Author Correction: 3-ketodihydrosphingosine reductase mutation induces steatosis and hepatic injury in zebrafish

An amendment to this paper has been published and can be accessed via a link at the top of the paper.

zebrafish kdsr I105R mutant that encodes a missense mutation in 3-ketodihydro-sphingosine reductase (kdsr) from a forward genetic screening to identify mutants with post-developmental liver disease 17 . Here, we use the kdsr I105R mutant to explore its role in the pathogenesis of hepatic injury. We found that accumulation of ceramides, Sph, and S1P resulted from activation of the lysosomal sphingolipid salvage pathway in the kdsr I105R mutant. Additionally, we found that oxidative stress by elevation of mitochondrial β-oxidation and ER stress in the kdsr I105R mutant can mediate mitochondrial cristae and liver injury. Through genetic interaction of kdsr and sphk2 mutations, we also found that sphk2-mediated S1P accumulation is a key factor in both oxidative and ER stress in the kdsr I105R mutant.

Results
kdsr I105R mutant zebrafish developed progressive liver injury and hepatic injury during post-developmental stage. From the previous forward genetic screening to identify zebrafish mutants with post-developmental liver defects 17 , we identified a mutant showing progression of liver defects ranging from hepatomegaly at 6 days post fertilization (dpf) to steatosis at 7 dpf, and to a more advanced hepatic injury thereafter ( Fig. 1A-C). We identified causative mutation by using whole genome sequencing of normal looking siblings and homozygous mutants (Supporting Fig. 1). The mutant carried a missense mutation in 3-ketodihydrosphingosine reductase (kdsr). The protein homology comparison with human KDSR showed that the zebrafish kdsr had high homology with human KDSR (81% identity and 93% positivity, Supporting Fig. 2). The secondary structure of human KDSR was previously reported 18 and the missense mutation tyrosine to guanine (T to G) caused the isoleucine (I) to arginine (R) change at the 105 amino acid residue of kdsr (Fig. 1D). Hepatocyte ballooning occurred at 7 dpf and advanced hepatic injury was identified at 8 dpf (Fig. 1C). We also noticed that all homozygous mutant died between 8 to 10 dpf. The relative mRNA expression of genes associated with inflammation, including tumor necrosis factor alpha (tnfa) and interleukin 1 beta (il1b); and fibrosis, collagen type 1 alpha 1a (col1a1a) were significantly elevated in the kdsr I105R mutants compared to controls (Fig. 1E). Thus, the kdsr mutant recapitulated characteristics found in hepatic injury in humans.
Inhibition of fatty acid synthase activity exacerbated steatosis phenotype rather than suppressing steatosis in the kdsr I105R mutant liver. To understand the mechanism of steatosis in the kdsr I105R mutant, we analyzed mRNA expression of proteins involved in lipid metabolism ( Fig. 2A). We found significant increases of srebp1, which regulates genes essential for lipogenesis, and fasn, a gene regulated by srebp1 that plays a role in palmitate synthesis from acetyl-CoA and malonyl-CoA. We also found significant increases of srebp2, which is essential for cholesterol biogenesis and the regulation of lpl expression, which encodes lipoprotein lipase, a key enzyme in lipid uptake (Fig. 3A). The hyperlipidemic phenotype in the kdsr I105R mutant (Fig. 1B) may have activated the expression of srebp2 and lpl expression to reduce plasma lipid level in the plasma rather than to cause liver steatosis. Since an increase in fasn expression may induce lipid accumulation in the liver, we treated mutants with fasnall, an inhibitor of fasn to determine whether the steatosis resulted from increase of lipogenesis. Interestingly, we found that the fasnall treatment exacerbated the steatosis phenotype in the kdsr I105R mutant rather than suppressing lipid accumulation (Fig. 2B). This result suggests that fasn upregulation was required to facilitate mitochondrial β-oxidation, because palmitate is required for lipid transport to the mitochondria through cpt1, and one of the substrates of fasn, malonyl-CoA, is known to inhibit mitochondrial β-oxidation 19 . Elevation of cpt1 expression further supports the enhancement of mitochondrial β-oxidation in the kdsr I105R mutant ( Fig. 2A), which is supported by the increase in oxygen consumption in the kdsr I105R mutants compared to control siblings at 7 dpf (Fig. 2C). Analysis of genes expressed in association with mitochondrial homeostasis revealed a significant increase in mfn1, important for mitochondrial fusion, opa1, which plays a role Lipid metabolism-and mitochondrial homeostasis-associated gene expression, and oxygen consumption analysis in larvae. Relative mRNA expression of genes associated with lipid metabolism include sterol regulatory element-binding protein 1 (srebp1), fatty acid synthase (fasn), srebp2, lipoprotein lipase (lpl), and carnitine-palmitoyltransferase I (cpt1) (A). Oil red O staining in control, kdsr I105R mutant, and kdsr I105R mutant siblings after fasnall treatment at 7 dpf; 1 µM fasnall treatment was performed from 4 dpf to 7 dpf (B). Images shown are representative of at least 10 in total. Scale bar in B = 100um. Average oxygen consumption rate was measured from 7 groups (3 larvae per group) of control and mutant siblings for 30 minutes (C). Relative mRNA expression of genes associated with mitochondrial homeostasis include dynamin related protein 1 (drp1), mitofusin 1 (mfn1), optic atrophy type 1 (opa1), peroxisome proliferator-activated receptor gamma coactivator 1-alpha (pgc1a), and NADH-ubiquinone oxidoreductase chain1 (nd1) (D). Error bars indicate standard deviation of the mean. *P ≤ 0.05, **P ≤ 0.005. in mitochondrial fusion and cristae stability, pgc1a, a master regulator of mitochondrial biogenesis 20 , and nd1, a gene expressed in the mitochondria and encodes a subunit of NADH dehydrogenase, while expression of drp1, a gene required for mitochondrial fission was not changed (Fig. 2D).
Ultrastructural analysis of mitochondrial cristae in the kdsr I105R mutant. Elevation of mitochondrial β-oxidation and oxygen consumption suggested that there may be accumulation of oxidative stress molecules in the mitochondria, which can induce mitochondrial injury. To test this possibility, we performed ultrastructural analysis wild-type control at 7 dpf and kdsr I105R mutant livers at 7 dpf and 7.5 dpf. We found that there was accumulation of mitochondria with less dense cristae structures in hepatocytes of mutant at 7 dpf ( Fig. 3A middle panel). There was also a more severely damaged and swollen mitochondrial phenotype in the kdsr I105R mutant at 7.5 dpf ( Fig. 3A right panel), suggesting progression of mitochondrial injury from less dense cristae damage to swollen and disrupted cristae phenotype. This could be a key factor for progression of liver phenotype from steatosis to hepatic injury phenotype in the kdsr I105R mutant. Higher magnification of mitochondria ( Fig. 3B), clearly showed cristae defects in the mutants. Activation of opa1 ( Fig. 2D) further supported damage of the cristae, because opa1 plays a role in stabilizing injured mitochondrial cristae 21 .
Kdsr I105R mutants accumulated sphingolipids through activation of the sphingolipid salvage pathway. Because kdsr is a key enzyme for the de novo synthesis of sphingolipids, we expected the missense mutation of kdsr to affect sphingolipid synthesis. Sphingolipidomics analysis data indicated accumulation of downstream components of sphingolipids, suggesting that the missense mutation might cause gain of kdsr function (Fig. 4A). To test our hypothesis, we generated a kdsr null mutant by CRISPR/Cas9 gene targeting in zebrafish, carrying premature stop codon by deletion/insertion in target region at exon 3 (Supporting Fig. 3). The kdsr cri null mutant showed the same phenotype as the kdsr I105R mutant (Supporting Fig. 3A) and histological analysis of the kdsr cri mutant showed that similar liver abnormality and steatosis observed in the kdsr I105R mutant (Supporting Fig. 3B). Additionally, the sphingolipid profile mirrored that of the kdsr I105R mutant (Supporting Fig. 3C,D). Complementation test was performed by crossing the heterozygous kdsr I105R mutant and the kdsr cri null mutant. We confirmed that the biallelic mutant (kdsr I105R/cri ) also developed the same phenotype as the kdsr I105R mutant (Supporting Fig. 4). We concluded that the missense mutation definitely induced loss of kdsr function. To understand how loss of kdsr function led to accumulation of downstream sphingolipids, we investigated the sphingolipid salvage pathway, which involves the degradation of SM and hexosyl-ceramides. SM species analysis showed significant decrease in the concentration of long chain SM species in the kdsr I105R mutant, including C16-SM, the most abundant SM (>4000 pmole per sample), C14-SM, C18-SM and C20:1-SM in the kdsr I105R mutant, while there were statistically significant increase of very long chain SM species such as C22-SM, C24:1-SM and C26:1-SM (Fig. 4B). We also found significant decrease of hexosyl-ceramides (glucosyl-or galactosyl-ceramides), including C16-, C18:1-, C18-, C20-, C26-Hexosyl-ceramides (Fig. 4C). This result suggests that C16-SM and hexosyl-ceramides were mainly used for the sphingolipid salvage pathway in the kdsr mutant zebrafish to compensate loss of sphingolipid de novo synthetic pathway. mRNA expression analysis showed a significant increase gba, which degrades glucosyl-ceramide into ceramides, asah1b, which is associated with lysosomal degradation of ceramides, and a decrease of asah2, a key enzyme in the plasma membrane, and acer2, which is found in the Golgi (Fig. 4D). Because SM is initially converted to ceramide by lysosomal acid sphingomyelinase encoded by smpd1, we also examined smpd1 expression and found no significant change in the kdsr I105R mutant (Fig. 4D). This result suggests that the lysosomal sphingolipid salvage pathway caused SM and hexosyl-ceramides degradation by transcriptional activation of gba and asah1b in the kdsr I105R mutant, and resulted in the accumulation of ceramides, Sph and S1P in the mutant. We also found transcriptional activation of sphk2 in the kdsr I105R mutant, while no change was detected in sphk1 expression (Fig. 4D). This result suggests that sphk2 is the major kinase involved in S1P accumulation in the mutant. In addition, mRNA expression of spl, the enzyme that mediates the degradation of S1P increased and possibly was induced to reduce accumulated of S1P in the kdsr I105R mutant (Fig. 4A). Further, we analyzed gene expression of salvage pathway related in the kdsr cri mutant to determine whether activation of sphingolipid salvage pathway is conserved in the kdsr null mutant (Supporting Fig. 3E). We found significant increases of asah1b and spl in the kdsr cri mutant, which is same as in the kdsr I105R mutant. Importantly, we found activation of smpd1, gba, ash1a and acer2 in the kdsr cri mutant, which suggesting kdsr cri mutant may have higher activity in salvage pathway compared to kdsr I105R mutant. Although both mutants have similar pathological defects, kdsr I105R mutant may have minimal activity of kdsr function.

Depletion of Sphk2 suppressed liver injury in the kdsr I105R mutant. A previous report showed that
sphk2 is the main sphingosine kinase in zebrafish embryo 22 . Because of transcriptional up-regulation of sphk2 in kdsr I105R mutant, and since accumulation of S1P has been reported to be involved in inflammation and steatohepatitis 10 , we tested whether the loss of sphk2 in the kdsr I105R mutant can suppress liver defects. We used an sphk2 wc1 mutant, a null mutant 23 . By crossing chimeric heterozygous (kdsr I105R/+ ; sphk2 wc1/+ ) mutants, we were able to obtain double homozygous mutants, which exhibited complete suppression of liver defects seen in the kdsr I105R mutant (Fig. 5A). Heterozygous kdsr I105R/+ , sphk2 wc1/+ mutants or homozygous sphk2 wc1 mutant had normal livers (data not shown). The qPCR analysis demonstrated that genes associated with inflammation (tnfa, il1b), tissue injury (a1at) and fibrosis (col1a1a) were all significantly reduced in the double mutant (Fig. 5B). We also investigated effects of sphk2 mutation on the sphingolipid salvage pathway. Significant suppression of mRNAs of spl and spp1, which functions to convert S1P to Sph, and asah1b was detected in kdsr I105R ; sphk2 wc1 double mutant, whereas no significant change was detected in asah2 levels. The sphk2 wc1 mutant showed significant decrease of asah2 expression. However, asah1a gene expression was elevated in both sphk2 mutant and kdsr I105R ; sphk2 wc1 double mutant (Fig. 5C). The qPCR results suggested that the salvage pathway is still activated in the kdsr I105R ; sphk2 wc1 mutant by elevating asah1a expression, a homolog of asah1b in zebrafish to maintain essential sphingolipids homeostasis for survival as additional compensatory mechanism. In addition, sphk2 depletion attenuated expression of genes (spl, spp1) involved in regulation of S1P levels. Thus, our results suggest that sphk2-mediated S1P accumulation might have a key role in the development of liver defects in the kdsr I105R mutant. Histological analysis of wt, kdsr I105R mutant and kdsr; sphk2 double homozygous mutant showed that suppression of liver defects by sphk2 depletion in the kdsr I105R mutant (Fig. 5D).
Sphk2 played a key role in oxidative stress and ER stress in the kdsr I105R mutant. Both oxidative stress and ER stress are major contributors to mitochondrial injury and liver disease 24,25 . We first analyzed the expression of genes activated by oxidative stress. We found significant elevation of oxidative stress-related genes in the kdsr I105R mutants compared to controls, suggesting that liver injury may be related to increased oxidative stress (Fig. 6A). Importantly, significant elevation of oxidative stress-related genes in the kdsr I105R mutant were suppressed by introducing the sphk2 null mutation into the kdsr I105R mutant (Fig. 6A). To investigate whether ER stress also contributed to liver defects in the kdsr I105R mutant, we analyzed the expression of genes associated with ER stress response in each control and mutant sibling, including atf4, and its target gene gadd45a. The Ire1, one of upstream component of nfkb transcription that functions as endonuclease to process xbp1, resulting in increased spliced form of xbp1 (xbp1-s) under ER stress condition. The atf6, and targets including xbp1, ddit3, edem1, bip, dnajc3, and grp94 were also analyzed. Additionally, bim, bida, and baxb were analyzed, which were known to be associated with ER stress-induced apoptosis through mitochondrial injury. Especially, bip, dnajc3, and grp94, molecular chaperons activated by unfolded protein folding response, are a main cause of ER stress. All of the tested genes in the ER stress pathway were all elevated in the kdsr I105R mutant (Fig. 6B). Although transcriptional activation of ire1 in the kdsr I105R mutant was not suppressed by sphk2 mutation, significant decrease of xbp1-s was found in both the sphk2 wc1 mutant and kdsr I105R ; sphk2 wc1 double mutant. This result suggests that sphk2 expression was required for endonuclease activity of Ire1. Transcriptional activation of atf6 was found in the sphk2 wc1 mutant; however, sphk2 depletion limited the activation of atf6 in the kdsr I105R mutant. Our result showed that even partial suppression of atf6 by sphk2 knockout was sufficient to suppress downstream target gene transcriptions including ddit3, edem1, bip, dnajc3, and grp94. These data suggest that sphk2 is a key mediator in the elevation of oxidative stress and ER stress in the kdsr I105R mutant.
Depletion of glutathione in the kdsr I105R mutant. Glutathione (GSH) is a key antioxidant that protects cellular components from oxidative stress and injury. A reduction in the ratio of GSH to GSSG indicates that cells have increased susceptibility to oxidative stress 26,27 . The increase in mRNA expression of components of the redox pathway (Fig. 6A) in the kdsr I105R mutant raised the possibility of severe oxidative stress. Therefore, we measured GSH, GSSG, and protein-thiol levels in the kdsr I105R mutant. We found a significant reduction in both GSH and GSSG levels (Fig. 7A,B). Although the GSH/GSSG ratio was not significantly decreased, the decrease in total GSH levels suggests that the kdsr I105R mutant might have limited capacity to protect proteins from ROS induced  (Fig. 7C). Further, a significant decrease in the protein-thiol levels suggested accumulation of oxidized proteins in the kdsr I105R mutant (Fig. 7D). In aggregate, these data suggest that GSH depletion/oxidative stress may play critical role in hepatocellular injury found in the kdsr I105R mutant.
To understand the mechanism of GSH depletion, we examined mRNA expression of gsr, an enzyme that converts GSSG to reduced GSH. We also examined genes associated with GSH synthesis including gss, which synthesizes GSH from r-glutamylcysteine and glycine; and gclc, which synthesizes r-glutamylcystein (Fig. 7E). Collectively, the genes involved in GSH synthesis were elevated in the kdsr I105R mutant, suggesting that GSH depletion was not a result of downregulation of those enzymes, and might be caused by limitation of substrate such as cysteine. Further, we treated control and mutant larvae with n-acetyl cysteine (NAC) to address whether NAC treatment can induce GSH synthesis and reduce ROS in kdsr mutants. The result showed that NAC treatment did not increased GSH level in both control and mutant larvae, which suggests that NAC treatment might not be effective in elevating GSH levels in vivo in zebrafish. However, we found a significant decrease of the oxidized form of GSH (GSSG) in control siblings compared to mutants. As a result, NAC treatment elevated GSH/ GSSG ratio in the control larvae, which led to ROS a decrease in ROS. However, NAC treatment was not effective in elevating GSH/GSSG ratio and suppressed ROS levels in the kdsr mutants (Supporting Fig. 5).

Hepatocellular injury in heterozygous kdsr I105R/+ mutant adults.
To investigate whether the heterozygous kdsr mutation alone leads to hepatocellular injury in adult zebrafish, we also analyzed livers of wild type and kdsr I105R/+ heterozygous mutant siblings. Ballooning of hepatocyte phenotype (Fig. 8A,B) and an increase of serum alanine-aminotransferase (ALT) levels (Fig. 8C) were identified in adults. Additionally, mRNA levels of tnfa, il1b and col1a1a expression were increased in the livers of kdsr I105R heterozygous mutants (Fig. 8D), similar to homozygous mutant larvae (Fig. 1E). To determine whether the heterozygous mutation also can enhance the sphinggolipid salvage pathway similar to homozygous mutant larvae, we analyzed genes involved in the sphingolipid salvage pathway. qPCR analysis revealed significant elevation of sphk2 and asah1a mRNAs in the liver (Fig. 8E). These data suggest that the liver injury phenotype of the kdsr I105R/+ adult may also result from S1P accumulation in a fashion similar to that of the kdsr I105R homozygous mutant.

Discussion
In this study, we have found that a novel kdsr I105R mutant had progression of liver disease from hepatomegaly to hepatic injury. Since KDSR deficiency was discovered about one year ago, liver abnormalities have not been addressed in KDSR human patients. It is possible that mutations in patients were hypomorphic, and biallelic mutations might not be enough to activate the sphingolipid salvage pathway. Our working model (Fig. 7F) showed that suppression of kdsr function by either homozygous or heterozygous mutations can enhance the sphingolipid salvage pathway, and that excess Sph is mainly phosphorylated by sphk2. Increase of S1P levels in turn might trigger oxidative stress via elevation of mitochondrial β-oxidation. Additional ER stress could be attributed to mitochondrial damage. Both stresses associated with liver disease were suppressed by loss of sphk2 function. Significant decrease of GSH in the kdsr mutant might sensitize kdsr mutant larvae against to oxidative stress. Mitochondrial injury could be upstream event preceding hepatic injury in the homozygous kdsr I105R mutant larvae and heterozygous kdsr I105R/+ mutant adults. Thus, we found how kdsr dysfunction can induce hepatic injury phenotype in both homozygous kdsr mutant larvae and heterozygous kdsr mutant adult fish. Our results suggest that genetic variant causing decrease of kdsr activity could be an underlying risk factor for development of liver disease and people who carry deleterious mutations in adult might be highly susceptible to liver disease such as steatohepatitis or advanced liver disease such as fibrosis or hepatocellular carcinoma. The progressive liver injury phenotype identified here was associated with progression of mitochondrial injury phenotype from less dense cristae structure to swollen mitochondria. The progressive liver defect was similar to the defect observed in the electron transfer flavoprotein alpha (etfa) mutant, a zebrafish model of multiple acyl-coA dehydrogenase deficiency 28 , suggesting that mitochondrial defects could be key factor in liver phenotype in the kdsr I105R mutant.
The homoeostasis of cellular sphingolipids are tightly controlled by both a de novo synthesis (initiated from palmitoyl-CoA and serine to produce ceramides) as well as a salvage catabolic pathway (initiated from complex sphingolipids such as glycosylated ceramides or sphingomyelins to produce ceramides) through the regulation of the intra-cellular levels of ceramide. Ceramide is the central molecule in the sphingolipid metabolic pathway that can be converted to various sphingolipid species 29 . Diet-induced alterations in ceramide via both de novo and salvage sphingolipid synthesis demonstrate that nutrition has the ability to alter sphingolipid metabolism and in turn downstream signaling pathways 30 . Enzymes of the salvage pathway have been implicated in dietary manipulations of ceramide levels. Ceramide acts as the central molecule in the sphingolipid metabolic pathway. A previous study showed that radio-labeled palmitate was found in the dihydrosphingosine, and then in the ceramides via de novo synthesis. However, they also found ceramides produced via salvage pathway at the same time. Thus, palmitic acid treatment can enhance ceramide formation through the both de novo and salvage pathway 31 . Furthermore, administration of high fat diet enhanced mRNA expression and activity of acid sphingomyelinase and neutral sphingomyelinase in rat liver 32 and mouse adipose tissue 33 . Pharmacological inhibition of acid sphingomyelinase inhibited ceramide induction by high fat diet in plasma and adipose tissue in mice 34 . However, compared to studies examining de novo sphingolipid synthesis, regulation of the salvage pathway has not been well-studied. Future experimentation that include modeling of sphingolipid metabolism can help understand the role the salvage pathway in mammals. To determine impact of external feeding in sphingolipid salvage pathway in zebrafish after consumed own egg yolk, we performed a diet experiment in control and kdsr mutant siblings at 7dpf. We found egg yolk feeding induced significant increase of gene expression involved in salvage pathway in both control and mutants (Supporting Fig. 6). Additionally, this result suggested that activation of salvage pathway in the kdsr mutant was not induced by starvation. Loss of kdsr function should in theory inhibit de novo synthesis of sphingolipids and deplete downstream sphingolipids. Due to the current lack of LC-MS methods to quantify the substrate of kdsr, 3-keto-dhSph, it is expected that significant amount of 3-keto-dhSph would be accumulated in the mutant. As 3-keto-dhSph and Sph are isomers, the kdsr mutant may instead produce 3-keto-dihydroceramide, which may have inhibitory effect on dihydroceramide desaturase, the enzyme that generates ceramide from dihydroceramide. This hypothesis may explain how endogenous dhSph and dihydroceramide, possibly transported from egg yolk accumulated in the kdsr mutant (Supporting Fig. 7). As 3-keto-sphingolipids have not been sufficiently studied, there is unfortunately no quantification methods available yet. Further, future studies on sphingolipids with the 3-keto moiety will be necessary to answer this question. In this study, we found that inhibition of the kdsr function activated the sphingolipid salvage pathway possibly to compensate for the inhibition of the de novo synthesis pathway. As a result, Sph, S1P and ceramides may have accumulated in the kdsr I105R mutant. Because the larvae could have used the egg yolk as nutrient by 6 dpf, probably there was no external source of sphingolipids without feeding the larvae, thus the salvage pathway would be the only way to produce downstream sphingolipid species. Previous studies have shown that accumulation of ceramides may induce steatohepatitis in [35][36][37][38] . However, our genetic study of the interaction of sphk2 and kdsr mutation suggested that S1P is likely the cause of the observed liver phenotype (Fig. 5).
Sphk1 was found to be necessary for S1P-mediated steatohepatitis in a high fat diet-induced liver disease in mice; however, the role of SphK2 was not investigated 10 . Notably, our findings raised the possibility that sphk2 is the major sphingosine kinase involved in steatohepatitis (Fig. 5B) and this finding is consistent with a previous report that showed sphk2 functions as the main sphingosine kinase for S1P production in zebrafish 22 . Importantly, a significant increase of SPHK2 expression was found in both steatosis and steatohepatitis cases of multiple patients 39 . Thus, an important finding from the current study is that sphk2 may play a key role in the development of hepatic injury associated with kdsr mutation and then steatohepatitis. The effect of high fat diet in the kdsr I105R mutant remains to be determined. Cellular GSH maintains the oxidation status of thiols in critical proteins and defends cells against to reactive oxygen species by its reducing capacity 40 . GSH depletion might reduce the buffering capacity of GSH against oxidative stress, which plays a key role in the aging process and the pathogenesis of many diseases 41,42 . Based on mRNA expression analysis, enzymes responsible for GSH synthesis and reduction are highly upregulated in the kdsr mutant, although total GSH level is significantly lower than control siblings (Fig. 7). This result suggests that GSH depletion may occur by depletion of a substrate(s) for GSH synthesis. Further investigation will be necessary to address the mechanism of GSH depletion. We propose that cysteine depletion might affect GSH levels in the kdsr mutant, since cysteine is tightly regulated in the liver for both GSH synthesis and protein synthesis 43 . We tested whether NAC treatment can elevate GSH synthesis, but NAC treatment did not elevate GSH in both control and kdsr mutants (Supporting Fig. 5). A previous study showed that NAC concentrations would have to exceed 1 mM, which is therapeutically unattainable in vivo to achieve maximum rates of GSH synthesis 44 . The NAC treatment might not be enough to elevate GSH synthesis in zebrafish same as in human. In this paper, we were not able to make a conclusion whether cysteine depletion is the main cause of GSH depletion, because sub-lethal dose of the NAC treatment did not elevate GSH levels in vivo. Further investigation will be necessary to determine the mechanism of GSH depletion in the kdsr mutant in the future.
Collectively, our findings indicate that kdsr deletion leads to compensatory activation of the sphingolipid salvage pathway and S1P accumulation, which can result in increased mitochondrial activity, oxidative stress, and ER stress and subsequent hepatocellular injury. The data point to the possibility that kdsr could be a novel genetic risk factor for steatosis and liver injury. The zebrafish strain used in this study was AB/TU. Adults were maintained at 28.5 °C and fed twice a day with brine shrimp and Tetramin flake (Tetra US, Blacksburg, VA). Embryos were obtained from natural mating and raised at 28.5 °C in egg water (0.3 mg of sea salt/L). Kdsr I105R and sphk2 wc123 lines were outcrossed with wild type AB/TU at least five times to reduce additional background mutations and maintenance. The kdsr I105R mutant were genotyped using 5′-GTGGTTCTTTGCATTTCTGTTGATGT-3′ and 5′-ATGGCGAAAGGATTTTATGAATTGTTAAACATAC -3′, 30 cycles of PCR with 56 °C annealing temperature and then digested with Hpy188I (R0617, NEB, inc.) for 1 hr. sphk2 wc1 was genotyped as described in the previous paper 23 . A kdsr null mutant was generated by CRISPR/Cas9 gene targeting in the laboratory. The guide RNA for kdsr was designed to target exon 3 of kdsr (5′-GGTTCAGGCTAAGAAAGAAGTGG-3′); T7-gRNA-kdsr nucleotides (5′-GAAATTAATACGACT CACTATAGGTTCAGGCTAAGAAAGAAGGTTTTAGAGCTAGAAATAGCAAGTTAAAAT) were used as template for gRNA synthesis. 50 to 100 pmole of guide RNA and 100 pmole of Cas9 RNA were co-injected at 1 cell-stage eggs. Injected embryos were raised and outcrossed with wild type to produce progeny. Each of the progenies was genotyped using 5′-GGTGTTACACAATTTGAAAACCATTTACC ACTG-3′ and 5′-TCCTTTGTTAAAGACATACATACT TGCTTATC-3′ primers, and deletions were confirmed by sequencing. Identified founder fish was used to generate stable lines. Siblings were used as controls.
Whole genome sequencing. Genomic DNA from 10 normal siblings and 10 homozygous mutants were used as template DNAs for whole genome sequencing. The MUSC Sequencing Core performed sequencing of samples using an Illumina HiSeq2500 Platform with 150 bp paired-end reads, resulting in approximately 10 fold genomic coverage. The sequencing results were uploaded to the SNPtrack Mapping server (http://genetics. bwh.harvard.edu/ snptrack/), and mapped mutations re-confirmed by sequencing and genotyping of individual mutants. The sequencing results were submitted to NCBI Short Read Archive (SRA) database under the BioSample accessions SAMN10247386 and SAMN10247387.

Oil Red O (ORO) Staining.
For whole mount staining at the larval stage, we used the method previously described 28 . Briefly, larvae were fixed in 4% PFA overnight. The same numbers of control and mutant larvae (5 to 10 larvae each) were processed together in the same tube. After staining, larvae were briefly rinsed in PBS-Tween and fixed in 4% PFA for 10 minutes. Larvae were mounted in glycerol prior to imaging. For ORO staining on transversely sectioned larvae, frozen sections with 10 µm thickness were dried at room temperature for 5 minutes. 150 µl of working ORO solution was added to slides and stained for 30 seconds. They were then washed with distilled water and mounted using 75% glycerol.
H&e staining. Embryos were fixed in 4% paraformaldehyde from overnight to two days at 4 °C. Fixed embryos were embedded in 1.2% agarose/5% sucrose, and saturated in 30% sucrose at 4 °C for 1 to 2 days. Blocks were frozen using liquid nitrogen. 10 µm sections were collected on microscope slides using a Leica cryostat. For adult liver histology, truncated bodies were fixed in 4% paraformaldehyde from overnight to two days at 4 °C and processed for embedding in paraffin. Paraffin sections were used for H&E staining, which was conducted in the Histo-Core lab at Medical University of South Carolina (MUSC). Images were taken with AxioCam ICC3 attached to Zeiss Axio-Imager M2.
Blood preparation and alanine aminotransferase (ALT) measurements. Zebrafish blood was obtained by minimally invasive blood collection method using a heparinized needle. The site for blood collection is along the body axis and posterior to the anus in the region of the dorsal aorta. Blood was collected from adult zebrafish, 20 hours after feeding and diluted 1:10 in PBS. The average volume of blood collected from a three-individual fish (average body weight = 0.6 g) was 25 µL. Plasma was separated by centrifugation for 15 minutes at 2,000 × g using a refrigerated centrifuge. 10 µl of plasma were transferred to 96-well plate and ALT was measured using a microplate-based ALT activity assay kit (Pointe Scientific, Cat. A7526).
Oxygen consumption assay of zebrafish larvae. Larval average oxygen consumption rate was determined using a sensor-dish reader (SDR) system (Loligo Systems, Viborg, DEN). Recordings were made once every 5 minutes for 1 hour using PreSens-SDR_v38 software (Loligo Systems, Viborg, DEN). Each well of 24-well optical fluorescence glass sensor microplate was filled with 125 µl of egg water and pre-ran for 20 minutes at 24 °C room. Three larvae were placed into wells of a 24-well plate, and then the plate was immediately sealed using parafilm and a silicone gel pad cover. Using PreSens-SDR_v38 software, dissolved oxygen amount was measured every 3 minutes for 30 minutes. Average oxygen consumption rate (nmol/L/min/fish) was calculated from change in O 2 concentration over time. Data presented are from 7 measurements made with 3 larvae of each control and mutant siblings.
Lipid analysis. Three sets of 7 day-old control siblings (n = 30) and mutant siblings (n = 30) were anesthetized and collected in 15 ml tube and kept in −80 °C before submission to LC/MS/MS analysis at the MUSC Lipidomics core facility as previously described 45 Table 1), and SsoAdvanced ™ Universal SYBR ® Green Supermix (Bio-rad, Cat. 172-5274). We used qPCR primers that were either used in previous studies in zebrafish [47][48][49][50][51][52][53][54] or designed and tested in our lab (Supporting Table 1). Glyceraldehyde-3-phosphate dehydrogenase (gapdh) was used as reference, and relative quantification was calculated using double delta Ct method. The qPCR was run in at least triplicate for each gene. Total RNA were isolated from 20 control siblings and 20 kdsr I105R homozygous mutant larvae. For multiple comparisons between wild type, kdsr I105R , sphk2 wc1 , and kdsr I105R ; sphk2 wc1 double homozygous mutants, kdsr I105R ; sphk2 wc1 double heterozygous mutants were crossed and siblings were anesthetized in 0.016% ethyl 3-aminobenzoate methanesulfonate salt (MS-222, Sigma-aldrich, E10521) with Ringer's solution. Tips of tails were dissected for genotyping and bodies were kept in −80 °C. After genotyping of sibling larvae at 7 dpf, 5 of each wild type, single and double mutants were collected and total RNA were extracted for cDNA synthesis.

Measurement of intracellular reduced thiol levels.
Triplicated groups of control and mutant larvae, each contained 5 larvae at 7 dpf stage were lysed in 75 ul of ice-cold lysis buffer [50 mM Tris-HCl (pH 7.5), 150 mM NaCl, 1% Triton, 1 mM EDTA, 1 mM EGTA, plus a protease inhibitor cocktail (Roche/Sigma)]. 2ul of each lysate were immediately subjected to the reduced thiol measurement by using thiol fluorescent probe IV (Millipore) as previously described 55 . Fluorescent intensities were detected at 400Ex/465Em by Microplate Reader. The thiol fluorescent probe IV with lysis buffer was used as a negative control.

Measurement of GSH and GSSG levels.
Quantitative determinations of GSH and GSSG levels were performed using the enzymatic-recycling method 56 . Triplicated group of each control mutant larvae were subjected to lysis. Each group contains 5 larvae at 7 dpf. Protein in the extracts from 5 larvae was precipitated by sulfosalicylic acid and the supernatant was then divided into two. For reduced GSH, the supernatant was incubated with the thiol fluorescent probe IV, and fluorescent intensities were measured at 400Ex/465Em. For total GSH (GSH + GSSG), the supernatant was neutralized by triethanolamine and incubated with the reduction system (containing NADPH and glutathione reductase) at 37 °C for 20 min. GSSG was calculated based on the results from reduced GSH and total GSH; the ratio of GSH GSSG / statistical tests. All evaluations were performed with MS Excel software. The Student t-test was used to test significant differences between groups. For sphingolipid analysis, groups of siblings (30 per each group) were used for lipid extraction. For the qPCR studies, groups of siblings (minimum 10 per each group) were used to generate cDNAs, and three sets of qPCR were analyzed per each target gene. P values less than 0.05 were considered statistically significant.