levels of neuronal activity. 332 We have previously shown in zebrafish that a brief exposure to hyperactivity-inducing drugs 333 such as the epileptogenic PTZ or wake-promoting caffeine induces a dose-dependent increase 334 in homeostatic rebound sleep following drug washout that is phenotypically and mechanistically 335 similar to rebound sleep following physical sleep deprivation (Reichert et al., 2019). Based on 336 their exaggerated intracellular Na + levels following exposure to PTZ, we predicted that dmist 337 mutants would also have increased rebound sleep in response to heightened neuronal activity. phenotype, but if Dmist and Atp1a3a act together in the same complex/pathway, the mutant 349 phenotypes should be non-additive. Indeed, dmist -/-; atp1a3a -/mutants have a sleep reduction 350 similar to that of atp1a3a -/mutants alone, consistent with a non-additive effect ( Figure 6F and 351 S6). Similar non-additivity can be also observed in the dmist -/-; atp1a3a +/animals, which, like 352 atp1a3a +/animals alone, have a milder sleep reduction, indicating that the lack of additivity 353 between dmist and atp1a3a phenotypes is unlikely due to a floor effect, since double 354 homozygous mutants can sleep even less ( Figure 6F). This genetic interaction data is 355 consistent with our hypothesis that Atp1a3a and Dmist act in the same pathway-the Na + /K + 356 pump--to influence sleep. 357

DISCUSSION 359
Genetic screening discovers dmist, a novel sleep-regulatory gene 360 Using a reverse genetic viral screening strategy, we discovered a short-sleeping mutant, 361 dmist, which has a disruption in a previously uncharacterized gene encoding a small 362 transmembrane peptide. Given that the dmist mutant appeared within the limited number of 26 363 lines that we screened, it is likely that many other sleep genes are still waiting to be discovered 364 in future screens. In zebrafish, one promising screening strategy will be to employ 365 CRISPR/Cas9 genome editing to systematically target candidate genes. Advances in the 366 efficiency of this technology now makes it feasible to perform a CRISPR "F0 screen" in which The similarity between Dmist and FXYD1 led us to directly manipulate the Na + ,K + ATPase 413 to test its importance in sleep. The Na + ,K + -ATPase is the major regulator of intracellular Na + in 414 all cells and, by actively exchanging two imported K + ions for three exported Na + ions, is 415 essential for determining cellular resting membrane potential (reviewed in Clausen et al., 416 2017). The Na + ,K + -ATPase consists of a catalytic alpha subunit (4 known isoforms, ATP1A1-417 4), a supporting beta subunit (3 isoforms, ATP1B1-3), and a regulatory gamma subunit (the 418 FXYD proteins). The alpha1 and alpha3 subunits are the predominant catalytic subunits in 419 neurons (alpha2 is mostly restricted to glia), although the alpha1 subunit is also used 420 ubiquitously in all tissues (McGrail et al., 1991). By mutating zebrafish orthologs of Atp1a3, we 421 therefore could test the neuronal-specific role of the Na + ,K + -ATPase in sleep. 422 Mutations in both zebrafish Atp1a3 orthologs increased waking locomotor behaviour during 423 the day. However, only mutations in atp1a3a, which is expressed brain-wide, but not in 424 atp1a3b, which is expressed in more restricted brain regions, led to changes in night-time 425 sleep. The atp1a3a mutants have a larger sleep reduction than dmist vir , dmist i8 , or fxyd1 28 426 mutants, which is expected since loss of a pump subunit should have a larger effect than the 427 loss of a modulatory subunit, as has been shown for other ion channels ( Together, the night-specific sleep phenotypes of dmist, fxyd1, and atp1a3a mutants point to 443 a role for the Na + ,K + -ATPase in boosting sleep at night. How might the alpha3 catalytic subunit 444 of the Na + /K + pump regulate sleep, and how could Dmist be involved? We found that Dmist is 445 required for proper maintenance of brain intracellular Na + levels at night but not during the day, 446 mirroring the timing of sleep disruption in dmist i8/i8 animals. This suggests that the decreased 447 night-time sleep of dmist mutants is due to a specific requirement for Dmist modulation of the 448 Na + /K + pump at night. However, we cannot exclude the possibility that Dmist's function is 449 required in only a subset of critical sleep/wake regulatory neurons during the day that then exchanges Na + ions for K + , the high intracellular Na + levels we observe in atp1a3a and dmist 467 mutants is likely accompanied by high extracellular K + . Although we can only speculate at this 468 time, a model in which extracellular ions that accumulate during wakefulness and then directly 469 signal onto sleep-regulatory neurons could provide a direct link between Na + ,K + ATPase 470 activity, neuronal firing, and sleep homeostasis. Such a model could also explain why 471 disruption of fxyd1 in non-neuronal cells also leads to a reduction in night-time sleep. 472 In addition to decreased night-time sleep, we also observed that dmist mutants have an 473 exaggerated sleep rebound response following the high, widespread neuronal activity induced 474 by the GABA-receptor antagonist, PTZ. Since both Atp1a3a and Dmist were essential for re-475 establishing proper brain intracellular Na + levels following PTZ exposure ( Figure 6A), we 476 speculate that the exaggerated sleep rebound is a consequence of increased neuronal 477 depolarization due to defective Na + pump activity. This is consistent with our previous 478 observations that the intensity of brain-wide neuronal activity impacts the magnitude of increased sleep rebound in response to exaggerated neuronal activity during the day? One 482 possibility is that Na + /K + pump complexes made up of different alpha and beta subunits may 483 be differentially required for maintaining Na + homeostasis under physiological conditions and 484 have different affinities for (or regulation by) Dmist. For example, the Atp1a1 subunit is 485 considered the Na + /K + pump workhorse in neurons, while Atp1a3, which has a lower affinity 486 for Na + ions, plays an essential role in repolarizing neurons when Na + rapidly increases during In conclusion, through a genetic screening strategy in zebrafish, we have identified a novel 499 brain expressed gene that encodes a small transmembrane protein regulator of night-time 500 sleep and wake behaviours. Future work will be required to uncover the precise signalling 501 dynamics by which Dmist regulates the Na + ,K + -ATPase and sleep. 502

503
Acknowledgements 504 The initial screen, discovery, and characterization of dreammist was conducted in the lab of 505 Alexander F Schier at Harvard University. We also would like to thank members of the Rihel 506 lab and other UCL zebrafish groups for helpful comments on experiments and the manuscript.  The dmist vir allele was generated in wild type line T/AB-5 (Varshney et al., 2013) and 845 outcrossed to Harvard AB. The dmist i8 , fxyd1 Δ28 , atp1a3a Δ19 , and atp1a3b Δ14 alleles were 846 generated and maintained at UCL on an AB/TL background. Both dmist i8 and dmist vir were out-847 crossed to the AB strain at UCL for at least 3 generations.  Circadian period for every larva was calculated using the Matlab findpeaks function on the 878 activity (delta-pixels) timeseries data with a minimum peak distance of 18 hours (1080 879 minutes). N-way ANOVA was calculated to evaluate differences between groups. 880 Code and data are available at https://github.com/ilbarlow/Dmist. 881

Adult behavioural tracking 882
Fish from a dmist i8/+ x dmist i8/+ cross were raised in a mixed gender tank to adulthood. 883 Zebrafish adults (aged 3-4 months) were randomly selected and tracked on a 14:10 light:dark 884 cycle (180 lux at water surface, lit from above) for three days as in (Chiu et al., 2016). In brief, 885 fish were placed into uncovered plastic chambers (7x12x8.5 cm; WxLxH) with small holes for 886 water exchange, and these were placed in a circulating water tank (46x54 cm with 4.5 cm water 887 height). This setup was supplied with fish water from the home aquarium heated to 28ºC and 888 were performed blind to genotype, which was determined by fin-clip after the experiment. 896 Females and males were originally analysed separately; since no significant gender effect was 897 found (two-way ANOVA, genotypeXgender), data from both genders were pooled for the final 898

analysis. 899
Genotyping 900 Prior to genotyping, adult fish were anaesthetised in 30 μg/ml MS-222, fin-clipped by cutting 901 a small section of the caudal fin, and then allowed to recover in fresh fish water. The dmist vir genotype was detected by PCR (standard conditions) using a cocktail of three 909 primers (0.36 mM final concentration each primer) to detect the wild type allele and viral 910 insertion (see Table 2) so that genotypes could be assigned according to size of bands 911 detected (dmist vir/vir 800 bp; dmist vir/+ 508 bp and 800bp; dmist +/+ 508 bp). 912 The dmist i8 genotype was assigned by KASP genotyping using allele-specific primers 913 The atp1a3b Δ14 genotype was assigned by PCR using MiSeq_atp1a3b primers (Table 2), 921 with the atp1a3b Δ14 allele running 14 bp faster than the atp1a3b + allele. 922 fxyd1 28 was assigned by KASP genotyping using allele-specific primers (fxyd1 28 allele 5'-923 or PCR using MiSeq_fxyd1 primers (see Table 2) followed by digestion with the restriction 927 enzyme DrdI, which yields bands at 138 bp and 133 bp for fxyd1 +/+ ; 138 bp, 133 bp and 271 928 bp for fxyd1 +/28 , and 243 bp for fxyd1 28 . 929

In situ hybridisation 936
Probes were designed to target the 3'UTR and entire open reading frame (ORF) of 937 dmist_Dr transcript using primers that amplified the target region from zebrafish cDNA under 938 standard PCR conditions (expected size 1325 bp; Table 2). The PCR product was cloned bleached for 30 min in the dark (0.05% formamide, 0.5X SSC, 6% H2O2) and then fixed in 4% 952 PFA for 30 min at room temperature. To image, the embryos were progressively rehydrated 953 into 0.1% PBTw, progressively sunk in to 80% glycerol, and imaged on a Nikon compound 954 microscope (Nikon Eclipse Ni, Leica MC190HD camera). 955

Sodium Green Assay 970
Cell permanent Sodium Green tetraacetate (Invitrogen, S6901) was prepared fresh from 971 frozen stock by dissolving in DMSO to 1 mM then diluting in fish water to a final concentration 972 of 10 µM. About 50 larvae (5-7 dpf) from atp1a3a 19/+ or dmist i8/+ in-crosses were placed in 973 wells of a 6 well plate, then most fish water was removed and replaced with 3 mL of the 10 µM 974 Sodium Green solution for two hours. During exposure, the plate was covered in foil and placed 975 in a 28°C incubator. For PTZ experiments, larvae were also exposed to 10 mM PTZ (diluted 976     Sleep (min/10 min) Waking activity (s/min)