Parallel Reaction Monitoring reveals structure-specific ceramide alterations in the zebrafish

Extensive characterisations of the zebrafish genome and proteome have established a foundation for the use of the zebrafish as a model organism; however, characterisation of the zebrafish lipidome has not been as comprehensive. In an effort to expand current knowledge of the zebrafish sphingolipidome, a Parallel Reaction Monitoring (PRM)-based liquid chromatography–mass spectrometry (LC–MS) method was developed to comprehensively quantify zebrafish ceramides. Comparison between zebrafish and a human cell line demonstrated remarkable overlap in ceramide composition, but also revealed a surprising lack of most sphingadiene-containing ceramides in the zebrafish. PRM analysis of zebrafish embryogenesis identified developmental stage-specific ceramide changes based on long chain base (LCB) length. A CRISPR-Cas9-generated zebrafish model of Farber disease exhibited reduced size, early mortality, and severe ceramide accumulation where the amplitude of ceramide change depended on both acyl chain and LCB lengths. Our method adds an additional level of detail to current understanding of the zebrafish lipidome, and could aid in the elucidation of structure-function associations in the context of lipid-related diseases.

isomers, m/z's of the major LCB fragments ([M + H-2H 2 O-acyl] + ) corresponding to d16:2 through d20:2, and d16:1 through d20:1 were manually extracted from the MS 2 spectra of each targeted parent ion. Additional extractions of d14, d15, d21 and d22 LCBs did not yield significant signal. Example chromatograms and MS 2 of C41:1 ceramide in 48 hours-post-fertilisation (hpf) zebrafish embryos are shown in Fig. 1c-e, illustrating the presence of four structural isomers with the same parent m/z.

Ceramide composition in zebrafish and humans.
To compare the ceramide profiles of zebrafish and humans, the optimised PRM protocol was applied toward adult zebrafish brain, 7 days-post-fertilisation (dpf) zebrafish larvae and a human embryonic kidney cell line (HEK293). Ceramide levels across these three sample types are summarised in Tables 1 and S2. The correlation coefficient was 0.99997 (y = 6165754.0x + 12551539.9) in the range of 100 fmol-1 nmol for the d18:1-d 7 /18:0 ceramide standard, and 0.99984 (y = 14326319.2x + 71700089.0) in the range of 100 fmol-1 nmol for the d18:1-d 7 /24:1 ceramide standard (Fig. S1). All PRM spectra were inspected and integrated manually, using the following criteria for inclusion: (1) presence of major (quantifier) LCB fragment within 5 ppm of theoretical m/z and within linear range based on the d18:1-d 7 /18:0 and d18:1-d 7 /24:1 standards, for all samples in data set, (2) presence of at least one additional ceramide-specific MS 2 fragment (loss-of-water fragment and/or a minor LCB fragment) within 5 ppm of theoretical m/z, in most samples of data set, (3) signal-to-noise ≥10, and 4) at least three consecutive non-zero data points across integrated peak. Ceramides that fulfilled all four criteria were quantified with internal standards; species that fulfilled all criteria but fell below the linear range of the tested standards were reported as trace (tr) ( Table 1). With the exception of Fig. 2a, all reported data were generated using only ceramides with non-trace values. Following data filtering based on inclusion criteria, 86 distinct LCB-acyl chain combinations were detected, 69 of which could be quantified (Tables 1 and S2). 48, 72 and 65 ceramide species were identified in zebrafish brain, larvae and HEK293 cells, respectively (Table 1 and Fig. 2a); of these species, 34, 51 and 57 were quantifiable (Table 1). 40 species were detected across all three sample types (Fig. 2a). The total ceramide content ranged from 1-3 nmol/mg protein, and the dominant LCB in all three sample types was d18:1 (Table 1), a consequence of the transfer of C16:0-CoA to L-serine via serine palmitoyltransferase to generate the precursor to the d18:1 LCB 40 . www.nature.com/scientificreports www.nature.com/scientificreports/ Monounsaturated (d16:1-d20:1 with no additional degrees of unsaturation) and diunsaturated ceramides were detected across all three sample types (Fig. 2b). Interestingly, while a significant portion of diunsaturated HEK293 ceramides contained the sphingadiene (d18:2) LCB, sphingadiene-containing ceramides were detected only in low amounts in zebrafish brain and larvae; instead, the major contributor toward the second degree of ceramide unsaturation in the zebrafish was the acyl chain (Fig. 2c). The presence of sphingadienes has been previously demonstrated in multiple model organisms and is unlikely to be specific to HEK293 cells 31,46-49 . Ceramide changes during zebrafish embryogenesis. Having defined the zebrafish ceramide profile relative to a human cell line, we proceeded to examine ceramide regulation in the context of early zebrafish development. Following initial stages of embryogenesis, hatching occurs between ~48-72 hpf, with organogenesis continuing into the larval stage (Fig. 3a) 50 . While prey-seeking behaviour is observable by ~4-5 dpf 51 , the developing larvae also rely on the yolk sac for nutrients, which is gradually depleted and rarely visible by 7 dpf (Fig. 3a). In the absence of external food source, yolk depletion triggers the initiation of a gluconeogenic program to accommodate increasing energy demand 6 .
As zebrafish development is accompanied by significant changes in energy expenditure and therefore metabolism, we used our PRM method to capture potential development stage-specific ceramide differences. Lipids from 48 hpf embryos, 4 dpf and 7 dpf larvae were extracted and subjected to PRM analysis. The majority of the previously detected ceramides (Table 1) were present across all three timepoints, with an increase in several d18:1 species over time ( Fig. 3b and Table S3). Surprisingly, a significant time-dependent reduction of d19:1 ceramides was also detected ( Fig. 3b and Table S3). The fraction of d19:1 ceramides decreased from 12% at 48 hpf to 1% at 7 dpf, accompanied by a rise in d18:1 ceramides from 74% to 90% (Fig. 3c). A smaller reduction (9% at 48 hpf to 5% at 7 dpf) was also observed for d16:1 ceramides (Fig. 3c). Taken together, these data point toward LCB-specific differences in ceramide profile during zebrafish embryogenesis. www.nature.com/scientificreports www.nature.com/scientificreports/ A zebrafish model of farber disease. While we have successfully quantified ceramides in wild-type zebrafish, the ease of genetic manipulation in this model organism supports the development of ceramide-related disease models. Farber disease is a lysosomal storage disorder characterised by loss-of-function mutations in acid ceramidase (ASAH1) that lead to ceramide accumulation, multiple-organ pathologies and early mortality 42 . Morpholino knockdown of the ASAH1 orthologue in zebrafish causes loss of motor neuron branching and increased cell death in the spinal cord 52 . To gain a more thorough understanding of the metabolic landscape in the context of Farber disease, we generated a zebrafish model of Farber disease using CRISPR-Cas9 mutagenesis.
Two zebrafish orthologues (asah1a, asah1b) of human ASAH1 were identified by database search 53 . The existence of two orthologues is a likely consequence of the additional round of whole genome duplication that occurred in teleosts relative to other vertebrates, such that 26% of zebrafish genes exist as ohnologue pairs 54 . Amino acid identity was 61% between Asah1a and ASAH1, and 62% between Asah1b and ASAH1 55,56 . To maximise mutagenesis efficiency, five CRISPR guides were designed against each of the two zebrafish ohnologues, and the mixture of ten guides was microinjected into embryos at the 1-cell stage. The guides were designed to be near residues G230 and R249 of Asah1a (NP_001006088), and G235 and R254 of Asah1b (NP_956871), as these are the conserved residues of G235 and R254 in human ASAH1 (NP_808592) 57 ; G235R and R254G mutations in ASAH1 are associated with significant loss of enzyme activity and varying degrees of disease severity [58][59][60][61] .
Initial DNA fragment analysis of ten injected embryos identified significant base pair shifts in all ten embryos, supporting a high efficiency of mutagenesis. Adult founder fish from the injected embryos were in-crossed to generate F1 fish. Following DNA fragment analysis and Sanger sequencing, F1 fish carrying mutations of interest were individually out-crossed with wild-type zebrafish to yield doubly heterozygous F2 populations (Fig. 4a). Heatmap of ceramide content in 48 hpf, 4 dpf and 7 dpf larvae, demonstrating time-dependent reduction in the majority of d19:1 (highlighted in red) ceramides (see Table S3  Relative to SKO siblings, all DKO zebrafish exhibited significantly reduced size (Fig. 4c,d) and early mortality at ~4 months. Mutagenesis of either asah1a or asah1b alone did not lead to size or lifespan differences relative to wild-type fish, suggesting that zebrafish Asah1a and Asah1b carry overlapping functions.

Ceramide alterations in farber disease.
To characterise the ceramide profile within the Farber disease model, brains were isolated from 3.5-month SKO and DKO zebrafish and subjected to PRM analysis. Significant ceramide accumulation was present for all detectable ceramides in the Farber disease model; representative species are shown in Fig. 5a (see Table S4 for additional ceramide quantifications). Total ceramide content was elevated by 15-fold in DKO zebrafish relative to SKO siblings (Fig. 5b). No change in ceramide content was detected in SKO fish relative to wild-type siblings (Fig. S4a,b), suggesting that preservation of any one of the two duplicate Asah1 enzymes is sufficient for maintaining physiological ceramide levels.
We also examined potential alterations in total monounsaturated, diunsaturated, even-and odd-chain ceramides. While total even-and odd-chain ceramides were elevated to similar degrees in DKO relative to SKO fish (Fig. 5c), total monounsaturated ceramides were increased to a larger degree (20-fold) than diunsaturated species (6-fold) (Fig. 5d). As our study is limited to brain, it is currently unknown if preferential elevation of monounsaturated ceramides in the zebrafish Farber disease model is tissue-specific.
Given our observation of LCB-dependent ceramide changes during zebrafish embryogenesis, we also examined the effects of LCB and acyl chain lengths on ceramide content in Farber disease. Interestingly, the amplitude of ceramide accumulation was LCB-dependent, as ceramides with longer LCBs were more elevated in DKO fish compared to those with shorter LCBs (Figs. 5e and S5a). In the case of acyl chain lengths, no consistent trends were observed for C14-C21 ceramides, but ceramides with C22 or longer acyl chains did not accumulate as much as C14-C21 species (Figs. 5f and S5b,c). Taken together, our data reveal that ceramide accumulation in the Farber disease zebrafish model is heavily biased toward species with longer LCBs and shorter acyl chains (Fig. S5c).

Discussion
A PRM method was optimised for detailed ceramide isomer quantification in zebrafish and mammalian samples. 45 ceramides were targeted for PRM analysis. Quantified LCBs ranged from d16 to d20, and acyl chains from C14 to C26. Taken together, 86 distinct LCB-acyl chain combinations were detected across different sample types and developmental stages.   Table S4 for quantification of additional ceramide species). www.nature.com/scientificreports www.nature.com/scientificreports/ Total ceramide content of HEK293 cells was 1.3 ± 0.1 (±SEM) nmol/mg protein (Table 1) and within range of known values for mammalian cells (0.4-5.5 nmol/mg protein) 31,44,[62][63][64] . Studies of zebrafish embryos, larvae and adult tissues have previously demonstrated the presence of over 30 ceramide species 16,31,65 . Total even-chain (C14-C26) ceramide content of TuAB zebrafish brain has been reported at 1.4 nmol/mg 65 , while a total brain ceramide content of 3.0 ± 0.2 nmol/mg protein was observed in our samples (Table 1); the observed differences could be attributable to potential differences in fish age and diet, as well as in the number of ceramides targeted. Our total ceramide content for 7 dpf zebrafish was 1.0 ± 0.1 nmol/mg protein (Table 1), or 145 ± 7 pmol/15 larvae. Previous measurements in 6 hpf-7 dpf zebrafish ranged from ~30 pmol/15 embryos to ~22 nmol/15 larvae 16,41,66 . These variabilities may be partially explained by differences in the ceramides targeted, larval age and strain, but could also be a consequence of sample processing. As both ionisation and extraction efficiencies depend on lipid structure, the use of different internal standards across different studies could lead to variability in quantification. Differences in chromatography and sample complexity, the latter a consequence of the extraction method, may also lead to different matrix effects around the analytes of interest and changes in signal response 67 .
Comparison of PRM data from zebrafish brain, larvae and HEK293 cells demonstrated a remarkable degree of species overlap, but also revealed some surprising differences such as the absence of the majority of d18:2 ceramides in the zebrafish. Sphingadiene-containing ceramides have been identified in multiple organisms including humans, mice, fruit flies and plants 31,[46][47][48][49] ; therefore, their absence in the zebrafish is unexpected and intriguing. To the best of our knowledge, this is the first demonstration of the sparsity of the sphingadiene LCB in the zebrafish model organism.
The first committed step of de novo ceramide biosynthesis involves the transfer of C16:0-CoA to L-serine via serine palmitoyltransferase (SPT) to generate 3-keto-dihydrosphingosine, which is reduced to dihydrosphingosine and then acylated by ceramide synthase to generate dihydroceramide; a 4,5-trans double bond is added by dihydroceramide desaturase as the last step of ceramide biosynthesis to yield mature ceramide 40 . While C16:0-CoA is the major SPT substrate, additional acyl-CoAs are also tolerated, including C16:1-CoA, which would ultimately give rise to the d18:2 ceramides that are present in mammals but mostly absent in zebrafish. Importantly, SPT catalytic activity depends on multiple subunits. In humans, SPTLC1 dimerises with either SPTLC2 or SPTLC3 to generate a fully functional enzyme complex, while two additional small activating subunits, SPTSSA and SPTSSB, enhance basal SPT activity and confer acyl-CoA specifity 68,69 . All five human SPT subunits have orthologues in the zebrafish, with amino acid identities of 61-85% (Table S5) 55 . Given different heterotrimers of SPTLCs and SPTSSs exhibit distinct acyl-CoA preferences, structural differences between human and zebrafish SPTs, as well as potential differences in subunit combinations, may lead to intolerance for the C16:1-CoA substrate and therefore the near absence of the d18:2 LCB in zebrafish. While the above hypothesis assumes origin of d18:2 ceramides from the C16:1-CoA substrate, there also exists the possibility of undiscovered sphingolipid or fatty acid desaturases capable of forming double bonds along different positions of the LCB 47 . Expansion of our current method toward the rest of the sphingolipidome could also help address the question of whether lack of d18:2 ceramides in the zebrafish is due to increased incorporation into additional sphingolipids.
While sphingadienes with 4,5-trans and 14,15-cis double bonds (Δ 4,14 ) have been identified in mammals 49,70 , Δ 4,6 -sphingadienes are found in fruit fly 47 . Lipidomic analysis of the sphingosine-1-phosphate lyase-deficient Drosophila line Sply revealed preferential upregulation of Δ 4,6 -sphingadienes in the thorax, where the degenerating flight muscles associated with Sply knockdown are located 47,71 . More than 90% of all sphingadiene-containing ceramides identified in human fibroblasts were found in the detergent-soluble fraction, suggesting that these lipids may function differently from the known role of ceramides as components of detergent-resistant lipid rafts 31 . Importantly, sphingadienes promoted apoptosis and autophagy of colon cancer cells and neuroblastoma via modulation of Akt signaling 72,73 . Given the biological relevance of sphingadienes in mammals and fruit fly, the near absence of these species in zebrafish implies the existence of alternative lipid processing and signalling pathways that are worth exploring if the zebrafish were to be further implemented toward the study of human metabolic disorders.
Our PRM analysis of zebrafish embryogenesis identified time-dependent increase of d18:1 ceramides, and reduction of d19:1 and d16:1 species. Sphingolipids are essential for all major aspects of cell function, and disrupted sphingolipid metabolism is associated with defects in organogenesis 74,75 . Zebrafish ceramides fluctuate over the first few days of development, rising in the embryo body while decreasing in the yolk 16 ; however, LCB-specific ceramide changes in the context of embryogenesis are not well understood. Our PRM analysis of zebrafish embryogenesis identified time-dependent increase of d18:1 ceramides, and reduction of d19:1 and d16:1 species. Given ceramides with different acyl chains may exert opposing functions in apoptosis 76 , our findings suggest that subtle variations in LCB length may also lead to different functions during development.
It is worth noting that as our sample isolation does not separate the yolk from the embryo body, it remains unclear if LCB length confers distinct cellular localisations; loss of d16:1 and d19:1 ceramides could thus be a direct readout for yolk depletion. Prevalence of the d18:1 LCB in mammals and zebrafish suggests that the 18-carbon length may be preferred in the context of cellular function, and more d18:1 ceramides may be required as development progresses. As the purpose of the egg yolk is predominantly fuel source, initial demands for specific LCB lengths may be less stringent in an effort to maximise yolk lipid content. As development continues, catabolism of yolk lipids leads to the biosynthesis of new d18:1 ceramides and additional, more structure-specific lipids to better suit the needs of the growing organism.
Given the necessity of ceramides for cellular function, mutations in enzymes of ceramide metabolism often result in fatal health outcomes. Reduced size and early mortality were observed in our zebrafish model of acid ceramidase deficiency (Farber disease). Size difference between asah1a/b −/− (DKO) and SKO populations is progressive and not apparent prior to three months of age, suggesting that this phenotype may be a consequence of progressive ceramide accumulation that significantly compromises health upon reaching a threshold. Given the frequent occurrence of progressive joint deformations and contractures in Farber disease 42  www.nature.com/scientificreports www.nature.com/scientificreports/ DKO zebrafish may be indicative of impaired skeletal development; small animal imaging techniques such as microCT have been successfully optimised for the zebrafish and could prove helpful toward addressing this question 77 . It is also worth noting that unlike other lysosomal storage disease models 11,33 , the majority of DKO zebrafish do not develop cachexia or progressive loss of locomotor skills despite rapid mortality around four months of age. Given the involvement of the heart and nervous system in Farber disease, the rapid loss in DKO fish survival could be due to sudden decline in organ function such as heart failure or fatal seizures; close monitoring of the DKO population and additional phenotypic characterisations could help uncover the cause of early death and shed light on the mechanisms of disease progression.
In additional to reduced size and lifespan, total brain ceramide content was increased by 15-fold in DKO zebrafish relative to SKO siblings. Given the wide range of ceramide functions, ceramide accumulation likely contributes toward early mortality via multiple pathways. Sphingolipids are required for early vertebrate development, and disrupted sphingolipid metabolism can lead to defects in organogenesis 74,75 . Morpholinos against the sphingosine-1-phosphate receptors s1pr1 and s1pr2 severely inhibit intersegmental vessel angiogenesis 65 , and delayed epiboly is observed in a maternal zygotic mutant of ceramide synthase 2b 41 . Zebrafish carrying a nonsense mutation in 3-ketodihydrosphingosine reductase, an enzyme of the de novo ceramide biosynthetic pathway, exhibits hepatosplenomegaly that progresses to steatosis and liver injury 66 . While our analysis has not been implemented in DKO and SKO larvae, the stable genetic mutations present in our model suggest that ceramide accumulation may begin in early development, resulting in defects in organogenesis that lead to reduced lifespan later in life. Ceramide accumulation may also perturb homeostasis by shifting cell fate toward apoptosis; in the mitochondrial apoptotic pathway, ceramides form channels in the mitochondrial outer membrane that cooperate with the proapoptotic proteins BAX and BAK to trigger membrane permeabilization 78 . Given the position of ceramide at the centre of the sphingolipid metabolic pathway, drastic alterations in ceramide levels could also translate to changes in additional sphingolipids such as sphingomyelins and hexosylceramides, all of which are critical components of lipid rafts and the myelin sheath 79,80 . Perturbations in either the level or acyl chain distribution of these sphingolipids may negatively impact myelin stability, thus contributing to potential pathologies within the central nervous system. Surprisingly, the extent of ceramide accumulations present in the DKO zebrafish depended on both LCB and acyl chain lengths; ceramides with longer LCBs were preferentially elevated, while a significantly smaller degree of ceramide accumulation occurred for all species with acyl chains longer than C21. Within the cell, multiple pathways are in place for ceramide generation 40 . While variations in LCB lengths are the consequence of different SPT subunit combinations 68,69 , diversity of acyl chain lengths is driven by ceramide synthases that catalyse the acylation of sphinganine to dihydroceramide 81 . The six known mammalian ceramide synthases (CERS1-6) exhibit distinct tissue localisations and acyl chain preferences 81 . Nine zebrafish Cers orthologues have been identified, with unique expression patterns during embryogenesis, and amino acid identities of 49-80% relative to human CERSs (Table S5) 55,82 . Given the involvement of multiple enzymes and pathways toward ceramide generation, loss of ceramidase activity may trigger multiple compensatory pathways in an effort to normalise ceramide levels. Selective downregulation of Spt heterotrimers and/or ceramide synthases in the absence of Asah1 may lead to the LCB-and acyl chain-specific ceramide changes observed in our zebrafish model.
The rapid reduction in ceramide accumulation that occurs past C21 could also be an indication of altered peroxisomal or mitochondrial function. During fatty acid β-oxidation in mammals, very long chain fatty acids (VLCFAs), typically C22-C26, are transported to the peroxisome, where they are oxidized to shorter-chain species that are trafficked to the mitochondria for further breakdown 83 . Short-to medium-chain FAs are typically catabolized in the mitochondria without peroxisomal involvement 84 . Loss-of-function mutations in the VLCFA transporter ABCD1 leads to X-linked adrenoleukodystrophy (ALD), one of the most frequently occurring peroxisomal disorders that involve progressive loss of myelin and early mortality; elevated VLCFA is detected in the blood of 99% of male patients, and is one of the diagnostic tools for ALD 85 . Given the function of the peroxisome in metabolite clearance, significant ceramide storage at the lysosome could trigger increased peroxisomal metabolism as a compensatory mechanism, thereby reducing the availability of the VLCFA pool for ceramide generation. An expansion of our PRM method toward other lipid families could help address the question of additional organelle involvement, as well as the roles of additional lipids in disease progression.
Animals. All zebrafish husbandry and experiment protocols were approved by and carried out in accordance with the Institutional Animal Care and Use Committee at Massachusetts General Hospital or University of Utah.
Sample collection. HEK293 cells were maintained at 37 °C and 5% CO 2 in Dulbecco's Modified Eagle Medium (high glucose) supplemented with 10% foetal bovine serum, penicillin and streptomycin. Cells were grown in 100 mm dishes, passaged at least three times and harvested at confluency (48-52 hours after last passage) www.nature.com/scientificreports www.nature.com/scientificreports/ by scraping. Zebrafish embryos and larvae were euthanised by cooling the Petri dish on ice, and the euthanised embryos or larvae transferred into Falcon tubes by pipetting. Adult TuAB zebrafish (12 months of age for all wild-type fish, 3.5 months for DKO and SKO fish due to early mortality in the former) were euthanised by immersion in ice-chilled water, following which the head was removed and the brain rapidly excised and flash frozen in liquid nitrogen prior to storage at −80 °C. Lipid extraction. All samples were extracted using a modified version of the Bligh-Dyer method 86 . Briefly, one adult zebrafish brain or 35-150 zebrafish larvae were homogenised 40 times on ice with a 7 mL glass dounce homogeniser (VWR, KT885300-0007) with pestle A in a mixture of 1.5 mL aqueous buffer (100 mM trisodium citrate, 1 M sodium chloride, pH 3.6), 1.5 mL methanol, and 3.0 mL chloroform containing internal standards. The resulting mixture was transferred into a glass vial with PTFE-lined cap (VWR, 66009-984), vortexed for 15 s, and centrifuged at 2000 g for 8 min to induce phase separation. The organic layer was retrieved with a Pasteur pipette, dried under a gentle stream of nitrogen and stored at −80 °C. HEK293 cells (~10 7 cells/sample) were homogenized by manual shaking for 30 s in glass vials with PTFE-lined caps, in the same volume of aqueous-organic mixture as zebrafish samples; subsequent steps were identical to zebrafish samples. All samples were analysed within two weeks. 1/6-1/4 of each sample was used for analysis. Protein concentrations for all samples were determined by Bradford assay.
Lipidomics. Parallel Reaction Monitoring was performed in positive ionisation mode on an Ultimate 3000 HPLC online with a Thermo q-Exactive Plus quadrupole-orbitrap mass spectrometer equipped with a heated electrospray ion source. Solvent A was 95: 5 water: methanol, 5 mM ammonium formate, 0.1% formic acid. Solvent B was 60: 35: 5 isopropanol: methanol: water, 5 mM ammonium formate, 0.1% formic acid. A Bio-Bond C4 column (Dikma, 5 µm, 50 × 4.6 mm) coupled to a Bio-Bond C4 guard cartridge (Dikma, 5 µm, 10 × 4.0 mm) was used. The gradient was held at 0% B between 0 and 5 min, raised to 20% B at 5.1 min, increased linearly from 20% to 100% B between 5.1 and 55 min, held at 100% B between 55 min and 63 min, returned to 0% B at 63.1 min, and held at 0% B until 70 min. Flow rate was 0.1 mL/min from 0 to 5 min, 0.4 mL/min between 5.1 min and 55 min, and 0.5 mL/min between 55.1 min and 70 min. Divert valve was set to waste for 1-1.5 min and 62-70 min, and to MS for rest of run. Spray voltage was 4.0 kV. Sheath, auxiliary and spare gases were 52.5, 13.75 and 2.75, respectively. Capillary temperature was 268.75 °C. S-lens RF level was 50. MS 2 was acquired with a resolution of 35000, target ion 2e5, maximum injection time 100 ms and isolation window 1.0 m/z. Stepped normalised collision energies were 20, 30. The m/z inclusion list is provided in Table S1. All data analyses were performed by manual peak integration in Thermo Xcalibur. Three internal standards were used for quantification. C32-C34, C35-C39 and C40-C44 ceramides were quantified using the d18:1-d 7 /15:0, d18:1-d 7 /18:0 and d18:1-d 7 /24:1 standards, respectively. Additional notes on chromatography are provided under Supplemental Discussion and Figs. S6-S10.
PCR products were purified using the QIAquick PCR Purification kit (Qiagen, 28104), and immediately used for in vitro transcription with the MEGAscript SP6 Transcription kit (ThermoFisher Scientific, AM1330) following the method of Gagnon et al. 10 . All reactions (10 μL/reaction) were incubated overnight at 37 °C and purified using RNA Clean & Concentrator kit (Zymo Research, R1013). Yield per guide was 250-1100 ng/μL. All guides were diluted to 200-500 ng/μL in nuclease-free water based on RNA concentration, aliquoted and stored at −80 °C.
Guide delivery. Frozen guides (eight guides total, 200-500 ng/μL, 2 μL aliquot per guide) were thawed on ice and 1 μL of each guide was removed and combined. Cas9 nuclease (10% of combined guide volume) was added to the combined guides (final Cas9 concentration ~2 μM), and the resulting solution incubated at room temperature for 10 min. Phenol Red solution (5% of Cas9 and combined guide volume) was added to improve visualisation during microinjection. ~1-2 nL of the guide-Cas9 solution was microinjected into each zebrafish embryo (TuAB strain) at the 1-cell stage.

Data availability
The datasets and model organism generated in this study are available from the corresponding author on reasonable request.