NCS1 overexpression restored mitochondrial activity and behavioral alterations in a zebrafish model of Wolfram syndrome

Wolfram syndrome (WS) is a rare neurodegenerative disease resulting in deafness, optic atrophy, diabetes, and neurological disorders. Currently, no treatment is available for patients. The mutated gene, WFS1, encodes an endoplasmic reticulum (ER) protein, Wolframin. We previously reported that Wolframin regulated the ER-mitochondria Ca2+ transfer and mitochondrial activity by protecting NCS1 from degradation in patients’ fibroblasts. We relied on a zebrafish model of WS, the wfs1abKO line, to analyze the functional and behavioral impact of NCS1 overexpression as a novel therapeutic strategy. The wfs1abKO line showed an increased locomotion in the visual motor and touch-escape responses. The absence of wfs1 did not impair the cellular unfolded protein response, in basal or tunicamycin-induced ER stress conditions. In contrast, metabolic analysis showed an increase in mitochondrial respiration in wfs1abKO larvae. Interestingly, overexpression of NCS1 using mRNA injection restored the alteration of mitochondrial respiration and hyperlocomotion. Taken together, these data validated the wfs1abKO zebrafish line as a pertinent experimental model of WS and confirmed the therapeutic potential of NCS1. The wfs1abKO line therefore appeared as an efficient model to identify novel therapeutic strategies, such as gene or pharmacological therapies targeting NCS1 that will correct or block WS symptoms.


INTRODUCTION
Rare diseases are defined, according to the European Union, as lifethreatening or chronically debilitating diseases which are of such low prevalence, less than 1/2,000 individuals, that they require special combined efforts to be addressed. Most of the time, and because of the limited market these diseases represent, no cure has been developed. Some treatments that only alleviate the symptoms are available to the affected individuals. Among this group of diseases, Wolfram syndrome (WS) is an autosomal recessive genetic disorder characterized by a diabetes mellitus, diabetes insipidus, optic nerve atrophy, hearing loss, and neurodegeneration. 1 The progression of the disease is severe and leads, ultimately, to the premature death of the affected individuals at around 35 years of age with severe neurological disabilities, including bulbar dysfunction and organic brain syndrome. 2 As of today, only symptomatic or substitutive therapies are available to the patients. Therefore, there is an urgent need for a cure that would stop the progression of the pathology.
More than 200 variants in the WFS1 gene are responsible for WS type 1, which represents 99% of WS cases, while <1% are due to variants in CISD2 gene. [3][4][5] WFS1 encodes a putative 890-amino acid protein, Wolframin, localized into the endoplasmic reticulum (ER) membrane. 6,7 Fonseca et al. have shown in rodent and human cells that WFS1 has a critical function in the regulation of ER stress signaling and prevents secretory cells, such as pancreatic b cells, from dysfunction and premature death caused by hyperactivation of ER stress signaling through its interaction with the transcription factor ATF6a. 8 Since this initial study, accumulating data have established the regulative role of WFS1 on ER stress signaling. 9 One other important function of WFS1 is its role on regulation of mitochondrial calcium transfer through the mitochondria-associated ER membranes (MAMs). These MAMs are involved in a plethora of cellular functions, such as Ca 2+ homeostasis, ER unfolded protein response (UPR), autophagy, and apoptosis. 10 Interestingly, under physiological conditions, WFS1 interacts with the Ca 2+ sensor NCS1 in the MAMs and prevents its degradation by forming a complex with the inositol-1,4,5 triphosphate receptor (IP 3 R) to activate ER-mitochondria Ca 2+ transfer. 11 The functionality of the tripartite complex WFS1/NCS1/IP 3 R ensures a proper Ca 2+ transfer from the ER to the mitochondria and the activation of mitochondrial oxidative respiration. In the WS context, we have shown that the decrease of WFS1 in patient fibroblasts is associated with NCS1 downregulation, leading to a decrease in ERmitochondria Ca 2+ transfer, triggering in fine cell death. 11 NCS1 is the mammalian ortholog of Frequenin in Drosophila, a protein encoded by the FREQ gene in humans. 12 NCS1 is one of the EF hand Ca 2+ sensors that has a high affinity for cellular calcium and thereby functions as a potent transducer of intracellular calcium fluctuations. 13 In mammalian cells, NCS1 promotes exocytosis from dense core vesicles in both neurons and neuroendocrine cells, 14 regulates synaptic transmission, 15 and promotes control of the growth dynamics of nerve endings. [15][16][17] NCS1 also regulates associative learning and memory in C. elegans, 18 and promotes exploration, synaptic plasticity, and rapid acquisition of spatial memory in mice. 19 Very interestingly, NCS1 is highly expressed in retinal ganglion cells, which are highly affected in WS patients. 20 In addition, as highlighted by recent studies, NCS1 plays an essential functional role as a calcium sensor, in pancreatic b cells, key cells in a context of diabetes mellitus. 21,22 In a previous study, we have shown that not only NCS1 knockdown in control fibroblasts recapitulated the metabolic deficits observed in fibroblasts from patients with WS but also that its overexpression in the latter rescued the defective mitochondrial phenotype. 11 A study corroborated our results in different species and cell types, namely rat insulinoma, where Wfs1 has been knocked out. The overexpression of NCS1 in these cells restored calcium homeostasis deficit and protein kinase B/Akt signaling and, subsequently, cell viability and glucose-stimulated insulin secretion. 22 Considered together, these information conferred to NCS1 a unique therapeutic potential in WS pathology.
Zebrafish (Danio rerio) models have proven efficient as animal model of human diseases to find new treatments, particularly in rare genetic diseases. 23 They represent a unique combination of different characteristics, such as short generation time, transparency, small size, and sequenced genome close to the human one, that confer great advantages to this model in therapeutic research.
As most of the genes in the zebrafish genome, wfs1 is duplicated, both copies wfs1a and wfs1b sharing, respectively, 52.5% and 52.8% similarities with human WFS1. In a previous study, we described two wfs1 mutant zebrafish lines, specific to each gene, wfs1a C825X and wfs1b W493X . These lines mimicked some of the symptoms of WS pathology, including visual, ER stress, and mitochondrial alterations. 24 To fully abolish any Wfs1 functional protein, we intercrossed the two lines mentioned above to create the double-mutant line, wfs1a C825X Â wfs1b W493X , termed wfs1ab KO hereafter. The characterization of this line revealed alterations relevant for the study of WS and allowed us to explore the potential therapeutic impact of NCS1 in a WS pathophysiological context.

Characterization of wfs1ab mutant zebrafish lines
To validate the double-mutant strain, relative wfs1a and wfs1b mRNA levels were measured by quantitative PCR (qPCR) in the wfs1ab KO mutant line ( Figure S1A). Both genes were significantly downregulated in the double-mutant line, each mutation leading to some mRNA decay, as described previously for each single-mutant line. 24 The protein expression level of wfs1 could not be assessed, as no commercial antibody specifically labeling the zebrafish protein is available. However, due to the premature stop codons, we predicted that both proteins, when expressed, were not functional.
We assessed the impact of the absence of both Wfs1 proteins on zebrafish larvae morphology using different parameters (detailed in Figure S1B). The fish were measured at 5 dpf using a binocular loupe and analyzed using the Fiji software ( Figures S1C-S1H). Overall, the mutations did not impact the grossly observable morphology of the larvae, with the exception of an increased ear area ( Figure S1F) in wfs1ab KO fish compared with the controls.

Downregulation of Wfs1 impact visual and locomotor functions of wfs1ab KO zebrafish
The visual motor response (VMR) assay measured the locomotor response of the larvae induced after visual stimulation. Experimentally, larvae at 5 dpf were placed individually in each well of a 96-well plate and their individual activity was measured by videotracking. A training phase of 30 min in the dark (light OFF) allowed the larvae to acclimate to their new environment, followed by two 10-min duration light ON (100% light) and light OFF phases, with a 10-min intertrial time interval. The locomotion profiles are shown in Figure 1A. The locomotion of the wfs1ab KO larvae increased drastically compared with wfs1ab WT controls during the training phase ( Figure 1B), ON ( Figure 1C), and OFF periods ( Figure 1D), highlighting a hyperlocomotor response in wfs1ab KO larvae. The hyperlocomotion is still present at 7 dpf but disappeared at 9 dpf ( Figures S3A-S3H).
The visual acuity of the mutant zebrafish was measured using the optokinetic response (OKR). Movements of larva eyes were provoked by rotating black-and-white strips retroprojected on a fixed cylinder and videotracked ( Figure 1E). The number of eye saccades within 2-min time intervals, reflecting the visual acuity of the immobilized 5-dpf larva, was not altered for wfs1ab KO larvae compared with wfs1ab WT controls ( Figure 1F).
The reflex motor response was also measured, using the touch-evoked escape behavior ( Figure 1G). After stimulus, the wfs1ab KO larvae showed a significant increase of the traveled distance ( Figure 1H), suggesting a motor alteration that could be related to the increased locomotor response previously observed in the VMR assay.
We also assessed the acoustic startle response (ASR) induced by a repetitive 1-s duration noise ( Figure S2) and measured using the quantity of movement ( Figure S2A). The quantity of movement during both the training ( Figure S2B) and baseline ( Figure S2C) phases was increased for wfs1ab KO larvae compared with wfs1ab WT controls. The wfs1 mutations had, however, no impact on the quantity of movement after stimulation ( Figure S2D).
Absence of functional Wfs1 disrupted eye development of zebrafish larvae As the larvae exhibited visual alterations in the VMR assay, an immunohistochemical analysis of the developing eye was performed at 5 dpf in the wfs1ab KO mutant line ( Figure 2). Quantification of cell numbers from immunostained cryosections of the whole retina showed a nonsignificant decrease of the density of retinal ganglion cells in wfs1ab KO compared with wfs1ab WT controls (Figures 2A and 2B), without alteration of the thickness of the ganglion cells layer ( Figure 2C). The density of red and green cones, labeled with Zpr-1 antibody ( Figures 2D  and 2E), as well as the density of rods, labeled with Rho4d2 antibody ( Figures 2F and 2G), were decreased in wfs1ab KO larvae. The level of glucose did not change in wfs1ab KO larvae ( Figure 2I).

ER stress response in wfs1ab KO zebrafish
The expression level of genes involved in the regulation cascade of the ER stress was analyzed by qPCR in the wfs1ab KO mutant zebrafish at 5 dpf ( Figure 3A). No alteration of the expression level of UPR pathway inducers (sigmar1, bip, hsp90b1), primary effectors of the UPR pathway (ire1, atf6, xbp1s, xbp1us) and secondary effectors (perk, eif2s1, atf4a, atf4b, chop) was noted, with the notable exception of eif2s1, slightly but significantly decreased in wfs1ab KO larvae compared with wfs1ab WT controls ( Figure 3A). Western blot analyses were carried out, only with a handful of antibodies showing a specific labeling in zebrafish tissues. We observed that, out of all the ER stress-related proteins tested, only Chop was significantly downregulated in wfs1ab KO larvae compared with wfs1ab WT controls ( Figures 3B  and 3C).
Larvae were treated with tunicamycin, 2 mg/mL for 24 h, to analyze the UPR in condition of ER stress ( Figures 3D and 3E). The mRNA level of all genes was significantly increased in the wfs1ab KO larvae and wfs1ab WT controls treated after tunicamycin, with the notable exception of eif2s1 in control larvae. The genotype had a mild impact on the gene's response to tunicamycin. Induction of hsp90b1 and, to a lesser extent, perk tended to show a milder increase while eif2s1 expression was significantly increased in ER stress conditions in wfs1ab KO larvae ( Figures 3D and 3E). These data suggested that Wfs1 inactivation in zebrafish had only a mild impact, limited to some markers of the www.moleculartherapy.org PERK-dependent pathway, while IRE1-and ATF6-dependent pathways were unaffected.
Mitochondrial functions are altered in wfs1ab KO zebrafish Mitochondrial oxidative respiration rates were measured in vivo in 5-dpf larvae using the Seahorse XFe24 analyzer to determine the consequences of Wfs1 invalidation on mitochondrial functionality ( Figure 4). The oxygen consumption rate (OCR) ( Figure 4A) was measured before and after the addition of: (1) the ATP synthase inhibitor oligomycin, to measure the ATP production-related OCR; (2) the uncoupling agent FCCP, to measure the maximal OCR; and (3) antimycin A and rotenone, to measure proton leak and non-mitochondrial OCR ( Figure 4A). The wfs1ab KO line showed a significant increase in basal respiration ( Figure 4B), ATP production-related OCR ( Figure 4C), and maximal respiration ( Figure 4D) compared with control larvae. No difference was found in proton leak ( Figure 4E) or non-mitochondrial respiration ( Figure 4F).   Figure S4. (D) Larvae were exposed to vehicle DMSO solution (Veh) or tunicamycin 2 mg/mL (Tunica) for 24 h. zeif2a and stain-free were used as a loading control in qPCR and western blot analyses, respectively. The relative expression levels of mRNA after tunicamycin treatment are indicated as the fold change from the respective wfs1ab WT /DMSO or wfs1ab KO /DMSO control group, when higher than 2-fold. Data are expressed as mean ± SEM, n = 5-9 in each group. *p < 0.05 versus wfs1ab WT in (A and B). *p < 0.05, **p < 0.01, ***p < 0.001 versus DMSO treatment in (A and B) and Tukey's multiple comparison test in (D).
Overexpression of Ncs1 restored the locomotion and mitochondrial alterations of the wfs1ab KO zebrafish We previously reported that, in WS patient fibroblasts, the overexpression of NCS1 was able to restore all the key physiological deficits responsible of the severity of the disease. 11 Therefore, we examined the impact of NCS1 overexpression in wfs1ab WT and wfs1ab KO zebrafish lines. mCherry or Ncs1 RNAs were injected into the eggs and Ncs1 expression was assessed by western blot analysis at 5 dpf. A significant 2-fold overexpression of the protein level was measured in both lines (Figures 5A and 5B). Once the overexpression was validated, we analyzed the motor behavior functions of the Ncs1-or . Analysis of mitochondrial respiration in 5 dpf wfs1 KO zebrafish larvae (A) Oxygen consumption rate (OCR) profiles of wfs1ab WT and wfs1ab KO zebrafish larvae at 5 dpf during the assay and calculated, (B) basal respiration rate, (C) ATP production-related OCR, (D) maximal respiration, (E) proton leak, and (F) non-mitochondrial respiration. Data are mean from triplicates and show mean ± SEM from n = 28 animals per genotype. Olig, oligomycin; FCCP, cyanide-p-trifluoromethoxyphenylhydrazone; Rot/AA, rotenone + antimycin A. *p < 0.05, **p < 0.01 versus wfs1ab WT , unpaired t test.
mcherry RNA-injected lines using the VMR assay ( Figure 5C). The locomotion profiles were measured ( Figures 5D-5F). No significant difference in the distance traveled during the light alternation sequences was measured during the training or ON phases ( Figures 5D and 5E), but mCherry RNA-injected wfs1ab KO larvae showed a significantly higher response during the OFF phase compared with wfs1 WT larvae, recapitulating the phenotype observed in non-injected fish. On the other hand, Ncs1 overexpression completely prevented the hyperlocomotion observed in wfs1ab KO larvae ( Figure 5F).
We also measured the impact of Ncs1 overexpression on mitochondrial function. The OCR profiles were analyzed using the Seahorse analyzer as described previously ( Figure 6A). The mcherry RNA-injected wfs1ab KO larvae exhibited statistically significant increases of basal respiration ( Figure 6B), maximal respiration ( Figure 6D), and a trend for ATP production-related OCR (p = 0.237; Figure 6C), compared with controls, confirming the previously observed effects (Figure 4). No significant impact of the absence of Wfs1 on proton leak and non-mitochondrial OCR were measured ( Figures 6E and 6F).
The injection of Ncs1 non-significantly prevented the increase in basal respiration ( Figure 6B) and significantly prevented the increase in maximal respiration (Figure 6D) down to the levels observed in the control wfs1ab WT line. These observations confirmed in vivo the essential functional role played by Ncs1 in a WS pathophysiological context.

DISCUSSION
The aim of the present study was to characterize a zebrafish model of WS, with non-functional Wfs1a nor Wfs1b proteins, and to challenge the hypothesis that an overexpression of Ncs1 in vivo would restore most of the observed behavioral and cellular alterations shown by the mutant line. Contrarily to single mutation models published previously (wfs1a L692X , 25 wfs1a C825X , 24 and wfs1b W493X24,25 ), the complete abolition of Wfs1 functional protein in the wfs1ab KO line was expected to avoid any functional redundancy and compensatory mechanisms and to mimic as closely as possible the human pathology and underlying molecular pathophysiological mechanisms.
Based on our previous study, 24 we intercrossed the wfs1a C825X and wfs1b W493X mutant lines to generate a double-mutant zebrafish line. To assess the role of Wfs1 protein in zebrafish, we initially selected two lines from the ZFIN repository, which were generated through ENU mutagenesis. 26 Both lines carry a point mutation coding for a premature stop codon, potentially leading to non-functional proteins. Moreover, the chosen variants, wfs1a C825X and wfs1b W493X , should mimic human mutations E752X or Q819X for Wfs1a and Q486X for Wfs1b. 27,28 WS is a highly heterogeneous disease in its clinical presentation as well as its genetic causes. Even though only one gene is responsible for WS type 1, more than 200 variants have been reported. 27 So far, studies have failed to establish a clear phenotype-genotype correlation, most likely due to the molecular complexity of WFS1 gene, the inter-and intra-familial divergences of the clinical characteristics, and the rarity of the disease, leading to small size cohorts. 27,29,30 The type and/or location of the variant cannot predict the associated phenotype; however, the variants have been classified according to their gene and protein alterations. 27 Based on this classification, we can hypothesize that wfs1a C825X and wfs1b W493X might lead to a complete degradation of the Wfs1 proteins.
The absence of Wfs1 protein had no effect on the gross morphology of the zebrafish 5-dpf larva, except for an increased size of the ear area compared with wfs1ab WT control. This anomaly was previously observed in the wfs1a C825X larva. 24 We could not, however, exclude some transient developmental deficits in early stages, as suggested by the study of Cairns et al. in single-mutant zebrafish at age up to 80 hpf. 25 As for both single-gene mutant larvae, no difference in auditory function during the noise stimulation was observed using the ASR test, suggesting that the morphological anomaly measured in both lines was not correlated to a functional alteration. 24 In the retina, in contrast to what we previously reported for both mutant lines wfs1a C825X and wfs1b W493X , 24 the number of cones and rods decreased in wfs1ab KO larvae at 5 dpf, suggesting that the photoreceptors were altered. Coherently, a visual alteration was estimated in the VMR test. Cairns et al. reported that in older fish, 4 and 12 months old, both wfs1a and wfs1b mutant fish exhibited a decrease of the retinal ganglion cell density, associated with thinner retinal ganglion cell layer and in ganglion cell layer. 25 Taken together, these data suggested that the double-mutant line might undergo accelerated visual degeneration process.
Patients with WS develop both neurodevelopmental and neurodegenerative phenotypes. While this study is focused on Wolfram syndrome zebrafish model larvae and the associated neurodevelopmental defects, we can speculate that older fish may present with neurodegeneration. The study carried out by Cairns et al. in other WS zebrafish models highlighted the evolution of the pathophysiology of the disease with the progression of the visual deficit in 4-and 12-month-old fish at both physiological (wfs1b W493X ) and cellular levels (wfs1a L692X and wfs1b W493X ). 25 In a 15-month-old WS rat model, in addition to axonal degeneration and disorganization of the myelin in the optic nerves leading to an optic atrophy, a decrease of medullary volume of the brain was reported, showing similarities with the brainstem neurodegeneration observed in WS patients. 31 These two independent studies support our hypothesis, and further investigations will be necessary to conclude if our zebrafish model, in addition to neurodevelopmental defects, presents neurodegeneration. However, we need to consider that zebrafish is also a useful model to study regeneration. Indeed, zebrafish has demonstrated good ability to regenerate axons, 32 retina, 33 and pancreas 34 (see also Massoz et al. 35 ). In fact, in our model, the disappearance of hyperactivity in 9-dpf larvae ( Figure S3) may be explained either by a regenerative process of neurons, or by the fact that the hyperactivity phenotype is due to a developmental delay that was corrected at 9 dpf, as it has been shown in another model by Del Pozo et al. 36 Consequently, the potential regeneration might be a limitation to the study of neurodegeneration.
WFS1 was described as a modulator of UPR in different models. 8,[37][38][39][40] In the mutant wfs1a line, we previously observed significant alterations in the PERK pathway and the mutant wfs1b line showed altered IRE1 and PERK pathways under basal conditions and under ER stress conditions induced by tunicamycin treatment. 24 Therefore, we measured the impact of the abrogation of wfs1 expression in 5-dpf wfs1ab KO larvae on the different UPR markers. We observed that, under basal conditions, wfs1 genes loss of function only impacted the expression level of eif2s1 and chop. Under ER stress conditions, marginal effects were observed on markers of the PERK pathway, but globally it appeared that the mutations have little impact on ER stress. This questioned the precise role of WFS1 in this critical cellular process in the pathology of WS. The data suggested that the WFS1 loss of function consequence is not an altered activation of the UPR but rather an impairment of the mitochondrial function we demonstrated in human patient fibroblasts, 11 and in Wfs1 DExon8 mutant mice. 41 WFS1 is indeed also involved in the regulation of the mitochondrial activity. We previously demonstrated that WFS1 deficiency in human fibroblasts was associated with decreased Ca 2+ uptake by mitochon-dria, reduced mitochondrial contact with the ER, and decreased mitochondrial respiration. 11 Therefore, we analyzed the impact of the total loss of function of wfs1 genes on the different parameters of mitochondrial physiology in the living zebrafish at 5 dpf. wfs1ab KO mutant larvae showed an increase in basal, ATP-linked, and maximal respirations, but no difference in proton leak and non-mitochondrial respiration compared with controls. The double-mutant line therefore appeared more altered than single-gene mutants that showed no deficit, in wfs1a C825X larvae, or transient decreases of basal, ATP-linked, and maximal respirations, in wfs1b W493X at 2 dpf but not 5 dpf. 24 This could be explained by direct compensatory mechanisms between wfs1a and wfs1b genes in single-mutant models. In the double mutant, neither Wfs1a nor Wfs1b are functional and no compensation or altered expression would interfere. Intriguingly, the results in wfs1ab KO mutant zebrafish did not mimic those obtained in WS patient fibroblasts. Indeed, in fibroblasts, a decrease in the mitochondrial activity was observed, whereas an increase of the mitochondrial activity is observed in the wfs1ab KO zebrafish. 11 One possible explanation is that, in vitro, experimental conditions, including glucose concentration or neurotransmitter levels, are controlled, which is not the case in vivo. Increased glucose concentration leads to increased OCR in cell. 42 No differences in basal glucose level were observed in wfs1ab KO mutant zebrafish; however, additional analysis, such as glucose tolerance test and glucose level at later time points, may help completely rule out this hypothesis. Another hypothesis is that the dopamine concentration is altered in the wfs1ab KO mutant, mimicking what was observed in Wfs1-deficient mice following high K + challenge. 43 Since it has been shown that an L-DOPA treatment inhibited complex I and ATP levels in the striatum, 44 and dopamine treatment in neuronal cell line induced a reduction in cellular ATP levels, 45 the deficit in neurotransmitter would result in vivo in wfs1ab KO mutant zebrafish, in an increase in OCR. Further experiments would be needed to check this hypothesis, such as measuring the levels in dopamine and its metabolites in 5-dpf zebrafish larvae, which remains a tricky experiment.
As mentioned above, ER stress plays a key role in WS physiopathology leading to therapeutic development targeting mainly ER stress modulation and regulation. This encompasses chemical chaperone, ER calcium stabilizers, and mitochondrial modulator. 46 However, we and others broaden the spectrum of potential underlying mechanisms, including now calcium dysregulation, as elegantly highlighted by two recent reviews. 47,48 Our choice of NCS1 as a relevant target falls within the scope of this new strategy as we expect NCS1 overexpression to restore the associated Ca 2+ pathways. In addition, deciphering WFS1-dependent calcium mechanisms may provide additional therapeutic targets. NCS1 is a calcium sensor expressed predominantly in neurons that regulates many cellular functions, including neurotransmission, synaptic plasticity, exocytosis and endocytosis, neuronal growth, neuroprotection, and nuclear Ca 2+ regulation. Interestingly, NCS1 exerts similar functions as WFS1: both proteins regulate cytosolic Ca 2+ levels and IP3R-dependent ER-Ca 2+ release 20,32-34 modulate neuronal morphology and neurodevelopment, 49,50 mediate neuroprotection, 51 and have been implicated www.moleculartherapy.org in neurodegenerative diseases and psychiatric disorders beyond WS. [52][53][54] In human fibroblasts, NCS1 knockdown impaired Ca 2+ homeostasis and mitochondrial function, demonstrating that NCS1 modulated [Ca 2+ ] m uptake and could be a target to maintain mitochondrial function and MAM integrity. In addition, the overexpression of NCS1 in patient fibroblasts restored the defective mitochondrial phenotype of WFS1-deficient cells. 11 Moreover, in rat insulinoma cells, the overexpression of NCS1 in wfs1 KO cells was also shown to restore calcium homeostasis including ATP-evoked ER calcium release and resting cytosolic calcium. 22 Therefore, we investigated if the overexpression of Ncs1 would bypass Wfs1 activation in our WS zebrafish model. Like the wfs1 gene, there are two copies of ncs1 in zebrafish: ncs1a and ncs1b. We decided to inject the murine Ncs1 gene, using a sequence that has been optimized for its expression in the zebrafish. 55 Even though ncs1 is not downregulated in our wfs1ab KO mutant zebrafish, murine Ncs1 overexpression, using mRNA microinjection at the embryonic stage, seems to compensate the absence of Wfs1 proteins and counteracts its negative impacts. We cannot exclude, though, that Ncs1-mediated effects would still be through its interaction with Wfs1. Indeed, Ncs1 interacts, in vivo, with the N-terminal part (amino acids 1-311) of Wfs1. 11 As stop codons for both lines are located after this domain (C825X and W493X for wfs1a and wfs1b, respectively), our mutants may still express the N-terminal parts of both Wfs1 proteins.
Interestingly, the locomotor hyperactivity of 5-dpf wfs1ab KO larvae, measured using the VMR in the OFF phase, was reversed after overexpression of Ncs1. More specifically, the abrogation of Wfs1 impacts the mitochondrial activity of 5-dpf wfs1ab KO larvae with a significant increase in basal and maximal respiration compared with controls. These alterations were partially or totally reversed by the overexpression of Ncs1. Also, the trends for increased ATP production and nonmitochondrial respiration in wfs1ab KO larvae were attenuated by overexpression of Ncs1. We can conclude that the absence of functional Wfs1 led to a mitochondrial alteration and resulted in a locomotor change, i.e., hyperlocomotion. Notably, all these alterations were compensated by the overexpression of Ncs1. Restoration of the locomotor and visual behaviors in mutant larvae appeared directly correlated to the initial restoration of the mitochondrial alteration induced by the overexpression of Ncs1 observed by measuring the OCR.
To date no treatment of WS is available and the on-going clinical trials have not proven efficient yet. 56 Several models for WS have been developed, including IPSC-derived cells, 57 Drosophila, 37 zebrafish, 24,25 and rodents (mouse 41,58-60 and rat 31 ). None of these models present with all the characteristics of the human pathology and the use of one model over another is dictated by different parameters. Zebrafish is a powerful model to screen drugs in vivo. Its inexpensive husbandry and maintenance associated with its small size and high fecundity facilitate its use for large-scale high-throughput drug screening. However, some drawbacks need to be taken in account, such as anatomical and physiological divergences, duplica-tion of the genome that might prevent efficient gene therapy, as well as a limited panel of complex behavioral analysis. 23 This model still remains a model of choice for early-stage high-throughput drug screening, complementary to rodent models. With our double-mutant model, we overcome the limitation due to the duplication of the zebrafish genome, allowing for gene therapy strategy. Moreover, the VMR defect, characteristic of our fish, is a readout that can be easily and extensively used in the context of drug screening.
While the use of NCS1 is currently explored in other indications such as cancer and fragile X syndrome, 61-63 developing therapies with NCS1 as a target in the context of WS is new, and only cell-based studies have been performed so far. 11,22 In the context of rat insulinoma cells, deficient for Wfs1, ibudilast, a canonical PDE4 inhibitor and NCS1 binding drug, has been used as an efficient treatment to restore calcium homeostasis, cell viability, and glucose-stimulated insulin secretion. 22 The effective use of a drug targeting NCS1, such as ibudilast, strengthens our strategy to use NCS1 as a therapeutical target. Our study is the first one to use successfully NCS1 in a whole organism. Based on our results and already published studies, we can envision both gene therapy and a pharmaceutical approach with small molecules.
In conclusion, we have characterized a zebrafish model for WS, bearing a double mutation on wfs1a and wfs1b genes, which will be a pertinent model for drug screening. We have demonstrated in vivo the therapeutic potential of NCS1 overexpression as a first stone to pave the way to an efficient therapeutic strategy that could halt the progression of WS disease.

Zebrafish lines and husbandry
This study followed the recommendations of the ARRRIVE guidelines 64 and the European Union Directive 2010/63. The wfs1ab KO mutant line was generated by cross-breeding the two lines wfs1a C825X and wfs1b W493X , as described in Crouzier et al. 24 Adult zebrafish were bred and maintained under standard conditions in an automated fish tank system (ZebTEC, Tecniplast, Louviers, France) at 28 C, pH 7, conductivity around 500 mS, and with a 14:10 h light:dark cycle. Eggs were obtained by natural spawning and maintained in fish water at 28 C. Each experimental procedure was carried out in triplicate and larvae were from three different crosses.

Chemical treatment
To induce ER stress, larvae were incubated, at 4 dpf, for 24 h with 2 mg/mL of tunicamycin (sc-3606, Santa Cruz Biotechnology, Dallas, TX), and diluted in 0.1% dimethylsulfoxide (DMSO) directly in the fish water. Control larvae were treated with 0.1% DMSO diluted in fish water.

mRNA injection
Transcripts encoding mouse wild-type Ncs1 were generated after insertion of the murine Ncs1 coding sequence in the pcDNA3.1(+) plasmid (construct engineered by GenScript, Leiden, the Netherlands). The sequence was optimized for a better translation in the zebrafish (sequence available upon request). The Ncs1 mRNAs were transcribed from Not1-linearized pcDNA3.1(+) using the T7 polymerase with the mMESSAGE Machine kit (Ambion, Austin, TX) and purified accordingly to the manufacturer's instructions. Eight hundred nanograms ng of Ncs1 RNAs, in a 1 nL volume, was microinjected into one-cell-to two-cell-stage embryos according to standard protocols. mCherry RNAs from pCS2+-mCherry were used as an injection control (generous gift from Dr G. Lutfalla, LPHI, Montpellier, France).

RT-PCR and quantitative real-time PCR
At 5 dpf, total RNA from 20 whole wfs1ab KO mutant larvae and their associated controls were extracted using a Nucleospin RNA Kit (Macherey-Nagel, Hoerdt, France) according to the manufacturer's instructions. RNA concentration and purity were evaluated using the Agilent RNA 6000 Nano Kit (Agilent Technologies, Santa Clara, CA). RNA samples (1 mg) were denatured for 5 min at 70 C and reverse transcribed into cDNA for 1 h at 37 C, using M-MLV reverse transcriptase (Promega, Madison, WI). Primer sequences are detailed in the supplemental information in Table S1. Control reactions were conducted with sterile water to determine signal background and DNA contamination. The standard curve of each gene was confirmed to be in a linear range, while zef1a gene was selected as reference.

VMR assay
The locomotor activity of zebrafish larvae was quantified by VMR using an infrared (IR) tracking system (Zebralab, Viewpoint, Lissieux, France) as described previously. 65 In brief, at 5 dpf, larvae were transferred into a 96-well plate (Whatman, no. 7701-1651) with 300 mL fish water and the locomotor behavior was monitored with an automated videotracking device (Zebrabox, ViewPoint). The response to light changes was recorded by an IR camera under IR light illumination. The light protocol was as follows: 30 min of acclimatization in the dark (0% light intensity), then two cycles of 10-min duration light ON (100% light intensity) or light OFF (0%) periods. The total duration of the experiment was 70 min. Activity during the experiment was measured in mm/s. The values obtained during OFF periods were subtracted for each larva from their values registered during ON periods to remove inter-and intragroup variability in basic locomotion.

ASR assay
The locomotor activity of zebrafish larvae was quantified using ASR in the ZebraBox. Experimental conditions were similar to those used in the VMR assay. The experiment consisted, first, in acclimating larvae during 30 min with no sound (35 dB ambient), followed by a 1-s stimulation with white sound at 90 dB, repeated three times with an intertrial time interval of 5 min. The quantity of movements during the entire experiment was measured for each larva. Baseline activity levels were subtracted from the activity levels during the sound stimulations (2 min before each stimulation) to normalize the values.

OKR assay
At 5-dpf zebrafish larvae were immersed per group of four in a Petri dish (35 mm diameter) containing 2.5% methylcellulose (no. 9004-65-3, Sigma Aldrich, St. Louis, MI). Larvae were placed dorsal up and forming an X to avoid touching and interfering with each other. All measurements were done in the afternoon between 2:00 p.m. and 6:00 p.m. The room temperature was 28 C and the light was OFF. Visual system performance of larval zebrafish was assessed using a videotracking device (VisioBox, ViewPoint). Forty 6-mm-wide black and white strips were projected at 2 rpm for 1 min clockwise and then 1 min anti-clockwise. The IR illuminated larvae from the bottom and responses were tracked using an FL3-U3-32S2M 1/2.8-inch Monochrome camera (FLEA3, Flir) at 25 frames/s. The number of saccades was manually recorded (PHIVisualize software) and the average number of saccades per 2 min quantified.

Touch-escape response
To measure the touch-escape, 5-dpf larvae were transferred to a rail developed with a 3D printer (18 Â 0.4 cm) with 200 mL of fish water and placed at the extremity of the rail. The tail of larvae was touched with a tip and the distance traveled was measured during 5 s. The same procedure was repeated three times per larvae, each test separated by 1 min to reduce stress. Each larva was assessed individually and the three values per larva were averaged. The light was ON and the temperature was 28 C.

Cell counts
Cone cells, immunolabeled with Zpr-1 antibody, were quantified individually and the total area of rod outer segments, immunolabeled with Rho4d2 antibody, was evaluated and both measures were normalized to the length of the associated retina. Ganglion cells, highlighted by DAPI counterstaining, were counted in three specific www.moleculartherapy.org regions of the retina, consistent from one sample to another. The total number of ganglion cells from all regions was averaged per larva. The thickness of the ganglion cell layer was also quantified in the same specific regions of the retina.
Measurement of total free glucose levels At 5 dpf, 10 larvae were homogenized in 5 mL of NaCl 0.9% using a hand homogenizer, and centrifuged at 21,694 Â g for 2 min. The supernatant (1.5 mL) was placed on a glucometer strip (Accu-Chek Performa). Total free glucose levels were measured using a glucometer (Accu-Chek Performa Nano) and two measurements were averaged per sample.

Seahorse XFe24 MitoStress test
The OCR of 5-dpf larvae was measured with the Seahorse XFe24 extracellular flux analyzer (Agilent, Santa Clara, CA). Larvae were singly placed in wells of a Seahorse XFe24 spheroid microplate, containing 500 mL of fish water. A grid was placed manually on larvae to keep them at the bottom of the wells throughout the experiment. The blanks were two empty wells of the plate. Four basal cycles readings were recorded, then five recording cycles following oligomycin (25 mM) injection, five recordings cycles after carbonyl cyanide-p-trifluoromethoxyphenylhydrazone (FCCP) (8 mM) injection and nine recordings cycles after rotenone + antimycin A (1.5 mM) injection. Calculations for specific parameters (non-mitochondrial respiration, basal respiration, maximal respiration, proton leak, and ATP production) were made. The room temperature was controlled at 28 C. Measurements of total zebrafish OCR were started immediately and performed according to the manufacturer's instructions.

Statistical analyses
Data are expressed as mean ± SEM. Statistical significance between groups was determined by unpaired Student's t test or two-way ANOVA. The levels of statistical significance considered were: *p < 0.05, **p < 0.01, and ***p < 0.001. Statistical analyses were performed using the Prism v.7.0 software (GraphPad, San Diego, CA). Two-way ANOVA statistical values for Figure 5 (Table S2) and for Figure 6 (Table S3) are presented in the supplemental information.

DATA AND MATERIALS AVAILABILITY
All of the data and materials are available upon request.